ha_tokudb.cc 307 KB
Newer Older
1
#ifdef USE_PRAGMA_IMPLEMENTATION
2
#pragma implementation          // gcc: Class implementation
3 4 5
#endif

#define MYSQL_SERVER 1
6
#include "toku_mysql_priv.h"
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7
#include "hatoku_cmp.h"
Zardosht Kasheff's avatar
Zardosht Kasheff committed
8 9 10 11 12 13
extern "C" {
#include "stdint.h"
#if defined(_WIN32)
#include "misc.h"
#endif
}
14
#include "hatoku_assert.h"
15

16 17 18 19 20 21 22 23
static inline void *thd_data_get(THD *thd, int slot) {
    return thd->ha_data[slot].ha_ptr;
}

static inline void thd_data_set(THD *thd, int slot, void *data) {
    thd->ha_data[slot].ha_ptr = data;
}

24 25 26 27 28
#undef PACKAGE
#undef VERSION
#undef HAVE_DTRACE
#undef _DTRACE_VERSION

29
//#include "tokudb_config.h"
30 31 32 33 34 35 36

/* We define DTRACE after mysql_priv.h in case it disabled dtrace in the main server */
#ifdef HAVE_DTRACE
#define _DTRACE_VERSION 1
#else
#endif

Zardosht Kasheff's avatar
Zardosht Kasheff committed
37
#include "hatoku_defines.h"
38
#include "ha_tokudb.h"
Zardosht Kasheff's avatar
Zardosht Kasheff committed
39
#include "hatoku_hton.h"
40 41
#include <mysql/plugin.h>

Zardosht Kasheff's avatar
Zardosht Kasheff committed
42 43 44 45
static const char *ha_tokudb_exts[] = {
    ha_tokudb_ext,
    NullS
};
46

47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68
// XXX 4530 get read or write locks on the key file array,
// so that functions not synchronized by mysql table locks
// don't race. 
//
// ha_tokudb::info() does not take any table locks
// so it needs to grab a read lock. any other function writing
// to the key file array needs to grab a write lock.
static void share_key_file_rdlock(TOKUDB_SHARE * share)
{
    rw_rdlock(&share->key_file_lock);
}

static void share_key_file_wrlock(TOKUDB_SHARE * share)
{
    rw_wrlock(&share->key_file_lock);
}

static void share_key_file_unlock(TOKUDB_SHARE * share)
{
    rw_unlock(&share->key_file_lock);
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
69 70 71
//
// This offset is calculated starting from AFTER the NULL bytes
//
72
inline u_int32_t get_fixed_field_size(KEY_AND_COL_INFO* kc_info, TABLE_SHARE* table_share, uint keynr) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
73 74
    uint offset = 0;
    for (uint i = 0; i < table_share->fields; i++) {
75 76
        if (kc_info->field_lengths[i] && !bitmap_is_set(&kc_info->key_filters[keynr],i)) {
            offset += kc_info->field_lengths[i];
Zardosht Kasheff's avatar
Zardosht Kasheff committed
77 78 79 80 81 82
        }
    }
    return offset;
}


83
inline u_int32_t get_len_of_offsets(KEY_AND_COL_INFO* kc_info, TABLE_SHARE* table_share, uint keynr) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
84 85
    uint len = 0;
    for (uint i = 0; i < table_share->fields; i++) {
86 87
        if (kc_info->length_bytes[i] && !bitmap_is_set(&kc_info->key_filters[keynr],i)) {
            len += kc_info->num_offset_bytes;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
88 89 90 91 92 93
        }
    }
    return len;
}


94
static int allocate_key_and_col_info ( TABLE_SHARE* table_share, KEY_AND_COL_INFO* kc_info) {
95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113
    int error;
    //
    // initialize all of the bitmaps
    //
    for (uint i = 0; i < MAX_KEY + 1; i++) {
        error = bitmap_init(
            &kc_info->key_filters[i],
            NULL,
            table_share->fields,
            false
            );
        if (error) {
            goto exit;
        }
    }
    
    //
    // create the field lengths
    //
114
    kc_info->field_lengths = (u_int16_t *)my_malloc(table_share->fields*sizeof(u_int16_t), MYF(MY_WME | MY_ZEROFILL));
115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135
    kc_info->length_bytes= (uchar *)my_malloc(table_share->fields, MYF(MY_WME | MY_ZEROFILL));
    kc_info->blob_fields= (u_int32_t *)my_malloc(table_share->fields*sizeof(u_int32_t), MYF(MY_WME | MY_ZEROFILL));
    
    if (kc_info->field_lengths == NULL || 
        kc_info->length_bytes == NULL || 
        kc_info->blob_fields == NULL ) {
        error = ENOMEM;
        goto exit;
    }
exit:
    if (error) {
        for (uint i = 0; MAX_KEY + 1; i++) {
            bitmap_free(&kc_info->key_filters[i]);
        }
        my_free(kc_info->field_lengths, MYF(MY_ALLOW_ZERO_PTR));
        my_free(kc_info->length_bytes, MYF(MY_ALLOW_ZERO_PTR));
        my_free(kc_info->blob_fields, MYF(MY_ALLOW_ZERO_PTR));
    }
    return error;
}

136 137 138 139 140
/** @brief
    Simple lock controls. The "share" it creates is a structure we will
    pass to each tokudb handler. Do you have to have one of these? Well, you have
    pieces that are used for locking, and they are needed to function.
*/
141
static TOKUDB_SHARE *get_share(const char *table_name, TABLE_SHARE* table_share) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
142 143
    TOKUDB_SHARE *share = NULL;
    int error = 0;
144
    uint length;
145

146 147 148
    pthread_mutex_lock(&tokudb_mutex);
    length = (uint) strlen(table_name);

149
    if (!(share = (TOKUDB_SHARE *) my_hash_search(&tokudb_open_tables, (uchar *) table_name, length))) {
150 151
        char *tmp_name;

Zardosht Kasheff's avatar
Zardosht Kasheff committed
152 153 154 155
        //
        // create share and fill it with all zeroes
        // hence, all pointers are initialized to NULL
        //
156 157 158 159 160
        if (!(share = (TOKUDB_SHARE *) 
            my_multi_malloc(MYF(MY_WME | MY_ZEROFILL), 
                            &share, sizeof(*share),
                            &tmp_name, length + 1, 
                            NullS))) {
161 162 163 164 165 166 167
            pthread_mutex_unlock(&tokudb_mutex);
            return NULL;
        }
        share->use_count = 0;
        share->table_name_length = length;
        share->table_name = tmp_name;
        strmov(share->table_name, table_name);
168

169 170
        error = allocate_key_and_col_info(table_share, &share->kc_info);
        if (error) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
171 172 173
            goto exit;
        }

Zardosht Kasheff's avatar
Zardosht Kasheff committed
174
        bzero((void *) share->key_file, sizeof(share->key_file));
175

Zardosht Kasheff's avatar
Zardosht Kasheff committed
176 177 178 179
        error = my_hash_insert(&tokudb_open_tables, (uchar *) share);
        if (error) {
            goto exit;
        }
180 181
        thr_lock_init(&share->lock);
        pthread_mutex_init(&share->mutex, MY_MUTEX_INIT_FAST);
182 183
        // XXX 4530 initialize the key file lock
        my_rwlock_init(&share->key_file_lock, 0);
184
        my_rwlock_init(&share->num_DBs_lock, 0);
185 186
    }

Zardosht Kasheff's avatar
Zardosht Kasheff committed
187 188 189 190 191 192
exit:
    if (error) {
        pthread_mutex_destroy(&share->mutex);
        my_free((uchar *) share, MYF(0));
        share = NULL;
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
193
    pthread_mutex_unlock(&tokudb_mutex);
194
    return share;
195 196
}

197 198 199 200 201 202 203 204

void free_key_and_col_info (KEY_AND_COL_INFO* kc_info) {
    for (uint i = 0; i < MAX_KEY+1; i++) {
        bitmap_free(&kc_info->key_filters[i]);
    }
    
    for (uint i = 0; i < MAX_KEY+1; i++) {
        my_free(kc_info->cp_info[i], MYF(MY_ALLOW_ZERO_PTR));
205
        kc_info->cp_info[i] = NULL; // 3144
206 207 208 209 210 211 212
    }
    
    my_free(kc_info->field_lengths, MYF(MY_ALLOW_ZERO_PTR));
    my_free(kc_info->length_bytes, MYF(MY_ALLOW_ZERO_PTR));
    my_free(kc_info->blob_fields, MYF(MY_ALLOW_ZERO_PTR));
}

213
static int free_share(TOKUDB_SHARE * share, bool mutex_is_locked) {
214 215 216 217 218 219 220 221 222
    int error, result = 0;

    pthread_mutex_lock(&tokudb_mutex);

    if (mutex_is_locked)
        pthread_mutex_unlock(&share->mutex);
    if (!--share->use_count) {
        DBUG_PRINT("info", ("share->use_count %u", share->use_count));

Zardosht Kasheff's avatar
Zardosht Kasheff committed
223 224 225 226 227 228 229 230 231 232 233
        //
        // number of open DB's may not be equal to number of keys we have because add_index
        // may have added some. So, we loop through entire array and close any non-NULL value
        // It is imperative that we reset a DB to NULL once we are done with it.
        //
        for (uint i = 0; i < sizeof(share->key_file)/sizeof(share->key_file[0]); i++) {
            if (tokudb_debug & TOKUDB_DEBUG_OPEN) {
                TOKUDB_TRACE("dbclose:%p\n", share->key_file[i]);
            }
            if (share->key_file[i]) { 
                error = share->key_file[i]->close(share->key_file[i], 0);
234
                assert(error == 0);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
235 236 237 238 239
                if (error) {
                    result = error;
                }
                share->key_file[i] = NULL;
            }
240
        }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
241

242
        free_key_and_col_info(&share->kc_info);
243

Zardosht Kasheff's avatar
Zardosht Kasheff committed
244
        if (share->status_block && (error = share->status_block->close(share->status_block, 0))) {
245
            assert(error == 0);
246
            result = error;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
247 248
        }
        
249

250
        my_hash_delete(&tokudb_open_tables, (uchar *) share);
251 252
        thr_lock_delete(&share->lock);
        pthread_mutex_destroy(&share->mutex);
253
        rwlock_destroy(&share->num_DBs_lock);
254 255 256 257 258
        my_free((uchar *) share, MYF(0));
    }
    pthread_mutex_unlock(&tokudb_mutex);

    return result;
259 260
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
261

Zardosht Kasheff's avatar
Zardosht Kasheff committed
262
#define HANDLE_INVALID_CURSOR() \
Zardosht Kasheff's avatar
Zardosht Kasheff committed
263 264 265 266 267
    if (cursor == NULL) { \
        error = last_cursor_error; \
        goto cleanup; \
    }

268 269 270 271
const char *ha_tokudb::table_type() const {
    extern const char * const tokudb_hton_name;
    return tokudb_hton_name;
} 
272

273 274 275
const char *ha_tokudb::index_type(uint inx) {
    return "BTREE";
}
276

Zardosht Kasheff's avatar
Zardosht Kasheff committed
277 278 279
/* 
 *  returns NULL terminated file extension string
 */
280
const char **ha_tokudb::bas_ext() const {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
281 282
    TOKUDB_DBUG_ENTER("ha_tokudb::bas_ext");
    DBUG_RETURN(ha_tokudb_exts);
283 284
}

285 286 287 288 289 290 291 292
static inline bool is_insert_ignore (THD* thd) {
    //
    // from http://lists.mysql.com/internals/37735
    //
    return thd->lex->ignore && thd->lex->duplicates == DUP_ERROR;
}

static inline bool is_replace_into(THD* thd) {
293
    return thd->lex->duplicates == DUP_REPLACE;
294 295
}

296
static inline bool do_ignore_flag_optimization(THD* thd, TABLE* table, bool opt_eligible) {
297 298
    uint pk_insert_mode = get_pk_insert_mode(thd);
    return ( 
299
        opt_eligible && 
300
        (is_replace_into(thd) || is_insert_ignore(thd)) && 
301 302 303 304 305
        ((!table->triggers && pk_insert_mode < 2) || pk_insert_mode == 0)
        );
}

ulonglong ha_tokudb::table_flags() const {
306
    return (table && do_ignore_flag_optimization(ha_thd(), table, share->replace_into_fast) ? 
307
        int_table_flags | HA_BINLOG_STMT_CAPABLE : 
Rich Prohaska's avatar
Rich Prohaska committed
308
#if defined(HA_GENERAL_ONLINE)
309
        int_table_flags | HA_BINLOG_ROW_CAPABLE | HA_BINLOG_STMT_CAPABLE | HA_ONLINE_ALTER);
310 311 312
#else
        int_table_flags | HA_BINLOG_ROW_CAPABLE | HA_BINLOG_STMT_CAPABLE);
#endif
313 314
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
315 316 317 318
//
// Returns a bit mask of capabilities of the key or its part specified by 
// the arguments. The capabilities are defined in sql/handler.h.
//
319
ulong ha_tokudb::index_flags(uint idx, uint part, bool all_parts) const {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
320
    TOKUDB_DBUG_ENTER("ha_tokudb::index_flags");
321
    assert(table_share);
322
    ulong flags = (HA_READ_NEXT | HA_READ_PREV | HA_READ_ORDER | HA_KEYREAD_ONLY | HA_READ_RANGE);
323 324 325
    if (table_share->key_info[idx].flags & HA_CLUSTERING) {
        flags |= HA_CLUSTERED_INDEX;
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
326
    DBUG_RETURN(flags);
327 328
}

Yoni Fogel's avatar
Yoni Fogel committed
329

Zardosht Kasheff's avatar
Zardosht Kasheff committed
330 331 332 333 334 335 336 337 338 339
//
// struct that will be used as a context for smart DBT callbacks
// contains parameters needed to complete the smart DBT cursor call
//
typedef struct smart_dbt_info {
    ha_tokudb* ha; //instance to ha_tokudb needed for reading the row
    uchar* buf; // output buffer where row will be written
    uint keynr; // index into share->key_file that represents DB we are currently operating on
} *SMART_DBT_INFO;

340 341 342 343 344 345 346
typedef struct smart_dbt_bf_info {
    ha_tokudb* ha;
    bool need_val;
    int direction;
    THD* thd;
} *SMART_DBT_BF_INFO;

347 348 349 350 351 352
typedef struct index_read_info {
    struct smart_dbt_info smart_dbt_info;
    int cmp;
    DBT* orig_key;
} *INDEX_READ_INFO;

353

354 355 356 357
int ai_poll_fun(void *extra, float progress) {
    LOADER_CONTEXT context = (LOADER_CONTEXT)extra;
    if (context->thd->killed) {
        sprintf(context->write_status_msg, "The process has been killed, aborting add index.");
358
        return ER_ABORTING_CONNECTION;
359 360 361 362 363 364
    }
    sprintf(context->write_status_msg, "Adding of indexes about %.1f%% done", progress*100);
    thd_proc_info(context->thd, context->write_status_msg);
    return 0;
}

365 366 367 368
int poll_fun(void *extra, float progress) {
    LOADER_CONTEXT context = (LOADER_CONTEXT)extra;
    if (context->thd->killed) {
        sprintf(context->write_status_msg, "The process has been killed, aborting bulk load.");
369
        return ER_ABORTING_CONNECTION;
370
    }
371
    sprintf(context->write_status_msg, "Loading of data about %.1f%% done", progress*100);
372 373 374 375
    thd_proc_info(context->thd, context->write_status_msg);
    return 0;
}

376 377 378 379 380 381 382 383 384 385 386
struct hot_poll_fun_extra {
    uint current_table;
    uint num_tables;
};

int hot_poll_fun(void *extra, float progress) {
    HOT_OPTIMIZE_CONTEXT context = (HOT_OPTIMIZE_CONTEXT)extra;
    if (context->thd->killed) {
        sprintf(context->write_status_msg, "The process has been killed, aborting hot optimize.");
        return ER_ABORTING_CONNECTION;
    }
387
    sprintf(context->write_status_msg, "Optimization of index %u of %u about %.lf%% done", context->current_table + 1, context->num_tables, progress*100);
388 389 390 391
    thd_proc_info(context->thd, context->write_status_msg);
    return 0;
}

392 393 394 395 396 397 398 399 400 401 402 403 404 405

void loader_ai_err_fun(DB *db, int i, int err, DBT *key, DBT *val, void *error_extra) {
    LOADER_CONTEXT context = (LOADER_CONTEXT)error_extra;
    assert(context->ha);
    context->ha->set_loader_error(err);
}

void loader_dup_fun(DB *db, int i, int err, DBT *key, DBT *val, void *error_extra) {
    LOADER_CONTEXT context = (LOADER_CONTEXT)error_extra;
    assert(context->ha);
    context->ha->set_loader_error(err);
    if (err == DB_KEYEXIST) {
        context->ha->set_dup_value_for_pk(key);
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
406 407
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
408 409 410 411 412 413 414
//
// smart DBT callback function for optimize
// in optimize, we want to flatten DB by doing
// a full table scan. Therefore, we don't
// want to actually do anything with the data, hence
// callback does nothing
//
415
static int smart_dbt_do_nothing (DBT const *key, DBT  const *row, void *context) {
Barry Perlman's avatar
Barry Perlman committed
416
  return 0;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
417 418
}

419 420 421
static int smart_dbt_metacallback (DBT const *key, DBT  const *row, void *context) {
    DBT* val = (DBT *)context;
    val->data = my_malloc(row->size, MYF(MY_WME|MY_ZEROFILL));
422 423 424 425
    if (val->data == NULL) return ENOMEM;
    memcpy(val->data, row->data, row->size);
    val->size = row->size;
    return 0;
426 427
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
428

429 430 431
static int
smart_dbt_callback_rowread_ptquery (DBT const *key, DBT  const *row, void *context) {
    SMART_DBT_INFO info = (SMART_DBT_INFO)context;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
432
    info->ha->extract_hidden_primary_key(info->keynr, key);
433
    return info->ha->read_row_callback(info->buf,info->keynr,row,key);
434 435
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
436 437 438
//
// Smart DBT callback function in case where we have a covering index
//
Barry Perlman's avatar
Barry Perlman committed
439 440
static int
smart_dbt_callback_keyread(DBT const *key, DBT  const *row, void *context) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
441
    SMART_DBT_INFO info = (SMART_DBT_INFO)context;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
442 443
    info->ha->extract_hidden_primary_key(info->keynr, key);
    info->ha->read_key_only(info->buf,info->keynr,key);
Barry Perlman's avatar
Barry Perlman committed
444
    return 0;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
445 446 447 448 449
}

//
// Smart DBT callback function in case where we do NOT have a covering index
//
Barry Perlman's avatar
Barry Perlman committed
450 451
static int
smart_dbt_callback_rowread(DBT const *key, DBT  const *row, void *context) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
452
    int error = 0;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
453
    SMART_DBT_INFO info = (SMART_DBT_INFO)context;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
454
    info->ha->extract_hidden_primary_key(info->keynr, key);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
455 456
    error = info->ha->read_primary_key(info->buf,info->keynr,row,key);
    return error;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
457 458
}

459 460 461 462 463 464 465 466 467 468 469 470 471
//
// Smart DBT callback function in case where we have a covering index
//
static int
smart_dbt_callback_ir_keyread(DBT const *key, DBT  const *row, void *context) {
    INDEX_READ_INFO ir_info = (INDEX_READ_INFO)context;
    ir_info->cmp = ir_info->smart_dbt_info.ha->prefix_cmp_dbts(ir_info->smart_dbt_info.keynr, ir_info->orig_key, key);
    if (ir_info->cmp) {
        return 0;
    }
    return smart_dbt_callback_keyread(key, row, &ir_info->smart_dbt_info);
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
472 473 474 475 476 477 478 479
static int
smart_dbt_callback_lookup(DBT const *key, DBT  const *row, void *context) {
    INDEX_READ_INFO ir_info = (INDEX_READ_INFO)context;
    ir_info->cmp = ir_info->smart_dbt_info.ha->prefix_cmp_dbts(ir_info->smart_dbt_info.keynr, ir_info->orig_key, key);
    return 0;
}


480 481 482 483 484 485 486 487 488 489 490 491 492
//
// Smart DBT callback function in case where we do NOT have a covering index
//
static int
smart_dbt_callback_ir_rowread(DBT const *key, DBT  const *row, void *context) {
    INDEX_READ_INFO ir_info = (INDEX_READ_INFO)context;
    ir_info->cmp = ir_info->smart_dbt_info.ha->prefix_cmp_dbts(ir_info->smart_dbt_info.keynr, ir_info->orig_key, key);
    if (ir_info->cmp) {
        return 0;
    }
    return smart_dbt_callback_rowread(key, row, &ir_info->smart_dbt_info);
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
493 494 495 496 497
//
// macro for Smart DBT callback function, 
// so we do not need to put this long line of code in multiple places
//
#define SMART_DBT_CALLBACK ( this->key_read ? smart_dbt_callback_keyread : smart_dbt_callback_rowread ) 
498
#define SMART_DBT_IR_CALLBACK ( this->key_read ? smart_dbt_callback_ir_keyread : smart_dbt_callback_ir_rowread ) 
Zardosht Kasheff's avatar
Zardosht Kasheff committed
499 500


Zardosht Kasheff's avatar
Zardosht Kasheff committed
501 502 503 504
//
// macro that modifies read flag for cursor operations depending on whether
// we have preacquired lock or not
//
505
#define SET_PRELOCK_FLAG(flg) ((flg) | (range_lock_grabbed ? (use_write_locks ? DB_PRELOCKED_WRITE : DB_PRELOCKED) : 0))
Zardosht Kasheff's avatar
Zardosht Kasheff committed
506 507 508 509 510 511 512 513 514 515 516 517 518

//
// This method retrieves the value of the auto increment column of a record in MySQL format
// This was basically taken from MyISAM
// Parameters:
//              type - the type of the auto increment column (e.g. int, float, double...)
//              offset - offset into the record where the auto increment column is stored
//      [in]    record - MySQL row whose auto increment value we want to extract
// Returns:
//      The value of the auto increment column in record
//
ulonglong retrieve_auto_increment(uint16 type, uint32 offset,const uchar *record)
{
Zardosht Kasheff's avatar
Zardosht Kasheff committed
519
    const uchar *key;     /* Key */
Zardosht Kasheff's avatar
Zardosht Kasheff committed
520 521
    ulonglong   unsigned_autoinc = 0;  /* Unsigned auto-increment */
    longlong      signed_autoinc = 0;  /* Signed auto-increment */
Zardosht Kasheff's avatar
Zardosht Kasheff committed
522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565
    enum { unsigned_type, signed_type } autoinc_type;
    float float_tmp;   /* Temporary variable */
    double double_tmp; /* Temporary variable */

    key = ((uchar *) record) + offset;

    /* Set default autoincrement type */
    autoinc_type = unsigned_type;

    switch (type) {
    case HA_KEYTYPE_INT8:
        signed_autoinc   = (longlong) *(char*)key;
        autoinc_type     = signed_type;
        break;

    case HA_KEYTYPE_BINARY:
        unsigned_autoinc = (ulonglong) *(uchar*) key;
        break;

    case HA_KEYTYPE_SHORT_INT:
        signed_autoinc   = (longlong) sint2korr(key);
        autoinc_type     = signed_type;
        break;

    case HA_KEYTYPE_USHORT_INT:
        unsigned_autoinc = (ulonglong) uint2korr(key);
        break;

    case HA_KEYTYPE_LONG_INT:
        signed_autoinc   = (longlong) sint4korr(key);
        autoinc_type     = signed_type;
        break;

    case HA_KEYTYPE_ULONG_INT:
        unsigned_autoinc = (ulonglong) uint4korr(key);
        break;

    case HA_KEYTYPE_INT24:
        signed_autoinc   = (longlong) sint3korr(key);
        autoinc_type     = signed_type;
        break;

    case HA_KEYTYPE_UINT24:
        unsigned_autoinc = (ulonglong) uint3korr(key);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
566
    break;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601

    case HA_KEYTYPE_LONGLONG:
        signed_autoinc   = sint8korr(key);
        autoinc_type     = signed_type;
        break;

    case HA_KEYTYPE_ULONGLONG:
        unsigned_autoinc = uint8korr(key);
        break;

    /* The remaining two cases should not be used but are included for 
       compatibility */
    case HA_KEYTYPE_FLOAT:                      
        float4get(float_tmp, key);  /* Note: float4get is a macro */
        signed_autoinc   = (longlong) float_tmp;
        autoinc_type     = signed_type;
        break;

    case HA_KEYTYPE_DOUBLE:
        float8get(double_tmp, key); /* Note: float8get is a macro */
        signed_autoinc   = (longlong) double_tmp;
        autoinc_type     = signed_type;
        break;

    default:
        DBUG_ASSERT(0);
        unsigned_autoinc = 0;
    }

    if (signed_autoinc < 0) {
        signed_autoinc = 0;
    }

    return autoinc_type == unsigned_type ?  
           unsigned_autoinc : (ulonglong) signed_autoinc;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
602 603 604
}


Zardosht Kasheff's avatar
Zardosht Kasheff committed
605 606 607

inline bool
is_null_field( TABLE* table, Field* field, const uchar* record) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
608
    uint null_offset;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
609
    bool ret_val;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
610 611
    if (!field->null_ptr) {
        ret_val = false;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
612
        goto exitpt;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
613 614
    }
    null_offset = get_null_offset(table,field);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
615 616 617
    ret_val = (record[null_offset] & field->null_bit) ? true: false;

exitpt:
Zardosht Kasheff's avatar
Zardosht Kasheff committed
618 619 620 621 622
    return ret_val;
}

inline ulong field_offset(Field* field, TABLE* table) {
    return((ulong) (field->ptr - table->record[0]));
Zardosht Kasheff's avatar
Zardosht Kasheff committed
623 624
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
625 626 627 628
inline HA_TOKU_ISO_LEVEL tx_to_toku_iso(ulong tx_isolation) {
    if (tx_isolation == ISO_READ_UNCOMMITTED) {
        return hatoku_iso_read_uncommitted;
    }
629 630 631
    else if (tx_isolation == ISO_READ_COMMITTED) {
        return hatoku_iso_read_committed;
    }
632 633 634
    else if (tx_isolation == ISO_REPEATABLE_READ) {
        return hatoku_iso_repeatable_read;
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
635 636 637 638 639 640 641
    else {
        return hatoku_iso_serializable;
    }
}

inline u_int32_t toku_iso_to_txn_flag (HA_TOKU_ISO_LEVEL lvl) {
    if (lvl == hatoku_iso_read_uncommitted) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
642
        return DB_READ_UNCOMMITTED;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
643
    }
644 645 646
    else if (lvl == hatoku_iso_read_committed) {
        return DB_READ_COMMITTED;
    }
647 648 649
    else if (lvl == hatoku_iso_repeatable_read) {
        return DB_TXN_SNAPSHOT;
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
650 651 652 653 654
    else {
        return 0;
    }
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
655 656


Zardosht Kasheff's avatar
Zardosht Kasheff committed
657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787
int filter_key_part_compare (const void* left, const void* right) {
    FILTER_KEY_PART_INFO* left_part= (FILTER_KEY_PART_INFO *)left;
    FILTER_KEY_PART_INFO* right_part = (FILTER_KEY_PART_INFO *)right;
    return left_part->offset - right_part->offset;
}


//
// Be very careful with parameters passed to this function. Who knows
// if key, table have proper info set. I had to verify by checking
// in the debugger.
//
void set_key_filter(MY_BITMAP* key_filter, KEY* key, TABLE* table, bool get_offset_from_keypart) {
    FILTER_KEY_PART_INFO parts[MAX_REF_PARTS];
    uint curr_skip_index = 0;

    for (uint i = 0; i < key->key_parts; i++) {
        //
        // horrendous hack due to bugs in mysql, basically
        // we cannot always reliably get the offset from the same source
        //
        parts[i].offset = get_offset_from_keypart ? key->key_part[i].offset : field_offset(key->key_part[i].field, table);
        parts[i].part_index = i;
    }
    qsort(
        parts, // start of array
        key->key_parts, //num elements
        sizeof(*parts), //size of each element
        filter_key_part_compare
        );

    for (uint i = 0; i < table->s->fields; i++) {
        Field* field = table->field[i];
        uint curr_field_offset = field_offset(field, table);
        if (curr_skip_index < key->key_parts) {
            uint curr_skip_offset = 0;
            curr_skip_offset = parts[curr_skip_index].offset;
            if (curr_skip_offset == curr_field_offset) {
                //
                // we have hit a field that is a portion of the primary key
                //
                uint curr_key_index = parts[curr_skip_index].part_index;
                curr_skip_index++;
                //
                // only choose to continue over the key if the key's length matches the field's length
                // otherwise, we may have a situation where the column is a varchar(10), the
                // key is only the first 3 characters, and we end up losing the last 7 bytes of the
                // column
                //
                TOKU_TYPE toku_type;
                toku_type = mysql_to_toku_type(field);
                switch(toku_type) {
                case(toku_type_blob):
                    break;
                case(toku_type_varbinary):
                case(toku_type_varstring):
                case(toku_type_fixbinary):
                case(toku_type_fixstring):
                    if (key->key_part[curr_key_index].length == field->field_length) {
                        bitmap_set_bit(key_filter,i);
                    }
                    break;
                default:
                    bitmap_set_bit(key_filter,i);
                    break;
                }
            }
        }
    }
}


inline uchar* pack_fixed_field(
    uchar* to_tokudb,
    const uchar* from_mysql,
    u_int32_t num_bytes
    )
{
    switch (num_bytes) {
    case (1):
        memcpy(to_tokudb, from_mysql, 1);
        break;
    case (2):
        memcpy(to_tokudb, from_mysql, 2);
        break;
    case (3):
        memcpy(to_tokudb, from_mysql, 3);
        break;
    case (4):
        memcpy(to_tokudb, from_mysql, 4);
        break;
    case (8):
        memcpy(to_tokudb, from_mysql, 8);
        break;
    default:
        memcpy(to_tokudb, from_mysql, num_bytes);
        break;
    }
    return to_tokudb+num_bytes;
}

inline const uchar* unpack_fixed_field(
    uchar* to_mysql,
    const uchar* from_tokudb,
    u_int32_t num_bytes
    )
{
    switch (num_bytes) {
    case (1):
        memcpy(to_mysql, from_tokudb, 1);
        break;
    case (2):
        memcpy(to_mysql, from_tokudb, 2);
        break;
    case (3):
        memcpy(to_mysql, from_tokudb, 3);
        break;
    case (4):
        memcpy(to_mysql, from_tokudb, 4);
        break;
    case (8):
        memcpy(to_mysql, from_tokudb, 8);
        break;
    default:
        memcpy(to_mysql, from_tokudb, num_bytes);
        break;
    }
    return from_tokudb+num_bytes;
}


788
inline uchar* write_var_field(
Zardosht Kasheff's avatar
Zardosht Kasheff committed
789
    uchar* to_tokudb_offset_ptr, //location where offset data is going to be written
790 791 792 793 794
    uchar* to_tokudb_data, // location where data is going to be written
    uchar* to_tokudb_offset_start, //location where offset starts, IS THIS A BAD NAME????
    const uchar * data, // the data to write
    u_int32_t data_length, // length of data to write
    u_int32_t offset_bytes // number of offset bytes
Zardosht Kasheff's avatar
Zardosht Kasheff committed
795 796
    )
{
797
    memcpy(to_tokudb_data, data, data_length);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
798 799 800
    //
    // for offset, we pack the offset where the data ENDS!
    //
801
    u_int32_t offset = to_tokudb_data + data_length - to_tokudb_offset_start;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
802 803 804 805 806 807 808 809 810 811 812 813 814 815
    switch(offset_bytes) {
    case (1):
        to_tokudb_offset_ptr[0] = (uchar)offset;
        break;
    case (2):
        int2store(to_tokudb_offset_ptr,offset);
        break;
    default:
        assert(false);
        break;
    }
    return to_tokudb_data + data_length;
}

816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855
inline u_int32_t get_var_data_length(
    const uchar * from_mysql, 
    u_int32_t mysql_length_bytes 
    ) 
{
    u_int32_t data_length;
    switch(mysql_length_bytes) {
    case(1):
        data_length = from_mysql[0];
        break;
    case(2):
        data_length = uint2korr(from_mysql);
        break;
    default:
        assert(false);
        break;
    }
    return data_length;
}

inline uchar* pack_var_field(
    uchar* to_tokudb_offset_ptr, //location where offset data is going to be written
    uchar* to_tokudb_data, // pointer to where tokudb data should be written
    uchar* to_tokudb_offset_start, //location where data starts, IS THIS A BAD NAME????
    const uchar * from_mysql, // mysql data
    u_int32_t mysql_length_bytes, //number of bytes used to store length in from_mysql
    u_int32_t offset_bytes //number of offset_bytes used in tokudb row
    )
{
    uint data_length = get_var_data_length(from_mysql, mysql_length_bytes);    
    return write_var_field(
        to_tokudb_offset_ptr,
        to_tokudb_data,
        to_tokudb_offset_start,
        from_mysql + mysql_length_bytes,
        data_length,
        offset_bytes
        );
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882
inline void unpack_var_field(
    uchar* to_mysql,
    const uchar* from_tokudb_data,
    u_int32_t from_tokudb_data_len,
    u_int32_t mysql_length_bytes
    )
{
    //
    // store the length
    //
    switch (mysql_length_bytes) {
    case(1):
        to_mysql[0] = (uchar)from_tokudb_data_len;
        break;
    case(2):
        int2store(to_mysql, from_tokudb_data_len);
        break;
    default:
        assert(false);
        break;
    }
    //
    // store the data
    //
    memcpy(to_mysql+mysql_length_bytes, from_tokudb_data, from_tokudb_data_len);
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917
uchar* pack_toku_field_blob(
    uchar* to_tokudb,
    const uchar* from_mysql,
    Field* field
    )
{
    u_int32_t len_bytes = field->row_pack_length();
    u_int32_t length = 0;
    uchar* data_ptr = NULL;
    memcpy(to_tokudb, from_mysql, len_bytes);

    switch (len_bytes) {
    case (1):
        length = (u_int32_t)(*from_mysql);
        break;
    case (2):
        length = uint2korr(from_mysql);
        break;
    case (3):
        length = uint3korr(from_mysql);
        break;
    case (4):
        length = uint4korr(from_mysql);
        break;
    default:
        assert(false);
    }

    if (length > 0) {
        memcpy_fixed((uchar *)(&data_ptr), from_mysql + len_bytes, sizeof(uchar*));
        memcpy(to_tokudb + len_bytes, data_ptr, length);
    }
    return (to_tokudb + len_bytes + length);
}

918

919
static int add_table_to_metadata(const char *name, TABLE* table, DB_TXN* txn) {
920 921 922 923
    int error = 0;
    DBT key;
    DBT val;
    uchar hidden_primary_key = (table->s->primary_key >= MAX_KEY);
924
    assert(txn);
925 926 927 928 929 930 931 932 933 934 935 936
    
    bzero((void *)&key, sizeof(key));
    bzero((void *)&val, sizeof(val));
    key.data = (void *)name;
    key.size = strlen(name) + 1;
    val.data = &hidden_primary_key;
    val.size = sizeof(hidden_primary_key);
    error = metadata_db->put(
        metadata_db,
        txn,
        &key,
        &val,
937
        0
938 939 940 941
        );
    return error;
}

942
static int drop_table_from_metadata(const char *name, DB_TXN* txn) {
943 944 945
    int error = 0;
    DBT key;
    DBT data;
946
    assert(txn);
947 948 949 950 951 952 953 954 955 956 957 958 959
    bzero((void *)&key, sizeof(key));
    bzero((void *)&data, sizeof(data));
    key.data = (void *)name;
    key.size = strlen(name) + 1;
    error = metadata_db->del(
        metadata_db, 
        txn, 
        &key , 
        DB_DELETE_ANY
        );
    return error;
}

960
static int rename_table_in_metadata(const char *from, const char *to, DB_TXN* txn) {
961 962
    int error = 0;
    DBT from_key;
963 964
    DBT to_key;
    DBT val;
965
    assert(txn);
966 967 968 969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984
    
    bzero((void *)&from_key, sizeof(from_key));
    bzero((void *)&to_key, sizeof(to_key));
    bzero((void *)&val, sizeof(val));
    from_key.data = (void *)from;
    from_key.size = strlen(from) + 1;
    to_key.data = (void *)to;
    to_key.size = strlen(to) + 1;
    
    error = metadata_db->getf_set(
        metadata_db, 
        txn, 
        0, 
        &from_key, 
        smart_dbt_metacallback, 
        &val
        );

    if (error) {
985
        goto cleanup;
986 987 988 989 990 991 992
    }

    error = metadata_db->put(
        metadata_db,
        txn,
        &to_key,
        &val,
993
        0
994
        );
995 996 997
    if (error) {
        goto cleanup;
    }
998

999
    error = metadata_db->del(
1000 1001 1002 1003 1004
        metadata_db, 
        txn, 
        &from_key, 
        DB_DELETE_ANY
        );
1005 1006 1007
    if (error) {
        goto cleanup;
    }
1008 1009 1010 1011 1012 1013 1014 1015 1016 1017

    error = 0;

cleanup:
    my_free(val.data, MYF(MY_ALLOW_ZERO_PTR));

    return error;
}


1018
static int check_table_in_metadata(const char *name, bool* table_found, DB_TXN* txn) {
1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042 1043 1044 1045 1046
    int error = 0;
    DBT key;
    pthread_mutex_lock(&tokudb_meta_mutex);
    bzero((void *)&key, sizeof(key));
    key.data = (void *)name;
    key.size = strlen(name) + 1;
    
    error = metadata_db->getf_set(
        metadata_db, 
        txn, 
        0, 
        &key, 
        smart_dbt_do_nothing, 
        NULL
        );

    if (error == 0) {
        *table_found = true;
    }
    else if (error == DB_NOTFOUND){
        *table_found = false;
        error = 0;
    }

    pthread_mutex_unlock(&tokudb_meta_mutex);
    return error;
}

1047 1048 1049 1050 1051 1052 1053 1054 1055 1056
int create_tokudb_trx_data_instance(tokudb_trx_data** out_trx) {
    int error;
    tokudb_trx_data* trx = NULL;
    trx = (tokudb_trx_data *) my_malloc(sizeof(*trx), MYF(MY_ZEROFILL));
    if (!trx) {
        error = ENOMEM;
        goto cleanup;
    }

    *out_trx = trx;
1057
    error = 0;
1058 1059 1060 1061
cleanup:
    return error;
}

1062 1063

inline int tokudb_generate_row(
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1064 1065 1066 1067 1068
    DB *dest_db, 
    DB *src_db,
    DBT *dest_key, 
    DBT *dest_val,
    const DBT *src_key, 
1069
    const DBT *src_val
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1070 1071
    ) 
{
1072 1073
    int error;

Zardosht Kasheff's avatar
Zardosht Kasheff committed
1074 1075 1076 1077 1078
    DB* curr_db = dest_db;
    uchar* row_desc = NULL;
    u_int32_t desc_size;
    uchar* buff = NULL;
    u_int32_t max_key_len = 0;
1079
    
1080
    row_desc = (uchar *)curr_db->descriptor->dbt.data;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1081 1082 1083 1084 1085 1086 1087 1088
    row_desc += (*(u_int32_t *)row_desc);
    desc_size = (*(u_int32_t *)row_desc) - 4;
    row_desc += 4;
    
    if (is_key_pk(row_desc, desc_size)) {
        if (dest_key->flags == DB_DBT_REALLOC && dest_key->data != NULL) {
            free(dest_key->data);
        }
1089
        if (dest_val != NULL) {
1090 1091 1092
            if (dest_val->flags == DB_DBT_REALLOC && dest_val->data != NULL) {
                free(dest_val->data);
            }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1093 1094 1095
        }
        dest_key->data = src_key->data;
        dest_key->size = src_key->size;
1096
        dest_key->flags = 0;
1097
        if (dest_val != NULL) {
1098 1099 1100 1101
            dest_val->data = src_val->data;
            dest_val->size = src_val->size;
            dest_val->flags = 0;
        }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1102 1103 1104
        error = 0;
        goto cleanup;
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1105 1106 1107 1108 1109 1110 1111 1112 1113
    // at this point, we need to create the key/val and set it
    // in the DBTs
    if (dest_key->flags == 0) {
        dest_key->ulen = 0;
        dest_key->size = 0;
        dest_key->data = NULL;
        dest_key->flags = DB_DBT_REALLOC;
    }
    if (dest_key->flags == DB_DBT_REALLOC) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1114 1115
        max_key_len = max_key_size_from_desc(row_desc, desc_size);
        max_key_len += src_key->size;
1116
        
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1117 1118 1119 1120 1121 1122 1123
        if (max_key_len > dest_key->ulen) {
            void* old_ptr = dest_key->data;
            void* new_ptr = NULL;
            new_ptr = realloc(old_ptr, max_key_len);
            assert(new_ptr);
            dest_key->data = new_ptr;
            dest_key->ulen = max_key_len;
1124
        }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1125 1126 1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148 1149 1150 1151

        buff = (uchar *)dest_key->data;
        assert(buff != NULL && max_key_len > 0);
    }
    else {
        assert(false);
    }

    dest_key->size = pack_key_from_desc(
        buff,
        row_desc,
        desc_size,
        src_key,
        src_val
        );
    assert(dest_key->ulen >= dest_key->size);
    if (tokudb_debug & TOKUDB_DEBUG_CHECK_KEY && !max_key_len) {
        max_key_len = max_key_size_from_desc(row_desc, desc_size);
        max_key_len += src_key->size;
    }
    if (max_key_len) {
        assert(max_key_len >= dest_key->size);
    }

    row_desc += desc_size;
    desc_size = (*(u_int32_t *)row_desc) - 4;
    row_desc += 4;
1152
    if (dest_val != NULL) {
1153 1154
        if (!is_key_clustering(row_desc, desc_size)) {
            dest_val->size = 0;
1155 1156
        }
        else {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1157 1158 1159 1160 1161 1162
            uchar* buff = NULL;            
            if (dest_val->flags == 0) {
                dest_val->ulen = 0;
                dest_val->size = 0;
                dest_val->data = NULL;
                dest_val->flags = DB_DBT_REALLOC;
1163
            }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1164
            if (dest_val->flags == DB_DBT_REALLOC){
1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176 1177 1178 1179 1180 1181 1182 1183 1184 1185
                if (dest_val->ulen < src_val->size) {
                    void* old_ptr = dest_val->data;
                    void* new_ptr = NULL;
                    new_ptr = realloc(old_ptr, src_val->size);
                    assert(new_ptr);
                    dest_val->data = new_ptr;
                    dest_val->ulen = src_val->size;
                }
                buff = (uchar *)dest_val->data;
                assert(buff != NULL);
            }
            else {
                assert(false);
            }
            dest_val->size = pack_clustering_val_from_desc(
                buff,
                row_desc,
                desc_size,
                src_val
                );
            assert(dest_val->ulen >= dest_val->size);
1186 1187 1188
        }
    }
    error = 0;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1189
cleanup:
1190 1191 1192
    return error;
}

1193 1194 1195 1196 1197
int generate_row_for_del(
    DB *dest_db, 
    DB *src_db,
    DBT *dest_key,
    const DBT *src_key, 
1198 1199
    const DBT *src_val
    )
1200 1201 1202 1203 1204 1205 1206
{
    return tokudb_generate_row(
        dest_db,
        src_db,
        dest_key,
        NULL,
        src_key,
1207
        src_val
1208 1209 1210 1211 1212 1213 1214 1215 1216 1217
        );
}


int generate_row_for_put(
    DB *dest_db, 
    DB *src_db,
    DBT *dest_key, 
    DBT *dest_val,
    const DBT *src_key, 
1218
    const DBT *src_val
1219 1220 1221 1222 1223 1224 1225 1226
    ) 
{
    return tokudb_generate_row(
        dest_db,
        src_db,
        dest_key,
        dest_val,
        src_key,
1227
        src_val
1228 1229 1230
        );
}

1231

Zardosht Kasheff's avatar
Zardosht Kasheff committed
1232 1233 1234
ha_tokudb::ha_tokudb(handlerton * hton, TABLE_SHARE * table_arg):handler(hton, table_arg) 
    // flags defined in sql\handler.h
{
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1235
    int_table_flags = HA_REC_NOT_IN_SEQ  | HA_NULL_IN_KEY | HA_CAN_INDEX_BLOBS | HA_PRIMARY_KEY_IN_READ_INDEX | 
1236
                    HA_FILE_BASED | HA_AUTO_PART_KEY | HA_TABLE_SCAN_ON_INDEX;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1237 1238
    alloc_ptr = NULL;
    rec_buff = NULL;
1239
    rec_update_buff = NULL;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1240
    transaction = NULL;
1241
    is_fast_alter_running = false;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1242 1243 1244 1245 1246 1247 1248 1249 1250 1251 1252 1253 1254 1255
    cursor = NULL;
    fixed_cols_for_query = NULL;
    var_cols_for_query = NULL;
    num_fixed_cols_for_query = 0;
    num_var_cols_for_query = 0;
    unpack_entire_row = true;
    read_blobs = false;
    read_key = false;
    added_rows = 0;
    deleted_rows = 0;
    last_dup_key = UINT_MAX;
    using_ignore = 0;
    last_cursor_error = 0;
    range_lock_grabbed = false;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1256 1257
    blob_buff = NULL;
    num_blob_bytes = 0;
1258 1259
    delay_updating_ai_metadata = false;
    ai_metadata_update_required = false;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1260 1261
    bzero(mult_key_dbt, sizeof(mult_key_dbt));
    bzero(mult_rec_dbt, sizeof(mult_rec_dbt));
1262 1263 1264
    loader = NULL;
    abort_loader = false;
    bzero(&lc, sizeof(lc));
1265
    lock.type = TL_IGNORE;
1266
    for (u_int32_t i = 0; i < MAX_KEY+1; i++) {
1267
        mult_put_flags[i] = 0;
1268
        mult_del_flags[i] = DB_DELETE_ANY;
1269 1270
        mult_dbt_flags[i] = DB_DBT_REALLOC;
    }
1271 1272
    num_DBs_locked_in_bulk = false;
    lock_count = 0;
1273
    use_write_locks = false;
1274 1275 1276 1277 1278 1279 1280
    range_query_buff = NULL;
    size_range_query_buff = 0;
    bytes_used_in_range_query_buff = 0;
    curr_range_query_buff_offset = 0;
    doing_bulk_fetch = false;
    prelocked_left_range_size = 0;
    prelocked_right_range_size = 0;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1281 1282
}

1283 1284 1285
ha_tokudb::~ha_tokudb() {
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
1286 1287 1288 1289 1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303 1304 1305 1306 1307 1308 1309
//
// states if table has an auto increment column, if so, sets index where auto inc column is to index
// Parameters:
//      [out]   index - if auto inc exists, then this param is set to where it exists in table, if not, then unchanged
// Returns:
//      true if auto inc column exists, false otherwise
//
bool ha_tokudb::has_auto_increment_flag(uint* index) {
    //
    // check to see if we have auto increment field
    //
    bool ai_found = false;
    uint ai_index = 0;
    for (uint i = 0; i < table_share->fields; i++, ai_index++) {
        Field* field = table->field[i];
        if (field->flags & AUTO_INCREMENT_FLAG) {
            ai_found = true;
            *index = ai_index;
            break;
        }
    }
    return ai_found;
}

1310
int open_status_dictionary(DB** ptr, const char* name, DB_TXN* txn) {
1311 1312 1313 1314 1315 1316 1317 1318 1319 1320 1321 1322 1323 1324 1325 1326 1327 1328 1329 1330 1331 1332 1333 1334 1335
    int error;
    char* newname = NULL;
    uint open_mode = DB_THREAD;
    newname = (char *)my_malloc(
        get_max_dict_name_path_length(name), 
        MYF(MY_WME)
        );
    if (newname == NULL) {
        error = ENOMEM;
        goto cleanup;
    }
    make_name(newname, name, "status");
    if (tokudb_debug & TOKUDB_DEBUG_OPEN) {
        TOKUDB_TRACE("open:%s\n", newname);
    }
    error = db_create(ptr, db_env, 0);
    if (error) { goto cleanup; }
    
    error = (*ptr)->open((*ptr), txn, newname, NULL, DB_BTREE, open_mode, 0);
    if (error) { 
        goto cleanup; 
    }
cleanup:
    if (error) {
        if (*ptr) {
1336 1337
            int r = (*ptr)->close(*ptr, 0);
            assert(r==0);
1338 1339 1340 1341 1342 1343 1344
            *ptr = NULL;
        }
    }
    my_free(newname, MYF(MY_ALLOW_ZERO_PTR));
    return error;
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
1345
int ha_tokudb::open_main_dictionary(const char* name, bool is_read_only, DB_TXN* txn) {
1346 1347
    int error;    
    char* newname = NULL;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1348
    uint open_flags = (is_read_only ? DB_RDONLY : 0) | DB_THREAD;
1349 1350 1351 1352 1353 1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366 1367

    assert(share->file == NULL);
    assert(share->key_file[primary_key] == NULL);

    newname = (char *)my_malloc(
        get_max_dict_name_path_length(name),
        MYF(MY_WME|MY_ZEROFILL)
        );
    if (newname == NULL) { 
        error = ENOMEM;
        goto exit;
    }
    make_name(newname, name, "main");

    error = db_create(&share->file, db_env, 0);
    if (error) {
        goto exit;
    }
    share->key_file[primary_key] = share->file;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1368

1369 1370 1371 1372 1373 1374 1375 1376
    error = share->file->open(share->file, txn, newname, NULL, DB_BTREE, open_flags, 0);
    if (error) {
        goto exit;
    }
    
    if (tokudb_debug & TOKUDB_DEBUG_OPEN) {
        TOKUDB_TRACE("open:%s:file=%p\n", newname, share->file);
    }
1377 1378

    error = 0;
1379 1380 1381
exit:
    if (error) {
        if (share->file) {
1382
            int r = share->file->close(
1383 1384 1385
                share->file,
                0
                );
1386
            assert(r==0);
1387 1388 1389 1390 1391 1392 1393
            share->file = NULL;
            share->key_file[primary_key] = NULL;
        }
    }
    my_free(newname, MYF(MY_ALLOW_ZERO_PTR));
    return error;
}
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1394

Zardosht Kasheff's avatar
Zardosht Kasheff committed
1395 1396 1397
//
// Open a secondary table, the key will be a secondary index, the data will be a primary key
//
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1398
int ha_tokudb::open_secondary_dictionary(DB** ptr, KEY* key_info, const char* name, bool is_read_only, DB_TXN* txn) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1399
    int error = ENOSYS;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1400
    char dict_name[MAX_DICT_NAME_LEN];
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1401
    uint open_flags = (is_read_only ? DB_RDONLY : 0) | DB_THREAD;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1402
    char* newname = NULL;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1403 1404 1405 1406 1407 1408
    uint newname_len = 0;
    
    sprintf(dict_name, "key-%s", key_info->name);

    newname_len = get_max_dict_name_path_length(name);
    newname = (char *)my_malloc(newname_len, MYF(MY_WME|MY_ZEROFILL));
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1409 1410 1411 1412
    if (newname == NULL) {
        error = ENOMEM;
        goto cleanup;
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1413
    make_name(newname, name, dict_name);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1414 1415 1416 1417 1418 1419


    if ((error = db_create(ptr, db_env, 0))) {
        my_errno = error;
        goto cleanup;
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1420

Zardosht Kasheff's avatar
Zardosht Kasheff committed
1421

1422
    if ((error = (*ptr)->open(*ptr, txn, newname, NULL, DB_BTREE, open_flags, 0))) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1423 1424 1425 1426 1427 1428 1429
        my_errno = error;
        goto cleanup;
    }
    if (tokudb_debug & TOKUDB_DEBUG_OPEN) {
        TOKUDB_TRACE("open:%s:file=%p\n", newname, *ptr);
    }
cleanup:
1430 1431
    if (error) {
        if (*ptr) {
1432 1433
            int r = (*ptr)->close(*ptr, 0);
            assert(r==0);
1434 1435 1436
            *ptr = NULL;
        }
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1437
    my_free(newname, MYF(MY_ALLOW_ZERO_PTR));
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1438 1439 1440
    return error;
}

1441
int initialize_col_pack_info(KEY_AND_COL_INFO* kc_info, TABLE_SHARE* table_share, uint keynr) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1442 1443 1444 1445
    int error = ENOSYS;
    //
    // set up the cp_info
    //
1446 1447
    assert(kc_info->cp_info[keynr] == NULL);
    kc_info->cp_info[keynr] = (COL_PACK_INFO *)my_malloc(
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1448 1449 1450
        table_share->fields*sizeof(COL_PACK_INFO), 
        MYF(MY_WME | MY_ZEROFILL)
        );
1451
    if (kc_info->cp_info[keynr] == NULL) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1452 1453 1454 1455 1456 1457 1458
        error = ENOMEM;
        goto exit;
    }
    {
    u_int32_t curr_fixed_offset = 0;
    u_int32_t curr_var_index = 0;
    for (uint j = 0; j < table_share->fields; j++) {
1459
        COL_PACK_INFO* curr = &kc_info->cp_info[keynr][j];
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1460 1461 1462 1463
        //
        // need to set the offsets / indexes
        // offsets are calculated AFTER the NULL bytes
        //
1464 1465
        if (!bitmap_is_set(&kc_info->key_filters[keynr],j)) {
            if (kc_info->field_lengths[j]) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1466
                curr->col_pack_val = curr_fixed_offset;
1467
                curr_fixed_offset += kc_info->field_lengths[j];
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1468
            }
1469
            else if (kc_info->length_bytes[j]) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1470 1471 1472 1473 1474 1475 1476 1477 1478
                curr->col_pack_val = curr_var_index;
                curr_var_index++;
            }
        }
    }
    
    //
    // set up the mcp_info
    //
1479
    kc_info->mcp_info[keynr].fixed_field_size = get_fixed_field_size(
1480
        kc_info,
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1481 1482 1483
        table_share,
        keynr
        );
1484 1485
    kc_info->mcp_info[keynr].len_of_offsets = get_len_of_offsets(
        kc_info,
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1486 1487 1488 1489 1490 1491 1492 1493 1494 1495
        table_share,
        keynr
        );

    error = 0;
    }
exit:
    return error;
}

1496 1497 1498 1499 1500 1501 1502 1503
// reset the kc_info state at keynr
static void reset_key_and_col_info(KEY_AND_COL_INFO *kc_info, uint keynr) {
    bitmap_clear_all(&kc_info->key_filters[keynr]);
    my_free(kc_info->cp_info[keynr], MYF(MY_ALLOW_ZERO_PTR));
    kc_info->cp_info[keynr] = NULL;
    kc_info->mcp_info[keynr] = (MULTI_COL_PACK_INFO) { 0, 0 };
}

1504
int initialize_key_and_col_info(TABLE_SHARE* table_share, TABLE* table, KEY_AND_COL_INFO* kc_info, uint hidden_primary_key, uint primary_key) {
1505
    int error = 0;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1506
    u_int32_t curr_blob_field_index = 0;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1507
    u_int32_t max_var_bytes = 0;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1508 1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522
    //
    // fill in the field lengths. 0 means it is a variable sized field length
    // fill in length_bytes, 0 means it is fixed or blob
    //
    for (uint i = 0; i < table_share->fields; i++) {
        Field* field = table_share->field[i];
        TOKU_TYPE toku_type = mysql_to_toku_type(field);
        uint32 pack_length = 0;
        switch (toku_type) {
        case toku_type_int:
        case toku_type_double:
        case toku_type_float:
        case toku_type_fixbinary:
        case toku_type_fixstring:
            pack_length = field->pack_length();
1523 1524
            assert(pack_length < 1<<16);
            kc_info->field_lengths[i] = (u_int16_t)pack_length;
1525
            kc_info->length_bytes[i] = 0;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1526 1527
            break;
        case toku_type_blob:
1528 1529 1530
            kc_info->field_lengths[i] = 0;
            kc_info->length_bytes[i] = 0;
            kc_info->blob_fields[curr_blob_field_index] = i;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1531 1532 1533 1534 1535 1536 1537
            curr_blob_field_index++;
            break;
        case toku_type_varstring:
        case toku_type_varbinary:
            //
            // meaning it is variable sized
            //
1538 1539
            kc_info->field_lengths[i] = 0;
            kc_info->length_bytes[i] = (uchar)((Field_varstring *)field)->length_bytes;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1540
            max_var_bytes += field->field_length;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1541 1542 1543 1544 1545
            break;
        default:
            assert(false);
        }
    }
1546
    kc_info->num_blobs = curr_blob_field_index;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1547

Zardosht Kasheff's avatar
Zardosht Kasheff committed
1548 1549 1550 1551 1552 1553
    //
    // initialize share->num_offset_bytes
    // because MAX_REF_LENGTH is 65536, we
    // can safely set num_offset_bytes to 1 or 2
    //
    if (max_var_bytes < 256) {
1554
        kc_info->num_offset_bytes = 1;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1555 1556
    }
    else {
1557
        kc_info->num_offset_bytes = 2;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1558 1559 1560
    }


Zardosht Kasheff's avatar
Zardosht Kasheff committed
1561 1562 1563 1564 1565 1566 1567
    for (uint i = 0; i < table_share->keys + test(hidden_primary_key); i++) {
        //
        // do the cluster/primary key filtering calculations
        //
        if (! (i==primary_key && hidden_primary_key) ){        
            if ( i == primary_key ) {
                set_key_filter(
1568
                    &kc_info->key_filters[primary_key],
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1569 1570 1571 1572 1573
                    &table_share->key_info[primary_key],
                    table,
                    true
                    );
            }
1574
            else {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1575
                set_key_filter(
1576
                    &kc_info->key_filters[i],
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1577 1578 1579 1580 1581 1582
                    &table_share->key_info[i],
                    table,
                    true
                    );
                if (!hidden_primary_key) {
                    set_key_filter(
1583
                        &kc_info->key_filters[i],
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1584 1585 1586 1587 1588 1589 1590 1591
                        &table_share->key_info[primary_key],
                        table,
                        true
                        );
                }
            }
        }
        if (i == primary_key || table_share->key_info[i].flags & HA_CLUSTERING) {
1592
            error = initialize_col_pack_info(kc_info,table_share,i);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1593 1594 1595 1596 1597 1598
            if (error) {
                goto exit;
            }
        }

    }
1599 1600 1601
exit:
    return error;
}
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1602

1603 1604
bool ha_tokudb::can_replace_into_be_fast(TABLE_SHARE* table_share, KEY_AND_COL_INFO* kc_info, uint pk) {
    uint curr_num_DBs = table_share->keys + test(hidden_primary_key);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1605 1606 1607 1608 1609 1610
    bool ret_val;
    if (curr_num_DBs == 1) {
        ret_val = true;
        goto exit;
    }
    ret_val = true;
1611 1612 1613 1614 1615 1616 1617 1618 1619 1620 1621 1622 1623 1624 1625 1626
    for (uint curr_index = 0; curr_index < table_share->keys; curr_index++) {
        if (curr_index == pk) continue;
        KEY* curr_key_info = &table_share->key_info[curr_index];
        for (uint i = 0; i < curr_key_info->key_parts; i++) {
            uint16 curr_field_index = curr_key_info->key_part[i].field->field_index;
            if (!bitmap_is_set(&kc_info->key_filters[curr_index],curr_field_index)) {
                ret_val = false;
                goto exit;
            }
            if (bitmap_is_set(&kc_info->key_filters[curr_index], curr_field_index) &&
                !bitmap_is_set(&kc_info->key_filters[pk], curr_field_index)) {
                ret_val = false;
                goto exit;
            }
            
        }
1627
    }
1628 1629 1630
exit:
    return ret_val;
}
1631 1632 1633 1634 1635 1636 1637 1638 1639

int ha_tokudb::initialize_share(
    const char* name,
    int mode
    )
{
    int error = 0;
    u_int64_t num_rows = 0;
    bool table_exists;
1640 1641
    DB_TXN* txn = NULL;
    bool do_commit = false;
1642
    THD* thd = ha_thd();
1643 1644
    tokudb_trx_data *trx = NULL;
    trx = (tokudb_trx_data *) thd_data_get(ha_thd(), tokudb_hton->slot);
1645
    if (thd_sql_command(thd) == SQLCOM_CREATE_TABLE && trx && trx->sub_sp_level) {
1646 1647 1648 1649 1650 1651 1652 1653
        txn = trx->sub_sp_level;
    }
    else {
        do_commit = true;
        error = db_env->txn_begin(db_env, 0, &txn, 0);
        if (error) { goto exit; }
    }

1654 1655 1656
    DBUG_PRINT("info", ("share->use_count %u", share->use_count));

    table_exists = true;
1657
    error = check_table_in_metadata(name, &table_exists, txn);
1658 1659 1660 1661 1662 1663 1664 1665 1666

    if (error) {
        goto exit;
    }
    if (!table_exists) {
        sql_print_error("table %s does not exist in metadata, was it moved from someplace else? Not opening table", name);
        error = HA_ADMIN_FAILED;
        goto exit;
    }
1667

1668
    error = get_status(txn);
1669 1670 1671
    if (error) {
        goto exit;
    }
1672
    if (share->version != HA_TOKU_VERSION) {
1673 1674 1675 1676 1677 1678 1679 1680 1681
        error = ENOSYS;
        goto exit;
    }

    //
    // verify frm file is what we expect it to be
    // only for tables that are not partitioned
    //
    if (table->part_info == NULL) {
1682
        error = verify_frm_data(table->s->path.str, txn);
1683 1684 1685 1686
        if (error) {
            goto exit;
        }
    }
1687 1688 1689 1690 1691 1692 1693
    error = initialize_key_and_col_info(
        table_share,
        table, 
        &share->kc_info,
        hidden_primary_key,
        primary_key
        );
1694
    if (error) { goto exit; }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1695
    
1696
    error = open_main_dictionary(name, mode == O_RDONLY, txn);
1697 1698 1699
    if (error) { goto exit; }

    share->has_unique_keys = false;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1700
    /* Open other keys;  These are part of the share structure */
1701 1702 1703 1704
    for (uint i = 0; i < table_share->keys + test(hidden_primary_key); i++) {
        if (table_share->key_info[i].flags & HA_NOSAME) {
            share->has_unique_keys = true;
        }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1705
        if (i != primary_key) {
1706
            error = open_secondary_dictionary(
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1707 1708 1709
                &share->key_file[i],
                &table_share->key_info[i],
                name,
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1710
                mode == O_RDONLY,
1711
                txn
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1712 1713 1714 1715 1716 1717
                );
            if (error) {
                goto exit;
            }
        }
    }
1718 1719 1720 1721 1722 1723
    share->replace_into_fast = can_replace_into_be_fast(
        table_share, 
        &share->kc_info, 
        primary_key
        );
        
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1724 1725 1726 1727 1728 1729 1730 1731 1732 1733 1734 1735 1736 1737 1738
    if (!hidden_primary_key) {
        //
        // We need to set the ref_length to start at 5, to account for
        // the "infinity byte" in keys, and for placing the DBT size in the first four bytes
        //
        ref_length = sizeof(u_int32_t) + sizeof(uchar);
        KEY_PART_INFO *key_part = table->key_info[primary_key].key_part;
        KEY_PART_INFO *end = key_part + table->key_info[primary_key].key_parts;
        for (; key_part != end; key_part++) {
            ref_length += key_part->field->max_packed_col_length(key_part->length);
        }
        share->status |= STATUS_PRIMARY_KEY_INIT;
    }
    share->ref_length = ref_length;

1739
    error = estimate_num_rows(share->file,&num_rows, txn);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1740 1741 1742 1743 1744 1745 1746 1747 1748 1749 1750 1751 1752 1753 1754 1755 1756
    //
    // estimate_num_rows should not fail under normal conditions
    //
    if (error == 0) {
        share->rows = num_rows;
    }
    else {
        goto exit;
    }
    //
    // initialize auto increment data
    //
    share->has_auto_inc = has_auto_increment_flag(&share->ai_field_index);
    if (share->has_auto_inc) {
        init_auto_increment();
    }

1757
    if (may_table_be_empty(NULL)) {
1758 1759 1760 1761 1762 1763
        share->try_table_lock = true;
    }
    else {
        share->try_table_lock = false;
    }

1764 1765
    share->num_DBs = table_share->keys + test(hidden_primary_key);

Zardosht Kasheff's avatar
Zardosht Kasheff committed
1766 1767
    error = 0;
exit:
1768 1769 1770
    if (do_commit && txn) {
        commit_txn(txn,0);
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1771 1772 1773
    return error;
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
1774 1775


Zardosht Kasheff's avatar
Zardosht Kasheff committed
1776 1777 1778 1779 1780 1781 1782 1783 1784 1785 1786
//
// Creates and opens a handle to a table which already exists in a tokudb
// database.
// Parameters:
//      [in]   name - table name
//             mode - seems to specify if table is read only
//             test_if_locked - unused
// Returns:
//      0 on success
//      1 on error
//
1787
int ha_tokudb::open(const char *name, int mode, uint test_if_locked) {
1788
    TOKUDB_DBUG_ENTER("ha_tokudb::open %p %s", this, name);
1789
    THD* thd = ha_thd();
1790

Zardosht Kasheff's avatar
Zardosht Kasheff committed
1791 1792
    int error = 0;
    int ret_val = 0;
1793
    uint curr_num_DBs = 0;
1794

Zardosht Kasheff's avatar
Zardosht Kasheff committed
1795 1796 1797
    transaction = NULL;
    cursor = NULL;

Zardosht Kasheff's avatar
Zardosht Kasheff committed
1798

1799 1800 1801 1802 1803 1804
    /* Open primary key */
    hidden_primary_key = 0;
    if ((primary_key = table_share->primary_key) >= MAX_KEY) {
        // No primary key
        primary_key = table_share->keys;
        key_used_on_scan = MAX_KEY;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1805 1806
        hidden_primary_key = TOKUDB_HIDDEN_PRIMARY_KEY_LENGTH;
        ref_length = TOKUDB_HIDDEN_PRIMARY_KEY_LENGTH + sizeof(u_int32_t);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1807
    } 
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1808
    else {
1809
        key_used_on_scan = primary_key;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1810
    }
1811
    curr_num_DBs = table_share->keys + test(hidden_primary_key);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1812

1813
    /* Need some extra memory in case of packed keys */
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1814
    // the "+ 1" is for the first byte that states +/- infinity
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1815 1816
    // multiply everything by 2 to account for clustered keys having a key and primary key together
    max_key_length = 2*(table_share->max_key_length + MAX_REF_PARTS * 3 + sizeof(uchar));
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1817 1818 1819
    alloc_ptr = my_multi_malloc(MYF(MY_WME),
        &key_buff, max_key_length, 
        &key_buff2, max_key_length, 
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1820
        &key_buff3, max_key_length, 
1821 1822
        &prelocked_left_range, max_key_length, 
        &prelocked_right_range, max_key_length, 
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1823 1824 1825 1826 1827 1828 1829 1830
        &primary_key_buff, (hidden_primary_key ? 0 : max_key_length),
        &fixed_cols_for_query, table_share->fields*sizeof(u_int32_t),
        &var_cols_for_query, table_share->fields*sizeof(u_int32_t),
        NullS
        );
    if (alloc_ptr == NULL) {
        ret_val = 1;
        goto exit;
1831 1832
    }

1833 1834 1835 1836 1837 1838 1839
    size_range_query_buff = get_tokudb_read_buf_size(thd);
    range_query_buff = (uchar *)my_malloc(size_range_query_buff, MYF(MY_WME));
    if (range_query_buff == NULL) {
        ret_val = 1;
        goto exit;
    }

Zardosht Kasheff's avatar
Zardosht Kasheff committed
1840
    alloced_rec_buff_length = table_share->rec_buff_length + table_share->fields;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1841
    rec_buff = (uchar *) my_malloc(alloced_rec_buff_length, MYF(MY_WME));
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1842 1843 1844
    if (rec_buff == NULL) {
        ret_val = 1;
        goto exit;
1845
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1846

1847 1848 1849 1850 1851 1852 1853
    alloced_update_rec_buff_length = alloced_rec_buff_length;
    rec_update_buff = (uchar *) my_malloc(alloced_update_rec_buff_length, MYF(MY_WME));
    if (rec_update_buff == NULL) {
        ret_val = 1;
        goto exit;
    }

Zardosht Kasheff's avatar
Zardosht Kasheff committed
1854 1855
    for (u_int32_t i = 0; i < sizeof(mult_key_dbt)/sizeof(mult_key_dbt[0]); i++) {
        mult_key_dbt[i].flags = DB_DBT_REALLOC;
1856 1857 1858
    }

    for (u_int32_t i = 0; i < curr_num_DBs; i++) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1859
        mult_rec_dbt[i].flags = DB_DBT_REALLOC;
1860 1861
    }

Zardosht Kasheff's avatar
Zardosht Kasheff committed
1862 1863 1864 1865 1866
    /* Init shared structure */
    share = get_share(name, table_share);
    if (share == NULL) {
        ret_val = 1;
        goto exit;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1867 1868
    }

1869 1870 1871 1872
    thr_lock_data_init(&share->lock, &lock, NULL);

    /* Fill in shared structure, if needed */
    pthread_mutex_lock(&share->mutex);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1873
    if (tokudb_debug & TOKUDB_DEBUG_OPEN) {
1874 1875
        TOKUDB_TRACE("tokudbopen:%p:share=%p:file=%p:table=%p:table->s=%p:%d\n", 
                     this, share, share->file, table, table->s, share->use_count);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1876
    }
1877
    if (!share->use_count++) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1878 1879 1880 1881 1882
        ret_val = initialize_share(
            name,
            mode
            );
        if (ret_val) {
1883
            free_share(share, 1);
1884
            goto exit;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1885
        }
1886 1887 1888 1889
    }
    ref_length = share->ref_length;     // If second open
    pthread_mutex_unlock(&share->mutex);

Zardosht Kasheff's avatar
Zardosht Kasheff committed
1890
    key_read = false;
1891
    stats.block_size = 1<<20;    // QQQ Tokudb DB block size
1892

Zardosht Kasheff's avatar
Zardosht Kasheff committed
1893 1894
    init_hidden_prim_key_info();

1895 1896
    info(HA_STATUS_NO_LOCK | HA_STATUS_VARIABLE | HA_STATUS_CONST);

Zardosht Kasheff's avatar
Zardosht Kasheff committed
1897 1898
exit:
    if (ret_val) {
1899 1900
        my_free(range_query_buff, MYF(MY_ALLOW_ZERO_PTR));
        range_query_buff = NULL;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1901 1902 1903 1904
        my_free(alloc_ptr, MYF(MY_ALLOW_ZERO_PTR));
        alloc_ptr = NULL;
        my_free(rec_buff, MYF(MY_ALLOW_ZERO_PTR));
        rec_buff = NULL;
1905 1906
        my_free(rec_update_buff, MYF(MY_ALLOW_ZERO_PTR));
        rec_update_buff = NULL;
1907
        
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1908 1909 1910 1911 1912
        if (error) {
            my_errno = error;
        }
    }
    TOKUDB_DBUG_RETURN(ret_val);
1913 1914
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
1915 1916 1917 1918 1919 1920 1921 1922 1923
//
// estimate the number of rows in a DB
// Parameters:
//      [in]    db - DB whose number of rows will be estimated
//      [out]   num_rows - number of estimated rows in db
// Returns:
//      0 on success
//      error otherwise
//
1924
int ha_tokudb::estimate_num_rows(DB* db, u_int64_t* num_rows, DB_TXN* txn) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1925 1926 1927
    int error = ENOSYS;
    DBC* crsr = NULL;
    bool do_commit = false;
1928
    DB_BTREE_STAT64 dict_stats;
1929
    DB_TXN* txn_to_use = NULL;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1930

1931 1932
    if (txn == NULL) {
        error = db_env->txn_begin(db_env, 0, &txn_to_use, DB_READ_UNCOMMITTED);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1933 1934 1935
        if (error) goto cleanup;
        do_commit = true;
    }
1936 1937 1938
    else {
        txn_to_use = txn;
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1939

1940 1941
    error = db->stat64(
        share->file, 
1942
        txn_to_use, 
1943
        &dict_stats
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1944
        );
1945
    if (error) { goto cleanup; }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1946

1947
    *num_rows = dict_stats.bt_ndata;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1948 1949
    error = 0;
cleanup:
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1950
    if (crsr != NULL) {
1951 1952
        int r = crsr->c_close(crsr);
        assert(r==0);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1953 1954
        crsr = NULL;
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1955
    if (do_commit) {
1956 1957
        commit_txn(txn_to_use, 0);
        txn_to_use = NULL;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1958
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
1959 1960 1961
    return error;
}

1962

1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005
int ha_tokudb::write_to_status(DB* db, HA_METADATA_KEY curr_key_data, void* data, uint size, DB_TXN* txn ){
    return write_metadata(db, &curr_key_data, sizeof(curr_key_data), data, size, txn);
}


int ha_tokudb::remove_metadata(DB* db, void* key_data, uint key_size, DB_TXN* transaction){
    int error;
    DBT key;
    DB_TXN* txn = NULL;
    bool do_commit = false;
    //
    // transaction to be used for putting metadata into status.tokudb
    //
    if (transaction == NULL) {
        error = db_env->txn_begin(db_env, 0, &txn, 0);
        if (error) { 
            goto cleanup;
        }
        do_commit = true;
    }
    else {
        txn = transaction;
    }

    bzero(&key, sizeof(key));
    key.data = key_data;
    key.size = key_size;
    error = db->del(db, txn, &key, DB_DELETE_ANY);
    if (error) { 
        goto cleanup; 
    }
    
    error = 0;
cleanup:
    if (do_commit && txn) {
        if (!error) {
            commit_txn(txn, DB_TXN_NOSYNC);
        }
        else {
            abort_txn(txn);
        }
    }
    return error;
2006 2007
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
2008 2009 2010
//
// helper function to write a piece of metadata in to status.tokudb
//
2011
int ha_tokudb::write_metadata(DB* db, void* key_data, uint key_size, void* val_data, uint val_size, DB_TXN* transaction ){
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2012 2013 2014 2015
    int error;
    DBT key;
    DBT value;
    DB_TXN* txn = NULL;
2016
    bool do_commit = false;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2017 2018 2019
    //
    // transaction to be used for putting metadata into status.tokudb
    //
2020 2021 2022 2023 2024 2025 2026 2027 2028
    if (transaction == NULL) {
        error = db_env->txn_begin(db_env, 0, &txn, 0);
        if (error) { 
            goto cleanup;
        }
        do_commit = true;
    }
    else {
        txn = transaction;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2029 2030 2031 2032
    }

    bzero(&key, sizeof(key));
    bzero(&value, sizeof(value));
2033 2034 2035 2036
    key.data = key_data;
    key.size = key_size;
    value.data = val_data;
    value.size = val_size;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2037 2038 2039 2040 2041 2042 2043
    error = db->put(db, txn, &key, &value, 0);
    if (error) { 
        goto cleanup; 
    }
    
    error = 0;
cleanup:
2044
    if (do_commit && txn) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2045
        if (!error) {
2046
            commit_txn(txn, DB_TXN_NOSYNC);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2047 2048
        }
        else {
2049
            abort_txn(txn);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2050 2051 2052 2053 2054
        }
    }
    return error;
}

2055 2056 2057 2058 2059 2060 2061 2062 2063 2064 2065 2066 2067 2068 2069 2070 2071 2072
int ha_tokudb::write_frm_data(DB* db, DB_TXN* txn, const char* frm_name) {
    uchar* frm_data = NULL;
    size_t frm_len = 0;
    int error = 0;
    TOKUDB_DBUG_ENTER("ha_tokudb::write_frm_data, %s", frm_name);

    error = readfrm(frm_name,&frm_data,&frm_len);
    if (error) { goto cleanup; }
    
    error = write_to_status(db,hatoku_frm_data,frm_data,(uint)frm_len, txn);
    if (error) { goto cleanup; }

    error = 0;
cleanup:
    my_free(frm_data, MYF(MY_ALLOW_ZERO_PTR));
    TOKUDB_DBUG_RETURN(error);
}

2073 2074 2075 2076
static int
smart_dbt_callback_verify_frm (DBT const *key, DBT  const *row, void *context) {
    DBT* stored_frm = (DBT *)context;
    stored_frm->size = row->size;
2077 2078 2079
    stored_frm->data = (uchar *)my_malloc(row->size, MYF(MY_WME));
    assert(stored_frm->data);
    memcpy(stored_frm->data, row->data, row->size);
2080 2081 2082
    return 0;
}

2083
int ha_tokudb::verify_frm_data(const char* frm_name, DB_TXN* txn) {
2084 2085
    uchar* mysql_frm_data = NULL;
    size_t mysql_frm_len = 0;
2086
    DBT key, stored_frm;
2087 2088 2089 2090 2091
    int error = 0;
    HA_METADATA_KEY curr_key = hatoku_frm_data;
    TOKUDB_DBUG_ENTER("ha_tokudb::verify_frm_data %s", frm_name);

    bzero(&key, sizeof(key));
2092
    bzero(&stored_frm, sizeof(&stored_frm));
2093 2094 2095 2096 2097 2098
    // get the frm data from MySQL
    error = readfrm(frm_name,&mysql_frm_data,&mysql_frm_len);
    if (error) { goto cleanup; }

    key.data = &curr_key;
    key.size = sizeof(curr_key);
2099
    error = share->status_block->getf_set(
2100
        share->status_block, 
2101 2102
        txn,
        0,
2103
        &key, 
2104 2105
        smart_dbt_callback_verify_frm, 
        &stored_frm
2106 2107 2108 2109 2110 2111 2112 2113 2114 2115 2116 2117 2118 2119
        );
    if (error == DB_NOTFOUND) {
        // if not found, write it
        error = write_frm_data(
            share->status_block,
            txn,
            frm_name
            );
        goto cleanup;
    }
    else if (error) {
        goto cleanup;
    }

2120 2121
    if (stored_frm.size != mysql_frm_len || 
        memcmp(stored_frm.data, mysql_frm_data, stored_frm.size))
2122 2123 2124 2125
    {
        error = HA_ERR_TABLE_DEF_CHANGED;
        goto cleanup;
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2126

2127 2128 2129
    error = 0;
cleanup:
    my_free(mysql_frm_data, MYF(MY_ALLOW_ZERO_PTR));
2130
    my_free(stored_frm.data, MYF(MY_ALLOW_ZERO_PTR));
2131 2132
    TOKUDB_DBUG_RETURN(error);
}
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2133 2134 2135 2136 2137 2138 2139 2140 2141 2142 2143

//
// Updates status.tokudb with a new max value used for the auto increment column
// Parameters:
//      [in]    db - this will always be status.tokudb
//              val - value to store
//  Returns:
//      0 on success, error otherwise
//
//
int ha_tokudb::update_max_auto_inc(DB* db, ulonglong val){
2144
    return write_to_status(db,hatoku_max_ai,&val,sizeof(val), NULL);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2145 2146 2147 2148 2149 2150 2151 2152 2153 2154 2155 2156 2157
}

//
// Writes the initial auto increment value, as specified by create table
// so if a user does "create table t1 (a int auto_increment, primary key (a)) auto_increment=100",
// then the value 100 will be stored here in val
// Parameters:
//      [in]    db - this will always be status.tokudb
//              val - value to store
//  Returns:
//      0 on success, error otherwise
//
//
2158 2159
int ha_tokudb::write_auto_inc_create(DB* db, ulonglong val, DB_TXN* txn){
    return write_to_status(db,hatoku_ai_create_value,&val,sizeof(val), txn);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2160
}
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2161 2162


Zardosht Kasheff's avatar
Zardosht Kasheff committed
2163 2164 2165
//
// Closes a handle to a table. 
//
2166
int ha_tokudb::close(void) {
2167
    TOKUDB_DBUG_ENTER("ha_tokudb::close %p", this);
2168
    TOKUDB_DBUG_RETURN(__close(0));
2169 2170 2171
}

int ha_tokudb::__close(int mutex_is_locked) {
2172
    TOKUDB_DBUG_ENTER("ha_tokudb::__close %p", this);
2173
    if (tokudb_debug & TOKUDB_DEBUG_OPEN) 
2174
        TOKUDB_TRACE("close:%p\n", this);
2175
    my_free(rec_buff, MYF(MY_ALLOW_ZERO_PTR));
2176
    my_free(rec_update_buff, MYF(MY_ALLOW_ZERO_PTR));
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2177
    my_free(blob_buff, MYF(MY_ALLOW_ZERO_PTR));
2178
    my_free(alloc_ptr, MYF(MY_ALLOW_ZERO_PTR));
2179
    my_free(range_query_buff, MYF(MY_ALLOW_ZERO_PTR));
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2180
    for (u_int32_t i = 0; i < sizeof(mult_rec_dbt)/sizeof(mult_rec_dbt[0]); i++) {
2181 2182 2183 2184
        if (mult_rec_dbt[i].flags == DB_DBT_REALLOC &&
                mult_rec_dbt[i].data != NULL) {
            free(mult_rec_dbt[i].data);
        }
2185
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2186
    for (u_int32_t i = 0; i < sizeof(mult_key_dbt)/sizeof(mult_key_dbt[0]); i++) {
2187 2188 2189 2190
        if (mult_key_dbt[i].flags == DB_DBT_REALLOC &&
                mult_key_dbt[i].data != NULL) {
            free(mult_key_dbt[i].data);
        }
2191
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2192
    rec_buff = NULL;
2193
    rec_update_buff = NULL;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2194
    alloc_ptr = NULL;
2195
    ha_tokudb::reset();
2196
    TOKUDB_DBUG_RETURN(free_share(share, mutex_is_locked));
2197 2198
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
2199 2200 2201 2202 2203 2204
//
// Reallocate record buffer (rec_buff) if needed
// If not needed, does nothing
// Parameters:
//          length - size of buffer required for rec_buff
//
2205
bool ha_tokudb::fix_rec_buff_for_blob(ulong length) {
2206
    if (!rec_buff || (length > alloced_rec_buff_length)) {
2207
        uchar *newptr;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2208
        if (!(newptr = (uchar *) my_realloc((void *) rec_buff, length, MYF(MY_ALLOW_ZERO_PTR))))
2209 2210 2211 2212 2213
            return 1;
        rec_buff = newptr;
        alloced_rec_buff_length = length;
    }
    return 0;
2214 2215
}

2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232
//
// Reallocate record buffer (rec_buff) if needed
// If not needed, does nothing
// Parameters:
//          length - size of buffer required for rec_buff
//
bool ha_tokudb::fix_rec_update_buff_for_blob(ulong length) {
    if (!rec_update_buff || (length > alloced_update_rec_buff_length)) {
        uchar *newptr;
        if (!(newptr = (uchar *) my_realloc((void *) rec_update_buff, length, MYF(MY_ALLOW_ZERO_PTR))))
            return 1;
        rec_update_buff= newptr;
        alloced_update_rec_buff_length = length;
    }
    return 0;
}

2233
/* Calculate max length needed for row */
2234 2235 2236 2237 2238
ulong ha_tokudb::max_row_length(const uchar * buf) {
    ulong length = table_share->reclength + table_share->fields * 2;
    uint *ptr, *end;
    for (ptr = table_share->blob_field, end = ptr + table_share->blob_fields; ptr != end; ptr++) {
        Field_blob *blob = ((Field_blob *) table->field[*ptr]);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2239
        length += blob->get_length((uchar *) (buf + field_offset(blob, table))) + 2;
2240 2241
    }
    return length;
2242 2243 2244 2245
}

/*
*/
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2246 2247 2248 2249 2250 2251 2252 2253 2254 2255
//
// take the row passed in as a DBT*, and convert it into a row in MySQL format in record
// Pack a row for storage.
// If the row is of fixed length, just store the  row 'as is'.
// If not, we will generate a packed row suitable for storage.
// This will only fail if we don't have enough memory to pack the row,
// which may only happen in rows with blobs, as the default row length is
// pre-allocated.
// Parameters:
//      [out]   row - row stored in DBT to be converted
2256
//      [out]   buf - buffer where row is packed
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2257 2258
//      [in]    record - row in MySQL format
//
2259

2260
int ha_tokudb::pack_row_in_buff(
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2261
    DBT * row, 
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2262
    const uchar* record,
2263 2264
    uint index,
    uchar* row_buff
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2265 2266
    ) 
{
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2267 2268 2269 2270
    uchar* fixed_field_ptr = NULL;
    uchar* var_field_offset_ptr = NULL;
    uchar* start_field_data_ptr = NULL;
    uchar* var_field_data_ptr = NULL;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2271
    int r = ENOSYS;
2272
    bzero((void *) row, sizeof(*row));
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2273 2274 2275

    my_bitmap_map *old_map = dbug_tmp_use_all_columns(table, table->write_set);
    
2276

2277
    /* Copy null bits */
2278 2279
    memcpy(row_buff, record, table_share->null_bytes);
    fixed_field_ptr = row_buff + table_share->null_bytes;
2280
    var_field_offset_ptr = fixed_field_ptr + share->kc_info.mcp_info[index].fixed_field_size;
2281 2282
    start_field_data_ptr = var_field_offset_ptr + share->kc_info.mcp_info[index].len_of_offsets;
    var_field_data_ptr = var_field_offset_ptr + share->kc_info.mcp_info[index].len_of_offsets;
2283

Zardosht Kasheff's avatar
Zardosht Kasheff committed
2284 2285 2286
    //
    // assert that when the hidden primary key exists, primary_key_offsets is NULL
    //
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2287 2288 2289
    for (uint i = 0; i < table_share->fields; i++) {
        Field* field = table->field[i];
        uint curr_field_offset = field_offset(field, table);
2290
        if (bitmap_is_set(&share->kc_info.key_filters[index],i)) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2291 2292
            continue;
        }
2293
        if (share->kc_info.field_lengths[i]) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2294 2295 2296
            fixed_field_ptr = pack_fixed_field(
                fixed_field_ptr,
                record + curr_field_offset, 
2297
                share->kc_info.field_lengths[i]
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2298 2299
                );
        }
2300
        else if (share->kc_info.length_bytes[i]) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2301 2302 2303 2304 2305
            var_field_data_ptr = pack_var_field(
                var_field_offset_ptr,
                var_field_data_ptr,
                start_field_data_ptr,
                record + curr_field_offset,
2306 2307
                share->kc_info.length_bytes[i],
                share->kc_info.num_offset_bytes
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2308
                );
2309
            var_field_offset_ptr += share->kc_info.num_offset_bytes;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2310 2311 2312
        }
    }

2313 2314
    for (uint i = 0; i < share->kc_info.num_blobs; i++) {
        Field* field = table->field[share->kc_info.blob_fields[i]];
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2315
        var_field_data_ptr = pack_toku_field_blob(
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2316
            var_field_data_ptr,
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2317 2318
            record + field_offset(field, table),
            field
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2319
            );
2320
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2321

2322 2323
    row->data = row_buff;
    row->size = (size_t) (var_field_data_ptr - row_buff);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2324 2325 2326 2327
    r = 0;

    dbug_tmp_restore_column_map(table->write_set, old_map);
    return r;
2328 2329
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
2330

2331 2332 2333 2334 2335 2336 2337 2338 2339 2340 2341 2342 2343 2344 2345 2346 2347 2348 2349
int ha_tokudb::pack_row(
    DBT * row, 
    const uchar* record,
    uint index
    )
{
    return pack_row_in_buff(row,record,index,rec_buff);
}

int ha_tokudb::pack_old_row_for_update(
    DBT * row, 
    const uchar* record,
    uint index
    )
{
    return pack_row_in_buff(row,record,index,rec_update_buff);
}


Zardosht Kasheff's avatar
Zardosht Kasheff committed
2350
int ha_tokudb::unpack_blobs(
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2351 2352
    uchar* record,
    const uchar* from_tokudb_blob,
2353 2354
    u_int32_t num_bytes,
    bool check_bitmap
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2355 2356
    )
{
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2357 2358 2359
    uint error = 0;
    uchar* ptr = NULL;
    const uchar* buff = NULL;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2360
    //
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2361
    // assert that num_bytes > 0 iff share->num_blobs > 0
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2362
    //
2363
    assert( !((share->kc_info.num_blobs == 0) && (num_bytes > 0)) );
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2364 2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375
    if (num_bytes > num_blob_bytes) {
        ptr = (uchar *)my_realloc((void *)blob_buff, num_bytes, MYF(MY_ALLOW_ZERO_PTR));
        if (ptr == NULL) {
            error = ENOMEM;
            goto exit;
        }
        blob_buff = ptr;
        num_blob_bytes = num_bytes;
    }
    
    memcpy(blob_buff, from_tokudb_blob, num_bytes);
    buff= blob_buff;
2376
    for (uint i = 0; i < share->kc_info.num_blobs; i++) {
2377 2378 2379 2380 2381 2382
        u_int32_t curr_field_index = share->kc_info.blob_fields[i]; 
        bool skip = check_bitmap ? 
            !(bitmap_is_set(table->read_set,curr_field_index) || 
                bitmap_is_set(table->write_set,curr_field_index)) : 
            false;
        Field* field = table->field[curr_field_index];
2383
        u_int32_t len_bytes = field->row_pack_length();
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2384
        buff = unpack_toku_field_blob(
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2385
            record + field_offset(field, table),
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2386
            buff,
2387
            len_bytes,
2388
            skip
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2389 2390 2391
            );
    }

Zardosht Kasheff's avatar
Zardosht Kasheff committed
2392 2393 2394
    error = 0;
exit:
    return error;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2395 2396
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
2397 2398 2399 2400 2401 2402
//
// take the row passed in as a DBT*, and convert it into a row in MySQL format in record
// Parameters:
//      [out]   record - row in MySQL format
//      [in]    row - row stored in DBT to be converted
//
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2403
int ha_tokudb::unpack_row(
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2404 2405
    uchar* record, 
    DBT const *row, 
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2406 2407
    DBT const *key,
    uint index
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2408 2409
    ) 
{
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2410 2411 2412 2413
    //
    // two cases, fixed length row, and variable length row
    // fixed length row is first below
    //
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2414
    /* Copy null bits */
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2415
    int error = 0;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2416 2417 2418 2419 2420 2421
    const uchar* fixed_field_ptr = (const uchar *) row->data;
    const uchar* var_field_offset_ptr = NULL;
    const uchar* var_field_data_ptr = NULL;
    u_int32_t data_end_offset = 0;
    memcpy(record, fixed_field_ptr, table_share->null_bytes);
    fixed_field_ptr += table_share->null_bytes;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2422

2423
    var_field_offset_ptr = fixed_field_ptr + share->kc_info.mcp_info[index].fixed_field_size;
2424
    var_field_data_ptr = var_field_offset_ptr + share->kc_info.mcp_info[index].len_of_offsets;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2425

Zardosht Kasheff's avatar
Zardosht Kasheff committed
2426 2427 2428 2429 2430 2431 2432 2433 2434 2435 2436 2437 2438 2439 2440 2441 2442 2443 2444
    //
    // unpack the key, if necessary
    //
    if (!(hidden_primary_key && index == primary_key)) {
        unpack_key(record,key,index);
    }

    u_int32_t last_offset = 0;
    //
    // we have two methods of unpacking, one if we need to unpack the entire row
    // the second if we unpack a subset of the entire row
    // first method here is if we unpack the entire row
    //
    if (unpack_entire_row) {
        //
        // fill in parts of record that are not part of the key
        //
        for (uint i = 0; i < table_share->fields; i++) {
            Field* field = table->field[i];
2445
            if (bitmap_is_set(&share->kc_info.key_filters[index],i)) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2446 2447
                continue;
            }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2448

2449
            if (share->kc_info.field_lengths[i]) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2450 2451 2452
                fixed_field_ptr = unpack_fixed_field(
                    record + field_offset(field, table),
                    fixed_field_ptr,
2453
                    share->kc_info.field_lengths[i]
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2454 2455
                    );
            }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2456
            //
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2457 2458
            // here, we DO modify var_field_data_ptr or var_field_offset_ptr
            // as we unpack variable sized fields
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2459
            //
2460 2461
            else if (share->kc_info.length_bytes[i]) {
                switch (share->kc_info.num_offset_bytes) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2462 2463 2464 2465 2466 2467 2468 2469 2470 2471 2472 2473 2474 2475
                case (1):
                    data_end_offset = var_field_offset_ptr[0];
                    break;
                case (2):
                    data_end_offset = uint2korr(var_field_offset_ptr);
                    break;
                default:
                    assert(false);
                    break;
                }
                unpack_var_field(
                    record + field_offset(field, table),
                    var_field_data_ptr,
                    data_end_offset - last_offset,
2476
                    share->kc_info.length_bytes[i]
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2477
                    );
2478
                var_field_offset_ptr += share->kc_info.num_offset_bytes;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2479 2480
                var_field_data_ptr += data_end_offset - last_offset;
                last_offset = data_end_offset;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2481
            }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2482
        }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2483
        error = unpack_blobs(
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2484 2485
            record,
            var_field_data_ptr,
2486 2487
            row->size - (u_int32_t)(var_field_data_ptr - (const uchar *)row->data),
            false
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2488
            );
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2489 2490 2491
        if (error) {
            goto exit;
        }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2492
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2493 2494 2495 2496
    //
    // in this case, we unpack only what is specified 
    // in fixed_cols_for_query and var_cols_for_query
    //
2497
    else {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2498 2499 2500 2501 2502 2503 2504 2505
        //
        // first the fixed fields
        //
        for (u_int32_t i = 0; i < num_fixed_cols_for_query; i++) {
            uint field_index = fixed_cols_for_query[i];
            Field* field = table->field[field_index];
            unpack_fixed_field(
                record + field_offset(field, table),
2506 2507
                fixed_field_ptr + share->kc_info.cp_info[index][field_index].col_pack_val,
                share->kc_info.field_lengths[field_index]
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2508
                );
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2509 2510 2511
        }

        //
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2512 2513
        // now the var fields
        // here, we do NOT modify var_field_data_ptr or var_field_offset_ptr
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2514
        //
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2515 2516 2517
        for (u_int32_t i = 0; i < num_var_cols_for_query; i++) {
            uint field_index = var_cols_for_query[i];
            Field* field = table->field[field_index];
2518
            u_int32_t var_field_index = share->kc_info.cp_info[index][field_index].col_pack_val;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2519
            u_int32_t data_start_offset;
2520
            u_int32_t field_len;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2521
            
2522 2523 2524 2525 2526 2527 2528
            get_var_field_info(
                &field_len, 
                &data_start_offset, 
                var_field_index, 
                var_field_offset_ptr, 
                share->kc_info.num_offset_bytes
                );
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2529 2530 2531 2532

            unpack_var_field(
                record + field_offset(field, table),
                var_field_data_ptr + data_start_offset,
2533 2534
                field_len,
                share->kc_info.length_bytes[field_index]
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2535 2536 2537
                );
        }

Zardosht Kasheff's avatar
Zardosht Kasheff committed
2538 2539 2540 2541
        if (read_blobs) {
            //
            // now the blobs
            //
2542 2543 2544 2545 2546 2547
            get_blob_field_info(
                &data_end_offset, 
                share->kc_info.mcp_info[index].len_of_offsets,
                var_field_data_ptr, 
                share->kc_info.num_offset_bytes
                );
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2548

2549
            var_field_data_ptr += data_end_offset;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2550
            error = unpack_blobs(
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2551 2552
                record,
                var_field_data_ptr,
2553 2554
                row->size - (u_int32_t)(var_field_data_ptr - (const uchar *)row->data),
                true
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2555
                );
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2556 2557 2558
            if (error) {
                goto exit;
            }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2559
        }
2560
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2561 2562 2563
    error = 0;
exit:
    return error;
2564 2565
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
2566 2567 2568 2569 2570 2571
u_int32_t ha_tokudb::place_key_into_mysql_buff(
    KEY* key_info, 
    uchar * record, 
    uchar* data
    ) 
{
2572
    KEY_PART_INFO *key_part = key_info->key_part, *end = key_part + key_info->key_parts;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2573
    uchar *pos = data;
2574 2575

    for (; key_part != end; key_part++) {
2576 2577
        if (key_part->field->null_bit) {
            uint null_offset = get_null_offset(table, key_part->field);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2578
            if (*pos++ == NULL_COL_VAL) { // Null value
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2579 2580 2581 2582
                //
                // We don't need to reset the record data as we will not access it
                // if the null data is set
                //            
2583
                record[null_offset] |= key_part->field->null_bit;
2584 2585
                continue;
            }
2586
            record[null_offset] &= ~key_part->field->null_bit;
2587
        }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2588 2589 2590 2591
        //
        // HOPEFULLY TEMPORARY
        //
        assert(table->s->db_low_byte_first);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2592
        pos = unpack_toku_key_field(
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2593
            record + field_offset(key_part->field, table),
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2594
            pos,
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2595 2596
            key_part->field,
            key_part->length
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2597
            );
2598
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2599
    return pos-data;
2600 2601
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
2602
//
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2603
// Store the key and the primary key into the row
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2604
// Parameters:
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2605 2606 2607 2608
//      [out]   record - key stored in MySQL format
//      [in]    key - key stored in DBT to be converted
//              index -index into key_file that represents the DB 
//                  unpacking a key of
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2609
//
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2610 2611 2612
void ha_tokudb::unpack_key(uchar * record, DBT const *key, uint index) {
    u_int32_t bytes_read;
    uchar *pos = (uchar *) key->data + 1;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2613 2614 2615 2616 2617
    bytes_read = place_key_into_mysql_buff(
        &table->key_info[index], 
        record, 
        pos
        );
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2618
    if( (index != primary_key) && !hidden_primary_key) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2619 2620 2621
        //
        // also unpack primary key
        //
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2622 2623 2624 2625 2626
        place_key_into_mysql_buff(
            &table->key_info[primary_key], 
            record, 
            pos+bytes_read
            );
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2627 2628
    }
}
2629

Zardosht Kasheff's avatar
Zardosht Kasheff committed
2630 2631 2632 2633 2634 2635 2636 2637
u_int32_t ha_tokudb::place_key_into_dbt_buff(
    KEY* key_info, 
    uchar * buff, 
    const uchar * record, 
    bool* has_null, 
    int key_length
    ) 
{
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2638 2639 2640
    KEY_PART_INFO *key_part = key_info->key_part;
    KEY_PART_INFO *end = key_part + key_info->key_parts;
    uchar* curr_buff = buff;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2641
    *has_null = false;
2642
    for (; key_part != end && key_length > 0; key_part++) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2643 2644 2645 2646 2647 2648
        //
        // accessing key_part->field->null_bit instead off key_part->null_bit
        // because key_part->null_bit is not set in add_index
        // filed ticket 862 to look into this
        //
        if (key_part->field->null_bit) {
2649
            /* Store 0 if the key part is a NULL part */
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2650
            uint null_offset = get_null_offset(table, key_part->field);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2651
            if (record[null_offset] & key_part->field->null_bit) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2652
                *curr_buff++ = NULL_COL_VAL;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2653
                *has_null = true;
2654 2655
                continue;
            }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2656
            *curr_buff++ = NONNULL_COL_VAL;        // Store NOT NULL marker
2657
        }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2658
        //
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2659 2660 2661 2662
        // HOPEFULLY TEMPORARY
        //
        assert(table->s->db_low_byte_first);
        //
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2663 2664 2665 2666
        // accessing field_offset(key_part->field) instead off key_part->offset
        // because key_part->offset is SET INCORRECTLY in add_index
        // filed ticket 862 to look into this
        //
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2667
        curr_buff = pack_toku_key_field(
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2668
            curr_buff,
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2669
            (uchar *) (record + field_offset(key_part->field, table)),
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2670 2671
            key_part->field,
            key_part->length
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2672
            );
2673 2674
        key_length -= key_part->length;
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2675 2676 2677 2678 2679 2680 2681 2682 2683 2684 2685 2686 2687 2688 2689 2690 2691 2692 2693
    return curr_buff - buff;
}



//
// Create a packed key from a row. This key will be written as such
// to the index tree.  This will never fail as the key buffer is pre-allocated.
// Parameters:
//      [out]   key - DBT that holds the key
//      [in]    key_info - holds data about the key, such as it's length and offset into record
//      [out]   buff - buffer that will hold the data for key (unless 
//                  we have a hidden primary key)
//      [in]    record - row from which to create the key
//              key_length - currently set to MAX_KEY_LENGTH, is it size of buff?
// Returns:
//      the parameter key
//

Zardosht Kasheff's avatar
Zardosht Kasheff committed
2694
DBT* ha_tokudb::create_dbt_key_from_key(
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2695
    DBT * key,
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2696
    KEY* key_info, 
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2697
    uchar * buff,
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2698
    const uchar * record, 
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2699 2700
    bool* has_null,
    bool dont_pack_pk,
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2701 2702 2703
    int key_length
    ) 
{
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2704 2705 2706 2707 2708 2709 2710 2711
    u_int32_t size = 0;
    uchar* tmp_buff = buff;
    my_bitmap_map *old_map = dbug_tmp_use_all_columns(table, table->write_set);

    key->data = buff;

    //
    // first put the "infinity" byte at beginning. States if missing columns are implicitly
2712
    // positive infinity or negative infinity or zero. For this, because we are creating key
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2713 2714 2715
    // from a row, there is no way that columns can be missing, so in practice,
    // this will be meaningless. Might as well put in a value
    //
2716
    *tmp_buff++ = COL_ZERO;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2717
    size++;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2718 2719 2720 2721 2722 2723 2724
    size += place_key_into_dbt_buff(
        key_info, 
        tmp_buff, 
        record, 
        has_null, 
        key_length
        );
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2725
    if (!dont_pack_pk) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743
        tmp_buff = buff + size;
        if (hidden_primary_key) {
            memcpy_fixed(tmp_buff, current_ident, TOKUDB_HIDDEN_PRIMARY_KEY_LENGTH);
            size += TOKUDB_HIDDEN_PRIMARY_KEY_LENGTH;
        }
        else {
            bool tmp_bool = false;
            size += place_key_into_dbt_buff(
                &table->key_info[primary_key], 
                tmp_buff, 
                record, 
                &tmp_bool, 
                MAX_KEY_LENGTH //this parameter does not matter
                );
        }
    }

    key->size = size;
2744 2745
    DBUG_DUMP("key", (uchar *) key->data, key->size);
    dbug_tmp_restore_column_map(table->write_set, old_map);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758
    return key;
}


//
// Create a packed key from a row. This key will be written as such
// to the index tree.  This will never fail as the key buffer is pre-allocated.
// Parameters:
//      [out]   key - DBT that holds the key
//              keynr - index for which to create the key
//      [out]   buff - buffer that will hold the data for key (unless 
//                  we have a hidden primary key)
//      [in]    record - row from which to create the key
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2759
//      [out]   has_null - says if the key has a NULL value for one of its columns
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2760 2761 2762 2763
//              key_length - currently set to MAX_KEY_LENGTH, is it size of buff?
// Returns:
//      the parameter key
//
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2764 2765 2766 2767 2768 2769 2770 2771 2772
DBT *ha_tokudb::create_dbt_key_from_table(
    DBT * key, 
    uint keynr, 
    uchar * buff, 
    const uchar * record, 
    bool* has_null, 
    int key_length
    ) 
{
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2773 2774 2775
    TOKUDB_DBUG_ENTER("ha_tokudb::create_dbt_key_from_table");
    bzero((void *) key, sizeof(*key));
    if (hidden_primary_key && keynr == primary_key) {
2776 2777
        key->data = buff;
        memcpy(buff, &current_ident, TOKUDB_HIDDEN_PRIMARY_KEY_LENGTH);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2778
        key->size = TOKUDB_HIDDEN_PRIMARY_KEY_LENGTH;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2779
        *has_null = false;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2780 2781
        DBUG_RETURN(key);
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2782
    DBUG_RETURN(create_dbt_key_from_key(key, &table->key_info[keynr],buff,record, has_null, (keynr == primary_key), key_length));
2783 2784
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
2785 2786 2787 2788 2789 2790 2791 2792 2793 2794 2795 2796
DBT* ha_tokudb::create_dbt_key_for_lookup(
    DBT * key, 
    KEY* key_info, 
    uchar * buff, 
    const uchar * record, 
    bool* has_null, 
    int key_length
    )
{
    TOKUDB_DBUG_ENTER("ha_tokudb::create_dbt_key_from_lookup");
    DBUG_RETURN(create_dbt_key_from_key(key, key_info, buff, record, has_null, true, key_length));    
}
2797

Zardosht Kasheff's avatar
Zardosht Kasheff committed
2798 2799 2800 2801 2802 2803 2804 2805 2806 2807 2808 2809
//
// Create a packed key from from a MySQL unpacked key (like the one that is
// sent from the index_read() This key is to be used to read a row
// Parameters:
//      [out]   key - DBT that holds the key
//              keynr - index for which to pack the key
//      [out]   buff - buffer that will hold the data for key
//      [in]    key_ptr - MySQL unpacked key
//              key_length - length of key_ptr
// Returns:
//      the parameter key
//
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2810 2811 2812 2813 2814 2815
DBT *ha_tokudb::pack_key(
    DBT * key, 
    uint keynr, 
    uchar * buff, 
    const uchar * key_ptr, 
    uint key_length, 
2816
    int8_t inf_byte
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2817 2818
    ) 
{
2819
    TOKUDB_DBUG_ENTER("ha_tokudb::pack_key");
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2820
    KEY *key_info = &table->key_info[keynr];
2821 2822 2823 2824 2825 2826 2827
    KEY_PART_INFO *key_part = key_info->key_part;
    KEY_PART_INFO *end = key_part + key_info->key_parts;
    my_bitmap_map *old_map = dbug_tmp_use_all_columns(table, table->write_set);

    bzero((void *) key, sizeof(*key));
    key->data = buff;

Zardosht Kasheff's avatar
Zardosht Kasheff committed
2828 2829 2830 2831
    //
    // first put the "infinity" byte at beginning. States if missing columns are implicitly
    // positive infinity or negative infinity
    //
2832
    *buff++ = (uchar)inf_byte;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2833

2834 2835 2836
    for (; key_part != end && (int) key_length > 0; key_part++) {
        uint offset = 0;
        if (key_part->null_bit) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2837 2838
            if (!(*key_ptr == 0)) {
                *buff++ = NULL_COL_VAL;
2839 2840 2841 2842
                key_length -= key_part->store_length;
                key_ptr += key_part->store_length;
                continue;
            }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2843
            *buff++ = NONNULL_COL_VAL;
2844 2845
            offset = 1;         // Data is at key_ptr+1
        }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2846 2847
        assert(table->s->db_low_byte_first);

Zardosht Kasheff's avatar
Zardosht Kasheff committed
2848
        buff = pack_key_toku_key_field(
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2849
            buff,
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2850
            (uchar *) key_ptr + offset,
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2851 2852
            key_part->field,
            key_part->length
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2853
            );
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2854
        
2855 2856 2857 2858 2859 2860 2861
        key_ptr += key_part->store_length;
        key_length -= key_part->store_length;
    }
    key->size = (buff - (uchar *) key->data);
    DBUG_DUMP("key", (uchar *) key->data, key->size);
    dbug_tmp_restore_column_map(table->write_set, old_map);
    DBUG_RETURN(key);
2862 2863
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
2864 2865 2866 2867 2868
//
// get max used hidden primary key value
//
void ha_tokudb::init_hidden_prim_key_info() {
    TOKUDB_DBUG_ENTER("ha_tokudb::init_prim_key_info");
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2869
    pthread_mutex_lock(&share->mutex);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2870
    if (!(share->status & STATUS_PRIMARY_KEY_INIT)) {
2871 2872 2873 2874 2875 2876 2877 2878 2879 2880 2881 2882 2883 2884 2885 2886 2887 2888 2889 2890 2891 2892 2893 2894 2895 2896 2897
        int error = 0;
        THD* thd = ha_thd();
        DB_TXN* txn = NULL;
        DBC* c = NULL;
        tokudb_trx_data *trx = NULL;
        trx = (tokudb_trx_data *) thd_data_get(ha_thd(), tokudb_hton->slot);
        bool do_commit = false;
        if (thd_sql_command(thd) == SQLCOM_CREATE_TABLE && trx && trx->sub_sp_level) {
            txn = trx->sub_sp_level;
        }
        else {
            do_commit = true;
            error = db_env->txn_begin(db_env, 0, &txn, 0);
            assert(error == 0);
        }
        
        error = share->key_file[primary_key]->cursor(
            share->key_file[primary_key],
            txn,
            &c,
            0
            );
        assert(error == 0);
        DBT key,val;        
        bzero(&key, sizeof(key));
        bzero(&val, sizeof(val));
        error = c->c_get(c, &key, &val, DB_LAST);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2898
        if (error == 0) {
2899 2900 2901 2902 2903 2904 2905
            assert(key.size == TOKUDB_HIDDEN_PRIMARY_KEY_LENGTH);
            share->auto_ident = hpk_char_to_num((uchar *)key.data);
        }
        error = c->c_close(c);
        assert(error == 0);
        if (do_commit) {
            commit_txn(txn, 0);
2906
        }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2907
        share->status |= STATUS_PRIMARY_KEY_INIT;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2908 2909
    }
    pthread_mutex_unlock(&share->mutex);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2910 2911
    DBUG_VOID_RETURN;
}
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2912

Zardosht Kasheff's avatar
Zardosht Kasheff committed
2913 2914 2915 2916 2917


/** @brief
    Get metadata info stored in status.tokudb
    */
2918
int ha_tokudb::get_status(DB_TXN* txn) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2919
    TOKUDB_DBUG_ENTER("ha_tokudb::get_status");
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2920 2921 2922
    DBT key, value;
    HA_METADATA_KEY curr_key;
    int error;
2923

Zardosht Kasheff's avatar
Zardosht Kasheff committed
2924 2925 2926 2927
    //
    // open status.tokudb
    //
    if (!share->status_block) {
2928 2929 2930
        error = open_status_dictionary(
            &share->status_block, 
            share->table_name, 
2931
            txn
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2932
            );
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2933 2934 2935
        if (error) { 
            goto cleanup; 
        }
2936
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2937
    
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2938 2939 2940 2941 2942 2943 2944
    //
    // transaction to be used for putting metadata into status.tokudb
    //
    bzero(&key, sizeof(key));
    bzero(&value, sizeof(value));
    key.data = &curr_key;
    key.size = sizeof(curr_key);
2945
    value.flags = DB_DBT_USERMEM;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2946

2947 2948 2949 2950 2951 2952
    assert(share->status_block);
    //
    // get version
    //
    value.ulen = sizeof(share->version);
    value.data = &share->version;
2953
    curr_key = hatoku_new_version;
2954 2955 2956 2957 2958 2959 2960 2961
    error = share->status_block->get(
        share->status_block, 
        txn, 
        &key, 
        &value, 
        0
        );
    if (error == DB_NOTFOUND) {
2962 2963 2964 2965 2966 2967 2968 2969 2970 2971 2972 2973 2974 2975 2976 2977 2978 2979 2980 2981 2982 2983 2984 2985 2986 2987 2988 2989 2990 2991 2992 2993 2994
        //
        // hack to keep handle the issues of going back and forth
        // between 5.0.3 to 5.0.4
        // the problem with going back and forth
        // is with storing the frm file, 5.0.4 stores it, 5.0.3 does not
        // so, if a user goes back and forth and alters the schema
        // the frm stored can get out of sync with the schema of the table
        // This can cause issues.
        // To take care of this, we are doing this versioning work here.
        // We change the key that stores the version. 
        // In 5.0.3, it is hatoku_old_version, in 5.0.4 it is hatoku_new_version
        // When we encounter a table that does not have hatoku_new_version
        // set, we give it the right one, and overwrite the old one with zero.
        // This ensures that 5.0.3 cannot open the table. Once it has been opened by 5.0.4
        //
        uint dummy_version = 0;
        share->version = HA_TOKU_ORIG_VERSION;
        error = write_to_status(
            share->status_block, 
            hatoku_new_version,
            &share->version,
            sizeof(share->version), 
            txn
            );
        if (error) { goto cleanup; }
        error = write_to_status(
            share->status_block, 
            hatoku_old_version,
            &dummy_version,
            sizeof(dummy_version), 
            txn
            );
        if (error) { goto cleanup; }
2995 2996 2997 2998
    }
    else if (error || value.size != sizeof(share->version)) {
        if (error == 0) {
            error = HA_ERR_INTERNAL_ERROR;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
2999
        }
3000 3001 3002 3003 3004 3005 3006 3007 3008 3009 3010 3011 3012 3013 3014 3015 3016 3017 3018 3019 3020
        goto cleanup;
    }
    //
    // get capabilities
    //
    curr_key = hatoku_capabilities;
    value.ulen = sizeof(share->capabilities);
    value.data = &share->capabilities;
    error = share->status_block->get(
        share->status_block, 
        txn, 
        &key, 
        &value, 
        0
        );
    if (error == DB_NOTFOUND) {
        share->capabilities= 0;
    }
    else if (error || value.size != sizeof(share->version)) {
        if (error == 0) {
            error = HA_ERR_INTERNAL_ERROR;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3021
        }
3022
        goto cleanup;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3023
    }
3024
    
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3025 3026 3027
    error = 0;
cleanup:
    TOKUDB_DBUG_RETURN(error);
3028 3029
}

3030 3031 3032
/** @brief
    Return an estimated of the number of rows in the table.
    Used when sorting to allocate buffers and by the optimizer.
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3033
    This is used in filesort.cc. 
3034
*/
3035
ha_rows ha_tokudb::estimate_rows_upper_bound() {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3036
    TOKUDB_DBUG_ENTER("ha_tokudb::estimate_rows_upper_bound");
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3037
    DBUG_RETURN(share->rows + HA_TOKUDB_EXTRA_ROWS);
3038
}
3039

Zardosht Kasheff's avatar
Zardosht Kasheff committed
3040 3041 3042 3043
//
// Function that compares two primary keys that were saved as part of rnd_pos
// and ::position
//
3044
int ha_tokudb::cmp_ref(const uchar * ref1, const uchar * ref2) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3045 3046
    int ret_val = 0;
    ret_val = tokudb_compare_two_keys(
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3047 3048 3049
        ref1 + sizeof(u_int32_t),
        *(u_int32_t *)ref1,
        ref2 + sizeof(u_int32_t),
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3050
        *(u_int32_t *)ref2,
3051 3052
        (uchar *)share->file->descriptor->dbt.data + 4,
        *(u_int32_t *)share->file->descriptor->dbt.data - 4,
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3053
        false
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3054 3055
        );
    return ret_val;
3056 3057
}

3058
bool ha_tokudb::check_if_incompatible_data(HA_CREATE_INFO * info, uint table_changes) {
3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070 3071
  //
  // This is a horrendous hack for now, as copied by InnoDB.
  // This states that if the auto increment create field has changed,
  // via a "alter table foo auto_increment=new_val", that this
  // change is incompatible, and to rebuild the entire table
  // This will need to be fixed
  //
  if ((info->used_fields & HA_CREATE_USED_AUTO) &&
      info->auto_increment_value != 0) {

    return COMPATIBLE_DATA_NO;
  }
  if (table_changes != IS_EQUAL_YES)
3072
    return COMPATIBLE_DATA_NO;
3073
  return COMPATIBLE_DATA_YES;
3074
}
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3075

3076 3077 3078 3079 3080 3081 3082 3083 3084 3085 3086 3087
//
// Method that is called before the beginning of many calls
// to insert rows (ha_tokudb::write_row). There is no guarantee
// that start_bulk_insert is called, however there is a guarantee
// that if start_bulk_insert is called, then end_bulk_insert may be
// called as well.
// Parameters:
//      [in]    rows - an estimate of the number of rows that will be inserted
//                     if number of rows is unknown (such as if doing 
//                     "insert into foo select * from bar), then rows 
//                     will be 0
//
3088 3089 3090 3091 3092
//
// This function returns true if the table MAY be empty.
// It is NOT meant to be a 100% check for emptiness.
// This is used for a bulk load optimization.
//
3093
bool ha_tokudb::may_table_be_empty(DB_TXN *txn) {
3094 3095 3096
    int error;
    bool ret_val = false;
    DBC* tmp_cursor = NULL;
3097
    DB_TXN* tmp_txn = NULL;
3098

3099 3100 3101 3102 3103 3104
    if (txn == NULL) {
        error = db_env->txn_begin(db_env, 0, &tmp_txn, 0);
        if (error) {
            goto cleanup;
        }
        txn = tmp_txn;
3105 3106 3107 3108 3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120
    }

    error = share->file->cursor(share->file, txn, &tmp_cursor, 0);
    if (error) {
        goto cleanup;
    }
    error = tmp_cursor->c_getf_next(tmp_cursor, 0, smart_dbt_do_nothing, NULL);
    if (error == DB_NOTFOUND) {
        ret_val = true;
    }
    else {
        ret_val = false;
    }
    error = 0;
cleanup:
    if (tmp_cursor) {
3121 3122
        int r = tmp_cursor->c_close(tmp_cursor);
        assert(r==0);
3123 3124
        tmp_cursor = NULL;
    }
3125 3126 3127
    if (tmp_txn) {
        commit_txn(tmp_txn, 0);
        tmp_txn = NULL;
3128 3129 3130 3131
    }
    return ret_val;
}

3132
void ha_tokudb::start_bulk_insert(ha_rows rows) {
3133 3134
    TOKUDB_DBUG_ENTER("ha_tokudb::start_bulk_insert");
    THD* thd = ha_thd();
3135
    tokudb_trx_data* trx = (tokudb_trx_data *) thd_data_get(thd, tokudb_hton->slot);
3136
    delay_updating_ai_metadata = true;
3137
    ai_metadata_update_required = false;
3138
    abort_loader = false;
3139 3140 3141 3142 3143 3144
    
    rw_rdlock(&share->num_DBs_lock);
    uint curr_num_DBs = table->s->keys + test(hidden_primary_key);
    num_DBs_locked_in_bulk = true;
    lock_count = 0;
    
3145
    if (share->try_table_lock) {
3146
        if (get_prelock_empty(thd) && may_table_be_empty(transaction)) {
3147
            if (using_ignore || get_load_save_space(thd)) {
3148 3149 3150 3151 3152 3153 3154 3155 3156 3157 3158 3159 3160 3161 3162 3163 3164 3165
                acquire_table_lock(transaction, lock_write);
            }
            else {
                mult_dbt_flags[primary_key] = 0;
                if (!thd_test_options(thd, OPTION_RELAXED_UNIQUE_CHECKS) && !hidden_primary_key) {
                    mult_put_flags[primary_key] = DB_NOOVERWRITE;
                }
                int error = db_env->create_loader(
                    db_env, 
                    transaction, 
                    &loader, 
                    NULL, // no src_db needed
                    curr_num_DBs, 
                    share->key_file, 
                    mult_put_flags,
                    mult_dbt_flags,
                    0
                    );
3166 3167 3168 3169
                if (error) { 
                    assert(loader == NULL);
                    goto exit_try_table_lock;
                }
3170 3171 3172 3173 3174 3175 3176 3177 3178

                lc.thd = thd;
                lc.ha = this;
                
                error = loader->set_poll_function(loader, poll_fun, &lc);
                assert(!error);

                error = loader->set_error_callback(loader, loader_dup_fun, &lc);
                assert(!error);
3179 3180

                trx->stmt_progress.using_loader = true;
3181
            }
3182
        }
3183
    exit_try_table_lock:
3184
        pthread_mutex_lock(&share->mutex);
3185
        share->try_table_lock = false; // RFP what good is the mutex?
3186 3187
        pthread_mutex_unlock(&share->mutex);
    }
3188
    DBUG_VOID_RETURN;
3189 3190 3191 3192 3193 3194 3195
}

//
// Method that is called at the end of many calls to insert rows
// (ha_tokudb::write_row). If start_bulk_insert is called, then
// this is guaranteed to be called.
//
3196
int ha_tokudb::end_bulk_insert(bool abort) {
3197
    TOKUDB_DBUG_ENTER("ha_tokudb::end_bulk_insert");
3198
    int error = 0;
3199
    THD* thd = ha_thd();
3200
    tokudb_trx_data* trx = (tokudb_trx_data *) thd_data_get(thd, tokudb_hton->slot);
3201
    bool using_loader = (loader != NULL);
3202 3203 3204 3205
    if (ai_metadata_update_required) {
        pthread_mutex_lock(&share->mutex);
        error = update_max_auto_inc(share->status_block, share->last_auto_increment);
        pthread_mutex_unlock(&share->mutex);
3206
        if (error) { goto cleanup; }
3207 3208 3209
    }
    delay_updating_ai_metadata = false;
    ai_metadata_update_required = false;
3210 3211
    loader_error = 0;
    if (loader) {
3212
        if (!abort_loader && !thd->killed) {
3213 3214
            error = loader->close(loader);
            loader = NULL;
3215 3216 3217 3218 3219 3220
            if (error) { 
                if (thd->killed) {
                    my_error(ER_QUERY_INTERRUPTED, MYF(0));
                }
                goto cleanup; 
            }
3221 3222 3223 3224

            for (uint i = 0; i < table_share->keys; i++) {
                if (table_share->key_info[i].flags & HA_NOSAME) {
                    bool is_unique;
3225 3226 3227
                    if (i == primary_key) {
                        continue;
                    }
3228 3229 3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240 3241 3242 3243
                    error = is_index_unique(
                        &is_unique, 
                        transaction, 
                        share->key_file[i], 
                        &table->key_info[i]
                        );
                    if (error) goto cleanup;
                    if (!is_unique) {
                        error = HA_ERR_FOUND_DUPP_KEY;
                        last_dup_key = i;
                        goto cleanup;
                    }
                }
            }
        }
        else {
3244 3245
            error = sprintf(write_status_msg, "aborting bulk load"); 
            thd_proc_info(thd, write_status_msg);
3246 3247
            loader->abort(loader);
            loader = NULL;
3248
            share->try_table_lock = true;
3249 3250 3251 3252
        }
    }

cleanup:
3253 3254 3255 3256 3257 3258
    if (num_DBs_locked_in_bulk) {
        rw_unlock(&share->num_DBs_lock);
    }
    num_DBs_locked_in_bulk = false;
    lock_count = 0;

3259
    if (loader) {
3260 3261
        error = sprintf(write_status_msg, "aborting bulk load"); 
        thd_proc_info(thd, write_status_msg);
3262 3263 3264 3265 3266 3267 3268
        loader->abort(loader);
        loader = NULL;
    }
    abort_loader = false;
    bzero(&lc,sizeof(lc));
    if (error || loader_error) {
        my_errno = error ? error : loader_error;
3269 3270 3271
        if (using_loader) {
            share->try_table_lock = true;
        }
3272
    }
3273
    trx->stmt_progress.using_loader = false;
3274
    TOKUDB_DBUG_RETURN(error ? error : loader_error);
3275 3276
}

3277 3278 3279
int ha_tokudb::end_bulk_insert() {
    return end_bulk_insert( false );
}
3280

3281 3282 3283 3284 3285
int ha_tokudb::is_index_unique(bool* is_unique, DB_TXN* txn, DB* db, KEY* key_info) {
    int error;
    DBC* tmp_cursor1 = NULL;
    DBC* tmp_cursor2 = NULL;
    DBT key1, key2, val, packed_key1, packed_key2;
3286
    u_int64_t cnt = 0;
3287
    char status_msg[MAX_ALIAS_NAME + 200]; //buffer of 200 should be a good upper bound.
3288
    THD* thd = ha_thd();
3289 3290 3291 3292 3293 3294 3295 3296 3297 3298 3299
    bzero(&key1, sizeof(key1));
    bzero(&key2, sizeof(key2));
    bzero(&val, sizeof(val));
    bzero(&packed_key1, sizeof(packed_key1));
    bzero(&packed_key2, sizeof(packed_key2));
    *is_unique = true;
    
    error = db->cursor(
        db, 
        txn, 
        &tmp_cursor1, 
3300
        DB_SERIALIZABLE
3301 3302 3303 3304 3305 3306 3307
        );
    if (error) { goto cleanup; }

    error = db->cursor(
        db, 
        txn, 
        &tmp_cursor2,
3308
        DB_SERIALIZABLE
3309 3310 3311 3312 3313 3314 3315 3316 3317 3318 3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331 3332 3333 3334 3335 3336 3337 3338 3339 3340 3341 3342 3343 3344 3345 3346 3347 3348 3349 3350 3351 3352 3353 3354 3355 3356 3357 3358 3359 3360 3361 3362 3363 3364 3365 3366 3367 3368 3369 3370 3371 3372 3373 3374 3375 3376 3377 3378
        );
    if (error) { goto cleanup; }

    
    error = tmp_cursor1->c_get(
        tmp_cursor1, 
        &key1, 
        &val, 
        DB_NEXT
        );
    if (error == DB_NOTFOUND) {
        *is_unique = true;
        error = 0;
        goto cleanup;
    }
    else if (error) { goto cleanup; }
    error = tmp_cursor2->c_get(
        tmp_cursor2, 
        &key2, 
        &val, 
        DB_NEXT
        );
    if (error) { goto cleanup; }

    error = tmp_cursor2->c_get(
        tmp_cursor2, 
        &key2, 
        &val, 
        DB_NEXT
        );
    if (error == DB_NOTFOUND) {
        *is_unique = true;
        error = 0;
        goto cleanup;
    }
    else if (error) { goto cleanup; }

    while (error != DB_NOTFOUND) {
        bool has_null1;
        bool has_null2;
        int cmp;
        place_key_into_mysql_buff(
            key_info,
            table->record[0], 
            (uchar *) key1.data + 1
            );
        place_key_into_mysql_buff(
            key_info,
            table->record[1], 
            (uchar *) key2.data + 1
            );
        
        create_dbt_key_for_lookup(
            &packed_key1,
            key_info,
            key_buff,
            table->record[0],
            &has_null1
            );
        create_dbt_key_for_lookup(
            &packed_key2,
            key_info,
            key_buff2,
            table->record[1],
            &has_null2
            );

        if (!has_null1 && !has_null2) {
            cmp = tokudb_prefix_cmp_dbt_key(db, &packed_key1, &packed_key2);
            if (cmp == 0) {
3379 3380 3381 3382 3383 3384
                memcpy(key_buff, key1.data, key1.size);
                place_key_into_mysql_buff(
                    key_info,
                    table->record[0], 
                    (uchar *) key_buff + 1
                    );
3385 3386 3387 3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403
                *is_unique = false;
                break;
            }
        }

        error = tmp_cursor1->c_get(
            tmp_cursor1, 
            &key1, 
            &val, 
            DB_NEXT
            );
        if (error) { goto cleanup; }
        error = tmp_cursor2->c_get(
            tmp_cursor2, 
            &key2, 
            &val, 
            DB_NEXT
            );
        if (error && (error != DB_NOTFOUND)) { goto cleanup; }
3404 3405

        cnt++;
3406 3407 3408 3409
        if ((cnt % 10000) == 0) {
            sprintf(
                status_msg, 
                "Verifying index uniqueness: Checked %llu of %llu rows in key-%s.", 
3410
                (long long unsigned) cnt, 
3411 3412 3413
                share->rows, 
                key_info->name);
            thd_proc_info(thd, status_msg);
3414 3415 3416 3417 3418 3419
            if (thd->killed) {
                my_error(ER_QUERY_INTERRUPTED, MYF(0));
                error = ER_QUERY_INTERRUPTED;
                goto cleanup;
            }
        }
3420 3421 3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435
    }

    error = 0;

cleanup:
    if (tmp_cursor1) {
        tmp_cursor1->c_close(tmp_cursor1);
        tmp_cursor1 = NULL;
    }
    if (tmp_cursor2) {
        tmp_cursor2->c_close(tmp_cursor2);
        tmp_cursor2 = NULL;
    }
    return error;
}

3436 3437 3438 3439 3440 3441 3442 3443 3444 3445 3446 3447 3448 3449 3450 3451 3452 3453 3454 3455 3456 3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468
int ha_tokudb::is_val_unique(bool* is_unique, uchar* record, KEY* key_info, uint dict_index, DB_TXN* txn) {
    DBT key;
    int error = 0;
    bool has_null;
    DBC* tmp_cursor = NULL;
    struct index_read_info ir_info;
    struct smart_dbt_info info;
    bzero((void *)&key, sizeof(key));
    info.ha = this;
    info.buf = NULL;
    info.keynr = dict_index;

    ir_info.smart_dbt_info = info;
    
    create_dbt_key_for_lookup(
        &key,
        key_info,
        key_buff3,
        record,
        &has_null
        );
    ir_info.orig_key = &key;

    if (has_null) {
        error = 0;
        *is_unique = true;
        goto cleanup;
    }
    
    error = share->key_file[dict_index]->cursor(
        share->key_file[dict_index], 
        txn, 
        &tmp_cursor, 
3469
        DB_SERIALIZABLE
3470 3471 3472 3473 3474 3475 3476 3477 3478 3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490 3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512 3513
        );
    if (error) { goto cleanup; }

    error = tmp_cursor->c_getf_set_range(
        tmp_cursor, 
        0, 
        &key, 
        smart_dbt_callback_lookup, 
        &ir_info
        );
    if (error == DB_NOTFOUND) {
        *is_unique = true;
        error = 0;
        goto cleanup;
    }
    else if (error) {
        goto cleanup;
    }
    if (ir_info.cmp) {
        *is_unique = true;
    }
    else {
        *is_unique = false;
    }
    error = 0;

cleanup:
    if (tmp_cursor) {
        int r = tmp_cursor->c_close(tmp_cursor);
        assert(r==0);
        tmp_cursor = NULL;
    }
    return error;
}

int ha_tokudb::do_uniqueness_checks(uchar* record, DB_TXN* txn, THD* thd) {
    int error;
    //
    // first do uniqueness checks
    //
    if (share->has_unique_keys && !thd_test_options(thd, OPTION_RELAXED_UNIQUE_CHECKS)) {
        for (uint keynr = 0; keynr < table_share->keys; keynr++) {
            bool is_unique_key = table->key_info[keynr].flags & HA_NOSAME;
            bool is_unique = false;
3514 3515 3516 3517 3518 3519
            //
            // don't need to do check for primary key
            //
            if (keynr == primary_key) {
                continue;
            }
3520 3521 3522 3523 3524 3525 3526 3527 3528 3529 3530 3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541
            if (!is_unique_key) {
                continue;
            }
            //
            // if unique key, check uniqueness constraint
            // but, we do not need to check it if the key has a null
            // and we do not need to check it if unique_checks is off
            //
            error = is_val_unique(&is_unique, record, &table->key_info[keynr], keynr, txn);
            if (error) { goto cleanup; }
            if (!is_unique) {
                error = DB_KEYEXIST;
                last_dup_key = keynr;
                goto cleanup;
            }
        }
    }    
    error = 0;
cleanup:
    return error;
}

3542
void ha_tokudb::test_row_packing(uchar* record, DBT* pk_key, DBT* pk_val) {
3543 3544 3545 3546 3547 3548 3549 3550 3551 3552
    int error;
    DBT row, key;
    //
    // variables for testing key packing, only used in some debug modes
    //
    uchar* tmp_pk_key_data = NULL;
    uchar* tmp_pk_val_data = NULL;
    DBT tmp_pk_key;
    DBT tmp_pk_val;
    bool has_null;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3553
    int cmp;
3554 3555 3556 3557 3558 3559 3560 3561 3562 3563 3564 3565 3566 3567 3568 3569 3570 3571 3572 3573 3574 3575 3576 3577 3578 3579 3580

    bzero(&tmp_pk_key, sizeof(DBT));
    bzero(&tmp_pk_val, sizeof(DBT));

    //
    //use for testing the packing of keys
    //
    tmp_pk_key_data = (uchar *)my_malloc(pk_key->size, MYF(MY_WME));
    assert(tmp_pk_key_data);
    tmp_pk_val_data = (uchar *)my_malloc(pk_val->size, MYF(MY_WME));
    assert(tmp_pk_val_data);
    memcpy(tmp_pk_key_data, pk_key->data, pk_key->size);
    memcpy(tmp_pk_val_data, pk_val->data, pk_val->size);
    tmp_pk_key.data = tmp_pk_key_data;
    tmp_pk_key.size = pk_key->size;
    tmp_pk_val.data = tmp_pk_val_data;
    tmp_pk_val.size = pk_val->size;

    for (uint keynr = 0; keynr < table_share->keys; keynr++) {
        u_int32_t tmp_num_bytes = 0;
        uchar* row_desc = NULL;
        u_int32_t desc_size = 0;
        
        if (keynr == primary_key) {
            continue;
        }

Zardosht Kasheff's avatar
Zardosht Kasheff committed
3581
        create_dbt_key_from_table(&key, keynr, key_buff2, record, &has_null); 
3582 3583 3584 3585

        //
        // TEST
        //
3586
        row_desc = (uchar *)share->key_file[keynr]->descriptor->dbt.data;
3587 3588 3589 3590 3591 3592 3593 3594 3595 3596 3597
        row_desc += (*(u_int32_t *)row_desc);
        desc_size = (*(u_int32_t *)row_desc) - 4;
        row_desc += 4;
        tmp_num_bytes = pack_key_from_desc(
            key_buff3,
            row_desc,
            desc_size,
            &tmp_pk_key,
            &tmp_pk_val
            );
        assert(tmp_num_bytes == key.size);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3598
        cmp = memcmp(key_buff3,key_buff2,tmp_num_bytes);
3599 3600 3601 3602 3603 3604
        assert(cmp == 0);

        //
        // test key packing of clustering keys
        //
        if (table->key_info[keynr].flags & HA_CLUSTERING) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3605
            error = pack_row(&row, (const uchar *) record, keynr);
3606
            assert(error == 0);
3607 3608 3609
            uchar* tmp_buff = NULL;
            tmp_buff = (uchar *)my_malloc(alloced_rec_buff_length,MYF(MY_WME));
            assert(tmp_buff);
3610
            row_desc = (uchar *)share->key_file[keynr]->descriptor->dbt.data;
3611 3612 3613 3614 3615 3616 3617 3618 3619 3620 3621
            row_desc += (*(u_int32_t *)row_desc);
            row_desc += (*(u_int32_t *)row_desc);
            desc_size = (*(u_int32_t *)row_desc) - 4;
            row_desc += 4;
            tmp_num_bytes = pack_clustering_val_from_desc(
                tmp_buff,
                row_desc,
                desc_size,
                &tmp_pk_val
                );
            assert(tmp_num_bytes == row.size);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3622
            cmp = memcmp(tmp_buff,rec_buff,tmp_num_bytes);
3623 3624 3625 3626 3627
            assert(cmp == 0);
            my_free(tmp_buff,MYF(MY_ALLOW_ZERO_PTR));
        }
    }

3628 3629 3630 3631
    //
    // copy stuff back out
    //
    error = pack_row(pk_val, (const uchar *) record, primary_key);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3632 3633 3634
    assert(pk_val->size == tmp_pk_val.size);
    cmp = memcmp(pk_val->data, tmp_pk_val_data, pk_val->size);    
    assert( cmp == 0);
3635

3636 3637 3638 3639
    my_free(tmp_pk_key_data,MYF(MY_ALLOW_ZERO_PTR));
    my_free(tmp_pk_val_data,MYF(MY_ALLOW_ZERO_PTR));
}

3640 3641 3642
//
// set the put flags for the main dictionary
//
3643 3644
void ha_tokudb::set_main_dict_put_flags(
    THD* thd, 
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3645
    bool opt_eligible,
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3646
    u_int32_t* put_flags
3647 3648
    ) 
{
3649
    u_int32_t old_prelock_flags = 0;
3650 3651
    uint curr_num_DBs = table->s->keys + test(hidden_primary_key);
    bool in_hot_index = share->num_DBs > curr_num_DBs;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3652 3653
    bool using_ignore_flag_opt = do_ignore_flag_optimization(
            thd, table, share->replace_into_fast);
3654
    //
3655
    // optimization for "REPLACE INTO..." (and "INSERT IGNORE") command
3656
    // if the command is "REPLACE INTO" and the only table
3657 3658
    // is the main table (or all indexes are a subset of the pk), 
    // then we can simply insert the element
3659 3660 3661
    // with DB_YESOVERWRITE. If the element does not exist,
    // it will act as a normal insert, and if it does exist, it 
    // will act as a replace, which is exactly what REPLACE INTO is supposed
3662
    // to do. We cannot do this if otherwise, because then we lose
3663 3664
    // consistency between indexes
    //
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3665 3666
    if (hidden_primary_key) 
    {
3667
        *put_flags = old_prelock_flags;
3668
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3669 3670
    else if (thd_test_options(thd, OPTION_RELAXED_UNIQUE_CHECKS)
            && !is_replace_into(thd) && !is_insert_ignore(thd))
3671
    {
3672
        *put_flags = old_prelock_flags;
3673
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3674 3675
    else if (using_ignore_flag_opt && is_replace_into(thd) 
            && !in_hot_index)
3676
    {
3677
        *put_flags = old_prelock_flags;
3678
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3679
    else if (opt_eligible && using_ignore_flag_opt && is_insert_ignore(thd) 
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3680
            && !in_hot_index)
3681
    {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3682
        *put_flags = DB_NOOVERWRITE_NO_ERROR | old_prelock_flags;
3683 3684 3685
    }
    else 
    {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3686
        *put_flags = DB_NOOVERWRITE | old_prelock_flags;
3687
    }
3688 3689 3690 3691
}

int ha_tokudb::insert_row_to_main_dictionary(uchar* record, DBT* pk_key, DBT* pk_val, DB_TXN* txn) {
    int error = 0;
3692
    u_int32_t put_flags = mult_put_flags[primary_key];
3693 3694 3695 3696 3697
    THD *thd = ha_thd();
    uint curr_num_DBs = table->s->keys + test(hidden_primary_key);

    assert(curr_num_DBs == 1);
    
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3698
    set_main_dict_put_flags(thd, true, &put_flags);
3699

3700 3701 3702 3703 3704 3705 3706
    error = share->file->put(
        share->file, 
        txn, 
        pk_key,
        pk_val, 
        put_flags
        );
3707 3708 3709 3710 3711 3712 3713

    if (error) {
        last_dup_key = primary_key;
        goto cleanup;
    }

cleanup:
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3714 3715 3716
    return error;
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
3717
int ha_tokudb::insert_rows_to_dictionaries_mult(DBT* pk_key, DBT* pk_val, DB_TXN* txn, THD* thd) {
3718
    int error = 0;
3719
    uint curr_num_DBs = share->num_DBs;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3720
    set_main_dict_put_flags(thd, true, &mult_put_flags[primary_key]);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731 3732 3733 3734 3735 3736 3737 3738 3739 3740 3741 3742 3743 3744 3745 3746 3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767
    uint32_t i, flags = mult_put_flags[primary_key];

    // the insert ignore optimization uses DB_NOOVERWRITE_NO_ERROR, 
    // which is not allowed with env->put_multiple. 
    // we have to insert the rows one by one in this case.
    if (flags & DB_NOOVERWRITE_NO_ERROR) {
        DB * src_db = share->key_file[primary_key];
        for (i = 0; i < curr_num_DBs; i++) {
            DB * db = share->key_file[i];
            if (i == primary_key) {
                // if it's the primary key, insert the rows
                // as they are.
                error = db->put(db, txn, pk_key, pk_val, flags);
            } else {
                // generate a row for secondary keys.
                // use our multi put key/rec buffers
                // just as the ydb layer would have in
                // env->put_multiple(), except that
                // we will just do a put() right away.
                error = tokudb_generate_row(db, src_db,
                        &mult_key_dbt[i], &mult_rec_dbt[i], 
                        pk_key, pk_val);
                if (error != 0) {
                    goto out;
                }
                error = db->put(db, txn, &mult_key_dbt[i], 
                        &mult_rec_dbt[i], flags);
            }
            if (error != 0) {
                goto out;
            }
        }
    } else {
        // not insert ignore, so we can use put multiple
        error = db_env->put_multiple(
            db_env, 
            share->key_file[primary_key], 
            txn, 
            pk_key, 
            pk_val,
            curr_num_DBs, 
            share->key_file, 
            mult_key_dbt,
            mult_rec_dbt,
            mult_put_flags
            );
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3768

Zardosht Kasheff's avatar
Zardosht Kasheff committed
3769
out:
3770 3771 3772 3773 3774 3775
    //
    // We break if we hit an error, unless it is a dup key error
    // and MySQL told us to ignore duplicate key errors
    //
    if (error) {
        last_dup_key = primary_key;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3776 3777 3778
    }
    return error;
}
3779

3780 3781
volatile int ha_tokudb_write_row_wait = 0; // debug

Zardosht Kasheff's avatar
Zardosht Kasheff committed
3782 3783 3784 3785 3786 3787 3788 3789
//
// Stores a row in the table, called when handling an INSERT query
// Parameters:
//      [in]    record - a row in MySQL format
// Returns:
//      0 on success
//      error otherwise
//
3790
int ha_tokudb::write_row(uchar * record) {
3791
    TOKUDB_DBUG_ENTER("ha_tokudb::write_row");
3792 3793 3794

    while (ha_tokudb_write_row_wait) sleep(1); // debug

3795
    DBT row, prim_key;
3796
    int error;
3797
    THD *thd = ha_thd();
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3798
    bool has_null;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3799 3800
    DB_TXN* sub_trans = NULL;
    DB_TXN* txn = NULL;
3801
    tokudb_trx_data *trx = NULL;
3802
    uint curr_num_DBs;
3803
    bool create_sub_trans = false;
3804

Zardosht Kasheff's avatar
Zardosht Kasheff committed
3805 3806 3807 3808
    //
    // some crap that needs to be done because MySQL does not properly abstract
    // this work away from us, namely filling in auto increment and setting auto timestamp
    //
3809
    statistic_increment(table->in_use->status_var.ha_write_count, &LOCK_status);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3810
    if (table->timestamp_field_type & TIMESTAMP_AUTO_SET_ON_INSERT) {
3811
        table->timestamp_field->set_time();
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3812 3813
    }
    if (table->next_number_field && record == table->record[0]) {
3814
        update_auto_increment();
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3815
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3816 3817 3818 3819 3820 3821 3822 3823 3824 3825 3826

    //
    // check to see if some value for the auto increment column that is bigger
    // than anything else til now is being used. If so, update the metadata to reflect it
    // the goal here is we never want to have a dup key error due to a bad increment
    // of the auto inc field.
    //
    if (share->has_auto_inc && record == table->record[0]) {
        pthread_mutex_lock(&share->mutex);
        ulonglong curr_auto_inc = retrieve_auto_increment(
            table->field[share->ai_field_index]->key_type(), 
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3827
            field_offset(table->field[share->ai_field_index], table),
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3828 3829 3830
            record
            );
        if (curr_auto_inc > share->last_auto_increment) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3831
            share->last_auto_increment = curr_auto_inc;
3832 3833
            if (delay_updating_ai_metadata) {
                ai_metadata_update_required = true;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3834
            }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3835 3836 3837 3838
            else {
                update_max_auto_inc(share->status_block, share->last_auto_increment);
            }
         }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3839 3840 3841
        pthread_mutex_unlock(&share->mutex);
    }

3842 3843 3844
    //
    // grab reader lock on numDBs_lock
    //
3845 3846 3847 3848 3849 3850 3851 3852 3853 3854 3855
    if (!num_DBs_locked_in_bulk) {
        rw_rdlock(&share->num_DBs_lock);
    }
    else {
        lock_count++;
        if (lock_count >= 2000) {
            rw_unlock(&share->num_DBs_lock);
            rw_rdlock(&share->num_DBs_lock);
            lock_count = 0;
        }
    }
3856
    curr_num_DBs = share->num_DBs;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3857 3858 3859 3860
    
    if (hidden_primary_key) {
        get_auto_primary_key(current_ident);
    }
3861

3862 3863 3864
    if (table_share->blob_fields) {
        if (fix_rec_buff_for_blob(max_row_length(record))) {
            error = HA_ERR_OUT_OF_MEM;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3865 3866 3867
            goto cleanup;
        }
    }
3868

Zardosht Kasheff's avatar
Zardosht Kasheff committed
3869 3870
    create_dbt_key_from_table(&prim_key, primary_key, primary_key_buff, record, &has_null);
    if ((error = pack_row(&row, (const uchar *) record, primary_key))){
3871
        goto cleanup;
3872 3873
    }

3874
    create_sub_trans = (using_ignore && !(do_ignore_flag_optimization(thd,table,share->replace_into_fast)));
3875
    if (create_sub_trans) {
3876
        error = db_env->txn_begin(db_env, transaction, &sub_trans, DB_INHERIT_ISOLATION);
3877 3878 3879
        if (error) {
            goto cleanup;
        }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3880
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3881
    
3882
    txn = create_sub_trans ? sub_trans : transaction;    
3883 3884

    if (tokudb_debug & TOKUDB_DEBUG_CHECK_KEY) {
3885
        test_row_packing(record,&prim_key,&row);
3886
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3887

3888 3889 3890 3891 3892 3893
    if (loader) {
        error = loader->put(loader, &prim_key, &row);
        if (error) {
            abort_loader = true;
            goto cleanup;
        }
3894 3895
    }
    else {
3896 3897 3898 3899 3900 3901 3902
        if (curr_num_DBs == 1) {
            error = insert_row_to_main_dictionary(record,&prim_key, &row, txn);
            if (error) { goto cleanup; }
        }
        else {
            error = do_uniqueness_checks(record, txn, thd);
            if (error) { goto cleanup; }
3903

3904 3905 3906
            error = insert_rows_to_dictionaries_mult(&prim_key, &row, txn, thd);
            if (error) { goto cleanup; }
        }
3907
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3908

3909
    trx = (tokudb_trx_data *) thd_data_get(thd, tokudb_hton->slot);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3910 3911
    if (!error) {
        added_rows++;
3912 3913
        trx->stmt_progress.inserted++;
        track_progress(thd);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3914 3915
    }
cleanup:
3916 3917 3918
    if (!num_DBs_locked_in_bulk) {
       rw_unlock(&share->num_DBs_lock);
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3919
    if (error == DB_KEYEXIST) {
3920
        error = HA_ERR_FOUND_DUPP_KEY;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3921
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3922 3923 3924 3925 3926
    if (sub_trans) {
        // no point in recording error value of abort.
        // nothing we can do about it anyway and it is not what
        // we want to return.
        if (error) {
3927
            abort_txn(sub_trans);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3928 3929
        }
        else {
3930
            commit_txn(sub_trans, DB_TXN_NOSYNC);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3931 3932
        }
    }
3933
    TOKUDB_DBUG_RETURN(error);
3934 3935 3936 3937 3938 3939 3940 3941 3942 3943 3944 3945 3946 3947 3948 3949 3950 3951 3952 3953
}

/* Compare if a key in a row has changed */
int ha_tokudb::key_cmp(uint keynr, const uchar * old_row, const uchar * new_row) {
    KEY_PART_INFO *key_part = table->key_info[keynr].key_part;
    KEY_PART_INFO *end = key_part + table->key_info[keynr].key_parts;

    for (; key_part != end; key_part++) {
        if (key_part->null_bit) {
            if ((old_row[key_part->null_offset] & key_part->null_bit) != (new_row[key_part->null_offset] & key_part->null_bit))
                return 1;
        }
        if (key_part->key_part_flag & (HA_BLOB_PART | HA_VAR_LENGTH_PART)) {

            if (key_part->field->cmp_binary((uchar *) (old_row + key_part->offset), (uchar *) (new_row + key_part->offset), (ulong) key_part->length))
                return 1;
        } else {
            if (memcmp(old_row + key_part->offset, new_row + key_part->offset, key_part->length))
                return 1;
        }
3954
    }
3955
    return 0;
3956 3957
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
3958 3959 3960 3961 3962 3963 3964 3965 3966
//
// Updates a row in the table, called when handling an UPDATE query
// Parameters:
//      [in]    old_row - row to be updated, in MySQL format
//      [in]    new_row - new row, in MySQL format
// Returns:
//      0 on success
//      error otherwise
//
3967
int ha_tokudb::update_row(const uchar * old_row, uchar * new_row) {
3968
    TOKUDB_DBUG_ENTER("update_row");
3969
    DBT prim_key, old_prim_key, prim_row, old_prim_row;
3970 3971
    int error;
    bool primary_key_changed;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3972 3973
    bool has_null;
    THD* thd = ha_thd();
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3974 3975
    DB_TXN* sub_trans = NULL;
    DB_TXN* txn = NULL;
3976
    tokudb_trx_data* trx = (tokudb_trx_data *) thd_data_get(thd, tokudb_hton->slot);
3977
    uint curr_num_DBs;
3978 3979

    LINT_INIT(error);
3980 3981 3982
    bzero((void *) &prim_key, sizeof(prim_key));
    bzero((void *) &old_prim_key, sizeof(old_prim_key));
    bzero((void *) &prim_row, sizeof(prim_row));
3983
    bzero((void *) &old_prim_row, sizeof(old_prim_row));
3984 3985


3986
    statistic_increment(table->in_use->status_var.ha_update_count, &LOCK_status);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3987
    if (table->timestamp_field_type & TIMESTAMP_AUTO_SET_ON_UPDATE) {
3988
        table->timestamp_field->set_time();
Zardosht Kasheff's avatar
Zardosht Kasheff committed
3989
    }
3990

Zardosht Kasheff's avatar
Zardosht Kasheff committed
3991 3992 3993 3994 3995 3996 3997 3998 3999 4000
    //
    // check to see if some value for the auto increment column that is bigger
    // than anything else til now is being used. If so, update the metadata to reflect it
    // the goal here is we never want to have a dup key error due to a bad increment
    // of the auto inc field.
    //
    if (share->has_auto_inc && new_row == table->record[0]) {
        pthread_mutex_lock(&share->mutex);
        ulonglong curr_auto_inc = retrieve_auto_increment(
            table->field[share->ai_field_index]->key_type(), 
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4001
            field_offset(table->field[share->ai_field_index], table),
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4002 4003 4004 4005 4006 4007 4008 4009 4010 4011 4012
            new_row
            );
        if (curr_auto_inc > share->last_auto_increment) {
            error = update_max_auto_inc(share->status_block, curr_auto_inc);
            if (!error) {
                share->last_auto_increment = curr_auto_inc;
            }
        }
        pthread_mutex_unlock(&share->mutex);
    }

4013 4014 4015 4016 4017 4018
    //
    // grab reader lock on numDBs_lock
    //
    rw_rdlock(&share->num_DBs_lock);
    curr_num_DBs = share->num_DBs;

4019 4020 4021 4022 4023
    if (using_ignore) {
        error = db_env->txn_begin(db_env, transaction, &sub_trans, DB_INHERIT_ISOLATION);
        if (error) {
            goto cleanup;
        }
4024
    }
4025
    txn = using_ignore ? sub_trans : transaction;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4026 4027


4028 4029 4030 4031 4032 4033
    if (hidden_primary_key) {
        primary_key_changed = 0;
        bzero((void *) &prim_key, sizeof(prim_key));
        prim_key.data = (void *) current_ident;
        prim_key.size = TOKUDB_HIDDEN_PRIMARY_KEY_LENGTH;
        old_prim_key = prim_key;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4034 4035
    } 
    else {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4036
        create_dbt_key_from_table(&prim_key, primary_key, key_buff, new_row, &has_null);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4037
        if ((primary_key_changed = key_cmp(primary_key, old_row, new_row))) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4038
            create_dbt_key_from_table(&old_prim_key, primary_key, primary_key_buff, old_row, &has_null);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4039 4040
        }
        else {
4041
            old_prim_key = prim_key;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4042
        }
4043 4044
    }

4045 4046 4047 4048 4049
    //
    // do uniqueness checks
    //
    if (share->has_unique_keys && !thd_test_options(thd, OPTION_RELAXED_UNIQUE_CHECKS)) {
        for (uint keynr = 0; keynr < table_share->keys; keynr++) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4050
            bool is_unique_key = table->key_info[keynr].flags & HA_NOSAME;
4051 4052
            if (keynr == primary_key) {
                continue;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4053
            }
4054 4055 4056 4057 4058 4059 4060 4061 4062 4063 4064
            if (is_unique_key) {
                bool key_changed = key_cmp(keynr, old_row, new_row);
                if (key_changed) {
                    bool is_unique;
                    error = is_val_unique(&is_unique, new_row, &table->key_info[keynr], keynr, txn);
                    if (error) goto cleanup;
                    if (!is_unique) {
                        error = DB_KEYEXIST;
                        last_dup_key = keynr;
                        goto cleanup;
                    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4065
                }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4066
            }
4067 4068 4069 4070 4071 4072 4073 4074 4075 4076 4077
        }
    }
    
    if (table_share->blob_fields) {
        if (fix_rec_buff_for_blob(max_row_length(new_row))) {
            error = HA_ERR_OUT_OF_MEM;
            goto cleanup;
        }
        if (fix_rec_update_buff_for_blob(max_row_length(old_row))) {
            error = HA_ERR_OUT_OF_MEM;
            goto cleanup;
4078 4079
        }
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4080

4081 4082 4083 4084 4085 4086
    error = pack_row(&prim_row, new_row, primary_key);
    if (error) { goto cleanup; }

    error = pack_old_row_for_update(&old_prim_row, old_row, primary_key);
    if (error) { goto cleanup; }

Zardosht Kasheff's avatar
Zardosht Kasheff committed
4087
    set_main_dict_put_flags(thd, false, &mult_put_flags[primary_key]);
4088 4089 4090 4091 4092 4093 4094 4095 4096 4097 4098 4099 4100 4101 4102 4103 4104 4105

    error = db_env->update_multiple(
        db_env, 
        share->key_file[primary_key], 
        txn,
        &old_prim_key, 
        &old_prim_row,
        &prim_key, 
        &prim_row,
        curr_num_DBs, 
        share->key_file,
        mult_put_flags,
        2*curr_num_DBs, 
        mult_key_dbt,
        curr_num_DBs, 
        mult_rec_dbt
        );
    
4106 4107 4108 4109
    if (error == DB_KEYEXIST) {
        last_dup_key = primary_key;
    }    
    else if (!error) {
4110 4111
        trx->stmt_progress.updated++;
        track_progress(thd);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4112
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4113 4114


Zardosht Kasheff's avatar
Zardosht Kasheff committed
4115
cleanup:
4116
    rw_unlock(&share->num_DBs_lock);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4117
    if (error == DB_KEYEXIST) {
4118
        error = HA_ERR_FOUND_DUPP_KEY;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4119
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4120 4121 4122 4123 4124
    if (sub_trans) {
        // no point in recording error value of abort.
        // nothing we can do about it anyway and it is not what
        // we want to return.
        if (error) {
4125
            abort_txn(sub_trans);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4126 4127
        }
        else {
4128
            commit_txn(sub_trans, DB_TXN_NOSYNC);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4129 4130
        }
    }
4131
    TOKUDB_DBUG_RETURN(error);
4132 4133
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
4134
//
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4135
// Deletes a row in the table, called when handling a DELETE query
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4136 4137 4138 4139 4140 4141
// Parameters:
//      [in]    record - row to be deleted, in MySQL format
// Returns:
//      0 on success
//      error otherwise
//
4142
int ha_tokudb::delete_row(const uchar * record) {
4143
    TOKUDB_DBUG_ENTER("ha_tokudb::delete_row");
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4144
    int error = ENOSYS;
4145
    DBT row, prim_key;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4146
    bool has_null;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4147
    THD* thd = ha_thd();
4148
    uint curr_num_DBs;
4149
    tokudb_trx_data* trx = (tokudb_trx_data *) thd_data_get(thd, tokudb_hton->slot);;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4150

4151 4152
    statistic_increment(table->in_use->status_var.ha_delete_count, &LOCK_status);

4153 4154 4155 4156 4157 4158
    //
    // grab reader lock on numDBs_lock
    //
    rw_rdlock(&share->num_DBs_lock);
    curr_num_DBs = share->num_DBs;

Zardosht Kasheff's avatar
Zardosht Kasheff committed
4159
    create_dbt_key_from_table(&prim_key, primary_key, key_buff, record, &has_null);
4160 4161 4162 4163 4164
    if (table_share->blob_fields) {
        if (fix_rec_buff_for_blob(max_row_length(record))) {
            error = HA_ERR_OUT_OF_MEM;
            goto cleanup;
        }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4165
    }
4166 4167 4168
    if ((error = pack_row(&row, (const uchar *) record, primary_key))){
        goto cleanup;
    }
4169 4170 4171 4172 4173 4174 4175 4176 4177 4178 4179 4180

    error = db_env->del_multiple(
        db_env, 
        share->key_file[primary_key], 
        transaction, 
        &prim_key, 
        &row,
        curr_num_DBs, 
        share->key_file, 
        mult_key_dbt,
        mult_del_flags
        );
4181

Zardosht Kasheff's avatar
Zardosht Kasheff committed
4182 4183
    if (error) {
        DBUG_PRINT("error", ("Got error %d", error));
4184
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4185
    else {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4186
        deleted_rows++;
4187 4188
        trx->stmt_progress.deleted++;
        track_progress(thd);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4189
    }
4190
cleanup:
4191
    rw_unlock(&share->num_DBs_lock);
4192
    TOKUDB_DBUG_RETURN(error);
4193 4194
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
4195 4196 4197 4198 4199 4200 4201 4202 4203 4204 4205 4206 4207 4208 4209 4210
//
// takes as input table->read_set and table->write_set
// and puts list of field indexes that need to be read in
// unpack_row in the member variables fixed_cols_for_query
// and var_cols_for_query
//
void ha_tokudb::set_query_columns(uint keynr) {
    u_int32_t curr_fixed_col_index = 0;
    u_int32_t curr_var_col_index = 0;
    read_key = false;
    read_blobs = false;
    //
    // i know this is probably confusing and will need to be explained better
    //
    uint key_index = 0;

Zardosht Kasheff's avatar
Zardosht Kasheff committed
4211
    if (keynr == primary_key || keynr == MAX_KEY) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4212 4213 4214 4215 4216 4217 4218 4219 4220 4221
        key_index = primary_key;
    }
    else {
        key_index = (table->key_info[keynr].flags & HA_CLUSTERING ? keynr : primary_key);
    }
    for (uint i = 0; i < table_share->fields; i++) {
        if (bitmap_is_set(table->read_set,i) || 
            bitmap_is_set(table->write_set,i)
            ) 
        {
4222
            if (bitmap_is_set(&share->kc_info.key_filters[key_index],i)) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4223 4224 4225 4226 4227 4228
                read_key = true;
            }
            else {
                //
                // if fixed field length
                //
4229
                if (share->kc_info.field_lengths[i] != 0) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4230 4231 4232 4233 4234 4235 4236 4237 4238
                    //
                    // save the offset into the list
                    //
                    fixed_cols_for_query[curr_fixed_col_index] = i;
                    curr_fixed_col_index++;
                }
                //
                // varchar or varbinary
                //
4239
                else if (share->kc_info.length_bytes[i] != 0) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4240 4241 4242 4243 4244 4245 4246 4247 4248 4249 4250 4251 4252 4253 4254 4255
                    var_cols_for_query[curr_var_col_index] = i;
                    curr_var_col_index++;
                }
                //
                // it is a blob
                //
                else {
                    read_blobs = true;
                }
            }
        }
    }
    num_fixed_cols_for_query = curr_fixed_col_index;
    num_var_cols_for_query = curr_var_col_index;
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
4256
void ha_tokudb::column_bitmaps_signal() {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4257 4258 4259 4260
    //
    // if we have max number of indexes, then MAX_KEY == primary_key
    //
    if (active_index != MAX_KEY || active_index == primary_key) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4261 4262 4263
        set_query_columns(active_index);
    }
}
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4264

Zardosht Kasheff's avatar
Zardosht Kasheff committed
4265 4266 4267 4268 4269 4270 4271 4272
//
// Notification that a scan of entire secondary table is about
// to take place. Will pre acquire table read lock
// Returns:
//      0 on success
//      error otherwise
//
int ha_tokudb::prepare_index_scan() {
4273
    int error = 0;
4274
    HANDLE_INVALID_CURSOR();
4275
    error = prelock_range(NULL, NULL);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4276 4277 4278 4279 4280 4281 4282 4283 4284
    if (error) { last_cursor_error = error; goto cleanup; }

    range_lock_grabbed = true;
    error = 0;
cleanup:
    return error;
}


Zardosht Kasheff's avatar
Zardosht Kasheff committed
4285 4286 4287 4288 4289 4290 4291
//
// Notification that a range query getting all elements that equal a key
//  to take place. Will pre acquire read lock
// Returns:
//      0 on success
//      error otherwise
//
4292
int ha_tokudb::prepare_index_key_scan(const uchar * key, uint key_len) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4293 4294
    int error = 0;
    DBT start_key, end_key;
4295
    THD* thd = ha_thd();
4296
    HANDLE_INVALID_CURSOR();
4297 4298 4299 4300
    pack_key(&start_key, active_index, prelocked_left_range, key, key_len, COL_NEG_INF);
    prelocked_left_range_size = start_key.size;
    pack_key(&end_key, active_index, prelocked_right_range, key, key_len, COL_POS_INF);
    prelocked_right_range_size = end_key.size;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4301

4302 4303 4304 4305 4306 4307
    error = cursor->c_pre_acquire_range_lock(
        cursor, 
        &start_key, 
        &end_key 
        );

Zardosht Kasheff's avatar
Zardosht Kasheff committed
4308 4309 4310 4311 4312
    if (error){ 
        goto cleanup; 
    }

    range_lock_grabbed = true;
4313
    doing_bulk_fetch = (thd_sql_command(thd) == SQLCOM_SELECT);
4314 4315
    bulk_fetch_iteration = 0;
    rows_fetched_using_bulk_fetch = 0;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4316 4317 4318 4319 4320
    error = 0;
cleanup:
    if (error) {
        last_cursor_error = error;
        //
4321 4322
        // cursor should be initialized here, but in case it is not, 
        // we still check
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4323 4324
        //
        if (cursor) {
4325 4326
            int r = cursor->c_close(cursor);
            assert(r==0);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4327 4328 4329 4330 4331 4332
            cursor = NULL;
        }
    }
    return error;
}

4333 4334 4335 4336
void ha_tokudb::invalidate_bulk_fetch() {
    bytes_used_in_range_query_buff= 0;
    curr_range_query_buff_offset = 0;
}
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4337

4338 4339
volatile int ha_tokudb_index_init_wait = 0; // debug

Zardosht Kasheff's avatar
Zardosht Kasheff committed
4340 4341 4342 4343 4344 4345 4346 4347 4348
//
// Initializes local cursor on DB with index keynr
// Parameters:
//          keynr - key (index) number
//          sorted - 1 if result MUST be sorted according to index
// Returns:
//      0 on success
//      error otherwise
//
4349
int ha_tokudb::index_init(uint keynr, bool sorted) {
4350
    TOKUDB_DBUG_ENTER("ha_tokudb::index_init %p %d", this, keynr);
4351 4352 4353

    while (ha_tokudb_index_init_wait) sleep(1); // debug

4354
    int error;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4355
    THD* thd = ha_thd(); 
4356 4357 4358 4359 4360 4361 4362 4363
    DBUG_PRINT("enter", ("table: '%s'  key: %d", table_share->table_name.str, keynr));

    /*
       Under some very rare conditions (like full joins) we may already have
       an active cursor at this point
     */
    if (cursor) {
        DBUG_PRINT("note", ("Closing active cursor"));
4364 4365
        int r = cursor->c_close(cursor);
        assert(r==0);
4366 4367
    }
    active_index = keynr;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4368
    last_cursor_error = 0;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4369
    range_lock_grabbed = false;
4370
    DBUG_ASSERT(keynr <= table->s->keys);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4371
    DBUG_ASSERT(share->key_file[keynr]);
4372
    cursor_flags = get_cursor_isolation_flags(lock.type, thd);
4373
    if (use_write_locks) {
4374
        cursor_flags |= DB_RMW;
4375 4376 4377 4378
    }
    if (get_disable_prefetching(thd)) {
        cursor_flags |= DBC_DISABLE_PREFETCHING;
    }
4379 4380
    if ((error = share->key_file[keynr]->cursor(share->key_file[keynr], transaction, &cursor, cursor_flags))) {
        if (error == TOKUDB_MVCC_DICTIONARY_TOO_NEW) {
4381
            error = HA_ERR_TABLE_DEF_CHANGED;
4382 4383
            my_error(ER_TABLE_DEF_CHANGED, MYF(0));
        }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4384 4385 4386 4387
        if (error == DB_LOCK_NOTGRANTED) {
            error = HA_ERR_LOCK_WAIT_TIMEOUT;
            my_error(ER_LOCK_WAIT_TIMEOUT, MYF(0));
        }
4388
        table->status = STATUS_NOT_FOUND;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4389
        last_cursor_error = error;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4390
        cursor = NULL;             // Safety
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4391
        goto exit;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4392
    }
4393
    bzero((void *) &last_key, sizeof(last_key));
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4394 4395 4396 4397 4398 4399 4400 4401

    if (thd_sql_command(thd) == SQLCOM_SELECT) {
        set_query_columns(keynr);
        unpack_entire_row = false;
    }
    else {
        unpack_entire_row = true;
    }
4402 4403
    invalidate_bulk_fetch();
    doing_bulk_fetch = false;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4404 4405
    error = 0;
exit:
4406
    TOKUDB_DBUG_RETURN(error);
4407 4408
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
4409 4410 4411
//
// closes the local cursor
//
4412
int ha_tokudb::index_end() {
4413
    TOKUDB_DBUG_ENTER("ha_tokudb::index_end %p", this);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4414
    range_lock_grabbed = false;
4415 4416
    if (cursor) {
        DBUG_PRINT("enter", ("table: '%s'", table_share->table_name.str));
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4417 4418
        int r = cursor->c_close(cursor);
        assert(r==0);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4419
        cursor = NULL;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4420
        last_cursor_error = 0;
4421
    }
4422
    active_index = MAX_KEY;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4423 4424 4425 4426 4427 4428 4429 4430 4431

    //
    // reset query variables
    //
    unpack_entire_row = true;
    read_blobs = true;
    read_key = true;
    num_fixed_cols_for_query = 0;
    num_var_cols_for_query = 0;
4432 4433 4434 4435

    invalidate_bulk_fetch();
    doing_bulk_fetch = false;

Zardosht Kasheff's avatar
Zardosht Kasheff committed
4436
    TOKUDB_DBUG_RETURN(0);
4437 4438
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
4439 4440 4441

int ha_tokudb::handle_cursor_error(int error, int err_to_return, uint keynr) {
    TOKUDB_DBUG_ENTER("ha_tokudb::handle_cursor_error");
4442
    if (error) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4443 4444 4445
        if (error == DB_LOCK_NOTGRANTED) {
            error = HA_ERR_LOCK_WAIT_TIMEOUT;
        }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4446
        last_cursor_error = error;
4447
        table->status = STATUS_NOT_FOUND;
4448 4449
        int r = cursor->c_close(cursor);
        assert(r==0);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4450
        cursor = NULL;
Barry Perlman's avatar
Barry Perlman committed
4451
        if (error == DB_NOTFOUND) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4452
            error = err_to_return;
4453
            if ((share->key_file[keynr]->cursor(share->key_file[keynr], transaction, &cursor, cursor_flags))) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4454 4455 4456
                cursor = NULL;             // Safety
            }
        }
4457
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4458 4459 4460 4461 4462 4463 4464 4465 4466 4467
    TOKUDB_DBUG_RETURN(error);
}


//
// Helper function for read_row and smart_dbt_callback_xxx functions
// When using a hidden primary key, upon reading a row, 
// we set the current_ident field to whatever the primary key we retrieved
// was
//
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4468
void ha_tokudb::extract_hidden_primary_key(uint keynr, DBT const *found_key) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4469
    //
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4470
    // extract hidden primary key to current_ident
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4471 4472
    //
    if (hidden_primary_key) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4473 4474 4475
        if (keynr == primary_key) {
            memcpy_fixed(current_ident, (char *) found_key->data, TOKUDB_HIDDEN_PRIMARY_KEY_LENGTH);
        }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4476
        //
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4477
        // if secondary key, hidden primary key is at end of found_key
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4478
        //
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4479
        else {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4480 4481 4482 4483 4484 4485
            memcpy_fixed(
                current_ident, 
                (char *) found_key->data + found_key->size - TOKUDB_HIDDEN_PRIMARY_KEY_LENGTH, 
                TOKUDB_HIDDEN_PRIMARY_KEY_LENGTH
                );
        }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4486
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4487 4488
}

4489

4490
int ha_tokudb::read_row_callback (uchar * buf, uint keynr, DBT const *row, DBT const *found_key) {
4491
    assert(keynr == primary_key);
4492
    return unpack_row(buf, row,found_key, keynr);
4493 4494
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
4495 4496 4497 4498 4499 4500 4501 4502 4503 4504
//
// Reads the contents of row and found_key, DBT's retrieved from the DB associated to keynr, into buf
// This function assumes that we are using a covering index, as a result, if keynr is the primary key,
// we do not read row into buf
// Parameters:
//      [out]   buf - buffer for the row, in MySQL format
//              keynr - index into key_file that represents DB we are currently operating on.
//      [in]    row - the row that has been read from the preceding DB call
//      [in]    found_key - key used to retrieve the row
//
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4505
void ha_tokudb::read_key_only(uchar * buf, uint keynr, DBT const *found_key) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4506 4507
    TOKUDB_DBUG_ENTER("ha_tokudb::read_key_only");
    table->status = 0;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4508 4509 4510 4511
    //
    // only case when we do not unpack the key is if we are dealing with the main dictionary
    // of a table with a hidden primary key
    //
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4512 4513 4514
    if (!(hidden_primary_key && keynr == primary_key)) {
        unpack_key(buf, found_key, keynr);
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4515 4516 4517 4518 4519 4520 4521 4522 4523 4524 4525 4526 4527
    DBUG_VOID_RETURN;
}

//
// Helper function used to try to retrieve the entire row
// If keynr is associated with the main table, reads contents of found_key and row into buf, otherwise,
// makes copy of primary key and saves it to last_key. This can later be used to retrieve the entire row
// Parameters:
//      [out]   buf - buffer for the row, in MySQL format
//              keynr - index into key_file that represents DB we are currently operating on.
//      [in]    row - the row that has been read from the preceding DB call
//      [in]    found_key - key used to retrieve the row
//
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4528
int ha_tokudb::read_primary_key(uchar * buf, uint keynr, DBT const *row, DBT const *found_key) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4529
    TOKUDB_DBUG_ENTER("ha_tokudb::read_primary_key");
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4530
    int error = 0;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4531
    table->status = 0;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4532 4533 4534 4535
    //
    // case where we read from secondary table that is not clustered
    //
    if (keynr != primary_key && !(table->key_info[keynr].flags & HA_CLUSTERING)) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4536
        bool has_null;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4537
        //
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4538 4539
        // create a DBT that has the same data as row, this is inefficient
        // extract_hidden_primary_key MUST have been called before this
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4540 4541
        //
        bzero((void *) &last_key, sizeof(last_key));
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4542 4543 4544 4545 4546 4547 4548 4549 4550 4551
        if (!hidden_primary_key) {
            unpack_key(buf, found_key, keynr);
        }
        create_dbt_key_from_table(
            &last_key, 
            primary_key,
            key_buff,
            buf,
            &has_null
            );
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4552
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4553
    //
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4554
    // else read from clustered/primary key
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4555
    //
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4556
    else {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4557 4558
        error = unpack_row(buf, row, found_key, keynr);
        if (error) { goto exit; }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4559 4560
    }
    if (found_key) { DBUG_DUMP("read row key", (uchar *) found_key->data, found_key->size); }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4561 4562 4563
    error = 0;
exit:
    TOKUDB_DBUG_RETURN(error);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4564 4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 4575
}

//
// This function reads an entire row into buf. This function also assumes that
// the key needed to retrieve the row is stored in the member variable last_key
// Parameters:
//      [out]   buf - buffer for the row, in MySQL format
// Returns:
//      0 on success, error otherwise
//
int ha_tokudb::read_full_row(uchar * buf) {
    TOKUDB_DBUG_ENTER("ha_tokudb::read_full_row");
4576
    int error = 0;
4577 4578 4579 4580
    struct smart_dbt_info info;
    info.ha = this;
    info.buf = buf;
    info.keynr = primary_key;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4581
    //
4582
    // assumes key is stored in this->last_key
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4583
    //
4584 4585 4586 4587 4588 4589 4590 4591 4592 4593

    error = share->file->getf_set(
        share->file, 
        transaction, 
        cursor_flags, 
        &last_key, 
        smart_dbt_callback_rowread_ptquery, 
        &info
        );

4594
    if (error) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4595 4596 4597
        if (error == DB_LOCK_NOTGRANTED) {
            error = HA_ERR_LOCK_WAIT_TIMEOUT;
        }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4598 4599 4600 4601
        table->status = STATUS_NOT_FOUND;
        TOKUDB_DBUG_RETURN(error == DB_NOTFOUND ? HA_ERR_CRASHED : error);
    }

Zardosht Kasheff's avatar
Zardosht Kasheff committed
4602
    TOKUDB_DBUG_RETURN(error);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4603 4604 4605
}


Zardosht Kasheff's avatar
Zardosht Kasheff committed
4606 4607 4608 4609 4610 4611 4612 4613 4614 4615 4616 4617
// 
// Reads the next row matching to the key, on success, advances cursor 
// Parameters: 
//      [out]   buf - buffer for the next row, in MySQL format 
//      [in]     key - key value 
//                keylen - length of key 
// Returns: 
//      0 on success 
//      HA_ERR_END_OF_FILE if not found 
//      error otherwise 
// 
int ha_tokudb::index_next_same(uchar * buf, const uchar * key, uint keylen) { 
4618 4619 4620
    TOKUDB_DBUG_ENTER("ha_tokudb::index_next_same");
    statistic_increment(table->in_use->status_var.ha_read_next_count, &LOCK_status);

Zardosht Kasheff's avatar
Zardosht Kasheff committed
4621 4622 4623 4624
    DBT curr_key;
    DBT found_key;
    bool has_null;
    int cmp;
4625
    int error = get_next(buf, 1);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4626 4627 4628 4629 4630 4631
    if (error) {
        goto cleanup;
    }
    //
    // now do the comparison
    //
4632
    pack_key(&curr_key, active_index, key_buff2, key, keylen, COL_ZERO);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4633 4634 4635 4636 4637
    create_dbt_key_from_table(&found_key,active_index,key_buff3,buf,&has_null);
    cmp = tokudb_prefix_cmp_dbt_key(share->key_file[active_index], &curr_key, &found_key);
    if (cmp) {
        error = HA_ERR_END_OF_FILE; 
    }
4638 4639 4640 4641

cleanup: 
    TOKUDB_DBUG_RETURN(error);
} 
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4642 4643


Zardosht Kasheff's avatar
Zardosht Kasheff committed
4644 4645 4646 4647 4648 4649 4650 4651 4652 4653 4654 4655 4656 4657 4658 4659 4660
//
// According to InnoDB handlerton: Positions an index cursor to the index 
// specified in keynr. Fetches the row if any
// Parameters:
//      [out]       buf - buffer for the  returned row
//      [in]         key - key value, according to InnoDB, if NULL, 
//                              position cursor at start or end of index,
//                              not sure if this is done now
//                    key_len - length of key
//                    find_flag - according to InnoDB, search flags from my_base.h
// Returns:
//      0 on success
//      HA_ERR_KEY_NOT_FOUND if not found (per InnoDB), 
//          we seem to return HA_ERR_END_OF_FILE if find_flag != HA_READ_KEY_EXACT
//          TODO: investigate this for correctness
//      error otherwise
//
4661 4662
int ha_tokudb::index_read(uchar * buf, const uchar * key, uint key_len, enum ha_rkey_function find_flag) {
    TOKUDB_DBUG_ENTER("ha_tokudb::index_read %p find %d", this, find_flag);
4663
    invalidate_bulk_fetch();
4664
    // TOKUDB_DBUG_DUMP("key=", key, key_len);
4665
    DBT row;
4666
    DBT lookup_key;
4667
    int error = 0;    
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4668
    u_int32_t flags = 0;
4669 4670
    THD* thd = ha_thd();
    tokudb_trx_data* trx = (tokudb_trx_data *) thd_data_get(thd, tokudb_hton->slot);;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4671
    struct smart_dbt_info info;
4672
    struct index_read_info ir_info;
4673

Zardosht Kasheff's avatar
Zardosht Kasheff committed
4674
    HANDLE_INVALID_CURSOR();
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4675

4676 4677 4678
    table->in_use->status_var.ha_read_key_count++;
    bzero((void *) &row, sizeof(row));

Zardosht Kasheff's avatar
Zardosht Kasheff committed
4679 4680 4681 4682
    info.ha = this;
    info.buf = buf;
    info.keynr = active_index;

4683
    ir_info.smart_dbt_info = info;
4684
    ir_info.cmp = 0;
4685

4686
    flags = SET_PRELOCK_FLAG(0);
4687
    switch (find_flag) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4688
    case HA_READ_KEY_EXACT: /* Find first record else error */
4689
        pack_key(&lookup_key, active_index, key_buff3, key, key_len, COL_NEG_INF);
4690
        ir_info.orig_key = &lookup_key;
4691 4692 4693

        error = cursor->c_getf_set_range(cursor, flags,
                &lookup_key, SMART_DBT_IR_CALLBACK, &ir_info);
4694 4695
        if (ir_info.cmp) {
            error = DB_NOTFOUND;
4696
        }
4697
        break;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4698
    case HA_READ_AFTER_KEY: /* Find next rec. after key-record */
4699
        pack_key(&lookup_key, active_index, key_buff3, key, key_len, COL_POS_INF);
4700 4701
        error = cursor->c_getf_set_range(cursor, flags,
                &lookup_key, SMART_DBT_CALLBACK, &info);
4702
        break;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4703
    case HA_READ_BEFORE_KEY: /* Find next rec. before key-record */
4704
        pack_key(&lookup_key, active_index, key_buff3, key, key_len, COL_NEG_INF);
4705 4706
        error = cursor->c_getf_set_range_reverse(cursor, flags, 
                &lookup_key, SMART_DBT_CALLBACK, &info);
4707
        break;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4708
    case HA_READ_KEY_OR_NEXT: /* Record or next record */
4709
        pack_key(&lookup_key, active_index, key_buff3, key, key_len, COL_NEG_INF);
4710 4711
        error = cursor->c_getf_set_range(cursor, flags,
                &lookup_key, SMART_DBT_CALLBACK, &info);
4712
        break;
4713 4714 4715
    //
    // This case does not seem to ever be used, it is ok for it to be slow
    //
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4716
    case HA_READ_KEY_OR_PREV: /* Record or previous */
4717
        pack_key(&lookup_key, active_index, key_buff3, key, key_len, COL_NEG_INF);
4718
        ir_info.orig_key = &lookup_key;
4719 4720
        error = cursor->c_getf_set_range(cursor, flags,
                &lookup_key, SMART_DBT_IR_CALLBACK, &ir_info);
4721
        if (error == DB_NOTFOUND) {
4722 4723
            error = cursor->c_getf_last(cursor, flags, SMART_DBT_CALLBACK, &info);
        }
4724 4725 4726
        else if (ir_info.cmp) {
            error = cursor->c_getf_prev(cursor, flags, SMART_DBT_CALLBACK, &info);
        }
4727
        break;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4728
    case HA_READ_PREFIX_LAST_OR_PREV: /* Last or prev key with the same prefix */
4729
        pack_key(&lookup_key, active_index, key_buff3, key, key_len, COL_POS_INF);
4730 4731
        error = cursor->c_getf_set_range_reverse(cursor, flags, 
                    &lookup_key, SMART_DBT_CALLBACK, &info);
4732
        break;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4733
    case HA_READ_PREFIX_LAST:
4734
        pack_key(&lookup_key, active_index, key_buff3, key, key_len, COL_POS_INF);
4735
        ir_info.orig_key = &lookup_key;
4736
        error = cursor->c_getf_set_range_reverse(cursor, flags, &lookup_key, SMART_DBT_IR_CALLBACK, &ir_info);
4737 4738
        if (ir_info.cmp) {
            error = DB_NOTFOUND;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4739 4740
        }
        break;
4741
    default:
4742
        TOKUDB_TRACE("unsupported:%d\n", find_flag);
4743
        error = HA_ERR_UNSUPPORTED;
4744
        break;
4745
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4746
    error = handle_cursor_error(error,HA_ERR_KEY_NOT_FOUND,active_index);
4747
    if (!error && !key_read && active_index != primary_key && !(table->key_info[active_index].flags & HA_CLUSTERING)) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4748
        error = read_full_row(buf);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4749
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4750 4751 4752 4753
    
    if (error && (tokudb_debug & TOKUDB_DEBUG_ERROR)) {
        TOKUDB_TRACE("error:%d:%d\n", error, find_flag);
    }
4754 4755 4756
    trx->stmt_progress.queried++;
    track_progress(thd);

Zardosht Kasheff's avatar
Zardosht Kasheff committed
4757
cleanup:
4758
    TOKUDB_DBUG_RETURN(error);
4759 4760
}

4761

4762 4763 4764 4765 4766 4767 4768 4769 4770 4771 4772 4773 4774 4775 4776 4777 4778 4779 4780
int ha_tokudb::read_data_from_range_query_buff(uchar* buf, bool need_val) {
    // buffer has the next row, get it from there
    int error;
    uchar* curr_pos = range_query_buff+curr_range_query_buff_offset;
    DBT curr_key;
    bzero((void *) &curr_key, sizeof(curr_key));
    
    // get key info
    u_int32_t key_size = *(u_int32_t *)curr_pos;
    curr_pos += sizeof(key_size);
    uchar* curr_key_buff = curr_pos;
    curr_pos += key_size;
    
    curr_key.data = curr_key_buff;
    curr_key.size = key_size;
    
    // if this is a covering index, this is all we need
    if (this->key_read) {
        assert(!need_val);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
4781
        extract_hidden_primary_key(active_index, &curr_key);
4782 4783 4784 4785 4786 4787 4788 4789 4790 4791 4792 4793 4794 4795 4796 4797 4798
        read_key_only(buf, active_index, &curr_key);
        error = 0;
    }
    // we need to get more data
    else {
        DBT curr_val;
        bzero((void *) &curr_val, sizeof(curr_val));
        uchar* curr_val_buff = NULL;
        u_int32_t val_size = 0;
        // in this case, we don't have a val, we are simply extracting the pk
        if (!need_val) {
            curr_val.data = curr_val_buff;
            curr_val.size = val_size;
            extract_hidden_primary_key(active_index, &curr_key);
            error = read_primary_key( buf, active_index, &curr_val, &curr_key);
        }
        else {
4799
            extract_hidden_primary_key(active_index, &curr_key);
4800 4801 4802 4803 4804 4805 4806 4807 4808 4809 4810 4811 4812 4813 4814 4815 4816 4817 4818 4819 4820 4821 4822 4823 4824 4825 4826 4827 4828 4829 4830 4831 4832 4833 4834 4835 4836 4837 4838 4839 4840 4841 4842 4843 4844 4845 4846 4847 4848 4849 4850 4851 4852 4853 4854 4855 4856 4857 4858 4859 4860 4861 4862 4863 4864 4865 4866 4867 4868 4869 4870
            // need to extract a val and place it into buf
            if (unpack_entire_row) {
                // get val info
                val_size = *(u_int32_t *)curr_pos;
                curr_pos += sizeof(val_size);
                curr_val_buff = curr_pos;
                curr_pos += val_size;
                curr_val.data = curr_val_buff;
                curr_val.size = val_size;
                error = unpack_row(buf,&curr_val, &curr_key, active_index);
            }
            else {
                if (!(hidden_primary_key && active_index == primary_key)) {
                    unpack_key(buf,&curr_key,active_index);
                }
                // read rows we care about

                // first the null bytes;
                memcpy(buf, curr_pos, table_share->null_bytes);
                curr_pos += table_share->null_bytes;

                // now the fixed sized rows                
                for (u_int32_t i = 0; i < num_fixed_cols_for_query; i++) {
                    uint field_index = fixed_cols_for_query[i];
                    Field* field = table->field[field_index];
                    unpack_fixed_field(
                        buf + field_offset(field, table),
                        curr_pos,
                        share->kc_info.field_lengths[field_index]
                        );
                    curr_pos += share->kc_info.field_lengths[field_index];
                }
                // now the variable sized rows
                for (u_int32_t i = 0; i < num_var_cols_for_query; i++) {
                    uint field_index = var_cols_for_query[i];
                    Field* field = table->field[field_index];
                    u_int32_t field_len = *(u_int32_t *)curr_pos;
                    curr_pos += sizeof(field_len);
                    unpack_var_field(
                        buf + field_offset(field, table),
                        curr_pos,
                        field_len,
                        share->kc_info.length_bytes[field_index]
                        );
                    curr_pos += field_len;
                }
                // now the blobs
                if (read_blobs) {
                    u_int32_t blob_size = *(u_int32_t *)curr_pos;
                    curr_pos += sizeof(blob_size);
                    error = unpack_blobs(
                        buf,
                        curr_pos,
                        blob_size,
                        true
                        );
                    curr_pos += blob_size;
                    if (error) {
                        invalidate_bulk_fetch();
                        goto exit;
                    }
                }
                error = 0;
            }
        }
    }
    
    curr_range_query_buff_offset = curr_pos - range_query_buff;
exit:
    return error;
}
4871

4872 4873 4874 4875 4876 4877 4878 4879 4880 4881 4882 4883 4884 4885 4886 4887 4888 4889 4890 4891 4892 4893 4894 4895 4896 4897 4898 4899 4900 4901 4902 4903 4904 4905 4906 4907 4908 4909 4910 4911 4912 4913 4914 4915 4916
static int
smart_dbt_bf_callback(DBT const *key, DBT  const *row, void *context) {
    SMART_DBT_BF_INFO info = (SMART_DBT_BF_INFO)context;
    return info->ha->fill_range_query_buf(info->need_val, key, row, info->direction, info->thd);
}

// fill in the range query buf for bulk fetch
int ha_tokudb::fill_range_query_buf(
    bool need_val, 
    DBT const *key, 
    DBT  const *row, 
    int direction,
    THD* thd
    ) {
    int error;
    //
    // first put the value into range_query_buf
    //
    u_int32_t size_remaining = size_range_query_buff - bytes_used_in_range_query_buff;
    u_int32_t size_needed;
    u_int32_t user_defined_size = get_tokudb_read_buf_size(thd);
    uchar* curr_pos = NULL;
    if (need_val) {
        if (unpack_entire_row) {
            size_needed = 2*sizeof(u_int32_t) + key->size + row->size;
        }
        else {
            // this is an upper bound
            size_needed = sizeof(u_int32_t) + // size of key length
                          key->size + row->size + //key and row
                          num_var_cols_for_query*(sizeof(u_int32_t)) + //lengths of varchars stored
                          sizeof(u_int32_t); //length of blobs
        }
    }
    else {
        size_needed = sizeof(u_int32_t) + key->size;
    }
    if (size_remaining < size_needed) {
        range_query_buff = (uchar *)my_realloc(
            (void *)range_query_buff, 
            bytes_used_in_range_query_buff+size_needed, 
            MYF(MY_WME)
            );
        if (range_query_buff == NULL) {
            error = ENOMEM;
4917
            invalidate_bulk_fetch();
4918 4919
            goto cleanup;
        }
4920
        size_range_query_buff = bytes_used_in_range_query_buff+size_needed;
4921 4922 4923 4924 4925 4926 4927 4928 4929 4930 4931 4932 4933 4934 4935 4936 4937 4938 4939 4940 4941 4942 4943 4944 4945 4946 4947 4948 4949 4950 4951 4952 4953 4954 4955 4956 4957 4958 4959 4960 4961 4962 4963 4964 4965 4966 4967 4968 4969 4970 4971 4972 4973 4974 4975 4976 4977 4978 4979 4980 4981 4982 4983 4984 4985 4986 4987 4988 4989 4990 4991 4992 4993 4994 4995 4996 4997 4998 4999
    }
    //
    // now we know we have the size, let's fill the buffer, starting with the key
    //
    curr_pos = range_query_buff + bytes_used_in_range_query_buff;

    *(u_int32_t *)curr_pos = key->size;
    curr_pos += sizeof(u_int32_t);
    memcpy(curr_pos, key->data, key->size);
    curr_pos += key->size;
    if (need_val) {
        if (unpack_entire_row) {
            *(u_int32_t *)curr_pos = row->size;
            curr_pos += sizeof(u_int32_t);
            memcpy(curr_pos, row->data, row->size);
            curr_pos += row->size;
        }
        else {
            // need to unpack just the data we care about
            const uchar* fixed_field_ptr = (const uchar *) row->data;
            fixed_field_ptr += table_share->null_bytes;

            const uchar* var_field_offset_ptr = NULL;
            const uchar* var_field_data_ptr = NULL;
            
            var_field_offset_ptr = fixed_field_ptr + share->kc_info.mcp_info[active_index].fixed_field_size;
            var_field_data_ptr = var_field_offset_ptr + share->kc_info.mcp_info[active_index].len_of_offsets;

            // first the null bytes
            memcpy(curr_pos, row->data, table_share->null_bytes);
            curr_pos += table_share->null_bytes;
            // now the fixed fields
            //
            // first the fixed fields
            //
            for (u_int32_t i = 0; i < num_fixed_cols_for_query; i++) {
                uint field_index = fixed_cols_for_query[i];
                memcpy(
                    curr_pos, 
                    fixed_field_ptr + share->kc_info.cp_info[active_index][field_index].col_pack_val,
                    share->kc_info.field_lengths[field_index]
                    );
                curr_pos += share->kc_info.field_lengths[field_index];
            }
            
            //
            // now the var fields
            //
            for (u_int32_t i = 0; i < num_var_cols_for_query; i++) {
                uint field_index = var_cols_for_query[i];
                u_int32_t var_field_index = share->kc_info.cp_info[active_index][field_index].col_pack_val;
                u_int32_t data_start_offset;
                u_int32_t field_len;
                
                get_var_field_info(
                    &field_len, 
                    &data_start_offset, 
                    var_field_index, 
                    var_field_offset_ptr, 
                    share->kc_info.num_offset_bytes
                    );
                memcpy(curr_pos, &field_len, sizeof(field_len));
                curr_pos += sizeof(field_len);
                memcpy(curr_pos, var_field_data_ptr + data_start_offset, field_len);
                curr_pos += field_len;
            }
            
            if (read_blobs) {
                u_int32_t blob_offset = 0;
                u_int32_t data_size = 0;
                //
                // now the blobs
                //
                get_blob_field_info(
                    &blob_offset, 
                    share->kc_info.mcp_info[active_index].len_of_offsets,
                    var_field_data_ptr, 
                    share->kc_info.num_offset_bytes
                    );
5000
                data_size = row->size - blob_offset - (u_int32_t)(var_field_data_ptr - (const uchar *)row->data);
5001 5002 5003 5004 5005 5006 5007 5008 5009 5010 5011 5012 5013 5014 5015 5016
                memcpy(curr_pos, &data_size, sizeof(data_size));
                curr_pos += sizeof(data_size);
                memcpy(curr_pos, var_field_data_ptr + blob_offset, data_size);
                curr_pos += data_size;
            }
        }
    }

    bytes_used_in_range_query_buff = curr_pos - range_query_buff;
    assert(bytes_used_in_range_query_buff <= size_range_query_buff);

    //
    // now determine if we should continue with the bulk fetch
    // we want to stop under these conditions:
    //  - we overran the prelocked range
    //  - we are close to the end of the buffer
5017 5018 5019 5020 5021 5022 5023 5024 5025 5026 5027 5028 5029 5030 5031 5032
    //  - we have fetched an exponential amount of rows with
    //  respect to the bulk fetch iteration, which is initialized 
    //  to 0 in index_init() and prelock_range().

    rows_fetched_using_bulk_fetch++;
    // if the iteration is less than the number of possible shifts on
    // a 64 bit integer, check that we haven't exceeded this iterations
    // row fetch upper bound.
    if (bulk_fetch_iteration < HA_TOKU_BULK_FETCH_ITERATION_MAX) {
        uint64_t row_fetch_upper_bound = 1LLU << bulk_fetch_iteration;
        assert(row_fetch_upper_bound > 0);
        if (rows_fetched_using_bulk_fetch >= row_fetch_upper_bound) { 
            error = 0;
            goto cleanup;
        }
    }
5033 5034 5035 5036 5037 5038 5039 5040 5041 5042 5043 5044 5045 5046 5047 5048 5049 5050 5051 5052 5053 5054 5055 5056 5057 5058 5059 5060 5061 5062 5063 5064 5065 5066 5067 5068 5069 5070 5071 5072 5073 5074 5075 5076 5077 5078

    if (bytes_used_in_range_query_buff + table_share->rec_buff_length > user_defined_size) {
        error = 0;
        goto cleanup;
    }
    if (direction > 0) {
        // compare what we got to the right endpoint of prelocked range
        // because we are searching keys in ascending order
        if (prelocked_right_range_size == 0) {
            error = TOKUDB_CURSOR_CONTINUE;
            goto cleanup;
        }
        DBT right_range;
        bzero(&right_range, sizeof(right_range));
        right_range.size = prelocked_right_range_size;
        right_range.data = prelocked_right_range;
        int cmp = tokudb_cmp_dbt_key(
            share->key_file[active_index], 
            key, 
            &right_range
            );
        error = (cmp > 0) ? 0 : TOKUDB_CURSOR_CONTINUE;
    }
    else {
        // compare what we got to the left endpoint of prelocked range
        // because we are searching keys in descending order
        if (prelocked_left_range_size == 0) {
            error = TOKUDB_CURSOR_CONTINUE;
            goto cleanup;
        }
        DBT left_range;
        bzero(&left_range, sizeof(left_range));
        left_range.size = prelocked_left_range_size;
        left_range.data = prelocked_left_range;
        int cmp = tokudb_cmp_dbt_key(
            share->key_file[active_index], 
            key, 
            &left_range
            );
        error = (cmp < 0) ? 0 : TOKUDB_CURSOR_CONTINUE;
    }
cleanup:
    return error;
}

int ha_tokudb::get_next(uchar* buf, int direction) {
5079
    int error = 0; 
5080
    u_int32_t flags = SET_PRELOCK_FLAG(0);
5081 5082
    THD* thd = ha_thd();
    tokudb_trx_data* trx = (tokudb_trx_data *) thd_data_get(thd, tokudb_hton->slot);;
5083
    bool need_val;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5084 5085
    HANDLE_INVALID_CURSOR();

5086 5087 5088 5089 5090 5091 5092
    // we need to read the val of what we retrieve if
    // we do NOT have a covering index AND we are using a clustering secondary
    // key
    need_val = (this->key_read == 0) && 
                (active_index == primary_key || 
                 table->key_info[active_index].flags & HA_CLUSTERING
                       );
5093

5094 5095
    if ((bytes_used_in_range_query_buff - curr_range_query_buff_offset) > 0) {
        error = read_data_from_range_query_buff(buf, need_val);
5096
    }
5097 5098 5099 5100 5101 5102 5103 5104 5105 5106 5107 5108
    else {
        invalidate_bulk_fetch();
        if (doing_bulk_fetch) {
            struct smart_dbt_bf_info bf_info;
            bf_info.ha = this;
            // you need the val if you have a clustering index and key_read is not 0;
            bf_info.direction = direction;
            bf_info.thd = ha_thd();
            bf_info.need_val = need_val;
            //
            // call c_getf_next with purpose of filling in range_query_buff
            //
5109
            rows_fetched_using_bulk_fetch = 0;
5110
            if (direction > 0) {
5111 5112
                error = cursor->c_getf_next(cursor, flags,
                        smart_dbt_bf_callback, &bf_info);
5113 5114
            }
            else {
5115 5116
                error = cursor->c_getf_prev(cursor, flags,
                        smart_dbt_bf_callback, &bf_info);
5117
            }
5118 5119 5120
            if (bulk_fetch_iteration < HA_TOKU_BULK_FETCH_ITERATION_MAX) {
                bulk_fetch_iteration++;
            }
5121 5122 5123 5124 5125 5126 5127 5128 5129 5130 5131 5132 5133 5134 5135

            error = handle_cursor_error(error, HA_ERR_END_OF_FILE,active_index);
            if (error) { goto cleanup; }
            
            //
            // now that range_query_buff is filled, read an element
            //
            error = read_data_from_range_query_buff(buf, need_val);
        }
        else {
            struct smart_dbt_info info;
            info.ha = this;
            info.buf = buf;
            info.keynr = active_index;
            
5136 5137 5138 5139
            error = cursor->c_getf_next(cursor, flags,
                    SMART_DBT_CALLBACK, &info);
            error = handle_cursor_error(error, HA_ERR_END_OF_FILE, 
                    active_index);
5140 5141 5142
        }
    }

Zardosht Kasheff's avatar
Zardosht Kasheff committed
5143
    //
5144 5145 5146 5147 5148 5149
    // at this point, one of two things has happened
    // either we have unpacked the data into buf, and we 
    // are done, or we have unpacked the primary key
    // into last_key, and we use the code below to
    // read the full row by doing a point query into the 
    // main table.
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5150
    //
5151
    
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5152
    if (!error && !key_read && (active_index != primary_key) && !(table->key_info[active_index].flags & HA_CLUSTERING) ) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5153 5154
        error = read_full_row(buf);
    }
5155 5156
    trx->stmt_progress.queried++;
    track_progress(thd);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5157
cleanup:
5158 5159 5160 5161 5162 5163 5164 5165 5166 5167 5168 5169 5170 5171 5172 5173 5174
    return error;
}


//
// Reads the next row from the active index (cursor) into buf, and advances cursor
// Parameters:
//      [out]   buf - buffer for the next row, in MySQL format
// Returns:
//      0 on success
//      HA_ERR_END_OF_FILE if not found
//      error otherwise
//
int ha_tokudb::index_next(uchar * buf) {
    TOKUDB_DBUG_ENTER("ha_tokudb::index_next");
    statistic_increment(table->in_use->status_var.ha_read_next_count, &LOCK_status);
    int error = get_next(buf, 1);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5175
    TOKUDB_DBUG_RETURN(error);
5176 5177
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
5178 5179 5180 5181 5182 5183

int ha_tokudb::index_read_last(uchar * buf, const uchar * key, uint key_len) {
    return(index_read(buf, key, key_len, HA_READ_PREFIX_LAST));    
}


Zardosht Kasheff's avatar
Zardosht Kasheff committed
5184 5185 5186 5187 5188 5189 5190 5191 5192
//
// Reads the previous row from the active index (cursor) into buf, and advances cursor
// Parameters:
//      [out]   buf - buffer for the next row, in MySQL format
// Returns:
//      0 on success
//      HA_ERR_END_OF_FILE if not found
//      error otherwise
//
5193
int ha_tokudb::index_prev(uchar * buf) {
5194
    TOKUDB_DBUG_ENTER("ha_tokudb::index_prev");
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5195
    statistic_increment(table->in_use->status_var.ha_read_next_count, &LOCK_status);
5196
    int error = get_next(buf, -1);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5197
    TOKUDB_DBUG_RETURN(error);
5198 5199
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
5200 5201 5202 5203 5204 5205 5206 5207 5208
//
// Reads the first row from the active index (cursor) into buf, and advances cursor
// Parameters:
//      [out]   buf - buffer for the next row, in MySQL format
// Returns:
//      0 on success
//      HA_ERR_END_OF_FILE if not found
//      error otherwise
//
5209
int ha_tokudb::index_first(uchar * buf) {
5210
    TOKUDB_DBUG_ENTER("ha_tokudb::index_first");
5211
    invalidate_bulk_fetch();
5212
    int error = 0;
5213
    struct smart_dbt_info info;
5214
    u_int32_t flags = SET_PRELOCK_FLAG(0);
5215 5216
    THD* thd = ha_thd();
    tokudb_trx_data* trx = (tokudb_trx_data *) thd_data_get(thd, tokudb_hton->slot);;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5217
    HANDLE_INVALID_CURSOR();
5218

5219
    statistic_increment(table->in_use->status_var.ha_read_first_count, &LOCK_status);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5220

5221 5222 5223 5224
    info.ha = this;
    info.buf = buf;
    info.keynr = active_index;

5225 5226
    error = cursor->c_getf_first(cursor, flags,
            SMART_DBT_CALLBACK, &info);
5227 5228
    error = handle_cursor_error(error,HA_ERR_END_OF_FILE,active_index);

5229 5230 5231 5232 5233 5234
    //
    // still need to get entire contents of the row if operation done on
    // secondary DB and it was NOT a covering index
    //
    if (!error && !key_read && (active_index != primary_key) && !(table->key_info[active_index].flags & HA_CLUSTERING) ) {
        error = read_full_row(buf);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5235
    }
5236 5237 5238
    trx->stmt_progress.queried++;
    track_progress(thd);
    
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5239
cleanup:
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5240
    TOKUDB_DBUG_RETURN(error);
5241 5242
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
5243 5244 5245 5246 5247 5248 5249 5250 5251
//
// Reads the last row from the active index (cursor) into buf, and advances cursor
// Parameters:
//      [out]   buf - buffer for the next row, in MySQL format
// Returns:
//      0 on success
//      HA_ERR_END_OF_FILE if not found
//      error otherwise
//
5252
int ha_tokudb::index_last(uchar * buf) {
5253
    TOKUDB_DBUG_ENTER("ha_tokudb::index_last");
5254
    invalidate_bulk_fetch();
5255
    int error = 0;
5256
    struct smart_dbt_info info;
5257
    u_int32_t flags = SET_PRELOCK_FLAG(0);
5258 5259
    THD* thd = ha_thd();
    tokudb_trx_data* trx = (tokudb_trx_data *) thd_data_get(thd, tokudb_hton->slot);;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5260
    HANDLE_INVALID_CURSOR();
5261

5262
    statistic_increment(table->in_use->status_var.ha_read_last_count, &LOCK_status);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5263

5264 5265 5266 5267
    info.ha = this;
    info.buf = buf;
    info.keynr = active_index;

5268 5269
    error = cursor->c_getf_last(cursor, flags,
            SMART_DBT_CALLBACK, &info);
5270
    error = handle_cursor_error(error,HA_ERR_END_OF_FILE,active_index);
5271 5272 5273 5274 5275 5276
    //
    // still need to get entire contents of the row if operation done on
    // secondary DB and it was NOT a covering index
    //
    if (!error && !key_read && (active_index != primary_key) && !(table->key_info[active_index].flags & HA_CLUSTERING) ) {
        error = read_full_row(buf);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5277
    }
5278 5279 5280 5281 5282

    if (trx) {
        trx->stmt_progress.queried++;
    }
    track_progress(thd);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5283
cleanup:
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5284
    TOKUDB_DBUG_RETURN(error);
5285 5286
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
5287 5288 5289 5290 5291 5292 5293 5294
//
// Initialize a scan of the table (which is why index_init is called on primary_key)
// Parameters:
//          scan - unused
// Returns:
//      0 on success
//      error otherwise
//
5295
int ha_tokudb::rnd_init(bool scan) {
5296
    TOKUDB_DBUG_ENTER("ha_tokudb::rnd_init");
5297
    int error = 0;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5298
    range_lock_grabbed = false;
5299 5300 5301
    error = index_init(primary_key, 0);
    if (error) { goto cleanup;}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
5302
    if (scan) {
5303
        error = prelock_range(NULL, NULL);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5304
        if (error) { goto cleanup; }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5305
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5306
    //
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5307
    // only want to set range_lock_grabbed to true after index_init
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5308 5309 5310 5311
    // successfully executed for two reasons:
    // 1) index_init will reset it to false anyway
    // 2) if it fails, we don't want prelocking on,
    //
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5312
    if (scan) { range_lock_grabbed = true; }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5313
    error = 0;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5314
cleanup:
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5315 5316 5317 5318
    if (error) { 
        index_end();
        last_cursor_error = error; 
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5319
    TOKUDB_DBUG_RETURN(error);
5320 5321
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
5322 5323 5324
//
// End a scan of the table
//
5325
int ha_tokudb::rnd_end() {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5326
    TOKUDB_DBUG_ENTER("ha_tokudb::rnd_end");
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5327
    range_lock_grabbed = false;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5328
    TOKUDB_DBUG_RETURN(index_end());
5329 5330
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
5331

Zardosht Kasheff's avatar
Zardosht Kasheff committed
5332 5333 5334 5335 5336 5337 5338 5339 5340
//
// Read the next row in a table scan
// Parameters:
//      [out]   buf - buffer for the next row, in MySQL format
// Returns:
//      0 on success
//      HA_ERR_END_OF_FILE if not found
//      error otherwise
//
5341
int ha_tokudb::rnd_next(uchar * buf) {
5342
    TOKUDB_DBUG_ENTER("ha_tokudb::ha_tokudb::rnd_next");
5343
    statistic_increment(table->in_use->status_var.ha_read_rnd_next_count, &LOCK_status);
5344
    int error = get_next(buf, 1);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5345
    TOKUDB_DBUG_RETURN(error);
5346 5347 5348
}


5349 5350 5351
void ha_tokudb::track_progress(THD* thd) {
    tokudb_trx_data* trx = (tokudb_trx_data *) thd_data_get(thd, tokudb_hton->slot);
    if (trx) {
5352
        ulonglong num_written = trx->stmt_progress.inserted + trx->stmt_progress.updated + trx->stmt_progress.deleted;
5353
        bool update_status = 
5354
            (trx->stmt_progress.queried && tokudb_read_status_frequency && (trx->stmt_progress.queried % tokudb_read_status_frequency) == 0) ||
5355
            (num_written && tokudb_write_status_frequency && (num_written % tokudb_write_status_frequency) == 0);
5356
        if (update_status) {
5357 5358 5359 5360
            char *next_status = write_status_msg;
            bool first = true;
            int r;
            if (trx->stmt_progress.queried) {
5361 5362
                r = sprintf(next_status, "Queried about %llu row%s", trx->stmt_progress.queried, trx->stmt_progress.queried == 1 ? "" : "s"); 
                assert(r >= 0);
5363 5364 5365 5366
                next_status += r;
                first = false;
            }
            if (trx->stmt_progress.inserted) {
5367 5368 5369 5370 5371 5372
                if (trx->stmt_progress.using_loader) {
                    r = sprintf(next_status, "%sFetched about %llu row%s, loading data still remains", first ? "" : ", ", trx->stmt_progress.inserted, trx->stmt_progress.inserted == 1 ? "" : "s"); 
                }
                else {
                    r = sprintf(next_status, "%sInserted about %llu row%s", first ? "" : ", ", trx->stmt_progress.inserted, trx->stmt_progress.inserted == 1 ? "" : "s"); 
                }
5373
                assert(r >= 0);
5374 5375 5376 5377
                next_status += r;
                first = false;
            }
            if (trx->stmt_progress.updated) {
5378 5379
                r = sprintf(next_status, "%sUpdated about %llu row%s", first ? "" : ", ", trx->stmt_progress.updated, trx->stmt_progress.updated == 1 ? "" : "s"); 
                assert(r >= 0);
5380 5381 5382 5383
                next_status += r;
                first = false;
            }
            if (trx->stmt_progress.deleted) {
5384 5385
                r = sprintf(next_status, "%sDeleted about %llu row%s", first ? "" : ", ", trx->stmt_progress.deleted, trx->stmt_progress.deleted == 1 ? "" : "s"); 
                assert(r >= 0);
5386 5387 5388
                next_status += r;
                first = false;
            }
5389 5390
            if (!first)
	        thd_proc_info(thd, write_status_msg);
5391 5392 5393 5394 5395
        }
    }
}


5396
DBT *ha_tokudb::get_pos(DBT * to, uchar * pos) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5397
    TOKUDB_DBUG_ENTER("ha_tokudb::get_pos");
5398 5399
    /* We don't need to set app_data here */
    bzero((void *) to, sizeof(*to));
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5400 5401
    to->data = pos + sizeof(u_int32_t);
    to->size = *(u_int32_t *)pos;
5402
    DBUG_DUMP("key", (const uchar *) to->data, to->size);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5403
    DBUG_RETURN(to);
5404 5405
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
5406
//
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5407
// Retrieves a row with based on the primary key saved in pos
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5408 5409 5410 5411 5412
// Returns:
//      0 on success
//      HA_ERR_KEY_NOT_FOUND if not found
//      error otherwise
//
5413
int ha_tokudb::rnd_pos(uchar * buf, uchar * pos) {
5414
    TOKUDB_DBUG_ENTER("ha_tokudb::rnd_pos");
5415
    DBT db_pos;
5416
    int error = 0;
5417 5418 5419 5420 5421
    struct smart_dbt_info info;
    bool old_unpack_entire_row = unpack_entire_row;
    DBT* key = get_pos(&db_pos, pos); 

    unpack_entire_row = true;
5422 5423
    statistic_increment(table->in_use->status_var.ha_read_rnd_count, &LOCK_status);
    active_index = MAX_KEY;
5424 5425 5426 5427 5428

    info.ha = this;
    info.buf = buf;
    info.keynr = primary_key;

5429 5430 5431
    error = share->file->getf_set(share->file, transaction, 
            get_cursor_isolation_flags(lock.type, ha_thd()), 
            key, smart_dbt_callback_rowread_ptquery, &info);
5432

Barry Perlman's avatar
Barry Perlman committed
5433
    if (error == DB_NOTFOUND) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5434 5435 5436 5437
        error = HA_ERR_KEY_NOT_FOUND;
        goto cleanup;
    }
cleanup:
5438
    unpack_entire_row = old_unpack_entire_row;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5439
    TOKUDB_DBUG_RETURN(error);
5440 5441
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
5442
int ha_tokudb::prelock_range( const key_range *start_key, const key_range *end_key) {
5443
    TOKUDB_DBUG_ENTER("ha_tokudb::prelock_range");
5444
    THD* thd = ha_thd(); 
5445 5446

    int error = 0;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5447 5448
    DBT start_dbt_key;
    DBT end_dbt_key;
5449 5450
    uchar* start_key_buff  = prelocked_left_range;
    uchar* end_key_buff = prelocked_right_range;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5451

Zardosht Kasheff's avatar
Zardosht Kasheff committed
5452 5453 5454
    bzero((void *) &start_dbt_key, sizeof(start_dbt_key));
    bzero((void *) &end_dbt_key, sizeof(end_dbt_key));

5455
    HANDLE_INVALID_CURSOR();
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5456 5457 5458
    if (start_key) {
        switch (start_key->flag) {
        case HA_READ_AFTER_KEY:
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5459
            pack_key(&start_dbt_key, active_index, start_key_buff, start_key->key, start_key->length, COL_POS_INF);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5460 5461
            break;
        default:
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5462
            pack_key(&start_dbt_key, active_index, start_key_buff, start_key->key, start_key->length, COL_NEG_INF);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5463 5464
            break;
        }
5465 5466 5467 5468
        prelocked_left_range_size = start_dbt_key.size;
    }
    else {
        prelocked_left_range_size = 0;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5469
    }
5470

Zardosht Kasheff's avatar
Zardosht Kasheff committed
5471 5472 5473
    if (end_key) {
        switch (end_key->flag) {
        case HA_READ_BEFORE_KEY:
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5474
            pack_key(&end_dbt_key, active_index, end_key_buff, end_key->key, end_key->length, COL_NEG_INF);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5475 5476
            break;
        default:
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5477
            pack_key(&end_dbt_key, active_index, end_key_buff, end_key->key, end_key->length, COL_POS_INF);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5478
            break;
5479
        }        
5480 5481 5482 5483
        prelocked_right_range_size = end_dbt_key.size;
    }
    else {
        prelocked_right_range_size = 0;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5484 5485
    }

5486 5487 5488 5489 5490
    error = cursor->c_pre_acquire_range_lock(
        cursor, 
        start_key ? &start_dbt_key : share->key_file[active_index]->dbt_neg_infty(), 
        end_key ? &end_dbt_key : share->key_file[active_index]->dbt_pos_infty()
        );
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5491 5492 5493 5494 5495 5496
    if (error){ 
        last_cursor_error = error;
        //
        // cursor should be initialized here, but in case it is not, we still check
        //
        if (cursor) {
5497 5498
            int r = cursor->c_close(cursor);
            assert(r==0);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5499 5500 5501 5502
            cursor = NULL;
        }
        goto cleanup; 
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5503

5504 5505 5506 5507 5508
    //
    // at this point, determine if we will be doing bulk fetch
    // as of now, only do it if we are doing a select
    //
    doing_bulk_fetch = (thd_sql_command(thd) == SQLCOM_SELECT);
5509 5510
    bulk_fetch_iteration = 0;
    rows_fetched_using_bulk_fetch = 0;
5511

Zardosht Kasheff's avatar
Zardosht Kasheff committed
5512 5513
cleanup:
    TOKUDB_DBUG_RETURN(error);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5514
}
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5515

5516 5517 5518 5519 5520 5521
//
// Prelock range if possible, start_key is leftmost, end_key is rightmost
// whether scanning forward or backward.  This function is called by MySQL
// for backward range queries (in QUICK_SELECT_DESC::get_next). 
// Forward scans use read_range_first()/read_range_next().
//
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5522
int ha_tokudb::prepare_range_scan( const key_range *start_key, const key_range *end_key) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5523 5524 5525 5526 5527
    int error = prelock_range(start_key, end_key);
    if (!error) {
        range_lock_grabbed = true;
    }
    return error;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5528 5529 5530 5531 5532 5533 5534 5535
}

int ha_tokudb::read_range_first(
    const key_range *start_key,
    const key_range *end_key,
    bool eq_range, 
    bool sorted) 
{
5536
    int error = prelock_range(start_key, end_key);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5537 5538 5539 5540 5541 5542 5543
    if (error) { goto cleanup; }
    range_lock_grabbed = true;
    
    error = handler::read_range_first(start_key, end_key, eq_range, sorted);
cleanup:
    return error;
}
5544

Zardosht Kasheff's avatar
Zardosht Kasheff committed
5545 5546
int ha_tokudb::read_range_next()
{
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5547 5548 5549 5550
    TOKUDB_DBUG_ENTER("ha_tokudb::read_range_next");
    int error;
    error = handler::read_range_next();
    if (error) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5551
        range_lock_grabbed = false;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5552 5553
    }
    TOKUDB_DBUG_RETURN(error);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5554 5555 5556 5557
}



5558 5559 5560 5561 5562 5563 5564 5565 5566 5567 5568 5569 5570 5571 5572 5573 5574 5575 5576 5577 5578
/*
  Set a reference to the current record in (ref,ref_length).

  SYNOPSIS
  ha_tokudb::position()
  record                      The current record buffer

  DESCRIPTION
  The BDB handler stores the primary key in (ref,ref_length).
  There is either an explicit primary key, or an implicit (hidden)
  primary key.
  During open(), 'ref_length' is calculated as the maximum primary
  key length. When an actual key is shorter than that, the rest of
  the buffer must be cleared out. The row cannot be identified, if
  garbage follows behind the end of the key. There is no length
  field for the current key, so that the whole ref_length is used
  for comparison.

  RETURN
  nothing
*/
5579
void ha_tokudb::position(const uchar * record) {
5580
    TOKUDB_DBUG_ENTER("ha_tokudb::position");
5581 5582
    DBT key;
    if (hidden_primary_key) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5583 5584
        DBUG_ASSERT(ref_length == (TOKUDB_HIDDEN_PRIMARY_KEY_LENGTH + sizeof(u_int32_t)));
        memcpy_fixed(ref + sizeof(u_int32_t), current_ident, TOKUDB_HIDDEN_PRIMARY_KEY_LENGTH);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5585
        *(u_int32_t *)ref = TOKUDB_HIDDEN_PRIMARY_KEY_LENGTH;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5586 5587
    } 
    else {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5588 5589 5590 5591 5592 5593 5594 5595 5596
        bool has_null;
        //
        // save the data
        //
        create_dbt_key_from_table(&key, primary_key, ref + sizeof(u_int32_t), record, &has_null);
        //
        // save the size of data in the first four bytes of ref
        //
        memcpy(ref, &key.size, sizeof(u_int32_t));
5597 5598 5599 5600
    }
    DBUG_VOID_RETURN;
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
5601 5602 5603 5604 5605 5606
//
// Per InnoDB: Returns statistics information of the table to the MySQL interpreter,
// in various fields of the handle object. 
// Return:
//      0, always success
//
5607
int ha_tokudb::info(uint flag) {
5608
    TOKUDB_DBUG_ENTER("ha_tokudb::info %p %d %lld", this, flag, (long long) share->rows);
5609
    int error;
5610
    bool key_file_lock_taken = false;
5611 5612 5613
    DB_TXN* txn = NULL;
    uint curr_num_DBs = table->s->keys + test(hidden_primary_key);
    DB_BTREE_STAT64 dict_stats;
5614 5615
    if (flag & HA_STATUS_VARIABLE) {
        // Just to get optimizations right
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5616
        stats.records = share->rows + share->rows_from_locked_table;
5617 5618 5619
        if (stats.records == 0) {
            stats.records++;
        }
5620
        stats.deleted = 0;
5621
        if (!(flag & HA_STATUS_NO_LOCK)) {
5622
            u_int64_t num_rows = 0;
5623 5624
            TOKU_DB_FRAGMENTATION_S frag_info;
            memset(&frag_info, 0, sizeof frag_info);
5625

5626 5627 5628
            error = db_env->txn_begin(db_env, NULL, &txn, DB_READ_UNCOMMITTED);
            if (error) { goto cleanup; }

5629 5630 5631 5632 5633 5634
            // XXX 4530 lock the key file lock for reading
            share_key_file_rdlock(share);
            key_file_lock_taken = true;
            // we should always have a primary key
            assert(share->file != NULL);

5635 5636 5637 5638
            error = estimate_num_rows(share->file,&num_rows, txn);
            if (error == 0) {
                share->rows = num_rows;
                stats.records = num_rows;
5639 5640 5641
                if (stats.records == 0) {
                    stats.records++;
                }
5642 5643 5644 5645
            }
            else {
                goto cleanup;
            }
5646 5647 5648 5649 5650 5651
            error = share->file->get_fragmentation(
                share->file,
                &frag_info
                );
            if (error) { goto cleanup; }
            stats.delete_length = frag_info.unused_bytes;
5652

5653 5654 5655 5656 5657 5658
            error = share->file->stat64(
                share->file, 
                txn, 
                &dict_stats
                );
            if (error) { goto cleanup; }
5659
            
5660 5661
            stats.create_time = dict_stats.bt_create_time_sec;
            stats.update_time = dict_stats.bt_modify_time_sec;
5662
            stats.check_time = dict_stats.bt_verify_time_sec;
5663
            stats.data_file_length = dict_stats.bt_dsize;
5664 5665 5666 5667 5668 5669 5670 5671 5672 5673 5674 5675 5676 5677 5678 5679
            if (hidden_primary_key) {
                //
                // in this case, we have a hidden primary key, do not
                // want to report space taken up by the hidden primary key to the user
                //
                u_int64_t hpk_space = TOKUDB_HIDDEN_PRIMARY_KEY_LENGTH*dict_stats.bt_ndata;
                stats.data_file_length = (hpk_space > stats.data_file_length) ? 0 : stats.data_file_length - hpk_space;
            }
            else {
                //
                // one infinity byte per key needs to be subtracted
                //
                u_int64_t inf_byte_space = dict_stats.bt_ndata;
                stats.data_file_length = (inf_byte_space > stats.data_file_length) ? 0 : stats.data_file_length - inf_byte_space;
            }

5680
            stats.mean_rec_length = stats.records ? (ulong)(stats.data_file_length/stats.records) : 0;
5681
            stats.index_file_length = 0;
5682 5683 5684 5685 5686 5687 5688 5689 5690 5691 5692
            // curr_num_DBs is the number of keys we have, according
            // to the mysql layer. if drop index is running concurrently
            // with info() (it can, because info does not take table locks),
            // then it could be the case that one of the dbs was dropped
            // and set to NULL before mysql was able to set table->s->keys
            // accordingly. 
            //
            // we should just ignore any DB * that is NULL. 
            //
            // this solution is much simpler than trying to maintain an 
            // accurate number of valid keys at the handlerton layer.
5693
            for (uint i = 0; i < curr_num_DBs; i++) {
5694 5695
                // skip the primary key, skip dropped indexes
                if (i == primary_key || share->key_file[i] == NULL) {
5696 5697 5698 5699 5700 5701 5702 5703 5704
                    continue;
                }
                error = share->key_file[i]->stat64(
                    share->key_file[i], 
                    txn, 
                    &dict_stats
                    );
                if (error) { goto cleanup; }
                stats.index_file_length += dict_stats.bt_dsize;
5705 5706 5707 5708 5709 5710 5711

                error = share->file->get_fragmentation(
                    share->file,
                    &frag_info
                    );
                if (error) { goto cleanup; }
                stats.delete_length += frag_info.unused_bytes;
5712 5713
            }
        }
5714
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5715
    if ((flag & HA_STATUS_CONST)) {
5716
        stats.max_data_file_length=  9223372036854775807ULL;
5717
        for (uint i = 0; i < table_share->keys; i++) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5718
            table->key_info[i].rec_per_key[table->key_info[i].key_parts - 1] = 0;
5719 5720 5721
        }
    }
    /* Don't return key if we got an error for the internal primary key */
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5722
    if (flag & HA_STATUS_ERRKEY && last_dup_key < table_share->keys) {
5723
        errkey = last_dup_key;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5724
    }    
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5725
    if (flag & HA_STATUS_AUTO && table->found_next_number_field) {        
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5726 5727
        THD *thd= table->in_use;
        struct system_variables *variables= &thd->variables;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5728 5729
        stats.auto_increment_value = share->last_auto_increment + variables->auto_increment_increment;
    }
5730 5731
    error = 0;
cleanup:
5732 5733 5734 5735
    // XXX 4530 unlock the key file lock if it was taken
    if (key_file_lock_taken) {
        share_key_file_unlock(share);
    }
5736
    if (txn != NULL) {
5737
        commit_txn(txn, DB_TXN_NOSYNC);
5738 5739 5740
        txn = NULL;
    }
    TOKUDB_DBUG_RETURN(error);
5741 5742
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
5743 5744 5745
//
//  Per InnoDB: Tells something additional to the handler about how to do things.
//
5746
int ha_tokudb::extra(enum ha_extra_function operation) {
5747
    TOKUDB_DBUG_ENTER("extra %p %d", this, operation);
5748 5749 5750 5751 5752 5753 5754 5755 5756 5757 5758 5759 5760 5761 5762 5763 5764 5765 5766
    switch (operation) {
    case HA_EXTRA_RESET_STATE:
        reset();
        break;
    case HA_EXTRA_KEYREAD:
        key_read = 1;           // Query satisfied with key
        break;
    case HA_EXTRA_NO_KEYREAD:
        key_read = 0;
        break;
    case HA_EXTRA_IGNORE_DUP_KEY:
        using_ignore = 1;
        break;
    case HA_EXTRA_NO_IGNORE_DUP_KEY:
        using_ignore = 0;
        break;
    default:
        break;
    }
5767
    TOKUDB_DBUG_RETURN(0);
5768 5769 5770
}

int ha_tokudb::reset(void) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5771
    TOKUDB_DBUG_ENTER("ha_tokudb::reset");
5772 5773
    key_read = 0;
    using_ignore = 0;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5774
    TOKUDB_DBUG_RETURN(0);
5775 5776
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
5777 5778 5779 5780 5781 5782 5783 5784 5785 5786 5787 5788 5789

//
// helper function that iterates through all DB's 
// and grabs a lock (either read or write, but not both)
// Parameters:
//      [in]    trans - transaction to be used to pre acquire the lock
//              lt - type of lock to get, either lock_read or lock_write
//  Returns:
//      0 on success
//      error otherwise
//
int ha_tokudb::acquire_table_lock (DB_TXN* trans, TABLE_LOCK_TYPE lt) {
    int error = ENOSYS;
5790 5791
    rw_rdlock(&share->num_DBs_lock);
    uint curr_num_DBs = share->num_DBs;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5792
    if (lt == lock_read) {
5793 5794
        error = 0;
        goto cleanup;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5795 5796
    }
    else if (lt == lock_write) {
5797 5798
        if (tokudb_debug & TOKUDB_DEBUG_LOCK)
            TOKUDB_TRACE("%s\n", __FUNCTION__);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5799 5800 5801
        for (uint i = 0; i < curr_num_DBs; i++) {
            DB* db = share->key_file[i];
            error = db->pre_acquire_table_lock(db, trans);
5802 5803 5804
            if (error == EINVAL) 
                TOKUDB_TRACE("%s %d db=%p trans=%p\n", __FUNCTION__, i, db, trans);
            if (error) break;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5805
        }
5806 5807 5808
        if (tokudb_debug & TOKUDB_DEBUG_LOCK)
            TOKUDB_TRACE("%s error=%d\n", __FUNCTION__, error);
        if (error) goto cleanup;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5809 5810 5811 5812 5813 5814 5815 5816
    }
    else {
        error = ENOSYS;
        goto cleanup;
    }

    error = 0;
cleanup:
5817
    rw_unlock(&share->num_DBs_lock);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5818 5819 5820 5821
    return error;
}


5822 5823 5824 5825
int ha_tokudb::create_txn(THD* thd, tokudb_trx_data* trx) {
    int error;
    ulong tx_isolation = thd_tx_isolation(thd);
    HA_TOKU_ISO_LEVEL toku_iso_level = tx_to_toku_iso(tx_isolation);
5826 5827
    bool is_autocommit = !thd_test_options(
            thd, OPTION_NOT_AUTOCOMMIT | OPTION_BEGIN);
5828 5829

    /* First table lock, start transaction */
5830
    if (thd_test_options(thd, OPTION_NOT_AUTOCOMMIT | OPTION_BEGIN) && 
5831 5832 5833
         !trx->all &&
         (thd_sql_command(thd) != SQLCOM_CREATE_TABLE) &&
         (thd_sql_command(thd) != SQLCOM_DROP_TABLE) &&
5834 5835
         (thd_sql_command(thd) != SQLCOM_DROP_INDEX) &&
         (thd_sql_command(thd) != SQLCOM_CREATE_INDEX) &&
5836 5837
         (thd_sql_command(thd) != SQLCOM_ALTER_TABLE)) {
        /* QQQ We have to start a master transaction */
5838
        // DBUG_PRINT("trans", ("starting transaction all "));
5839 5840 5841 5842 5843
        if ((error = db_env->txn_begin(db_env, NULL, &trx->all, toku_iso_to_txn_flag(toku_iso_level)))) {
            trx->tokudb_lock_count--;      // We didn't get the lock
            goto cleanup;
        }
        if (tokudb_debug & TOKUDB_DEBUG_TXN) {
5844
            TOKUDB_TRACE("just created master:%p\n", trx->all);
5845 5846 5847 5848 5849 5850 5851 5852 5853 5854 5855
        }
        trx->sp_level = trx->all;
        trans_register_ha(thd, TRUE, tokudb_hton);
    }
    DBUG_PRINT("trans", ("starting transaction stmt"));
    if (trx->stmt) { 
        if (tokudb_debug & TOKUDB_DEBUG_TXN) {
            TOKUDB_TRACE("warning:stmt=%p\n", trx->stmt);
        }
    }
    u_int32_t txn_begin_flags;
5856
    if (trx->all == NULL) {
5857
        txn_begin_flags = toku_iso_to_txn_flag(toku_iso_level);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5858
        if (txn_begin_flags == 0 && is_autocommit && thd_sql_command(thd) == SQLCOM_SELECT) {
5859 5860
            txn_begin_flags = DB_TXN_SNAPSHOT;
        }
5861 5862
    }
    else {
5863
        txn_begin_flags = DB_INHERIT_ISOLATION;
5864 5865 5866 5867 5868 5869
    }
    if ((error = db_env->txn_begin(db_env, trx->sp_level, &trx->stmt, txn_begin_flags))) {
        /* We leave the possible master transaction open */
        trx->tokudb_lock_count--;  // We didn't get the lock
        goto cleanup;
    }
5870
    trx->sub_sp_level = trx->stmt;
5871
    if (tokudb_debug & TOKUDB_DEBUG_TXN) {
5872
        TOKUDB_TRACE("just created stmt:%p:%p\n", trx->sp_level, trx->stmt);
5873
    }
5874
    reset_stmt_progress(&trx->stmt_progress);
5875 5876 5877 5878 5879
    trans_register_ha(thd, FALSE, tokudb_hton);
cleanup:
    return error;
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
5880

5881 5882 5883 5884 5885 5886 5887 5888
/*
  As MySQL will execute an external lock for every new table it uses
  we can use this to start the transactions.
  If we are in auto_commit mode we just need to start a transaction
  for the statement to be able to rollback the statement.
  If not, we have to start a master transaction if there doesn't exist
  one from before.
*/
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5889 5890 5891 5892 5893 5894 5895 5896
//
// Parameters:
//      [in]    thd - handle to the user thread
//              lock_type - the type of lock
// Returns:
//      0 on success
//      error otherwise
//
5897
int ha_tokudb::external_lock(THD * thd, int lock_type) {
5898 5899 5900
    TOKUDB_DBUG_ENTER("ha_tokudb::external_lock cmd=%d %d", thd_sql_command(thd), lock_type);
    if (tokudb_debug & TOKUDB_DEBUG_LOCK)
        TOKUDB_TRACE("%s cmd=%d %d\n", __FUNCTION__, thd_sql_command(thd), lock_type);
5901

Zardosht Kasheff's avatar
Zardosht Kasheff committed
5902 5903 5904
    int error = 0;
    tokudb_trx_data *trx = NULL;
    trx = (tokudb_trx_data *) thd_data_get(thd, tokudb_hton->slot);
5905
    if (!trx) {
5906 5907
        error = create_tokudb_trx_data_instance(&trx);
        if (error) { goto cleanup; }
5908
        thd_data_set(thd, tokudb_hton->slot, trx);
5909
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5910 5911
    if (trx->all == NULL) {
        trx->sp_level = NULL;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5912
    }
5913
    if (lock_type != F_UNLCK) {
5914
        is_fast_alter_running = false;
5915
        use_write_locks = false;
5916
        if (lock_type == F_WRLCK)
5917
            use_write_locks = true;
5918 5919
        if (!trx->tokudb_lock_count++) {
            DBUG_ASSERT(trx->stmt == 0);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5920
            transaction = NULL;    // Safety
5921 5922
            error = create_txn(thd, trx);
            if (error) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5923
                goto cleanup;
5924 5925
            }
        }
5926 5927
        assert(thd->in_sub_stmt == 0);
        transaction = trx->sub_sp_level;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5928 5929
    }
    else {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5930 5931 5932 5933 5934 5935 5936 5937 5938 5939 5940
        pthread_mutex_lock(&share->mutex);
        // hate dealing with comparison of signed vs unsigned, so doing this
        if (deleted_rows > added_rows && share->rows < (deleted_rows - added_rows)) {
            share->rows = 0;
        }
        else {
            share->rows += (added_rows - deleted_rows);
        }
        pthread_mutex_unlock(&share->mutex);
        added_rows = 0;
        deleted_rows = 0;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5941
        share->rows_from_locked_table = 0;
5942 5943 5944 5945 5946 5947 5948 5949
        if (!--trx->tokudb_lock_count) {
            if (trx->stmt) {
                /*
                   F_UNLCK is done without a transaction commit / rollback.
                   This happens if the thread didn't update any rows
                   We must in this case commit the work to keep the row locks
                 */
                DBUG_PRINT("trans", ("commiting non-updating transaction"));
5950
                reset_stmt_progress(&trx->stmt_progress);
5951 5952 5953 5954 5955 5956 5957 5958
                if (!is_fast_alter_running) {
                    commit_txn(trx->stmt, 0);
                    if (tokudb_debug & TOKUDB_DEBUG_TXN) {
                        TOKUDB_TRACE("commit:%p:%d\n", trx->stmt, error);
                    }
                    trx->stmt = NULL;
                    trx->sub_sp_level = NULL;
                }
5959 5960
            }
        }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5961
        transaction = NULL;
5962
        is_fast_alter_running = false;
5963
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5964
cleanup:
5965 5966
    if (tokudb_debug & TOKUDB_DEBUG_LOCK)
        TOKUDB_TRACE("%s error=%d\n", __FUNCTION__, error);
5967
    TOKUDB_DBUG_RETURN(error);
5968 5969 5970 5971 5972 5973 5974 5975 5976
}


/*
  When using LOCK TABLE's external_lock is only called when the actual
  TABLE LOCK is done.
  Under LOCK TABLES, each used tables will force a call to start_stmt.
*/

5977
int ha_tokudb::start_stmt(THD * thd, thr_lock_type lock_type) {
5978
    TOKUDB_DBUG_ENTER("ha_tokudb::start_stmt cmd=%d %d", thd_sql_command(thd), lock_type);
5979
    int error = 0;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
5980 5981


5982
    tokudb_trx_data *trx = (tokudb_trx_data *) thd_data_get(thd, tokudb_hton->slot);
5983 5984 5985 5986 5987 5988 5989
    DBUG_ASSERT(trx);
    /*
       note that trx->stmt may have been already initialized as start_stmt()
       is called for *each table* not for each storage engine,
       and there could be many bdb tables referenced in the query
     */
    if (!trx->stmt) {
5990 5991 5992 5993
        error = create_txn(thd, trx);
        if (error) {
            goto cleanup;
        }
5994
    }
5995 5996 5997 5998 5999
    else {
        if (tokudb_debug & TOKUDB_DEBUG_TXN) {
            TOKUDB_TRACE("trx->stmt already existed\n");
        }
    }
6000 6001 6002 6003 6004 6005 6006 6007 6008 6009 6010
    //
    // we know we are in lock tables
    // attempt to grab a table lock
    // if fail, continue, do not return error
    // This is because a failure ok, it simply means
    // another active transaction has some locks.
    // That other transaction modify this table
    // until it is unlocked, therefore having acquire_table_lock
    // potentially grab some locks but not all is ok.
    //
    if (lock.type <= TL_READ_NO_INSERT) {
6011
        acquire_table_lock(trx->sub_sp_level,lock_read);
6012 6013
    }
    else {
6014 6015 6016 6017 6018 6019
        if (!(thd_sql_command(thd) == SQLCOM_CREATE_INDEX ||
            thd_sql_command(thd) == SQLCOM_ALTER_TABLE ||
            thd_sql_command(thd) == SQLCOM_DROP_INDEX ||
            thd_sql_command(thd) == SQLCOM_TRUNCATE)) {
            acquire_table_lock(trx->sub_sp_level,lock_write);
        }
6020
    }    
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6021 6022 6023
    if (added_rows > deleted_rows) {
        share->rows_from_locked_table = added_rows - deleted_rows;
    }
6024 6025
    transaction = trx->sub_sp_level;
    trans_register_ha(thd, FALSE, tokudb_hton);
6026
cleanup:
6027
    TOKUDB_DBUG_RETURN(error);
6028 6029
}

6030 6031 6032 6033 6034

u_int32_t ha_tokudb::get_cursor_isolation_flags(enum thr_lock_type lock_type, THD* thd) {
    uint sql_command = thd_sql_command(thd);
    bool in_lock_tables = thd_in_lock_tables(thd);

6035 6036 6037 6038 6039 6040 6041
    //
    // following InnoDB's lead and having checksum command use a snapshot read if told
    //
    if (sql_command == SQLCOM_CHECKSUM) {
        return 0;
    }
    else if ((lock_type == TL_READ && in_lock_tables) || 
6042 6043 6044
             (lock_type == TL_READ_HIGH_PRIORITY && in_lock_tables) || 
             sql_command != SQLCOM_SELECT ||
             (sql_command == SQLCOM_SELECT && lock_type >= TL_WRITE_ALLOW_WRITE)) { // select for update 
6045 6046 6047 6048 6049 6050 6051
        return DB_SERIALIZABLE;
    }
    else {
        return 0;
    }
}

6052 6053 6054 6055 6056 6057 6058 6059 6060 6061 6062 6063 6064 6065 6066 6067 6068 6069 6070 6071 6072 6073 6074 6075 6076 6077 6078 6079
/*
  The idea with handler::store_lock() is the following:

  The statement decided which locks we should need for the table
  for updates/deletes/inserts we get WRITE locks, for SELECT... we get
  read locks.

  Before adding the lock into the table lock handler (see thr_lock.c)
  mysqld calls store lock with the requested locks.  Store lock can now
  modify a write lock to a read lock (or some other lock), ignore the
  lock (if we don't want to use MySQL table locks at all) or add locks
  for many tables (like we do when we are using a MERGE handler).

  Tokudb DB changes all WRITE locks to TL_WRITE_ALLOW_WRITE (which
  signals that we are doing WRITES, but we are still allowing other
  reader's and writer's.

  When releasing locks, store_lock() are also called. In this case one
  usually doesn't have to do anything.

  In some exceptional cases MySQL may send a request for a TL_IGNORE;
  This means that we are requesting the same lock as last time and this
  should also be ignored. (This may happen when someone does a flush
  table when we have opened a part of the tables, in which case mysqld
  closes and reopens the tables and tries to get the same locks at last
  time).  In the future we will probably try to remove this.
*/

6080
THR_LOCK_DATA **ha_tokudb::store_lock(THD * thd, THR_LOCK_DATA ** to, enum thr_lock_type lock_type) {
6081
    TOKUDB_DBUG_ENTER("ha_tokudb::store_lock, lock_type=%d cmd=%d", lock_type, thd_sql_command(thd));
6082
    if (tokudb_debug & TOKUDB_DEBUG_LOCK) {
6083
        TOKUDB_TRACE("%s lock_type=%d cmd=%d\n", __FUNCTION__, lock_type, thd_sql_command(thd));
6084
    }
6085 6086 6087 6088 6089 6090 6091 6092 6093 6094

    if (lock_type != TL_IGNORE && lock.type == TL_UNLOCK) {
        // if creating a hot index
        if (get_create_index_online(thd) && thd_sql_command(thd)== SQLCOM_CREATE_INDEX) {
            rw_rdlock(&share->num_DBs_lock);
            if (share->num_DBs == (table->s->keys + test(hidden_primary_key))) {
                lock_type = TL_WRITE_ALLOW_WRITE;
            }
            lock.type = lock_type;
            rw_unlock(&share->num_DBs_lock);
6095 6096 6097 6098 6099 6100
        } 
        // force alter table to lock out other readers
        else if (thd_sql_command(thd)== SQLCOM_CREATE_INDEX || 
                 thd_sql_command(thd)== SQLCOM_ALTER_TABLE ||
                 thd_sql_command(thd)== SQLCOM_DROP_INDEX) {
            lock_type = TL_WRITE;
6101
            lock.type = lock_type;
6102 6103
        }
        else {
6104 6105 6106 6107 6108 6109
            // If we are not doing a LOCK TABLE, then allow multiple writers
            if ((lock_type >= TL_WRITE_CONCURRENT_INSERT && lock_type <= TL_WRITE) && 
                !thd->in_lock_tables && thd_sql_command(thd) != SQLCOM_TRUNCATE && !thd_tablespace_op(thd)) {
                lock_type = TL_WRITE_ALLOW_WRITE;
            }
            lock.type = lock_type;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6110
        }
6111
    }
6112
    *to++ = &lock;
6113 6114
    if (tokudb_debug & TOKUDB_DEBUG_LOCK)
        TOKUDB_TRACE("%s lock_type=%d\n", __FUNCTION__, lock_type);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6115
    DBUG_RETURN(to);
6116 6117
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
6118 6119 6120 6121 6122 6123 6124
int toku_dbt_up(DB*,
                                 u_int32_t old_version, const DBT *old_descriptor, const DBT *old_key, const DBT *old_val,
                                 u_int32_t new_version, const DBT *new_descriptor, const DBT *new_key, const DBT *new_val) {
    assert(false);
    return 0;
}

6125 6126 6127 6128 6129 6130 6131 6132 6133
static int create_sub_table(
    const char *table_name, 
    DBT* row_descriptor, 
    DB_TXN* txn, 
    uint32_t block_size, 
    uint32_t read_block_size, 
    bool is_hot_index
    ) 
{
6134
    TOKUDB_DBUG_ENTER("create_sub_table");
6135
    int error;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6136
    DB *file = NULL;
6137
    u_int32_t create_flags;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6138
    
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6139 6140 6141 6142 6143 6144 6145 6146
    
    error = db_create(&file, db_env, 0);
    if (error) {
        DBUG_PRINT("error", ("Got error: %d when creating table", error));
        my_errno = error;
        goto exit;
    }
        
6147

6148 6149
    if (block_size != 0) {
        error = file->set_pagesize(file, block_size);
6150
        if (error != 0) {
6151
            DBUG_PRINT("error", ("Got error: %d when setting block size %u for table '%s'", error, block_size, table_name));
6152 6153 6154
            goto exit;
        }
    }
6155 6156 6157 6158 6159 6160 6161
    if (read_block_size != 0) {
        error = file->set_readpagesize(file, read_block_size);
        if (error != 0) {
            DBUG_PRINT("error", ("Got error: %d when setting read block size %u for table '%s'", error, read_block_size, table_name));
            goto exit;
        }
    }
6162 6163 6164

    create_flags = DB_THREAD | DB_CREATE | DB_EXCL | (is_hot_index ? DB_IS_HOT_INDEX : 0);    
    error = file->open(file, txn, table_name, NULL, DB_BTREE, create_flags, my_umask);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6165 6166 6167 6168
    if (error) {
        DBUG_PRINT("error", ("Got error: %d when opening table '%s'", error, table_name));
        goto exit;
    } 
6169

6170 6171 6172 6173 6174 6175
    error = file->change_descriptor(file, txn, row_descriptor, (is_hot_index ? DB_IS_HOT_INDEX : 0));
    if (error) {
        DBUG_PRINT("error", ("Got error: %d when setting row descriptor for table '%s'", error, table_name));
        goto exit;
    }

Zardosht Kasheff's avatar
Zardosht Kasheff committed
6176 6177
    error = 0;
exit:
6178
    if (file) {
6179 6180
        int r = file->close(file, 0);
        assert(r==0);
6181
    }
6182
    TOKUDB_DBUG_RETURN(error);
6183 6184
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
6185 6186 6187 6188 6189 6190 6191
void ha_tokudb::update_create_info(HA_CREATE_INFO* create_info) {
    if (share->has_auto_inc) {
        info(HA_STATUS_AUTO);
        create_info->auto_increment_value = stats.auto_increment_value;
    }
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
6192
//
6193 6194 6195 6196
// removes key name from status.tokudb.
// needed for when we are dropping indexes, so that 
// during drop table, we do not attempt to remove already dropped
// indexes because we did not keep status.tokudb in sync with list of indexes.
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6197
//
6198
int ha_tokudb::remove_key_name_from_status(DB* status_block, char* key_name, DB_TXN* txn) {
6199
    int error;
6200 6201 6202 6203 6204 6205 6206 6207 6208 6209 6210 6211 6212 6213 6214 6215 6216 6217 6218
    uchar status_key_info[FN_REFLEN + sizeof(HA_METADATA_KEY)];
    HA_METADATA_KEY md_key = hatoku_key_name;
    memcpy(status_key_info, &md_key, sizeof(HA_METADATA_KEY));
    //
    // put index name in status.tokudb
    // 
    memcpy(
        status_key_info + sizeof(HA_METADATA_KEY), 
        key_name, 
        strlen(key_name) + 1
        );
    error = remove_metadata(
        status_block,
        status_key_info,
        sizeof(HA_METADATA_KEY) + strlen(key_name) + 1,
        txn
        );
    return error;
}
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6219

6220 6221 6222 6223 6224 6225 6226 6227 6228 6229 6230 6231 6232 6233 6234 6235 6236 6237 6238 6239 6240 6241 6242 6243 6244 6245 6246
//
// writes the key name in status.tokudb, so that we may later delete or rename
// the dictionary associated with key_name
//
int ha_tokudb::write_key_name_to_status(DB* status_block, char* key_name, DB_TXN* txn) {
    int error;
    uchar status_key_info[FN_REFLEN + sizeof(HA_METADATA_KEY)];
    HA_METADATA_KEY md_key = hatoku_key_name;
    memcpy(status_key_info, &md_key, sizeof(HA_METADATA_KEY));
    //
    // put index name in status.tokudb
    // 
    memcpy(
        status_key_info + sizeof(HA_METADATA_KEY), 
        key_name, 
        strlen(key_name) + 1
        );
    error = write_metadata(
        status_block,
        status_key_info,
        sizeof(HA_METADATA_KEY) + strlen(key_name) + 1,
        NULL,
        0,
        txn
        );
    return error;
}
6247

6248 6249 6250 6251 6252
//
// some tracing moved out of ha_tokudb::create, because ::create was getting cluttered
//
void ha_tokudb::trace_create_table_info(const char *name, TABLE * form) {
    uint i;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6253 6254 6255
    //
    // tracing information about what type of table we are creating
    //
6256 6257 6258 6259 6260 6261 6262 6263 6264 6265 6266 6267 6268 6269 6270 6271 6272
    if (tokudb_debug & TOKUDB_DEBUG_OPEN) {
        for (i = 0; i < form->s->fields; i++) {
            Field *field = form->s->field[i];
            TOKUDB_TRACE("field:%d:%s:type=%d:flags=%x\n", i, field->field_name, field->type(), field->flags);
        }
        for (i = 0; i < form->s->keys; i++) {
            KEY *key = &form->s->key_info[i];
            TOKUDB_TRACE("key:%d:%s:%d\n", i, key->name, key->key_parts);
            uint p;
            for (p = 0; p < key->key_parts; p++) {
                KEY_PART_INFO *key_part = &key->key_part[p];
                Field *field = key_part->field;
                TOKUDB_TRACE("key:%d:%d:length=%d:%s:type=%d:flags=%x\n",
                             i, p, key_part->length, field->field_name, field->type(), field->flags);
            }
        }
    }
6273 6274
}

6275 6276 6277 6278 6279 6280 6281 6282 6283 6284 6285 6286 6287 6288 6289 6290 6291 6292 6293 6294 6295 6296 6297 6298 6299 6300 6301 6302 6303 6304 6305 6306 6307 6308 6309 6310 6311 6312 6313 6314 6315 6316 6317 6318 6319 6320 6321 6322 6323 6324 6325 6326 6327
u_int32_t get_max_desc_size(KEY_AND_COL_INFO* kc_info, TABLE* form) {
    u_int32_t max_row_desc_buff_size;
    max_row_desc_buff_size = 2*(form->s->fields * 6)+10; // upper bound of key comparison descriptor
    max_row_desc_buff_size += get_max_secondary_key_pack_desc_size(kc_info); // upper bound for sec. key part
    max_row_desc_buff_size += get_max_clustering_val_pack_desc_size(form->s); // upper bound for clustering val part
    return max_row_desc_buff_size;
}

u_int32_t create_secondary_key_descriptor(
    uchar* buf,
    KEY* key_info,
    KEY* prim_key,
    uint hpk,
    TABLE* form,
    uint primary_key,
    u_int32_t keynr,
    KEY_AND_COL_INFO* kc_info    
    ) 
{
    uchar* ptr = NULL;

    ptr = buf;
    ptr += create_toku_key_descriptor(
        ptr,
        false,
        key_info,
        hpk,
        prim_key
        );

    ptr += create_toku_secondary_key_pack_descriptor(
        ptr,
        hpk,
        primary_key,
        form->s,
        form,
        kc_info,
        key_info,
        prim_key
        );

    ptr += create_toku_clustering_val_pack_descriptor(
        ptr,
        primary_key,
        form->s,
        kc_info,
        keynr,
        key_info->flags & HA_CLUSTERING
        );
    return ptr - buf;
}


6328 6329 6330
//
// creates dictionary for secondary index, with key description key_info, all using txn
//
6331 6332 6333 6334 6335 6336 6337 6338 6339
int ha_tokudb::create_secondary_dictionary(
    const char* name, TABLE* form, 
    KEY* key_info, 
    DB_TXN* txn, 
    KEY_AND_COL_INFO* kc_info, 
    u_int32_t keynr,
    bool is_hot_index
    ) 
{
6340 6341 6342 6343 6344 6345
    int error;
    DBT row_descriptor;
    uchar* row_desc_buff = NULL;
    char* newname = NULL;
    KEY* prim_key = NULL;
    char dict_name[MAX_DICT_NAME_LEN];
6346
    u_int32_t max_row_desc_buff_size;
6347
    uint hpk= (form->s->primary_key >= MAX_KEY) ? TOKUDB_HIDDEN_PRIMARY_KEY_LENGTH : 0;
6348
    uint32_t block_size;
6349 6350
    uint32_t read_block_size;
    THD* thd = ha_thd();
6351

6352
    bzero(&row_descriptor, sizeof(row_descriptor));
6353
    
6354
    max_row_desc_buff_size = get_max_desc_size(kc_info,form);
6355 6356

    row_desc_buff = (uchar *)my_malloc(max_row_desc_buff_size, MYF(MY_WME));
6357 6358 6359 6360 6361 6362 6363 6364 6365 6366 6367 6368 6369 6370
    if (row_desc_buff == NULL){ error = ENOMEM; goto cleanup;}

    newname = (char *)my_malloc(get_max_dict_name_path_length(name),MYF(MY_WME));
    if (newname == NULL){ error = ENOMEM; goto cleanup;}

    sprintf(dict_name, "key-%s", key_info->name);
    make_name(newname, name, dict_name);

    prim_key = (hpk) ? NULL : &form->s->key_info[primary_key];

    //
    // setup the row descriptor
    //
    row_descriptor.data = row_desc_buff;
6371 6372 6373
    //
    // save data necessary for key comparisons
    //
6374
    row_descriptor.size = create_secondary_key_descriptor(
6375 6376
        row_desc_buff,
        key_info,
6377
        prim_key,
6378
        hpk,
6379 6380 6381
        form,
        primary_key,
        keynr,
6382
        kc_info    
6383 6384 6385
        );
    assert(row_descriptor.size <= max_row_desc_buff_size);

6386 6387 6388 6389
    block_size = key_info->block_size << 10;
    if (block_size == 0) {
        block_size = get_tokudb_block_size(thd);
    }
6390
    read_block_size = get_tokudb_read_block_size(thd);
6391

6392
    error = create_sub_table(newname, &row_descriptor, txn, block_size, read_block_size, is_hot_index);
6393 6394 6395 6396 6397 6398
cleanup:    
    my_free(newname, MYF(MY_ALLOW_ZERO_PTR));
    my_free(row_desc_buff, MYF(MY_ALLOW_ZERO_PTR));
    return error;
}

6399 6400 6401 6402 6403 6404 6405 6406 6407 6408 6409 6410 6411 6412 6413 6414 6415 6416 6417 6418 6419 6420 6421 6422 6423 6424 6425 6426 6427 6428 6429 6430 6431 6432

u_int32_t create_main_key_descriptor(
    uchar* buf,
    KEY* prim_key,
    uint hpk,
    uint primary_key,
    TABLE* form,
    KEY_AND_COL_INFO* kc_info
    ) 
{
    uchar* ptr = buf;
    ptr += create_toku_key_descriptor(
        ptr, 
        hpk,
        prim_key,
        false,
        NULL
        );
    
    ptr += create_toku_main_key_pack_descriptor(
        ptr
        );

    ptr += create_toku_clustering_val_pack_descriptor(
        ptr,
        primary_key,
        form->s,
        kc_info,
        primary_key,
        false
        );
    return ptr - buf;
}

6433 6434 6435 6436
//
// create and close the main dictionarr with name of "name" using table form, all within
// transaction txn.
//
6437
int ha_tokudb::create_main_dictionary(const char* name, TABLE* form, DB_TXN* txn, KEY_AND_COL_INFO* kc_info) {
6438 6439 6440 6441 6442
    int error;
    DBT row_descriptor;
    uchar* row_desc_buff = NULL;
    char* newname = NULL;
    KEY* prim_key = NULL;
6443
    u_int32_t max_row_desc_buff_size;
6444
    uint hpk= (form->s->primary_key >= MAX_KEY) ? TOKUDB_HIDDEN_PRIMARY_KEY_LENGTH : 0;
6445
    uint32_t block_size;
6446 6447
    uint32_t read_block_size;
    THD* thd = ha_thd();
6448 6449

    bzero(&row_descriptor, sizeof(row_descriptor));
6450
    max_row_desc_buff_size = get_max_desc_size(kc_info, form);
6451 6452

    row_desc_buff = (uchar *)my_malloc(max_row_desc_buff_size, MYF(MY_WME));
6453 6454 6455 6456
    if (row_desc_buff == NULL){ error = ENOMEM; goto cleanup;}

    newname = (char *)my_malloc(get_max_dict_name_path_length(name),MYF(MY_WME));
    if (newname == NULL){ error = ENOMEM; goto cleanup;}
6457

6458
    make_name(newname, name, "main");
6459 6460

    prim_key = (hpk) ? NULL : &form->s->key_info[primary_key];
6461

Zardosht Kasheff's avatar
Zardosht Kasheff committed
6462 6463 6464 6465
    //
    // setup the row descriptor
    //
    row_descriptor.data = row_desc_buff;
6466 6467 6468
    //
    // save data necessary for key comparisons
    //
6469 6470
    row_descriptor.size = create_main_key_descriptor(
        row_desc_buff,
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6471
        prim_key,
6472
        hpk,
6473
        primary_key,
6474 6475
        form,
        kc_info
6476 6477
        );
    assert(row_descriptor.size <= max_row_desc_buff_size);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6478

6479 6480 6481 6482 6483 6484
    block_size = 0;
    if (prim_key)
        block_size = prim_key->block_size << 10;
    if (block_size == 0) {
        block_size = get_tokudb_block_size(thd);
    }
6485
    read_block_size = get_tokudb_read_block_size(thd);
6486

6487
    /* Create the main table that will hold the real rows */
6488
    error = create_sub_table(newname, &row_descriptor, txn, block_size, read_block_size, false);
6489 6490 6491 6492 6493 6494 6495 6496 6497 6498 6499 6500 6501 6502 6503 6504 6505 6506 6507 6508 6509 6510 6511
cleanup:    
    my_free(newname, MYF(MY_ALLOW_ZERO_PTR));
    my_free(row_desc_buff, MYF(MY_ALLOW_ZERO_PTR));
    return error;
}

//
// Creates a new table
// Parameters:
//      [in]    name - table name
//      [in]    form - info on table, columns and indexes
//      [in]    create_info - more info on table, CURRENTLY UNUSED
// Returns:
//      0 on success
//      error otherwise
//
int ha_tokudb::create(const char *name, TABLE * form, HA_CREATE_INFO * create_info) {
    TOKUDB_DBUG_ENTER("ha_tokudb::create");
    int error;
    DB *status_block = NULL;
    uint version;
    uint capabilities;
    DB_TXN* txn = NULL;
6512
    bool do_commit = false;
6513
    char* newname = NULL;
6514
    KEY_AND_COL_INFO kc_info;
6515
    tokudb_trx_data *trx = NULL;
6516
    THD* thd = ha_thd();
6517
    bool create_from_engine= (create_info->table_options & HA_OPTION_CREATE_FROM_ENGINE);
6518
    bzero(&kc_info, sizeof(kc_info));
6519

6520
    pthread_mutex_lock(&tokudb_meta_mutex);
6521

6522 6523
    trx = (tokudb_trx_data *) thd_data_get(ha_thd(), tokudb_hton->slot);

6524 6525 6526 6527 6528 6529 6530
    if (create_from_engine) {
        // table already exists, nothing to do
        error = 0;
        goto cleanup;
    }
    

6531 6532
    newname = (char *)my_malloc(get_max_dict_name_path_length(name),MYF(MY_WME));
    if (newname == NULL){ error = ENOMEM; goto cleanup;}
6533

6534
    if (thd_sql_command(thd) == SQLCOM_CREATE_TABLE && trx && trx->sub_sp_level) {
6535 6536 6537 6538 6539 6540 6541 6542
        txn = trx->sub_sp_level;
    }
    else {
        do_commit = true;
        error = db_env->txn_begin(db_env, 0, &txn, 0);
        if (error) { goto cleanup; }        
    }
    
6543 6544
    primary_key = form->s->primary_key;
    hidden_primary_key = (primary_key  >= MAX_KEY) ? TOKUDB_HIDDEN_PRIMARY_KEY_LENGTH : 0;
6545 6546 6547
    if (hidden_primary_key) {
        primary_key = form->s->keys;
    }
6548 6549 6550 6551

    /* do some tracing */
    trace_create_table_info(name,form);

Zardosht Kasheff's avatar
Zardosht Kasheff committed
6552
    /* Create status.tokudb and save relevant metadata */
6553
    make_name(newname, name, "status");
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6554

6555 6556 6557
    error = db_create(&status_block, db_env, 0);
    if (error) { goto cleanup; }

6558
    error = status_block->open(status_block, txn, newname, NULL, DB_BTREE, DB_CREATE | DB_EXCL, 0);
6559
    if (error) { goto cleanup; }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6560

6561 6562 6563
    version = HA_TOKU_VERSION;
    capabilities = HA_TOKU_CAP;
    
6564
    error = write_to_status(status_block, hatoku_new_version,&version,sizeof(version), txn);
6565
    if (error) { goto cleanup; }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6566

6567
    error = write_to_status(status_block, hatoku_capabilities,&capabilities,sizeof(capabilities), txn);
6568 6569
    if (error) { goto cleanup; }

6570
    error = write_auto_inc_create(status_block, create_info->auto_increment_value, txn);
6571
    if (error) { goto cleanup; }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6572

6573 6574 6575 6576 6577
    // only for tables that are not partitioned
    if (form->part_info == NULL) {
        error = write_frm_data(status_block, txn, form->s->path.str);
        if (error) { goto cleanup; }
    }
6578 6579 6580 6581 6582 6583 6584 6585 6586 6587 6588
    error = allocate_key_and_col_info(form->s, &kc_info);
    if (error) { goto cleanup; }

    error = initialize_key_and_col_info(
        form->s, 
        form,
        &kc_info,
        hidden_primary_key,
        primary_key
        );
    if (error) { goto cleanup; }
6589

6590
    error = create_main_dictionary(name, form, txn, &kc_info);
6591 6592 6593 6594
    if (error) {
        goto cleanup;
    }

6595 6596 6597

    for (uint i = 0; i < form->s->keys; i++) {
        if (i != primary_key) {
6598
            error = create_secondary_dictionary(name, form, &form->key_info[i], txn, &kc_info, i, false);
6599 6600 6601 6602 6603 6604 6605 6606 6607 6608 6609 6610
            if (error) {
                goto cleanup;
            }

            error = write_key_name_to_status(status_block, form->s->key_info[i].name, txn);
            if (error) { goto cleanup; }
        }
    }

    error = add_table_to_metadata(name, form, txn);
    if (error) { goto cleanup; }

6611
    error = 0;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6612 6613
cleanup:
    if (status_block != NULL) {
6614 6615
        int r = status_block->close(status_block, 0);
        assert(r==0);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6616
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6617
    free_key_and_col_info(&kc_info);
6618
    if (do_commit && txn) {
6619 6620 6621 6622 6623 6624
        if (error) {
            abort_txn(txn);
        }
        else {
            commit_txn(txn,0);
        }
6625
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6626
    my_free(newname, MYF(MY_ALLOW_ZERO_PTR));
6627
    pthread_mutex_unlock(&tokudb_meta_mutex);
6628
    TOKUDB_DBUG_RETURN(error);
6629 6630
}

6631 6632 6633 6634 6635 6636 6637 6638 6639 6640 6641 6642 6643
int ha_tokudb::discard_or_import_tablespace(my_bool discard) {
    /*
    if (discard) {
        my_errno=HA_ERR_WRONG_COMMAND;
        return my_errno;
    }
    return add_table_to_metadata(share->table_name);
    */
    my_errno=HA_ERR_WRONG_COMMAND;
    return my_errno;
}


Zardosht Kasheff's avatar
Zardosht Kasheff committed
6644
//
6645 6646 6647 6648
// deletes from_name or renames from_name to to_name, all using transaction txn.
// is_delete specifies which we are doing
// is_key specifies if it is a secondary index (and hence a "key-" needs to be prepended) or
// if it is not a secondary index
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6649
//
6650
int ha_tokudb::delete_or_rename_dictionary( const char* from_name, const char* to_name, const char* secondary_name, bool is_key, DB_TXN* txn, bool is_delete) {
6651 6652 6653 6654 6655 6656 6657 6658 6659 6660 6661 6662 6663 6664 6665 6666 6667 6668 6669 6670 6671 6672 6673 6674 6675 6676 6677 6678 6679 6680 6681 6682 6683 6684 6685 6686 6687 6688 6689 6690 6691 6692 6693 6694 6695 6696 6697 6698 6699 6700
    int error;
    char dict_name[MAX_DICT_NAME_LEN];
    char* new_from_name = NULL;
    char* new_to_name = NULL;
    assert(txn);
    
    new_from_name = (char *)my_malloc(
        get_max_dict_name_path_length(from_name), 
        MYF(MY_WME)
        );
    if (new_from_name == NULL) {
        error = ENOMEM;
        goto cleanup;
    }
    if (!is_delete) {
        assert(to_name);
        new_to_name = (char *)my_malloc(
            get_max_dict_name_path_length(to_name), 
            MYF(MY_WME)
            );
        if (new_to_name == NULL) {
            error = ENOMEM;
            goto cleanup;
        }
    }
    
    if (is_key) {
        sprintf(dict_name, "key-%s", secondary_name);
        make_name(new_from_name, from_name, dict_name);
    }
    else {
        make_name(new_from_name, from_name, secondary_name);
    }
    if (!is_delete) {
        if (is_key) {
            sprintf(dict_name, "key-%s", secondary_name);
            make_name(new_to_name, to_name, dict_name);
        }
        else {
            make_name(new_to_name, to_name, secondary_name);
        }
    }

    if (is_delete) {    
        error = db_env->dbremove(db_env, txn, new_from_name, NULL, 0);
    }
    else {
        error = db_env->dbrename(db_env, txn, new_from_name, NULL, new_to_name, 0);
    }
    if (error) { goto cleanup; }
6701

6702 6703 6704 6705 6706 6707 6708 6709 6710 6711 6712 6713
cleanup:
    my_free(new_from_name, MYF(MY_ALLOW_ZERO_PTR));
    my_free(new_to_name, MYF(MY_ALLOW_ZERO_PTR));
    return error;
}


//
// deletes or renames a table. if is_delete is true, then we delete, and to_name can be NULL
// if is_delete is false, then to_name must be non-NULL, as we are renaming the table.
//
int ha_tokudb::delete_or_rename_table (const char* from_name, const char* to_name, bool is_delete) {
6714
    int error;
6715
    DB* status_db = NULL;
6716 6717 6718 6719 6720 6721
    DBC* status_cursor = NULL;
    DB_TXN* txn = NULL;
    DBT curr_key;
    DBT curr_val;
    bzero(&curr_key, sizeof(curr_key));
    bzero(&curr_val, sizeof(curr_val));
6722
    pthread_mutex_lock(&tokudb_meta_mutex);
6723

6724 6725 6726
    error = db_env->txn_begin(db_env, 0, &txn, 0);
    if (error) { goto cleanup; }

6727
    //
6728
    // modify metadata db
6729
    //
6730 6731
    if (is_delete) {
        error = drop_table_from_metadata(from_name, txn);
6732
    }
6733 6734
    else {
        error = rename_table_in_metadata(from_name, to_name, txn);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6735
    }
6736 6737 6738 6739 6740 6741 6742 6743 6744 6745 6746 6747 6748 6749 6750 6751 6752 6753 6754 6755 6756 6757 6758 6759 6760 6761 6762 6763 6764 6765 6766 6767 6768 6769 6770 6771 6772 6773
    if (error) { goto cleanup; }

    //
    // open status db,
    // create cursor,
    // for each name read out of there, create a db and delete or rename it
    //
    error = open_status_dictionary(&status_db, from_name, txn);
    if (error) { goto cleanup; }

    error = status_db->cursor(status_db, txn, &status_cursor, 0);
    if (error) { goto cleanup; }

    while (error != DB_NOTFOUND) {
        error = status_cursor->c_get(
            status_cursor,
            &curr_key,
            &curr_val,
            DB_NEXT
            );
        if (error && error != DB_NOTFOUND) { goto cleanup; }
        if (error == DB_NOTFOUND) { break; }

        HA_METADATA_KEY mk = *(HA_METADATA_KEY *)curr_key.data;
        if (mk != hatoku_key_name) {
            continue;
        }
        error = delete_or_rename_dictionary(from_name, to_name, (char *)((char *)curr_key.data + sizeof(HA_METADATA_KEY)), true, txn, is_delete);
        if (error) { goto cleanup; }
    }

    //
    // delete or rename main.tokudb
    //
    error = delete_or_rename_dictionary(from_name, to_name, "main", false, txn, is_delete);
    if (error) { goto cleanup; }

    error = status_cursor->c_close(status_cursor);
6774
    assert(error==0);
6775
    status_cursor = NULL;
6776
    if (error) { goto cleanup; }
6777 6778

    error = status_db->close(status_db, 0);
6779
    assert(error == 0);
6780 6781 6782 6783 6784 6785 6786 6787
    status_db = NULL;
    
    //
    // delete or rename status.tokudb
    //
    error = delete_or_rename_dictionary(from_name, to_name, "status", false, txn, is_delete);
    if (error) { goto cleanup; }

6788
    my_errno = error;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6789
cleanup:
6790
    if (status_cursor) {
6791 6792
        int r = status_cursor->c_close(status_cursor);
        assert(r==0);
6793 6794
    }
    if (status_db) {
6795 6796
        int r = status_db->close(status_db, 0);
        assert(r==0);
6797 6798 6799 6800 6801 6802 6803 6804
    }
    if (txn) {
        if (error) {
            abort_txn(txn);
        }
        else {
            commit_txn(txn, 0);
        }
6805
    }
6806
    pthread_mutex_unlock(&tokudb_meta_mutex);
6807 6808 6809 6810 6811 6812 6813 6814 6815 6816 6817 6818 6819 6820
    return error;
}


//
// Drops table
// Parameters:
//      [in]    name - name of table to be deleted
// Returns:
//      0 on success
//      error otherwise
//
int ha_tokudb::delete_table(const char *name) {
    TOKUDB_DBUG_ENTER("ha_tokudb::delete_table");
6821 6822
    int error;
    error = delete_or_rename_table(name, NULL, true);
6823
    if (error == DB_LOCK_NOTGRANTED && ((tokudb_debug & TOKUDB_DEBUG_HIDE_DDL_LOCK_ERRORS) == 0)) {
6824 6825 6826 6827 6828
        sql_print_error("Could not delete table %s because \
another transaction has accessed the table. \
To drop the table, make sure no transactions touch the table.", name);
    }
    TOKUDB_DBUG_RETURN(error);
6829 6830 6831
}


Zardosht Kasheff's avatar
Zardosht Kasheff committed
6832 6833 6834 6835 6836 6837 6838 6839 6840
//
// renames table from "from" to "to"
// Parameters:
//      [in]    name - old name of table
//      [in]    to - new name of table
// Returns:
//      0 on success
//      error otherwise
//
6841
int ha_tokudb::rename_table(const char *from, const char *to) {
6842
    TOKUDB_DBUG_ENTER("%s %s %s", __FUNCTION__, from, to);
6843
    int error;
6844
    error = delete_or_rename_table(from, to, false);
6845
    if (error == DB_LOCK_NOTGRANTED && ((tokudb_debug & TOKUDB_DEBUG_HIDE_DDL_LOCK_ERRORS) == 0)) {
6846 6847 6848 6849
        sql_print_error("Could not rename table from %s to %s because \
another transaction has accessed the table. \
To rename the table, make sure no transactions touch the table.", from, to);
    }
6850
    TOKUDB_DBUG_RETURN(error);
6851 6852 6853 6854
}


/*
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6855
  Returns estimate on number of seeks it will take to read through the table
6856 6857 6858
  This is to be comparable to the number returned by records_in_range so
  that we can decide if we should scan the table or use keys.
*/
6859
/// QQQ why divide by 3
6860
double ha_tokudb::scan_time() {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6861
    TOKUDB_DBUG_ENTER("ha_tokudb::scan_time");
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6862
    double ret_val = (double)stats.records / 3;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6863
    DBUG_RETURN(ret_val);
6864 6865
}

6866 6867
double ha_tokudb::keyread_time(uint index, uint ranges, ha_rows rows)
{
6868
  if ((table->key_info[index].flags & HA_CLUSTERING) || (index == primary_key)) {
6869 6870 6871 6872 6873 6874 6875 6876 6877 6878 6879 6880 6881 6882 6883 6884 6885
    return read_time(index, ranges, rows);
  }
  /*
    It is assumed that we will read trough the whole key range and that all
    key blocks are half full (normally things are much better). It is also
    assumed that each time we read the next key from the index, the handler
    performs a random seek, thus the cost is proportional to the number of
    blocks read. This model does not take into account clustered indexes -
    engines that support that (e.g. InnoDB) may want to overwrite this method.
  */
  double keys_per_block= (stats.block_size/2.0/
                          (table->key_info[index].key_length +
                           ref_length) + 1);
  return (rows + keys_per_block - 1)/ keys_per_block;
}


Zardosht Kasheff's avatar
Zardosht Kasheff committed
6886 6887 6888 6889 6890 6891 6892 6893 6894 6895 6896 6897
//
// Calculate the time it takes to read a set of ranges through an index
// This enables us to optimize reads for clustered indexes.
// Implementation pulled from InnoDB
// Parameters:
//          index - index to use
//          ranges - number of ranges
//          rows - estimated number of rows in the range
// Returns:
//      estimated time measured in disk seeks
//
double ha_tokudb::read_time(
6898 6899 6900 6901
    uint    index,
    uint    ranges,
    ha_rows rows
    )
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6902 6903 6904
{
    double total_scan;
    double ret_val; 
6905
    bool is_primary = (index == primary_key);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6906
    bool is_clustering;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6907

Zardosht Kasheff's avatar
Zardosht Kasheff committed
6908 6909 6910 6911 6912 6913 6914
    //
    // in case for hidden primary key, this is called
    //
    if (index >= table_share->keys) {
        ret_val = handler::read_time(index, ranges, rows);
        goto cleanup;
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6915 6916
    
    is_clustering = (table->key_info[index].flags & HA_CLUSTERING);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6917 6918


Zardosht Kasheff's avatar
Zardosht Kasheff committed
6919 6920 6921
    //
    // if it is not the primary key, and it is not a clustering key, then return handler::read_time
    //
6922
    if (!(is_primary || is_clustering)) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6923 6924 6925 6926
        ret_val = handler::read_time(index, ranges, rows);
        goto cleanup;
    }

Zardosht Kasheff's avatar
Zardosht Kasheff committed
6927 6928 6929
    //
    // for primary key and for clustered keys, return a fraction of scan_time()
    //
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6930 6931 6932
    total_scan = scan_time();

    if (stats.records < rows) {
6933
        ret_val = is_clustering ? total_scan + 0.00001 : total_scan;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6934 6935 6936 6937 6938 6939 6940
        goto cleanup;
    }

    //
    // one disk seek per range plus the proportional scan time of the rows
    //
    ret_val = (ranges + (double) rows / (double) stats.records * total_scan);
6941
    ret_val = is_clustering ? ret_val + 0.00001 : ret_val;
6942
    
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6943 6944 6945 6946 6947
cleanup:
    return ret_val;
}


Zardosht Kasheff's avatar
Zardosht Kasheff committed
6948
//
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6949 6950 6951 6952 6953 6954 6955 6956 6957 6958 6959
// Estimates the number of index records in a range. In case of errors, return
//   HA_TOKUDB_RANGE_COUNT instead of HA_POS_ERROR. This was behavior
//   when we got the handlerton from MySQL.
// Parameters:
//              keynr -index to use 
//      [in]    start_key - low end of the range
//      [in]    end_key - high end of the range
// Returns:
//      0 - There are no matching keys in the given range
//      number > 0 - There are approximately number matching rows in the range
//      HA_POS_ERROR - Something is wrong with the index tree
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6960
//
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6961
ha_rows ha_tokudb::records_in_range(uint keynr, key_range* start_key, key_range* end_key) {
6962
    TOKUDB_DBUG_ENTER("ha_tokudb::records_in_range");
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6963
    DBT key;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6964
    ha_rows ret_val = HA_TOKUDB_RANGE_COUNT;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6965
    DB *kfile = share->key_file[keynr];
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6966
    u_int64_t less, equal, greater;
6967
    u_int64_t total_rows_estimate = HA_TOKUDB_RANGE_COUNT;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6968 6969
    u_int64_t start_rows, end_rows, rows;
    int is_exact;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6970
    int error;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6971
    uchar inf_byte;
6972

Zardosht Kasheff's avatar
Zardosht Kasheff committed
6973
    //
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6974
    // get start_rows and end_rows values so that we can estimate range
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6975 6976 6977 6978
    // when calling key_range64, the only value we can trust is the value for less
    // The reason is that the key being passed in may be a prefix of keys in the DB
    // As a result, equal may be 0 and greater may actually be equal+greater
    // So, we call key_range64 on the key, and the key that is after it.
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6979
    //
6980 6981 6982 6983 6984 6985
    if (!start_key && !end_key) {
        error = estimate_num_rows(kfile, &end_rows, transaction);
        if (error) {
            ret_val = HA_TOKUDB_RANGE_COUNT;
            goto cleanup;
        }
6986 6987
        ret_val = (end_rows <= 1) ? 1 : end_rows;
        goto cleanup;
6988
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6989
    if (start_key) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
6990 6991 6992 6993 6994 6995 6996 6997 6998 6999
        inf_byte = (start_key->flag == HA_READ_KEY_EXACT) ? 
            COL_NEG_INF : COL_POS_INF;
        pack_key(
            &key, 
            keynr, 
            key_buff, 
            start_key->key, 
            start_key->length, 
            inf_byte
            ); 
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7000
        error = kfile->key_range64(
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7001 7002
            kfile, 
            transaction, 
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7003
            &key,
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7004 7005 7006 7007
            &less,
            &equal,
            &greater,
            &is_exact
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7008 7009 7010 7011 7012
            );
        if (error) {
            ret_val = HA_TOKUDB_RANGE_COUNT;
            goto cleanup;
        }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7013
        start_rows= less;
7014
        total_rows_estimate = less + equal + greater;
7015
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7016
    else {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7017
        start_rows= 0;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7018 7019 7020
    }

    if (end_key) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7021 7022 7023 7024 7025 7026 7027 7028 7029 7030
        inf_byte = (end_key->flag == HA_READ_BEFORE_KEY) ?
            COL_NEG_INF : COL_POS_INF;
        pack_key(
            &key, 
            keynr, 
            key_buff, 
            end_key->key, 
            end_key->length, 
            inf_byte
            );
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7031
        error = kfile->key_range64(
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7032 7033
            kfile, 
            transaction, 
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7034
            &key,
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7035 7036 7037 7038
            &less,
            &equal,
            &greater,
            &is_exact
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7039 7040 7041 7042 7043
            );
        if (error) {
            ret_val = HA_TOKUDB_RANGE_COUNT;
            goto cleanup;
        }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7044
        end_rows= less;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7045 7046
    }
    else {
7047 7048 7049 7050 7051
        //
        // first if-clause ensures that start_key is non-NULL
        //
        assert(start_key);
        end_rows = total_rows_estimate;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7052 7053
    }

Zardosht Kasheff's avatar
Zardosht Kasheff committed
7054
    rows = (end_rows > start_rows) ? end_rows - start_rows : 1;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7055

Zardosht Kasheff's avatar
Zardosht Kasheff committed
7056
    //
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7057 7058
    // MySQL thinks a return value of 0 means there are exactly 0 rows
    // Therefore, always return non-zero so this assumption is not made
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7059
    //
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7060
    ret_val = (ha_rows) (rows <= 1 ? 1 : rows);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7061
cleanup:
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7062
    DBUG_RETURN(ret_val);
7063 7064
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
7065 7066

//
7067 7068 7069
// Initializes the auto-increment data in the local "share" object to the
// greater of two values: what's stored in the metadata or the last inserted
// auto-increment field (if auto-increment field is the first field of a key).
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7070 7071 7072 7073 7074 7075 7076 7077 7078 7079
//
void ha_tokudb::init_auto_increment() {
    DBT key;
    DBT value;
    int error;
    HA_METADATA_KEY key_val = hatoku_max_ai;
    bzero(&key, sizeof(key));
    bzero(&value, sizeof(value));
    key.data = &key_val;
    key.size = sizeof(key_val);
7080
    value.flags = DB_DBT_USERMEM;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7081 7082 7083 7084 7085 7086 7087 7088 7089 7090 7091 7092
    DB_TXN* txn = NULL;

    error = db_env->txn_begin(db_env, 0, &txn, 0);
    if (error) {
        share->last_auto_increment = 0;    
    }
    else {
        //
        // First retrieve hatoku_max_ai, which is max value used by auto increment
        // column so far, the max value could have been auto generated (e.g. insert (NULL))
        // or it could have been manually inserted by user (e.g. insert (345))
        //
7093 7094
        value.ulen = sizeof(share->last_auto_increment);
        value.data = &share->last_auto_increment;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7095 7096 7097 7098 7099 7100 7101 7102
        error = share->status_block->get(
            share->status_block, 
            txn, 
            &key, 
            &value, 
            0
            );
        
7103
        if (error || value.size != sizeof(share->last_auto_increment)) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7104 7105
            share->last_auto_increment = 0;
        }
7106

Zardosht Kasheff's avatar
Zardosht Kasheff committed
7107 7108 7109 7110 7111 7112
        //
        // Now retrieve the initial auto increment value, as specified by create table
        // so if a user does "create table t1 (a int auto_increment, primary key (a)) auto_increment=100",
        // then the value 100 should be stored here
        //
        key_val = hatoku_ai_create_value;
7113 7114
        value.ulen = sizeof(share->auto_inc_create_value);
        value.data = &share->auto_inc_create_value;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7115 7116 7117 7118 7119 7120 7121 7122
        error = share->status_block->get(
            share->status_block, 
            txn, 
            &key, 
            &value, 
            0
            );
        
7123
        if (error || value.size != sizeof(share->auto_inc_create_value)) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7124 7125 7126
            share->auto_inc_create_value = 0;
        }

7127
        commit_txn(txn, 0);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7128 7129 7130 7131 7132 7133
    }
    if (tokudb_debug & TOKUDB_DEBUG_AUTO_INCREMENT) {
        TOKUDB_TRACE("init auto increment:%lld\n", share->last_auto_increment);
    }
}

7134
void ha_tokudb::get_auto_increment(ulonglong offset, ulonglong increment, ulonglong nb_desired_values, ulonglong * first_value, ulonglong * nb_reserved_values) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7135
    TOKUDB_DBUG_ENTER("ha_tokudb::get_auto_increment");
7136 7137
    ulonglong nr;

7138
    pthread_mutex_lock(&share->mutex);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7139 7140 7141 7142

    if (share->auto_inc_create_value > share->last_auto_increment) {
        nr = share->auto_inc_create_value;
        share->last_auto_increment = share->auto_inc_create_value;
7143
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7144 7145 7146 7147
    else {
        nr = share->last_auto_increment + increment;
    }
    share->last_auto_increment = nr + (nb_desired_values - 1)*increment;
7148 7149
    if (delay_updating_ai_metadata) {
        ai_metadata_update_required = true;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7150 7151 7152 7153
    }
    else {
        update_max_auto_inc(share->status_block, share->last_auto_increment);
    }
7154

Zardosht Kasheff's avatar
Zardosht Kasheff committed
7155
    if (tokudb_debug & TOKUDB_DEBUG_AUTO_INCREMENT) {
7156 7157
        TOKUDB_TRACE("get_auto_increment(%lld,%lld,%lld):got:%lld:%lld\n",
                     offset, increment, nb_desired_values, nr, nb_desired_values);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7158
    }
7159
    *first_value = nr;
7160
    *nb_reserved_values = nb_desired_values;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7161
    pthread_mutex_unlock(&share->mutex);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7162
    DBUG_VOID_RETURN;
7163 7164
}

7165 7166 7167 7168
bool ha_tokudb::is_optimize_blocking() {
    return false;
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
7169 7170 7171 7172
bool ha_tokudb::is_auto_inc_singleton(){
    return false;
}

7173 7174
volatile int ha_tokudb_tokudb_add_index_wait = 0; // debug

7175 7176 7177 7178
//
// Internal function called by ha_tokudb::add_index and ha_tokudb::alter_table_phase2
// With a transaction, drops dictionaries associated with indexes in key_num
//
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7179 7180 7181 7182 7183 7184 7185 7186 7187 7188 7189 7190
//
// Adds indexes to the table. Takes the array of KEY passed in key_info, and creates
// DB's that will go at the end of share->key_file. THE IMPLICIT ASSUMPTION HERE is
// that the table will be modified and that these added keys will be appended to the end
// of the array table->key_info
// Parameters:
//      [in]    table_arg - table that is being modified, seems to be identical to this->table
//      [in]    key_info - array of KEY's to be added
//              num_of_keys - number of keys to be added, number of elements in key_info
//  Returns:
//      0 on success, error otherwise
//
7191 7192 7193 7194 7195 7196 7197 7198 7199 7200
int ha_tokudb::tokudb_add_index(
    TABLE *table_arg, 
    KEY *key_info, 
    uint num_of_keys, 
    DB_TXN* txn, 
    bool* inc_num_DBs,
    bool* modified_DBs
    ) 
{
    TOKUDB_DBUG_ENTER("ha_tokudb::tokudb_add_index");
7201 7202 7203
    
    while (ha_tokudb_tokudb_add_index_wait) sleep(1); // debug

Zardosht Kasheff's avatar
Zardosht Kasheff committed
7204 7205 7206 7207
    int error;
    uint curr_index = 0;
    DBC* tmp_cursor = NULL;
    int cursor_ret_val = 0;
7208
    DBT curr_pk_key, curr_pk_val;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7209
    THD* thd = ha_thd(); 
7210
    DB_LOADER* loader = NULL;
7211
    DB_INDEXER* indexer = NULL;
7212
    bool loader_use_puts = get_load_save_space(thd);
7213
    bool use_hot_index = (lock.type == TL_WRITE_ALLOW_WRITE);
7214
    u_int32_t loader_flags = loader_use_puts ? LOADER_USE_PUTS : 0;
7215 7216
    u_int32_t indexer_flags = 0;
    u_int32_t mult_db_flags[MAX_KEY + 1] = {0};
7217 7218
    u_int32_t mult_put_flags[MAX_KEY + 1];
    u_int32_t mult_dbt_flags[MAX_KEY + 1];
7219
    bool creating_hot_index = false;
7220 7221
    struct loader_context lc;
    memset(&lc, 0, sizeof lc);
7222 7223 7224
    lc.thd = thd;
    lc.ha = this;
    loader_error = 0;
7225
    bool rw_lock_taken = false;
7226 7227
    *inc_num_DBs = false;
    *modified_DBs = false;
7228 7229
    invalidate_bulk_fetch();
    unpack_entire_row = true; // for bulk fetching rows
7230
    for (u_int32_t i = 0; i < MAX_KEY+1; i++) {
7231
        mult_put_flags[i] = 0;
7232 7233
        mult_dbt_flags[i] = DB_DBT_REALLOC;
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7234 7235 7236 7237
    //
    // number of DB files we have open currently, before add_index is executed
    //
    uint curr_num_DBs = table_arg->s->keys + test(hidden_primary_key);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7238

Zardosht Kasheff's avatar
Zardosht Kasheff committed
7239 7240 7241 7242 7243 7244 7245
    //
    // status message to be shown in "show process list"
    //
    char status_msg[MAX_ALIAS_NAME + 200]; //buffer of 200 should be a good upper bound.
    ulonglong num_processed = 0; //variable that stores number of elements inserted thus far
    thd_proc_info(thd, "Adding indexes");

Zardosht Kasheff's avatar
Zardosht Kasheff committed
7246 7247 7248 7249 7250
    
    //
    // in unpack_row, MySQL passes a buffer that is this long,
    // so this length should be good enough for us as well
    //
7251 7252
    bzero((void *) &curr_pk_key, sizeof(curr_pk_key));
    bzero((void *) &curr_pk_val, sizeof(curr_pk_val));
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7253 7254 7255 7256 7257 7258 7259 7260 7261 7262 7263 7264 7265 7266 7267 7268

    //
    // The files for secondary tables are derived from the name of keys
    // If we try to add a key with the same name as an already existing key,
    // We can crash. So here we check if any of the keys added has the same
    // name of an existing key, and if so, we fail gracefully
    //
    for (uint i = 0; i < num_of_keys; i++) {
        for (uint j = 0; j < table_arg->s->keys; j++) {
            if (strcmp(key_info[i].name, table_arg->s->key_info[j].name) == 0) {
                error = HA_ERR_WRONG_COMMAND;
                goto cleanup;
            }
        }
    }
    
7269 7270
    rw_wrlock(&share->num_DBs_lock);
    rw_lock_taken = true;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7271 7272 7273 7274
    //
    // open all the DB files and set the appropriate variables in share
    // they go to the end of share->key_file
    //
7275
    creating_hot_index = use_hot_index && num_of_keys == 1 && (key_info[0].flags & HA_NOSAME) == 0;
7276 7277 7278 7279 7280 7281 7282
    if (use_hot_index && (share->num_DBs > curr_num_DBs)) {
        //
        // already have hot index in progress, get out
        //
        error = HA_ERR_INTERNAL_ERROR;
        goto cleanup;
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7283
    curr_index = curr_num_DBs;
7284
    *modified_DBs = true;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7285
    for (uint i = 0; i < num_of_keys; i++, curr_index++) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7286 7287
        if (key_info[i].flags & HA_CLUSTERING) {
            set_key_filter(
7288
                &share->kc_info.key_filters[curr_index],
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7289 7290 7291 7292 7293 7294
                &key_info[i],
                table_arg,
                false
                );                
            if (!hidden_primary_key) {
                set_key_filter(
7295
                    &share->kc_info.key_filters[curr_index],
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7296 7297 7298 7299 7300 7301
                    &table_arg->key_info[primary_key],
                    table_arg,
                    false
                    );
            }

7302
            error = initialize_col_pack_info(&share->kc_info,table_arg->s,curr_index);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7303 7304 7305 7306 7307 7308
            if (error) {
                goto cleanup;
            }
        }


7309
        error = create_secondary_dictionary(share->table_name, table_arg, &key_info[i], txn, &share->kc_info, curr_index, creating_hot_index);
7310 7311
        if (error) { goto cleanup; }

7312
        error = open_secondary_dictionary(
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7313 7314 7315
            &share->key_file[curr_index], 
            &key_info[i],
            share->table_name,
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7316
            false,
7317
            txn
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7318 7319 7320 7321
            );
        if (error) { goto cleanup; }
    }
    
7322
    if (creating_hot_index) {
7323
        share->num_DBs++;
7324
        *inc_num_DBs = true;
7325 7326 7327 7328 7329 7330 7331 7332 7333 7334 7335
        error = db_env->create_indexer(
            db_env,
            txn,
            &indexer,
            share->file,
            num_of_keys,
            &share->key_file[curr_num_DBs],
            mult_db_flags,
            indexer_flags
            );
        if (error) { goto cleanup; }
7336

7337 7338
        error = indexer->set_poll_function(indexer, ai_poll_fun, &lc);
        if (error) { goto cleanup; }
7339

7340 7341
        error = indexer->set_error_callback(indexer, loader_ai_err_fun, &lc);
        if (error) { goto cleanup; }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7342

7343 7344 7345 7346 7347
        rw_unlock(&share->num_DBs_lock);
        rw_lock_taken = false;
        
        error = indexer->build(indexer);
        if (error) { goto cleanup; }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7348

7349 7350 7351
        error = indexer->close(indexer);
        if (error) { goto cleanup; }
        indexer = NULL;
7352
    }
7353 7354 7355
    else {
        rw_unlock(&share->num_DBs_lock);
        rw_lock_taken = false;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7356 7357
        prelocked_right_range_size = 0;
        prelocked_left_range_size = 0;
7358 7359 7360 7361 7362 7363 7364
        struct smart_dbt_bf_info bf_info;
        bf_info.ha = this;
        // you need the val if you have a clustering index and key_read is not 0;
        bf_info.direction = 1;
        bf_info.thd = ha_thd();
        bf_info.need_val = TRUE;

7365 7366 7367 7368 7369 7370 7371 7372 7373 7374 7375 7376
        error = db_env->create_loader(
            db_env, 
            txn, 
            &loader, 
            NULL, // no src_db needed
            num_of_keys, 
            &share->key_file[curr_num_DBs], 
            mult_put_flags,
            mult_dbt_flags,
            loader_flags
            );
        if (error) { goto cleanup; }
7377

7378 7379
        error = loader->set_poll_function(loader, poll_fun, &lc);
        if (error) { goto cleanup; }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7380

7381 7382 7383 7384 7385
        error = loader->set_error_callback(loader, loader_ai_err_fun, &lc);
        if (error) { goto cleanup; }
        //
        // scan primary table, create each secondary key, add to each DB
        //    
7386
        if ((error = share->file->cursor(share->file, txn, &tmp_cursor, DB_SERIALIZABLE))) {
7387
            tmp_cursor = NULL;             // Safety
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7388 7389
            goto cleanup;
        }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7390

7391 7392 7393 7394 7395
        //
        // grab some locks to make this go faster
        // first a global read lock on the main DB, because
        // we intend to scan the entire thing
        //
7396 7397 7398 7399 7400
        error = tmp_cursor->c_pre_acquire_range_lock(
            tmp_cursor,
            share->file->dbt_neg_infty(),
            share->file->dbt_pos_infty()
            );
7401
        if (error) { goto cleanup; }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7402

7403 7404 7405 7406 7407
        // set the bulk fetch iteration to its max so that adding an
        // index fills the bulk fetch buffer every time. we do not
        // want it to grow exponentially fast.
        rows_fetched_using_bulk_fetch = 0;
        bulk_fetch_iteration = HA_TOKU_BULK_FETCH_ITERATION_MAX;
7408
        cursor_ret_val = tmp_cursor->c_getf_next(tmp_cursor, DB_PRELOCKED,smart_dbt_bf_callback, &bf_info);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7409

7410 7411
        while (cursor_ret_val != DB_NOTFOUND || ((bytes_used_in_range_query_buff - curr_range_query_buff_offset) > 0)) {
            if ((bytes_used_in_range_query_buff - curr_range_query_buff_offset) == 0) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7412
                invalidate_bulk_fetch(); // reset the buffers
7413 7414 7415 7416 7417
                cursor_ret_val = tmp_cursor->c_getf_next(tmp_cursor, DB_PRELOCKED, smart_dbt_bf_callback, &bf_info);
                if (cursor_ret_val != DB_NOTFOUND && cursor_ret_val != 0) {
                    error = cursor_ret_val;
                    goto cleanup;
                }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7418
            }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7419 7420
            // do this check in case the the c_getf_next did not put anything into the buffer because
            // there was no more data
7421 7422 7423
            if ((bytes_used_in_range_query_buff - curr_range_query_buff_offset) == 0) {
                break;
            }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7424
            // at this point, we know the range query buffer has at least one key/val pair
7425 7426 7427 7428 7429 7430 7431 7432 7433 7434 7435 7436 7437 7438 7439 7440 7441
            uchar* curr_pos = range_query_buff+curr_range_query_buff_offset;
            
            u_int32_t key_size = *(u_int32_t *)curr_pos;    
            curr_pos += sizeof(key_size);    
            uchar* curr_key_buff = curr_pos;    
            curr_pos += key_size;        
            curr_pk_key.data = curr_key_buff;    
            curr_pk_key.size = key_size;
            
            u_int32_t val_size = *(u_int32_t *)curr_pos;    
            curr_pos += sizeof(val_size);    
            uchar* curr_val_buff = curr_pos;    
            curr_pos += val_size;        
            curr_pk_val.data = curr_val_buff;    
            curr_pk_val.size = val_size;
            
            curr_range_query_buff_offset = curr_pos - range_query_buff;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7442

7443 7444
            error = loader->put(loader, &curr_pk_key, &curr_pk_val);
            if (error) { goto cleanup; }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7445

7446 7447 7448 7449 7450 7451 7452 7453 7454 7455 7456 7457 7458 7459 7460 7461 7462 7463 7464 7465 7466 7467 7468 7469
            num_processed++; 

            if ((num_processed % 1000) == 0) {
                if (loader_use_puts) {
                    sprintf(status_msg, "Adding indexes: Processed %llu of about %llu rows.", num_processed, (long long unsigned) share->rows);
                }
                else {
                    sprintf(status_msg, "Adding indexes: Fetched %llu of about %llu rows, loading of data still remains.", num_processed, (long long unsigned) share->rows);
                }
                thd_proc_info(thd, status_msg);
                if (thd->killed) {
                    error = ER_ABORTING_CONNECTION;
                    goto cleanup;
                }
            }
        }
        error = tmp_cursor->c_close(tmp_cursor);
        assert(error==0);
        tmp_cursor = NULL;

        error = loader->close(loader);
        loader = NULL;
        if (error) goto cleanup;
    }
7470 7471 7472 7473 7474 7475 7476 7477 7478 7479 7480 7481 7482 7483
    curr_index = curr_num_DBs;
    for (uint i = 0; i < num_of_keys; i++, curr_index++) {
        if (key_info[i].flags & HA_NOSAME) {
            bool is_unique;
            error = is_index_unique(
                &is_unique, 
                txn, 
                share->key_file[curr_index], 
                &key_info[i]
                );
            if (error) goto cleanup;
            if (!is_unique) {
                error = HA_ERR_FOUND_DUPP_KEY;
                last_dup_key = i;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7484 7485 7486 7487 7488
                goto cleanup;
            }
        }
    }

7489 7490 7491 7492 7493 7494
    //
    // We have an accurate row count, might as well update share->rows
    //
    pthread_mutex_lock(&share->mutex);
    share->rows = num_processed;
    pthread_mutex_unlock(&share->mutex);
7495 7496 7497 7498 7499 7500 7501 7502 7503

    //
    // now write stuff to status.tokudb
    //
    pthread_mutex_lock(&share->mutex);
    for (uint i = 0; i < num_of_keys; i++) {
        write_key_name_to_status(share->status_block, key_info[i].name, txn);
    }
    pthread_mutex_unlock(&share->mutex);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7504 7505 7506
    
    error = 0;
cleanup:
7507 7508 7509 7510
    if (rw_lock_taken) {
        rw_unlock(&share->num_DBs_lock);
        rw_lock_taken = false;
    }
7511
    if (tmp_cursor) {            
7512 7513
        int r = tmp_cursor->c_close(tmp_cursor);
        assert(r==0);
7514 7515
        tmp_cursor = NULL;
    }
7516
    if (loader != NULL) {
7517 7518
        sprintf(status_msg, "aborting creation of indexes.");
        thd_proc_info(thd, status_msg);
7519 7520
        loader->abort(loader);
    }
7521 7522 7523 7524 7525
    if (indexer != NULL) {
        sprintf(status_msg, "aborting creation of indexes.");
        thd_proc_info(thd, status_msg);
        indexer->abort(indexer);
    }
7526 7527
    if (error == DB_LOCK_NOTGRANTED && ((tokudb_debug & TOKUDB_DEBUG_HIDE_DDL_LOCK_ERRORS) == 0)) {
        sql_print_error("Could not add indexes to table %s because \
7528 7529
another transaction has accessed the table. \
To add indexes, make sure no transactions touch the table.", share->table_name);
7530
    }
7531
    TOKUDB_DBUG_RETURN(error ? error : loader_error);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7532 7533
}

7534 7535 7536 7537
//
// Internal function called by ha_tokudb::add_index and ha_tokudb::alter_table_phase2
// Closes added indexes in case of error in error path of add_index and alter_table_phase2
//
7538 7539 7540 7541 7542 7543 7544 7545 7546 7547 7548 7549 7550 7551 7552 7553 7554 7555 7556 7557 7558 7559 7560 7561 7562 7563 7564 7565 7566 7567 7568 7569 7570
void ha_tokudb::restore_add_index(TABLE* table_arg, uint num_of_keys, bool incremented_numDBs, bool modified_DBs) {
    uint curr_num_DBs = table_arg->s->keys + test(hidden_primary_key);
    uint curr_index = 0;

    //
    // need to restore num_DBs, and we have to do it before we close the dictionaries
    // so that there is not a window 
    //
    if (incremented_numDBs) {
        rw_wrlock(&share->num_DBs_lock);
        share->num_DBs--;
    }
    if (modified_DBs) {
        curr_index = curr_num_DBs;
        for (uint i = 0; i < num_of_keys; i++, curr_index++) {
            reset_key_and_col_info(&share->kc_info, curr_index);
        }
        curr_index = curr_num_DBs;
        for (uint i = 0; i < num_of_keys; i++, curr_index++) {
            if (share->key_file[curr_index]) {
                int r = share->key_file[curr_index]->close(
                    share->key_file[curr_index],
                    0
                    );
                assert(r==0);
                share->key_file[curr_index] = NULL;
            }
        }
    }
    if (incremented_numDBs) {
        rw_unlock(&share->num_DBs_lock);
    }
}
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7571

7572 7573 7574 7575 7576 7577 7578
int ha_tokudb::add_index(TABLE *table_arg, KEY *key_info, uint num_of_keys) {
    TOKUDB_DBUG_ENTER("ha_tokudb::add_index");
    DB_TXN* txn = NULL;
    int error;
    bool incremented_numDBs = false;
    bool modified_DBs = false;
    
7579 7580
    error = db_env->txn_begin(db_env, 0, &txn, 0);
    if (error) { goto cleanup; }
7581 7582 7583 7584 7585 7586 7587 7588 7589 7590 7591 7592 7593

    error = tokudb_add_index(
        table_arg,
        key_info,
        num_of_keys,
        txn,
        &incremented_numDBs,
        &modified_DBs
        );
    if (error) { goto cleanup; }
    
cleanup:
    if (error) {
7594 7595 7596 7597
        if (txn) {
            restore_add_index(table_arg, num_of_keys, incremented_numDBs, modified_DBs);
            abort_txn(txn);
        }
7598 7599 7600 7601 7602 7603 7604
    }
    else {
      commit_txn(txn, 0);
    }
    TOKUDB_DBUG_RETURN(error);
}

7605 7606
volatile int ha_tokudb_drop_indexes_wait = 0; // debug

7607 7608 7609 7610
//
// Internal function called by ha_tokudb::prepare_drop_index and ha_tokudb::alter_table_phase2
// With a transaction, drops dictionaries associated with indexes in key_num
//
7611 7612
int ha_tokudb::drop_indexes(TABLE *table_arg, uint *key_num, uint num_of_keys, DB_TXN* txn) {
    TOKUDB_DBUG_ENTER("ha_tokudb::drop_indexes");
7613

7614 7615 7616
    // XXX 4530 lock the key file lock for writing.
    share_key_file_wrlock(share);

7617 7618
    while (ha_tokudb_drop_indexes_wait) sleep(1); // debug

7619
    int error = 0;
7620 7621 7622 7623 7624 7625 7626
    for (uint i = 0; i < num_of_keys; i++) {
        uint curr_index = key_num[i];
        error = share->key_file[curr_index]->pre_acquire_fileops_lock(share->key_file[curr_index],txn);
        if (error != 0) {
            goto cleanup;
        }
    }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7627 7628
    for (uint i = 0; i < num_of_keys; i++) {
        uint curr_index = key_num[i];
7629 7630
        int r = share->key_file[curr_index]->close(share->key_file[curr_index],0);
        assert(r==0);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7631 7632
        share->key_file[curr_index] = NULL;

7633 7634 7635 7636 7637
        error = remove_key_name_from_status(share->status_block, table_arg->key_info[curr_index].name, txn);
        if (error) { goto cleanup; }
        
        error = delete_or_rename_dictionary(share->table_name, NULL, table_arg->key_info[curr_index].name, true, txn, true);
        if (error) { goto cleanup; }
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7638
    }
7639

Zardosht Kasheff's avatar
Zardosht Kasheff committed
7640
cleanup:
7641 7642
    if (error == DB_LOCK_NOTGRANTED && ((tokudb_debug & TOKUDB_DEBUG_HIDE_DDL_LOCK_ERRORS) == 0)) {
        sql_print_error("Could not drop indexes from table %s because \
7643 7644
another transaction has accessed the table. \
To drop indexes, make sure no transactions touch the table.", share->table_name);
7645
    }
7646 7647
    // XXX 4530 unlock the key file lock
    share_key_file_unlock(share);
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7648 7649 7650
    TOKUDB_DBUG_RETURN(error);
}

7651 7652 7653 7654
//
// Internal function called by ha_tokudb::prepare_drop_index and ha_tokudb::alter_table_phase2
// Restores dropped indexes in case of error in error path of prepare_drop_index and alter_table_phase2
//
7655
void ha_tokudb::restore_drop_indexes(TABLE *table_arg, uint *key_num, uint num_of_keys) {
7656 7657 7658
    // XXX 4530 lock the key file lock for writing
    share_key_file_wrlock(share);

7659 7660 7661 7662 7663 7664 7665 7666 7667 7668 7669 7670 7671 7672 7673 7674 7675
    //
    // reopen closed dictionaries
    //
    for (uint i = 0; i < num_of_keys; i++) {
        int r;
        uint curr_index = key_num[i];
        if (share->key_file[curr_index] == NULL) {
            r = open_secondary_dictionary(
                &share->key_file[curr_index], 
                &table_share->key_info[curr_index],
                share->table_name,
                false, // 
                NULL
                );
            assert(!r);
        }
    }            
7676 7677
    // XXX 4530 unlock the key file lock
    share_key_file_unlock(share);
7678
}
7679 7680 7681

volatile int ha_tokudb_prepare_drop_index_wait = 0; //debug

7682 7683 7684 7685 7686 7687 7688 7689 7690 7691 7692 7693 7694 7695 7696 7697 7698 7699 7700
//
// Prepares to drop indexes to the table. For each value, i, in the array key_num,
// table->key_info[i] is a key that is to be dropped.
//  ***********NOTE*******************
// Although prepare_drop_index is supposed to just get the DB's ready for removal,
// and not actually do the removal, we are doing it here and not in final_drop_index
// For the flags we expose in alter_table_flags, namely xxx_NO_WRITES, this is allowed
// Changes for "future-proofing" this so that it works when we have the equivalent flags
// that are not NO_WRITES are not worth it at the moments
// Parameters:
//      [in]    table_arg - table that is being modified, seems to be identical to this->table
//      [in]    key_num - array of indexes that specify which keys of the array table->key_info
//                  are to be dropped
//              num_of_keys - size of array, key_num
//  Returns:
//      0 on success, error otherwise
//
int ha_tokudb::prepare_drop_index(TABLE *table_arg, uint *key_num, uint num_of_keys) {
    TOKUDB_DBUG_ENTER("ha_tokudb::prepare_drop_index");
7701 7702 7703

    while (ha_tokudb_prepare_drop_index_wait) sleep(1); // debug

7704 7705 7706 7707 7708 7709 7710 7711 7712 7713 7714 7715 7716 7717 7718 7719 7720 7721 7722 7723 7724 7725
    int error;
    DB_TXN* txn = NULL;

    error = db_env->txn_begin(db_env, 0, &txn, 0);
    if (error) { goto cleanup; }
    
    error = drop_indexes(table_arg, key_num, num_of_keys, txn);
    if (error) { goto cleanup; }

cleanup:
    if (txn) {
        if (error) {
            abort_txn(txn);
            restore_drop_indexes(table_arg, key_num, num_of_keys);
        }
        else {
            commit_txn(txn,0);
        }
    }
    TOKUDB_DBUG_RETURN(error);
}

7726
volatile int ha_tokudb_final_drop_index_wait = 0; // debug
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7727 7728 7729 7730 7731 7732 7733 7734 7735 7736

//  ***********NOTE*******************
// Although prepare_drop_index is supposed to just get the DB's ready for removal,
// and not actually do the removal, we are doing it here and not in final_drop_index
// For the flags we expose in alter_table_flags, namely xxx_NO_WRITES, this is allowed
// Changes for "future-proofing" this so that it works when we have the equivalent flags
// that are not NO_WRITES are not worth it at the moments, therefore, we can make
// this function just return
int ha_tokudb::final_drop_index(TABLE *table_arg) {
    TOKUDB_DBUG_ENTER("ha_tokudb::final_drop_index");
7737
    while (ha_tokudb_final_drop_index_wait) sleep(1); // debug
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7738 7739 7740
    TOKUDB_DBUG_RETURN(0);
}

7741
void ha_tokudb::print_error(int error, myf errflag) {
7742
    if (error == DB_LOCK_DEADLOCK)
7743
        error = HA_ERR_LOCK_DEADLOCK;
7744
    if (error == DB_LOCK_NOTGRANTED)
7745
        error = HA_ERR_LOCK_WAIT_TIMEOUT;
7746
#if defined(HA_ERR_DISK_FULL)
7747 7748 7749
    if (error == ENOSPC) {
        error = HA_ERR_DISK_FULL;
    }
7750
#endif
7751 7752 7753
    if (error == DB_KEYEXIST) {
        error = HA_ERR_FOUND_DUPP_KEY;
    }
7754
#if defined(HA_ALTER_ERROR)
7755 7756
    if (error == HA_ALTER_ERROR) {
        error = HA_ERR_UNSUPPORTED;
7757 7758
    }
#endif
7759 7760
    // TODO: should rename debug code to something better
    // just reusing this so that tests don' start complaining
7761
#if MYSQL_VERSION_ID < 50500
7762
    if ((tokudb_debug & TOKUDB_DEBUG_HIDE_DDL_LOCK_ERRORS) == 0) {
7763
        THD* thd = ha_thd();
7764 7765 7766
        if (get_log_client_errors(thd)) {
            sql_print_error("query \"%s\" returned handler error %d", thd->query_string.str, error);
        }
7767
    }
7768
#endif
7769 7770 7771
    handler::print_error(error, errflag);
}

Zardosht Kasheff's avatar
Zardosht Kasheff committed
7772
#if 0 // QQQ use default
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7773 7774 7775 7776
//
// This function will probably need to be redone from scratch
// if we ever choose to implement it
//
7777
int ha_tokudb::analyze(THD * thd, HA_CHECK_OPT * check_opt) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7778 7779 7780 7781 7782 7783 7784 7785 7786 7787 7788 7789 7790 7791 7792 7793 7794 7795 7796 7797 7798 7799 7800 7801 7802 7803 7804 7805 7806 7807 7808 7809 7810 7811 7812 7813 7814
    uint i;
    DB_BTREE_STAT *stat = 0;
    DB_TXN_STAT *txn_stat_ptr = 0;
    tokudb_trx_data *trx = (tokudb_trx_data *) thd->ha_data[tokudb_hton->slot];
    DBUG_ASSERT(trx);

    for (i = 0; i < table_share->keys; i++) {
        if (stat) {
            free(stat);
            stat = 0;
        }
        if ((key_file[i]->stat) (key_file[i], trx->all, (void *) &stat, 0))
            goto err;
        share->rec_per_key[i] = (stat->bt_ndata / (stat->bt_nkeys ? stat->bt_nkeys : 1));
    }
    /* A hidden primary key is not in key_file[] */
    if (hidden_primary_key) {
        if (stat) {
            free(stat);
            stat = 0;
        }
        if ((file->stat) (file, trx->all, (void *) &stat, 0))
            goto err;
    }
    pthread_mutex_lock(&share->mutex);
    share->status |= STATUS_TOKUDB_ANALYZE;        // Save status on close
    share->version++;           // Update stat in table
    pthread_mutex_unlock(&share->mutex);
    update_status(share, table);        // Write status to file
    if (stat)
        free(stat);
    return ((share->status & STATUS_TOKUDB_ANALYZE) ? HA_ADMIN_FAILED : HA_ADMIN_OK);

  err:
    if (stat)
        free(stat);
    return HA_ADMIN_FAILED;
7815
}
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7816
#endif
7817

7818 7819
volatile int ha_tokudb_optimize_wait = 0; // debug

Zardosht Kasheff's avatar
Zardosht Kasheff committed
7820 7821 7822
//
// flatten all DB's in this table, to do so, just do a full scan on every DB
//
7823
int ha_tokudb::optimize(THD * thd, HA_CHECK_OPT * check_opt) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7824
    TOKUDB_DBUG_ENTER("ha_tokudb::optimize");
7825 7826 7827
    
    while (ha_tokudb_optimize_wait) sleep(1); // debug

Zardosht Kasheff's avatar
Zardosht Kasheff committed
7828 7829 7830
    int error;
    uint curr_num_DBs = table->s->keys + test(hidden_primary_key);
    //
7831
    // for each DB, run optimize and hot_optimize
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7832 7833
    //
    for (uint i = 0; i < curr_num_DBs; i++) {
7834 7835
        DB* db = share->key_file[i];
        error = db->optimize(db);
7836 7837 7838
        if (error) {
            goto cleanup;
        }
7839 7840 7841
        struct hot_optimize_context hc;
        memset(&hc, 0, sizeof hc);
        hc.thd = thd;
7842
        hc.write_status_msg = this->write_status_msg;
7843 7844 7845 7846
        hc.ha = this;
        hc.current_table = i;
        hc.num_tables = curr_num_DBs;
        error = db->hot_optimize(db, hot_poll_fun, &hc);
7847
        if (error) {
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7848 7849 7850 7851 7852 7853
            goto cleanup;
        }
    }

    error = 0;
cleanup:
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7854
    TOKUDB_DBUG_RETURN(error);
7855 7856
}

7857 7858 7859 7860 7861 7862 7863 7864 7865 7866
//
// truncate's dictionary associated with keynr index using transaction txn
// does so by deleting and then recreating the dictionary in the context
// of a transaction
//
int ha_tokudb::truncate_dictionary( uint keynr, DB_TXN* txn ) {
    int error;
    bool is_pk = (keynr == primary_key);
    
    error = share->key_file[keynr]->close(share->key_file[keynr], 0);
7867
    assert(error == 0);
7868 7869 7870 7871 7872 7873 7874 7875 7876

    share->key_file[keynr] = NULL;
    if (is_pk) { share->file = NULL; }

    if (is_pk) {
        error = delete_or_rename_dictionary(
            share->table_name, 
            NULL,
            "main", 
7877
            false, //is_key
7878 7879 7880 7881 7882 7883 7884 7885 7886 7887
            txn,
            true // is a delete
            );
        if (error) { goto cleanup; }
    }
    else {
        error = delete_or_rename_dictionary(
            share->table_name, 
            NULL,
            table_share->key_info[keynr].name, 
7888
            true, //is_key
7889 7890 7891 7892 7893 7894 7895
            txn,
            true // is a delete
            );
        if (error) { goto cleanup; }
    }

    if (is_pk) {
7896
        error = create_main_dictionary(share->table_name, table, txn, &share->kc_info);
7897 7898 7899 7900 7901 7902
    }
    else {
        error = create_secondary_dictionary(
            share->table_name, 
            table, 
            &table_share->key_info[keynr], 
7903 7904
            txn,
            &share->kc_info,
7905 7906
            keynr,
            false
7907 7908 7909 7910 7911 7912 7913 7914
            );
    }
    if (error) { goto cleanup; }

cleanup:
    return error;
}

7915 7916 7917 7918 7919 7920 7921 7922 7923

//
// for 5.5
//
int ha_tokudb::truncate() {
    return delete_all_rows();
}


7924 7925 7926 7927 7928 7929 7930 7931 7932 7933 7934 7935
// delete all rows from a table
//
// effects: delete all of the rows in the main dictionary and all of the
// indices.  this must be atomic, so we use the statement transaction
// for all of the truncate operations.
// locks:  if we have an exclusive table write lock, all of the concurrency
// issues go away.
// returns: 0 if success

int ha_tokudb::delete_all_rows() {
    TOKUDB_DBUG_ENTER("delete_all_rows");
    int error = 0;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7936
    uint curr_num_DBs = 0;
7937 7938
    DB_TXN* txn = NULL;

7939 7940 7941
    // XXX 4530 lock the key file array for writing
    share_key_file_wrlock(share);

7942 7943
    error = db_env->txn_begin(db_env, 0, &txn, 0);
    if (error) { goto cleanup; }
7944

7945
    if (thd_sql_command(ha_thd()) != SQLCOM_TRUNCATE) {
7946
        share->try_table_lock = true;
7947 7948 7949 7950
        error = HA_ERR_WRONG_COMMAND;
        goto cleanup;
    }

Zardosht Kasheff's avatar
Zardosht Kasheff committed
7951
    curr_num_DBs = table->s->keys + test(hidden_primary_key);
7952 7953 7954 7955 7956 7957
    for (uint i = 0; i < curr_num_DBs; i++) {
        error = share->key_file[i]->pre_acquire_fileops_lock(
            share->key_file[i], 
            txn
            );
        if (error) { goto cleanup; }
7958 7959 7960 7961 7962
        error = share->key_file[i]->pre_acquire_table_lock(
            share->key_file[i], 
            txn
            );
        if (error) { goto cleanup; }
7963
    }
7964
    for (uint i = 0; i < curr_num_DBs; i++) {
7965 7966
        error = truncate_dictionary(i, txn);
        if (error) { goto cleanup; }
7967 7968 7969
    }

    // zap the row count
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7970
    if (error == 0) {
7971
        share->rows = 0;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7972
    }
7973

7974
    share->try_table_lock = true;
Zardosht Kasheff's avatar
Zardosht Kasheff committed
7975
cleanup:
7976 7977 7978 7979 7980 7981 7982 7983 7984
    if (txn) {
        if (error) {
            abort_txn(txn);
        }
        else {
            commit_txn(txn,0);
        }
    }

7985 7986 7987 7988 7989
    if (error == DB_LOCK_NOTGRANTED && ((tokudb_debug & TOKUDB_DEBUG_HIDE_DDL_LOCK_ERRORS) == 0)) {
        sql_print_error("Could not truncate table %s because another transaction has accessed the \
        table. To truncate the table, make sure no transactions touch the table.", 
        share->table_name);
    }
7990 7991 7992 7993 7994 7995 7996 7997 7998 7999 8000
    //
    // regardless of errors, need to reopen the DB's
    //    
    for (uint i = 0; i < curr_num_DBs; i++) {
        int r = 0;
        if (share->key_file[i] == NULL) {
            if (i != primary_key) {
                r = open_secondary_dictionary(
                    &share->key_file[i], 
                    &table_share->key_info[i],
                    share->table_name,
Zardosht Kasheff's avatar
Zardosht Kasheff committed
8001
                    false, // 
8002 8003 8004 8005 8006 8007 8008
                    NULL
                    );
                assert(!r);
            }
            else {
                r = open_main_dictionary(
                    share->table_name, 
Zardosht Kasheff's avatar
Zardosht Kasheff committed
8009
                    false, 
8010 8011 8012 8013 8014 8015
                    NULL
                    );
                assert(!r);
            }
        }
    }
8016 8017
    // XXX 4530 unlock the key file lock
    share_key_file_unlock(share);
8018 8019 8020
    TOKUDB_DBUG_RETURN(error);
}

8021 8022 8023 8024 8025 8026 8027 8028
void ha_tokudb::set_loader_error(int err) {
    loader_error = err;
}
void ha_tokudb::set_dup_value_for_pk(DBT* key) {
    assert(!hidden_primary_key);
    unpack_key(table->record[0],key,primary_key);
    last_dup_key = primary_key;
}
8029

8030 8031 8032 8033 8034 8035
//
// MySQL sets the null_bit as a number that you can bit-wise AND a byte to
// to evaluate whether a field is NULL or not. This value is a power of 2, from
// 2^0 to 2^7. We return the position of the bit within the byte, which is
// lg null_bit
//
8036 8037 8038 8039 8040 8041 8042 8043 8044 8045 8046 8047 8048 8049 8050 8051 8052 8053 8054 8055 8056 8057 8058 8059 8060 8061 8062 8063 8064 8065 8066 8067 8068
inline u_int32_t get_null_bit_position(u_int32_t null_bit) {
    u_int32_t retval = 0;
    switch(null_bit) {
    case (1):
        retval = 0;
        break;
    case (2):
        retval = 1;
        break;
    case (4):
        retval = 2;
        break;
    case (8):
        retval = 3;
        break;
    case (16):
        retval = 4;
        break;
    case (32):
        retval = 5;
        break;
    case (64):
        retval = 6;
        break;
    case (128):
        retval = 7;
        break;        
    default:
        assert(false);
    }
    return retval;
}

8069 8070 8071
//
// checks whether the bit at index pos in data is set or not
//
8072 8073 8074 8075 8076 8077 8078
inline bool is_overall_null_position_set(uchar* data, u_int32_t pos) {
    u_int32_t offset = pos/8;
    uchar remainder = pos%8; 
    uchar null_bit = 1<<remainder;
    return ((data[offset] & null_bit) != 0);
}

8079 8080 8081
//
// sets the bit at index pos in data to 1 if is_null, 0 otherwise
// 
8082 8083 8084 8085 8086 8087 8088 8089 8090 8091 8092 8093
inline void set_overall_null_position(uchar* data, u_int32_t pos, bool is_null) {
    u_int32_t offset = pos/8;
    uchar remainder = pos%8;
    uchar null_bit = 1<<remainder;
    if (is_null) {
        data[offset] |= null_bit;
    }
    else {
        data[offset] &= ~null_bit;
    }
}

8094 8095 8096
//
// returns the index of the null bit of field. 
//
8097 8098 8099 8100 8101 8102 8103 8104 8105 8106 8107 8108 8109 8110 8111 8112 8113 8114 8115 8116 8117 8118 8119 8120 8121 8122 8123 8124 8125 8126 8127 8128 8129 8130 8131 8132 8133 8134 8135 8136 8137 8138 8139 8140 8141 8142 8143 8144 8145 8146 8147 8148 8149 8150 8151 8152 8153 8154 8155 8156 8157 8158 8159 8160 8161 8162 8163 8164 8165
inline u_int32_t get_overall_null_bit_position(TABLE* table, Field* field) {
    u_int32_t offset = get_null_offset(table, field);
    u_int32_t null_bit = field->null_bit;
    return offset*8 + get_null_bit_position(null_bit);
}


bool are_null_bits_in_order(TABLE* table) {
    u_int32_t curr_null_pos = 0;
    bool first = true;
    bool retval = true;
    for (uint i = 0; i < table->s->fields; i++) {
        Field* curr_field = table->field[i];
        bool nullable = (curr_field->null_bit != 0);
        if (nullable) {
            u_int32_t pos = get_overall_null_bit_position(
                table,
                curr_field
                );
            if (!first && pos != curr_null_pos+1){
                retval = false;
                break;
            }
            first = false;
            curr_null_pos = pos;
        }
    }
    return retval;
}

u_int32_t get_first_null_bit_pos(TABLE* table) {
    u_int32_t table_pos = 0;
    for (uint i = 0; i < table->s->fields; i++) {
        Field* curr_field = table->field[i];
        bool nullable = (curr_field->null_bit != 0);
        if (nullable) {
            table_pos = get_overall_null_bit_position(
                table,
                curr_field
                );
            break;
        }
    }
    return table_pos;
}

bool is_column_default_null(TABLE* src_table, u_int32_t field_index) {
    Field* curr_field = src_table->field[field_index];
    bool is_null_default = false;
    bool nullable = curr_field->null_bit != 0;
    if (nullable) {
        u_int32_t null_bit_position = get_overall_null_bit_position(src_table, curr_field);
        is_null_default = is_overall_null_position_set(
            src_table->s->default_values,
            null_bit_position
            );
    }
    return is_null_default;
}

bool columns_have_default_null_blobs(
    u_int32_t* changed_columns,
    u_int32_t num_changed_columns,
    TABLE* table 
) {
    bool retval = true;
    for (u_int32_t i = 0; i < num_changed_columns; i++) {
        Field* curr_field = table->field[changed_columns[i]];
        TOKU_TYPE field_type = mysql_to_toku_type (curr_field);
8166
        if (field_type == toku_type_blob && !is_column_default_null(table,changed_columns[i])) {
8167 8168 8169 8170 8171 8172 8173
            retval = false;
            break;
        }
    }
    return retval;
}

8174
bool tables_have_same_keys(TABLE* table, TABLE* altered_table, bool print_error, bool check_field_index) {
8175 8176 8177 8178 8179 8180 8181 8182 8183 8184 8185 8186 8187 8188 8189 8190 8191 8192 8193 8194 8195 8196 8197 8198 8199 8200 8201 8202 8203 8204 8205 8206 8207 8208 8209 8210 8211 8212 8213 8214 8215 8216 8217 8218 8219 8220 8221 8222 8223 8224 8225 8226 8227 8228 8229 8230 8231 8232 8233 8234 8235 8236 8237 8238 8239 8240 8241 8242 8243 8244 8245 8246 8247 8248 8249 8250 8251 8252 8253 8254 8255 8256 8257 8258 8259 8260
    bool retval;
    if (table->s->keys != altered_table->s->keys) {
        if (print_error) {
            sql_print_error("tables have different number of keys");
        }
        retval = false;
        goto cleanup;
    }
    if (table->s->primary_key != altered_table->s->primary_key) {
        if (print_error) {
            sql_print_error(
                "Tables have different primary keys, %d %d", 
                table->s->primary_key,
                altered_table->s->primary_key
                );
        }
        retval = false;
        goto cleanup;
    }
    for (u_int32_t i=0; i < table->s->keys; i++) {
        KEY* curr_orig_key = &table->key_info[i];
        KEY* curr_altered_key = &altered_table->key_info[i];
        if (strcmp(curr_orig_key->name, curr_altered_key->name)) {
            if (print_error) {
                sql_print_error(
                    "key %d has different name, %s %s", 
                    i, 
                    curr_orig_key->name,
                    curr_altered_key->name
                    );
            }
            retval = false;
            goto cleanup;
        }
        if (((curr_orig_key->flags & HA_CLUSTERING) == 0) != ((curr_altered_key->flags & HA_CLUSTERING) == 0)) {
            if (print_error) {
                sql_print_error(
                    "keys disagree on if they are clustering, %d, %d",
                    curr_orig_key->key_parts,
                    curr_altered_key->key_parts
                    );
            }
            retval = false;
            goto cleanup;
        }
        if (((curr_orig_key->flags & HA_NOSAME) == 0) != ((curr_altered_key->flags & HA_NOSAME) == 0)) {
            if (print_error) {
                sql_print_error(
                    "keys disagree on if they are unique, %d, %d",
                    curr_orig_key->key_parts,
                    curr_altered_key->key_parts
                    );
            }
            retval = false;
            goto cleanup;
        }
        if (curr_orig_key->key_parts != curr_altered_key->key_parts) {
            if (print_error) {
                sql_print_error(
                    "keys have different number of parts, %d, %d",
                    curr_orig_key->key_parts,
                    curr_altered_key->key_parts
                    );
            }
            retval = false;
            goto cleanup;
        }
        //
        // now verify that each field in the key is the same
        //
        for (u_int32_t j = 0; j < curr_orig_key->key_parts; j++) {
            KEY_PART_INFO* curr_orig_part = &curr_orig_key->key_part[j];
            KEY_PART_INFO* curr_altered_part = &curr_altered_key->key_part[j];
            Field* curr_orig_field = curr_orig_part->field;
            Field* curr_altered_field = curr_altered_part->field;
            if (curr_orig_part->length != curr_altered_part->length) {
                if (print_error) {
                    sql_print_error(
                        "Key %s has different length at index %d", 
                        curr_orig_key->name, 
                        j
                        );
                }
                retval = false;
                goto cleanup;
            }
8261 8262 8263 8264 8265 8266 8267
            bool are_fields_same;
            are_fields_same = (check_field_index) ? 
                (curr_orig_part->fieldnr == curr_altered_part->fieldnr && 
                 fields_are_same_type(curr_orig_field, curr_altered_field)) :
                (are_two_fields_same(curr_orig_field,curr_altered_field));
                
            if (!are_fields_same) {
8268 8269 8270 8271 8272 8273 8274 8275 8276 8277 8278 8279 8280 8281 8282 8283 8284 8285
                if (print_error) {
                    sql_print_error(
                        "Key %s has different field at index %d", 
                        curr_orig_key->name, 
                        j
                        );
                }
                retval = false;
                goto cleanup;
            }
        }
    }

    retval = true;
cleanup:
    return retval;
}

8286 8287
#if defined(HA_GENERAL_ONLINE)

8288 8289 8290 8291 8292 8293 8294
void ha_tokudb::print_alter_info(
    TABLE *altered_table,
    HA_CREATE_INFO *create_info,
    HA_ALTER_FLAGS *alter_flags,
    uint table_changes
    )
{
8295
    printf("***are keys of two tables same? %d\n", tables_have_same_keys(table,altered_table,false, false));
8296 8297 8298 8299 8300 8301 8302 8303 8304 8305 8306 8307 8308 8309 8310 8311 8312 8313 8314 8315 8316 8317 8318 8319 8320 8321 8322 8323 8324 8325 8326 8327 8328 8329 8330 8331 8332 8333 8334 8335 8336 8337 8338 8339 8340 8341 8342 8343 8344 8345 8346 8347 8348 8349 8350 8351 8352 8353 8354 8355 8356 8357 8358 8359 8360 8361
    printf("***alter flags set ***\n");
    for (uint i = 0; i < HA_MAX_ALTER_FLAGS; i++) {
      if (alter_flags->is_set(i)) {
        printf("flag: %d\n", i);
      }
    }
    //
    // everyone calculates data by doing some default_values - record[0], but I do not see why
    // that is necessary
    //
    printf("******\n");
    printf("***orig table***\n");
    for (uint i = 0; i < table->s->fields; i++) {
      //
      // make sure to use table->field, and NOT table->s->field
      //
      Field* curr_field = table->field[i];
      uint null_offset = get_null_offset(table, curr_field);
      printf(
          "name: %s, nullable: %d, null_offset: %d, is_null_field: %d, is_null %d, \n", 
          curr_field->field_name, 
          curr_field->null_bit,
          null_offset,
          (curr_field->null_ptr != NULL),
          (curr_field->null_ptr != NULL) ? table->s->default_values[null_offset] & curr_field->null_bit : 0xffffffff
          );
    }
    printf("******\n");
    printf("***altered table***\n");
    for (uint i = 0; i < altered_table->s->fields; i++) {
      Field* curr_field = altered_table->field[i];
      uint null_offset = get_null_offset(altered_table, curr_field);
      printf(
         "name: %s, nullable: %d, null_offset: %d, is_null_field: %d, is_null %d, \n", 
         curr_field->field_name, 
         curr_field->null_bit,
         null_offset,
         (curr_field->null_ptr != NULL),
         (curr_field->null_ptr != NULL) ? altered_table->s->default_values[null_offset] & curr_field->null_bit : 0xffffffff
         );
    }
    printf("******\n");
}


int find_changed_columns(
    u_int32_t* changed_columns,
    u_int32_t* num_changed_columns,
    TABLE* smaller_table, 
    TABLE* bigger_table
    ) 
{
    uint curr_new_col_index = 0;
    uint i = 0;
    int retval;
    u_int32_t curr_num_changed_columns=0;
    assert(bigger_table->s->fields > smaller_table->s->fields);
    for (i = 0; i < smaller_table->s->fields; i++, curr_new_col_index++) {
        if (curr_new_col_index >= bigger_table->s->fields) {
            sql_print_error("error in determining changed columns");
            retval = 1;
            goto cleanup;
        }
        Field* curr_field_in_new = bigger_table->field[curr_new_col_index];
        Field* curr_field_in_orig = smaller_table->field[i];
        while (!fields_have_same_name(curr_field_in_orig, curr_field_in_new)) {
8362 8363 8364 8365
            changed_columns[curr_num_changed_columns] = curr_new_col_index;
            curr_num_changed_columns++;
            curr_new_col_index++;
            curr_field_in_new = bigger_table->field[curr_new_col_index];
8366 8367 8368 8369 8370 8371 8372 8373 8374 8375 8376 8377 8378 8379 8380 8381 8382 8383 8384 8385 8386 8387 8388 8389 8390 8391 8392 8393 8394
            if (curr_new_col_index >= bigger_table->s->fields) {
                sql_print_error("error in determining changed columns");
                retval = 1;
                goto cleanup;
            }
        }
        // at this point, curr_field_in_orig and curr_field_in_new should be the same, let's verify
        // make sure the two fields that have the same name are ok
        if (!are_two_fields_same(curr_field_in_orig, curr_field_in_new)) {
            sql_print_error(
                "Two fields that were supposedly the same are not: \
                %s in original, %s in new", 
                curr_field_in_orig->field_name,
                curr_field_in_new->field_name
                );
            retval = 1;
            goto cleanup;
        }
    }
    for (i = curr_new_col_index; i < bigger_table->s->fields; i++) {
        changed_columns[curr_num_changed_columns] = i;
        curr_num_changed_columns++;
    }
    *num_changed_columns = curr_num_changed_columns;
    retval = 0;
cleanup:
    return retval;
}

8395 8396 8397 8398 8399 8400 8401 8402 8403 8404 8405 8406 8407 8408 8409 8410 8411 8412 8413 8414 8415 8416 8417 8418 8419 8420 8421 8422 8423 8424 8425 8426 8427 8428 8429 8430 8431 8432 8433 8434 8435 8436 8437 8438 8439 8440 8441 8442 8443 8444 8445 8446 8447 8448 8449 8450 8451 8452 8453
bool column_rename_supported(
    HA_ALTER_INFO* alter_info, 
    TABLE* orig_table, 
    TABLE* new_table
    ) 
{
    bool retval = false;
    bool keys_same_for_cr;
    uint num_fields_with_different_names = 0;
    uint field_with_different_name = orig_table->s->fields;
    if (orig_table->s->fields != new_table->s->fields) {
        retval = false;
        goto cleanup;
    }
    if (alter_info->contains_first_or_after) {
        retval = false;
        goto cleanup;
    }

    for (uint i = 0; i < orig_table->s->fields; i++) {
        Field* orig_field = orig_table->field[i];
        Field* new_field = new_table->field[i];
        if (!fields_are_same_type(orig_field, new_field)) {
            retval = false;
            goto cleanup;
        }
        if (!fields_have_same_name(orig_field, new_field)) {
            num_fields_with_different_names++;
            field_with_different_name = i;
        }
    }
    // only allow one renamed field
    if (num_fields_with_different_names != 1) {
        retval = false;
        goto cleanup;
    }
    assert(field_with_different_name < orig_table->s->fields);
    //
    // at this point, we have verified that the two tables have
    // the same field types and with ONLY one field with a different name. 
    // We have also identified the field with the different name
    //
    // Now we need to check the indexes
    //
    keys_same_for_cr = tables_have_same_keys(
        orig_table,
        new_table,
        false,
        true
        );
    if (!keys_same_for_cr) {
        retval = false;
        goto cleanup;
    }
    retval = true;
cleanup:
    return retval;
}

8454 8455 8456
int ha_tokudb::check_if_supported_alter(TABLE *altered_table,
    HA_CREATE_INFO *create_info,
    HA_ALTER_FLAGS *alter_flags,
8457
    HA_ALTER_INFO  *alter_info,
8458 8459 8460 8461 8462
    uint table_changes)
{
    TOKUDB_DBUG_ENTER("check_if_supported_alter");
    int retval;
    THD* thd = ha_thd(); 
8463
    bool keys_same = tables_have_same_keys(table,altered_table, false, false);
8464 8465 8466


    if (tokudb_debug & TOKUDB_DEBUG_ALTER_TABLE_INFO) {
8467
      printf("has after or first %d\n", alter_info->contains_first_or_after);
8468 8469 8470 8471
        print_alter_info(altered_table, create_info, alter_flags, table_changes);
    }
    bool has_added_columns = alter_flags->is_set(HA_ADD_COLUMN);
    bool has_dropped_columns = alter_flags->is_set(HA_DROP_COLUMN);
8472 8473
    bool has_column_rename = alter_flags->is_set(HA_CHANGE_COLUMN) && 
                             alter_flags->is_set(HA_ALTER_COLUMN_NAME);
8474 8475 8476 8477 8478
    //
    // We do not check for changes to foreign keys or primary keys. They are not supported
    // Changing the primary key implies changing keys in all dictionaries. that is why we don't
    // try to make it fast
    //
8479 8480 8481 8482 8483 8484 8485
    bool has_indexing_changes = alter_flags->is_set(HA_DROP_INDEX) || 
                                alter_flags->is_set(HA_DROP_UNIQUE_INDEX) ||
                                alter_flags->is_set(HA_ADD_INDEX) ||
                                alter_flags->is_set(HA_ADD_UNIQUE_INDEX);
    bool has_non_indexing_changes = false;
    bool has_non_dropped_changes = false;
    bool has_non_added_changes = false;
8486
    bool has_non_column_rename_changes = false;
8487 8488 8489 8490 8491 8492 8493 8494 8495 8496 8497 8498 8499
    for (uint i = 0; i < HA_MAX_ALTER_FLAGS; i++) {
        if (i == HA_DROP_INDEX ||
            i == HA_DROP_UNIQUE_INDEX ||
            i == HA_ADD_INDEX ||
            i == HA_ADD_UNIQUE_INDEX)
        {
            continue;
        }
        if (alter_flags->is_set(i)) {
            has_non_indexing_changes = true;
            break;
        }
    }
8500 8501 8502 8503 8504 8505 8506 8507 8508 8509 8510
    for (uint i = 0; i < HA_MAX_ALTER_FLAGS; i++) {
        if (i == HA_ALTER_COLUMN_NAME||
            i == HA_CHANGE_COLUMN)
        {
            continue;
        }
        if (alter_flags->is_set(i)) {
            has_non_column_rename_changes = true;
            break;
        }
    }
8511 8512 8513 8514 8515 8516 8517 8518 8519 8520 8521 8522 8523 8524 8525 8526 8527 8528 8529 8530 8531 8532 8533 8534 8535 8536 8537 8538 8539 8540
    for (uint i = 0; i < HA_MAX_ALTER_FLAGS; i++) {
        if (i == HA_DROP_COLUMN) {
            continue;
        }
        if (keys_same && 
            (i == HA_ALTER_INDEX || i == HA_ALTER_UNIQUE_INDEX || i == HA_ALTER_PK_INDEX)) {
            continue;
        }
        if (alter_flags->is_set(i)) {
            has_non_dropped_changes = true;
            break;
        }
    }
    for (uint i = 0; i < HA_MAX_ALTER_FLAGS; i++) {
        if (i == HA_ADD_COLUMN) {
            continue;
        }
        if (keys_same && 
            (i == HA_ALTER_INDEX || i == HA_ALTER_UNIQUE_INDEX || i == HA_ALTER_PK_INDEX)) {
            continue;
        }
        if (alter_flags->is_set(i)) {
            has_non_added_changes = true;
            break;
        }
    }

    if (tokudb_debug & TOKUDB_DEBUG_ALTER_TABLE_INFO) {
        printf("has indexing changes %d, has non indexing changes %d\n", has_indexing_changes, has_non_indexing_changes);
    }
8541 8542 8543 8544 8545 8546 8547 8548
#ifdef MARIADB_BASE_VERSION
#if MYSQL_VERSION_ID >= 50203
    if (table->s->vfields || altered_table->s->vfields) {
      retval = HA_ALTER_ERROR;
      goto cleanup;
    }
#endif
#endif
8549 8550 8551 8552 8553 8554 8555 8556 8557 8558 8559 8560 8561 8562 8563 8564 8565 8566 8567 8568 8569 8570 8571 8572 8573 8574 8575 8576 8577 8578 8579 8580 8581 8582 8583 8584 8585 8586 8587 8588 8589 8590 8591 8592 8593 8594 8595 8596 8597 8598 8599 8600 8601 8602 8603 8604 8605 8606 8607 8608 8609 8610 8611 8612 8613 8614 8615 8616 8617 8618 8619 8620 8621 8622 8623 8624 8625 8626 8627 8628 8629 8630 8631
    if (table->s->tmp_table != NO_TMP_TABLE) {
      retval = (get_disable_slow_alter(thd)) ? HA_ALTER_ERROR : HA_ALTER_NOT_SUPPORTED;
      goto cleanup;
    }
    if (!(are_null_bits_in_order(table) && 
          are_null_bits_in_order(altered_table)
          )
       ) 
    {
        sql_print_error("Problems parsing null bits of the original and altered table");
        retval = (get_disable_slow_alter(thd)) ? HA_ALTER_ERROR : HA_ALTER_NOT_SUPPORTED;
        goto cleanup;
    }
    if (has_added_columns && !has_non_added_changes) {
        u_int32_t added_columns[altered_table->s->fields];
        u_int32_t num_added_columns = 0;
        int r = find_changed_columns(
            added_columns,
            &num_added_columns,
            table,
            altered_table
            );
        if (r) {
            retval = (get_disable_slow_alter(thd)) ? HA_ALTER_ERROR : HA_ALTER_NOT_SUPPORTED;
            goto cleanup;
        }
        if (!columns_have_default_null_blobs(
            added_columns,
            num_added_columns,
            altered_table
            )) 
        {
            sql_print_error("unexpectedly, an added column has a non-null default");
            retval = HA_ALTER_ERROR;
            goto cleanup;
        }
        if (tokudb_debug & TOKUDB_DEBUG_ALTER_TABLE_INFO) {
            for (u_int32_t i = 0; i < num_added_columns; i++) {
                u_int32_t curr_added_index = added_columns[i];
                Field* curr_added_field = altered_table->field[curr_added_index];
                printf(
                    "Added column: index %d, name %s\n", 
                    curr_added_index, 
                    curr_added_field->field_name
                    );
            }
        }
    }
    if (has_dropped_columns && !has_non_dropped_changes) {
        u_int32_t dropped_columns[table->s->fields];
        u_int32_t num_dropped_columns = 0;
        int r = find_changed_columns(
            dropped_columns,
            &num_dropped_columns,
            altered_table,
            table
            );
        if (r) {
            retval = (get_disable_slow_alter(thd)) ? HA_ALTER_ERROR : HA_ALTER_NOT_SUPPORTED;
            goto cleanup;
        }
        if (tokudb_debug & TOKUDB_DEBUG_ALTER_TABLE_INFO) {
            for (u_int32_t i = 0; i < num_dropped_columns; i++) {
                u_int32_t curr_dropped_index = dropped_columns[i];
                Field* curr_dropped_field = table->field[curr_dropped_index];
                printf(
                    "Dropped column: index %d, name %s\n", 
                    curr_dropped_index, 
                    curr_dropped_field->field_name
                    );
            }
        }
    }
    
    if (has_indexing_changes && !has_non_indexing_changes) {
        retval = HA_ALTER_SUPPORTED_WAIT_LOCK;
    }
    else if (has_dropped_columns && !has_non_dropped_changes) {
        retval = HA_ALTER_SUPPORTED_WAIT_LOCK;
    }
    else if (has_added_columns && !has_non_added_changes) {
        retval = HA_ALTER_SUPPORTED_WAIT_LOCK;
    }
8632 8633 8634 8635 8636 8637 8638 8639 8640 8641 8642 8643 8644 8645 8646 8647 8648 8649 8650 8651 8652 8653 8654
    else if (has_column_rename && !has_non_column_rename_changes) {
        // we have identified a possible column rename, 
        // but let's do some more checks

        // we will only allow an hcr if there are no changes
        // in column positions
        if (alter_info->contains_first_or_after) {
            retval = (get_disable_slow_alter(thd)) ? HA_ALTER_ERROR : HA_ALTER_NOT_SUPPORTED;
            goto cleanup;
        }

        // now need to verify that one and only one column
        // has changed only its name. If we find anything to
        // the contrary, we don't allow it, also check indexes

        bool cr_supported = column_rename_supported(alter_info, table, altered_table);
        if (cr_supported) {
            retval = HA_ALTER_SUPPORTED_WAIT_LOCK;
        }
        else {
            retval = (get_disable_slow_alter(thd)) ? HA_ALTER_ERROR : HA_ALTER_NOT_SUPPORTED;
        }
    }
8655 8656 8657 8658 8659 8660 8661 8662 8663 8664 8665 8666 8667 8668 8669 8670 8671 8672 8673 8674 8675 8676 8677 8678 8679 8680 8681 8682 8683 8684 8685 8686 8687 8688 8689 8690 8691 8692 8693 8694 8695 8696 8697 8698 8699 8700 8701 8702 8703 8704 8705 8706 8707 8708 8709 8710 8711 8712 8713 8714 8715 8716 8717 8718 8719 8720 8721 8722 8723 8724 8725 8726 8727 8728 8729 8730 8731 8732 8733 8734 8735 8736 8737 8738 8739 8740 8741 8742 8743 8744 8745 8746 8747 8748 8749 8750 8751 8752 8753 8754 8755 8756 8757 8758 8759 8760 8761 8762 8763 8764 8765 8766 8767 8768 8769 8770 8771 8772 8773 8774 8775 8776 8777 8778 8779 8780 8781 8782 8783 8784 8785 8786 8787 8788 8789 8790 8791 8792 8793 8794 8795 8796 8797 8798 8799 8800 8801 8802 8803 8804 8805 8806 8807 8808 8809 8810 8811 8812 8813 8814 8815 8816 8817 8818 8819 8820 8821 8822 8823 8824 8825 8826 8827 8828 8829 8830 8831 8832 8833 8834 8835 8836 8837 8838 8839 8840 8841 8842 8843 8844 8845 8846 8847 8848 8849 8850 8851 8852 8853 8854 8855 8856 8857 8858 8859 8860 8861 8862 8863 8864 8865 8866 8867 8868 8869 8870 8871 8872 8873 8874 8875 8876 8877 8878 8879 8880 8881 8882 8883 8884 8885 8886 8887 8888 8889 8890 8891 8892 8893 8894 8895 8896 8897 8898 8899 8900 8901 8902 8903 8904 8905 8906 8907 8908 8909 8910 8911 8912 8913 8914 8915 8916 8917 8918 8919 8920 8921 8922 8923 8924 8925 8926 8927 8928 8929 8930 8931 8932 8933 8934 8935 8936 8937 8938 8939 8940 8941 8942 8943 8944 8945 8946 8947 8948 8949 8950 8951 8952 8953 8954 8955 8956 8957 8958 8959 8960 8961 8962 8963 8964 8965 8966 8967 8968 8969 8970 8971 8972 8973 8974 8975 8976 8977 8978 8979 8980 8981 8982 8983 8984 8985 8986 8987 8988 8989 8990 8991 8992 8993 8994 8995 8996 8997 8998 8999 9000 9001 9002 9003 9004 9005 9006 9007 9008 9009 9010 9011 9012 9013 9014 9015 9016 9017 9018 9019 9020 9021 9022 9023 9024 9025 9026 9027 9028 9029 9030 9031 9032 9033 9034 9035 9036 9037 9038 9039 9040 9041 9042 9043 9044 9045 9046 9047 9048 9049 9050 9051 9052 9053 9054 9055 9056 9057 9058 9059 9060 9061 9062 9063 9064 9065 9066 9067 9068 9069 9070 9071 9072 9073 9074 9075 9076 9077 9078 9079 9080 9081 9082 9083 9084 9085 9086 9087 9088 9089
    else { 
        retval = (get_disable_slow_alter(thd)) ? HA_ALTER_ERROR : HA_ALTER_NOT_SUPPORTED;
    }
cleanup:
    DBUG_RETURN(retval);
}

#define UP_COL_ADD_OR_DROP 0

#define COL_DROP 0xaa
#define COL_ADD 0xbb

#define COL_FIXED 0xcc
#define COL_VAR 0xdd
#define COL_BLOB 0xee



#define STATIC_ROW_MUTATOR_SIZE 1+8+2+8+8+8

/*
how much space do I need for the mutators?
static stuff first:
1 - UP_COL_ADD_OR_DROP
8 - old null, new null
2 - old num_offset, new num_offset
8 - old fixed_field size, new fixed_field_size
8 - old and new length of offsets
8 - old and new starting null bit position
TOTAL: 27

dynamic stuff:
4 - number of columns
for each column:
1 - add or drop
1 - is nullable
4 - if nullable, position
1 - if add, whether default is null or not
1 - if fixed, var, or not
 for fixed, entire default
 for var, 4 bytes length, then entire default
 for blob, nothing
So, an upperbound is 4 + num_fields(12) + all default stuff

static blob stuff:
4 - num blobs
1 byte for each num blobs in old table
So, an upperbound is 4 + kc_info->num_blobs

dynamic blob stuff:
for each blob added:
1 - state if we are adding or dropping
4 - blob index
if add, 1 len bytes
 at most, 4 0's
So, upperbound is num_blobs(1+4+1+4) = num_columns*10
*/
u_int32_t fill_static_row_mutator(
    uchar* buf, 
    TABLE* orig_table,
    TABLE* altered_table,
    KEY_AND_COL_INFO* orig_kc_info,
    KEY_AND_COL_INFO* altered_kc_info,
    u_int32_t keynr
    ) 
{
    //
    // start packing extra
    //
    uchar* pos = buf;
    // says what the operation is
    pos[0] = UP_COL_ADD_OR_DROP;
    pos++;
    
    //
    // null byte information
    //
    memcpy(pos, &orig_table->s->null_bytes, sizeof(orig_table->s->null_bytes));
    pos += sizeof(orig_table->s->null_bytes);
    memcpy(pos, &altered_table->s->null_bytes, sizeof(orig_table->s->null_bytes));
    pos += sizeof(altered_table->s->null_bytes);
    
    //
    // num_offset_bytes
    //
    assert(orig_kc_info->num_offset_bytes <= 2);
    pos[0] = orig_kc_info->num_offset_bytes;
    pos++;
    assert(altered_kc_info->num_offset_bytes <= 2);
    pos[0] = altered_kc_info->num_offset_bytes;
    pos++;
    
    //
    // size of fixed fields
    //
    u_int32_t fixed_field_size = orig_kc_info->mcp_info[keynr].fixed_field_size;
    memcpy(pos, &fixed_field_size, sizeof(fixed_field_size));
    pos += sizeof(fixed_field_size);
    fixed_field_size = altered_kc_info->mcp_info[keynr].fixed_field_size;
    memcpy(pos, &fixed_field_size, sizeof(fixed_field_size));
    pos += sizeof(fixed_field_size);
    
    //
    // length of offsets
    //
    u_int32_t len_of_offsets = orig_kc_info->mcp_info[keynr].len_of_offsets;
    memcpy(pos, &len_of_offsets, sizeof(len_of_offsets));
    pos += sizeof(len_of_offsets);
    len_of_offsets = altered_kc_info->mcp_info[keynr].len_of_offsets;
    memcpy(pos, &len_of_offsets, sizeof(len_of_offsets));
    pos += sizeof(len_of_offsets);

    u_int32_t orig_start_null_pos = get_first_null_bit_pos(orig_table);
    memcpy(pos, &orig_start_null_pos, sizeof(orig_start_null_pos));
    pos += sizeof(orig_start_null_pos);
    u_int32_t altered_start_null_pos = get_first_null_bit_pos(altered_table);
    memcpy(pos, &altered_start_null_pos, sizeof(altered_start_null_pos));
    pos += sizeof(altered_start_null_pos);

    assert((pos-buf) == STATIC_ROW_MUTATOR_SIZE);
    return pos - buf;
}


u_int32_t fill_dynamic_row_mutator(
    uchar* buf,
    u_int32_t* columns, 
    u_int32_t num_columns,
    TABLE* src_table,
    KEY_AND_COL_INFO* src_kc_info,
    u_int32_t keynr,
    bool is_add,
    bool* out_has_blobs
    ) 
{
    uchar* pos = buf;
    bool has_blobs = false;
    u_int32_t cols = num_columns;
    memcpy(pos, &cols, sizeof(cols));
    pos += sizeof(cols);
    for (u_int32_t i = 0; i < num_columns; i++) {
        u_int32_t curr_index = columns[i];
        Field* curr_field = src_table->field[curr_index];
    
        pos[0] = is_add ? COL_ADD : COL_DROP;
        pos++;
        //
        // NULL bit information
        //
        bool is_null_default = false;
        bool nullable = curr_field->null_bit != 0;
        if (!nullable) {
            pos[0] = 0;
            pos++;
        }
        else {
            pos[0] = 1;
            pos++;
            // write position of null byte that is to be removed
            u_int32_t null_bit_position = get_overall_null_bit_position(src_table, curr_field);
            memcpy(pos, &null_bit_position, sizeof(null_bit_position));
            pos += sizeof(null_bit_position);
            //
            // if adding a column, write the value of the default null_bit
            //
            if (is_add) {
                is_null_default = is_overall_null_position_set(
                    src_table->s->default_values,
                    null_bit_position
                    );
                pos[0] = is_null_default ? 1 : 0;
                pos++;
            }
        }
        if (src_kc_info->field_lengths[curr_index] != 0) {
            // we have a fixed field being dropped
            // store the offset and the number of bytes
            pos[0] = COL_FIXED;
            pos++;
            //store the offset
            u_int32_t fixed_field_offset = src_kc_info->cp_info[keynr][curr_index].col_pack_val;
            memcpy(pos, &fixed_field_offset, sizeof(fixed_field_offset));
            pos += sizeof(fixed_field_offset);
            //store the number of bytes
            u_int32_t num_bytes = src_kc_info->field_lengths[curr_index];
            memcpy(pos, &num_bytes, sizeof(num_bytes));
            pos += sizeof(num_bytes);
            if (is_add && !is_null_default) {
                uint curr_field_offset = field_offset(curr_field, src_table);
                memcpy(
                    pos, 
                    src_table->s->default_values + curr_field_offset, 
                    num_bytes
                    );
                pos += num_bytes;
            }
        }
        else if (src_kc_info->length_bytes[curr_index] != 0) {
            pos[0] = COL_VAR;
            pos++;
            //store the index of the variable column
            u_int32_t var_field_index = src_kc_info->cp_info[keynr][curr_index].col_pack_val;
            memcpy(pos, &var_field_index, sizeof(var_field_index));
            pos += sizeof(var_field_index);
            if (is_add && !is_null_default) {
                uint curr_field_offset = field_offset(curr_field, src_table);
                u_int32_t len_bytes = src_kc_info->length_bytes[curr_index];
                u_int32_t data_length = get_var_data_length(
                    src_table->s->default_values + curr_field_offset,
                    len_bytes
                    );
                memcpy(pos, &data_length, sizeof(data_length));
                pos += sizeof(data_length);
                memcpy(
                    pos, 
                    src_table->s->default_values + curr_field_offset + len_bytes,
                    data_length
                    );
                pos += data_length;
            }
        }
        else {
            pos[0] = COL_BLOB;
            pos++;
            has_blobs = true;
        }
    }
    *out_has_blobs = has_blobs;
    return pos-buf;
}


u_int32_t fill_static_blob_row_mutator(
    uchar* buf,
    TABLE* src_table,
    KEY_AND_COL_INFO* src_kc_info
    ) 
{
    uchar* pos = buf;
    // copy number of blobs
    memcpy(pos, &src_kc_info->num_blobs, sizeof(src_kc_info->num_blobs));
    pos += sizeof(src_kc_info->num_blobs);
    // copy length bytes for each blob
    for (u_int32_t i = 0; i < src_kc_info->num_blobs; i++) {
        u_int32_t curr_field_index = src_kc_info->blob_fields[i]; 
        Field* field = src_table->field[curr_field_index];
        u_int32_t len_bytes = field->row_pack_length();
        assert(len_bytes <= 4);
        pos[0] = len_bytes;
        pos++;
    }
    
    return pos-buf;
}

u_int32_t fill_dynamic_blob_row_mutator(
    uchar* buf,
    u_int32_t* columns, 
    u_int32_t num_columns,
    TABLE* src_table,
    KEY_AND_COL_INFO* src_kc_info,
    bool is_add
    ) 
{
    uchar* pos = buf;
    for (u_int32_t i = 0; i < num_columns; i++) {
        u_int32_t curr_field_index = columns[i];
        Field* curr_field = src_table->field[curr_field_index];
        if (src_kc_info->field_lengths[curr_field_index] == 0 && 
            src_kc_info->length_bytes[curr_field_index]== 0
            ) 
        {
            // find out which blob it is
            u_int32_t blob_index = src_kc_info->num_blobs;
            for (u_int32_t j = 0; j < src_kc_info->num_blobs; j++) {
                if (curr_field_index  == src_kc_info->blob_fields[j]) {
                    blob_index = j;
                    break;
                }
            }
            // assert we found blob in list
            assert(blob_index < src_kc_info->num_blobs);
            pos[0] = is_add ? COL_ADD : COL_DROP;
            pos++;
            memcpy(pos, &blob_index, sizeof(blob_index));
            pos += sizeof(blob_index);
            if (is_add) {
                bool is_null_default = is_column_default_null(
                    src_table,
                    curr_field_index
                    );

                u_int32_t len_bytes = curr_field->row_pack_length();
                assert(len_bytes <= 4);
                pos[0] = len_bytes;
                pos++;

                if (is_null_default) {
                    // create a zero length blob field that can be directly copied in
                    bzero(pos,len_bytes);
                    pos += len_bytes;
                }
                else {
                    // in future, if is_null_default can be 0, we will have a default value placed here
                    // for now, in MySQL, we can only have blob fields that are null by default
                    // in check_if_supported_alter, we verify that all blob fields have null by default,
                    // so, we can assert this here.
                    assert(is_null_default);
                }
            }
        }
        else {
            // not a blob, continue
            continue;
        }
    }
    return pos-buf;
}

// TODO: carefully review to make sure that the right information is used
// TODO: namely, when do we get stuff from share->kc_info and when we get
// TODO: it from altered_kc_info, and when is keynr associated with the right thing
u_int32_t ha_tokudb::fill_row_mutator(
    uchar* buf, 
    u_int32_t* columns, 
    u_int32_t num_columns,
    TABLE* altered_table,
    KEY_AND_COL_INFO* altered_kc_info,
    u_int32_t keynr,
    bool is_add
    ) 
{
    if (tokudb_debug & TOKUDB_DEBUG_ALTER_TABLE_INFO) {
        printf("*****some info:*************\n");
        printf(
            "old things: num_null_bytes %d, num_offset_bytes %d, fixed_field_size %d, fixed_field_size %d\n",
            table->s->null_bytes,
            share->kc_info.num_offset_bytes,
            share->kc_info.mcp_info[keynr].fixed_field_size,
            share->kc_info.mcp_info[keynr].len_of_offsets
            );
        printf(
            "new things: num_null_bytes %d, num_offset_bytes %d, fixed_field_size %d, fixed_field_size %d\n",
            altered_table->s->null_bytes,
            altered_kc_info->num_offset_bytes,
            altered_kc_info->mcp_info[keynr].fixed_field_size,
            altered_kc_info->mcp_info[keynr].len_of_offsets
            );
        printf("****************************\n");
    }
    uchar* pos = buf;
    bool has_blobs = false;
    pos += fill_static_row_mutator(
        pos,
        table,
        altered_table,
        &share->kc_info,
        altered_kc_info,
        keynr
        );
    
    if (is_add) {
        pos += fill_dynamic_row_mutator(
            pos,
            columns,
            num_columns,
            altered_table,
            altered_kc_info,
            keynr,
            is_add,
            &has_blobs
            );
    }
    else {
        pos += fill_dynamic_row_mutator(
            pos,
            columns,
            num_columns,
            table,
            &share->kc_info,
            keynr,
            is_add,
            &has_blobs
            );
    }
    if (has_blobs) {
        pos += fill_static_blob_row_mutator(
            pos,
            table,
            &share->kc_info
            );
        if (is_add) {
            pos += fill_dynamic_blob_row_mutator(
                pos,
                columns,
                num_columns,
                altered_table,
                altered_kc_info,
                is_add
                );
        }
        else {
            pos += fill_dynamic_blob_row_mutator(
                pos,
                columns,
                num_columns,
                table,
                &share->kc_info,
                is_add
                );
        }
    }
    return pos-buf;
}

int ha_tokudb::alter_table_phase2(
    THD *thd,
    TABLE *altered_table,
    HA_CREATE_INFO *create_info,
    HA_ALTER_INFO *alter_info,
    HA_ALTER_FLAGS *alter_flags
    )
{
    TOKUDB_DBUG_ENTER("ha_tokudb::alter_table_phase2");
    int error;
    DB_TXN* txn = NULL;
    bool incremented_numDBs = false;
    bool modified_DBs = false;
    bool has_dropped_columns = alter_flags->is_set(HA_DROP_COLUMN);
    bool has_added_columns = alter_flags->is_set(HA_ADD_COLUMN);
    KEY_AND_COL_INFO altered_kc_info;
    bzero(&altered_kc_info, sizeof(altered_kc_info));
    u_int32_t max_new_desc_size = 0;
    uchar* row_desc_buff = NULL;
    uchar* column_extra = NULL; 
9090 9091
    bool dropping_indexes = alter_info->index_drop_count > 0 && !tables_have_same_keys(table,altered_table,false, false);
    bool adding_indexes = alter_info->index_add_count > 0 && !tables_have_same_keys(table,altered_table,false, false);
9092 9093 9094 9095 9096 9097 9098 9099 9100 9101 9102 9103 9104 9105 9106 9107 9108 9109 9110 9111 9112 9113 9114 9115 9116 9117 9118 9119 9120 9121 9122 9123 9124 9125 9126 9127 9128 9129 9130 9131 9132 9133 9134 9135 9136 9137 9138 9139 9140 9141 9142 9143 9144 9145 9146 9147 9148 9149 9150 9151 9152 9153 9154 9155 9156 9157 9158 9159 9160 9161 9162 9163 9164 9165 9166 9167 9168 9169 9170 9171 9172 9173 9174
    tokudb_trx_data* trx = (tokudb_trx_data *) thd_data_get(thd, tokudb_hton->slot);

    is_fast_alter_running = true;

    if (!trx || 
        (trx->all != NULL) || 
        (trx->sp_level != NULL) ||
        (trx->stmt == NULL) ||
        (trx->sub_sp_level != trx->stmt)
       )
    {
      error = HA_ERR_UNSUPPORTED;
      goto cleanup;
    }
    txn = trx->stmt;

    error = allocate_key_and_col_info(altered_table->s, &altered_kc_info);
    if (error) { goto cleanup; }

    max_new_desc_size = get_max_desc_size(&altered_kc_info, altered_table);
    row_desc_buff = (uchar *)my_malloc(max_new_desc_size, MYF(MY_WME));
    if (row_desc_buff == NULL){ error = ENOMEM; goto cleanup;}

    // drop indexes
    if (dropping_indexes) {
        error = drop_indexes(table, alter_info->index_drop_buffer, alter_info->index_drop_count, txn);
        if (error) { goto cleanup; }
    }

    // add indexes
    if (adding_indexes) {
        KEY           *key_info;
        KEY           *key;
        uint          *idx_p;
        uint          *idx_end_p;
        KEY_PART_INFO *key_part;
        KEY_PART_INFO *part_end;
        /* The add_index() method takes an array of KEY structs. */
        key_info= (KEY*) thd->alloc(sizeof(KEY) * alter_info->index_add_count);
        key= key_info;
        for (idx_p= alter_info->index_add_buffer, idx_end_p= idx_p + alter_info->index_add_count;
             idx_p < idx_end_p;
             idx_p++, key++)
        {
          /* Copy the KEY struct. */
          *key= alter_info->key_info_buffer[*idx_p];
          /* Fix the key parts. */
          part_end= key->key_part + key->key_parts;
          for (key_part= key->key_part; key_part < part_end; key_part++)
            key_part->field = table->field[key_part->fieldnr];
        }
        error = tokudb_add_index(
            table, 
            key_info,
            alter_info->index_add_count,
            txn,
            &incremented_numDBs,
            &modified_DBs
            );
        if (error) { 
            // hack for now, in case of duplicate key error, 
            // because at the moment we cannot display the right key
            // information to the user, so that he knows potentially what went
            // wrong.
            last_dup_key = MAX_KEY;
            goto cleanup;
        }
    }

    if (has_dropped_columns || has_added_columns) {
        DBT column_dbt;
        bzero(&column_dbt, sizeof(DBT));
        u_int32_t max_column_extra_size;
        u_int32_t num_column_extra;
        u_int32_t columns[table->s->fields + altered_table->s->fields]; // set size such that we know it is big enough for both cases
        u_int32_t num_columns = 0;
        u_int32_t curr_num_DBs = table->s->keys + test(hidden_primary_key);
        memset(columns, 0, sizeof(columns));

        if (has_added_columns && has_dropped_columns) {
            error = HA_ERR_UNSUPPORTED;
            goto cleanup;
        }
9175
        if (!tables_have_same_keys(table, altered_table, true, false)) {
9176 9177 9178 9179 9180 9181 9182 9183 9184 9185 9186 9187 9188 9189 9190 9191 9192 9193 9194 9195 9196 9197 9198 9199 9200 9201 9202 9203 9204 9205 9206 9207 9208 9209 9210 9211 9212 9213 9214 9215 9216 9217 9218 9219 9220 9221 9222 9223 9224 9225 9226 9227 9228 9229 9230 9231 9232 9233 9234 9235 9236 9237 9238 9239 9240 9241 9242 9243 9244 9245 9246 9247 9248 9249 9250 9251 9252 9253 9254 9255 9256 9257 9258 9259 9260 9261 9262 9263 9264 9265 9266 9267 9268 9269 9270 9271 9272 9273 9274 9275
            error = HA_ERR_UNSUPPORTED;
            goto cleanup;
        }

        error = initialize_key_and_col_info(
            altered_table->s, 
            altered_table,
            &altered_kc_info,
            hidden_primary_key,
            primary_key
            );
        if (error) { goto cleanup; }

        // generate the array of columns
        if (has_dropped_columns) {
            find_changed_columns(
                columns,
                &num_columns,
                altered_table,
                table
                );
        }
        if (has_added_columns) {
            find_changed_columns(
                columns,
                &num_columns,
                table,
                altered_table
                );
        }
        max_column_extra_size = 
            STATIC_ROW_MUTATOR_SIZE + //max static row_mutator
            4 + num_columns*(1+1+4+1+1+4) + altered_table->s->reclength + // max dynamic row_mutator
            (4 + share->kc_info.num_blobs) + // max static blob size
            (num_columns*(1+4+1+4)); // max dynamic blob size
        column_extra = (uchar *)my_malloc(max_column_extra_size, MYF(MY_WME));
        if (column_extra == NULL) { error = ENOMEM; goto cleanup; }

        for (u_int32_t i = 0; i < curr_num_DBs; i++) {
            DBT row_descriptor;
            bzero(&row_descriptor, sizeof(row_descriptor));
            KEY* prim_key = (hidden_primary_key) ? NULL : &altered_table->s->key_info[primary_key];
            KEY* key_info = &altered_table->key_info[i];
            if (i == primary_key) {
                row_descriptor.size = create_main_key_descriptor(
                    row_desc_buff,
                    prim_key,
                    hidden_primary_key,
                    primary_key,
                    altered_table,
                    &altered_kc_info
                    );
                    row_descriptor.data = row_desc_buff;
            }
            else {
                row_descriptor.size = create_secondary_key_descriptor(
                    row_desc_buff,
                    key_info,
                    prim_key,
                    hidden_primary_key,
                    altered_table,
                    primary_key,
                    i,
                    &altered_kc_info
                    );
                row_descriptor.data = row_desc_buff;
            }
            error = share->key_file[i]->change_descriptor(
                share->key_file[i],
                txn,
                &row_descriptor,
                0
                );
            if (error) { goto cleanup; }
            
            if (i == primary_key || table_share->key_info[i].flags & HA_CLUSTERING) {
                num_column_extra = fill_row_mutator(
                    column_extra,
                    columns,
                    num_columns,
                    altered_table,
                    &altered_kc_info,
                    i,
                    has_added_columns // true if adding columns, otherwise is a drop
                    );
                
                column_dbt.data = column_extra;
                column_dbt.size = num_column_extra;
                DBUG_ASSERT(num_column_extra <= max_column_extra_size);
                
                error = share->key_file[i]->update_broadcast(
                    share->key_file[i],
                    txn,
                    &column_dbt,
                    DB_IS_RESETTING_OP
                    );
                if (error) { goto cleanup; }
            }
        }
    }
9276 9277 9278 9279 9280 9281 9282

    // update frm file    
    // only for tables that are not partitioned
    if (altered_table->part_info == NULL) {
        error = write_frm_data(share->status_block, txn, altered_table->s->path.str);
        if (error) { goto cleanup; }
    }    
9283 9284 9285 9286 9287 9288 9289 9290 9291 9292 9293 9294 9295 9296 9297 9298 9299 9300 9301 9302 9303 9304 9305 9306 9307 9308 9309 9310 9311 9312 9313 9314 9315 9316 9317 9318 9319 9320 9321 9322 9323 9324 9325 9326 9327 9328 9329 9330 9331 9332 9333 9334 9335 9336 9337 9338 9339 9340 9341 9342 9343 9344 9345 9346 9347 9348 9349 9350 9351 9352 9353 9354 9355 9356 9357 9358 9359 9360 9361 9362 9363 9364 9365 9366 9367 9368 9369 9370 9371 9372 9373 9374 9375 9376 9377 9378 9379 9380 9381 9382 9383 9384 9385 9386 9387 9388 9389 9390 9391 9392 9393 9394 9395 9396 9397 9398 9399 9400 9401 9402 9403 9404 9405 9406 9407 9408 9409 9410 9411 9412 9413 9414 9415 9416 9417 9418 9419 9420 9421 9422 9423 9424 9425 9426 9427 9428 9429 9430 9431 9432 9433 9434 9435 9436 9437 9438 9439 9440 9441 9442 9443 9444 9445 9446 9447 9448 9449 9450 9451 9452 9453 9454 9455 9456 9457 9458 9459 9460 9461 9462 9463 9464 9465 9466 9467 9468 9469 9470 9471 9472 9473 9474 9475 9476 9477 9478 9479 9480 9481 9482 9483 9484 9485 9486 9487 9488 9489 9490 9491 9492 9493 9494 9495 9496 9497 9498 9499 9500 9501 9502 9503 9504 9505 9506 9507 9508 9509 9510 9511 9512 9513 9514 9515 9516 9517 9518 9519 9520 9521 9522 9523 9524 9525 9526 9527 9528 9529 9530 9531 9532 9533 9534 9535 9536 9537 9538 9539 9540 9541 9542 9543 9544 9545 9546 9547 9548 9549 9550 9551 9552 9553 9554 9555 9556 9557 9558 9559 9560 9561 9562 9563 9564 9565 9566 9567 9568 9569 9570 9571 9572 9573 9574 9575 9576 9577 9578 9579 9580 9581 9582 9583 9584 9585 9586 9587 9588 9589 9590 9591 9592 9593 9594 9595 9596 9597 9598 9599 9600 9601 9602 9603 9604 9605 9606 9607 9608 9609 9610 9611 9612 9613 9614 9615 9616 9617 9618 9619 9620 9621 9622 9623 9624 9625 9626 9627 9628 9629 9630 9631 9632 9633 9634 9635 9636 9637 9638 9639 9640 9641 9642 9643 9644 9645 9646 9647 9648 9649 9650 9651 9652 9653 9654 9655 9656 9657 9658 9659 9660 9661 9662 9663 9664 9665 9666 9667 9668 9669 9670 9671 9672 9673 9674 9675 9676 9677 9678 9679 9680 9681 9682 9683 9684 9685 9686 9687 9688 9689 9690 9691 9692 9693 9694 9695 9696 9697 9698 9699 9700 9701 9702 9703 9704 9705 9706 9707 9708 9709 9710 9711 9712 9713 9714 9715 9716 9717 9718 9719 9720 9721 9722 9723 9724 9725 9726 9727 9728 9729 9730 9731 9732 9733 9734 9735 9736 9737 9738 9739 9740 9741 9742 9743 9744 9745 9746 9747 9748 9749 9750 9751 9752 9753 9754 9755 9756 9757 9758 9759 9760 9761 9762 9763 9764 9765 9766 9767 9768 9769 9770 9771 9772 9773 9774 9775 9776 9777 9778 9779 9780 9781 9782 9783 9784 9785 9786 9787 9788 9789 9790 9791 9792 9793 9794 9795 9796 9797 9798 9799 9800 9801 9802 9803 9804 9805 9806 9807 9808 9809 9810 9811 9812 9813 9814 9815 9816
    if (thd->killed) {
        error = ER_ABORTING_CONNECTION;
        goto cleanup;
    }

    error = 0;    
cleanup:
    free_key_and_col_info(&altered_kc_info);
    my_free(row_desc_buff, MYF(MY_ALLOW_ZERO_PTR));
    my_free(column_extra, MYF(MY_ALLOW_ZERO_PTR));
    if (txn) {
        if (error) {
            if (adding_indexes) {
                restore_add_index(table, alter_info->index_add_count, incremented_numDBs, modified_DBs);
            }
            abort_txn(txn);
            trx->stmt = NULL;
            trx->sub_sp_level = NULL;
            if (dropping_indexes) {
                restore_drop_indexes(table, alter_info->index_drop_buffer, alter_info->index_drop_count);
            }
        }
    }
    TOKUDB_DBUG_RETURN(error);
}

inline void copy_null_bits(
    u_int32_t start_old_pos,
    u_int32_t start_new_pos,
    u_int32_t num_bits,
    uchar* old_null_bytes,
    uchar* new_null_bytes
    ) 
{
    for (u_int32_t i = 0; i < num_bits; i++) {
        u_int32_t curr_old_pos = i + start_old_pos;
        u_int32_t curr_new_pos = i + start_new_pos;
        // copy over old null bytes
        if (is_overall_null_position_set(old_null_bytes,curr_old_pos)) {
            set_overall_null_position(new_null_bytes,curr_new_pos,true);
        }
        else {
            set_overall_null_position(new_null_bytes,curr_new_pos,false);
        }
    }
}

inline void copy_var_fields(
    u_int32_t start_old_num_var_field, //index of var fields that we should start writing
    u_int32_t num_var_fields, // number of var fields to copy
    uchar* old_var_field_offset_ptr, //static ptr to where offset bytes begin in old row
    uchar old_num_offset_bytes, //number of offset bytes used in old row
    uchar* start_new_var_field_data_ptr, // where the new var data should be written
    uchar* start_new_var_field_offset_ptr, // where the new var offsets should be written
    uchar* new_var_field_data_ptr, // pointer to beginning of var fields in new row
    uchar* old_var_field_data_ptr, // pointer to beginning of var fields in old row
    u_int32_t new_num_offset_bytes, // number of offset bytes used in new row
    u_int32_t* num_data_bytes_written,
    u_int32_t* num_offset_bytes_written
    ) 
{
    uchar* curr_new_var_field_data_ptr = start_new_var_field_data_ptr;
    uchar* curr_new_var_field_offset_ptr = start_new_var_field_offset_ptr;
    for (u_int32_t i = 0; i < num_var_fields; i++) {
        u_int32_t field_len;
        u_int32_t start_read_offset;
        u_int32_t curr_old = i + start_old_num_var_field;
        uchar* data_to_copy = NULL;
        // get the length and pointer to data that needs to be copied
        get_var_field_info(
            &field_len, 
            &start_read_offset, 
            curr_old, 
            old_var_field_offset_ptr, 
            old_num_offset_bytes
            );
        data_to_copy = old_var_field_data_ptr + start_read_offset;
        // now need to copy field_len bytes starting from data_to_copy
        curr_new_var_field_data_ptr = write_var_field(
            curr_new_var_field_offset_ptr,
            curr_new_var_field_data_ptr,
            new_var_field_data_ptr,
            data_to_copy,
            field_len,
            new_num_offset_bytes
            );
        curr_new_var_field_offset_ptr += new_num_offset_bytes;
    }
    *num_data_bytes_written = (u_int32_t)(curr_new_var_field_data_ptr - start_new_var_field_data_ptr);
    *num_offset_bytes_written = (u_int32_t)(curr_new_var_field_offset_ptr - start_new_var_field_offset_ptr);
}

inline u_int32_t copy_toku_blob(uchar* to_ptr, uchar* from_ptr, u_int32_t len_bytes, bool skip) {
    u_int32_t length = 0;
    if (!skip) {
        memcpy(to_ptr, from_ptr, len_bytes);
    }
    length = get_blob_field_len(from_ptr,len_bytes);
    if (!skip) {
        memcpy(to_ptr + len_bytes, from_ptr + len_bytes, length);
    }
    return (length + len_bytes);
}

int tokudb_update_fun(
    DB* db,
    const DBT *key,
    const DBT *old_val, 
    const DBT *extra,
    void (*set_val)(const DBT *new_val, void *set_extra),
    void *set_extra
    ) 
{
    u_int32_t max_num_bytes;
    u_int32_t num_columns;
    DBT new_val;
    u_int32_t num_bytes_left;
    u_int32_t num_var_fields_to_copy;
    u_int32_t num_data_bytes_written = 0;
    u_int32_t num_offset_bytes_written = 0;
    int error;
    bzero(&new_val, sizeof(DBT));
    uchar operation;
    uchar* new_val_data = NULL;
    uchar* extra_pos = NULL;
    uchar* extra_pos_start = NULL;
    //
    // info for pointers into rows
    //
    u_int32_t old_num_null_bytes;
    u_int32_t new_num_null_bytes;
    uchar old_num_offset_bytes;
    uchar new_num_offset_bytes;
    u_int32_t old_fixed_field_size;
    u_int32_t new_fixed_field_size;
    u_int32_t old_len_of_offsets;
    u_int32_t new_len_of_offsets;

    uchar* old_fixed_field_ptr = NULL;
    uchar* new_fixed_field_ptr = NULL;
    u_int32_t curr_old_fixed_offset;
    u_int32_t curr_new_fixed_offset;

    uchar* old_null_bytes = NULL;
    uchar* new_null_bytes = NULL;
    u_int32_t curr_old_null_pos;
    u_int32_t curr_new_null_pos;    
    u_int32_t old_null_bits_left;
    u_int32_t new_null_bits_left;
    u_int32_t overall_null_bits_left;

    u_int32_t old_num_var_fields;
    u_int32_t new_num_var_fields;
    u_int32_t curr_old_num_var_field;
    u_int32_t curr_new_num_var_field;
    uchar* old_var_field_offset_ptr = NULL;
    uchar* new_var_field_offset_ptr = NULL;
    uchar* curr_new_var_field_offset_ptr = NULL;
    uchar* old_var_field_data_ptr = NULL;
    uchar* new_var_field_data_ptr = NULL;
    uchar* curr_new_var_field_data_ptr = NULL;

    u_int32_t start_blob_offset;
    uchar* start_blob_ptr;
    u_int32_t num_blob_bytes;

    // came across a delete, nothing to update
    if (old_val == NULL) {
        error = 0;
        goto cleanup;
    }

    extra_pos_start = (uchar *)extra->data;
    extra_pos = (uchar *)extra->data;

    operation = extra_pos[0];
    extra_pos++;
    assert(operation == UP_COL_ADD_OR_DROP);

    memcpy(&old_num_null_bytes, extra_pos, sizeof(u_int32_t));
    extra_pos += sizeof(u_int32_t);
    memcpy(&new_num_null_bytes, extra_pos, sizeof(u_int32_t));
    extra_pos += sizeof(u_int32_t);

    old_num_offset_bytes = extra_pos[0];
    extra_pos++;
    new_num_offset_bytes = extra_pos[0];
    extra_pos++;

    memcpy(&old_fixed_field_size, extra_pos, sizeof(u_int32_t));
    extra_pos += sizeof(u_int32_t);
    memcpy(&new_fixed_field_size, extra_pos, sizeof(u_int32_t));
    extra_pos += sizeof(u_int32_t);

    memcpy(&old_len_of_offsets, extra_pos, sizeof(u_int32_t));
    extra_pos += sizeof(u_int32_t);
    memcpy(&new_len_of_offsets, extra_pos, sizeof(u_int32_t));
    extra_pos += sizeof(u_int32_t);

    max_num_bytes = old_val->size + extra->size + new_len_of_offsets + new_fixed_field_size;
    new_val_data = (uchar *)my_malloc(
        max_num_bytes, 
        MYF(MY_FAE)
        );
    if (new_val_data == NULL) { goto cleanup; }

    old_fixed_field_ptr = (uchar *) old_val->data;
    old_fixed_field_ptr += old_num_null_bytes;
    new_fixed_field_ptr = new_val_data + new_num_null_bytes;
    curr_old_fixed_offset = 0;
    curr_new_fixed_offset = 0;

    old_num_var_fields = old_len_of_offsets/old_num_offset_bytes;
    new_num_var_fields = new_len_of_offsets/new_num_offset_bytes;
    // following fields will change as we write the variable data
    old_var_field_offset_ptr = old_fixed_field_ptr + old_fixed_field_size;
    new_var_field_offset_ptr = new_fixed_field_ptr + new_fixed_field_size;
    old_var_field_data_ptr = old_var_field_offset_ptr + old_len_of_offsets;
    new_var_field_data_ptr = new_var_field_offset_ptr + new_len_of_offsets;
    curr_new_var_field_offset_ptr = new_var_field_offset_ptr;
    curr_new_var_field_data_ptr = new_var_field_data_ptr;
    curr_old_num_var_field = 0;
    curr_new_num_var_field = 0;

    old_null_bytes = (uchar *)old_val->data;
    new_null_bytes = new_val_data;

    
    memcpy(&curr_old_null_pos, extra_pos, sizeof(u_int32_t));
    extra_pos += sizeof(u_int32_t);
    memcpy(&curr_new_null_pos, extra_pos, sizeof(u_int32_t));
    extra_pos += sizeof(u_int32_t);

    memcpy(&num_columns, extra_pos, sizeof(num_columns));
    extra_pos += sizeof(num_columns);
    
    //
    // now go through and apply the change into new_val_data
    //
    for (u_int32_t i = 0; i < num_columns; i++) {
        uchar op_type = extra_pos[0];
        bool is_null_default = false;
        extra_pos++;

        assert(op_type == COL_DROP || op_type == COL_ADD);
        bool nullable = (extra_pos[0] != 0);
        extra_pos++;
        if (nullable) {
            u_int32_t null_bit_position;
            memcpy(&null_bit_position, extra_pos, sizeof(u_int32_t));
            extra_pos += sizeof(u_int32_t);
            u_int32_t num_bits;
            if (op_type == COL_DROP) {
                assert(curr_old_null_pos <= null_bit_position);
                num_bits = null_bit_position - curr_old_null_pos;
            }
            else {
                assert(curr_new_null_pos <= null_bit_position);
                num_bits = null_bit_position - curr_new_null_pos;
            }
            copy_null_bits(
                curr_old_null_pos,
                curr_new_null_pos,
                num_bits,
                old_null_bytes,
                new_null_bytes
                );
            // update the positions
            curr_new_null_pos += num_bits;
            curr_old_null_pos += num_bits;
            if (op_type == COL_DROP) {
                curr_old_null_pos++; // account for dropped column
            }
            else {
                is_null_default = (extra_pos[0] != 0);
                extra_pos++;
                set_overall_null_position(
                    new_null_bytes,
                    null_bit_position,
                    is_null_default
                    );
                curr_new_null_pos++; //account for added column
            }
        }
        uchar col_type = extra_pos[0];
        extra_pos++;
        if (col_type == COL_FIXED) {
            u_int32_t col_offset;
            u_int32_t col_size;
            u_int32_t num_bytes_to_copy;
            memcpy(&col_offset, extra_pos, sizeof(u_int32_t));
            extra_pos += sizeof(u_int32_t);
            memcpy(&col_size, extra_pos, sizeof(u_int32_t));
            extra_pos += sizeof(u_int32_t);

            if (op_type == COL_DROP) {
                num_bytes_to_copy = col_offset - curr_old_fixed_offset;
            }
            else {
                num_bytes_to_copy = col_offset - curr_new_fixed_offset;
            }
            memcpy(
                new_fixed_field_ptr + curr_new_fixed_offset,
                old_fixed_field_ptr + curr_old_fixed_offset, 
                num_bytes_to_copy
                );
            curr_old_fixed_offset += num_bytes_to_copy;
            curr_new_fixed_offset += num_bytes_to_copy;
            if (op_type == COL_DROP) {
                // move old_fixed_offset val to skip OVER column that is being dropped
                curr_old_fixed_offset += col_size;
            }
            else {
                if (is_null_default) {
                    // copy zeroes
                    bzero(new_fixed_field_ptr + curr_new_fixed_offset, col_size);
                }
                else {
                    // copy data from extra_pos into new row
                    memcpy(
                        new_fixed_field_ptr + curr_new_fixed_offset,
                        extra_pos,
                        col_size
                        );
                    extra_pos += col_size;
                }
                curr_new_fixed_offset += col_size;
            }
            
        }
        else if (col_type == COL_VAR) {
            u_int32_t var_col_index;
            memcpy(&var_col_index, extra_pos, sizeof(u_int32_t));
            extra_pos += sizeof(u_int32_t);
            if (op_type == COL_DROP) {
                num_var_fields_to_copy = var_col_index - curr_old_num_var_field;
            }
            else {
                num_var_fields_to_copy = var_col_index - curr_new_num_var_field;
            }
            copy_var_fields(
                curr_old_num_var_field,
                num_var_fields_to_copy,
                old_var_field_offset_ptr,
                old_num_offset_bytes,
                curr_new_var_field_data_ptr,
                curr_new_var_field_offset_ptr,
                new_var_field_data_ptr, // pointer to beginning of var fields in new row
                old_var_field_data_ptr, // pointer to beginning of var fields in old row
                new_num_offset_bytes, // number of offset bytes used in new row
                &num_data_bytes_written,
                &num_offset_bytes_written
                );
            curr_new_var_field_data_ptr += num_data_bytes_written;
            curr_new_var_field_offset_ptr += num_offset_bytes_written;
            curr_new_num_var_field += num_var_fields_to_copy;
            curr_old_num_var_field += num_var_fields_to_copy;
            if (op_type == COL_DROP) {
                curr_old_num_var_field++; // skip over dropped field
            }
            else {
                if (is_null_default) {
                    curr_new_var_field_data_ptr = write_var_field(
                        curr_new_var_field_offset_ptr,
                        curr_new_var_field_data_ptr,
                        new_var_field_data_ptr,
                        NULL, //copying no data
                        0, //copying 0 bytes
                        new_num_offset_bytes
                        );
                    curr_new_var_field_offset_ptr += new_num_offset_bytes;
                }
                else {
                    u_int32_t data_length;
                    memcpy(&data_length, extra_pos, sizeof(data_length));
                    extra_pos += sizeof(data_length);
                    curr_new_var_field_data_ptr = write_var_field(
                        curr_new_var_field_offset_ptr,
                        curr_new_var_field_data_ptr,
                        new_var_field_data_ptr,
                        extra_pos, //copying data from mutator
                        data_length, //copying data_length bytes
                        new_num_offset_bytes
                        );
                    extra_pos += data_length;
                    curr_new_var_field_offset_ptr += new_num_offset_bytes;
                }
                curr_new_num_var_field++; //account for added column
            }
        }
        else if (col_type == COL_BLOB) {
            // handle blob data later
            continue;
        }
        else {
            assert(false);
        }
    }
    // finish copying the null stuff
    old_null_bits_left = 8*old_num_null_bytes - curr_old_null_pos;
    new_null_bits_left = 8*new_num_null_bytes - curr_new_null_pos;
    overall_null_bits_left = old_null_bits_left;
    set_if_smaller(overall_null_bits_left, new_null_bits_left);
    copy_null_bits(
        curr_old_null_pos,
        curr_new_null_pos,
        overall_null_bits_left,
        old_null_bytes,
        new_null_bytes
        );
    // finish copying fixed field stuff
    num_bytes_left = old_fixed_field_size - curr_old_fixed_offset;
    memcpy(
        new_fixed_field_ptr + curr_new_fixed_offset,
        old_fixed_field_ptr + curr_old_fixed_offset, 
        num_bytes_left
        );
    curr_old_fixed_offset += num_bytes_left;
    curr_new_fixed_offset += num_bytes_left;
    // sanity check
    assert(curr_new_fixed_offset == new_fixed_field_size);

    // finish copying var field stuff
    num_var_fields_to_copy = old_num_var_fields - curr_old_num_var_field;
    copy_var_fields(
        curr_old_num_var_field,
        num_var_fields_to_copy,
        old_var_field_offset_ptr,
        old_num_offset_bytes,
        curr_new_var_field_data_ptr,
        curr_new_var_field_offset_ptr,
        new_var_field_data_ptr, // pointer to beginning of var fields in new row
        old_var_field_data_ptr, // pointer to beginning of var fields in old row
        new_num_offset_bytes, // number of offset bytes used in new row
        &num_data_bytes_written,
        &num_offset_bytes_written
        );
    curr_new_var_field_offset_ptr += num_offset_bytes_written;
    curr_new_var_field_data_ptr += num_data_bytes_written;
    // sanity check
    assert(curr_new_var_field_offset_ptr == new_var_field_data_ptr);

    // start handling blobs
    get_blob_field_info(
        &start_blob_offset, 
        old_len_of_offsets,
        old_var_field_data_ptr,
        old_num_offset_bytes
        );
    start_blob_ptr = old_var_field_data_ptr + start_blob_offset;
    // if nothing else in extra, then there are no blobs to add or drop, so can copy blobs straight
    if ((extra_pos - extra_pos_start) == extra->size) {
        num_blob_bytes = old_val->size - (start_blob_ptr - old_null_bytes);
        memcpy(curr_new_var_field_data_ptr, start_blob_ptr, num_blob_bytes);
        curr_new_var_field_data_ptr += num_blob_bytes;
    }
    // else, there is blob information to process
    else {
        uchar* len_bytes = NULL;
        u_int32_t curr_old_blob = 0;
        u_int32_t curr_new_blob = 0;
        u_int32_t num_old_blobs = 0;
        uchar* curr_old_blob_ptr = start_blob_ptr;
        memcpy(&num_old_blobs, extra_pos, sizeof(num_old_blobs));
        extra_pos += sizeof(num_old_blobs);
        len_bytes = extra_pos;
        extra_pos += num_old_blobs;
        // copy over blob fields one by one
        while ((extra_pos - extra_pos_start) < extra->size) {
            uchar op_type = extra_pos[0];
            extra_pos++;
            u_int32_t num_blobs_to_copy = 0;
            u_int32_t blob_index;
            memcpy(&blob_index, extra_pos, sizeof(blob_index));
            extra_pos += sizeof(blob_index);
            assert (op_type == COL_DROP || op_type == COL_ADD);
            if (op_type == COL_DROP) {
                num_blobs_to_copy = blob_index - curr_old_blob;
            }
            else {
                num_blobs_to_copy = blob_index - curr_new_blob;
            }
            for (u_int32_t i = 0; i < num_blobs_to_copy; i++) {
                u_int32_t num_bytes_written = copy_toku_blob(
                    curr_new_var_field_data_ptr,
                    curr_old_blob_ptr,
                    len_bytes[curr_old_blob + i],
                    false
                    );
                curr_old_blob_ptr += num_bytes_written;
                curr_new_var_field_data_ptr += num_bytes_written;
            }
            curr_old_blob += num_blobs_to_copy;
            curr_new_blob += num_blobs_to_copy;
            if (op_type == COL_DROP) {
                // skip over blob in row
                u_int32_t num_bytes = copy_toku_blob(
                    NULL,
                    curr_old_blob_ptr,
                    len_bytes[curr_old_blob],
                    true
                    );
                curr_old_blob++;
                curr_old_blob_ptr += num_bytes;
            }
            else {
                // copy new data
                u_int32_t new_len_bytes = extra_pos[0];
                extra_pos++;
                u_int32_t num_bytes = copy_toku_blob(
                    curr_new_var_field_data_ptr,
                    extra_pos,
                    new_len_bytes,
                    false
                    );
                curr_new_blob++;
                curr_new_var_field_data_ptr += num_bytes;
                extra_pos += num_bytes;
            }                
        }
        num_blob_bytes = old_val->size - (curr_old_blob_ptr - old_null_bytes);
        memcpy(curr_new_var_field_data_ptr, curr_old_blob_ptr, num_blob_bytes);
        curr_new_var_field_data_ptr += num_blob_bytes;
    }
    new_val.data = new_val_data;
    new_val.size = curr_new_var_field_data_ptr - new_val_data;
    set_val(&new_val, set_extra);
    
    error = 0;
cleanup:
    my_free(new_val_data, MYF(MY_ALLOW_ZERO_PTR));
    return error;    
}

9817 9818
#endif

9819 9820 9821 9822 9823 9824 9825 9826 9827 9828 9829 9830 9831 9832
struct check_context {
    THD *thd;
};

static int
ha_tokudb_check_progress(void *extra, float progress) {
    struct check_context *context = (struct check_context *) extra;
    int result = 0;
    if (context->thd->killed)
        result = ER_ABORTING_CONNECTION;
    return result;
}

static void
9833
ha_tokudb_check_info(THD *thd, TABLE *table, const char *msg) {
9834
    if (thd->vio_ok()) {
9835 9836
        char tablename[256];
        snprintf(tablename, sizeof tablename, "%s.%s", table->s->db.str, table->s->table_name.str);
9837 9838 9839 9840 9841 9842 9843 9844 9845
        thd->protocol->prepare_for_resend();
        thd->protocol->store(tablename, strlen(tablename), system_charset_info);
        thd->protocol->store("check", 5, system_charset_info);
        thd->protocol->store("info", 4, system_charset_info);
        thd->protocol->store(msg, strlen(msg), system_charset_info);
        thd->protocol->write();
    }
}

9846
volatile int ha_tokudb_check_verbose = 0; // debug
9847
volatile int ha_tokudb_check_wait = 0; // debug
9848 9849 9850 9851 9852

int
ha_tokudb::check(THD *thd, HA_CHECK_OPT *check_opt) {
    TOKUDB_DBUG_ENTER("check");

9853
    while (ha_tokudb_check_wait) sleep(1); // debug
9854

9855 9856 9857
    const char *old_proc_info = thd->proc_info;
    thd_proc_info(thd, "tokudb::check");

9858 9859 9860 9861 9862 9863 9864 9865 9866 9867 9868 9869 9870 9871 9872 9873
    int result = HA_ADMIN_OK;
    int r;

    int keep_going = 1;
    if (check_opt->flags & T_QUICK) {
        keep_going = 0;
    }
    if (check_opt->flags & T_EXTEND) {
        keep_going = 1;
    }

    r = acquire_table_lock(transaction, lock_write);
    if (r != 0)
        result = HA_ADMIN_INTERNAL_ERROR;
    if (result == HA_ADMIN_OK) {
        uint32_t num_DBs = table_share->keys + test(hidden_primary_key);
9874 9875
        time_t now;
        char timebuf[32];
9876
        snprintf(write_status_msg, sizeof write_status_msg, "%s primary=%d num=%d", share->table_name, primary_key, num_DBs);
9877
        if (ha_tokudb_check_verbose) {
9878
            ha_tokudb_check_info(thd, table, write_status_msg);
9879
            now = time(0);
9880
            fprintf(stderr, "%.24s ha_tokudb::check %s\n", ctime_r(&now, timebuf), write_status_msg);
9881
        }
9882 9883 9884 9885 9886
        for (uint i = 0; i < num_DBs; i++) {
            time_t now;
            DB *db = share->key_file[i];
            const char *kname = table_share->key_info[i].name;
            if (i == primary_key)
9887
                kname = "primary"; // hidden primary key does not set name
9888
            snprintf(write_status_msg, sizeof write_status_msg, "%s key=%s %u", share->table_name, kname, i);
9889
            thd_proc_info(thd, write_status_msg);
9890
            if (ha_tokudb_check_verbose) {
9891
                ha_tokudb_check_info(thd, table, write_status_msg);
9892
                now = time(0);
9893
                fprintf(stderr, "%.24s ha_tokudb::check %s\n", ctime_r(&now, timebuf), write_status_msg);
9894
            }
9895
            struct check_context check_context = { thd };
9896
            r = db->verify_with_progress(db, ha_tokudb_check_progress, &check_context, ha_tokudb_check_verbose, keep_going);
9897
            snprintf(write_status_msg, sizeof write_status_msg, "%s key=%s %u result=%d", share->table_name, kname, i, r);
9898
            thd_proc_info(thd, write_status_msg);
9899
            if (ha_tokudb_check_verbose) {
9900
                ha_tokudb_check_info(thd, table, write_status_msg);
9901
                now = time(0);
9902
                fprintf(stderr, "%.24s ha_tokudb::check %s\n", ctime_r(&now, timebuf), write_status_msg);
9903
            }
9904
            if (result == HA_ADMIN_OK && r != 0) {
9905
                result = HA_ADMIN_CORRUPT;
9906 9907 9908
                if (!keep_going)
                    break;
            }
9909 9910 9911 9912 9913
        }
    }
    thd_proc_info(thd, old_proc_info);
    TOKUDB_DBUG_RETURN(result);
}