/
row0ftsort.cc
1813 lines (1495 loc) · 48.5 KB
/
row0ftsort.cc
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
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
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
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
/*****************************************************************************
Copyright (c) 2010, 2016, Oracle and/or its affiliates. All Rights Reserved.
Copyright (c) 2015, 2019, MariaDB Corporation.
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; version 2 of the License.
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.,
51 Franklin Street, Suite 500, Boston, MA 02110-1335 USA
*****************************************************************************/
/**************************************************//**
@file row/row0ftsort.cc
Create Full Text Index with (parallel) merge sort
Created 10/13/2010 Jimmy Yang
*******************************************************/
#include "row0ftsort.h"
#include "dict0dict.h"
#include "row0merge.h"
#include "row0row.h"
#include "btr0cur.h"
#include "fts0plugin.h"
#include "log0crypt.h"
/** Read the next record to buffer N.
@param N index into array of merge info structure */
#define ROW_MERGE_READ_GET_NEXT(N) \
do { \
b[N] = row_merge_read_rec( \
block[N], buf[N], b[N], index, \
fd[N], &foffs[N], &mrec[N], offsets[N], \
crypt_block[N], space); \
if (UNIV_UNLIKELY(!b[N])) { \
if (mrec[N]) { \
goto exit; \
} \
} \
} while (0)
/** Parallel sort degree */
ulong fts_sort_pll_degree = 2;
/*********************************************************************//**
Create a temporary "fts sort index" used to merge sort the
tokenized doc string. The index has three "fields":
1) Tokenized word,
2) Doc ID (depend on number of records to sort, it can be a 4 bytes or 8 bytes
integer value)
3) Word's position in original doc.
@see fts_create_one_index_table()
@return dict_index_t structure for the fts sort index */
dict_index_t*
row_merge_create_fts_sort_index(
/*============================*/
dict_index_t* index, /*!< in: Original FTS index
based on which this sort index
is created */
dict_table_t* table, /*!< in,out: table that FTS index
is being created on */
ibool* opt_doc_id_size)
/*!< out: whether to use 4 bytes
instead of 8 bytes integer to
store Doc ID during sort */
{
dict_index_t* new_index;
dict_field_t* field;
dict_field_t* idx_field;
CHARSET_INFO* charset;
// FIXME: This name shouldn't be hard coded here.
new_index = dict_mem_index_create(table, "tmp_fts_idx", DICT_FTS, 3);
new_index->id = index->id;
new_index->n_uniq = FTS_NUM_FIELDS_SORT;
new_index->n_def = FTS_NUM_FIELDS_SORT;
new_index->cached = TRUE;
new_index->parser = index->parser;
idx_field = dict_index_get_nth_field(index, 0);
charset = fts_index_get_charset(index);
/* The first field is on the Tokenized Word */
field = dict_index_get_nth_field(new_index, 0);
field->name = NULL;
field->prefix_len = 0;
field->col = new(mem_heap_zalloc(new_index->heap, sizeof(dict_col_t)))
dict_col_t();
field->col->prtype = idx_field->col->prtype | DATA_NOT_NULL;
field->col->mtype = charset == &my_charset_latin1
? DATA_VARCHAR : DATA_VARMYSQL;
field->col->mbminlen = idx_field->col->mbminlen;
field->col->mbmaxlen = idx_field->col->mbmaxlen;
field->col->len = HA_FT_MAXCHARLEN * unsigned(field->col->mbmaxlen);
field->fixed_len = 0;
/* Doc ID */
field = dict_index_get_nth_field(new_index, 1);
field->name = NULL;
field->prefix_len = 0;
field->col = new(mem_heap_zalloc(new_index->heap, sizeof(dict_col_t)))
dict_col_t();
field->col->mtype = DATA_INT;
*opt_doc_id_size = FALSE;
/* Check whether we can use 4 bytes instead of 8 bytes integer
field to hold the Doc ID, thus reduce the overall sort size */
if (DICT_TF2_FLAG_IS_SET(table, DICT_TF2_FTS_ADD_DOC_ID)) {
/* If Doc ID column is being added by this create
index, then just check the number of rows in the table */
if (dict_table_get_n_rows(table) < MAX_DOC_ID_OPT_VAL) {
*opt_doc_id_size = TRUE;
}
} else {
doc_id_t max_doc_id;
/* If the Doc ID column is supplied by user, then
check the maximum Doc ID in the table */
max_doc_id = fts_get_max_doc_id((dict_table_t*) table);
if (max_doc_id && max_doc_id < MAX_DOC_ID_OPT_VAL) {
*opt_doc_id_size = TRUE;
}
}
if (*opt_doc_id_size) {
field->col->len = sizeof(ib_uint32_t);
field->fixed_len = sizeof(ib_uint32_t);
} else {
field->col->len = FTS_DOC_ID_LEN;
field->fixed_len = FTS_DOC_ID_LEN;
}
field->col->prtype = DATA_NOT_NULL | DATA_BINARY_TYPE;
/* The third field is on the word's position in the original doc */
field = dict_index_get_nth_field(new_index, 2);
field->name = NULL;
field->prefix_len = 0;
field->col = new(mem_heap_zalloc(new_index->heap, sizeof(dict_col_t)))
dict_col_t();
field->col->mtype = DATA_INT;
field->col->len = 4 ;
field->fixed_len = 4;
field->col->prtype = DATA_NOT_NULL;
return(new_index);
}
/*********************************************************************//**
Initialize FTS parallel sort structures.
@return TRUE if all successful */
ibool
row_fts_psort_info_init(
/*====================*/
trx_t* trx, /*!< in: transaction */
row_merge_dup_t* dup, /*!< in,own: descriptor of
FTS index being created */
const dict_table_t* new_table,/*!< in: table on which indexes are
created */
ibool opt_doc_id_size,
/*!< in: whether to use 4 bytes
instead of 8 bytes integer to
store Doc ID during sort */
fts_psort_t** psort, /*!< out: parallel sort info to be
instantiated */
fts_psort_t** merge) /*!< out: parallel merge info
to be instantiated */
{
ulint i;
ulint j;
fts_psort_common_t* common_info = NULL;
fts_psort_t* psort_info = NULL;
fts_psort_t* merge_info = NULL;
ulint block_size;
ibool ret = TRUE;
bool encrypted = false;
block_size = 3 * srv_sort_buf_size;
*psort = psort_info = static_cast<fts_psort_t*>(ut_zalloc_nokey(
fts_sort_pll_degree * sizeof *psort_info));
if (!psort_info) {
ut_free(dup);
return(FALSE);
}
/* Common Info for all sort threads */
common_info = static_cast<fts_psort_common_t*>(
ut_malloc_nokey(sizeof *common_info));
if (!common_info) {
ut_free(dup);
ut_free(psort_info);
return(FALSE);
}
common_info->dup = dup;
common_info->new_table = (dict_table_t*) new_table;
common_info->trx = trx;
common_info->all_info = psort_info;
common_info->sort_event = os_event_create(0);
common_info->merge_event = os_event_create(0);
common_info->opt_doc_id_size = opt_doc_id_size;
if (log_tmp_is_encrypted()) {
encrypted = true;
}
ut_ad(trx->mysql_thd != NULL);
const char* path = thd_innodb_tmpdir(trx->mysql_thd);
/* There will be FTS_NUM_AUX_INDEX number of "sort buckets" for
each parallel sort thread. Each "sort bucket" holds records for
a particular "FTS index partition" */
for (j = 0; j < fts_sort_pll_degree; j++) {
UT_LIST_INIT(
psort_info[j].fts_doc_list, &fts_doc_item_t::doc_list);
for (i = 0; i < FTS_NUM_AUX_INDEX; i++) {
psort_info[j].merge_file[i] =
static_cast<merge_file_t*>(
ut_zalloc_nokey(sizeof(merge_file_t)));
if (!psort_info[j].merge_file[i]) {
ret = FALSE;
goto func_exit;
}
psort_info[j].merge_buf[i] = row_merge_buf_create(
dup->index);
if (row_merge_file_create(psort_info[j].merge_file[i],
path) == OS_FILE_CLOSED) {
goto func_exit;
}
/* Need to align memory for O_DIRECT write */
psort_info[j].block_alloc[i] =
static_cast<row_merge_block_t*>(ut_malloc_nokey(
block_size + 1024));
psort_info[j].merge_block[i] =
static_cast<row_merge_block_t*>(
ut_align(
psort_info[j].block_alloc[i], 1024));
if (!psort_info[j].merge_block[i]) {
ret = FALSE;
goto func_exit;
}
/* If tablespace is encrypted, allocate additional buffer for
encryption/decryption. */
if (encrypted) {
/* Need to align memory for O_DIRECT write */
psort_info[j].crypt_alloc[i] =
static_cast<row_merge_block_t*>(ut_malloc_nokey(
block_size + 1024));
psort_info[j].crypt_block[i] =
static_cast<row_merge_block_t*>(
ut_align(
psort_info[j].crypt_alloc[i], 1024));
if (!psort_info[j].crypt_block[i]) {
ret = FALSE;
goto func_exit;
}
} else {
psort_info[j].crypt_alloc[i] = NULL;
psort_info[j].crypt_block[i] = NULL;
}
}
psort_info[j].child_status = 0;
psort_info[j].state = 0;
psort_info[j].psort_common = common_info;
psort_info[j].error = DB_SUCCESS;
psort_info[j].memory_used = 0;
mutex_create(LATCH_ID_FTS_PLL_TOKENIZE, &psort_info[j].mutex);
}
/* Initialize merge_info structures parallel merge and insert
into auxiliary FTS tables (FTS_INDEX_TABLE) */
*merge = merge_info = static_cast<fts_psort_t*>(
ut_malloc_nokey(FTS_NUM_AUX_INDEX * sizeof *merge_info));
for (j = 0; j < FTS_NUM_AUX_INDEX; j++) {
merge_info[j].child_status = 0;
merge_info[j].state = 0;
merge_info[j].psort_common = common_info;
}
func_exit:
if (!ret) {
row_fts_psort_info_destroy(psort_info, merge_info);
}
return(ret);
}
/*********************************************************************//**
Clean up and deallocate FTS parallel sort structures, and close the
merge sort files */
void
row_fts_psort_info_destroy(
/*=======================*/
fts_psort_t* psort_info, /*!< parallel sort info */
fts_psort_t* merge_info) /*!< parallel merge info */
{
ulint i;
ulint j;
if (psort_info) {
for (j = 0; j < fts_sort_pll_degree; j++) {
for (i = 0; i < FTS_NUM_AUX_INDEX; i++) {
if (psort_info[j].merge_file[i]) {
row_merge_file_destroy(
psort_info[j].merge_file[i]);
}
ut_free(psort_info[j].block_alloc[i]);
ut_free(psort_info[j].merge_file[i]);
if (psort_info[j].crypt_alloc[i]) {
ut_free(psort_info[j].crypt_alloc[i]);
}
}
mutex_free(&psort_info[j].mutex);
}
os_event_destroy(merge_info[0].psort_common->sort_event);
os_event_destroy(merge_info[0].psort_common->merge_event);
ut_free(merge_info[0].psort_common->dup);
ut_free(merge_info[0].psort_common);
ut_free(psort_info);
}
ut_free(merge_info);
}
/*********************************************************************//**
Free up merge buffers when merge sort is done */
void
row_fts_free_pll_merge_buf(
/*=======================*/
fts_psort_t* psort_info) /*!< in: parallel sort info */
{
ulint j;
ulint i;
if (!psort_info) {
return;
}
for (j = 0; j < fts_sort_pll_degree; j++) {
for (i = 0; i < FTS_NUM_AUX_INDEX; i++) {
row_merge_buf_free(psort_info[j].merge_buf[i]);
}
}
return;
}
/*********************************************************************//**
FTS plugin parser 'myql_add_word' callback function for row merge.
Refer to 'st_mysql_ftparser_param' for more detail.
@return always returns 0 */
static
int
row_merge_fts_doc_add_word_for_parser(
/*==================================*/
MYSQL_FTPARSER_PARAM *param, /* in: parser paramter */
const char *word, /* in: token word */
int word_len, /* in: word len */
MYSQL_FTPARSER_BOOLEAN_INFO* boolean_info) /* in: boolean info */
{
fts_string_t str;
fts_tokenize_ctx_t* t_ctx;
row_fts_token_t* fts_token;
byte* ptr;
ut_ad(param);
ut_ad(param->mysql_ftparam);
ut_ad(word);
ut_ad(boolean_info);
t_ctx = static_cast<fts_tokenize_ctx_t*>(param->mysql_ftparam);
ut_ad(t_ctx);
str.f_str = (byte*)(word);
str.f_len = ulint(word_len);
str.f_n_char = fts_get_token_size(
(CHARSET_INFO*)param->cs, word, ulint(word_len));
/* JAN: TODO: MySQL 5.7 FTS
ut_ad(boolean_info->position >= 0);
*/
ptr = static_cast<byte*>(ut_malloc_nokey(sizeof(row_fts_token_t)
+ sizeof(fts_string_t) + str.f_len));
fts_token = reinterpret_cast<row_fts_token_t*>(ptr);
fts_token->text = reinterpret_cast<fts_string_t*>(
ptr + sizeof(row_fts_token_t));
fts_token->text->f_str = static_cast<byte*>(
ptr + sizeof(row_fts_token_t) + sizeof(fts_string_t));
fts_token->text->f_len = str.f_len;
fts_token->text->f_n_char = str.f_n_char;
memcpy(fts_token->text->f_str, str.f_str, str.f_len);
/* JAN: TODO: MySQL 5.7 FTS
fts_token->position = boolean_info->position;
*/
/* Add token to list */
UT_LIST_ADD_LAST(t_ctx->fts_token_list, fts_token);
return(0);
}
/*********************************************************************//**
Tokenize by fts plugin parser */
static
void
row_merge_fts_doc_tokenize_by_parser(
/*=================================*/
fts_doc_t* doc, /* in: doc to tokenize */
st_mysql_ftparser* parser, /* in: plugin parser instance */
fts_tokenize_ctx_t* t_ctx) /* in/out: tokenize ctx instance */
{
MYSQL_FTPARSER_PARAM param;
ut_a(parser);
/* Set paramters for param */
param.mysql_parse = fts_tokenize_document_internal;
param.mysql_add_word = row_merge_fts_doc_add_word_for_parser;
param.mysql_ftparam = t_ctx;
param.cs = doc->charset;
param.doc = reinterpret_cast<char*>(doc->text.f_str);
param.length = static_cast<int>(doc->text.f_len);
param.mode= MYSQL_FTPARSER_SIMPLE_MODE;
PARSER_INIT(parser, ¶m);
/* We assume parse returns successfully here. */
parser->parse(¶m);
PARSER_DEINIT(parser, ¶m);
}
/*********************************************************************//**
Tokenize incoming text data and add to the sort buffer.
@see row_merge_buf_encode()
@return TRUE if the record passed, FALSE if out of space */
static
ibool
row_merge_fts_doc_tokenize(
/*=======================*/
row_merge_buf_t** sort_buf, /*!< in/out: sort buffer */
doc_id_t doc_id, /*!< in: Doc ID */
fts_doc_t* doc, /*!< in: Doc to be tokenized */
merge_file_t** merge_file, /*!< in/out: merge file */
ibool opt_doc_id_size,/*!< in: whether to use 4 bytes
instead of 8 bytes integer to
store Doc ID during sort*/
fts_tokenize_ctx_t* t_ctx) /*!< in/out: tokenize context */
{
ulint inc = 0;
fts_string_t str;
ulint len;
row_merge_buf_t* buf;
dfield_t* field;
fts_string_t t_str;
ibool buf_full = FALSE;
byte str_buf[FTS_MAX_WORD_LEN + 1];
ulint data_size[FTS_NUM_AUX_INDEX];
ulint n_tuple[FTS_NUM_AUX_INDEX];
st_mysql_ftparser* parser;
t_str.f_n_char = 0;
t_ctx->buf_used = 0;
memset(n_tuple, 0, FTS_NUM_AUX_INDEX * sizeof(ulint));
memset(data_size, 0, FTS_NUM_AUX_INDEX * sizeof(ulint));
parser = sort_buf[0]->index->parser;
/* Tokenize the data and add each word string, its corresponding
doc id and position to sort buffer */
while (t_ctx->processed_len < doc->text.f_len) {
ulint idx = 0;
ulint cur_len;
doc_id_t write_doc_id;
row_fts_token_t* fts_token = NULL;
if (parser != NULL) {
if (t_ctx->processed_len == 0) {
UT_LIST_INIT(t_ctx->fts_token_list, &row_fts_token_t::token_list);
/* Parse the whole doc and cache tokens */
row_merge_fts_doc_tokenize_by_parser(doc,
parser, t_ctx);
/* Just indictate we have parsed all the word */
t_ctx->processed_len += 1;
}
/* Then get a token */
fts_token = UT_LIST_GET_FIRST(t_ctx->fts_token_list);
if (fts_token) {
str.f_len = fts_token->text->f_len;
str.f_n_char = fts_token->text->f_n_char;
str.f_str = fts_token->text->f_str;
} else {
ut_ad(UT_LIST_GET_LEN(t_ctx->fts_token_list) == 0);
/* Reach the end of the list */
t_ctx->processed_len = doc->text.f_len;
break;
}
} else {
inc = innobase_mysql_fts_get_token(
doc->charset,
doc->text.f_str + t_ctx->processed_len,
doc->text.f_str + doc->text.f_len, &str);
ut_a(inc > 0);
}
/* Ignore string whose character number is less than
"fts_min_token_size" or more than "fts_max_token_size" */
if (!fts_check_token(&str, NULL, NULL)) {
if (parser != NULL) {
UT_LIST_REMOVE(t_ctx->fts_token_list, fts_token);
ut_free(fts_token);
} else {
t_ctx->processed_len += inc;
}
continue;
}
t_str.f_len = innobase_fts_casedn_str(
doc->charset, (char*) str.f_str, str.f_len,
(char*) &str_buf, FTS_MAX_WORD_LEN + 1);
t_str.f_str = (byte*) &str_buf;
/* if "cached_stopword" is defined, ignore words in the
stopword list */
if (!fts_check_token(&str, t_ctx->cached_stopword,
doc->charset)) {
if (parser != NULL) {
UT_LIST_REMOVE(t_ctx->fts_token_list, fts_token);
ut_free(fts_token);
} else {
t_ctx->processed_len += inc;
}
continue;
}
/* There are FTS_NUM_AUX_INDEX auxiliary tables, find
out which sort buffer to put this word record in */
t_ctx->buf_used = fts_select_index(
doc->charset, t_str.f_str, t_str.f_len);
buf = sort_buf[t_ctx->buf_used];
ut_a(t_ctx->buf_used < FTS_NUM_AUX_INDEX);
idx = t_ctx->buf_used;
mtuple_t* mtuple = &buf->tuples[buf->n_tuples + n_tuple[idx]];
field = mtuple->fields = static_cast<dfield_t*>(
mem_heap_alloc(buf->heap,
FTS_NUM_FIELDS_SORT * sizeof *field));
/* The first field is the tokenized word */
dfield_set_data(field, t_str.f_str, t_str.f_len);
len = dfield_get_len(field);
dict_col_copy_type(dict_index_get_nth_col(buf->index, 0), &field->type);
field->type.prtype |= DATA_NOT_NULL;
ut_ad(len <= field->type.len);
/* For the temporary file, row_merge_buf_encode() uses
1 byte for representing the number of extra_size bytes.
This number will always be 1, because for this 3-field index
consisting of one variable-size column, extra_size will always
be 1 or 2, which can be encoded in one byte.
The extra_size is 1 byte if the length of the
variable-length column is less than 128 bytes or the
maximum length is less than 256 bytes. */
/* One variable length column, word with its lenght less than
fts_max_token_size, add one extra size and one extra byte.
Since the max length for FTS token now is larger than 255,
so we will need to signify length byte itself, so only 1 to 128
bytes can be used for 1 bytes, larger than that 2 bytes. */
if (len < 128 || field->type.len < 256) {
/* Extra size is one byte. */
cur_len = 2 + len;
} else {
/* Extra size is two bytes. */
cur_len = 3 + len;
}
dfield_dup(field, buf->heap);
field++;
/* The second field is the Doc ID */
ib_uint32_t doc_id_32_bit;
if (!opt_doc_id_size) {
fts_write_doc_id((byte*) &write_doc_id, doc_id);
dfield_set_data(
field, &write_doc_id, sizeof(write_doc_id));
} else {
mach_write_to_4(
(byte*) &doc_id_32_bit, (ib_uint32_t) doc_id);
dfield_set_data(
field, &doc_id_32_bit, sizeof(doc_id_32_bit));
}
len = field->len;
ut_ad(len == FTS_DOC_ID_LEN || len == sizeof(ib_uint32_t));
field->type.mtype = DATA_INT;
field->type.prtype = DATA_NOT_NULL | DATA_BINARY_TYPE;
field->type.len = len;
field->type.mbminlen = 0;
field->type.mbmaxlen = 0;
cur_len += len;
dfield_dup(field, buf->heap);
++field;
/* The third field is the position.
MySQL 5.7 changed the fulltext parser plugin interface
by adding MYSQL_FTPARSER_BOOLEAN_INFO::position.
Below we assume that the field is always 0. */
ulint pos = t_ctx->init_pos;
byte position[4];
if (parser == NULL) {
pos += t_ctx->processed_len + inc - str.f_len;
}
len = 4;
mach_write_to_4(position, pos);
dfield_set_data(field, &position, len);
field->type.mtype = DATA_INT;
field->type.prtype = DATA_NOT_NULL;
field->type.len = len;
field->type.mbminlen = 0;
field->type.mbmaxlen = 0;
cur_len += len;
dfield_dup(field, buf->heap);
/* Reserve one byte for the end marker of row_merge_block_t */
if (buf->total_size + data_size[idx] + cur_len
>= srv_sort_buf_size - 1) {
buf_full = TRUE;
break;
}
/* Increment the number of tuples */
n_tuple[idx]++;
if (parser != NULL) {
UT_LIST_REMOVE(t_ctx->fts_token_list, fts_token);
ut_free(fts_token);
} else {
t_ctx->processed_len += inc;
}
data_size[idx] += cur_len;
}
/* Update the data length and the number of new word tuples
added in this round of tokenization */
for (ulint i = 0; i < FTS_NUM_AUX_INDEX; i++) {
/* The computation of total_size below assumes that no
delete-mark flags will be stored and that all fields
are NOT NULL and fixed-length. */
sort_buf[i]->total_size += data_size[i];
sort_buf[i]->n_tuples += n_tuple[i];
merge_file[i]->n_rec += n_tuple[i];
t_ctx->rows_added[i] += n_tuple[i];
}
if (!buf_full) {
/* we pad one byte between text accross two fields */
t_ctx->init_pos += doc->text.f_len + 1;
}
return(!buf_full);
}
/*********************************************************************//**
Get next doc item from fts_doc_list */
UNIV_INLINE
void
row_merge_fts_get_next_doc_item(
/*============================*/
fts_psort_t* psort_info, /*!< in: psort_info */
fts_doc_item_t** doc_item) /*!< in/out: doc item */
{
if (*doc_item != NULL) {
ut_free(*doc_item);
}
mutex_enter(&psort_info->mutex);
*doc_item = UT_LIST_GET_FIRST(psort_info->fts_doc_list);
if (*doc_item != NULL) {
UT_LIST_REMOVE(psort_info->fts_doc_list, *doc_item);
ut_ad(psort_info->memory_used >= sizeof(fts_doc_item_t)
+ (*doc_item)->field->len);
psort_info->memory_used -= sizeof(fts_doc_item_t)
+ (*doc_item)->field->len;
}
mutex_exit(&psort_info->mutex);
}
/*********************************************************************//**
Function performs parallel tokenization of the incoming doc strings.
It also performs the initial in memory sort of the parsed records.
@return OS_THREAD_DUMMY_RETURN */
static
os_thread_ret_t
DECLARE_THREAD(fts_parallel_tokenization)(
/*======================*/
void* arg) /*!< in: psort_info for the thread */
{
fts_psort_t* psort_info = (fts_psort_t*) arg;
ulint i;
fts_doc_item_t* doc_item = NULL;
row_merge_buf_t** buf;
ibool processed = FALSE;
merge_file_t** merge_file;
row_merge_block_t** block;
row_merge_block_t** crypt_block;
pfs_os_file_t tmpfd[FTS_NUM_AUX_INDEX];
ulint mycount[FTS_NUM_AUX_INDEX];
ib_uint64_t total_rec = 0;
ulint num_doc_processed = 0;
doc_id_t last_doc_id = 0;
mem_heap_t* blob_heap = NULL;
fts_doc_t doc;
dict_table_t* table = psort_info->psort_common->new_table;
fts_tokenize_ctx_t t_ctx;
ulint retried = 0;
dberr_t error = DB_SUCCESS;
ut_ad(psort_info->psort_common->trx->mysql_thd != NULL);
/* const char* path = thd_innodb_tmpdir(
psort_info->psort_common->trx->mysql_thd);
*/
ut_ad(psort_info->psort_common->trx->mysql_thd != NULL);
const char* path = thd_innodb_tmpdir(
psort_info->psort_common->trx->mysql_thd);
ut_ad(psort_info);
buf = psort_info->merge_buf;
merge_file = psort_info->merge_file;
blob_heap = mem_heap_create(512);
memset(&doc, 0, sizeof(doc));
memset(mycount, 0, FTS_NUM_AUX_INDEX * sizeof(int));
doc.charset = fts_index_get_charset(
psort_info->psort_common->dup->index);
block = psort_info->merge_block;
crypt_block = psort_info->crypt_block;
const page_size_t& page_size = dict_table_page_size(table);
row_merge_fts_get_next_doc_item(psort_info, &doc_item);
t_ctx.cached_stopword = table->fts->cache->stopword_info.cached_stopword;
processed = TRUE;
loop:
while (doc_item) {
dfield_t* dfield = doc_item->field;
last_doc_id = doc_item->doc_id;
ut_ad (dfield->data != NULL
&& dfield_get_len(dfield) != UNIV_SQL_NULL);
/* If finish processing the last item, update "doc" with
strings in the doc_item, otherwise continue processing last
item */
if (processed) {
byte* data;
ulint data_len;
dfield = doc_item->field;
data = static_cast<byte*>(dfield_get_data(dfield));
data_len = dfield_get_len(dfield);
if (dfield_is_ext(dfield)) {
doc.text.f_str =
btr_copy_externally_stored_field(
&doc.text.f_len, data,
page_size, data_len, blob_heap);
} else {
doc.text.f_str = data;
doc.text.f_len = data_len;
}
doc.tokens = 0;
t_ctx.processed_len = 0;
} else {
/* Not yet finish processing the "doc" on hand,
continue processing it */
ut_ad(doc.text.f_str);
ut_ad(t_ctx.processed_len < doc.text.f_len);
}
processed = row_merge_fts_doc_tokenize(
buf, doc_item->doc_id, &doc,
merge_file, psort_info->psort_common->opt_doc_id_size,
&t_ctx);
/* Current sort buffer full, need to recycle */
if (!processed) {
ut_ad(t_ctx.processed_len < doc.text.f_len);
ut_ad(t_ctx.rows_added[t_ctx.buf_used]);
break;
}
num_doc_processed++;
if (fts_enable_diag_print && num_doc_processed % 10000 == 1) {
ib::info() << "Number of documents processed: "
<< num_doc_processed;
#ifdef FTS_INTERNAL_DIAG_PRINT
for (i = 0; i < FTS_NUM_AUX_INDEX; i++) {
ib::info() << "ID " << psort_info->psort_id
<< ", partition " << i << ", word "
<< mycount[i];
}
#endif
}
mem_heap_empty(blob_heap);
row_merge_fts_get_next_doc_item(psort_info, &doc_item);
if (doc_item && last_doc_id != doc_item->doc_id) {
t_ctx.init_pos = 0;
}
}
/* If we run out of current sort buffer, need to sort
and flush the sort buffer to disk */
if (t_ctx.rows_added[t_ctx.buf_used] && !processed) {
row_merge_buf_sort(buf[t_ctx.buf_used], NULL);
row_merge_buf_write(buf[t_ctx.buf_used],
merge_file[t_ctx.buf_used],
block[t_ctx.buf_used]);
if (!row_merge_write(merge_file[t_ctx.buf_used]->fd,
merge_file[t_ctx.buf_used]->offset++,
block[t_ctx.buf_used],
crypt_block[t_ctx.buf_used],
table->space_id)) {
error = DB_TEMP_FILE_WRITE_FAIL;
goto func_exit;
}
UNIV_MEM_INVALID(block[t_ctx.buf_used], srv_sort_buf_size);
buf[t_ctx.buf_used] = row_merge_buf_empty(buf[t_ctx.buf_used]);
mycount[t_ctx.buf_used] += t_ctx.rows_added[t_ctx.buf_used];
t_ctx.rows_added[t_ctx.buf_used] = 0;
ut_a(doc_item);
goto loop;
}
/* Parent done scanning, and if finish processing all the docs, exit */
if (psort_info->state == FTS_PARENT_COMPLETE) {
if (UT_LIST_GET_LEN(psort_info->fts_doc_list) == 0) {
goto exit;
} else if (retried > 10000) {
ut_ad(!doc_item);
/* retried too many times and cannot get new record */
ib::error() << "FTS parallel sort processed "
<< num_doc_processed
<< " records, the sort queue has "
<< UT_LIST_GET_LEN(psort_info->fts_doc_list)
<< " records. But sort cannot get the next"
" records";
goto exit;
}
} else if (psort_info->state == FTS_PARENT_EXITING) {
/* Parent abort */
goto func_exit;
}
if (doc_item == NULL) {
os_thread_yield();
}
row_merge_fts_get_next_doc_item(psort_info, &doc_item);
if (doc_item != NULL) {
if (last_doc_id != doc_item->doc_id) {
t_ctx.init_pos = 0;
}
retried = 0;
} else if (psort_info->state == FTS_PARENT_COMPLETE) {
retried++;
}
goto loop;
exit:
/* Do a final sort of the last (or latest) batch of records
in block memory. Flush them to temp file if records cannot
be hold in one block memory */
for (i = 0; i < FTS_NUM_AUX_INDEX; i++) {
if (t_ctx.rows_added[i]) {
row_merge_buf_sort(buf[i], NULL);
row_merge_buf_write(
buf[i], merge_file[i], block[i]);
/* Write to temp file, only if records have
been flushed to temp file before (offset > 0):
The pseudo code for sort is following:
while (there are rows) {
tokenize rows, put result in block[]
if (block[] runs out) {
sort rows;
write to temp file with
row_merge_write();
offset++;
}
}
# write out the last batch
if (offset > 0) {
row_merge_write();
offset++;
} else {
# no need to write anything
offset stay as 0
}
so if merge_file[i]->offset is 0 when we come to
here as the last batch, this means rows have
never flush to temp file, it can be held all in
memory */
if (merge_file[i]->offset != 0) {
if (!row_merge_write(merge_file[i]->fd,
merge_file[i]->offset++,
block[i],
crypt_block[i],
table->space_id)) {
error = DB_TEMP_FILE_WRITE_FAIL;
goto func_exit;
}
UNIV_MEM_INVALID(block[i], srv_sort_buf_size);
if (crypt_block[i]) {
UNIV_MEM_INVALID(crypt_block[i],