Refs Tokutek/ft-index#46 Killed dmt_functor<> template. Added dmtwriter_t...

Refs Tokutek/ft-index#46 Killed dmt_functor<> template.  Added dmtwriter_t template parameter to dmt and some documentation

ft/dmt-wrapper.cc

@@ -97,28 +97,6 @@ PATENT RIGHTS GRANT:
 #include <util/mempool.h>
 #include "dmt-wrapper.h"
 
-namespace toku {
-template<>
-class dmt_functor<DMTVALUE> {
-    public:
-        size_t get_dmtdatain_t_size(void) const {
-            return sizeof(DMTVALUE);
-        }
-        void write_dmtdata_t_to(DMTVALUE *const dest) const {
-            *dest = value;
-        }
-
-        dmt_functor(DMTVALUE _value)
-            : value(_value) {}
-        dmt_functor(const uint32_t size UU(), DMTVALUE *const src)
-            : value(*src) {
-            paranoid_invariant(size == sizeof(DMTVALUE));
-        }
-    private:
-        const DMTVALUE value;
-};
-}
-
 int
 toku_dmt_create_steal_sorted_array(DMT *dmtp, DMTVALUE **valuesp, uint32_t numvalues, uint32_t capacity) {
     //TODO: implement using create_steal_sorted_array when it exists
@@ -167,7 +145,7 @@ int toku_dmt_create_from_sorted_array(DMT *dmtp, DMTVALUE *values, uint32_t numv
 }
 
 int toku_dmt_insert_at(DMT dmt, DMTVALUE value, uint32_t index) {
-    toku::dmt_functor<DMTVALUE> functor(value);
+    dmt_wrapper_internal::dmtvalue_writer functor(value);
     return dmt->insert_at(functor, index);
 }
 
@@ -222,7 +200,7 @@ int call_heftor(const uint32_t size, const DMTVALUE &v, const heftor &htor) {
 
 int toku_dmt_insert(DMT dmt, DMTVALUE value, int(*h)(DMTVALUE, void*v), void *v, uint32_t *index) {
     struct heftor htor = { .h = h, .v = v };
-    toku::dmt_functor<DMTVALUE> functor(value);
+    dmt_wrapper_internal::dmtvalue_writer functor(value);
     return dmt->insert<heftor, call_heftor>(functor, htor, index);
 }
 

ft/dmt-wrapper.h

@@ -143,8 +143,32 @@ PATENT RIGHTS GRANT:
 
 //typedef struct value *DMTVALUE; // A slight improvement over using void*.
 #include <util/dmt.h>
+
 typedef void *DMTVALUE;
-typedef toku::dmt<DMTVALUE> *DMT;
+
+namespace dmt_wrapper_internal {
+class dmtvalue_writer {
+    public:
+        size_t get_size(void) const {
+            return sizeof(DMTVALUE);
+        }
+        void write_to(DMTVALUE *const dest) const {
+            *dest = value;
+        }
+
+        dmtvalue_writer(DMTVALUE _value)
+            : value(_value) {}
+        dmtvalue_writer(const uint32_t size UU(), DMTVALUE *const src)
+            : value(*src) {
+            paranoid_invariant(size == sizeof(DMTVALUE));
+        }
+    private:
+        const DMTVALUE value;
+};
+};
+
+
+typedef toku::dmt<DMTVALUE, DMTVALUE, dmt_wrapper_internal::dmtvalue_writer> *DMT;
 
 
 int toku_dmt_create (DMT *dmtp);

ft/ft-flusher.cc

@@ -689,16 +689,16 @@ ftleaf_get_split_loc(
     switch (split_mode) {
     case SPLIT_LEFT_HEAVY: {
         *num_left_bns = node->n_children;
-        *num_left_les = BLB_DATA(node, *num_left_bns - 1)->omt_size();
+        *num_left_les = BLB_DATA(node, *num_left_bns - 1)->dmt_size();
         if (*num_left_les == 0) {
             *num_left_bns = node->n_children - 1;
-            *num_left_les = BLB_DATA(node, *num_left_bns - 1)->omt_size();
+            *num_left_les = BLB_DATA(node, *num_left_bns - 1)->dmt_size();
         }
         goto exit;
     }
     case SPLIT_RIGHT_HEAVY: {
         *num_left_bns = 1;
-        *num_left_les = BLB_DATA(node, 0)->omt_size() ? 1 : 0;
+        *num_left_les = BLB_DATA(node, 0)->dmt_size() ? 1 : 0;
         goto exit;
     }
     case SPLIT_EVENLY: {
@@ -707,8 +707,8 @@ ftleaf_get_split_loc(
         uint64_t sumlesizes = ftleaf_disk_size(node);
         uint32_t size_so_far = 0;
         for (int i = 0; i < node->n_children; i++) {
-            BN_DATA bd = BLB_DATA(node, i);
-            uint32_t n_leafentries = bd->omt_size();
+            bn_data* bd = BLB_DATA(node, i);
+            uint32_t n_leafentries = bd->dmt_size();
             for (uint32_t j=0; j < n_leafentries; j++) {
                 size_t size_this_le;
                 int rr = bd->fetch_klpair_disksize(j, &size_this_le);
@@ -725,7 +725,7 @@ ftleaf_get_split_loc(
                             (*num_left_les)--;
                         } else if (*num_left_bns > 1) {
                             (*num_left_bns)--;
-                            *num_left_les = BLB_DATA(node, *num_left_bns - 1)->omt_size();
+                            *num_left_les = BLB_DATA(node, *num_left_bns - 1)->dmt_size();
                         } else {
                             // we are trying to split a leaf with only one
                             // leafentry in it
@@ -851,7 +851,7 @@ ftleaf_split(
     ftleaf_get_split_loc(node, split_mode, &num_left_bns, &num_left_les);
     {
         // did we split right on the boundary between basement nodes?
-        const bool split_on_boundary = (num_left_les == 0) || (num_left_les == (int) BLB_DATA(node, num_left_bns - 1)->omt_size());
+        const bool split_on_boundary = (num_left_les == 0) || (num_left_les == (int) BLB_DATA(node, num_left_bns - 1)->dmt_size());
         // Now we know where we are going to break it
         // the two nodes will have a total of n_children+1 basement nodes
         // and n_children-1 pivots
@@ -912,7 +912,7 @@ ftleaf_split(
             move_leafentries(BLB(B, curr_dest_bn_index),
                              BLB(node, curr_src_bn_index),
                              num_left_les,         // first row to be moved to B
-                             BLB_DATA(node, curr_src_bn_index)->omt_size()  // number of rows in basement to be split
+                             BLB_DATA(node, curr_src_bn_index)->dmt_size()  // number of rows in basement to be split
                              );
             BLB_MAX_MSN_APPLIED(B, curr_dest_bn_index) = BLB_MAX_MSN_APPLIED(node, curr_src_bn_index);
             curr_dest_bn_index++;
@@ -954,10 +954,10 @@ ftleaf_split(
                 toku_destroy_dbt(&node->childkeys[num_left_bns - 1]);
             }
         } else if (splitk) {
-            BN_DATA bd = BLB_DATA(node, num_left_bns - 1);
+            bn_data* bd = BLB_DATA(node, num_left_bns - 1);
             uint32_t keylen;
             void *key;
-            int rr = bd->fetch_le_key_and_len(bd->omt_size() - 1, &keylen, &key);
+            int rr = bd->fetch_key_and_len(bd->dmt_size() - 1, &keylen, &key);
             invariant_zero(rr);
             toku_memdup_dbt(splitk, key, keylen);
         }
@@ -1168,11 +1168,11 @@ merge_leaf_nodes(FTNODE a, FTNODE b)
     a->dirty = 1;
     b->dirty = 1;
 
-    BN_DATA a_last_bd = BLB_DATA(a, a->n_children-1);
+    bn_data* a_last_bd = BLB_DATA(a, a->n_children-1);
     // this bool states if the last basement node in a has any items or not
     // If it does, then it stays in the merge. If it does not, the last basement node
     // of a gets eliminated because we do not have a pivot to store for it (because it has no elements)
-    const bool a_has_tail = a_last_bd->omt_size() > 0;
+    const bool a_has_tail = a_last_bd->dmt_size() > 0;
 
     // move each basement node from b to a
     // move the pivots, adding one of what used to be max(a)
@@ -1199,7 +1199,7 @@ merge_leaf_nodes(FTNODE a, FTNODE b)
     if (a_has_tail) {
         uint32_t keylen;
         void *key;
-        int rr = a_last_bd->fetch_le_key_and_len(a_last_bd->omt_size() - 1, &keylen, &key);
+        int rr = a_last_bd->fetch_key_and_len(a_last_bd->dmt_size() - 1, &keylen, &key);
         invariant_zero(rr);
         toku_memdup_dbt(&a->childkeys[a->n_children-1], key, keylen);
         a->totalchildkeylens += keylen;

ft/ft-ops.cc

@@ -419,7 +419,7 @@ get_leaf_num_entries(FTNODE node) {
     int i;
     toku_assert_entire_node_in_memory(node);
     for ( i = 0; i < node->n_children; i++) {
-        result += BLB_DATA(node, i)->omt_size();
+        result += BLB_DATA(node, i)->dmt_size();
     }
     return result;
 }
@@ -1720,7 +1720,7 @@ toku_ft_bn_apply_cmd_once (
         oldsize = leafentry_memsize(le) + key_storage_size;
     }
     
-    // toku_le_apply_msg() may call mempool_malloc_from_omt() to allocate more space.
+    // toku_le_apply_msg() may call bn_data::mempool_malloc_and_update_dmt() to allocate more space.
     // That means le is guaranteed to not cause a sigsegv but it may point to a mempool that is
     // no longer in use.  We'll have to release the old mempool later.
     toku_le_apply_msg(
@@ -1910,7 +1910,7 @@ toku_ft_bn_apply_cmd (
     void* key = NULL;
     uint32_t keylen = 0;
 
-    uint32_t omt_size;
+    uint32_t dmt_size;
     int r;
     struct cmd_leafval_heaviside_extra be = {compare_fun, desc, cmd->u.id.key};
 
@@ -1922,9 +1922,9 @@ toku_ft_bn_apply_cmd (
     case FT_INSERT: {
         uint32_t idx;
         if (doing_seqinsert) {
-            idx = bn->data_buffer.omt_size();
+            idx = bn->data_buffer.dmt_size();
             DBT kdbt;
-            r = bn->data_buffer.fetch_le_key_and_len(idx-1, &kdbt.size, &kdbt.data);
+            r = bn->data_buffer.fetch_key_and_len(idx-1, &kdbt.size, &kdbt.data);
             if (r != 0) goto fz;
             int cmp = toku_cmd_leafval_heaviside(kdbt, be);
             if (cmp >= 0) goto fz;
@@ -1950,7 +1950,7 @@ toku_ft_bn_apply_cmd (
         // the leaf then it is sequential
         // window = min(32, number of leaf entries/16)
         {
-            uint32_t s = bn->data_buffer.omt_size();
+            uint32_t s = bn->data_buffer.dmt_size();
             uint32_t w = s / 16;
             if (w == 0) w = 1;
             if (w > 32) w = 32;
@@ -1985,8 +1985,8 @@ toku_ft_bn_apply_cmd (
     case FT_COMMIT_BROADCAST_ALL:
     case FT_OPTIMIZE:
         // Apply to all leafentries
-        omt_size = bn->data_buffer.omt_size();
-        for (uint32_t idx = 0; idx < omt_size; ) {
+        dmt_size = bn->data_buffer.dmt_size();
+        for (uint32_t idx = 0; idx < dmt_size; ) {
             DBT curr_keydbt;
             void* curr_keyp = NULL;
             uint32_t curr_keylen = 0;
@@ -2000,26 +2000,26 @@ toku_ft_bn_apply_cmd (
             if (!le_is_clean(storeddata)) { //If already clean, nothing to do.
                 toku_ft_bn_apply_cmd_once(bn, cmd, idx, storeddata, oldest_referenced_xid_known, gc_info, workdone, stats_to_update);
                 // at this point, we cannot trust cmd->u.id.key to be valid.
-                uint32_t new_omt_size = bn->data_buffer.omt_size();
-                if (new_omt_size != omt_size) {
-                    paranoid_invariant(new_omt_size+1 == omt_size);
+                uint32_t new_dmt_size = bn->data_buffer.dmt_size();
+                if (new_dmt_size != dmt_size) {
+                    paranoid_invariant(new_dmt_size+1 == dmt_size);
                     //Item was deleted.
                     deleted = 1;
                 }
             }
             if (deleted)
-                omt_size--;
+                dmt_size--;
             else
                 idx++;
         }
-        paranoid_invariant(bn->data_buffer.omt_size() == omt_size);
+        paranoid_invariant(bn->data_buffer.dmt_size() == dmt_size);
 
         break;
     case FT_COMMIT_BROADCAST_TXN:
     case FT_ABORT_BROADCAST_TXN:
         // Apply to all leafentries if txn is represented
-        omt_size = bn->data_buffer.omt_size();
-        for (uint32_t idx = 0; idx < omt_size; ) {
+        dmt_size = bn->data_buffer.dmt_size();
+        for (uint32_t idx = 0; idx < dmt_size; ) {
             DBT curr_keydbt;
             void* curr_keyp = NULL;
             uint32_t curr_keylen = 0;
@@ -2032,19 +2032,19 @@ toku_ft_bn_apply_cmd (
             int deleted = 0;
             if (le_has_xids(storeddata, cmd->xids)) {
                 toku_ft_bn_apply_cmd_once(bn, cmd, idx, storeddata, oldest_referenced_xid_known, gc_info, workdone, stats_to_update);
-                uint32_t new_omt_size = bn->data_buffer.omt_size();
-                if (new_omt_size != omt_size) {
-                    paranoid_invariant(new_omt_size+1 == omt_size);
+                uint32_t new_dmt_size = bn->data_buffer.dmt_size();
+                if (new_dmt_size != dmt_size) {
+                    paranoid_invariant(new_dmt_size+1 == dmt_size);
                     //Item was deleted.
                     deleted = 1;
                 }
             }
             if (deleted)
-                omt_size--;
+                dmt_size--;
             else
                 idx++;
         }
-        paranoid_invariant(bn->data_buffer.omt_size() == omt_size);
+        paranoid_invariant(bn->data_buffer.dmt_size() == dmt_size);
 
         break;
     case FT_UPDATE: {
@@ -2073,7 +2073,7 @@ toku_ft_bn_apply_cmd (
         // apply to all leafentries.
         uint32_t idx = 0;
         uint32_t num_leafentries_before;
-        while (idx < (num_leafentries_before = bn->data_buffer.omt_size())) {
+        while (idx < (num_leafentries_before = bn->data_buffer.dmt_size())) {
             void* curr_key = nullptr;
             uint32_t curr_keylen = 0;
             r = bn->data_buffer.fetch_klpair(idx, &storeddata, &curr_keylen, &curr_key);
@@ -2091,7 +2091,7 @@ toku_ft_bn_apply_cmd (
             r = do_update(update_fun, desc, bn, cmd, idx, storeddata, curr_key, curr_keylen, oldest_referenced_xid_known, gc_info, workdone, stats_to_update);
             assert_zero(r);
 
-            if (num_leafentries_before == bn->data_buffer.omt_size()) {
+            if (num_leafentries_before == bn->data_buffer.dmt_size()) {
                 // we didn't delete something, so increment the index.
                 idx++;
             }
@@ -2404,7 +2404,7 @@ basement_node_gc_all_les(BASEMENTNODE bn,
     int r = 0;
     uint32_t index = 0;
     uint32_t num_leafentries_before;
-    while (index < (num_leafentries_before = bn->data_buffer.omt_size())) {
+    while (index < (num_leafentries_before = bn->data_buffer.dmt_size())) {
         void* keyp = NULL;
         uint32_t keylen = 0;
         LEAFENTRY leaf_entry;
@@ -2423,7 +2423,7 @@ basement_node_gc_all_les(BASEMENTNODE bn,
             delta
             );
         // Check if the leaf entry was deleted or not.
-        if (num_leafentries_before == bn->data_buffer.omt_size()) {
+        if (num_leafentries_before == bn->data_buffer.dmt_size()) {
             ++index;
         }
     }
@@ -4929,7 +4929,7 @@ ok: ;
             switch (search->direction) {
             case FT_SEARCH_LEFT:
                 idx++;
-                if (idx >= bn->data_buffer.omt_size()) {
+                if (idx >= bn->data_buffer.dmt_size()) {
                     if (ftcursor->interrupt_cb && ftcursor->interrupt_cb(ftcursor->interrupt_cb_extra)) {
                         return TOKUDB_INTERRUPTED;
                     }
@@ -5604,7 +5604,7 @@ ft_cursor_shortcut (
     int r = 0;
     // if we are searching towards the end, limit is last element
     // if we are searching towards the beginning, limit is the first element
-    uint32_t limit = (direction > 0) ? (bd->omt_size() - 1) : 0;
+    uint32_t limit = (direction > 0) ? (bd->dmt_size() - 1) : 0;
 
     //Starting with the prev, find the first real (non-provdel) leafentry.
     while (index != limit) {
@@ -5895,7 +5895,7 @@ keysrange_in_leaf_partition (FT_HANDLE brt, FTNODE node,
         *less = idx_left;
         *equal_left = (r==0) ? 1 : 0;
 
-        uint32_t size = bn->data_buffer.omt_size();
+        uint32_t size = bn->data_buffer.dmt_size();
         uint32_t idx_right = size;
         r = -1;
         if (single_basement && key_right) {
@@ -6155,7 +6155,7 @@ static int get_key_after_bytes_in_basementnode(FT ft, BASEMENTNODE bn, const DBT
         assert(r == 0 || r == DB_NOTFOUND);
     }
     struct get_key_after_bytes_iterate_extra iter_extra = {skip_len, skipped, callback, cb_extra};
-    r = bn->data_buffer.omt_iterate_on_range<get_key_after_bytes_iterate_extra, get_key_after_bytes_iterate>(idx_left, bn->data_buffer.omt_size(), &iter_extra);
+    r = bn->data_buffer.dmt_iterate_on_range<get_key_after_bytes_iterate_extra, get_key_after_bytes_iterate>(idx_left, bn->data_buffer.dmt_size(), &iter_extra);
 
     // Invert the sense of r == 0 (meaning the iterate finished, which means we didn't find what we wanted)
     if (r == 1) {
@@ -6351,7 +6351,7 @@ toku_dump_ftnode (FILE *file, FT_HANDLE brt, BLOCKNUM blocknum, int depth, const
                              });
             }
             else {
-                int size = BLB_DATA(node, i)->omt_size();
+                int size = BLB_DATA(node, i)->dmt_size();
                 if (0)
                     for (int j=0; j<size; j++) {
                         LEAFENTRY le;
@@ -6531,9 +6531,9 @@ static bool is_empty_fast_iter (FT_HANDLE brt, FTNODE node) {
         }
         return 1;
     } else {
-        // leaf:  If the omt is empty, we are happy.
+        // leaf:  If the dmt is empty, we are happy.
         for (int i = 0; i < node->n_children; i++) {
-            if (BLB_DATA(node, i)->omt_size()) {
+            if (BLB_DATA(node, i)->dmt_size()) {
                 return false;
             }
         }

ft/ft-verify.cc

@@ -152,7 +152,7 @@ verify_msg_in_child_buffer(FT_HANDLE brt, enum ft_msg_type type, MSN msn, byteve
 static DBT
 get_ith_key_dbt (BASEMENTNODE bn, int i) {
     DBT kdbt;
-    int r = bn->data_buffer.fetch_le_key_and_len(i, &kdbt.size, &kdbt.data);
+    int r = bn->data_buffer.fetch_key_and_len(i, &kdbt.size, &kdbt.data);
     invariant_zero(r); // this is a bad failure if it happens.
     return kdbt;
 }
@@ -424,7 +424,7 @@ toku_verify_ftnode_internal(FT_HANDLE brt,
         }
         else {
             BASEMENTNODE bn = BLB(node, i);
-            for (uint32_t j = 0; j < bn->data_buffer.omt_size(); j++) {
+            for (uint32_t j = 0; j < bn->data_buffer.dmt_size(); j++) {
                 VERIFY_ASSERTION((rootmsn.msn >= this_msn.msn), 0, "leaf may have latest msn, but cannot be greater than root msn");
                 DBT kdbt = get_ith_key_dbt(bn, j);
                 if (curr_less_pivot) {

ft/ft.cc

@@ -1077,8 +1077,8 @@ garbage_helper(BLOCKNUM blocknum, int64_t UU(size), int64_t UU(address), void *e
         goto exit;
     }
     for (int i = 0; i < node->n_children; ++i) {
-        BN_DATA bd = BLB_DATA(node, i);
-        r = bd->omt_iterate<struct garbage_helper_extra, garbage_leafentry_helper>(info);
+        bn_data* bd = BLB_DATA(node, i);
+        r = bd->dmt_iterate<struct garbage_helper_extra, garbage_leafentry_helper>(info);
         if (r != 0) {
             goto exit;
         }

ft/ft_node-serialize.cc

@@ -375,10 +375,10 @@ serialize_ftnode_partition(FTNODE node, int i, struct sub_block *sb) {
     }
     else {
         unsigned char ch = FTNODE_PARTITION_OMT_LEAVES;
-        BN_DATA bd = BLB_DATA(node, i);
+        bn_data* bd = BLB_DATA(node, i);
 
         wbuf_nocrc_char(&wb, ch);
-        wbuf_nocrc_uint(&wb, bd->omt_size());
+        wbuf_nocrc_uint(&wb, bd->dmt_size());
 
         bd->prepare_to_serialize();
         bd->serialize_header(&wb);
@@ -386,7 +386,7 @@ serialize_ftnode_partition(FTNODE node, int i, struct sub_block *sb) {
             //
             // iterate over leafentries and place them into the buffer
             //
-            bd->omt_iterate<struct wbuf, wbufwriteleafentry>(&wb);
+            bd->dmt_iterate<struct wbuf, wbufwriteleafentry>(&wb);
         } else {
             bd->serialize_rest(&wb);
         }
@@ -552,7 +552,7 @@ rebalance_ftnode_leaf(FTNODE node, unsigned int basementnodesize)
     // Count number of leaf entries in this leaf (num_le).
     uint32_t num_le = 0;
     for (uint32_t i = 0; i < num_orig_basements; i++) {
-        num_le += BLB_DATA(node, i)->omt_size();
+        num_le += BLB_DATA(node, i)->dmt_size();
     }
 
     uint32_t num_alloc = num_le ? num_le : 1;  // simplify logic below by always having at least one entry per array
@@ -577,10 +577,10 @@ rebalance_ftnode_leaf(FTNODE node, unsigned int basementnodesize)
 
     uint32_t curr_le = 0;
     for (uint32_t i = 0; i < num_orig_basements; i++) {
-        BN_DATA bd = BLB_DATA(node, i);
+        bn_data* bd = BLB_DATA(node, i);
         struct array_info ai {.offset = curr_le, .le_array = leafpointers, .key_sizes_array = key_sizes, .key_ptr_array = key_pointers };
-        bd->omt_iterate<array_info, array_item>(&ai);
-        curr_le += bd->omt_size();
+        bd->dmt_iterate<array_info, array_item>(&ai);
+        curr_le += bd->dmt_size();
     }
 
     // Create an array that will store indexes of new pivots.
@@ -702,8 +702,8 @@ rebalance_ftnode_leaf(FTNODE node, unsigned int basementnodesize)
         uint32_t num_les_to_copy = num_les_this_bn[i];
         invariant(num_les_to_copy == num_in_bn); 
 
-        BN_DATA bd = BLB_DATA(node, i);
-        bd->replace_contents_with_clone_of_sorted_array(
+        bn_data* bd = BLB_DATA(node, i);
+        bd->set_contents_as_clone_of_sorted_array(
             num_les_to_copy,
             &key_pointers[baseindex_this_bn],
             &key_sizes[baseindex_this_bn],
@@ -1560,7 +1560,7 @@ deserialize_ftnode_partition(
         data_size -= rb.ndone; // remaining bytes of leafentry data
 
         BASEMENTNODE bn = BLB(node, childnum);
-        bn->data_buffer.initialize_from_data(num_entries, &rb, data_size, node->layout_version_read_from_disk);
+        bn->data_buffer.deserialize_from_rbuf(num_entries, &rb, data_size, node->layout_version_read_from_disk);
     }
     assert(rb.ndone == rb.size);
 exit:
@@ -2112,7 +2112,7 @@ deserialize_and_upgrade_leaf_node(FTNODE node,
         if (has_end_to_end_checksum) {
             data_size -= sizeof(uint32_t);
         }
-        bn->data_buffer.initialize_from_data(n_in_buf, rb, data_size, node->layout_version_read_from_disk);
+        bn->data_buffer.deserialize_from_rbuf(n_in_buf, rb, data_size, node->layout_version_read_from_disk);
     }
 
     // Whatever this is must be less than the MSNs of every message above

ft/ftloader.cc

@@ -2917,7 +2917,7 @@ static void add_pair_to_leafnode (struct leaf_buf *lbuf, unsigned char *key, int
     // #3588 TODO just make a clean ule and append it to the omt
     // #3588 TODO can do the rebalancing here and avoid a lot of work later
     FTNODE leafnode = lbuf->node;
-    uint32_t idx = BLB_DATA(leafnode, 0)->omt_size();
+    uint32_t idx = BLB_DATA(leafnode, 0)->dmt_size();
     DBT thekey = { .data = key, .size = (uint32_t) keylen }; 
     DBT theval = { .data = val, .size = (uint32_t) vallen };
     FT_MSG_S cmd = { .type = FT_INSERT,

ft/fttypes.h

@@ -234,7 +234,7 @@ typedef struct cachetable *CACHETABLE;
 typedef struct cachefile *CACHEFILE;
 typedef struct ctpair *PAIR;
 typedef class checkpointer *CHECKPOINTER;
-typedef class bn_data *BN_DATA;
+class bn_data;
 
 /* tree command types */
 enum ft_msg_type {

ft/tests/dmt-test2.cc

@@ -131,27 +131,26 @@ struct val_type {
 };
 
 namespace toku {
-template<>
-class dmt_functor<val_type> {
+class vwriter {
     public:
-        size_t get_dmtdatain_t_size(void) const {
+        size_t get_size(void) const {
             size_t len = strlen(v.c);
             invariant(len < sizeof(val_type));
             return len + 1;
         }
-        void write_dmtdata_t_to(val_type *const dest) const {
+        void write_to(val_type *const dest) const {
             strcpy(dest->c, v.c);
         }
 
-        dmt_functor(const char* c) {
+        vwriter(const char* c) {
             invariant(strlen(c) < sizeof(val_type));
             strcpy(v.c, c);
         }
 
-        dmt_functor(const uint32_t klpair_len, val_type *const src) {
+        vwriter(const uint32_t klpair_len, val_type *const src) {
             invariant(strlen(src->c) < sizeof(val_type));
             strcpy(v.c, src->c);
-            invariant(klpair_len == get_dmtdatain_t_size());
+            invariant(klpair_len == get_size());
         }
     private:
         val_type v;
@@ -159,8 +158,7 @@ class dmt_functor<val_type> {
 }
 
 /* Globals */
-typedef toku::dmt<val_type, val_type*> vdmt;
-typedef toku::dmt_functor<val_type> vfunctor;
+typedef toku::dmt<val_type, val_type*, toku::vwriter> vdmt;
 
 const unsigned int random_seed = 0xFEADACBA;
 
@@ -211,7 +209,7 @@ static void test_builder_fixed(uint32_t len, uint32_t num) {
     builder.create(num, num * len);
 
     for (uint32_t i = 0; i < num; i++) {
-        vfunctor vfun(data[i]);
+        vwriter vfun(data[i]);
         builder.append(vfun);
     }
     invariant(builder.value_length_is_fixed());
@@ -230,7 +228,7 @@ static void test_builder_fixed(uint32_t len, uint32_t num) {
         v2.delete_at(change);
         fail_one_verify(len, num, &v2);
 
-        vfunctor vfun(data[change]);
+        vwriter vfun(data[change]);
         v2.insert_at(vfun, change);
         verify(len, num, &v2);
         v2.destroy();
@@ -258,7 +256,7 @@ static void test_builder_variable(uint32_t len, uint32_t len2, uint32_t num) {
     builder.create(num, (num-1) * len + len2);
 
     for (uint32_t i = 0; i < num; i++) {
-        vfunctor vfun(data[i]);
+        vwriter vfun(data[i]);
         builder.append(vfun);
     }
     invariant(!builder.value_length_is_fixed());

ft/tests/ft-serialize-test.cc

@@ -357,7 +357,7 @@ test_serialize_leaf_check_msn(enum ftnode_verify_type bft, bool do_clone) {
             if (bn > 0) {
                 assert(dest_ndd[bn].start >= dest_ndd[bn-1].start + dest_ndd[bn-1].size);
             }
-            for (uint32_t i = 0; i < BLB_DATA(dn, bn)->omt_size(); i++) {
+            for (uint32_t i = 0; i < BLB_DATA(dn, bn)->dmt_size(); i++) {
                 LEAFENTRY curr_le;
                 uint32_t curr_keylen;
                 void* curr_key;
@@ -431,7 +431,7 @@ test_serialize_leaf_with_large_pivots(enum ftnode_verify_type bft, bool do_clone
         if (i < nrows-1) {
             uint32_t keylen;
             void* curr_key;
-            BLB_DATA(&sn, i)->fetch_le_key_and_len(0, &keylen, &curr_key);
+            BLB_DATA(&sn, i)->fetch_key_and_len(0, &keylen, &curr_key);
             toku_memdup_dbt(&sn.childkeys[i], curr_key, keylen);
         }
     }
@@ -499,8 +499,8 @@ test_serialize_leaf_with_large_pivots(enum ftnode_verify_type bft, bool do_clone
             if (bn > 0) {
                 assert(dest_ndd[bn].start >= dest_ndd[bn-1].start + dest_ndd[bn-1].size);
             }
-            assert(BLB_DATA(dn, bn)->omt_size() > 0);
-            for (uint32_t i = 0; i < BLB_DATA(dn, bn)->omt_size(); i++) {
+            assert(BLB_DATA(dn, bn)->dmt_size() > 0);
+            for (uint32_t i = 0; i < BLB_DATA(dn, bn)->dmt_size(); i++) {
                 LEAFENTRY curr_le;
                 uint32_t curr_keylen;
                 void* curr_key;
@@ -631,8 +631,8 @@ test_serialize_leaf_with_many_rows(enum ftnode_verify_type bft, bool do_clone) {
             if (bn > 0) {
                 assert(dest_ndd[bn].start >= dest_ndd[bn-1].start + dest_ndd[bn-1].size);
             }
-            assert(BLB_DATA(dn, bn)->omt_size() > 0);
-            for (uint32_t i = 0; i < BLB_DATA(dn, bn)->omt_size(); i++) {
+            assert(BLB_DATA(dn, bn)->dmt_size() > 0);
+            for (uint32_t i = 0; i < BLB_DATA(dn, bn)->dmt_size(); i++) {
                 LEAFENTRY curr_le;
                 uint32_t curr_keylen;
                 void* curr_key;
@@ -781,8 +781,8 @@ test_serialize_leaf_with_large_rows(enum ftnode_verify_type bft, bool do_clone)
             if (bn > 0) {
                 assert(dest_ndd[bn].start >= dest_ndd[bn-1].start + dest_ndd[bn-1].size);
             }
-            assert(BLB_DATA(dn, bn)->omt_size() > 0);
-            for (uint32_t i = 0; i < BLB_DATA(dn, bn)->omt_size(); i++) {
+            assert(BLB_DATA(dn, bn)->dmt_size() > 0);
+            for (uint32_t i = 0; i < BLB_DATA(dn, bn)->dmt_size(); i++) {
                 LEAFENTRY curr_le;
                 uint32_t curr_keylen;
                 void* curr_key;
@@ -919,7 +919,7 @@ test_serialize_leaf_with_empty_basement_nodes(enum ftnode_verify_type bft, bool
             if (bn > 0) {
                 assert(dest_ndd[bn].start >= dest_ndd[bn-1].start + dest_ndd[bn-1].size);
             }
-            for (uint32_t i = 0; i < BLB_DATA(dn, bn)->omt_size(); i++) {
+            for (uint32_t i = 0; i < BLB_DATA(dn, bn)->dmt_size(); i++) {
                 LEAFENTRY curr_le;
                 uint32_t curr_keylen;
                 void* curr_key;
@@ -1040,7 +1040,7 @@ test_serialize_leaf_with_multiple_empty_basement_nodes(enum ftnode_verify_type b
             if (i > 0) {
                 assert(dest_ndd[i].start >= dest_ndd[i-1].start + dest_ndd[i-1].size);
             }
-            assert(BLB_DATA(dn, i)->omt_size() == 0);
+            assert(BLB_DATA(dn, i)->dmt_size() == 0);
         }
     }
     toku_ftnode_free(&dn);

ft/tests/make-tree.cc

@@ -119,7 +119,7 @@ append_leaf(FTNODE leafnode, void *key, size_t keylen, void *val, size_t vallen)
     DBT theval; toku_fill_dbt(&theval, val, vallen);
 
     // get an index that we can use to create a new leaf entry
-    uint32_t idx = BLB_DATA(leafnode, 0)->omt_size();
+    uint32_t idx = BLB_DATA(leafnode, 0)->dmt_size();
 
     MSN msn = next_dummymsn();
 

ft/tests/orthopush-flush.cc

@@ -733,7 +733,7 @@ flush_to_leaf(FT_HANDLE t, bool make_leaf_up_to_date, bool use_flush) {
 
     int total_messages = 0;
     for (i = 0; i < 8; ++i) {
-        total_messages += BLB_DATA(child, i)->omt_size();
+        total_messages += BLB_DATA(child, i)->dmt_size();
     }
     assert(total_messages <= num_parent_messages + num_child_messages);
 
@@ -746,7 +746,7 @@ flush_to_leaf(FT_HANDLE t, bool make_leaf_up_to_date, bool use_flush) {
     memset(parent_messages_present, 0, sizeof parent_messages_present);
     memset(child_messages_present, 0, sizeof child_messages_present);
     for (int j = 0; j < 8; ++j) {
-        uint32_t len = BLB_DATA(child, j)->omt_size();
+        uint32_t len = BLB_DATA(child, j)->dmt_size();
         for (uint32_t idx = 0; idx < len; ++idx) {
             LEAFENTRY le;
             DBT keydbt, valdbt;
@@ -968,7 +968,7 @@ flush_to_leaf_with_keyrange(FT_HANDLE t, bool make_leaf_up_to_date) {
 
     int total_messages = 0;
     for (i = 0; i < 8; ++i) {
-        total_messages += BLB_DATA(child, i)->omt_size();
+        total_messages += BLB_DATA(child, i)->dmt_size();
     }
     assert(total_messages <= num_parent_messages + num_child_messages);
 
@@ -1144,10 +1144,10 @@ compare_apply_and_flush(FT_HANDLE t, bool make_leaf_up_to_date) {
     toku_ftnode_free(&parentnode);
 
     for (int j = 0; j < 8; ++j) {
-        BN_DATA first = BLB_DATA(child1, j);
-        BN_DATA second = BLB_DATA(child2, j);
-        uint32_t len = first->omt_size();
-        assert(len == second->omt_size());
+        bn_data* first = BLB_DATA(child1, j);
+        bn_data* second = BLB_DATA(child2, j);
+        uint32_t len = first->dmt_size();
+        assert(len == second->dmt_size());
         for (uint32_t idx = 0; idx < len; ++idx) {
             LEAFENTRY le1, le2;
             DBT key1dbt, val1dbt, key2dbt, val2dbt;

ft/tests/test-checkpoint-during-merge.cc

@@ -348,7 +348,7 @@ doit (int state) {
         assert(node->height == 0);
         assert(!node->dirty);
         assert(node->n_children == 1);
-        assert(BLB_DATA(node, 0)->omt_size() == 1);
+        assert(BLB_DATA(node, 0)->dmt_size() == 1);
         toku_unpin_ftnode_off_client_thread(c_ft->ft, node);
 
         toku_pin_ftnode_off_client_thread(
@@ -364,7 +364,7 @@ doit (int state) {
         assert(node->height == 0);
         assert(!node->dirty);
         assert(node->n_children == 1);
-        assert(BLB_DATA(node, 0)->omt_size() == 1);
+        assert(BLB_DATA(node, 0)->dmt_size() == 1);
         toku_unpin_ftnode_off_client_thread(c_ft->ft, node);
     }
     else if (state == ft_flush_aflter_merge || state == flt_flush_before_unpin_remove) {
@@ -381,7 +381,7 @@ doit (int state) {
         assert(node->height == 0);
         assert(!node->dirty);
         assert(node->n_children == 1);
-        assert(BLB_DATA(node, 0)->omt_size() == 2);
+        assert(BLB_DATA(node, 0)->dmt_size() == 2);
         toku_unpin_ftnode_off_client_thread(c_ft->ft, node);
     }
     else {

ft/tests/test-checkpoint-during-rebalance.cc

@@ -359,7 +359,7 @@ doit (int state) {
     assert(node->height == 0);
     assert(!node->dirty);
     assert(node->n_children == 1);
-    assert(BLB_DATA(node, 0)->omt_size() == 2);
+    assert(BLB_DATA(node, 0)->dmt_size() == 2);
     toku_unpin_ftnode_off_client_thread(c_ft->ft, node);
     
     toku_pin_ftnode_off_client_thread(
@@ -375,7 +375,7 @@ doit (int state) {
     assert(node->height == 0);
     assert(!node->dirty);
     assert(node->n_children == 1);
-    assert(BLB_DATA(node, 0)->omt_size() == 2);
+    assert(BLB_DATA(node, 0)->dmt_size() == 2);
     toku_unpin_ftnode_off_client_thread(c_ft->ft, node);
 
 

ft/tests/test-checkpoint-during-split.cc

@@ -342,7 +342,7 @@ doit (bool after_split) {
         assert(node->height == 0);
         assert(!node->dirty);
         assert(node->n_children == 1);
-        assert(BLB_DATA(node, 0)->omt_size() == 1);
+        assert(BLB_DATA(node, 0)->dmt_size() == 1);
         toku_unpin_ftnode_off_client_thread(c_ft->ft, node);
 
         toku_pin_ftnode_off_client_thread(
@@ -358,7 +358,7 @@ doit (bool after_split) {
         assert(node->height == 0);
         assert(!node->dirty);
         assert(node->n_children == 1);
-        assert(BLB_DATA(node, 0)->omt_size() == 1);
+        assert(BLB_DATA(node, 0)->dmt_size() == 1);
         toku_unpin_ftnode_off_client_thread(c_ft->ft, node);
     }
     else {
@@ -375,7 +375,7 @@ doit (bool after_split) {
         assert(node->height == 0);
         assert(!node->dirty);
         assert(node->n_children == 1);
-        assert(BLB_DATA(node, 0)->omt_size() == 2);
+        assert(BLB_DATA(node, 0)->dmt_size() == 2);
         toku_unpin_ftnode_off_client_thread(c_ft->ft, node);
     }
 

ft/tests/verify-bad-msn.cc

@@ -122,7 +122,7 @@ append_leaf(FTNODE leafnode, void *key, size_t keylen, void *val, size_t vallen)
     DBT theval; toku_fill_dbt(&theval, val, vallen);
 
     // get an index that we can use to create a new leaf entry
-    uint32_t idx = BLB_DATA(leafnode, 0)->omt_size();
+    uint32_t idx = BLB_DATA(leafnode, 0)->dmt_size();
 
     MSN msn = next_dummymsn();
 

ft/tests/verify-bad-pivots.cc

@@ -111,7 +111,7 @@ append_leaf(FTNODE leafnode, void *key, size_t keylen, void *val, size_t vallen)
     DBT theval; toku_fill_dbt(&theval, val, vallen);
 
     // get an index that we can use to create a new leaf entry
-    uint32_t idx = BLB_DATA(leafnode, 0)->omt_size();
+    uint32_t idx = BLB_DATA(leafnode, 0)->dmt_size();
 
     // apply an insert to the leaf node
     MSN msn = next_dummymsn();

ft/tests/verify-dup-in-leaf.cc

@@ -112,7 +112,7 @@ append_leaf(FTNODE leafnode, void *key, size_t keylen, void *val, size_t vallen)
     DBT theval; toku_fill_dbt(&theval, val, vallen);
 
     // get an index that we can use to create a new leaf entry
-    uint32_t idx = BLB_DATA(leafnode, 0)->omt_size();
+    uint32_t idx = BLB_DATA(leafnode, 0)->dmt_size();
 
     // apply an insert to the leaf node
     MSN msn = next_dummymsn();

ft/tests/verify-dup-pivots.cc

@@ -111,7 +111,7 @@ append_leaf(FTNODE leafnode, void *key, size_t keylen, void *val, size_t vallen)
     DBT theval; toku_fill_dbt(&theval, val, vallen);
 
     // get an index that we can use to create a new leaf entry
-    uint32_t idx = BLB_DATA(leafnode, 0)->omt_size();
+    uint32_t idx = BLB_DATA(leafnode, 0)->dmt_size();
 
     // apply an insert to the leaf node
     MSN msn = next_dummymsn();

ft/tests/verify-misrouted-msgs.cc

@@ -112,7 +112,7 @@ append_leaf(FTNODE leafnode, void *key, size_t keylen, void *val, size_t vallen)
     DBT theval; toku_fill_dbt(&theval, val, vallen);
 
     // get an index that we can use to create a new leaf entry
-    uint32_t idx = BLB_DATA(leafnode, 0)->omt_size();
+    uint32_t idx = BLB_DATA(leafnode, 0)->dmt_size();
 
     // apply an insert to the leaf node
     MSN msn = next_dummymsn();

ft/tests/verify-unsorted-leaf.cc

@@ -114,7 +114,7 @@ append_leaf(FTNODE leafnode, void *key, size_t keylen, void *val, size_t vallen)
     toku_fill_dbt(&theval, val, vallen);
 
     // get an index that we can use to create a new leaf entry
-    uint32_t idx = BLB_DATA(leafnode, 0)->omt_size();
+    uint32_t idx = BLB_DATA(leafnode, 0)->dmt_size();
 
     // apply an insert to the leaf node
     MSN msn = next_dummymsn();

ft/tests/verify-unsorted-pivots.cc

@@ -111,7 +111,7 @@ append_leaf(FTNODE leafnode, void *key, size_t keylen, void *val, size_t vallen)
     DBT theval; toku_fill_dbt(&theval, val, vallen);
 
     // get an index that we can use to create a new leaf entry
-    uint32_t idx = BLB_DATA(leafnode, 0)->omt_size();
+    uint32_t idx = BLB_DATA(leafnode, 0)->dmt_size();
 
     // apply an insert to the leaf node
     MSN msn = next_dummymsn();

ft/tokuftdump.cc

@@ -315,9 +315,9 @@ dump_node (int f, BLOCKNUM blocknum, FT h) {
             }
         } else {
             printf(" n_bytes_in_buffer= %" PRIu64 "", BLB_DATA(n, i)->get_disk_size());
-            printf(" items_in_buffer=%u\n", BLB_DATA(n, i)->omt_size());
+            printf(" items_in_buffer=%u\n", BLB_DATA(n, i)->dmt_size());
             if (dump_data) {
-                BLB_DATA(n, i)->omt_iterate<void, print_le>(NULL);
+                BLB_DATA(n, i)->dmt_iterate<void, print_le>(NULL);
             }
         }
     }

util/dmt.h

@@ -197,43 +197,40 @@ class dmt_node_templated {
 
 using namespace toku::dmt_internal;
 
-// Each data type used in a dmt requires a dmt_functor (allows you to insert/etc with dynamic sized types).
+// Each data type used in a dmt requires a dmt_writer class (allows you to insert/etc with dynamic sized types).
 // There is no default implementation.
-template<typename dmtdata_t>
-class dmt_functor {
-    // Ensures that if you forget to use partial specialization this compile error will remind you to use it.
-    // We would use static_assert(false, ...) here except that it would cause a compile error even if dmt_functor<>
-    // We instead use an expression that evaluates to false that the compiler won't evaluate unless dmt_functor<> is used.
-    static_assert(!std::is_same<dmtdata_t, dmtdata_t>::value, "Cannot use default dmt_functor<>. Use partial specialization.");
-    // Defines the interface:
-    static size_t get_dmtdata_t_size(const dmtdata_t &) { return 0; }
-    size_t get_dmtdatain_t_size(void) { return 0; }
-    void write_dmtdata_t_to(dmtdata_t *const dest) {}
-};
+// A dmtwriter instance handles reading/writing 'dmtdata_t's to/from the dmt.
+// The class must implement the following functions:
+//      The size required in a dmt for the dmtdata_t represented:
+//          size_t get_size(void) const;  
+//      Write the dmtdata_t to memory owned by a dmt:
+//          void write_to(dmtdata_t *const dest) const;
+//      Constructor (others are allowed, but this one is required)
+//          dmtwriter(const uint32_t dmtdata_t_len, dmtdata_t *const src)
 
 template<typename dmtdata_t,
-         typename dmtdataout_t=dmtdata_t
+         typename dmtdataout_t,
+         typename dmtwriter_t
         >
 class dmt {
 private:
     typedef dmt_node_templated<dmtdata_t> dmt_node;
-    typedef dmt_functor<dmtdata_t> dmtdatain_t;
 
 public:
     static const uint8_t ALIGNMENT = 4;
 
     class builder {
     public:
-        void append(const dmtdatain_t &value);
+        void append(const dmtwriter_t &value);
         void create(uint32_t n_values, uint32_t n_value_bytes);
         bool value_length_is_fixed(void);
-        void build(dmt<dmtdata_t, dmtdataout_t> *dest);
+        void build(dmt<dmtdata_t, dmtdataout_t, dmtwriter_t> *dest);
     private:
         uint32_t max_values;
         uint32_t max_value_bytes;
         node_offset *sorted_node_offsets;
         bool temp_valid;
-        dmt<dmtdata_t, dmtdataout_t> temp;
+        dmt<dmtdata_t, dmtdataout_t, dmtwriter_t> temp;
     };
 
     /**
@@ -306,7 +303,7 @@ class dmt {
      * Rationale: Some future implementation may be O(\log N) worst-case time, but O(\log N) amortized is good enough for now.
      */
     template<typename dmtcmp_t, int (*h)(const uint32_t size, const dmtdata_t &, const dmtcmp_t &)>
-    int insert(const dmtdatain_t &value, const dmtcmp_t &v, uint32_t *const idx);
+    int insert(const dmtwriter_t &value, const dmtcmp_t &v, uint32_t *const idx);
 
     /**
      * Effect: Increases indexes of all items at slot >= idx by 1.
@@ -318,7 +315,7 @@ class dmt {
      * Performance: time=O(\log N) amortized time.
      * Rationale: Some future implementation may be O(\log N) worst-case time, but O(\log N) amortized is good enough for now.
      */
-    int insert_at(const dmtdatain_t &value, const uint32_t idx);
+    int insert_at(const dmtwriter_t &value, const uint32_t idx);
 
     /**
      * Effect: Delete the item in slot idx.
@@ -557,9 +554,9 @@ class dmt {
 
     uint32_t nweight(const subtree &subtree) const;
 
-    node_offset node_malloc_and_set_value(const dmtdatain_t &value);
+    node_offset node_malloc_and_set_value(const dmtwriter_t &value);
 
-    void node_set_value(dmt_node *n, const dmtdatain_t &value);
+    void node_set_value(dmt_node *n, const dmtwriter_t &value);
 
     void node_free(const subtree &st);
 
@@ -567,15 +564,15 @@ class dmt {
 
     void convert_to_tree(void);
 
-    void maybe_resize_tree(const dmtdatain_t * value);
+    void maybe_resize_tree(const dmtwriter_t * value);
 
     bool will_need_rebalance(const subtree &subtree, const int leftmod, const int rightmod) const;
 
     __attribute__((nonnull))
-    void insert_internal(subtree *const subtreep, const dmtdatain_t &value, const uint32_t idx, subtree **const rebalance_subtree);
+    void insert_internal(subtree *const subtreep, const dmtwriter_t &value, const uint32_t idx, subtree **const rebalance_subtree);
 
     template<bool with_resize>
-    int insert_at_array_end(const dmtdatain_t& value_in);
+    int insert_at_array_end(const dmtwriter_t& value_in);
 
     dmtdata_t * alloc_array_value_end(void);