ha_ndbcluster.cc 124 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34
/* Copyright (C) 2000-2003 MySQL AB

  This program is free software; you can redistribute it and/or modify
  it under the terms of the GNU General Public License as published by
  the Free Software Foundation; either version 2 of the License, or
  (at your option) any later version.

  This program is distributed in the hope that it will be useful,
  but WITHOUT ANY WARRANTY; without even the implied warranty of
  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  GNU General Public License for more details.

  You should have received a copy of the GNU General Public License
  along with this program; if not, write to the Free Software
  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA 
*/

/*
  This file defines the NDB Cluster handler: the interface between MySQL and
  NDB Cluster
*/

#ifdef __GNUC__
#pragma implementation                          // gcc: Class implementation
#endif

#include "mysql_priv.h"

#ifdef HAVE_NDBCLUSTER_DB
#include <my_dir.h>
#include "ha_ndbcluster.h"
#include <ndbapi/NdbApi.hpp>
#include <ndbapi/NdbScanFilter.hpp>

35 36 37 38
// options from from mysqld.cc
extern my_bool opt_ndb_optimized_node_selection;
extern const char *opt_ndbcluster_connectstring;

39 40 41
// Default value for parallelism
static const int parallelism= 240;

42 43
// Default value for max number of transactions
// createable against NDB from this handler
44 45
static const int max_transactions= 256;

46 47
static const char *ha_ndb_ext=".ndb";

48
#define NDB_HIDDEN_PRIMARY_KEY_LENGTH 8
49

50 51

#define ERR_PRINT(err) \
52
  DBUG_PRINT("error", ("%d  message: %s", err.code, err.message))
53 54 55 56 57 58 59 60 61

#define ERR_RETURN(err)		         \
{				         \
  ERR_PRINT(err);		         \
  DBUG_RETURN(ndb_to_mysql_error(&err)); \
}

// Typedefs for long names
typedef NdbDictionary::Column NDBCOL;
joreland@mysql.com's avatar
joreland@mysql.com committed
62
typedef NdbDictionary::Table NDBTAB;
63 64 65
typedef NdbDictionary::Index  NDBINDEX;
typedef NdbDictionary::Dictionary  NDBDICT;

66
bool ndbcluster_inited= FALSE;
67

68
static Ndb* g_ndb= NULL;
69
static Ndb_cluster_connection* g_ndb_cluster_connection= NULL;
70

71 72 73 74 75 76 77 78 79 80 81 82 83 84 85
// Handler synchronization
pthread_mutex_t ndbcluster_mutex;

// Table lock handling
static HASH ndbcluster_open_tables;

static byte *ndbcluster_get_key(NDB_SHARE *share,uint *length,
                                my_bool not_used __attribute__((unused)));
static NDB_SHARE *get_share(const char *table_name);
static void free_share(NDB_SHARE *share);

static int packfrm(const void *data, uint len, const void **pack_data, uint *pack_len);
static int unpackfrm(const void **data, uint *len,
		     const void* pack_data);

86
static int ndb_get_table_statistics(Ndb*, const char *, 
87 88
				    Uint64* rows, Uint64* commits);

89

90 91 92 93 94 95
/*
  Dummy buffer to read zero pack_length fields
  which are mapped to 1 char
*/
static byte dummy_buf[1];

96 97 98 99 100 101 102 103
/*
  Error handling functions
*/

struct err_code_mapping
{
  int ndb_err;
  int my_err;
104
  int show_warning;
105 106 107 108
};

static const err_code_mapping err_map[]= 
{
109 110
  { 626, HA_ERR_KEY_NOT_FOUND, 0 },
  { 630, HA_ERR_FOUND_DUPP_KEY, 0 },
111
  { 893, HA_ERR_FOUND_DUPP_KEY, 0 },
112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133
  { 721, HA_ERR_TABLE_EXIST, 1 },
  { 4244, HA_ERR_TABLE_EXIST, 1 },

  { 709, HA_ERR_NO_SUCH_TABLE, 1 },
  { 284, HA_ERR_NO_SUCH_TABLE, 1 },

  { 266, HA_ERR_LOCK_WAIT_TIMEOUT, 1 },
  { 274, HA_ERR_LOCK_WAIT_TIMEOUT, 1 },
  { 296, HA_ERR_LOCK_WAIT_TIMEOUT, 1 },
  { 297, HA_ERR_LOCK_WAIT_TIMEOUT, 1 },
  { 237, HA_ERR_LOCK_WAIT_TIMEOUT, 1 },

  { 623, HA_ERR_RECORD_FILE_FULL, 1 },
  { 624, HA_ERR_RECORD_FILE_FULL, 1 },
  { 625, HA_ERR_RECORD_FILE_FULL, 1 },
  { 826, HA_ERR_RECORD_FILE_FULL, 1 },
  { 827, HA_ERR_RECORD_FILE_FULL, 1 },
  { 832, HA_ERR_RECORD_FILE_FULL, 1 },

  { 0, 1, 0 },

  { -1, -1, 1 }
134 135 136 137 138 139
};


static int ndb_to_mysql_error(const NdbError *err)
{
  uint i;
140 141
  for (i=0; err_map[i].ndb_err != err->code && err_map[i].my_err != -1; i++);
  if (err_map[i].show_warning)
142
  {
143 144 145 146
    // Push the NDB error message as warning
    push_warning_printf(current_thd, MYSQL_ERROR::WARN_LEVEL_ERROR,
			ER_GET_ERRMSG, ER(ER_GET_ERRMSG),
			err->code, err->message, "NDB");
147
  }
148 149
  if (err_map[i].my_err == -1)
    return err->code;
150 151 152 153
  return err_map[i].my_err;
}


154 155 156 157 158

inline
int execute_no_commit(ha_ndbcluster *h, NdbConnection *trans)
{
  int m_batch_execute= 0;
159 160
#ifdef NOT_USED
  if (m_batch_execute)
161
    return 0;
162
#endif
163
  return trans->execute(NoCommit,AbortOnError,h->m_force_send);
164 165 166 167 168 169
}

inline
int execute_commit(ha_ndbcluster *h, NdbConnection *trans)
{
  int m_batch_execute= 0;
170 171
#ifdef NOT_USED
  if (m_batch_execute)
172
    return 0;
173
#endif
174 175 176 177 178 179 180 181 182 183 184 185
  return trans->execute(Commit,AbortOnError,h->m_force_send);
}

inline
int execute_commit(THD *thd, NdbConnection *trans)
{
  int m_batch_execute= 0;
#ifdef NOT_USED
  if (m_batch_execute)
    return 0;
#endif
  return trans->execute(Commit,AbortOnError,thd->variables.ndb_force_send);
186 187 188 189 190 191
}

inline
int execute_no_commit_ie(ha_ndbcluster *h, NdbConnection *trans)
{
  int m_batch_execute= 0;
192 193
#ifdef NOT_USED
  if (m_batch_execute)
194
    return 0;
195
#endif
joreland@mysql.com's avatar
merge  
joreland@mysql.com committed
196
  return trans->execute(NoCommit, AO_IgnoreError,h->m_force_send);
197 198
}

199 200 201 202 203 204
/*
  Place holder for ha_ndbcluster thread specific data
*/

Thd_ndb::Thd_ndb()
{
205
  ndb= new Ndb(g_ndb_cluster_connection, "");
206 207
  lock_count= 0;
  count= 0;
208
  error= 0;
209 210 211 212
}

Thd_ndb::~Thd_ndb()
{
213 214
  if (ndb)
    delete ndb;
mskold@mysql.com's avatar
mskold@mysql.com committed
215
  ndb= 0;
216 217
}

218 219 220 221 222 223
inline
Ndb *ha_ndbcluster::get_ndb()
{
  return ((Thd_ndb*)current_thd->transaction.thd_ndb)->ndb;
}

224 225 226 227 228 229
/*
 * manage uncommitted insert/deletes during transactio to get records correct
 */

struct Ndb_table_local_info {
  int no_uncommitted_rows_count;
230
  ulong last_count;
231 232 233
  ha_rows records;
};

tomas@poseidon.ndb.mysql.com's avatar
tomas@poseidon.ndb.mysql.com committed
234 235 236 237 238 239 240 241 242 243
void ha_ndbcluster::set_rec_per_key()
{
  DBUG_ENTER("ha_ndbcluster::get_status_const");
  for (uint i=0 ; i < table->keys ; i++)
  {
    table->key_info[i].rec_per_key[table->key_info[i].key_parts-1]= 1;
  }
  DBUG_VOID_RETURN;
}

244 245
void ha_ndbcluster::records_update()
{
246 247
  if (m_ha_not_exact_count)
    return;
248 249 250 251 252
  DBUG_ENTER("ha_ndbcluster::records_update");
  struct Ndb_table_local_info *info= (struct Ndb_table_local_info *)m_table_info;
  DBUG_PRINT("info", ("id=%d, no_uncommitted_rows_count=%d",
		      ((const NDBTAB *)m_table)->getTableId(),
		      info->no_uncommitted_rows_count));
253
  //  if (info->records == ~(ha_rows)0)
254
  {
255
    Ndb *ndb= get_ndb();
256
    Uint64 rows;
257
    if(ndb_get_table_statistics(ndb, m_tabname, &rows, 0) == 0){
258 259 260
      info->records= rows;
    }
  }
261 262 263 264 265
  {
    THD *thd= current_thd;
    if (((Thd_ndb*)(thd->transaction.thd_ndb))->error)
      info->no_uncommitted_rows_count= 0;
  }
266 267 268 269
  records= info->records+ info->no_uncommitted_rows_count;
  DBUG_VOID_RETURN;
}

270 271
void ha_ndbcluster::no_uncommitted_rows_execute_failure()
{
272 273
  if (m_ha_not_exact_count)
    return;
274
  DBUG_ENTER("ha_ndbcluster::no_uncommitted_rows_execute_failure");
275 276
  THD *thd= current_thd;
  ((Thd_ndb*)(thd->transaction.thd_ndb))->error= 1;
277 278 279
  DBUG_VOID_RETURN;
}

280 281
void ha_ndbcluster::no_uncommitted_rows_init(THD *thd)
{
282 283
  if (m_ha_not_exact_count)
    return;
284 285 286
  DBUG_ENTER("ha_ndbcluster::no_uncommitted_rows_init");
  struct Ndb_table_local_info *info= (struct Ndb_table_local_info *)m_table_info;
  Thd_ndb *thd_ndb= (Thd_ndb *)thd->transaction.thd_ndb;
287
  if (info->last_count != thd_ndb->count)
288
  {
289
    info->last_count = thd_ndb->count;
290 291 292 293 294 295 296 297 298 299 300
    info->no_uncommitted_rows_count= 0;
    info->records= ~(ha_rows)0;
    DBUG_PRINT("info", ("id=%d, no_uncommitted_rows_count=%d",
			((const NDBTAB *)m_table)->getTableId(),
			info->no_uncommitted_rows_count));
  }
  DBUG_VOID_RETURN;
}

void ha_ndbcluster::no_uncommitted_rows_update(int c)
{
301 302
  if (m_ha_not_exact_count)
    return;
303
  DBUG_ENTER("ha_ndbcluster::no_uncommitted_rows_update");
304 305
  struct Ndb_table_local_info *info=
    (struct Ndb_table_local_info *)m_table_info;
306 307 308 309 310 311 312 313 314
  info->no_uncommitted_rows_count+= c;
  DBUG_PRINT("info", ("id=%d, no_uncommitted_rows_count=%d",
		      ((const NDBTAB *)m_table)->getTableId(),
		      info->no_uncommitted_rows_count));
  DBUG_VOID_RETURN;
}

void ha_ndbcluster::no_uncommitted_rows_reset(THD *thd)
{
315 316
  if (m_ha_not_exact_count)
    return;
317 318
  DBUG_ENTER("ha_ndbcluster::no_uncommitted_rows_reset");
  ((Thd_ndb*)(thd->transaction.thd_ndb))->count++;
319
  ((Thd_ndb*)(thd->transaction.thd_ndb))->error= 0;
320 321 322
  DBUG_VOID_RETURN;
}

323 324 325 326 327 328 329 330
/*
  Take care of the error that occured in NDB
  
  RETURN
    0	No error
    #   The mapped error code
*/

331

332 333
int ha_ndbcluster::ndb_err(NdbConnection *trans)
{
334
  int res;
335 336 337 338 339 340 341
  const NdbError err= trans->getNdbError();
  DBUG_ENTER("ndb_err");
  
  ERR_PRINT(err);
  switch (err.classification) {
  case NdbError::SchemaError:
  {
342 343
    Ndb *ndb= get_ndb();
    NDBDICT *dict= ndb->getDictionary();
344 345
    DBUG_PRINT("info", ("invalidateTable %s", m_tabname));
    dict->invalidateTable(m_tabname);
346
    table->version=0L;			/* Free when thread is ready */
347 348 349 350 351
    break;
  }
  default:
    break;
  }
352 353 354 355
  res= ndb_to_mysql_error(&err);
  DBUG_PRINT("info", ("transformed ndbcluster error %d to mysql error %d", 
		      err.code, res));
  if (res == HA_ERR_FOUND_DUPP_KEY)
356
    m_dupkey= table->primary_key;
357 358
  
  DBUG_RETURN(res);
359 360 361
}


362
/*
363
  Override the default get_error_message in order to add the 
364 365 366
  error message of NDB 
 */

367 368
bool ha_ndbcluster::get_error_message(int error, 
				      String *buf)
369
{
370
  DBUG_ENTER("ha_ndbcluster::get_error_message");
371
  DBUG_PRINT("enter", ("error: %d", error));
372

373
  Ndb *ndb= get_ndb();
374
  if (!ndb)
375
    DBUG_RETURN(FALSE);
376

377
  const NdbError err= ndb->getNdbError(error);
378 379 380 381
  bool temporary= err.status==NdbError::TemporaryError;
  buf->set(err.message, strlen(err.message), &my_charset_bin);
  DBUG_PRINT("exit", ("message: %s, temporary: %d", buf->ptr(), temporary));
  DBUG_RETURN(temporary);
382 383 384
}


pekka@mysql.com's avatar
pekka@mysql.com committed
385 386
/*
  Check if type is supported by NDB.
387 388
  TODO Use this once in open(), not in every operation

pekka@mysql.com's avatar
pekka@mysql.com committed
389 390 391 392 393
*/

static inline bool ndb_supported_type(enum_field_types type)
{
  switch (type) {
pekka@mysql.com's avatar
pekka@mysql.com committed
394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415
  case MYSQL_TYPE_DECIMAL:    
  case MYSQL_TYPE_TINY:        
  case MYSQL_TYPE_SHORT:
  case MYSQL_TYPE_LONG:
  case MYSQL_TYPE_INT24:       
  case MYSQL_TYPE_LONGLONG:
  case MYSQL_TYPE_FLOAT:
  case MYSQL_TYPE_DOUBLE:
  case MYSQL_TYPE_TIMESTAMP:
  case MYSQL_TYPE_DATETIME:    
  case MYSQL_TYPE_DATE:
  case MYSQL_TYPE_NEWDATE:
  case MYSQL_TYPE_TIME:        
  case MYSQL_TYPE_YEAR:        
  case MYSQL_TYPE_STRING:      
  case MYSQL_TYPE_VAR_STRING:
  case MYSQL_TYPE_TINY_BLOB:
  case MYSQL_TYPE_BLOB:    
  case MYSQL_TYPE_MEDIUM_BLOB:   
  case MYSQL_TYPE_LONG_BLOB:  
  case MYSQL_TYPE_ENUM:
  case MYSQL_TYPE_SET:         
416
    return TRUE;
pekka@mysql.com's avatar
pekka@mysql.com committed
417 418
  case MYSQL_TYPE_NULL:   
  case MYSQL_TYPE_GEOMETRY:
pekka@mysql.com's avatar
pekka@mysql.com committed
419
    break;
pekka@mysql.com's avatar
pekka@mysql.com committed
420
  }
421
  return FALSE;
pekka@mysql.com's avatar
pekka@mysql.com committed
422 423 424
}


425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451
/*
  Instruct NDB to set the value of the hidden primary key
*/

bool ha_ndbcluster::set_hidden_key(NdbOperation *ndb_op,
				   uint fieldnr, const byte *field_ptr)
{
  DBUG_ENTER("set_hidden_key");
  DBUG_RETURN(ndb_op->equal(fieldnr, (char*)field_ptr,
			    NDB_HIDDEN_PRIMARY_KEY_LENGTH) != 0);
}


/*
  Instruct NDB to set the value of one primary key attribute
*/

int ha_ndbcluster::set_ndb_key(NdbOperation *ndb_op, Field *field,
                               uint fieldnr, const byte *field_ptr)
{
  uint32 pack_len= field->pack_length();
  DBUG_ENTER("set_ndb_key");
  DBUG_PRINT("enter", ("%d: %s, ndb_type: %u, len=%d", 
                       fieldnr, field->field_name, field->type(),
                       pack_len));
  DBUG_DUMP("key", (char*)field_ptr, pack_len);
  
pekka@mysql.com's avatar
pekka@mysql.com committed
452 453 454 455 456
  if (ndb_supported_type(field->type()))
  {
    if (! (field->flags & BLOB_FLAG))
      // Common implementation for most field types
      DBUG_RETURN(ndb_op->equal(fieldnr, (char*) field_ptr, pack_len) != 0);
457
  }
pekka@mysql.com's avatar
pekka@mysql.com committed
458 459 460
  // Unhandled field types
  DBUG_PRINT("error", ("Field type %d not supported", field->type()));
  DBUG_RETURN(2);
461 462 463 464 465 466 467 468
}


/*
 Instruct NDB to set the value of one attribute
*/

int ha_ndbcluster::set_ndb_value(NdbOperation *ndb_op, Field *field, 
469
                                 uint fieldnr, bool *set_blob_value)
470 471 472 473 474 475 476 477
{
  const byte* field_ptr= field->ptr;
  uint32 pack_len=  field->pack_length();
  DBUG_ENTER("set_ndb_value");
  DBUG_PRINT("enter", ("%d: %s, type: %u, len=%d, is_null=%s", 
                       fieldnr, field->field_name, field->type(), 
                       pack_len, field->is_null()?"Y":"N"));
  DBUG_DUMP("value", (char*) field_ptr, pack_len);
pekka@mysql.com's avatar
pekka@mysql.com committed
478 479

  if (ndb_supported_type(field->type()))
480
  {
481 482 483 484 485 486 487
    // ndb currently does not support size 0
    const byte *empty_field= "";
    if (pack_len == 0)
    {
      pack_len= 1;
      field_ptr= empty_field;
    }
pekka@mysql.com's avatar
pekka@mysql.com committed
488 489 490 491 492 493 494 495 496 497
    if (! (field->flags & BLOB_FLAG))
    {
      if (field->is_null())
        // Set value to NULL
        DBUG_RETURN((ndb_op->setValue(fieldnr, (char*)NULL, pack_len) != 0));
      // Common implementation for most field types
      DBUG_RETURN(ndb_op->setValue(fieldnr, (char*)field_ptr, pack_len) != 0);
    }

    // Blob type
498
    NdbBlob *ndb_blob= ndb_op->getBlobHandle(fieldnr);
pekka@mysql.com's avatar
pekka@mysql.com committed
499 500 501 502 503 504 505 506 507 508 509 510
    if (ndb_blob != NULL)
    {
      if (field->is_null())
        DBUG_RETURN(ndb_blob->setNull() != 0);

      Field_blob *field_blob= (Field_blob*)field;

      // Get length and pointer to data
      uint32 blob_len= field_blob->get_length(field_ptr);
      char* blob_ptr= NULL;
      field_blob->get_ptr(&blob_ptr);

511 512 513
      // Looks like NULL ptr signals length 0 blob
      if (blob_ptr == NULL) {
        DBUG_ASSERT(blob_len == 0);
514
        blob_ptr= (char*)"";
515
      }
pekka@mysql.com's avatar
pekka@mysql.com committed
516 517 518 519 520

      DBUG_PRINT("value", ("set blob ptr=%x len=%u",
                           (unsigned)blob_ptr, blob_len));
      DBUG_DUMP("value", (char*)blob_ptr, min(blob_len, 26));

521
      if (set_blob_value)
522
	*set_blob_value= TRUE;
pekka@mysql.com's avatar
pekka@mysql.com committed
523 524 525 526
      // No callback needed to write value
      DBUG_RETURN(ndb_blob->setValue(blob_ptr, blob_len) != 0);
    }
    DBUG_RETURN(1);
527
  }
pekka@mysql.com's avatar
pekka@mysql.com committed
528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545
  // Unhandled field types
  DBUG_PRINT("error", ("Field type %d not supported", field->type()));
  DBUG_RETURN(2);
}


/*
  Callback to read all blob values.
  - not done in unpack_record because unpack_record is valid
    after execute(Commit) but reading blobs is not
  - may only generate read operations; they have to be executed
    somewhere before the data is available
  - due to single buffer for all blobs, we let the last blob
    process all blobs (last so that all are active)
  - null bit is still set in unpack_record
  - TODO allocate blob part aligned buffers
*/

546
NdbBlob::ActiveHook g_get_ndb_blobs_value;
pekka@mysql.com's avatar
pekka@mysql.com committed
547

548
int g_get_ndb_blobs_value(NdbBlob *ndb_blob, void *arg)
pekka@mysql.com's avatar
pekka@mysql.com committed
549
{
550
  DBUG_ENTER("g_get_ndb_blobs_value");
pekka@mysql.com's avatar
pekka@mysql.com committed
551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582
  if (ndb_blob->blobsNextBlob() != NULL)
    DBUG_RETURN(0);
  ha_ndbcluster *ha= (ha_ndbcluster *)arg;
  DBUG_RETURN(ha->get_ndb_blobs_value(ndb_blob));
}

int ha_ndbcluster::get_ndb_blobs_value(NdbBlob *last_ndb_blob)
{
  DBUG_ENTER("get_ndb_blobs_value");

  // Field has no field number so cannot use TABLE blob_field
  // Loop twice, first only counting total buffer size
  for (int loop= 0; loop <= 1; loop++)
  {
    uint32 offset= 0;
    for (uint i= 0; i < table->fields; i++)
    {
      Field *field= table->field[i];
      NdbValue value= m_value[i];
      if (value.ptr != NULL && (field->flags & BLOB_FLAG))
      {
        Field_blob *field_blob= (Field_blob *)field;
        NdbBlob *ndb_blob= value.blob;
        Uint64 blob_len= 0;
        if (ndb_blob->getLength(blob_len) != 0)
          DBUG_RETURN(-1);
        // Align to Uint64
        uint32 blob_size= blob_len;
        if (blob_size % 8 != 0)
          blob_size+= 8 - blob_size % 8;
        if (loop == 1)
        {
583
          char *buf= m_blobs_buffer + offset;
pekka@mysql.com's avatar
pekka@mysql.com committed
584 585
          uint32 len= 0xffffffff;  // Max uint32
          DBUG_PRINT("value", ("read blob ptr=%x len=%u",
joreland@mysql.com's avatar
joreland@mysql.com committed
586
                               (UintPtr)buf, (uint)blob_len));
pekka@mysql.com's avatar
pekka@mysql.com committed
587 588 589 590 591 592 593 594
          if (ndb_blob->readData(buf, len) != 0)
            DBUG_RETURN(-1);
          DBUG_ASSERT(len == blob_len);
          field_blob->set_ptr(len, buf);
        }
        offset+= blob_size;
      }
    }
595
    if (loop == 0 && offset > m_blobs_buffer_size)
pekka@mysql.com's avatar
pekka@mysql.com committed
596
    {
597 598
      my_free(m_blobs_buffer, MYF(MY_ALLOW_ZERO_PTR));
      m_blobs_buffer_size= 0;
pekka@mysql.com's avatar
pekka@mysql.com committed
599
      DBUG_PRINT("value", ("allocate blobs buffer size %u", offset));
600 601
      m_blobs_buffer= my_malloc(offset, MYF(MY_WME));
      if (m_blobs_buffer == NULL)
pekka@mysql.com's avatar
pekka@mysql.com committed
602
        DBUG_RETURN(-1);
603
      m_blobs_buffer_size= offset;
pekka@mysql.com's avatar
pekka@mysql.com committed
604
    }
605
  }
pekka@mysql.com's avatar
pekka@mysql.com committed
606
  DBUG_RETURN(0);
607 608 609 610 611
}


/*
  Instruct NDB to fetch one field
pekka@mysql.com's avatar
pekka@mysql.com committed
612 613
  - data is read directly into buffer provided by field
    if field is NULL, data is read into memory provided by NDBAPI
614 615
*/

pekka@mysql.com's avatar
pekka@mysql.com committed
616
int ha_ndbcluster::get_ndb_value(NdbOperation *ndb_op, Field *field,
617
                                 uint fieldnr, byte* buf)
618 619
{
  DBUG_ENTER("get_ndb_value");
pekka@mysql.com's avatar
pekka@mysql.com committed
620 621 622 623 624
  DBUG_PRINT("enter", ("fieldnr: %d flags: %o", fieldnr,
                       (int)(field != NULL ? field->flags : 0)));

  if (field != NULL)
  {
625
    DBUG_ASSERT(buf);
pekka@mysql.com's avatar
pekka@mysql.com committed
626 627 628 629
    if (ndb_supported_type(field->type()))
    {
      DBUG_ASSERT(field->ptr != NULL);
      if (! (field->flags & BLOB_FLAG))
630
      {	
631 632 633 634 635
	byte *field_buf;
	if (field->pack_length() != 0)
	  field_buf= buf + (field->ptr - table->record[0]);
	else
	  field_buf= dummy_buf;
636 637
        m_value[fieldnr].rec= ndb_op->getValue(fieldnr, 
					       field_buf);
pekka@mysql.com's avatar
pekka@mysql.com committed
638 639 640 641 642 643 644 645 646 647
        DBUG_RETURN(m_value[fieldnr].rec == NULL);
      }

      // Blob type
      NdbBlob *ndb_blob= ndb_op->getBlobHandle(fieldnr);
      m_value[fieldnr].blob= ndb_blob;
      if (ndb_blob != NULL)
      {
        // Set callback
        void *arg= (void *)this;
648
        DBUG_RETURN(ndb_blob->setActiveHook(g_get_ndb_blobs_value, arg) != 0);
pekka@mysql.com's avatar
pekka@mysql.com committed
649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668
      }
      DBUG_RETURN(1);
    }
    // Unhandled field types
    DBUG_PRINT("error", ("Field type %d not supported", field->type()));
    DBUG_RETURN(2);
  }

  // Used for hidden key only
  m_value[fieldnr].rec= ndb_op->getValue(fieldnr, NULL);
  DBUG_RETURN(m_value[fieldnr].rec == NULL);
}


/*
  Check if any set or get of blob value in current query.
*/
bool ha_ndbcluster::uses_blob_value(bool all_fields)
{
  if (table->blob_fields == 0)
669
    return FALSE;
pekka@mysql.com's avatar
pekka@mysql.com committed
670
  if (all_fields)
671
    return TRUE;
pekka@mysql.com's avatar
pekka@mysql.com committed
672 673 674
  {
    uint no_fields= table->fields;
    int i;
675
    THD *thd= table->in_use;
pekka@mysql.com's avatar
pekka@mysql.com committed
676 677 678 679 680 681
    // They always put blobs at the end..
    for (i= no_fields - 1; i >= 0; i--)
    {
      Field *field= table->field[i];
      if (thd->query_id == field->query_id)
      {
682
        return TRUE;
pekka@mysql.com's avatar
pekka@mysql.com committed
683 684 685
      }
    }
  }
686
  return FALSE;
687 688 689 690 691 692 693 694 695 696 697 698 699 700 701
}


/*
  Get metadata for this table from NDB 

  IMPLEMENTATION
    - save the NdbDictionary::Table for easy access
    - check that frm-file on disk is equal to frm-file
      of table accessed in NDB
    - build a list of the indexes for the table
*/

int ha_ndbcluster::get_metadata(const char *path)
{
702 703
  Ndb *ndb= get_ndb();
  NDBDICT *dict= ndb->getDictionary();
704 705
  const NDBTAB *tab;
  int error;
706
  bool invalidating_ndb_table= FALSE;
707

708 709 710
  DBUG_ENTER("get_metadata");
  DBUG_PRINT("enter", ("m_tabname: %s, path: %s", m_tabname, path));

711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728
  do {
    const void *data, *pack_data;
    uint length, pack_length;

    if (!(tab= dict->getTable(m_tabname)))
      ERR_RETURN(dict->getNdbError());
    DBUG_PRINT("info", ("Table schema version: %d", tab->getObjectVersion()));
    /*
      Compare FrmData in NDB with frm file from disk.
    */
    error= 0;
    if (readfrm(path, &data, &length) ||
	packfrm(data, length, &pack_data, &pack_length))
    {
      my_free((char*)data, MYF(MY_ALLOW_ZERO_PTR));
      my_free((char*)pack_data, MYF(MY_ALLOW_ZERO_PTR));
      DBUG_RETURN(1);
    }
729
    
730 731 732 733 734 735 736
    if ((pack_length != tab->getFrmLength()) || 
	(memcmp(pack_data, tab->getFrmData(), pack_length)))
    {
      if (!invalidating_ndb_table)
      {
	DBUG_PRINT("info", ("Invalidating table"));
	dict->invalidateTable(m_tabname);
737
	invalidating_ndb_table= TRUE;
738 739 740 741 742 743 744 745 746
      }
      else
      {
	DBUG_PRINT("error", 
		   ("metadata, pack_length: %d getFrmLength: %d memcmp: %d", 
		    pack_length, tab->getFrmLength(),
		    memcmp(pack_data, tab->getFrmData(), pack_length)));      
	DBUG_DUMP("pack_data", (char*)pack_data, pack_length);
	DBUG_DUMP("frm", (char*)tab->getFrmData(), tab->getFrmLength());
747
	error= 3;
748
	invalidating_ndb_table= FALSE;
749 750 751 752
      }
    }
    else
    {
753
      invalidating_ndb_table= FALSE;
754 755 756 757 758
    }
    my_free((char*)data, MYF(0));
    my_free((char*)pack_data, MYF(0));
  } while (invalidating_ndb_table);

759 760 761
  if (error)
    DBUG_RETURN(error);

tomas@poseidon.ndb.mysql.com's avatar
tomas@poseidon.ndb.mysql.com committed
762 763
  m_table= NULL;
  m_table_info= NULL;
764
  
765
  DBUG_RETURN(build_index_list(table, ILBP_OPEN));  
766
}
767

768

mskold@mysql.com's avatar
mskold@mysql.com committed
769
int ha_ndbcluster::build_index_list(TABLE *tab, enum ILBP phase)
770
{
771
  uint i;
772
  int error= 0;
773 774
  const char *name, *index_name;
  char unique_index_name[FN_LEN];
775
  static const char* unique_suffix= "$unique";
776 777
  KEY* key_info= tab->key_info;
  const char **key_name= tab->keynames.type_names;
778 779
  Ndb *ndb= get_ndb();
  NdbDictionary::Dictionary *dict= ndb->getDictionary();
780
  DBUG_ENTER("build_index_list");
781
  
782
  // Save information about all known indexes
783
  for (i= 0; i < tab->keys; i++, key_info++, key_name++)
784
  {
785
    index_name= *key_name;
786
    NDB_INDEX_TYPE idx_type= get_index_type_from_table(i);
787
    m_index[i].type= idx_type;
788
    if (idx_type == UNIQUE_ORDERED_INDEX || idx_type == UNIQUE_INDEX)
789
    {
790 791 792
      strxnmov(unique_index_name, FN_LEN, index_name, unique_suffix, NullS);
      DBUG_PRINT("info", ("Created unique index name \'%s\' for index %d",
			  unique_index_name, i));
793
    }
794 795 796
    // Create secondary indexes if in create phase
    if (phase == ILBP_CREATE)
    {
797 798
      DBUG_PRINT("info", ("Creating index %u: %s", i, index_name));      
      switch (idx_type){
799 800 801 802 803 804 805 806 807
	
      case PRIMARY_KEY_INDEX:
	// Do nothing, already created
	break;
      case PRIMARY_KEY_ORDERED_INDEX:
	error= create_ordered_index(index_name, key_info);
	break;
      case UNIQUE_ORDERED_INDEX:
	if (!(error= create_ordered_index(index_name, key_info)))
808
	  error= create_unique_index(unique_index_name, key_info);
809 810
	break;
      case UNIQUE_INDEX:
811 812
	if (!(error= check_index_fields_not_null(i)))
	  error= create_unique_index(unique_index_name, key_info);
813 814 815 816 817
	break;
      case ORDERED_INDEX:
	error= create_ordered_index(index_name, key_info);
	break;
      default:
818
	DBUG_ASSERT(FALSE);
819 820 821 822 823 824 825 826 827 828
	break;
      }
      if (error)
      {
	DBUG_PRINT("error", ("Failed to create index %u", i));
	drop_table();
	break;
      }
    }
    // Add handles to index objects
829
    if (idx_type != PRIMARY_KEY_INDEX && idx_type != UNIQUE_INDEX)
830
    {
831
      DBUG_PRINT("info", ("Get handle to index %s", index_name));
832
      const NDBINDEX *index= dict->getIndex(index_name, m_tabname);
833
      if (!index) DBUG_RETURN(1);
mskold@mysql.com's avatar
mskold@mysql.com committed
834
      m_index[i].index= (void *) index;
835
    }
836
    if (idx_type == UNIQUE_ORDERED_INDEX || idx_type == UNIQUE_INDEX)
837
    {
838 839
      DBUG_PRINT("info", ("Get handle to unique_index %s", unique_index_name));
      const NDBINDEX *index= dict->getIndex(unique_index_name, m_tabname);
840
      if (!index) DBUG_RETURN(1);
mskold@mysql.com's avatar
mskold@mysql.com committed
841
      m_index[i].unique_index= (void *) index;
842 843
    }      
  }
844 845
  
  DBUG_RETURN(error);
846 847
}

848

849 850 851 852
/*
  Decode the type of an index from information 
  provided in table object
*/
853
NDB_INDEX_TYPE ha_ndbcluster::get_index_type_from_table(uint inx) const
854
{
855 856 857
  bool is_hash_index=  (table->key_info[inx].algorithm == HA_KEY_ALG_HASH);
  if (inx == table->primary_key)
    return is_hash_index ? PRIMARY_KEY_INDEX : PRIMARY_KEY_ORDERED_INDEX;
858
  else
859 860
    return ((table->key_info[inx].flags & HA_NOSAME) ? 
	    (is_hash_index ? UNIQUE_INDEX : UNIQUE_ORDERED_INDEX) :
861 862
	    ORDERED_INDEX);
} 
863

864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883
int ha_ndbcluster::check_index_fields_not_null(uint inx)
{
  KEY* key_info= table->key_info + inx;
  KEY_PART_INFO* key_part= key_info->key_part;
  KEY_PART_INFO* end= key_part+key_info->key_parts;
  DBUG_ENTER("check_index_fields_not_null");
  
  for (; key_part != end; key_part++) 
    {
      Field* field= key_part->field;
      if (field->maybe_null())
      {
	my_printf_error(ER_NULL_COLUMN_IN_INDEX,ER(ER_NULL_COLUMN_IN_INDEX),
			MYF(0),field->field_name);
	DBUG_RETURN(ER_NULL_COLUMN_IN_INDEX);
      }
    }
  
  DBUG_RETURN(0);
}
884 885 886

void ha_ndbcluster::release_metadata()
{
887
  uint i;
888

889 890 891 892
  DBUG_ENTER("release_metadata");
  DBUG_PRINT("enter", ("m_tabname: %s", m_tabname));

  m_table= NULL;
tomas@poseidon.ndb.mysql.com's avatar
tomas@poseidon.ndb.mysql.com committed
893
  m_table_info= NULL;
894

895
  // Release index list 
896 897
  for (i= 0; i < MAX_KEY; i++)
  {
898 899
    m_index[i].unique_index= NULL;      
    m_index[i].index= NULL;      
900 901
  }

902 903 904
  DBUG_VOID_RETURN;
}

pekka@mysql.com's avatar
pekka@mysql.com committed
905
int ha_ndbcluster::get_ndb_lock_type(enum thr_lock_type type)
906
{
907
  if (type >= TL_WRITE_ALLOW_WRITE)
tomas@poseidon.ndb.mysql.com's avatar
tomas@poseidon.ndb.mysql.com committed
908
    return NdbOperation::LM_Exclusive;
909
  else if (uses_blob_value(m_retrieve_all_fields))
910
    return NdbOperation::LM_Read;
pekka@mysql.com's avatar
pekka@mysql.com committed
911
  else
tomas@poseidon.ndb.mysql.com's avatar
tomas@poseidon.ndb.mysql.com committed
912
    return NdbOperation::LM_CommittedRead;
913 914
}

915 916 917 918 919 920
static const ulong index_type_flags[]=
{
  /* UNDEFINED_INDEX */
  0,                         

  /* PRIMARY_KEY_INDEX */
921
  HA_ONLY_WHOLE_INDEX, 
922 923

  /* PRIMARY_KEY_ORDERED_INDEX */
924
  /* 
mskold@mysql.com's avatar
mskold@mysql.com committed
925
     Enable HA_KEYREAD_ONLY when "sorted" indexes are supported, 
926 927 928
     thus ORDERD BY clauses can be optimized by reading directly 
     through the index.
  */
mskold@mysql.com's avatar
mskold@mysql.com committed
929
  // HA_KEYREAD_ONLY | 
930 931 932
  HA_READ_NEXT |
  HA_READ_RANGE |
  HA_READ_ORDER,
933 934

  /* UNIQUE_INDEX */
935
  HA_ONLY_WHOLE_INDEX,
936

937
  /* UNIQUE_ORDERED_INDEX */
938 939 940
  HA_READ_NEXT |
  HA_READ_RANGE |
  HA_READ_ORDER,
941

942
  /* ORDERED_INDEX */
943 944 945
  HA_READ_NEXT |
  HA_READ_RANGE |
  HA_READ_ORDER
946 947 948 949 950 951 952
};

static const int index_flags_size= sizeof(index_type_flags)/sizeof(ulong);

inline NDB_INDEX_TYPE ha_ndbcluster::get_index_type(uint idx_no) const
{
  DBUG_ASSERT(idx_no < MAX_KEY);
953
  return m_index[idx_no].type;
954 955 956 957 958 959 960 961 962 963
}


/*
  Get the flags for an index

  RETURN
    flags depending on the type of the index.
*/

964 965
inline ulong ha_ndbcluster::index_flags(uint idx_no, uint part,
                                        bool all_parts) const 
966 967
{ 
  DBUG_ENTER("index_flags");
968
  DBUG_PRINT("info", ("idx_no: %d", idx_no));
969 970 971 972 973 974 975 976 977 978 979 980 981 982 983 984 985 986 987 988 989 990 991 992
  DBUG_ASSERT(get_index_type_from_table(idx_no) < index_flags_size);
  DBUG_RETURN(index_type_flags[get_index_type_from_table(idx_no)]);
}


int ha_ndbcluster::set_primary_key(NdbOperation *op, const byte *key)
{
  KEY* key_info= table->key_info + table->primary_key;
  KEY_PART_INFO* key_part= key_info->key_part;
  KEY_PART_INFO* end= key_part+key_info->key_parts;
  DBUG_ENTER("set_primary_key");

  for (; key_part != end; key_part++) 
  {
    Field* field= key_part->field;
    if (set_ndb_key(op, field, 
		    key_part->fieldnr-1, key))
      ERR_RETURN(op->getNdbError());
    key += key_part->length;
  }
  DBUG_RETURN(0);
}


993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010
int ha_ndbcluster::set_primary_key_from_old_data(NdbOperation *op, const byte *old_data)
{
  KEY* key_info= table->key_info + table->primary_key;
  KEY_PART_INFO* key_part= key_info->key_part;
  KEY_PART_INFO* end= key_part+key_info->key_parts;
  DBUG_ENTER("set_primary_key_from_old_data");

  for (; key_part != end; key_part++) 
  {
    Field* field= key_part->field;
    if (set_ndb_key(op, field, 
		    key_part->fieldnr-1, old_data+key_part->offset))
      ERR_RETURN(op->getNdbError());
  }
  DBUG_RETURN(0);
}


1011 1012 1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032
int ha_ndbcluster::set_primary_key(NdbOperation *op)
{
  DBUG_ENTER("set_primary_key");
  KEY* key_info= table->key_info + table->primary_key;
  KEY_PART_INFO* key_part= key_info->key_part;
  KEY_PART_INFO* end= key_part+key_info->key_parts;

  for (; key_part != end; key_part++) 
  {
    Field* field= key_part->field;
    if (set_ndb_key(op, field, 
                    key_part->fieldnr-1, field->ptr))
      ERR_RETURN(op->getNdbError());
  }
  DBUG_RETURN(0);
}


/*
  Read one record from NDB using primary key
*/

1033
int ha_ndbcluster::pk_read(const byte *key, uint key_len, byte *buf) 
1034 1035 1036 1037 1038 1039 1040 1041 1042
{
  uint no_fields= table->fields, i;
  NdbConnection *trans= m_active_trans;
  NdbOperation *op;
  THD *thd= current_thd;
  DBUG_ENTER("pk_read");
  DBUG_PRINT("enter", ("key_len: %u", key_len));
  DBUG_DUMP("key", (char*)key, key_len);

1043 1044
  NdbOperation::LockMode lm=
    (NdbOperation::LockMode)get_ndb_lock_type(m_lock.type);
joreland@mysql.com's avatar
joreland@mysql.com committed
1045
  if (!(op= trans->getNdbOperation((const NDBTAB *) m_table)) || 
1046
      op->readTuple(lm) != 0)
1047
    ERR_RETURN(trans->getNdbError());
1048 1049 1050 1051 1052 1053 1054

  if (table->primary_key == MAX_KEY) 
  {
    // This table has no primary key, use "hidden" primary key
    DBUG_PRINT("info", ("Using hidden key"));
    DBUG_DUMP("key", (char*)key, 8);    
    if (set_hidden_key(op, no_fields, key))
1055 1056
      ERR_RETURN(trans->getNdbError());

1057
    // Read key at the same time, for future reference
1058
    if (get_ndb_value(op, NULL, no_fields, NULL))
1059
      ERR_RETURN(trans->getNdbError());
1060 1061 1062 1063 1064 1065 1066 1067
  } 
  else 
  {
    int res;
    if ((res= set_primary_key(op, key)))
      return res;
  }
  
1068
  // Read all wanted non-key field(s) unless HA_EXTRA_RETRIEVE_ALL_COLS
1069 1070 1071
  for (i= 0; i < no_fields; i++) 
  {
    Field *field= table->field[i];
1072
    if ((thd->query_id == field->query_id) ||
1073 1074
	m_retrieve_all_fields ||
	(field->flags & PRI_KEY_FLAG) && m_retrieve_primary_key)
1075
    {
1076
      if (get_ndb_value(op, field, i, buf))
1077
	ERR_RETURN(trans->getNdbError());
1078 1079 1080 1081
    }
    else
    {
      // Attribute was not to be read
pekka@mysql.com's avatar
pekka@mysql.com committed
1082
      m_value[i].ptr= NULL;
1083 1084 1085
    }
  }
  
1086
  if (execute_no_commit_ie(this,trans) != 0) 
1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098
  {
    table->status= STATUS_NOT_FOUND;
    DBUG_RETURN(ndb_err(trans));
  }

  // The value have now been fetched from NDB  
  unpack_record(buf);
  table->status= 0;     
  DBUG_RETURN(0);
}


1099 1100 1101 1102 1103 1104 1105 1106 1107 1108 1109 1110
/*
  Read one complementing record from NDB using primary key from old_data
*/

int ha_ndbcluster::complemented_pk_read(const byte *old_data, byte *new_data)
{
  uint no_fields= table->fields, i;
  NdbConnection *trans= m_active_trans;
  NdbOperation *op;
  THD *thd= current_thd;
  DBUG_ENTER("complemented_pk_read");

1111
  if (m_retrieve_all_fields)
1112 1113 1114
    // We have allready retrieved all fields, nothing to complement
    DBUG_RETURN(0);

1115 1116
  NdbOperation::LockMode lm=
    (NdbOperation::LockMode)get_ndb_lock_type(m_lock.type);
joreland@mysql.com's avatar
joreland@mysql.com committed
1117
  if (!(op= trans->getNdbOperation((const NDBTAB *) m_table)) || 
1118
      op->readTuple(lm) != 0)
1119
    ERR_RETURN(trans->getNdbError());
1120 1121

    int res;
1122
    if ((res= set_primary_key_from_old_data(op, old_data)))
1123
      ERR_RETURN(trans->getNdbError());
1124 1125 1126 1127 1128 1129 1130 1131
    
  // Read all unreferenced non-key field(s)
  for (i= 0; i < no_fields; i++) 
  {
    Field *field= table->field[i];
    if (!(field->flags & PRI_KEY_FLAG) &&
	(thd->query_id != field->query_id))
    {
1132
      if (get_ndb_value(op, field, i, new_data))
1133
	ERR_RETURN(trans->getNdbError());
1134 1135 1136
    }
  }
  
1137
  if (execute_no_commit(this,trans) != 0) 
1138 1139 1140 1141 1142 1143 1144 1145 1146 1147 1148
  {
    table->status= STATUS_NOT_FOUND;
    DBUG_RETURN(ndb_err(trans));
  }

  // The value have now been fetched from NDB  
  unpack_record(new_data);
  table->status= 0;     
  DBUG_RETURN(0);
}

1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169 1170 1171 1172 1173 1174 1175 1176
/*
  Peek to check if a particular row already exists
*/

int ha_ndbcluster::peek_row()
{
  NdbConnection *trans= m_active_trans;
  NdbOperation *op;
  THD *thd= current_thd;
  DBUG_ENTER("peek_row");
                                                                                
  NdbOperation::LockMode lm=
    (NdbOperation::LockMode)get_ndb_lock_type(m_lock.type);
  if (!(op= trans->getNdbOperation((const NDBTAB *) m_table)) ||
      op->readTuple(lm) != 0)
    ERR_RETURN(trans->getNdbError());
                                                                                
  int res;
  if ((res= set_primary_key(op)))
    ERR_RETURN(trans->getNdbError());
                                                                                
  if (execute_no_commit_ie(this,trans) != 0)
    {
      table->status= STATUS_NOT_FOUND;
      DBUG_RETURN(ndb_err(trans));
    }                                                                                
  DBUG_RETURN(0);
}
1177

1178 1179 1180 1181 1182 1183 1184 1185 1186 1187 1188 1189 1190 1191 1192 1193 1194 1195
/*
  Read one record from NDB using unique secondary index
*/

int ha_ndbcluster::unique_index_read(const byte *key,
				     uint key_len, byte *buf)
{
  NdbConnection *trans= m_active_trans;
  NdbIndexOperation *op;
  THD *thd= current_thd;
  byte *key_ptr;
  KEY* key_info;
  KEY_PART_INFO *key_part, *end;
  uint i;
  DBUG_ENTER("unique_index_read");
  DBUG_PRINT("enter", ("key_len: %u, index: %u", key_len, active_index));
  DBUG_DUMP("key", (char*)key, key_len);
  
1196 1197
  NdbOperation::LockMode lm=
    (NdbOperation::LockMode)get_ndb_lock_type(m_lock.type);
1198 1199
  if (!(op= trans->getNdbIndexOperation((NDBINDEX *) 
					m_index[active_index].unique_index, 
joreland@mysql.com's avatar
joreland@mysql.com committed
1200
                                        (const NDBTAB *) m_table)) ||
1201
      op->readTuple(lm) != 0)
1202 1203 1204 1205 1206 1207 1208 1209 1210 1211
    ERR_RETURN(trans->getNdbError());
  
  // Set secondary index key(s)
  key_ptr= (byte *) key;
  key_info= table->key_info + active_index;
  DBUG_ASSERT(key_info->key_length == key_len);
  end= (key_part= key_info->key_part) + key_info->key_parts;

  for (i= 0; key_part != end; key_part++, i++) 
  {
1212 1213
    if (set_ndb_key(op, key_part->field, i, 
		    key_part->null_bit ? key_ptr + 1 : key_ptr))
1214
      ERR_RETURN(trans->getNdbError());
1215
    key_ptr+= key_part->store_length;
1216 1217 1218 1219 1220 1221 1222
  }

  // Get non-index attribute(s)
  for (i= 0; i < table->fields; i++) 
  {
    Field *field= table->field[i];
    if ((thd->query_id == field->query_id) ||
1223
        (field->flags & PRI_KEY_FLAG)) // && m_retrieve_primary_key ??
1224
    {
1225
      if (get_ndb_value(op, field, i, buf))
1226 1227 1228 1229 1230
        ERR_RETURN(op->getNdbError());
    }
    else
    {
      // Attribute was not to be read
pekka@mysql.com's avatar
pekka@mysql.com committed
1231
      m_value[i].ptr= NULL;
1232 1233 1234
    }
  }

1235
  if (execute_no_commit_ie(this,trans) != 0) 
1236 1237 1238 1239 1240 1241 1242 1243 1244 1245 1246
  {
    table->status= STATUS_NOT_FOUND;
    DBUG_RETURN(ndb_err(trans));
  }
  // The value have now been fetched from NDB
  unpack_record(buf);
  table->status= 0;
  DBUG_RETURN(0);
}

/*
1247 1248 1249 1250 1251 1252 1253 1254
  Get the next record of a started scan. Try to fetch
  it locally from NdbApi cached records if possible, 
  otherwise ask NDB for more.

  NOTE
  If this is a update/delete make sure to not contact 
  NDB before any pending ops have been sent to NDB.

1255 1256 1257 1258
*/

inline int ha_ndbcluster::next_result(byte *buf)
{  
1259
  int check;
1260 1261 1262
  NdbConnection *trans= m_active_trans;
  NdbResultSet *cursor= m_active_cursor; 
  DBUG_ENTER("next_result");
1263 1264 1265 1266 1267 1268 1269 1270

  if (!cursor)
    DBUG_RETURN(HA_ERR_END_OF_FILE);
    
  /* 
     If this an update or delete, call nextResult with false
     to process any records already cached in NdbApi
  */
1271
  bool contact_ndb= m_lock.type < TL_WRITE_ALLOW_WRITE;
1272 1273
  do {
    DBUG_PRINT("info", ("Call nextResult, contact_ndb: %d", contact_ndb));
pekka@mysql.com's avatar
pekka@mysql.com committed
1274 1275 1276
    /*
      We can only handle one tuple with blobs at a time.
    */
1277
    if (m_ops_pending && m_blobs_pending)
pekka@mysql.com's avatar
pekka@mysql.com committed
1278
    {
1279
      if (execute_no_commit(this,trans) != 0)
1280
	DBUG_RETURN(ndb_err(trans));
1281 1282
      m_ops_pending= 0;
      m_blobs_pending= FALSE;
pekka@mysql.com's avatar
pekka@mysql.com committed
1283
    }
1284
    check= cursor->nextResult(contact_ndb, m_force_send);
1285 1286 1287 1288
    if (check == 0)
    {
      // One more record found
      DBUG_PRINT("info", ("One more record found"));    
pekka@mysql.com's avatar
pekka@mysql.com committed
1289

1290 1291 1292 1293 1294 1295 1296 1297 1298 1299 1300 1301 1302 1303
      unpack_record(buf);
      table->status= 0;
      DBUG_RETURN(0);
    } 
    else if (check == 1 || check == 2)
    {
      // 1: No more records
      // 2: No more cached records

      /*
	Before fetching more rows and releasing lock(s),
	all pending update or delete operations should 
	be sent to NDB
      */
1304 1305
      DBUG_PRINT("info", ("ops_pending: %d", m_ops_pending));    
      if (m_ops_pending)
1306
      {
1307 1308
	//	if (current_thd->transaction.on)
	if (m_transaction_on)
1309 1310 1311 1312 1313 1314 1315 1316
	{
	  if (execute_no_commit(this,trans) != 0)
	    DBUG_RETURN(ndb_err(trans));
	}
	else
	{
	  if  (execute_commit(this,trans) != 0)
	    DBUG_RETURN(ndb_err(trans));
1317 1318
	  int res= trans->restart();
	  DBUG_ASSERT(res == 0);
1319
	}
1320
	m_ops_pending= 0;
1321
      }
1322 1323 1324 1325 1326
      
      contact_ndb= (check == 2);
    }
  } while (check == 2);
    
1327
  table->status= STATUS_NOT_FOUND;
1328 1329
  if (check == -1)
    DBUG_RETURN(ndb_err(trans));
1330 1331 1332 1333 1334 1335

  // No more records
  DBUG_PRINT("info", ("No more records"));
  DBUG_RETURN(HA_ERR_END_OF_FILE);
}

1336
/*
1337
  Set bounds for ordered index scan.
1338 1339
*/

joreland@mysql.com's avatar
joreland@mysql.com committed
1340
int ha_ndbcluster::set_bounds(NdbIndexScanOperation *op,
1341
			      const key_range *keys[2])
1342
{
1343 1344 1345 1346
  const KEY *const key_info= table->key_info + active_index;
  const uint key_parts= key_info->key_parts;
  uint key_tot_len[2];
  uint tot_len;
1347
  uint i, j;
1348 1349

  DBUG_ENTER("set_bounds");
1350
  DBUG_PRINT("info", ("key_parts=%d", key_parts));
1351

1352
  for (j= 0; j <= 1; j++)
1353
  {
1354 1355 1356 1357 1358 1359 1360 1361 1362 1363 1364 1365 1366
    const key_range *key= keys[j];
    if (key != NULL)
    {
      // for key->flag see ha_rkey_function
      DBUG_PRINT("info", ("key %d length=%d flag=%d",
                          j, key->length, key->flag));
      key_tot_len[j]= key->length;
    }
    else
    {
      DBUG_PRINT("info", ("key %d not present", j));
      key_tot_len[j]= 0;
    }
1367
  }
1368
  tot_len= 0;
1369

1370 1371 1372 1373 1374 1375 1376 1377 1378 1379 1380 1381 1382 1383 1384 1385 1386 1387 1388 1389 1390 1391 1392 1393 1394 1395 1396
  for (i= 0; i < key_parts; i++)
  {
    KEY_PART_INFO *key_part= &key_info->key_part[i];
    Field *field= key_part->field;
    uint part_len= key_part->length;
    uint part_store_len= key_part->store_length;
    // Info about each key part
    struct part_st {
      bool part_last;
      const key_range *key;
      const byte *part_ptr;
      bool part_null;
      int bound_type;
      const char* bound_ptr;
    };
    struct part_st part[2];

    for (j= 0; j <= 1; j++)
    {
      struct part_st &p = part[j];
      p.key= NULL;
      p.bound_type= -1;
      if (tot_len < key_tot_len[j])
      {
        p.part_last= (tot_len + part_store_len >= key_tot_len[j]);
        p.key= keys[j];
        p.part_ptr= &p.key->key[tot_len];
joreland@mysql.com's avatar
joreland@mysql.com committed
1397
        p.part_null= key_part->null_bit && *p.part_ptr;
1398
        p.bound_ptr= (const char *)
joreland@mysql.com's avatar
joreland@mysql.com committed
1399
          p.part_null ? 0 : key_part->null_bit ? p.part_ptr + 1 : p.part_ptr;
1400 1401 1402 1403 1404 1405 1406 1407 1408 1409 1410 1411 1412 1413 1414 1415 1416 1417 1418 1419 1420 1421 1422 1423 1424 1425 1426 1427 1428 1429 1430 1431 1432 1433 1434 1435 1436

        if (j == 0)
        {
          switch (p.key->flag)
          {
            case HA_READ_KEY_EXACT:
              p.bound_type= NdbIndexScanOperation::BoundEQ;
              break;
            case HA_READ_KEY_OR_NEXT:
              p.bound_type= NdbIndexScanOperation::BoundLE;
              break;
            case HA_READ_AFTER_KEY:
              if (! p.part_last)
                p.bound_type= NdbIndexScanOperation::BoundLE;
              else
                p.bound_type= NdbIndexScanOperation::BoundLT;
              break;
            default:
              break;
          }
        }
        if (j == 1) {
          switch (p.key->flag)
          {
            case HA_READ_BEFORE_KEY:
              if (! p.part_last)
                p.bound_type= NdbIndexScanOperation::BoundGE;
              else
                p.bound_type= NdbIndexScanOperation::BoundGT;
              break;
            case HA_READ_AFTER_KEY:     // weird
              p.bound_type= NdbIndexScanOperation::BoundGE;
              break;
            default:
              break;
          }
        }
1437

1438 1439 1440 1441 1442 1443 1444 1445 1446
        if (p.bound_type == -1)
        {
          DBUG_PRINT("error", ("key %d unknown flag %d", j, p.key->flag));
          DBUG_ASSERT(false);
          // Stop setting bounds but continue with what we have
          DBUG_RETURN(0);
        }
      }
    }
1447

1448 1449 1450 1451 1452 1453 1454 1455 1456 1457 1458 1459 1460 1461 1462 1463 1464
    // Seen with e.g. b = 1 and c > 1
    if (part[0].bound_type == NdbIndexScanOperation::BoundLE &&
        part[1].bound_type == NdbIndexScanOperation::BoundGE &&
        memcmp(part[0].part_ptr, part[1].part_ptr, part_store_len) == 0)
    {
      DBUG_PRINT("info", ("replace LE/GE pair by EQ"));
      part[0].bound_type= NdbIndexScanOperation::BoundEQ;
      part[1].bound_type= -1;
    }
    // Not seen but was in previous version
    if (part[0].bound_type == NdbIndexScanOperation::BoundEQ &&
        part[1].bound_type == NdbIndexScanOperation::BoundGE &&
        memcmp(part[0].part_ptr, part[1].part_ptr, part_store_len) == 0)
    {
      DBUG_PRINT("info", ("remove GE from EQ/GE pair"));
      part[1].bound_type= -1;
    }
1465

1466 1467 1468 1469 1470 1471 1472 1473 1474 1475 1476 1477
    for (j= 0; j <= 1; j++)
    {
      struct part_st &p = part[j];
      // Set bound if not done with this key
      if (p.key != NULL)
      {
        DBUG_PRINT("info", ("key %d:%d offset=%d length=%d last=%d bound=%d",
                            j, i, tot_len, part_len, p.part_last, p.bound_type));
        DBUG_DUMP("info", (const char*)p.part_ptr, part_store_len);

        // Set bound if not cancelled via type -1
        if (p.bound_type != -1)
1478 1479 1480 1481 1482
	{
	  char truncated_field_name[NDB_MAX_ATTR_NAME_SIZE];
	  strnmov(truncated_field_name,field->field_name,sizeof(truncated_field_name));
	  truncated_field_name[sizeof(truncated_field_name)-1]= '\0';
          if (op->setBound(truncated_field_name, p.bound_type, p.bound_ptr))
1483
            ERR_RETURN(op->getNdbError());
1484
	}
1485 1486 1487 1488
      }
    }

    tot_len+= part_store_len;
1489
  }
1490
  DBUG_RETURN(0);
1491
}
1492

joreland@mysql.com's avatar
joreland@mysql.com committed
1493 1494 1495 1496 1497 1498 1499 1500 1501 1502 1503 1504 1505 1506 1507
inline 
int ha_ndbcluster::define_read_attrs(byte* buf, NdbOperation* op)
{
  uint i;
  THD *thd= current_thd;
  NdbConnection *trans= m_active_trans;

  DBUG_ENTER("define_read_attrs");  

  // Define attributes to read
  for (i= 0; i < table->fields; i++) 
  {
    Field *field= table->field[i];
    if ((thd->query_id == field->query_id) ||
	(field->flags & PRI_KEY_FLAG) || 
1508
	m_retrieve_all_fields)
joreland@mysql.com's avatar
joreland@mysql.com committed
1509 1510 1511 1512 1513 1514 1515 1516 1517 1518 1519 1520 1521 1522 1523 1524 1525 1526 1527 1528 1529 1530 1531 1532 1533 1534 1535 1536 1537 1538
    {      
      if (get_ndb_value(op, field, i, buf))
	ERR_RETURN(op->getNdbError());
    } 
    else 
    {
      m_value[i].ptr= NULL;
    }
  }
    
  if (table->primary_key == MAX_KEY) 
  {
    DBUG_PRINT("info", ("Getting hidden key"));
    // Scanning table with no primary key
    int hidden_no= table->fields;      
#ifndef DBUG_OFF
    const NDBTAB *tab= (const NDBTAB *) m_table;    
    if (!tab->getColumn(hidden_no))
      DBUG_RETURN(1);
#endif
    if (get_ndb_value(op, NULL, hidden_no, NULL))
      ERR_RETURN(op->getNdbError());
  }

  if (execute_no_commit(this,trans) != 0)
    DBUG_RETURN(ndb_err(trans));
  DBUG_PRINT("exit", ("Scan started successfully"));
  DBUG_RETURN(next_result(buf));
} 

1539
/*
1540
  Start ordered index scan in NDB
1541 1542
*/

1543 1544 1545
int ha_ndbcluster::ordered_index_scan(const key_range *start_key,
				      const key_range *end_key,
				      bool sorted, byte* buf)
1546
{  
joreland@mysql.com's avatar
joreland@mysql.com committed
1547
  bool restart;
1548
  NdbConnection *trans= m_active_trans;
1549
  NdbResultSet *cursor;
joreland@mysql.com's avatar
joreland@mysql.com committed
1550
  NdbIndexScanOperation *op;
1551

1552
  DBUG_ENTER("ordered_index_scan");
1553
  DBUG_PRINT("enter", ("index: %u, sorted: %d", active_index, sorted));  
1554
  DBUG_PRINT("enter", ("Starting new ordered scan on %s", m_tabname));
1555

1556 1557
  // Check that sorted seems to be initialised
  DBUG_ASSERT(sorted == 0 || sorted == 1);
1558
  
1559
  if (m_active_cursor == 0)
joreland@mysql.com's avatar
joreland@mysql.com committed
1560 1561 1562 1563 1564 1565 1566 1567 1568 1569 1570 1571 1572 1573 1574 1575 1576
  {
    restart= false;
    NdbOperation::LockMode lm=
      (NdbOperation::LockMode)get_ndb_lock_type(m_lock.type);
    if (!(op= trans->getNdbIndexScanOperation((NDBINDEX *)
					      m_index[active_index].index, 
					      (const NDBTAB *) m_table)) ||
	!(cursor= op->readTuples(lm, 0, parallelism, sorted)))
      ERR_RETURN(trans->getNdbError());
    m_active_cursor= cursor;
  } else {
    restart= true;
    op= (NdbIndexScanOperation*)m_active_cursor->getOperation();
    
    DBUG_ASSERT(op->getSorted() == sorted);
    DBUG_ASSERT(op->getLockMode() == 
		(NdbOperation::LockMode)get_ndb_lock_type(m_lock.type));
1577
    if(op->reset_bounds(m_force_send))
joreland@mysql.com's avatar
joreland@mysql.com committed
1578 1579
      DBUG_RETURN(ndb_err(m_active_trans));
  }
1580

1581
  {
1582 1583 1584 1585
    const key_range *keys[2]= { start_key, end_key };
    int ret= set_bounds(op, keys);
    if (ret)
      DBUG_RETURN(ret);
1586
  }
1587 1588

  if (!restart)
joreland@mysql.com's avatar
joreland@mysql.com committed
1589 1590 1591 1592 1593 1594 1595 1596 1597 1598
  {
    DBUG_RETURN(define_read_attrs(buf, op));
  }
  else
  {
    if (execute_no_commit(this,trans) != 0)
      DBUG_RETURN(ndb_err(trans));
    
    DBUG_RETURN(next_result(buf));
  }
1599 1600 1601
} 

/*
1602 1603 1604 1605 1606 1607 1608 1609 1610 1611 1612
  Start a filtered scan in NDB.

  NOTE
  This function is here as an example of how to start a
  filtered scan. It should be possible to replace full_table_scan 
  with this function and make a best effort attempt 
  at filtering out the irrelevant data by converting the "items" 
  into interpreted instructions.
  This would speed up table scans where there is a limiting WHERE clause
  that doesn't match any index in the table.

1613 1614 1615 1616 1617 1618 1619
 */

int ha_ndbcluster::filtered_scan(const byte *key, uint key_len, 
				 byte *buf,
				 enum ha_rkey_function find_flag)
{  
  NdbConnection *trans= m_active_trans;
1620 1621
  NdbResultSet *cursor;
  NdbScanOperation *op;
1622 1623 1624 1625 1626 1627 1628

  DBUG_ENTER("filtered_scan");
  DBUG_PRINT("enter", ("key_len: %u, index: %u", 
                       key_len, active_index));
  DBUG_DUMP("key", (char*)key, key_len);  
  DBUG_PRINT("info", ("Starting a new filtered scan on %s",
		      m_tabname));
1629

1630 1631 1632 1633
  NdbOperation::LockMode lm=
    (NdbOperation::LockMode)get_ndb_lock_type(m_lock.type);
  if (!(op= trans->getNdbScanOperation((const NDBTAB *) m_table)) ||
      !(cursor= op->readTuples(lm, 0, parallelism)))
1634 1635
    ERR_RETURN(trans->getNdbError());
  m_active_cursor= cursor;
joreland@mysql.com's avatar
joreland@mysql.com committed
1636
  
1637 1638 1639 1640 1641 1642 1643 1644 1645 1646 1647 1648 1649 1650 1651
  {
    // Start scan filter
    NdbScanFilter sf(op);
    sf.begin();
      
    // Set filter using the supplied key data
    byte *key_ptr= (byte *) key;    
    uint tot_len= 0;
    KEY* key_info= table->key_info + active_index;
    for (uint k= 0; k < key_info->key_parts; k++) 
    {
      KEY_PART_INFO* key_part= key_info->key_part+k;
      Field* field= key_part->field;
      uint ndb_fieldnr= key_part->fieldnr-1;
      DBUG_PRINT("key_part", ("fieldnr: %d", ndb_fieldnr));
1652
      //const NDBCOL *col= ((const NDBTAB *) m_table)->getColumn(ndb_fieldnr);
1653 1654 1655 1656 1657 1658 1659 1660 1661 1662 1663 1664 1665 1666 1667 1668 1669 1670 1671 1672 1673 1674 1675 1676 1677 1678 1679 1680 1681
      uint32 field_len=  field->pack_length();
      DBUG_DUMP("key", (char*)key, field_len);
	
      DBUG_PRINT("info", ("Column %s, type: %d, len: %d", 
			  field->field_name, field->real_type(), field_len));
	
      // Define scan filter
      if (field->real_type() == MYSQL_TYPE_STRING)
	sf.eq(ndb_fieldnr, key_ptr, field_len);
      else 
      {
	if (field_len == 8)
	  sf.eq(ndb_fieldnr, (Uint64)*key_ptr);
	else if (field_len <= 4)
	  sf.eq(ndb_fieldnr, (Uint32)*key_ptr);
	else 
	  DBUG_RETURN(1);
      }
	
      key_ptr += field_len;
      tot_len += field_len;
	
      if (tot_len >= key_len)
	break;
    }
    // End scan filter
    sf.end();
  }

1682
  DBUG_RETURN(define_read_attrs(buf, op));
1683 1684 1685 1686
} 


/*
1687
  Start full table scan in NDB
1688 1689 1690 1691 1692 1693 1694
 */

int ha_ndbcluster::full_table_scan(byte *buf)
{
  uint i;
  NdbResultSet *cursor;
  NdbScanOperation *op;
1695
  NdbConnection *trans= m_active_trans;
1696 1697 1698 1699

  DBUG_ENTER("full_table_scan");  
  DBUG_PRINT("enter", ("Starting new scan on %s", m_tabname));

1700 1701 1702 1703
  NdbOperation::LockMode lm=
    (NdbOperation::LockMode)get_ndb_lock_type(m_lock.type);
  if (!(op=trans->getNdbScanOperation((const NDBTAB *) m_table)) ||
      !(cursor= op->readTuples(lm, 0, parallelism)))
1704 1705
    ERR_RETURN(trans->getNdbError());
  m_active_cursor= cursor;
1706 1707 1708
  DBUG_RETURN(define_read_attrs(buf, op));
}

1709 1710 1711 1712 1713
/*
  Insert one record into NDB
*/
int ha_ndbcluster::write_row(byte *record)
{
mskold@mysql.com's avatar
mskold@mysql.com committed
1714
  bool has_auto_increment;
1715 1716 1717 1718 1719
  uint i;
  NdbConnection *trans= m_active_trans;
  NdbOperation *op;
  int res;
  DBUG_ENTER("write_row");
1720

1721
  if(m_ignore_dup_key && table->primary_key != MAX_KEY)
1722
  {
1723 1724 1725 1726 1727 1728 1729 1730 1731
    int peek_res= peek_row();
    
    if (!peek_res) 
    {
      m_dupkey= table->primary_key;
      DBUG_RETURN(HA_ERR_FOUND_DUPP_KEY);
    }
    if (peek_res != HA_ERR_KEY_NOT_FOUND)
      DBUG_RETURN(peek_res);
1732
  }
1733 1734
  
  statistic_increment(ha_write_count,&LOCK_status);
tomas@poseidon.ndb.mysql.com's avatar
tomas@poseidon.ndb.mysql.com committed
1735 1736
  if (table->timestamp_field_type & TIMESTAMP_AUTO_SET_ON_INSERT)
    table->timestamp_field->set_time();
1737
  has_auto_increment= (table->next_number_field && record == table->record[0]);
1738

joreland@mysql.com's avatar
joreland@mysql.com committed
1739
  if (!(op= trans->getNdbOperation((const NDBTAB *) m_table)))
1740 1741 1742 1743 1744 1745 1746 1747 1748
    ERR_RETURN(trans->getNdbError());

  res= (m_use_write) ? op->writeTuple() :op->insertTuple(); 
  if (res != 0)
    ERR_RETURN(trans->getNdbError());  
 
  if (table->primary_key == MAX_KEY) 
  {
    // Table has hidden primary key
1749 1750
    Ndb *ndb= get_ndb();
    Uint64 auto_value= ndb->getAutoIncrementValue((const NDBTAB *) m_table);
1751 1752 1753 1754 1755 1756
    if (set_hidden_key(op, table->fields, (const byte*)&auto_value))
      ERR_RETURN(op->getNdbError());
  } 
  else 
  {
    int res;
1757

1758 1759
    if (has_auto_increment) 
    {
1760
      m_skip_auto_increment= FALSE;
1761
      update_auto_increment();
1762
      m_skip_auto_increment= !auto_increment_column_changed;
1763
    }
1764

1765 1766 1767 1768 1769
    if ((res= set_primary_key(op)))
      return res;
  }

  // Set non-key attribute(s)
1770
  bool set_blob_value= FALSE;
1771 1772 1773 1774
  for (i= 0; i < table->fields; i++) 
  {
    Field *field= table->field[i];
    if (!(field->flags & PRI_KEY_FLAG) &&
1775
	set_ndb_value(op, field, i, &set_blob_value))
1776
    {
1777
      m_skip_auto_increment= TRUE;
1778
      ERR_RETURN(op->getNdbError());
1779
    }
1780 1781 1782 1783 1784 1785 1786 1787 1788
  }

  /*
    Execute write operation
    NOTE When doing inserts with many values in 
    each INSERT statement it should not be necessary
    to NoCommit the transaction between each row.
    Find out how this is detected!
  */
1789
  m_rows_inserted++;
1790
  no_uncommitted_rows_update(1);
1791 1792 1793
  m_bulk_insert_not_flushed= TRUE;
  if ((m_rows_to_insert == 1) || 
      ((m_rows_inserted % m_bulk_insert_rows) == 0) ||
1794
      set_blob_value)
1795
  {
1796
    THD *thd= current_thd;
1797 1798 1799
    // Send rows to NDB
    DBUG_PRINT("info", ("Sending inserts to NDB, "\
			"rows_inserted:%d, bulk_insert_rows: %d", 
1800
			(int)m_rows_inserted, (int)m_bulk_insert_rows));
1801

1802
    m_bulk_insert_not_flushed= FALSE;
1803 1804
    //    if (thd->transaction.on)
    if (m_transaction_on)
1805
    {
1806
      if (execute_no_commit(this,trans) != 0)
1807
      {
1808
	m_skip_auto_increment= TRUE;
1809
	no_uncommitted_rows_execute_failure();
1810 1811
	DBUG_RETURN(ndb_err(trans));
      }
1812 1813
    }
    else
1814
    {
1815
      if (execute_commit(this,trans) != 0)
1816
      {
1817
	m_skip_auto_increment= TRUE;
1818
	no_uncommitted_rows_execute_failure();
1819
	DBUG_RETURN(ndb_err(trans));
1820
      }
1821 1822
      int res= trans->restart();
      DBUG_ASSERT(res == 0);
1823
    }
1824
  }
1825
  if ((has_auto_increment) && (m_skip_auto_increment))
mskold@mysql.com's avatar
mskold@mysql.com committed
1826
  {
1827
    Ndb *ndb= get_ndb();
1828
    Uint64 next_val= (Uint64) table->next_number_field->val_int() + 1;
mskold@mysql.com's avatar
mskold@mysql.com committed
1829
    DBUG_PRINT("info", 
1830 1831
	       ("Trying to set next auto increment value to %lu",
                (ulong) next_val));
1832
    if (ndb->setAutoIncrementValue((const NDBTAB *) m_table, next_val, TRUE))
mskold@mysql.com's avatar
mskold@mysql.com committed
1833 1834
      DBUG_PRINT("info", 
		 ("Setting next auto increment value to %u", next_val));  
1835
  }
1836
  m_skip_auto_increment= TRUE;
1837

1838 1839 1840 1841 1842 1843 1844 1845 1846 1847 1848 1849 1850 1851 1852 1853 1854 1855 1856 1857 1858 1859 1860 1861 1862 1863 1864 1865 1866 1867 1868 1869 1870 1871 1872 1873 1874 1875 1876 1877 1878 1879 1880 1881 1882 1883
  DBUG_RETURN(0);
}


/* Compare if a key in a row has changed */

int ha_ndbcluster::key_cmp(uint keynr, const byte * old_row,
			   const byte * 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))
    {

      if (key_part->field->cmp_binary((char*) (old_row + key_part->offset),
				      (char*) (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;
    }
  }
  return 0;
}

/*
  Update one record in NDB using primary key
*/

int ha_ndbcluster::update_row(const byte *old_data, byte *new_data)
{
  THD *thd= current_thd;
  NdbConnection *trans= m_active_trans;
1884
  NdbResultSet* cursor= m_active_cursor;
1885 1886 1887 1888 1889
  NdbOperation *op;
  uint i;
  DBUG_ENTER("update_row");
  
  statistic_increment(ha_update_count,&LOCK_status);
tomas@poseidon.ndb.mysql.com's avatar
tomas@poseidon.ndb.mysql.com committed
1890 1891 1892
  if (table->timestamp_field_type & TIMESTAMP_AUTO_SET_ON_UPDATE)
    table->timestamp_field->set_time();

1893
  /* Check for update of primary key for special handling */  
1894 1895
  if ((table->primary_key != MAX_KEY) &&
      (key_cmp(table->primary_key, old_data, new_data)))
1896
  {
1897
    int read_res, insert_res, delete_res;
1898

1899
    DBUG_PRINT("info", ("primary key update, doing pk read+insert+delete"));
1900
    // Get all old fields, since we optimize away fields not in query
1901
    read_res= complemented_pk_read(old_data, new_data);
1902 1903 1904 1905 1906 1907
    if (read_res)
    {
      DBUG_PRINT("info", ("pk read failed"));
      DBUG_RETURN(read_res);
    }
    // Insert new row
1908 1909
    insert_res= write_row(new_data);
    if (insert_res)
1910 1911 1912 1913
    {
      DBUG_PRINT("info", ("insert failed"));
      DBUG_RETURN(insert_res);
    }
1914 1915
    // Delete old row
    DBUG_PRINT("info", ("insert succeded"));
1916
    m_primary_key_update= TRUE;
1917
    delete_res= delete_row(old_data);
1918
    m_primary_key_update= FALSE;
1919 1920 1921 1922 1923 1924 1925 1926
    if (delete_res)
    {
      DBUG_PRINT("info", ("delete failed"));
      // Undo write_row(new_data)
      DBUG_RETURN(delete_row(new_data));
    }     
    DBUG_PRINT("info", ("insert+delete succeeded"));
    DBUG_RETURN(0);
1927
  }
1928

1929
  if (cursor)
1930
  {
1931 1932 1933 1934 1935 1936 1937 1938 1939 1940
    /*
      We are scanning records and want to update the record
      that was just found, call updateTuple on the cursor 
      to take over the lock to a new update operation
      And thus setting the primary key of the record from 
      the active record in cursor
    */
    DBUG_PRINT("info", ("Calling updateTuple on cursor"));
    if (!(op= cursor->updateTuple()))
      ERR_RETURN(trans->getNdbError());
1941
    m_ops_pending++;
1942
    if (uses_blob_value(FALSE))
1943
      m_blobs_pending= TRUE;
1944 1945 1946
  }
  else
  {  
joreland@mysql.com's avatar
joreland@mysql.com committed
1947
    if (!(op= trans->getNdbOperation((const NDBTAB *) m_table)) ||
1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958
	op->updateTuple() != 0)
      ERR_RETURN(trans->getNdbError());  
    
    if (table->primary_key == MAX_KEY) 
    {
      // This table has no primary key, use "hidden" primary key
      DBUG_PRINT("info", ("Using hidden key"));
      
      // Require that the PK for this record has previously been 
      // read into m_value
      uint no_fields= table->fields;
pekka@mysql.com's avatar
pekka@mysql.com committed
1959
      NdbRecAttr* rec= m_value[no_fields].rec;
1960 1961 1962 1963 1964 1965 1966 1967 1968
      DBUG_ASSERT(rec);
      DBUG_DUMP("key", (char*)rec->aRef(), NDB_HIDDEN_PRIMARY_KEY_LENGTH);
      
      if (set_hidden_key(op, no_fields, rec->aRef()))
	ERR_RETURN(op->getNdbError());
    } 
    else 
    {
      int res;
1969
      if ((res= set_primary_key_from_old_data(op, old_data)))
1970 1971
	DBUG_RETURN(res);
    }
1972 1973 1974 1975 1976 1977
  }

  // Set non-key attribute(s)
  for (i= 0; i < table->fields; i++) 
  {
    Field *field= table->field[i];
1978
    if (((thd->query_id == field->query_id) || m_retrieve_all_fields) &&
1979 1980 1981 1982
        (!(field->flags & PRI_KEY_FLAG)) &&
	set_ndb_value(op, field, i))
      ERR_RETURN(op->getNdbError());
  }
1983

1984
  // Execute update operation
1985
  if (!cursor && execute_no_commit(this,trans) != 0) {
1986
    no_uncommitted_rows_execute_failure();
1987
    DBUG_RETURN(ndb_err(trans));
1988
  }
1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000
  
  DBUG_RETURN(0);
}


/*
  Delete one record from NDB, using primary key 
*/

int ha_ndbcluster::delete_row(const byte *record)
{
  NdbConnection *trans= m_active_trans;
2001
  NdbResultSet* cursor= m_active_cursor;
2002 2003 2004 2005 2006
  NdbOperation *op;
  DBUG_ENTER("delete_row");

  statistic_increment(ha_delete_count,&LOCK_status);

2007
  if (cursor)
2008
  {
2009
    /*
2010
      We are scanning records and want to delete the record
2011
      that was just found, call deleteTuple on the cursor 
2012
      to take over the lock to a new delete operation
2013 2014 2015 2016 2017 2018
      And thus setting the primary key of the record from 
      the active record in cursor
    */
    DBUG_PRINT("info", ("Calling deleteTuple on cursor"));
    if (cursor->deleteTuple() != 0)
      ERR_RETURN(trans->getNdbError());     
2019
    m_ops_pending++;
2020

2021 2022
    no_uncommitted_rows_update(-1);

2023 2024 2025 2026
    // If deleting from cursor, NoCommit will be handled in next_result
    DBUG_RETURN(0);
  }
  else
2027
  {
2028
    
joreland@mysql.com's avatar
joreland@mysql.com committed
2029
    if (!(op=trans->getNdbOperation((const NDBTAB *) m_table)) || 
2030 2031 2032
	op->deleteTuple() != 0)
      ERR_RETURN(trans->getNdbError());
    
2033 2034
    no_uncommitted_rows_update(-1);
    
2035 2036 2037 2038 2039
    if (table->primary_key == MAX_KEY) 
    {
      // This table has no primary key, use "hidden" primary key
      DBUG_PRINT("info", ("Using hidden key"));
      uint no_fields= table->fields;
pekka@mysql.com's avatar
pekka@mysql.com committed
2040
      NdbRecAttr* rec= m_value[no_fields].rec;
2041 2042 2043 2044 2045 2046 2047 2048
      DBUG_ASSERT(rec != NULL);
      
      if (set_hidden_key(op, no_fields, rec->aRef()))
	ERR_RETURN(op->getNdbError());
    } 
    else 
    {
      int res;
2049 2050 2051 2052
      if ((res= (m_primary_key_update ?
		 set_primary_key_from_old_data(op, record)
		 : set_primary_key(op))))
	  return res;  
2053
    }
2054
  }
2055
  
2056
  // Execute delete operation
2057
  if (execute_no_commit(this,trans) != 0) {
2058
    no_uncommitted_rows_execute_failure();
2059
    DBUG_RETURN(ndb_err(trans));
2060
  }
2061 2062
  DBUG_RETURN(0);
}
2063
  
2064 2065 2066 2067 2068 2069 2070 2071 2072 2073 2074 2075 2076 2077 2078 2079 2080 2081
/*
  Unpack a record read from NDB 

  SYNOPSIS
    unpack_record()
    buf			Buffer to store read row

  NOTE
    The data for each row is read directly into the
    destination buffer. This function is primarily 
    called in order to check if any fields should be 
    set to null.
*/

void ha_ndbcluster::unpack_record(byte* buf)
{
  uint row_offset= (uint) (buf - table->record[0]);
  Field **field, **end;
pekka@mysql.com's avatar
pekka@mysql.com committed
2082
  NdbValue *value= m_value;
2083 2084 2085 2086 2087 2088 2089 2090
  DBUG_ENTER("unpack_record");
  
  // Set null flag(s)
  bzero(buf, table->null_bytes);
  for (field= table->field, end= field+table->fields;
       field < end;
       field++, value++)
  {
pekka@mysql.com's avatar
pekka@mysql.com committed
2091 2092 2093 2094 2095 2096 2097 2098 2099 2100
    if ((*value).ptr)
    {
      if (! ((*field)->flags & BLOB_FLAG))
      {
        if ((*value).rec->isNULL())
         (*field)->set_null(row_offset);
      }
      else
      {
        NdbBlob* ndb_blob= (*value).blob;
2101
        bool isNull= TRUE;
pekka@mysql.com's avatar
pekka@mysql.com committed
2102 2103 2104 2105 2106 2107
        int ret= ndb_blob->getNull(isNull);
        DBUG_ASSERT(ret == 0);
        if (isNull)
         (*field)->set_null(row_offset);
      }
    }
2108 2109 2110 2111 2112 2113 2114 2115
  }

#ifndef DBUG_OFF
  // Read and print all values that was fetched
  if (table->primary_key == MAX_KEY)
  {
    // Table with hidden primary key
    int hidden_no= table->fields;
joreland@mysql.com's avatar
joreland@mysql.com committed
2116
    const NDBTAB *tab= (const NDBTAB *) m_table;
2117
    const NDBCOL *hidden_col= tab->getColumn(hidden_no);
pekka@mysql.com's avatar
pekka@mysql.com committed
2118
    NdbRecAttr* rec= m_value[hidden_no].rec;
2119 2120 2121 2122 2123 2124 2125 2126 2127 2128 2129 2130 2131 2132 2133
    DBUG_ASSERT(rec);
    DBUG_PRINT("hidden", ("%d: %s \"%llu\"", hidden_no, 
                          hidden_col->getName(), rec->u_64_value()));
  } 
  print_results();
#endif
  DBUG_VOID_RETURN;
}

/*
  Utility function to print/dump the fetched field
 */

void ha_ndbcluster::print_results()
{
joreland@mysql.com's avatar
joreland@mysql.com committed
2134
  const NDBTAB *tab= (const NDBTAB*) m_table;
2135 2136 2137 2138 2139 2140 2141 2142 2143 2144
  DBUG_ENTER("print_results");

#ifndef DBUG_OFF
  if (!_db_on_)
    DBUG_VOID_RETURN;
  
  for (uint f=0; f<table->fields;f++)
  {
    Field *field;
    const NDBCOL *col;
pekka@mysql.com's avatar
pekka@mysql.com committed
2145
    NdbValue value;
2146

pekka@mysql.com's avatar
pekka@mysql.com committed
2147
    if (!(value= m_value[f]).ptr)
2148 2149 2150 2151 2152 2153 2154 2155
    {
      fprintf(DBUG_FILE, "Field %d was not read\n", f);
      continue;
    }
    field= table->field[f];
    DBUG_DUMP("field->ptr", (char*)field->ptr, field->pack_length());
    col= tab->getColumn(f);
    fprintf(DBUG_FILE, "%d: %s\t", f, col->getName());
pekka@mysql.com's avatar
pekka@mysql.com committed
2156 2157 2158

    NdbBlob *ndb_blob= NULL;
    if (! (field->flags & BLOB_FLAG))
2159
    {
pekka@mysql.com's avatar
pekka@mysql.com committed
2160 2161 2162 2163 2164 2165 2166 2167 2168
      if (value.rec->isNULL())
      {
        fprintf(DBUG_FILE, "NULL\n");
        continue;
      }
    }
    else
    {
      ndb_blob= value.blob;
2169
      bool isNull= TRUE;
pekka@mysql.com's avatar
pekka@mysql.com committed
2170 2171 2172 2173 2174
      ndb_blob->getNull(isNull);
      if (isNull) {
        fprintf(DBUG_FILE, "NULL\n");
        continue;
      }
2175 2176 2177 2178 2179 2180 2181 2182 2183 2184 2185 2186 2187 2188 2189 2190 2191 2192 2193 2194 2195 2196 2197 2198 2199 2200 2201 2202 2203 2204 2205 2206 2207 2208 2209 2210 2211 2212 2213 2214 2215 2216 2217 2218 2219 2220 2221 2222 2223 2224 2225 2226 2227 2228 2229 2230 2231 2232 2233 2234 2235 2236 2237 2238 2239 2240 2241 2242 2243 2244 2245
    }

    switch (col->getType()) {
    case NdbDictionary::Column::Tinyint: {
      char value= *field->ptr;
      fprintf(DBUG_FILE, "Tinyint\t%d", value);
      break;
    }
    case NdbDictionary::Column::Tinyunsigned: {
      unsigned char value= *field->ptr;
      fprintf(DBUG_FILE, "Tinyunsigned\t%u", value);
      break;
    }
    case NdbDictionary::Column::Smallint: {
      short value= *field->ptr;
      fprintf(DBUG_FILE, "Smallint\t%d", value);
      break;
    }
    case NdbDictionary::Column::Smallunsigned: {
      unsigned short value= *field->ptr;
      fprintf(DBUG_FILE, "Smallunsigned\t%u", value);
      break;
    }
    case NdbDictionary::Column::Mediumint: {
      byte value[3];
      memcpy(value, field->ptr, 3);
      fprintf(DBUG_FILE, "Mediumint\t%d,%d,%d", value[0], value[1], value[2]);
      break;
    }
    case NdbDictionary::Column::Mediumunsigned: {
      byte value[3];
      memcpy(value, field->ptr, 3);
      fprintf(DBUG_FILE, "Mediumunsigned\t%u,%u,%u", value[0], value[1], value[2]);
      break;
    }
    case NdbDictionary::Column::Int: {
      fprintf(DBUG_FILE, "Int\t%lld", field->val_int());
      break;
    }
    case NdbDictionary::Column::Unsigned: {
      Uint32 value= (Uint32) *field->ptr;
      fprintf(DBUG_FILE, "Unsigned\t%u", value);
      break;
    }
    case NdbDictionary::Column::Bigint: {
      Int64 value= (Int64) *field->ptr;
      fprintf(DBUG_FILE, "Bigint\t%lld", value);
      break;
    }
    case NdbDictionary::Column::Bigunsigned: {
      Uint64 value= (Uint64) *field->ptr;
      fprintf(DBUG_FILE, "Bigunsigned\t%llu", value);
      break;
    }
    case NdbDictionary::Column::Float: {
      float value= (float) *field->ptr;
      fprintf(DBUG_FILE, "Float\t%f", value);
      break;
    }
    case NdbDictionary::Column::Double: {
      double value= (double) *field->ptr;
      fprintf(DBUG_FILE, "Double\t%f", value);
      break;
    }
    case NdbDictionary::Column::Decimal: {
      char *value= field->ptr;

      fprintf(DBUG_FILE, "Decimal\t'%-*s'", field->pack_length(), value);
      break;
    }
    case NdbDictionary::Column::Char:{
2246 2247
      const char *value= (char *) field->ptr;
      fprintf(DBUG_FILE, "Char\t'%.*s'", field->pack_length(), value);
2248 2249 2250 2251 2252
      break;
    }
    case NdbDictionary::Column::Varchar:
    case NdbDictionary::Column::Binary:
    case NdbDictionary::Column::Varbinary: {
2253 2254
      const char *value= (char *) field->ptr;
      fprintf(DBUG_FILE, "Var\t'%.*s'", field->pack_length(), value);
2255 2256 2257 2258 2259 2260 2261
      break;
    }
    case NdbDictionary::Column::Datetime: {
      Uint64 value= (Uint64) *field->ptr;
      fprintf(DBUG_FILE, "Datetime\t%llu", value);
      break;
    }
2262
    case NdbDictionary::Column::Date: {
2263
      Uint64 value= (Uint64) *field->ptr;
2264
      fprintf(DBUG_FILE, "Date\t%llu", value);
2265 2266
      break;
    }
2267 2268 2269 2270 2271
    case NdbDictionary::Column::Time: {
      Uint64 value= (Uint64) *field->ptr;
      fprintf(DBUG_FILE, "Time\t%llu", value);
      break;
    }
pekka@mysql.com's avatar
pekka@mysql.com committed
2272 2273 2274 2275 2276 2277 2278 2279 2280 2281 2282 2283 2284 2285 2286
    case NdbDictionary::Column::Blob: {
      Uint64 len= 0;
      ndb_blob->getLength(len);
      fprintf(DBUG_FILE, "Blob\t[len=%u]", (unsigned)len);
      break;
    }
    case NdbDictionary::Column::Text: {
      Uint64 len= 0;
      ndb_blob->getLength(len);
      fprintf(DBUG_FILE, "Text\t[len=%u]", (unsigned)len);
      break;
    }
    case NdbDictionary::Column::Undefined:
      fprintf(DBUG_FILE, "Unknown type: %d", col->getType());
      break;
2287 2288 2289 2290 2291 2292 2293 2294 2295 2296 2297 2298 2299 2300 2301 2302 2303 2304 2305 2306
    }
    fprintf(DBUG_FILE, "\n");
    
  }
#endif
  DBUG_VOID_RETURN;
}


int ha_ndbcluster::index_init(uint index)
{
  DBUG_ENTER("index_init");
  DBUG_PRINT("enter", ("index: %u", index));
  DBUG_RETURN(handler::index_init(index));
}


int ha_ndbcluster::index_end()
{
  DBUG_ENTER("index_end");
2307
  DBUG_RETURN(close_scan());
2308 2309
}

2310 2311 2312 2313 2314 2315 2316 2317 2318 2319 2320 2321 2322 2323 2324 2325 2326 2327 2328 2329 2330 2331
/**
 * Check if key contains null
 */
static
int
check_null_in_key(const KEY* key_info, const byte *key, uint key_len)
{
  KEY_PART_INFO *curr_part, *end_part;
  const byte* end_ptr = key + key_len;
  curr_part= key_info->key_part;
  end_part= curr_part + key_info->key_parts;
  

  for (; curr_part != end_part && key < end_ptr; curr_part++)
  {
    if(curr_part->null_bit && *key)
      return 1;

    key += curr_part->store_length;
  }
  return 0;
}
2332 2333

int ha_ndbcluster::index_read(byte *buf,
2334 2335
			      const byte *key, uint key_len, 
			      enum ha_rkey_function find_flag)
2336 2337 2338 2339 2340
{
  DBUG_ENTER("index_read");
  DBUG_PRINT("enter", ("active_index: %u, key_len: %u, find_flag: %d", 
                       active_index, key_len, find_flag));

joreland@mysql.com's avatar
joreland@mysql.com committed
2341 2342 2343 2344 2345 2346 2347 2348
  int error;
  ndb_index_type type = get_index_type(active_index);
  const KEY* key_info = table->key_info+active_index;
  switch (type){
  case PRIMARY_KEY_ORDERED_INDEX:
  case PRIMARY_KEY_INDEX:
    if (find_flag == HA_READ_KEY_EXACT && key_info->key_length == key_len)
    {
2349 2350
      if(m_active_cursor && (error= close_scan()))
	DBUG_RETURN(error);
joreland@mysql.com's avatar
joreland@mysql.com committed
2351 2352 2353 2354 2355 2356 2357 2358 2359
      DBUG_RETURN(pk_read(key, key_len, buf));
    }
    else if (type == PRIMARY_KEY_INDEX)
    {
      DBUG_RETURN(1);
    }
    break;
  case UNIQUE_ORDERED_INDEX:
  case UNIQUE_INDEX:
2360 2361
    if (find_flag == HA_READ_KEY_EXACT && key_info->key_length == key_len &&
	!check_null_in_key(key_info, key, key_len))
joreland@mysql.com's avatar
joreland@mysql.com committed
2362
    {
2363 2364
      if(m_active_cursor && (error= close_scan()))
	DBUG_RETURN(error);
joreland@mysql.com's avatar
joreland@mysql.com committed
2365 2366 2367 2368 2369 2370 2371 2372 2373 2374 2375
      DBUG_RETURN(unique_index_read(key, key_len, buf));
    }
    else if (type == UNIQUE_INDEX)
    {
      DBUG_RETURN(1);
    }
    break;
  case ORDERED_INDEX:
    break;
  default:
  case UNDEFINED_INDEX:
2376
    DBUG_ASSERT(FALSE);
2377
    DBUG_RETURN(1);
joreland@mysql.com's avatar
joreland@mysql.com committed
2378 2379 2380
    break;
  }
  
2381
  key_range start_key;
joreland@mysql.com's avatar
joreland@mysql.com committed
2382 2383 2384
  start_key.key = key;
  start_key.length = key_len;
  start_key.flag = find_flag;
2385
  error= ordered_index_scan(&start_key, 0, TRUE, buf);  
joreland@mysql.com's avatar
joreland@mysql.com committed
2386
  DBUG_RETURN(error == HA_ERR_END_OF_FILE ? HA_ERR_KEY_NOT_FOUND : error);
2387 2388 2389 2390 2391 2392 2393 2394 2395 2396 2397 2398 2399 2400 2401 2402 2403 2404 2405
}


int ha_ndbcluster::index_read_idx(byte *buf, uint index_no, 
			      const byte *key, uint key_len, 
			      enum ha_rkey_function find_flag)
{
  statistic_increment(ha_read_key_count,&LOCK_status);
  DBUG_ENTER("index_read_idx");
  DBUG_PRINT("enter", ("index_no: %u, key_len: %u", index_no, key_len));  
  index_init(index_no);  
  DBUG_RETURN(index_read(buf, key, key_len, find_flag));
}


int ha_ndbcluster::index_next(byte *buf)
{
  DBUG_ENTER("index_next");

2406
  int error= 1;
2407
  statistic_increment(ha_read_next_count,&LOCK_status);
2408
  DBUG_RETURN(next_result(buf));
2409 2410 2411 2412 2413 2414 2415 2416 2417 2418 2419 2420 2421 2422 2423
}


int ha_ndbcluster::index_prev(byte *buf)
{
  DBUG_ENTER("index_prev");
  statistic_increment(ha_read_prev_count,&LOCK_status);
  DBUG_RETURN(1);
}


int ha_ndbcluster::index_first(byte *buf)
{
  DBUG_ENTER("index_first");
  statistic_increment(ha_read_first_count,&LOCK_status);
2424 2425 2426
  // Start the ordered index scan and fetch the first row

  // Only HA_READ_ORDER indexes get called by index_first
2427
  DBUG_RETURN(ordered_index_scan(0, 0, TRUE, buf));
2428 2429 2430 2431 2432 2433 2434
}


int ha_ndbcluster::index_last(byte *buf)
{
  DBUG_ENTER("index_last");
  statistic_increment(ha_read_last_count,&LOCK_status);
2435
  int res;
2436
  if((res= ordered_index_scan(0, 0, TRUE, buf)) == 0){
2437
    NdbResultSet *cursor= m_active_cursor; 
2438
    while((res= cursor->nextResult(TRUE, m_force_send)) == 0);
2439 2440 2441 2442 2443 2444
    if(res == 1){
      unpack_record(buf);
      table->status= 0;     
      DBUG_RETURN(0);
    }
  }
joreland@mysql.com's avatar
joreland@mysql.com committed
2445
  DBUG_RETURN(res);
2446 2447 2448
}


2449 2450 2451 2452 2453 2454 2455 2456 2457
inline
int ha_ndbcluster::read_range_first_to_buf(const key_range *start_key,
					   const key_range *end_key,
					   bool eq_range, bool sorted,
					   byte* buf)
{
  KEY* key_info;
  int error= 1; 
  DBUG_ENTER("ha_ndbcluster::read_range_first_to_buf");
2458
  DBUG_PRINT("info", ("eq_range: %d, sorted: %d", eq_range, sorted));
2459

2460
  switch (get_index_type(active_index)){
2461
  case PRIMARY_KEY_ORDERED_INDEX:
2462
  case PRIMARY_KEY_INDEX:
2463 2464 2465 2466
    key_info= table->key_info + active_index;
    if (start_key && 
	start_key->length == key_info->key_length &&
	start_key->flag == HA_READ_KEY_EXACT)
2467
    {
2468 2469
      if(m_active_cursor && (error= close_scan()))
	DBUG_RETURN(error);
2470 2471 2472
      error= pk_read(start_key->key, start_key->length, buf);      
      DBUG_RETURN(error == HA_ERR_KEY_NOT_FOUND ? HA_ERR_END_OF_FILE : error);
    }
2473
    break;
2474
  case UNIQUE_ORDERED_INDEX:
2475
  case UNIQUE_INDEX:
2476
    key_info= table->key_info + active_index;
2477 2478 2479
    if (start_key && start_key->length == key_info->key_length &&
	start_key->flag == HA_READ_KEY_EXACT && 
	!check_null_in_key(key_info, start_key->key, start_key->length))
2480
    {
2481 2482
      if(m_active_cursor && (error= close_scan()))
	DBUG_RETURN(error);
2483 2484 2485
      error= unique_index_read(start_key->key, start_key->length, buf);
      DBUG_RETURN(error == HA_ERR_KEY_NOT_FOUND ? HA_ERR_END_OF_FILE : error);
    }
2486 2487 2488 2489
    break;
  default:
    break;
  }
2490 2491

  // Start the ordered index scan and fetch the first row
joreland@mysql.com's avatar
joreland@mysql.com committed
2492
  error= ordered_index_scan(start_key, end_key, sorted, buf);
2493 2494 2495
  DBUG_RETURN(error);
}

2496

joreland@mysql.com's avatar
joreland@mysql.com committed
2497 2498 2499 2500 2501 2502 2503 2504 2505 2506 2507 2508 2509 2510
int ha_ndbcluster::read_range_first(const key_range *start_key,
				    const key_range *end_key,
				    bool eq_range, bool sorted)
{
  byte* buf= table->record[0];
  DBUG_ENTER("ha_ndbcluster::read_range_first");
  
  DBUG_RETURN(read_range_first_to_buf(start_key,
				      end_key,
				      eq_range, 
				      sorted,
				      buf));
}

2511
int ha_ndbcluster::read_range_next()
2512 2513 2514 2515 2516 2517
{
  DBUG_ENTER("ha_ndbcluster::read_range_next");
  DBUG_RETURN(next_result(table->record[0]));
}


2518 2519 2520 2521 2522
int ha_ndbcluster::rnd_init(bool scan)
{
  NdbResultSet *cursor= m_active_cursor;
  DBUG_ENTER("rnd_init");
  DBUG_PRINT("enter", ("scan: %d", scan));
2523
  // Check if scan is to be restarted
mskold@mysql.com's avatar
mskold@mysql.com committed
2524 2525 2526 2527
  if (cursor)
  {
    if (!scan)
      DBUG_RETURN(1);
2528
    int res= cursor->restart(m_force_send);
2529
    DBUG_ASSERT(res == 0);
mskold@mysql.com's avatar
mskold@mysql.com committed
2530
  }
2531 2532 2533 2534
  index_init(table->primary_key);
  DBUG_RETURN(0);
}

2535 2536 2537
int ha_ndbcluster::close_scan()
{
  NdbResultSet *cursor= m_active_cursor;
2538
  NdbConnection *trans= m_active_trans;
2539 2540 2541 2542 2543
  DBUG_ENTER("close_scan");

  if (!cursor)
    DBUG_RETURN(1);

2544
  
2545
  if (m_ops_pending)
2546 2547 2548 2549 2550
  {
    /*
      Take over any pending transactions to the 
      deleteing/updating transaction before closing the scan    
    */
2551
    DBUG_PRINT("info", ("ops_pending: %d", m_ops_pending));    
2552
    if (execute_no_commit(this,trans) != 0) {
2553
      no_uncommitted_rows_execute_failure();
2554
      DBUG_RETURN(ndb_err(trans));
2555
    }
2556
    m_ops_pending= 0;
2557 2558
  }
  
2559
  cursor->close(m_force_send);
2560
  m_active_cursor= NULL;
mskold@mysql.com's avatar
mskold@mysql.com committed
2561
  DBUG_RETURN(0);
2562
}
2563 2564 2565 2566

int ha_ndbcluster::rnd_end()
{
  DBUG_ENTER("rnd_end");
2567
  DBUG_RETURN(close_scan());
2568 2569 2570 2571 2572 2573 2574
}


int ha_ndbcluster::rnd_next(byte *buf)
{
  DBUG_ENTER("rnd_next");
  statistic_increment(ha_read_rnd_next_count, &LOCK_status);
2575

2576
  if (!m_active_cursor)
2577 2578
    DBUG_RETURN(full_table_scan(buf));
  DBUG_RETURN(next_result(buf));
2579 2580 2581 2582 2583 2584 2585 2586 2587 2588 2589 2590 2591 2592 2593 2594 2595 2596 2597 2598 2599 2600 2601 2602 2603 2604 2605 2606 2607 2608 2609 2610 2611 2612 2613 2614 2615 2616 2617 2618 2619 2620 2621 2622 2623 2624 2625 2626 2627 2628 2629 2630 2631 2632 2633 2634 2635 2636 2637 2638 2639
}


/*
  An "interesting" record has been found and it's pk 
  retrieved by calling position
  Now it's time to read the record from db once 
  again
*/

int ha_ndbcluster::rnd_pos(byte *buf, byte *pos)
{
  DBUG_ENTER("rnd_pos");
  statistic_increment(ha_read_rnd_count,&LOCK_status);
  // The primary key for the record is stored in pos
  // Perform a pk_read using primary key "index"
  DBUG_RETURN(pk_read(pos, ref_length, buf));  
}


/*
  Store the primary key of this record in ref 
  variable, so that the row can be retrieved again later
  using "reference" in rnd_pos
*/

void ha_ndbcluster::position(const byte *record)
{
  KEY *key_info;
  KEY_PART_INFO *key_part;
  KEY_PART_INFO *end;
  byte *buff;
  DBUG_ENTER("position");

  if (table->primary_key != MAX_KEY) 
  {
    key_info= table->key_info + table->primary_key;
    key_part= key_info->key_part;
    end= key_part + key_info->key_parts;
    buff= ref;
    
    for (; key_part != end; key_part++) 
    {
      if (key_part->null_bit) {
        /* Store 0 if the key part is a NULL part */      
        if (record[key_part->null_offset]
            & key_part->null_bit) {
          *buff++= 1;
          continue;
        }      
        *buff++= 0;
      }
      memcpy(buff, record + key_part->offset, key_part->length);
      buff += key_part->length;
    }
  } 
  else 
  {
    // No primary key, get hidden key
    DBUG_PRINT("info", ("Getting hidden key"));
    int hidden_no= table->fields;
pekka@mysql.com's avatar
pekka@mysql.com committed
2640
    NdbRecAttr* rec= m_value[hidden_no].rec;
joreland@mysql.com's avatar
joreland@mysql.com committed
2641
    const NDBTAB *tab= (const NDBTAB *) m_table;  
2642 2643 2644 2645 2646 2647 2648 2649 2650 2651 2652 2653 2654 2655 2656 2657 2658 2659 2660 2661 2662 2663 2664 2665 2666
    const NDBCOL *hidden_col= tab->getColumn(hidden_no);
    DBUG_ASSERT(hidden_col->getPrimaryKey() && 
                hidden_col->getAutoIncrement() &&
                rec != NULL && 
                ref_length == NDB_HIDDEN_PRIMARY_KEY_LENGTH);
    memcpy(ref, (const void*)rec->aRef(), ref_length);
  }
  
  DBUG_DUMP("ref", (char*)ref, ref_length);
  DBUG_VOID_RETURN;
}


void ha_ndbcluster::info(uint flag)
{
  DBUG_ENTER("info");
  DBUG_PRINT("enter", ("flag: %d", flag));
  
  if (flag & HA_STATUS_POS)
    DBUG_PRINT("info", ("HA_STATUS_POS"));
  if (flag & HA_STATUS_NO_LOCK)
    DBUG_PRINT("info", ("HA_STATUS_NO_LOCK"));
  if (flag & HA_STATUS_TIME)
    DBUG_PRINT("info", ("HA_STATUS_TIME"));
  if (flag & HA_STATUS_VARIABLE)
2667
  {
2668
    DBUG_PRINT("info", ("HA_STATUS_VARIABLE"));
2669 2670
    if (m_table_info)
    {
2671 2672 2673 2674
      if (m_ha_not_exact_count)
	records= 100;
      else
	records_update();
2675 2676 2677
    }
    else
    {
2678 2679 2680
      if ((my_errno= check_ndb_connection()))
        DBUG_VOID_RETURN;
      Ndb *ndb= get_ndb();
2681 2682
      Uint64 rows= 100;
      if (current_thd->variables.ndb_use_exact_count)
mskold@mysql.com's avatar
mskold@mysql.com committed
2683
	ndb_get_table_statistics(ndb, m_tabname, &rows, 0);
2684
      records= rows;
2685
    }
2686
  }
tomas@poseidon.ndb.mysql.com's avatar
tomas@poseidon.ndb.mysql.com committed
2687 2688 2689 2690 2691
  if (flag & HA_STATUS_CONST)
  {
    DBUG_PRINT("info", ("HA_STATUS_CONST"));
    set_rec_per_key();
  }
2692
  if (flag & HA_STATUS_ERRKEY)
2693
  {
2694
    DBUG_PRINT("info", ("HA_STATUS_ERRKEY"));
2695
    errkey= m_dupkey;
2696
  }
2697 2698 2699 2700 2701 2702 2703 2704 2705 2706 2707 2708 2709 2710 2711 2712 2713 2714 2715 2716 2717 2718 2719 2720 2721 2722 2723 2724 2725 2726 2727 2728 2729 2730 2731 2732 2733 2734 2735 2736 2737 2738 2739 2740 2741 2742 2743 2744 2745 2746 2747 2748 2749 2750 2751 2752 2753 2754 2755 2756 2757 2758 2759 2760 2761 2762 2763 2764 2765 2766 2767 2768 2769 2770 2771 2772 2773 2774 2775 2776 2777 2778 2779 2780 2781 2782 2783
  if (flag & HA_STATUS_AUTO)
    DBUG_PRINT("info", ("HA_STATUS_AUTO"));
  DBUG_VOID_RETURN;
}


int ha_ndbcluster::extra(enum ha_extra_function operation)
{
  DBUG_ENTER("extra");
  switch (operation) {
  case HA_EXTRA_NORMAL:              /* Optimize for space (def) */
    DBUG_PRINT("info", ("HA_EXTRA_NORMAL"));
    break;
  case HA_EXTRA_QUICK:                 /* Optimize for speed */
    DBUG_PRINT("info", ("HA_EXTRA_QUICK"));
    break;
  case HA_EXTRA_RESET:                 /* Reset database to after open */
    DBUG_PRINT("info", ("HA_EXTRA_RESET"));
    break;
  case HA_EXTRA_CACHE:                 /* Cash record in HA_rrnd() */
    DBUG_PRINT("info", ("HA_EXTRA_CACHE"));
    break;
  case HA_EXTRA_NO_CACHE:              /* End cacheing of records (def) */
    DBUG_PRINT("info", ("HA_EXTRA_NO_CACHE"));
    break;
  case HA_EXTRA_NO_READCHECK:          /* No readcheck on update */
    DBUG_PRINT("info", ("HA_EXTRA_NO_READCHECK"));
    break;
  case HA_EXTRA_READCHECK:             /* Use readcheck (def) */
    DBUG_PRINT("info", ("HA_EXTRA_READCHECK"));
    break;
  case HA_EXTRA_KEYREAD:               /* Read only key to database */
    DBUG_PRINT("info", ("HA_EXTRA_KEYREAD"));
    break;
  case HA_EXTRA_NO_KEYREAD:            /* Normal read of records (def) */
    DBUG_PRINT("info", ("HA_EXTRA_NO_KEYREAD"));
    break;
  case HA_EXTRA_NO_USER_CHANGE:        /* No user is allowed to write */
    DBUG_PRINT("info", ("HA_EXTRA_NO_USER_CHANGE"));
    break;
  case HA_EXTRA_KEY_CACHE:
    DBUG_PRINT("info", ("HA_EXTRA_KEY_CACHE"));
    break;
  case HA_EXTRA_NO_KEY_CACHE:
    DBUG_PRINT("info", ("HA_EXTRA_NO_KEY_CACHE"));
    break;
  case HA_EXTRA_WAIT_LOCK:            /* Wait until file is avalably (def) */
    DBUG_PRINT("info", ("HA_EXTRA_WAIT_LOCK"));
    break;
  case HA_EXTRA_NO_WAIT_LOCK:         /* If file is locked, return quickly */
    DBUG_PRINT("info", ("HA_EXTRA_NO_WAIT_LOCK"));
    break;
  case HA_EXTRA_WRITE_CACHE:           /* Use write cache in ha_write() */
    DBUG_PRINT("info", ("HA_EXTRA_WRITE_CACHE"));
    break;
  case HA_EXTRA_FLUSH_CACHE:           /* flush write_record_cache */
    DBUG_PRINT("info", ("HA_EXTRA_FLUSH_CACHE"));
    break;
  case HA_EXTRA_NO_KEYS:               /* Remove all update of keys */
    DBUG_PRINT("info", ("HA_EXTRA_NO_KEYS"));
    break;
  case HA_EXTRA_KEYREAD_CHANGE_POS:         /* Keyread, but change pos */
    DBUG_PRINT("info", ("HA_EXTRA_KEYREAD_CHANGE_POS")); /* xxxxchk -r must be used */
    break;                                  
  case HA_EXTRA_REMEMBER_POS:          /* Remember pos for next/prev */
    DBUG_PRINT("info", ("HA_EXTRA_REMEMBER_POS"));
    break;
  case HA_EXTRA_RESTORE_POS:
    DBUG_PRINT("info", ("HA_EXTRA_RESTORE_POS"));
    break;
  case HA_EXTRA_REINIT_CACHE:          /* init cache from current record */
    DBUG_PRINT("info", ("HA_EXTRA_REINIT_CACHE"));
    break;
  case HA_EXTRA_FORCE_REOPEN:          /* Datafile have changed on disk */
    DBUG_PRINT("info", ("HA_EXTRA_FORCE_REOPEN"));
    break;
  case HA_EXTRA_FLUSH:                 /* Flush tables to disk */
    DBUG_PRINT("info", ("HA_EXTRA_FLUSH"));
    break;
  case HA_EXTRA_NO_ROWS:               /* Don't write rows */
    DBUG_PRINT("info", ("HA_EXTRA_NO_ROWS"));
    break;
  case HA_EXTRA_RESET_STATE:           /* Reset positions */
    DBUG_PRINT("info", ("HA_EXTRA_RESET_STATE"));
    break;
  case HA_EXTRA_IGNORE_DUP_KEY:       /* Dup keys don't rollback everything*/
    DBUG_PRINT("info", ("HA_EXTRA_IGNORE_DUP_KEY"));
2784 2785 2786 2787 2788 2789
    if (current_thd->lex->sql_command == SQLCOM_REPLACE)
    {
      DBUG_PRINT("info", ("Turning ON use of write instead of insert"));
      m_use_write= TRUE;
    } else 
    {
2790 2791
      DBUG_PRINT("info", ("Ignoring duplicate key"));
      m_ignore_dup_key= TRUE;
mskold@mysql.com's avatar
mskold@mysql.com committed
2792
    }
2793 2794 2795 2796
    break;
  case HA_EXTRA_NO_IGNORE_DUP_KEY:
    DBUG_PRINT("info", ("HA_EXTRA_NO_IGNORE_DUP_KEY"));
    DBUG_PRINT("info", ("Turning OFF use of write instead of insert"));
2797
    m_use_write= FALSE;
2798
    m_ignore_dup_key= FALSE;
2799 2800 2801 2802 2803
    break;
  case HA_EXTRA_RETRIEVE_ALL_COLS:    /* Retrieve all columns, not just those
					 where field->query_id is the same as
					 the current query id */
    DBUG_PRINT("info", ("HA_EXTRA_RETRIEVE_ALL_COLS"));
2804
    m_retrieve_all_fields= TRUE;
2805 2806 2807 2808 2809 2810 2811 2812 2813 2814 2815 2816
    break;
  case HA_EXTRA_PREPARE_FOR_DELETE:
    DBUG_PRINT("info", ("HA_EXTRA_PREPARE_FOR_DELETE"));
    break;
  case HA_EXTRA_PREPARE_FOR_UPDATE:     /* Remove read cache if problems */
    DBUG_PRINT("info", ("HA_EXTRA_PREPARE_FOR_UPDATE"));
    break;
  case HA_EXTRA_PRELOAD_BUFFER_SIZE: 
    DBUG_PRINT("info", ("HA_EXTRA_PRELOAD_BUFFER_SIZE"));
    break;
  case HA_EXTRA_RETRIEVE_PRIMARY_KEY: 
    DBUG_PRINT("info", ("HA_EXTRA_RETRIEVE_PRIMARY_KEY"));
2817
    m_retrieve_primary_key= TRUE;
2818 2819 2820 2821 2822 2823 2824 2825 2826 2827 2828 2829 2830
    break;
  case HA_EXTRA_CHANGE_KEY_TO_UNIQUE: 
    DBUG_PRINT("info", ("HA_EXTRA_CHANGE_KEY_TO_UNIQUE"));
    break;
  case HA_EXTRA_CHANGE_KEY_TO_DUP: 
    DBUG_PRINT("info", ("HA_EXTRA_CHANGE_KEY_TO_DUP"));
    break;

  }
  
  DBUG_RETURN(0);
}

2831 2832 2833 2834 2835 2836 2837 2838 2839 2840 2841 2842 2843
/* 
   Start of an insert, remember number of rows to be inserted, it will
   be used in write_row and get_autoincrement to send an optimal number
   of rows in each roundtrip to the server

   SYNOPSIS
   rows     number of rows to insert, 0 if unknown

*/

void ha_ndbcluster::start_bulk_insert(ha_rows rows)
{
  int bytes, batch;
joreland@mysql.com's avatar
joreland@mysql.com committed
2844
  const NDBTAB *tab= (const NDBTAB *) m_table;    
2845 2846

  DBUG_ENTER("start_bulk_insert");
pekka@mysql.com's avatar
pekka@mysql.com committed
2847
  DBUG_PRINT("enter", ("rows: %d", (int)rows));
2848
  
2849 2850
  m_rows_inserted= 0;
  m_rows_to_insert= rows; 
2851 2852 2853 2854 2855 2856 2857 2858

  /* 
    Calculate how many rows that should be inserted
    per roundtrip to NDB. This is done in order to minimize the 
    number of roundtrips as much as possible. However performance will 
    degrade if too many bytes are inserted, thus it's limited by this 
    calculation.   
  */
2859
  const int bytesperbatch= 8192;
2860
  bytes= 12 + tab->getRowSizeInBytes() + 4 * tab->getNoOfColumns();
2861
  batch= bytesperbatch/bytes;
2862 2863
  batch= batch == 0 ? 1 : batch;
  DBUG_PRINT("info", ("batch: %d, bytes: %d", batch, bytes));
2864
  m_bulk_insert_rows= batch;
2865 2866 2867 2868 2869 2870 2871 2872 2873

  DBUG_VOID_RETURN;
}

/*
  End of an insert
 */
int ha_ndbcluster::end_bulk_insert()
{
2874 2875
  int error= 0;

2876
  DBUG_ENTER("end_bulk_insert");
2877
  // Check if last inserts need to be flushed
2878
  if (m_bulk_insert_not_flushed)
2879 2880 2881 2882 2883
  {
    NdbConnection *trans= m_active_trans;
    // Send rows to NDB
    DBUG_PRINT("info", ("Sending inserts to NDB, "\
                        "rows_inserted:%d, bulk_insert_rows: %d", 
2884 2885
                        m_rows_inserted, m_bulk_insert_rows)); 
    m_bulk_insert_not_flushed= FALSE;
2886
    if (execute_no_commit(this,trans) != 0) {
2887
      no_uncommitted_rows_execute_failure();
2888
      my_errno= error= ndb_err(trans);
2889
    }
2890 2891
  }

2892 2893
  m_rows_inserted= 0;
  m_rows_to_insert= 1;
2894
  DBUG_RETURN(error);
2895 2896
}

2897 2898 2899 2900

int ha_ndbcluster::extra_opt(enum ha_extra_function operation, ulong cache_size)
{
  DBUG_ENTER("extra_opt");
pekka@mysql.com's avatar
pekka@mysql.com committed
2901
  DBUG_PRINT("enter", ("cache_size: %lu", cache_size));
2902 2903 2904 2905 2906 2907 2908 2909 2910 2911 2912 2913 2914
  DBUG_RETURN(extra(operation));
}


int ha_ndbcluster::reset()
{
  DBUG_ENTER("reset");
  // Reset what?
  DBUG_RETURN(1);
}


const char **ha_ndbcluster::bas_ext() const
2915
{ static const char *ext[]= { ha_ndb_ext, NullS }; return ext; }
2916 2917 2918 2919 2920 2921 2922 2923 2924 2925


/*
  How many seeks it will take to read through the table
  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.
*/

double ha_ndbcluster::scan_time()
{
2926 2927 2928 2929 2930
  DBUG_ENTER("ha_ndbcluster::scan_time()");
  double res= rows2double(records*1000);
  DBUG_PRINT("exit", ("table: %s value: %f", 
		      m_tabname, res));
  DBUG_RETURN(res);
2931 2932 2933 2934 2935 2936 2937 2938 2939 2940
}


THR_LOCK_DATA **ha_ndbcluster::store_lock(THD *thd,
                                          THR_LOCK_DATA **to,
                                          enum thr_lock_type lock_type)
{
  DBUG_ENTER("store_lock");
  if (lock_type != TL_IGNORE && m_lock.type == TL_UNLOCK) 
  {
2941

2942 2943 2944
    /* If we are not doing a LOCK TABLE, then allow multiple
       writers */
    
2945 2946 2947
    /* Since NDB does not currently have table locks
       this is treated as a ordinary lock */

2948
    if ((lock_type >= TL_WRITE_ALLOW_WRITE &&
2949 2950 2951 2952 2953 2954 2955 2956 2957 2958 2959 2960 2961 2962 2963
         lock_type <= TL_WRITE) && !thd->in_lock_tables)      
      lock_type= TL_WRITE_ALLOW_WRITE;
    
    /* In queries of type INSERT INTO t1 SELECT ... FROM t2 ...
       MySQL would use the lock TL_READ_NO_INSERT on t2, and that
       would conflict with TL_WRITE_ALLOW_WRITE, blocking all inserts
       to t2. Convert the lock to a normal read lock to allow
       concurrent inserts to t2. */
    
    if (lock_type == TL_READ_NO_INSERT && !thd->in_lock_tables)
      lock_type= TL_READ;
    
    m_lock.type=lock_type;
  }
  *to++= &m_lock;
2964 2965

  DBUG_PRINT("exit", ("lock_type: %d", lock_type));
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 2995 2996 2997 2998 2999 3000 3001 3002 3003 3004 3005 3006 3007 3008
  
  DBUG_RETURN(to);
}

#ifndef DBUG_OFF
#define PRINT_OPTION_FLAGS(t) { \
      if (t->options & OPTION_NOT_AUTOCOMMIT) \
        DBUG_PRINT("thd->options", ("OPTION_NOT_AUTOCOMMIT")); \
      if (t->options & OPTION_BEGIN) \
        DBUG_PRINT("thd->options", ("OPTION_BEGIN")); \
      if (t->options & OPTION_TABLE_LOCK) \
        DBUG_PRINT("thd->options", ("OPTION_TABLE_LOCK")); \
}
#else
#define PRINT_OPTION_FLAGS(t)
#endif


/*
  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, this will be stored in transaction.stmt.
  If not, we have to start a master transaction if there doesn't exist
  one from before, this will be stored in transaction.all
 
  When a table lock is held one transaction will be started which holds
  the table lock and for each statement a hupp transaction will be started  
 */

int ha_ndbcluster::external_lock(THD *thd, int lock_type)
{
  int error=0;
  NdbConnection* trans= NULL;

  DBUG_ENTER("external_lock");
  /*
    Check that this handler instance has a connection
    set up to the Ndb object of thd
   */
  if (check_ndb_connection())
    DBUG_RETURN(1);
 
3009
  Thd_ndb *thd_ndb= (Thd_ndb*)thd->transaction.thd_ndb;
3010
  Ndb *ndb= thd_ndb->ndb;
3011 3012 3013 3014

  DBUG_PRINT("enter", ("transaction.thd_ndb->lock_count: %d", 
                       thd_ndb->lock_count));

3015 3016
  if (lock_type != F_UNLCK)
  {
3017
    DBUG_PRINT("info", ("lock_type != F_UNLCK"));
3018
    if (!thd_ndb->lock_count++)
3019 3020 3021 3022 3023 3024 3025 3026 3027
    {
      PRINT_OPTION_FLAGS(thd);

      if (!(thd->options & (OPTION_NOT_AUTOCOMMIT | OPTION_BEGIN | OPTION_TABLE_LOCK))) 
      {
        // Autocommit transaction
        DBUG_ASSERT(!thd->transaction.stmt.ndb_tid);
        DBUG_PRINT("trans",("Starting transaction stmt"));      

3028
        trans= ndb->startTransaction();
3029
        if (trans == NULL)
3030
          ERR_RETURN(ndb->getNdbError());
3031
	no_uncommitted_rows_reset(thd);
3032 3033 3034 3035 3036 3037 3038 3039 3040 3041
        thd->transaction.stmt.ndb_tid= trans;
      } 
      else 
      { 
        if (!thd->transaction.all.ndb_tid)
	{
          // Not autocommit transaction
          // A "master" transaction ha not been started yet
          DBUG_PRINT("trans",("starting transaction, all"));
          
3042
          trans= ndb->startTransaction();
3043
          if (trans == NULL)
3044
            ERR_RETURN(ndb->getNdbError());
3045
	  no_uncommitted_rows_reset(thd);
3046 3047 3048 3049 3050 3051 3052 3053 3054 3055 3056 3057 3058 3059 3060 3061 3062 3063 3064 3065 3066 3067 3068 3069 3070 3071 3072 3073 3074

          /*
            If this is the start of a LOCK TABLE, a table look 
            should be taken on the table in NDB
           
            Check if it should be read or write lock
           */
          if (thd->options & (OPTION_TABLE_LOCK))
	  {
            //lockThisTable();
            DBUG_PRINT("info", ("Locking the table..." ));
          }

          thd->transaction.all.ndb_tid= trans; 
        }
      }
    }
    /*
      This is the place to make sure this handler instance
      has a started transaction.
     
      The transaction is started by the first handler on which 
      MySQL Server calls external lock
     
      Other handlers in the same stmt or transaction should use 
      the same NDB transaction. This is done by setting up the m_active_trans
      pointer to point to the NDB transaction. 
     */

3075 3076 3077 3078 3079 3080 3081 3082
    // store thread specific data first to set the right context
    m_force_send=          thd->variables.ndb_force_send;
    m_ha_not_exact_count= !thd->variables.ndb_use_exact_count;
    m_autoincrement_prefetch= thd->variables.ndb_autoincrement_prefetch_sz;
    if (!thd->transaction.on)
      m_transaction_on= FALSE;
    else
      m_transaction_on= thd->variables.ndb_use_transactions;
3083
    //     m_use_local_query_cache= thd->variables.ndb_use_local_query_cache;
3084

3085 3086 3087 3088
    m_active_trans= thd->transaction.all.ndb_tid ? 
      (NdbConnection*)thd->transaction.all.ndb_tid:
      (NdbConnection*)thd->transaction.stmt.ndb_tid;
    DBUG_ASSERT(m_active_trans);
3089
    // Start of transaction
3090
    m_retrieve_all_fields= FALSE;
3091
    m_retrieve_primary_key= FALSE;
3092
    m_ops_pending= 0;    
3093
    {
3094
      NDBDICT *dict= ndb->getDictionary();
3095 3096 3097 3098 3099 3100 3101 3102
      const NDBTAB *tab;
      void *tab_info;
      if (!(tab= dict->getTable(m_tabname, &tab_info)))
	ERR_RETURN(dict->getNdbError());
      DBUG_PRINT("info", ("Table schema version: %d", tab->getObjectVersion()));
      m_table= (void *)tab;
      m_table_info= tab_info;
    }
3103
    no_uncommitted_rows_init(thd);
3104 3105 3106
  } 
  else 
  {
3107
    DBUG_PRINT("info", ("lock_type == F_UNLCK"));
3108
    if (!--thd_ndb->lock_count)
3109 3110 3111 3112 3113 3114 3115 3116 3117 3118 3119 3120
    {
      DBUG_PRINT("trans", ("Last external_lock"));
      PRINT_OPTION_FLAGS(thd);

      if (thd->transaction.stmt.ndb_tid)
      {
        /*
          Unlock is done without a transaction commit / rollback.
          This happens if the thread didn't update any rows
          We must in this case close the transaction to release resources
        */
        DBUG_PRINT("trans",("ending non-updating transaction"));
3121
        ndb->closeTransaction(m_active_trans);
3122 3123 3124
        thd->transaction.stmt.ndb_tid= 0;
      }
    }
tomas@poseidon.ndb.mysql.com's avatar
tomas@poseidon.ndb.mysql.com committed
3125 3126
    m_table= NULL;
    m_table_info= NULL;
3127 3128 3129 3130 3131 3132 3133 3134 3135
    /*
      This is the place to make sure this handler instance
      no longer are connected to the active transaction.

      And since the handler is no longer part of the transaction 
      it can't have open cursors, ops or blobs pending.
    */
    m_active_trans= NULL;    

3136 3137
    if (m_active_cursor)
      DBUG_PRINT("warning", ("m_active_cursor != NULL"));
3138 3139
    m_active_cursor= NULL;

3140
    if (m_blobs_pending)
3141
      DBUG_PRINT("warning", ("blobs_pending != 0"));
3142
    m_blobs_pending= 0;
3143
    
3144
    if (m_ops_pending)
3145
      DBUG_PRINT("warning", ("ops_pending != 0L"));
3146
    m_ops_pending= 0;
3147 3148 3149 3150 3151 3152 3153 3154
  }
  DBUG_RETURN(error);
}

/*
  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.
joreland@mysql.com's avatar
joreland@mysql.com committed
3155 3156
  Ndb doesn't currently support table locks, and will do ordinary
  startTransaction for each transaction/statement.
3157 3158 3159 3160 3161 3162 3163 3164 3165 3166
*/

int ha_ndbcluster::start_stmt(THD *thd)
{
  int error=0;
  DBUG_ENTER("start_stmt");
  PRINT_OPTION_FLAGS(thd);

  NdbConnection *trans= (NdbConnection*)thd->transaction.stmt.ndb_tid;
  if (!trans){
3167
    Ndb *ndb= ((Thd_ndb*)thd->transaction.thd_ndb)->ndb;
3168 3169 3170 3171
    DBUG_PRINT("trans",("Starting transaction stmt"));  
    
    NdbConnection *tablock_trans= 
      (NdbConnection*)thd->transaction.all.ndb_tid;
joreland@mysql.com's avatar
joreland@mysql.com committed
3172
    DBUG_PRINT("info", ("tablock_trans: %x", (UintPtr)tablock_trans));
joreland@mysql.com's avatar
joreland@mysql.com committed
3173
    DBUG_ASSERT(tablock_trans);
3174 3175
//    trans= ndb->hupp(tablock_trans);
    trans= ndb->startTransaction();
3176
    if (trans == NULL)
3177
      ERR_RETURN(ndb->getNdbError());
3178
    no_uncommitted_rows_reset(thd);
3179 3180 3181
    thd->transaction.stmt.ndb_tid= trans;
  }
  m_active_trans= trans;
3182

3183
  // Start of statement
3184
  m_retrieve_all_fields= FALSE;
3185
  m_retrieve_primary_key= FALSE;
3186
  m_ops_pending= 0;    
3187 3188 3189 3190 3191 3192 3193 3194 3195 3196 3197 3198
  
  DBUG_RETURN(error);
}


/*
  Commit a transaction started in NDB 
 */

int ndbcluster_commit(THD *thd, void *ndb_transaction)
{
  int res= 0;
3199
  Ndb *ndb= ((Thd_ndb*)thd->transaction.thd_ndb)->ndb;
3200 3201 3202 3203 3204 3205 3206 3207
  NdbConnection *trans= (NdbConnection*)ndb_transaction;

  DBUG_ENTER("ndbcluster_commit");
  DBUG_PRINT("transaction",("%s",
                            trans == thd->transaction.stmt.ndb_tid ? 
                            "stmt" : "all"));
  DBUG_ASSERT(ndb && trans);

3208
  if (execute_commit(thd,trans) != 0)
3209 3210
  {
    const NdbError err= trans->getNdbError();
3211
    const NdbOperation *error_op= trans->getNdbErrorOperation();
3212 3213
    ERR_PRINT(err);     
    res= ndb_to_mysql_error(&err);
3214
    if (res != -1) 
3215
      ndbcluster_print_error(res, error_op);
3216
  }
3217
  ndb->closeTransaction(trans);
3218 3219 3220 3221 3222 3223 3224 3225 3226 3227 3228
  DBUG_RETURN(res);
}


/*
  Rollback a transaction started in NDB
 */

int ndbcluster_rollback(THD *thd, void *ndb_transaction)
{
  int res= 0;
3229
  Ndb *ndb= ((Thd_ndb*)thd->transaction.thd_ndb)->ndb;
3230 3231 3232 3233 3234 3235 3236 3237 3238 3239 3240
  NdbConnection *trans= (NdbConnection*)ndb_transaction;

  DBUG_ENTER("ndbcluster_rollback");
  DBUG_PRINT("transaction",("%s",
                            trans == thd->transaction.stmt.ndb_tid ? 
                            "stmt" : "all"));
  DBUG_ASSERT(ndb && trans);

  if (trans->execute(Rollback) != 0)
  {
    const NdbError err= trans->getNdbError();
3241
    const NdbOperation *error_op= trans->getNdbErrorOperation();
3242 3243
    ERR_PRINT(err);     
    res= ndb_to_mysql_error(&err);
3244 3245
    if (res != -1) 
      ndbcluster_print_error(res, error_op);
3246 3247 3248 3249 3250 3251 3252
  }
  ndb->closeTransaction(trans);
  DBUG_RETURN(0);
}


/*
pekka@mysql.com's avatar
pekka@mysql.com committed
3253 3254 3255
  Define NDB column based on Field.
  Returns 0 or mysql error code.
  Not member of ha_ndbcluster because NDBCOL cannot be declared.
3256 3257
 */

pekka@mysql.com's avatar
pekka@mysql.com committed
3258 3259 3260
static int create_ndb_column(NDBCOL &col,
                             Field *field,
                             HA_CREATE_INFO *info)
3261
{
pekka@mysql.com's avatar
pekka@mysql.com committed
3262
  // Set name
3263 3264 3265 3266 3267 3268
  {
    char truncated_field_name[NDB_MAX_ATTR_NAME_SIZE];
    strnmov(truncated_field_name,field->field_name,sizeof(truncated_field_name));
    truncated_field_name[sizeof(truncated_field_name)-1]= '\0';
    col.setName(truncated_field_name);
  }
pekka@mysql.com's avatar
pekka@mysql.com committed
3269 3270
  // Get char set
  CHARSET_INFO *cs= field->charset();
pekka@mysql.com's avatar
pekka@mysql.com committed
3271 3272 3273 3274
  // Set type and sizes
  const enum enum_field_types mysql_type= field->real_type();
  switch (mysql_type) {
  // Numeric types
3275
  case MYSQL_TYPE_DECIMAL:    
pekka@mysql.com's avatar
pekka@mysql.com committed
3276 3277 3278
    col.setType(NDBCOL::Char);
    col.setLength(field->pack_length());
    break;
3279
  case MYSQL_TYPE_TINY:        
pekka@mysql.com's avatar
pekka@mysql.com committed
3280 3281 3282 3283 3284 3285
    if (field->flags & UNSIGNED_FLAG)
      col.setType(NDBCOL::Tinyunsigned);
    else
      col.setType(NDBCOL::Tinyint);
    col.setLength(1);
    break;
3286
  case MYSQL_TYPE_SHORT:
pekka@mysql.com's avatar
pekka@mysql.com committed
3287 3288 3289 3290 3291 3292
    if (field->flags & UNSIGNED_FLAG)
      col.setType(NDBCOL::Smallunsigned);
    else
      col.setType(NDBCOL::Smallint);
    col.setLength(1);
    break;
3293
  case MYSQL_TYPE_LONG:
pekka@mysql.com's avatar
pekka@mysql.com committed
3294 3295 3296 3297 3298 3299
    if (field->flags & UNSIGNED_FLAG)
      col.setType(NDBCOL::Unsigned);
    else
      col.setType(NDBCOL::Int);
    col.setLength(1);
    break;
3300
  case MYSQL_TYPE_INT24:       
pekka@mysql.com's avatar
pekka@mysql.com committed
3301 3302 3303 3304 3305 3306 3307 3308 3309 3310 3311 3312
    if (field->flags & UNSIGNED_FLAG)
      col.setType(NDBCOL::Mediumunsigned);
    else
      col.setType(NDBCOL::Mediumint);
    col.setLength(1);
    break;
  case MYSQL_TYPE_LONGLONG:
    if (field->flags & UNSIGNED_FLAG)
      col.setType(NDBCOL::Bigunsigned);
    else
      col.setType(NDBCOL::Bigint);
    col.setLength(1);
3313 3314
    break;
  case MYSQL_TYPE_FLOAT:
pekka@mysql.com's avatar
pekka@mysql.com committed
3315 3316 3317
    col.setType(NDBCOL::Float);
    col.setLength(1);
    break;
3318
  case MYSQL_TYPE_DOUBLE:
pekka@mysql.com's avatar
pekka@mysql.com committed
3319 3320 3321 3322 3323 3324 3325 3326 3327 3328 3329 3330 3331
    col.setType(NDBCOL::Double);
    col.setLength(1);
    break;
  // Date types
  case MYSQL_TYPE_TIMESTAMP:
    col.setType(NDBCOL::Unsigned);
    col.setLength(1);
    break;
  case MYSQL_TYPE_DATETIME:    
    col.setType(NDBCOL::Datetime);
    col.setLength(1);
    break;
  case MYSQL_TYPE_NEWDATE:
3332 3333 3334
    col.setType(NDBCOL::Date);
    col.setLength(1);
    break;
pekka@mysql.com's avatar
pekka@mysql.com committed
3335
  case MYSQL_TYPE_TIME:        
3336 3337 3338 3339
    col.setType(NDBCOL::Time);
    col.setLength(1);
    break;
  case MYSQL_TYPE_DATE: // ?
pekka@mysql.com's avatar
pekka@mysql.com committed
3340 3341 3342 3343 3344 3345 3346 3347
  case MYSQL_TYPE_YEAR:        
    col.setType(NDBCOL::Char);
    col.setLength(field->pack_length());
    break;
  // Char types
  case MYSQL_TYPE_STRING:      
    if (field->flags & BINARY_FLAG)
      col.setType(NDBCOL::Binary);
pekka@mysql.com's avatar
pekka@mysql.com committed
3348
    else {
pekka@mysql.com's avatar
pekka@mysql.com committed
3349
      col.setType(NDBCOL::Char);
pekka@mysql.com's avatar
pekka@mysql.com committed
3350 3351
      col.setCharset(cs);
    }
3352 3353 3354 3355
    if (field->pack_length() == 0)
      col.setLength(1); // currently ndb does not support size 0
    else
      col.setLength(field->pack_length());
pekka@mysql.com's avatar
pekka@mysql.com committed
3356 3357 3358 3359
    break;
  case MYSQL_TYPE_VAR_STRING:
    if (field->flags & BINARY_FLAG)
      col.setType(NDBCOL::Varbinary);
pekka@mysql.com's avatar
pekka@mysql.com committed
3360
    else {
pekka@mysql.com's avatar
pekka@mysql.com committed
3361
      col.setType(NDBCOL::Varchar);
pekka@mysql.com's avatar
pekka@mysql.com committed
3362 3363
      col.setCharset(cs);
    }
pekka@mysql.com's avatar
pekka@mysql.com committed
3364 3365 3366 3367 3368 3369 3370
    col.setLength(field->pack_length());
    break;
  // Blob types (all come in as MYSQL_TYPE_BLOB)
  mysql_type_tiny_blob:
  case MYSQL_TYPE_TINY_BLOB:
    if (field->flags & BINARY_FLAG)
      col.setType(NDBCOL::Blob);
pekka@mysql.com's avatar
pekka@mysql.com committed
3371
    else {
pekka@mysql.com's avatar
pekka@mysql.com committed
3372
      col.setType(NDBCOL::Text);
pekka@mysql.com's avatar
pekka@mysql.com committed
3373 3374
      col.setCharset(cs);
    }
pekka@mysql.com's avatar
pekka@mysql.com committed
3375 3376 3377 3378 3379 3380 3381 3382 3383
    col.setInlineSize(256);
    // No parts
    col.setPartSize(0);
    col.setStripeSize(0);
    break;
  mysql_type_blob:
  case MYSQL_TYPE_BLOB:    
    if (field->flags & BINARY_FLAG)
      col.setType(NDBCOL::Blob);
pekka@mysql.com's avatar
pekka@mysql.com committed
3384
    else {
pekka@mysql.com's avatar
pekka@mysql.com committed
3385
      col.setType(NDBCOL::Text);
pekka@mysql.com's avatar
pekka@mysql.com committed
3386 3387
      col.setCharset(cs);
    }
pekka@mysql.com's avatar
pekka@mysql.com committed
3388 3389 3390 3391 3392 3393 3394 3395 3396 3397 3398 3399 3400 3401 3402 3403 3404 3405
    // Use "<=" even if "<" is the exact condition
    if (field->max_length() <= (1 << 8))
      goto mysql_type_tiny_blob;
    else if (field->max_length() <= (1 << 16))
    {
      col.setInlineSize(256);
      col.setPartSize(2000);
      col.setStripeSize(16);
    }
    else if (field->max_length() <= (1 << 24))
      goto mysql_type_medium_blob;
    else
      goto mysql_type_long_blob;
    break;
  mysql_type_medium_blob:
  case MYSQL_TYPE_MEDIUM_BLOB:   
    if (field->flags & BINARY_FLAG)
      col.setType(NDBCOL::Blob);
pekka@mysql.com's avatar
pekka@mysql.com committed
3406
    else {
pekka@mysql.com's avatar
pekka@mysql.com committed
3407
      col.setType(NDBCOL::Text);
pekka@mysql.com's avatar
pekka@mysql.com committed
3408 3409
      col.setCharset(cs);
    }
pekka@mysql.com's avatar
pekka@mysql.com committed
3410 3411 3412 3413 3414 3415 3416 3417
    col.setInlineSize(256);
    col.setPartSize(4000);
    col.setStripeSize(8);
    break;
  mysql_type_long_blob:
  case MYSQL_TYPE_LONG_BLOB:  
    if (field->flags & BINARY_FLAG)
      col.setType(NDBCOL::Blob);
pekka@mysql.com's avatar
pekka@mysql.com committed
3418
    else {
pekka@mysql.com's avatar
pekka@mysql.com committed
3419
      col.setType(NDBCOL::Text);
pekka@mysql.com's avatar
pekka@mysql.com committed
3420 3421
      col.setCharset(cs);
    }
pekka@mysql.com's avatar
pekka@mysql.com committed
3422 3423 3424 3425 3426 3427 3428 3429 3430 3431 3432 3433 3434 3435 3436 3437 3438 3439 3440
    col.setInlineSize(256);
    col.setPartSize(8000);
    col.setStripeSize(4);
    break;
  // Other types
  case MYSQL_TYPE_ENUM:
    col.setType(NDBCOL::Char);
    col.setLength(field->pack_length());
    break;
  case MYSQL_TYPE_SET:         
    col.setType(NDBCOL::Char);
    col.setLength(field->pack_length());
    break;
  case MYSQL_TYPE_NULL:        
  case MYSQL_TYPE_GEOMETRY:
    goto mysql_type_unsupported;
  mysql_type_unsupported:
  default:
    return HA_ERR_UNSUPPORTED;
3441
  }
pekka@mysql.com's avatar
pekka@mysql.com committed
3442 3443 3444 3445 3446 3447 3448 3449
  // Set nullable and pk
  col.setNullable(field->maybe_null());
  col.setPrimaryKey(field->flags & PRI_KEY_FLAG);
  // Set autoincrement
  if (field->flags & AUTO_INCREMENT_FLAG) 
  {
    col.setAutoIncrement(TRUE);
    ulonglong value= info->auto_increment_value ?
3450
      info->auto_increment_value : (ulonglong) 1;
pekka@mysql.com's avatar
pekka@mysql.com committed
3451 3452
    DBUG_PRINT("info", ("Autoincrement key, initial: %llu", value));
    col.setAutoIncrementInitialValue(value);
3453
  }
pekka@mysql.com's avatar
pekka@mysql.com committed
3454
  else
3455
    col.setAutoIncrement(FALSE);
pekka@mysql.com's avatar
pekka@mysql.com committed
3456
  return 0;
3457 3458 3459 3460 3461 3462 3463 3464 3465 3466 3467 3468
}

/*
  Create a table in NDB Cluster
 */

int ha_ndbcluster::create(const char *name, 
			  TABLE *form, 
			  HA_CREATE_INFO *info)
{
  NDBTAB tab;
  NDBCOL col;
joreland@mysql.com's avatar
joreland@mysql.com committed
3469
  uint pack_length, length, i, pk_length= 0;
3470
  const void *data, *pack_data;
3471
  const char **key_names= form->keynames.type_names;
3472
  char name2[FN_HEADLEN];
3473
  bool create_from_engine= (info->table_options & HA_CREATE_FROM_ENGINE);
3474 3475 3476 3477 3478
   
  DBUG_ENTER("create");
  DBUG_PRINT("enter", ("name: %s", name));
  fn_format(name2, name, "", "",2);       // Remove the .frm extension
  set_dbname(name2);
3479 3480 3481 3482 3483 3484 3485 3486 3487 3488 3489 3490
  set_tabname(name2);    

  if (create_from_engine)
  {
    /*
      Table alreay exists in NDB and frm file has been created by 
      caller.
      Do Ndb specific stuff, such as create a .ndb file
    */
    my_errno= write_ndb_file();
    DBUG_RETURN(my_errno);
  }
3491 3492 3493 3494 3495 3496 3497 3498 3499 3500 3501 3502 3503 3504 3505 3506 3507 3508 3509 3510 3511 3512

  DBUG_PRINT("table", ("name: %s", m_tabname));  
  tab.setName(m_tabname);
  tab.setLogging(!(info->options & HA_LEX_CREATE_TMP_TABLE));    
   
  // Save frm data for this table
  if (readfrm(name, &data, &length))
    DBUG_RETURN(1);
  if (packfrm(data, length, &pack_data, &pack_length))
    DBUG_RETURN(2);
  
  DBUG_PRINT("info", ("setFrm data=%x, len=%d", pack_data, pack_length));
  tab.setFrm(pack_data, pack_length);      
  my_free((char*)data, MYF(0));
  my_free((char*)pack_data, MYF(0));
  
  for (i= 0; i < form->fields; i++) 
  {
    Field *field= form->field[i];
    DBUG_PRINT("info", ("name: %s, type: %u, pack_length: %d", 
                        field->field_name, field->real_type(),
			field->pack_length()));
3513
    if ((my_errno= create_ndb_column(col, field, info)))
pekka@mysql.com's avatar
pekka@mysql.com committed
3514
      DBUG_RETURN(my_errno);
3515
    tab.addColumn(col);
joreland@mysql.com's avatar
joreland@mysql.com committed
3516 3517
    if(col.getPrimaryKey())
      pk_length += (field->pack_length() + 3) / 4;
3518 3519 3520 3521 3522 3523 3524 3525 3526
  }
  
  // No primary key, create shadow key as 64 bit, auto increment  
  if (form->primary_key == MAX_KEY) 
  {
    DBUG_PRINT("info", ("Generating shadow key"));
    col.setName("$PK");
    col.setType(NdbDictionary::Column::Bigunsigned);
    col.setLength(1);
3527
    col.setNullable(FALSE);
3528 3529 3530
    col.setPrimaryKey(TRUE);
    col.setAutoIncrement(TRUE);
    tab.addColumn(col);
joreland@mysql.com's avatar
joreland@mysql.com committed
3531 3532 3533 3534 3535 3536 3537 3538 3539 3540 3541 3542 3543 3544 3545 3546 3547 3548 3549 3550 3551 3552 3553 3554 3555 3556 3557 3558 3559 3560 3561 3562 3563
    pk_length += 2;
  }
  
  // Make sure that blob tables don't have to big part size
  for (i= 0; i < form->fields; i++) 
  {
    /**
     * The extra +7 concists
     * 2 - words from pk in blob table
     * 5 - from extra words added by tup/dict??
     */
    switch (form->field[i]->real_type()) {
    case MYSQL_TYPE_BLOB:    
    case MYSQL_TYPE_MEDIUM_BLOB:   
    case MYSQL_TYPE_LONG_BLOB: 
    {
      NdbDictionary::Column * col = tab.getColumn(i);
      int size = pk_length + (col->getPartSize()+3)/4 + 7;
      if(size > NDB_MAX_TUPLE_SIZE_IN_WORDS && 
	 (pk_length+7) < NDB_MAX_TUPLE_SIZE_IN_WORDS)
      {
	size = NDB_MAX_TUPLE_SIZE_IN_WORDS - pk_length - 7;
	col->setPartSize(4*size);
      }
      /**
       * If size > NDB_MAX and pk_length+7 >= NDB_MAX
       *   then the table can't be created anyway, so skip
       *   changing part size, and have error later
       */ 
    }
    default:
      break;
    }
3564 3565
  }
  
3566
  if ((my_errno= check_ndb_connection()))
3567 3568 3569
    DBUG_RETURN(my_errno);
  
  // Create the table in NDB     
3570 3571
  Ndb *ndb= get_ndb();
  NDBDICT *dict= ndb->getDictionary();
3572
  if (dict->createTable(tab) != 0) 
3573 3574 3575 3576 3577 3578 3579 3580
  {
    const NdbError err= dict->getNdbError();
    ERR_PRINT(err);
    my_errno= ndb_to_mysql_error(&err);
    DBUG_RETURN(my_errno);
  }
  DBUG_PRINT("info", ("Table %s/%s created successfully", 
                      m_dbname, m_tabname));
3581

3582 3583
  // Create secondary indexes
  my_errno= build_index_list(form, ILBP_CREATE);
3584

3585 3586 3587
  if (!my_errno)
    my_errno= write_ndb_file();

3588 3589 3590 3591
  DBUG_RETURN(my_errno);
}


3592 3593 3594 3595
int ha_ndbcluster::create_ordered_index(const char *name, 
					KEY *key_info)
{
  DBUG_ENTER("create_ordered_index");
3596
  DBUG_RETURN(create_index(name, key_info, FALSE));
3597 3598 3599 3600 3601 3602
}

int ha_ndbcluster::create_unique_index(const char *name, 
				       KEY *key_info)
{

3603
  DBUG_ENTER("create_unique_index");
3604
  DBUG_RETURN(create_index(name, key_info, TRUE));
3605 3606 3607
}


3608 3609 3610 3611 3612
/*
  Create an index in NDB Cluster
 */

int ha_ndbcluster::create_index(const char *name, 
3613 3614 3615
				KEY *key_info,
				bool unique)
{
3616 3617
  Ndb *ndb= get_ndb();
  NdbDictionary::Dictionary *dict= ndb->getDictionary();
3618 3619 3620 3621 3622
  KEY_PART_INFO *key_part= key_info->key_part;
  KEY_PART_INFO *end= key_part + key_info->key_parts;
  
  DBUG_ENTER("create_index");
  DBUG_PRINT("enter", ("name: %s ", name));
3623

3624
  NdbDictionary::Index ndb_index(name);
3625
  if (unique)
3626 3627 3628 3629 3630
    ndb_index.setType(NdbDictionary::Index::UniqueHashIndex);
  else 
  {
    ndb_index.setType(NdbDictionary::Index::OrderedIndex);
    // TODO Only temporary ordered indexes supported
3631
    ndb_index.setLogging(FALSE); 
3632 3633 3634 3635 3636 3637 3638
  }
  ndb_index.setTable(m_tabname);

  for (; key_part != end; key_part++) 
  {
    Field *field= key_part->field;
    DBUG_PRINT("info", ("attr: %s", field->field_name));
3639 3640 3641 3642 3643 3644
    {
      char truncated_field_name[NDB_MAX_ATTR_NAME_SIZE];
      strnmov(truncated_field_name,field->field_name,sizeof(truncated_field_name));
      truncated_field_name[sizeof(truncated_field_name)-1]= '\0';
      ndb_index.addColumnName(truncated_field_name);
    }
3645 3646 3647 3648 3649 3650 3651 3652 3653 3654 3655 3656 3657 3658 3659 3660 3661
  }
  
  if (dict->createIndex(ndb_index))
    ERR_RETURN(dict->getNdbError());

  // Success
  DBUG_PRINT("info", ("Created index %s", name));
  DBUG_RETURN(0);  
}


/*
  Rename a table in NDB Cluster
*/

int ha_ndbcluster::rename_table(const char *from, const char *to)
{
3662
  NDBDICT *dict;
3663
  char new_tabname[FN_HEADLEN];
3664 3665
  const NDBTAB *orig_tab;
  int result;
3666 3667

  DBUG_ENTER("ha_ndbcluster::rename_table");
3668
  DBUG_PRINT("info", ("Renaming %s to %s", from, to));
3669 3670 3671 3672
  set_dbname(from);
  set_tabname(from);
  set_tabname(to, new_tabname);

3673 3674
  if (check_ndb_connection())
    DBUG_RETURN(my_errno= HA_ERR_NO_CONNECTION);
mskold@mysql.com's avatar
mskold@mysql.com committed
3675 3676 3677
  
  Ndb *ndb= get_ndb();
  dict= ndb->getDictionary();
3678 3679
  if (!(orig_tab= dict->getTable(m_tabname)))
    ERR_RETURN(dict->getNdbError());
3680

3681 3682 3683
  m_table= (void *)orig_tab;
  // Change current database to that of target table
  set_dbname(to);
mskold@mysql.com's avatar
mskold@mysql.com committed
3684
  ndb->setDatabaseName(m_dbname);
3685
  if (!(result= alter_table_name(new_tabname)))
3686
  {
3687 3688
    // Rename .ndb file
    result= handler::rename_table(from, to);
3689
  }
3690

3691 3692 3693 3694 3695 3696 3697 3698
  DBUG_RETURN(result);
}


/*
  Rename a table in NDB Cluster using alter table
 */

3699
int ha_ndbcluster::alter_table_name(const char *to)
3700
{
3701 3702
  Ndb *ndb= get_ndb();
  NDBDICT *dict= ndb->getDictionary();
3703 3704
  const NDBTAB *orig_tab= (const NDBTAB *) m_table;
  int ret;
3705 3706
  DBUG_ENTER("alter_table_name_table");

3707
  NdbDictionary::Table new_tab= *orig_tab;
3708 3709
  new_tab.setName(to);
  if (dict->alterTable(new_tab) != 0)
3710 3711 3712
    ERR_RETURN(dict->getNdbError());

  m_table= NULL;
tomas@poseidon.ndb.mysql.com's avatar
tomas@poseidon.ndb.mysql.com committed
3713
  m_table_info= NULL;
3714 3715 3716 3717 3718 3719 3720 3721 3722 3723 3724 3725 3726 3727 3728 3729 3730 3731
                                                                             
  DBUG_RETURN(0);
}


/*
  Delete a table from NDB Cluster
 */

int ha_ndbcluster::delete_table(const char *name)
{
  DBUG_ENTER("delete_table");
  DBUG_PRINT("enter", ("name: %s", name));
  set_dbname(name);
  set_tabname(name);
  
  if (check_ndb_connection())
    DBUG_RETURN(HA_ERR_NO_CONNECTION);
3732
  // Remove .ndb file
3733
  handler::delete_table(name);
3734 3735 3736 3737 3738 3739 3740 3741 3742 3743
  DBUG_RETURN(drop_table());
}


/*
  Drop a table in NDB Cluster
 */

int ha_ndbcluster::drop_table()
{
3744 3745 3746
  Ndb *ndb= get_ndb();
  NdbDictionary::Dictionary *dict= ndb->getDictionary();
  
3747 3748 3749 3750 3751 3752 3753 3754 3755 3756 3757 3758 3759 3760 3761 3762 3763 3764 3765 3766 3767 3768 3769 3770 3771 3772 3773 3774 3775
  DBUG_ENTER("drop_table");
  DBUG_PRINT("enter", ("Deleting %s", m_tabname));
  
  if (dict->dropTable(m_tabname)) 
  {
    const NdbError err= dict->getNdbError();
    if (err.code == 709)
      ; // 709: No such table existed
    else 
      ERR_RETURN(dict->getNdbError());
  }  
  release_metadata();
  DBUG_RETURN(0);
}


/*
  Drop a database in NDB Cluster
 */

int ndbcluster_drop_database(const char *path)
{
  DBUG_ENTER("ndbcluster_drop_database");
  // TODO drop all tables for this database
  DBUG_RETURN(1);
}


longlong ha_ndbcluster::get_auto_increment()
3776
{  
3777 3778
  DBUG_ENTER("get_auto_increment");
  DBUG_PRINT("enter", ("m_tabname: %s", m_tabname));
3779
  Ndb *ndb= get_ndb();
3780
  int cache_size= 
3781
    (m_rows_to_insert - m_rows_inserted < m_autoincrement_prefetch) ?
3782
    m_rows_to_insert - m_rows_inserted 
3783
    : (m_rows_to_insert > m_autoincrement_prefetch) ? 
3784
    m_rows_to_insert 
3785
    : m_autoincrement_prefetch;
3786
  Uint64 auto_value= 
3787
    (m_skip_auto_increment) ? 
3788 3789
    ndb->readAutoIncrementValue((const NDBTAB *) m_table)
    : ndb->getAutoIncrementValue((const NDBTAB *) m_table, cache_size);
3790
  DBUG_RETURN((longlong)auto_value);
3791 3792 3793 3794 3795 3796 3797 3798 3799 3800 3801 3802
}


/*
  Constructor for the NDB Cluster table handler 
 */

ha_ndbcluster::ha_ndbcluster(TABLE *table_arg):
  handler(table_arg),
  m_active_trans(NULL),
  m_active_cursor(NULL),
  m_table(NULL),
3803
  m_table_info(NULL),
3804
  m_table_flags(HA_REC_NOT_IN_SEQ |
3805
		HA_NULL_IN_KEY |
3806 3807
		HA_AUTO_PART_KEY |
		HA_NO_PREFIX_CHAR_KEYS),
3808
  m_share(0),
3809
  m_use_write(FALSE),
3810
  m_ignore_dup_key(FALSE),
3811 3812
  m_primary_key_update(FALSE),
  m_retrieve_all_fields(FALSE),
3813
  m_retrieve_primary_key(FALSE),
3814 3815 3816 3817 3818 3819 3820 3821 3822
  m_rows_to_insert(1),
  m_rows_inserted(0),
  m_bulk_insert_rows(1024),
  m_bulk_insert_not_flushed(FALSE),
  m_ops_pending(0),
  m_skip_auto_increment(TRUE),
  m_blobs_pending(0),
  m_blobs_buffer(0),
  m_blobs_buffer_size(0),
3823 3824 3825 3826
  m_dupkey((uint) -1),
  m_ha_not_exact_count(FALSE),
  m_force_send(TRUE),
  m_autoincrement_prefetch(32),
3827 3828
  m_transaction_on(TRUE),
  m_use_local_query_cache(FALSE)
3829
{ 
3830 3831
  int i;
  
3832 3833 3834 3835 3836
  DBUG_ENTER("ha_ndbcluster");

  m_tabname[0]= '\0';
  m_dbname[0]= '\0';

3837
  records= ~(ha_rows)0; // uninitialized
3838 3839
  block_size= 1024;

3840 3841
  for (i= 0; i < MAX_KEY; i++)
  {
3842 3843 3844
    m_index[i].type= UNDEFINED_INDEX;   
    m_index[i].unique_index= NULL;      
    m_index[i].index= NULL;      
3845 3846
  }

3847 3848 3849 3850 3851 3852 3853 3854 3855 3856 3857 3858
  DBUG_VOID_RETURN;
}


/*
  Destructor for NDB Cluster table handler
 */

ha_ndbcluster::~ha_ndbcluster() 
{
  DBUG_ENTER("~ha_ndbcluster");

3859 3860
  if (m_share)
    free_share(m_share);
3861
  release_metadata();
3862 3863
  my_free(m_blobs_buffer, MYF(MY_ALLOW_ZERO_PTR));
  m_blobs_buffer= 0;
3864 3865

  // Check for open cursor/transaction
3866 3867
  if (m_active_cursor) {
  }
3868
  DBUG_ASSERT(m_active_cursor == NULL);
3869 3870
  if (m_active_trans) {
  }
3871 3872 3873 3874 3875 3876 3877 3878 3879 3880 3881 3882 3883 3884
  DBUG_ASSERT(m_active_trans == NULL);

  DBUG_VOID_RETURN;
}


/*
  Open a table for further use
  - fetch metadata for this table from NDB
  - check that table exists
*/

int ha_ndbcluster::open(const char *name, int mode, uint test_if_locked)
{
tomas@poseidon.ndb.mysql.com's avatar
tomas@poseidon.ndb.mysql.com committed
3885
  int res;
3886 3887 3888 3889 3890 3891 3892 3893 3894 3895 3896 3897 3898 3899 3900 3901 3902 3903 3904 3905 3906 3907
  KEY *key;
  DBUG_ENTER("open");
  DBUG_PRINT("enter", ("name: %s mode: %d test_if_locked: %d",
                       name, mode, test_if_locked));
  
  // Setup ref_length to make room for the whole 
  // primary key to be written in the ref variable
  
  if (table->primary_key != MAX_KEY) 
  {
    key= table->key_info+table->primary_key;
    ref_length= key->key_length;
    DBUG_PRINT("info", (" ref_length: %d", ref_length));
  }
  // Init table lock structure 
  if (!(m_share=get_share(name)))
    DBUG_RETURN(1);
  thr_lock_data_init(&m_share->lock,&m_lock,(void*) 0);
  
  set_dbname(name);
  set_tabname(name);
  
3908 3909
  if (check_ndb_connection()) {
    free_share(m_share); m_share= 0;
3910
    DBUG_RETURN(HA_ERR_NO_CONNECTION);
3911
  }
3912
  
tomas@poseidon.ndb.mysql.com's avatar
tomas@poseidon.ndb.mysql.com committed
3913 3914 3915 3916 3917
  res= get_metadata(name);
  if (!res)
    info(HA_STATUS_VARIABLE | HA_STATUS_CONST);

  DBUG_RETURN(res);
3918 3919 3920 3921 3922 3923 3924 3925 3926 3927 3928
}


/*
  Close the table
  - release resources setup by open()
 */

int ha_ndbcluster::close(void)
{
  DBUG_ENTER("close");  
3929
  free_share(m_share); m_share= 0;
3930 3931 3932 3933 3934
  release_metadata();
  DBUG_RETURN(0);
}


3935
Thd_ndb* ha_ndbcluster::seize_thd_ndb()
3936
{
3937 3938
  Thd_ndb *thd_ndb;
  DBUG_ENTER("seize_thd_ndb");
3939

3940 3941 3942
  thd_ndb= new Thd_ndb();
  thd_ndb->ndb->getDictionary()->set_local_table_data_size(sizeof(Ndb_table_local_info));
  if (thd_ndb->ndb->init(max_transactions) != 0)
3943
  {
3944
    ERR_PRINT(thd_ndb->ndb->getNdbError());
3945 3946 3947 3948 3949 3950
    /*
      TODO 
      Alt.1 If init fails because to many allocated Ndb 
      wait on condition for a Ndb object to be released.
      Alt.2 Seize/release from pool, wait until next release 
    */
3951 3952
    delete thd_ndb;
    thd_ndb= NULL;
3953
  }
3954
  DBUG_RETURN(thd_ndb);
3955 3956 3957
}


3958
void ha_ndbcluster::release_thd_ndb(Thd_ndb* thd_ndb)
3959
{
3960 3961
  DBUG_ENTER("release_thd_ndb");
  delete thd_ndb;
3962 3963 3964 3965 3966
  DBUG_VOID_RETURN;
}


/*
magnus@neptunus.(none)'s avatar
magnus@neptunus.(none) committed
3967
  If this thread already has a Thd_ndb object allocated
3968
  in current THD, reuse it. Otherwise
magnus@neptunus.(none)'s avatar
magnus@neptunus.(none) committed
3969
  seize a Thd_ndb object, assign it to current THD and use it.
3970 3971 3972
 
*/

3973
Ndb* check_ndb_in_thd(THD* thd)
3974
{
3975
  DBUG_ENTER("check_ndb_in_thd");
3976
  Thd_ndb *thd_ndb= (Thd_ndb*)thd->transaction.thd_ndb;
3977
  
3978
  if (!thd_ndb)
3979
  {
magnus@neptunus.(none)'s avatar
magnus@neptunus.(none) committed
3980
    if (!(thd_ndb= ha_ndbcluster::seize_thd_ndb()))
3981
      DBUG_RETURN(NULL);
3982
    thd->transaction.thd_ndb= thd_ndb;
3983
  }
magnus@neptunus.(none)'s avatar
magnus@neptunus.(none) committed
3984
  DBUG_RETURN(thd_ndb->ndb);
3985 3986
}

magnus@neptunus.(none)'s avatar
magnus@neptunus.(none) committed
3987

3988

3989 3990 3991
int ha_ndbcluster::check_ndb_connection()
{
  THD* thd= current_thd;
3992
  Ndb *ndb;
3993 3994
  DBUG_ENTER("check_ndb_connection");
  
3995
  if (!(ndb= check_ndb_in_thd(thd)))
3996
    DBUG_RETURN(HA_ERR_NO_CONNECTION);
3997
  ndb->setDatabaseName(m_dbname);
3998 3999 4000
  DBUG_RETURN(0);
}

magnus@neptunus.(none)'s avatar
magnus@neptunus.(none) committed
4001

4002 4003
void ndbcluster_close_connection(THD *thd)
{
4004
  Thd_ndb *thd_ndb= (Thd_ndb*)thd->transaction.thd_ndb;
4005
  DBUG_ENTER("ndbcluster_close_connection");
4006 4007
  if (thd_ndb)
  {
4008
    ha_ndbcluster::release_thd_ndb(thd_ndb);
4009 4010
    thd->transaction.thd_ndb= NULL;
  }
4011 4012 4013 4014 4015 4016 4017 4018
  DBUG_VOID_RETURN;
}


/*
  Try to discover one table from NDB
 */

4019
int ndbcluster_discover(THD* thd, const char *db, const char *name,
4020 4021 4022 4023 4024
			const void** frmblob, uint* frmlen)
{
  uint len;
  const void* data;
  const NDBTAB* tab;
4025
  Ndb* ndb;
4026
  DBUG_ENTER("ndbcluster_discover");
4027
  DBUG_PRINT("enter", ("db: %s, name: %s", db, name)); 
4028

4029 4030 4031
  if (!(ndb= check_ndb_in_thd(thd)))
    DBUG_RETURN(HA_ERR_NO_CONNECTION);  
  ndb->setDatabaseName(db);
4032

4033
  NDBDICT* dict= ndb->getDictionary();
magnus@neptunus.(none)'s avatar
magnus@neptunus.(none) committed
4034
  dict->set_local_table_data_size(sizeof(Ndb_table_local_info));
4035 4036 4037 4038 4039 4040 4041
  dict->invalidateTable(name);
  if (!(tab= dict->getTable(name)))
  {    
    const NdbError err= dict->getNdbError();
    if (err.code == 709)
      DBUG_RETURN(1);
    ERR_RETURN(err);
4042 4043 4044 4045 4046 4047 4048 4049 4050 4051 4052 4053 4054 4055 4056 4057 4058 4059 4060 4061 4062 4063
  }
  
  DBUG_PRINT("info", ("Found table %s", tab->getName()));
  
  len= tab->getFrmLength();  
  if (len == 0 || tab->getFrmData() == NULL)
  {
    DBUG_PRINT("No frm data found",
               ("Table is probably created via NdbApi")); 
    DBUG_RETURN(2);
  }
  
  if (unpackfrm(&data, &len, tab->getFrmData()))
    DBUG_RETURN(3);

  *frmlen= len;
  *frmblob= data;
  
  DBUG_RETURN(0);
}

/*
4064 4065 4066
  Check if a table exists in NDB
   
 */
4067

4068 4069 4070 4071 4072 4073 4074 4075 4076 4077 4078 4079 4080 4081
int ndbcluster_table_exists(THD* thd, const char *db, const char *name)
{
  uint len;
  const void* data;
  const NDBTAB* tab;
  Ndb* ndb;
  DBUG_ENTER("ndbcluster_table_exists");
  DBUG_PRINT("enter", ("db: %s, name: %s", db, name)); 

  if (!(ndb= check_ndb_in_thd(thd)))
    DBUG_RETURN(HA_ERR_NO_CONNECTION);  
  ndb->setDatabaseName(db);

  NDBDICT* dict= ndb->getDictionary();
magnus@neptunus.(none)'s avatar
magnus@neptunus.(none) committed
4082
  dict->set_local_table_data_size(sizeof(Ndb_table_local_info));
4083 4084 4085 4086 4087 4088 4089 4090 4091 4092 4093 4094 4095 4096
  dict->invalidateTable(name);
  if (!(tab= dict->getTable(name)))
  {    
    const NdbError err= dict->getNdbError();
    if (err.code == 709)
      DBUG_RETURN(0);
    ERR_RETURN(err);
  }
  
  DBUG_PRINT("info", ("Found table %s", tab->getName()));
  DBUG_RETURN(1);
}


4097

4098 4099
extern "C" byte* tables_get_key(const char *entry, uint *length,
				my_bool not_used __attribute__((unused)))
4100 4101 4102 4103 4104 4105 4106
{
  *length= strlen(entry);
  return (byte*) entry;
}


int ndbcluster_find_files(THD *thd,const char *db,const char *path,
4107
			  const char *wild, bool dir, List<char> *files)
4108
{
4109 4110 4111
  DBUG_ENTER("ndbcluster_find_files");
  DBUG_PRINT("enter", ("db: %s", db));
  { // extra bracket to avoid gcc 2.95.3 warning
4112
  uint i;
4113
  Ndb* ndb;
4114
  char name[FN_REFLEN];
4115
  HASH ndb_tables, ok_tables;
4116
  NdbDictionary::Dictionary::List list;
4117 4118 4119 4120

  if (!(ndb= check_ndb_in_thd(thd)))
    DBUG_RETURN(HA_ERR_NO_CONNECTION);

4121
  if (dir)
4122
    DBUG_RETURN(0); // Discover of databases not yet supported
4123

4124
  // List tables in NDB
4125
  NDBDICT *dict= ndb->getDictionary();
4126 4127
  if (dict->listObjects(list, 
			NdbDictionary::Object::UserTable) != 0)
4128
    ERR_RETURN(dict->getNdbError());
4129

4130 4131 4132 4133 4134 4135 4136 4137 4138 4139 4140 4141 4142 4143 4144
  if (hash_init(&ndb_tables, system_charset_info,list.count,0,0,
		(hash_get_key)tables_get_key,0,0))
  {
    DBUG_PRINT("error", ("Failed to init HASH ndb_tables"));
    DBUG_RETURN(-1);
  }

  if (hash_init(&ok_tables, system_charset_info,32,0,0,
		(hash_get_key)tables_get_key,0,0))
  {
    DBUG_PRINT("error", ("Failed to init HASH ok_tables"));
    hash_free(&ndb_tables);
    DBUG_RETURN(-1);
  }  

4145 4146 4147
  for (i= 0 ; i < list.count ; i++)
  {
    NdbDictionary::Dictionary::List::Element& t= list.elements[i];
4148
    DBUG_PRINT("info", ("Found %s/%s in NDB", t.database, t.name));     
4149

4150 4151 4152 4153
    // Add only tables that belongs to db
    if (my_strcasecmp(system_charset_info, t.database, db))
      continue;

4154
    // Apply wildcard to list of tables in NDB
4155
    if (wild)
4156
    {
4157 4158 4159 4160 4161 4162 4163 4164
      if (lower_case_table_names)
      {
	if (wild_case_compare(files_charset_info, t.name, wild))
	  continue;
      }
      else if (wild_compare(t.name,wild,0))
	continue;
    }
4165 4166
    DBUG_PRINT("info", ("Inserting %s into ndb_tables hash", t.name));     
    my_hash_insert(&ndb_tables, (byte*)thd->strdup(t.name));
4167 4168
  }

4169 4170 4171 4172 4173 4174 4175 4176 4177 4178 4179 4180 4181 4182 4183
  char *file_name;
  List_iterator<char> it(*files);
  List<char> delete_list;
  while ((file_name=it++))
  {
    DBUG_PRINT("info", ("%s", file_name));     
    if (hash_search(&ndb_tables, file_name, strlen(file_name)))
    {
      DBUG_PRINT("info", ("%s existed in NDB _and_ on disk ", file_name));
      // File existed in NDB and as frm file, put in ok_tables list
      my_hash_insert(&ok_tables, (byte*)file_name);
      continue;
    }
    
    // File is not in NDB, check for .ndb file with this name
4184
    (void)strxnmov(name, FN_REFLEN, 
4185 4186
		   mysql_data_home,"/",db,"/",file_name,ha_ndb_ext,NullS);
    DBUG_PRINT("info", ("Check access for %s", name));
4187
    if (access(name, F_OK))
4188 4189 4190 4191
    {
      DBUG_PRINT("info", ("%s did not exist on disk", name));     
      // .ndb file did not exist on disk, another table type
      continue;
4192
    }
4193

4194 4195 4196 4197 4198 4199 4200 4201 4202 4203 4204
    DBUG_PRINT("info", ("%s existed on disk", name));     
    // The .ndb file exists on disk, but it's not in list of tables in ndb
    // Verify that handler agrees table is gone.
    if (ndbcluster_table_exists(thd, db, file_name) == 0)    
    {
      DBUG_PRINT("info", ("NDB says %s does not exists", file_name));     
      it.remove();
      // Put in list of tables to remove from disk
      delete_list.push_back(thd->strdup(file_name));
    }
  }
4205

4206 4207 4208 4209
  // Check for new files to discover
  DBUG_PRINT("info", ("Checking for new files to discover"));       
  List<char> create_list;
  for (i= 0 ; i < ndb_tables.records ; i++)
4210
  {
4211 4212
    file_name= hash_element(&ndb_tables, i);
    if (!hash_search(&ok_tables, file_name, strlen(file_name)))
4213
    {
4214 4215 4216 4217 4218 4219
      DBUG_PRINT("info", ("%s must be discovered", file_name));       
      // File is in list of ndb tables and not in ok_tables
      // This table need to be created
      create_list.push_back(thd->strdup(file_name));
    }
  }
4220

4221 4222
  // Lock mutex before deleting and creating frm files
  pthread_mutex_lock(&LOCK_open);
4223

4224 4225 4226 4227 4228 4229 4230 4231 4232 4233 4234
  if (!global_read_lock)
  {
    // Delete old files
    List_iterator_fast<char> it3(delete_list);
    while ((file_name=it3++))
    {  
      DBUG_PRINT("info", ("Remove table %s/%s",db, file_name ));
      // Delete the table and all related files
      TABLE_LIST table_list;
      bzero((char*) &table_list,sizeof(table_list));
      table_list.db= (char*) db;
4235
      table_list.alias=table_list.real_name=(char*)file_name;
4236
      (void)mysql_rm_table_part2(thd, &table_list, 
4237 4238 4239
				 /* if_exists */ TRUE, 
				 /* drop_temporary */ FALSE, 
				 /* dont_log_query*/ TRUE);
4240 4241 4242
    }
  }

4243 4244 4245 4246 4247
  // Create new files
  List_iterator_fast<char> it2(create_list);
  while ((file_name=it2++))
  {  
    DBUG_PRINT("info", ("Table %s need discovery", name));
4248
    if (ha_create_table_from_engine(thd, db, file_name, TRUE) == 0)
4249
      files->push_back(thd->strdup(file_name)); 
4250 4251 4252 4253 4254
  }

  pthread_mutex_unlock(&LOCK_open);      
  
  hash_free(&ok_tables);
4255
  hash_free(&ndb_tables);
4256
  } // extra bracket to avoid gcc 2.95.3 warning
4257
  DBUG_RETURN(0);    
4258 4259 4260 4261 4262 4263 4264 4265 4266 4267
}


/*
  Initialise all gloal variables before creating 
  a NDB Cluster table handler
 */

bool ndbcluster_init()
{
4268
  int res;
4269
  DBUG_ENTER("ndbcluster_init");
4270
  // Set connectstring if specified
4271 4272
  if (opt_ndbcluster_connectstring != 0)
    DBUG_PRINT("connectstring", ("%s", opt_ndbcluster_connectstring));     
4273
  if ((g_ndb_cluster_connection=
4274
       new Ndb_cluster_connection(opt_ndbcluster_connectstring)) == 0)
4275
  {
4276 4277
    DBUG_PRINT("error",("Ndb_cluster_connection(%s)",
			opt_ndbcluster_connectstring));
4278
    goto ndbcluster_init_error;
4279
  }
4280

4281 4282 4283
  g_ndb_cluster_connection->set_optimized_node_selection
    (opt_ndb_optimized_node_selection);

4284 4285
  // Create a Ndb object to open the connection  to NDB
  g_ndb= new Ndb(g_ndb_cluster_connection, "sys");
4286
  g_ndb->getDictionary()->set_local_table_data_size(sizeof(Ndb_table_local_info));
4287
  if (g_ndb->init() != 0)
4288 4289
  {
    ERR_PRINT (g_ndb->getNdbError());
4290
    goto ndbcluster_init_error;
4291
  }
4292

4293
  if ((res= g_ndb_cluster_connection->connect(0,0,0)) == 0)
4294
  {
4295 4296 4297
    DBUG_PRINT("info",("NDBCLUSTER storage engine at %s on port %d",
		       g_ndb_cluster_connection->get_connected_host(),
		       g_ndb_cluster_connection->get_connected_port()));
4298
    g_ndb_cluster_connection->wait_until_ready(10,0);
4299
  } 
4300
  else if(res == 1)
4301
  {
4302 4303
    if (g_ndb_cluster_connection->start_connect_thread()) {
      DBUG_PRINT("error", ("g_ndb_cluster_connection->start_connect_thread()"));
4304 4305 4306 4307 4308 4309
      goto ndbcluster_init_error;
    }
    {
      char buf[1024];
      DBUG_PRINT("info",("NDBCLUSTER storage engine not started, will connect using %s",
			 g_ndb_cluster_connection->get_connectstring(buf,sizeof(buf))));
4310
    }
4311
  }
4312
  else
4313 4314 4315
  {
    DBUG_ASSERT(res == -1);
    DBUG_PRINT("error", ("permanent error"));
4316
    goto ndbcluster_init_error;
4317 4318
  }
  
4319 4320 4321
  (void) hash_init(&ndbcluster_open_tables,system_charset_info,32,0,0,
                   (hash_get_key) ndbcluster_get_key,0,0);
  pthread_mutex_init(&ndbcluster_mutex,MY_MUTEX_INIT_FAST);
4322

4323 4324
  ndbcluster_inited= 1;
#ifdef USE_DISCOVER_ON_STARTUP
magnus@neptunus.(none)'s avatar
magnus@neptunus.(none) committed
4325
  if (ndb_discover_tables() != 0)
4326
    goto ndbcluster_init_error;    
4327
#endif
4328
  DBUG_RETURN(FALSE);
4329 4330 4331
 ndbcluster_init_error:
  ndbcluster_end();
  DBUG_RETURN(TRUE);
4332 4333 4334 4335 4336 4337 4338 4339 4340 4341 4342 4343
}


/*
  End use of the NDB Cluster table handler
  - free all global variables allocated by 
    ndcluster_init()
*/

bool ndbcluster_end()
{
  DBUG_ENTER("ndbcluster_end");
4344 4345
  if(g_ndb)
    delete g_ndb;
4346
  g_ndb= NULL;
4347 4348 4349
  if (g_ndb_cluster_connection)
    delete g_ndb_cluster_connection;
  g_ndb_cluster_connection= NULL;
4350 4351 4352 4353 4354 4355 4356 4357
  if (!ndbcluster_inited)
    DBUG_RETURN(0);
  hash_free(&ndbcluster_open_tables);
  pthread_mutex_destroy(&ndbcluster_mutex);
  ndbcluster_inited= 0;
  DBUG_RETURN(0);
}

4358 4359 4360 4361 4362
/*
  Static error print function called from
  static handler method ndbcluster_commit
  and ndbcluster_rollback
*/
4363 4364

void ndbcluster_print_error(int error, const NdbOperation *error_op)
4365
{
4366 4367
  DBUG_ENTER("ndbcluster_print_error");
  TABLE tab;
4368 4369
  const char *tab_name= (error_op) ? error_op->getTableName() : "";
  tab.table_name= (char *) tab_name;
4370
  ha_ndbcluster error_handler(&tab);
4371
  tab.file= &error_handler;
4372
  error_handler.print_error(error, MYF(0));
ndbdev@ndbmaster.mysql.com's avatar
ndbdev@ndbmaster.mysql.com committed
4373
  DBUG_VOID_RETURN;
4374
}
4375 4376 4377 4378 4379 4380 4381 4382 4383 4384 4385 4386 4387 4388 4389 4390 4391 4392 4393 4394 4395 4396 4397

/*
  Set m_tabname from full pathname to table file 
 */

void ha_ndbcluster::set_tabname(const char *path_name)
{
  char *end, *ptr;
  
  /* Scan name from the end */
  end= strend(path_name)-1;
  ptr= end;
  while (ptr >= path_name && *ptr != '\\' && *ptr != '/') {
    ptr--;
  }
  uint name_len= end - ptr;
  memcpy(m_tabname, ptr + 1, end - ptr);
  m_tabname[name_len]= '\0';
#ifdef __WIN__
  /* Put to lower case */
  ptr= m_tabname;
  
  while (*ptr != '\0') {
4398
    *ptr= tolower(*ptr);
4399 4400 4401 4402 4403 4404 4405 4406 4407 4408 4409 4410 4411 4412 4413
    ptr++;
  }
#endif
}

/**
 * Set a given location from full pathname to table file
 *
 */
void
ha_ndbcluster::set_tabname(const char *path_name, char * tabname)
{
  char *end, *ptr;
  
  /* Scan name from the end */
4414 4415
  end= strend(path_name)-1;
  ptr= end;
4416 4417 4418
  while (ptr >= path_name && *ptr != '\\' && *ptr != '/') {
    ptr--;
  }
4419
  uint name_len= end - ptr;
4420
  memcpy(tabname, ptr + 1, end - ptr);
4421
  tabname[name_len]= '\0';
4422 4423
#ifdef __WIN__
  /* Put to lower case */
4424
  ptr= tabname;
4425 4426 4427 4428 4429 4430 4431 4432 4433 4434 4435 4436 4437 4438 4439 4440 4441 4442 4443 4444 4445 4446 4447 4448 4449 4450 4451 4452 4453 4454 4455 4456 4457 4458 4459 4460 4461 4462 4463 4464 4465 4466 4467 4468 4469
  
  while (*ptr != '\0') {
    *ptr= tolower(*ptr);
    ptr++;
  }
#endif
}


/*
  Set m_dbname from full pathname to table file
 
 */

void ha_ndbcluster::set_dbname(const char *path_name)
{
  char *end, *ptr;
  
  /* Scan name from the end */
  ptr= strend(path_name)-1;
  while (ptr >= path_name && *ptr != '\\' && *ptr != '/') {
    ptr--;
  }
  ptr--;
  end= ptr;
  while (ptr >= path_name && *ptr != '\\' && *ptr != '/') {
    ptr--;
  }
  uint name_len= end - ptr;
  memcpy(m_dbname, ptr + 1, name_len);
  m_dbname[name_len]= '\0';
#ifdef __WIN__
  /* Put to lower case */
  
  ptr= m_dbname;
  
  while (*ptr != '\0') {
    *ptr= tolower(*ptr);
    ptr++;
  }
#endif
}


ha_rows 
4470 4471 4472 4473
ha_ndbcluster::records_in_range(uint inx, key_range *min_key,
                                key_range *max_key)
{
  KEY *key_info= table->key_info + inx;
4474
  uint key_length= key_info->key_length;
4475
  NDB_INDEX_TYPE idx_type= get_index_type(inx);  
4476 4477

  DBUG_ENTER("records_in_range");
4478 4479 4480 4481 4482 4483 4484 4485 4486 4487 4488 4489 4490 4491
  // Prevent partial read of hash indexes by returning HA_POS_ERROR
  if ((idx_type == UNIQUE_INDEX || idx_type == PRIMARY_KEY_INDEX) &&
      ((min_key && min_key->length < key_length) ||
       (max_key && max_key->length < key_length)))
    DBUG_RETURN(HA_POS_ERROR);
  
  // Read from hash index with full key
  // This is a "const" table which returns only one record!      
  if ((idx_type != ORDERED_INDEX) &&
      ((min_key && min_key->length == key_length) || 
       (max_key && max_key->length == key_length)))
    DBUG_RETURN(1);
  
  DBUG_RETURN(10); /* Good guess when you don't know anything */
4492 4493
}

4494 4495 4496 4497 4498 4499 4500 4501 4502 4503 4504 4505 4506 4507 4508 4509 4510 4511 4512 4513 4514 4515 4516 4517 4518 4519 4520 4521 4522 4523 4524 4525 4526 4527 4528 4529 4530
ulong ha_ndbcluster::table_flags(void) const
{
  if (m_ha_not_exact_count)
    return m_table_flags | HA_NOT_EXACT_COUNT;
  else
    return m_table_flags;
}
const char * ha_ndbcluster::table_type() const 
{
  return("ndbcluster");
}
uint ha_ndbcluster::max_supported_record_length() const
{ 
  return NDB_MAX_TUPLE_SIZE;
}
uint ha_ndbcluster::max_supported_keys() const
{
  return MAX_KEY;
}
uint ha_ndbcluster::max_supported_key_parts() const 
{
  return NDB_MAX_NO_OF_ATTRIBUTES_IN_KEY;
}
uint ha_ndbcluster::max_supported_key_length() const
{
  return NDB_MAX_KEY_SIZE;
}
bool ha_ndbcluster::low_byte_first() const
{ 
#ifdef WORDS_BIGENDIAN
  return FALSE;
#else
  return TRUE;
#endif
}
bool ha_ndbcluster::has_transactions()
{
4531
  return m_transaction_on;
4532 4533 4534 4535 4536 4537 4538 4539 4540 4541 4542 4543 4544 4545 4546 4547
}
const char* ha_ndbcluster::index_type(uint key_number)
{
  switch (get_index_type(key_number)) {
  case ORDERED_INDEX:
  case UNIQUE_ORDERED_INDEX:
  case PRIMARY_KEY_ORDERED_INDEX:
    return "BTREE";
  case UNIQUE_INDEX:
  case PRIMARY_KEY_INDEX:
  default:
    return "HASH";
  }
}
uint8 ha_ndbcluster::table_cache_type()
{
4548 4549 4550 4551
  if (m_use_local_query_cache)
    return HA_CACHE_TBL_TRANSACT;
  else
    return HA_CACHE_TBL_NOCACHE;
4552
}
4553 4554 4555 4556 4557 4558 4559 4560 4561 4562 4563 4564 4565 4566 4567 4568 4569 4570 4571 4572 4573 4574 4575 4576 4577 4578 4579 4580 4581 4582 4583 4584 4585 4586 4587 4588 4589 4590 4591 4592 4593 4594 4595 4596 4597 4598 4599 4600 4601 4602 4603 4604 4605 4606 4607 4608 4609 4610 4611 4612 4613 4614 4615 4616 4617 4618 4619 4620 4621 4622 4623 4624 4625 4626 4627 4628 4629 4630 4631 4632 4633 4634 4635 4636 4637 4638 4639 4640 4641 4642 4643

/*
  Handling the shared NDB_SHARE structure that is needed to 
  provide table locking.
  It's also used for sharing data with other NDB handlers
  in the same MySQL Server. There is currently not much
  data we want to or can share.
 */

static byte* ndbcluster_get_key(NDB_SHARE *share,uint *length,
				my_bool not_used __attribute__((unused)))
{
  *length=share->table_name_length;
  return (byte*) share->table_name;
}

static NDB_SHARE* get_share(const char *table_name)
{
  NDB_SHARE *share;
  pthread_mutex_lock(&ndbcluster_mutex);
  uint length=(uint) strlen(table_name);
  if (!(share=(NDB_SHARE*) hash_search(&ndbcluster_open_tables,
                                       (byte*) table_name,
                                       length)))
  {
    if ((share=(NDB_SHARE *) my_malloc(sizeof(*share)+length+1,
                                       MYF(MY_WME | MY_ZEROFILL))))
    {
      share->table_name_length=length;
      share->table_name=(char*) (share+1);
      strmov(share->table_name,table_name);
      if (my_hash_insert(&ndbcluster_open_tables, (byte*) share))
      {
        pthread_mutex_unlock(&ndbcluster_mutex);
        my_free((gptr) share,0);
        return 0;
      }
      thr_lock_init(&share->lock);
      pthread_mutex_init(&share->mutex,MY_MUTEX_INIT_FAST);
    }
  }
  share->use_count++;
  pthread_mutex_unlock(&ndbcluster_mutex);
  return share;
}


static void free_share(NDB_SHARE *share)
{
  pthread_mutex_lock(&ndbcluster_mutex);
  if (!--share->use_count)
  {
    hash_delete(&ndbcluster_open_tables, (byte*) share);
    thr_lock_delete(&share->lock);
    pthread_mutex_destroy(&share->mutex);
    my_free((gptr) share, MYF(0));
  }
  pthread_mutex_unlock(&ndbcluster_mutex);
}



/*
  Internal representation of the frm blob
   
*/

struct frm_blob_struct 
{
  struct frm_blob_header 
  {
    uint ver;      // Version of header
    uint orglen;   // Original length of compressed data
    uint complen;  // Compressed length of data, 0=uncompressed
  } head;
  char data[1];  
};



static int packfrm(const void *data, uint len, 
		   const void **pack_data, uint *pack_len)
{
  int error;
  ulong org_len, comp_len;
  uint blob_len;
  frm_blob_struct* blob;
  DBUG_ENTER("packfrm");
  DBUG_PRINT("enter", ("data: %x, len: %d", data, len));
  
  error= 1;
4644
  org_len= len;
4645 4646 4647 4648 4649 4650 4651 4652 4653 4654 4655 4656 4657 4658 4659 4660 4661 4662 4663
  if (my_compress((byte*)data, &org_len, &comp_len))
    goto err;
  
  DBUG_PRINT("info", ("org_len: %d, comp_len: %d", org_len, comp_len));
  DBUG_DUMP("compressed", (char*)data, org_len);
  
  error= 2;
  blob_len= sizeof(frm_blob_struct::frm_blob_header)+org_len;
  if (!(blob= (frm_blob_struct*) my_malloc(blob_len,MYF(MY_WME))))
    goto err;
  
  // Store compressed blob in machine independent format
  int4store((char*)(&blob->head.ver), 1);
  int4store((char*)(&blob->head.orglen), comp_len);
  int4store((char*)(&blob->head.complen), org_len);
  
  // Copy frm data into blob, already in machine independent format
  memcpy(blob->data, data, org_len);  
  
4664 4665 4666
  *pack_data= blob;
  *pack_len= blob_len;
  error= 0;
4667 4668 4669 4670 4671 4672 4673 4674 4675 4676 4677
  
  DBUG_PRINT("exit", ("pack_data: %x, pack_len: %d", *pack_data, *pack_len));
err:
  DBUG_RETURN(error);
  
}


static int unpackfrm(const void **unpack_data, uint *unpack_len,
		    const void *pack_data)
{
4678
   const frm_blob_struct *blob= (frm_blob_struct*)pack_data;
4679 4680 4681 4682 4683 4684 4685 4686 4687 4688 4689 4690 4691 4692 4693
   byte *data;
   ulong complen, orglen, ver;
   DBUG_ENTER("unpackfrm");
   DBUG_PRINT("enter", ("pack_data: %x", pack_data));

   complen=	uint4korr((char*)&blob->head.complen);
   orglen=	uint4korr((char*)&blob->head.orglen);
   ver=		uint4korr((char*)&blob->head.ver);
 
   DBUG_PRINT("blob",("ver: %d complen: %d orglen: %d",
 		     ver,complen,orglen));
   DBUG_DUMP("blob->data", (char*) blob->data, complen);
 
   if (ver != 1)
     DBUG_RETURN(1);
4694
   if (!(data= my_malloc(max(orglen, complen), MYF(MY_WME))))
4695 4696 4697 4698 4699 4700 4701 4702 4703
     DBUG_RETURN(2);
   memcpy(data, blob->data, complen);
 
   if (my_uncompress(data, &complen, &orglen))
   {
     my_free((char*)data, MYF(0));
     DBUG_RETURN(3);
   }

4704 4705
   *unpack_data= data;
   *unpack_len= complen;
4706 4707 4708 4709 4710

   DBUG_PRINT("exit", ("frmdata: %x, len: %d", *unpack_data, *unpack_len));

   DBUG_RETURN(0);
}
4711 4712 4713 4714 4715 4716 4717 4718

static 
int
ndb_get_table_statistics(Ndb* ndb, const char * table, 
			 Uint64* row_count, Uint64* commit_count)
{
  DBUG_ENTER("ndb_get_table_statistics");
  DBUG_PRINT("enter", ("table: %s", table));
4719
  NdbConnection* pTrans= ndb->startTransaction();
4720 4721 4722 4723
  do 
  {
    if (pTrans == NULL)
      break;
4724
      
4725 4726 4727 4728
    NdbScanOperation* pOp= pTrans->getNdbScanOperation(table);
    if (pOp == NULL)
      break;
    
tomas@poseidon.ndb.mysql.com's avatar
tomas@poseidon.ndb.mysql.com committed
4729
    NdbResultSet* rs= pOp->readTuples(NdbOperation::LM_CommittedRead); 
4730 4731 4732 4733 4734 4735 4736 4737 4738 4739 4740
    if (rs == 0)
      break;
    
    int check= pOp->interpret_exit_last_row();
    if (check == -1)
      break;
    
    Uint64 rows, commits;
    pOp->getValue(NdbDictionary::Column::ROW_COUNT, (char*)&rows);
    pOp->getValue(NdbDictionary::Column::COMMIT_COUNT, (char*)&commits);
    
4741
    check= pTrans->execute(NoCommit, AbortOnError, TRUE);
4742 4743 4744 4745 4746
    if (check == -1)
      break;
    
    Uint64 sum_rows= 0;
    Uint64 sum_commits= 0;
4747
    while((check= rs->nextResult(TRUE, TRUE)) == 0)
4748 4749 4750 4751 4752 4753 4754 4755
    {
      sum_rows+= rows;
      sum_commits+= commits;
    }
    
    if (check == -1)
      break;

4756 4757
    rs->close(TRUE);

4758 4759 4760 4761 4762 4763 4764 4765 4766
    ndb->closeTransaction(pTrans);
    if(row_count)
      * row_count= sum_rows;
    if(commit_count)
      * commit_count= sum_commits;
    DBUG_PRINT("exit", ("records: %u commits: %u", sum_rows, sum_commits));
    DBUG_RETURN(0);
  } while(0);

4767
  ndb->closeTransaction(pTrans);
4768 4769 4770 4771
  DBUG_PRINT("exit", ("failed"));
  DBUG_RETURN(-1);
}

4772 4773 4774 4775 4776 4777 4778 4779 4780 4781 4782 4783 4784 4785 4786 4787 4788 4789 4790 4791 4792 4793 4794 4795 4796 4797
/*
  Create a .ndb file to serve as a placeholder indicating 
  that the table with this name is a ndb table
*/

int ha_ndbcluster::write_ndb_file()
{
  File file;
  bool error=1;
  char path[FN_REFLEN];
  
  DBUG_ENTER("write_ndb_file");
  DBUG_PRINT("enter", ("db: %s, name: %s", m_dbname, m_tabname));

  (void)strxnmov(path, FN_REFLEN, 
		 mysql_data_home,"/",m_dbname,"/",m_tabname,ha_ndb_ext,NullS);

  if ((file=my_create(path, CREATE_MODE,O_RDWR | O_TRUNC,MYF(MY_WME))) >= 0)
  {
    // It's an empty file
    error=0;
    my_close(file,MYF(0));
  }
  DBUG_RETURN(error);
}

4798
#endif /* HAVE_NDBCLUSTER_DB */