Parent: [55e2fe] (diff)

Child: [b1ff34] (diff)

Download this file

textsplit.cpp    1321 lines (1184 with data), 39.9 kB

   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
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
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
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
/* Copyright (C) 2004 J.F.Dockes
* 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.
*/
#ifndef TEST_TEXTSPLIT
#include "autoconfig.h"
#include <assert.h>
#include <stdlib.h>
#include <iostream>
#include <string>
#include <algorithm>
#include <cstring>
#include <unordered_set>
#include "textsplit.h"
#include "log.h"
//#define UTF8ITER_CHECK
#include "utf8iter.h"
#include "uproplist.h"
#include "smallut.h"
// Decide if we treat katakana as western scripts, splitting into
// words instead of n-grams. This is not absurd (katakana is a kind of
// alphabet, albeit phonetic and syllabic and is mostly used to
// transcribe western words), but it does not work well because
// japanese uses separator-less compound katakana words, and because
// the plural terminaisons are irregular and would need a specialized
// stemmer. So we for now process katakana as the rest of cjk, using
// ngrams
#undef KATAKANA_AS_WORDS
using namespace std;
/**
* Splitting a text into words. The code in this file works with utf-8
* in a semi-clean way (see uproplist.h). Ascii still gets special
* treatment in the sense that many special characters can only be
* ascii (e.g. @, _,...). However, this compromise works quite well
* while being much more light-weight than a full-blown Unicode
* approach (ICU...)
*/
// Ascii character classes: we have three main groups, and then some chars
// are their own class because they want special handling.
//
// We have an array with 256 slots where we keep the character types.
// The array could be fully static, but we use a small function to fill it
// once.
// The array is actually a remnant of the original version which did no utf8.
// Only the lower 127 slots are now used, but keep it at 256
// because it makes some tests in the code simpler.
const unsigned int charclasses_size = 256;
enum CharClass {LETTER=256, SPACE=257, DIGIT=258, WILD=259,
A_ULETTER=260, A_LLETTER=261, SKIP=262};
static int charclasses[charclasses_size];
// Non-ascii UTF-8 characters are handled with sets holding all
// characters with interesting properties. This is far from full-blown
// management of Unicode properties, but seems to do the job well
// enough in most common cases
static vector<unsigned int> vpuncblocks;
static std::unordered_set<unsigned int> spunc;
static std::unordered_set<unsigned int> visiblewhite;
static std::unordered_set<unsigned int> sskip;
class CharClassInit {
public:
CharClassInit()
{
unsigned int i;
// Set default value for all: SPACE
for (i = 0 ; i < 256 ; i ++)
charclasses[i] = SPACE;
char digits[] = "0123456789";
for (i = 0; i < strlen(digits); i++)
charclasses[int(digits[i])] = DIGIT;
char upper[] = "ABCDEFGHIJKLMNOPQRSTUVWXYZ";
for (i = 0; i < strlen(upper); i++)
charclasses[int(upper[i])] = A_ULETTER;
char lower[] = "abcdefghijklmnopqrstuvwxyz";
for (i = 0; i < strlen(lower); i++)
charclasses[int(lower[i])] = A_LLETTER;
char wild[] = "*?[]";
for (i = 0; i < strlen(wild); i++)
charclasses[int(wild[i])] = WILD;
// Characters with special treatment:
//
// The first ones are mostly span-constructing "glue"
// characters, for example those typically allowing us to
// search for an email address as a whole (bob@isp.org instead
// of as a phrase "bob isp org"
//
// The case of the minus sign is a complicated one. It went
// from glue to non-glue to glue along Recoll versions.
// See minus-hyphen-dash.txt in doc/notes
char special[] = ".@+-#'_\n\r\f";
for (i = 0; i < strlen(special); i++)
charclasses[int(special[i])] = special[i];
for (i = 0; i < sizeof(unipunc) / sizeof(int); i++) {
spunc.insert(unipunc[i]);
}
spunc.insert((unsigned int)-1);
for (i = 0; i < sizeof(unipuncblocks) / sizeof(int); i++) {
vpuncblocks.push_back(unipuncblocks[i]);
}
assert((vpuncblocks.size() % 2) == 0);
for (i = 0; i < sizeof(avsbwht) / sizeof(int); i++) {
visiblewhite.insert(avsbwht[i]);
}
for (i = 0; i < sizeof(uniskip) / sizeof(int); i++) {
sskip.insert(uniskip[i]);
}
}
};
static const CharClassInit charClassInitInstance;
void TextSplit::backslashAsLetter(bool on) {
if (on) {
charclasses[int('\\')] = A_LLETTER;
} else {
charclasses[int('\\')] = SPACE;
}
}
static inline int whatcc(unsigned int c)
{
if (c <= 127) {
return charclasses[c];
} else {
if (c == 0x2010) {
// Special treatment for hyphen: handle as ascii minus. See
// doc/notes/minus-hyphen-dash.txt
return 0x2010;
} else if (sskip.find(c) != sskip.end()) {
return SKIP;
} else if (spunc.find(c) != spunc.end()) {
return SPACE;
} else {
vector<unsigned int>::iterator it =
lower_bound(vpuncblocks.begin(), vpuncblocks.end(), c);
if (it == vpuncblocks.end())
return LETTER;
if (c == *it)
return SPACE;
if ((it - vpuncblocks.begin()) % 2 == 1) {
return SPACE;
} else {
return LETTER;
}
}
}
}
// testing whatcc...
#if 0
unsigned int testvalues[] = {'a', '0', 0x80, 0xbf, 0xc0, 0x05c3, 0x1000,
0x2000, 0x2001, 0x206e, 0x206f, 0x20d0, 0x2399,
0x2400, 0x2401, 0x243f, 0x2440, 0xff65};
int ntest = sizeof(testvalues) / sizeof(int);
for (int i = 0; i < ntest; i++) {
int ret = whatcc(testvalues[i]);
printf("Tested value 0x%x, returned value %d %s\n",
testvalues[i], ret, ret == LETTER ? "LETTER" :
ret == SPACE ? "SPACE" : "OTHER");
}
#endif
// CJK Unicode character detection:
//
// 2E80..2EFF; CJK Radicals Supplement
// 3000..303F; CJK Symbols and Punctuation
// 3040..309F; Hiragana
// 30A0..30FF; Katakana
// 3100..312F; Bopomofo
// 3130..318F; Hangul Compatibility Jamo
// 3190..319F; Kanbun
// 31A0..31BF; Bopomofo Extended
// 31C0..31EF; CJK Strokes
// 31F0..31FF; Katakana Phonetic Extensions
// 3200..32FF; Enclosed CJK Letters and Months
// 3300..33FF; CJK Compatibility
// 3400..4DBF; CJK Unified Ideographs Extension A
// 4DC0..4DFF; Yijing Hexagram Symbols
// 4E00..9FFF; CJK Unified Ideographs
// A700..A71F; Modifier Tone Letters
// AC00..D7AF; Hangul Syllables
// F900..FAFF; CJK Compatibility Ideographs
// FE30..FE4F; CJK Compatibility Forms
// FF00..FFEF; Halfwidth and Fullwidth Forms
// 20000..2A6DF; CJK Unified Ideographs Extension B
// 2F800..2FA1F; CJK Compatibility Ideographs Supplement
#define UNICODE_IS_CJK(p) \
(((p) >= 0x2E80 && (p) <= 0x2EFF) || \
((p) >= 0x3000 && (p) <= 0x9FFF) || \
((p) >= 0xA700 && (p) <= 0xA71F) || \
((p) >= 0xAC00 && (p) <= 0xD7AF) || \
((p) >= 0xF900 && (p) <= 0xFAFF) || \
((p) >= 0xFE30 && (p) <= 0xFE4F) || \
((p) >= 0xFF00 && (p) <= 0xFFEF) || \
((p) >= 0x20000 && (p) <= 0x2A6DF) || \
((p) >= 0x2F800 && (p) <= 0x2FA1F))
// We should probably map 'fullwidth ascii variants' and 'halfwidth
// katakana variants' to something else. Look up "Kuromoji" Lucene
// filter, KuromojiNormalizeFilter.java
// 309F is Hiragana.
#ifdef KATAKANA_AS_WORDS
#define UNICODE_IS_KATAKANA(p) \
((p) != 0x309F && \
(((p) >= 0x3099 && (p) <= 0x30FF) || \
((p) >= 0x31F0 && (p) <= 0x31FF)))
#else
#define UNICODE_IS_KATAKANA(p) false
#endif
bool TextSplit::isCJK(int c)
{
return UNICODE_IS_CJK(c) && !UNICODE_IS_KATAKANA(c);
}
bool TextSplit::isKATAKANA(int c)
{
return UNICODE_IS_KATAKANA(c);
}
// This is used to detect katakana/other transitions, which must
// trigger a word split (there is not always a separator, and katakana
// is otherwise treated like other, in the same routine, unless cjk
// which has its span reader causing a word break)
enum CharSpanClass {CSC_CJK, CSC_KATAKANA, CSC_OTHER};
bool TextSplit::o_processCJK = true;
unsigned int TextSplit::o_CJKNgramLen = 2;
bool TextSplit::o_noNumbers = false;
bool TextSplit::o_deHyphenate = false;
// Final term checkpoint: do some checking (the kind which is simpler
// to do here than in the main loop), then send term to our client.
inline bool TextSplit::emitterm(bool isspan, string &w, int pos,
size_t btstart, size_t btend)
{
LOGDEB2("TextSplit::emitterm: [" << w << "] pos " << pos << "\n");
int l = int(w.length());
#ifdef TEXTSPLIT_STATS
// Update word length statistics. Do this before we filter out
// long words because stats are used to detect bad text
if (!isspan || m_wordLen == m_span.length())
m_stats.newsamp(m_wordChars);
#endif
if (l > 0 && l < m_maxWordLength) {
// 1 byte word: we index single ascii letters and digits, but
// nothing else. We might want to turn this into a test for a
// single utf8 character instead ?
if (l == 1) {
unsigned int c = ((unsigned int)w[0]) & 0xff;
if (charclasses[c] != A_ULETTER && charclasses[c] != A_LLETTER &&
charclasses[c] != DIGIT &&
(!(m_flags & TXTS_KEEPWILD) || charclasses[c] != WILD)
) {
//cerr << "ERASING single letter term " << c << endl;
return true;
}
}
if (pos != m_prevpos || l != m_prevlen) {
bool ret = takeword(w, pos, int(btstart), int(btend));
m_prevpos = pos;
m_prevlen = int(w.length());
return ret;
}
LOGDEB2("TextSplit::emitterm:dup: [" << w << "] pos " << pos << "\n");
}
return true;
}
// Check for an acronym/abbreviation ie I.B.M. This only works with
// ascii (no non-ascii utf-8 acronym are possible)
bool TextSplit::span_is_acronym(string *acronym)
{
bool acron = false;
if (m_wordLen != m_span.length() &&
m_span.length() > 2 && m_span.length() <= 20) {
acron = true;
// Check odd chars are '.'
for (unsigned int i = 1 ; i < m_span.length(); i += 2) {
if (m_span[i] != '.') {
acron = false;
break;
}
}
if (acron) {
// Check that even chars are letters
for (unsigned int i = 0 ; i < m_span.length(); i += 2) {
int c = m_span[i];
if (!((c >= 'a' && c <= 'z')||(c >= 'A' && c <= 'Z'))) {
acron = false;
break;
}
}
}
}
if (acron) {
for (unsigned int i = 0; i < m_span.length(); i += 2) {
*acronym += m_span[i];
}
}
return acron;
}
// Generate terms from span. Have to take into account the
// flags: ONLYSPANS, NOSPANS, noNumbers
bool TextSplit::words_from_span(size_t bp)
{
#if 0
cerr << "Span: [" << m_span << "] " << " w_i_s size: " <<
m_words_in_span.size() << " : ";
for (unsigned int i = 0; i < m_words_in_span.size(); i++) {
cerr << " [" << m_words_in_span[i].first << " " <<
m_words_in_span[i].second << "] ";
}
cerr << endl;
#endif
int spanwords = int(m_words_in_span.size());
// It seems that something like: tv_combo-sample_util.Po@am_quote
// can get the splitter to call doemit with a span of '@' and
// words_in_span==0, which then causes a crash when accessing
// words_in_span[0] if the stl assertions are active (e.g. Fedora
// RPM build). Not too sure what the right fix would be, but for
// now, just defend against it
if (spanwords == 0) {
return true;
}
int pos = m_spanpos;
// Byte position of the span start
size_t spboffs = bp - m_span.size();
if (o_deHyphenate && spanwords == 2 &&
m_span[m_words_in_span[0].second] == '-') {
unsigned int s0 = m_words_in_span[0].first;
unsigned int l0 = m_words_in_span[0].second - m_words_in_span[0].first;
unsigned int s1 = m_words_in_span[1].first;
unsigned int l1 = m_words_in_span[1].second - m_words_in_span[1].first;
string word = m_span.substr(s0, l0) + m_span.substr(s1, l1);
if (l0 && l1)
emitterm(false, word,
m_spanpos, spboffs, spboffs + m_words_in_span[1].second);
}
for (int i = 0;
i < ((m_flags&TXTS_ONLYSPANS) ? 1 : spanwords);
i++) {
int deb = m_words_in_span[i].first;
bool noposinc = m_words_in_span[i].second == deb;
for (int j = ((m_flags&TXTS_ONLYSPANS) ? spanwords-1 : i);
j < ((m_flags&TXTS_NOSPANS) ? i+1 : spanwords);
j++) {
int fin = m_words_in_span[j].second;
//cerr << "i " << i << " j " << j << " deb " << deb <<
//" fin " << fin << endl;
if (fin - deb > int(m_span.size()))
break;
string word(m_span.substr(deb, fin-deb));
if (!emitterm(j != i+1, word, pos, spboffs+deb, spboffs+fin))
return false;
}
if (!noposinc)
++pos;
}
return true;
}
/**
* A method called at word boundaries (different places in
* text_to_words()), to adjust the current state of the parser, and
* possibly generate term(s). While inside a span (words linked by
* glue characters), we just keep track of the word boundaries. Once
* actual white-space is reached, we get called with spanerase set to
* true, and we process the span, calling the emitterm() routine for
* each generated term.
*
* The object flags can modify our behaviour, deciding if we only emit
* single words (bill, recoll, org), only spans (bill@recoll.org), or
* words and spans (bill@recoll.org, recoll.org, jf, recoll...)
*
* @return true if ok, false for error. Splitting should stop in this case.
* @param spanerase Set if the current span is at its end. Process it.
* @param bp The current BYTE position in the stream
*/
inline bool TextSplit::doemit(bool spanerase, size_t _bp)
{
int bp = int(_bp);
LOGDEB2("TextSplit::doemit: sper " << spanerase << " bp " << bp <<
" spp " << m_spanpos << " spanwords " << m_words_in_span.size() <<
" wS " << m_wordStart << " wL " << m_wordLen << " inn " <<
m_inNumber << " span [" << m_span << "]\n");
if (m_wordLen) {
// We have a current word. Remember it
// Limit max span word count
if (m_words_in_span.size() >= 6) {
spanerase = true;
}
m_words_in_span.push_back(pair<int,int>(m_wordStart,
m_wordStart + m_wordLen));
m_wordpos++;
m_wordLen = m_wordChars = 0;
}
if (spanerase) {
// We encountered a span-terminating character. Produce terms.
string acronym;
if (span_is_acronym(&acronym)) {
if (!emitterm(false, acronym, m_spanpos, bp - m_span.length(), bp))
return false;
}
// Maybe trim at end. These are chars that we might keep
// inside a span, but not at the end.
while (m_span.length() > 0) {
switch (*(m_span.rbegin())) {
case '.':
case '-':
case ',':
case '@':
case '_':
case '\'':
m_span.resize(m_span.length()-1);
if (m_words_in_span.size() &&
m_words_in_span.back().second > int(m_span.size()))
m_words_in_span.back().second = int(m_span.size());
if (--bp < 0)
bp = 0;
break;
default:
goto breaktrimloop;
}
}
breaktrimloop:
if (!words_from_span(bp)) {
return false;
}
discardspan();
} else {
m_wordStart = int(m_span.length());
}
return true;
}
void TextSplit::discardspan()
{
m_span.clear();
m_words_in_span.clear();
m_spanpos = m_wordpos;
m_wordStart = 0;
m_wordLen = m_wordChars = 0;
}
static inline bool isalphanum(int what, unsigned int flgs)
{
return what == A_LLETTER || what == A_ULETTER ||
what == DIGIT || what == LETTER ||
((flgs & TextSplit::TXTS_KEEPWILD) && what == WILD);
}
static inline bool isdigit(int what, unsigned int flgs)
{
return what == DIGIT || ((flgs & TextSplit::TXTS_KEEPWILD) && what == WILD);
}
#ifdef TEXTSPLIT_STATS
#define STATS_INC_WORDCHARS ++m_wordChars
#else
#define STATS_INC_WORDCHARS
#endif
vector<CharFlags> splitFlags{
{TextSplit::TXTS_NOSPANS, "nospans"},
{TextSplit::TXTS_ONLYSPANS, "onlyspans"},
{TextSplit::TXTS_KEEPWILD, "keepwild"}
};
/**
* Splitting a text into terms to be indexed.
* We basically emit a word every time we see a separator, but some chars are
* handled specially so that special cases, ie, c++ and jfd@recoll.com etc,
* are handled properly,
*/
bool TextSplit::text_to_words(const string &in)
{
LOGDEB1("TextSplit::text_to_words: docjk " << o_processCJK << "(" <<
o_CJKNgramLen << ") " << flagsToString(splitFlags, m_flags) <<
" [" << in.substr(0,50) << "]\n");
if (in.empty())
return true;
// Reset the data members relative to splitting state
clearsplitstate();
bool pagepending = false;
bool softhyphenpending = false;
// Running count of non-alphanum chars. Reset when we see one;
int nonalnumcnt = 0;
Utf8Iter it(in);
int prev_csc = -1;
for (; !it.eof(); it++) {
unsigned int c = *it;
nonalnumcnt++;
if (c == (unsigned int)-1) {
LOGERR("Textsplit: error occurred while scanning UTF-8 string\n");
return false;
}
CharSpanClass csc;
if (UNICODE_IS_KATAKANA(c)) {
csc = CSC_KATAKANA;
} else if (UNICODE_IS_CJK(c)) {
csc = CSC_CJK;
} else {
csc = CSC_OTHER;
}
if (o_processCJK && csc == CSC_CJK) {
// CJK excluding Katakana character hit.
// Do like at EOF with the current non-cjk data.
if (m_wordLen || m_span.length()) {
if (!doemit(true, it.getBpos()))
return false;
}
// Hand off situation to the cjk routine.
if (!cjk_to_words(&it, &c)) {
LOGERR("Textsplit: scan error in cjk handler\n");
return false;
}
// Check for eof, else c contains the first non-cjk
// character after the cjk sequence, just go on.
if (it.eof())
break;
}
#ifdef KATAKANA_AS_WORDS
// Only needed if we have script transitions inside this
// routine, else the call to cjk_to_words does the job.
if (csc != prev_csc && (m_wordLen || m_span.length())) {
LOGDEB("csc " << csc << " pcsc " << prev_csc << " wl " <<
m_wordLen << " spl " << m_span.length() << endl);
if (!doemit(true, it.getBpos())) {
return false;
}
}
#endif
prev_csc = csc;
int cc = whatcc(c);
switch (cc) {
case SKIP:
// Special-case soft-hyphen. To work, this depends on the
// fact that only SKIP calls "continue" inside the
// switch. All the others will do the softhyphenpending
// reset after the switch
if (c == 0xad) {
softhyphenpending = true;
} else {
softhyphenpending = false;
}
// Skips the softhyphenpending reset
continue;
case DIGIT:
nonalnumcnt = 0;
if (m_wordLen == 0)
m_inNumber = true;
m_wordLen += it.appendchartostring(m_span);
STATS_INC_WORDCHARS;
break;
case SPACE:
nonalnumcnt = 0;
SPACE:
if (m_wordLen || m_span.length()) {
if (!doemit(true, it.getBpos()))
return false;
m_inNumber = false;
}
if (pagepending) {
pagepending = false;
newpage(m_wordpos);
}
break;
case WILD:
if (m_flags & TXTS_KEEPWILD)
goto NORMALCHAR;
else
goto SPACE;
break;
case '-':
case '+':
if (m_wordLen == 0) {
// + or - don't start a term except if this looks like
// it's going to be to be a number
if (isdigit(whatcc(it[it.getCpos()+1]), m_flags)) {
// -10
m_inNumber = true;
m_wordLen += it.appendchartostring(m_span);
STATS_INC_WORDCHARS;
break;
}
} else if (m_inNumber) {
if ((m_span[m_span.length() - 1] == 'e' ||
m_span[m_span.length() - 1] == 'E')) {
if (isdigit(whatcc(it[it.getCpos()+1]), m_flags)) {
m_wordLen += it.appendchartostring(m_span);
STATS_INC_WORDCHARS;
break;
}
}
} else {
if (cc == '+') {
int nextc = it[it.getCpos()+1];
if (nextc == '+' || nextc == -1 || visiblewhite.find(nextc)
!= visiblewhite.end()) {
// someword++[+...] !
m_wordLen += it.appendchartostring(m_span);
STATS_INC_WORDCHARS;
break;
}
} else {
// Treat '-' inside span as glue char
if (!doemit(false, it.getBpos()))
return false;
m_inNumber = false;
m_wordStart += it.appendchartostring(m_span);
break;
}
}
goto SPACE;
break;
case 0x2010:
// Hyphen is replaced with ascii minus
if (m_wordLen != 0) {
// Treat '-' inside span as glue char
if (!doemit(false, it.getBpos()))
return false;
m_inNumber = false;
m_span += '-';
m_wordStart++;
break;
}
goto SPACE;
case '.':
{
// Need a little lookahead here. At worse this gets the end null
int nextc = it[it.getCpos()+1];
int nextwhat = whatcc(nextc);
if (m_inNumber) {
if (!isdigit(nextwhat, m_flags))
goto SPACE;
m_wordLen += it.appendchartostring(m_span);
STATS_INC_WORDCHARS;
break;
} else {
// Found '.' while not in number
// Only letters and digits make sense after
if (!isalphanum(nextwhat, m_flags))
goto SPACE;
// Keep an initial '.' for catching .net, and .34 (aka
// 0.34) but this adds quite a few spurious terms !
if (m_span.length() == 0) {
// Check for number like .1
if (isdigit(nextwhat, m_flags)) {
m_inNumber = true;
m_wordLen += it.appendchartostring(m_span);
} else {
m_words_in_span.
push_back(pair<int,int>(m_wordStart, m_wordStart));
m_wordStart += it.appendchartostring(m_span);
}
STATS_INC_WORDCHARS;
break;
}
// '.' between words: span glue
if (m_wordLen) {
if (!doemit(false, it.getBpos()))
return false;
m_wordStart += it.appendchartostring(m_span);
}
}
}
break;
case '@':
case '_':
case '\'':
// If in word, potential span: o'brien, jf@dockes.org,
// else just ignore
if (m_wordLen) {
if (!doemit(false, it.getBpos()))
return false;
m_inNumber = false;
m_wordStart += it.appendchartostring(m_span);
}
break;
case '#': {
int w = whatcc(it[it.getCpos()+1]);
// Keep it only at the beginning of a word (hashtag),
if (m_wordLen == 0 && isalphanum(w, m_flags)) {
m_wordLen += it.appendchartostring(m_span);
STATS_INC_WORDCHARS;
break;
}
// or at the end (special case for c# ...)
if (m_wordLen > 0) {
if (w == SPACE || w == '\n' || w == '\r') {
m_wordLen += it.appendchartostring(m_span);
STATS_INC_WORDCHARS;
break;
}
}
goto SPACE;
}
break;
case '\n':
case '\r':
if (m_span.length() && *m_span.rbegin() == '-') {
// if '-' is the last char before end of line, we
// strip it. We have no way to know if this is added
// because of the line split or if it was part of an
// actual compound word (would need a dictionary to
// check). As soft-hyphen *should* be used if the '-'
// is not part of the text, it is better to properly
// process a real compound word, and produce wrong
// output from wrong text. The word-emitting routine
// will strip the trailing '-'.
goto SPACE;
} else if (softhyphenpending) {
// Don't reset soft-hyphen
continue;
} else {
// Normal case: EOL is white space
goto SPACE;
}
break;
case '\f':
pagepending = true;
goto SPACE;
break;
#ifdef RCL_SPLIT_CAMELCASE
// Camelcase handling.
// If we get uppercase ascii after lowercase ascii, emit word.
// This emits "camel" when hitting the 'C' of camelCase
// Not enabled by defaults as this makes phrase searches quite
// confusing.
// ie "MySQL manual" is matched by "MySQL manual" and
// "my sql manual" but not "mysql manual"
// A possibility would be to emit both my and sql at the
// same position. All non-phrase searches would work, and
// both "MySQL manual" and "mysql manual" phrases would
// match too. "my sql manual" would not match, but this is
// not an issue.
case A_ULETTER:
if (m_span.length() &&
charclasses[(unsigned char)m_span[m_span.length() - 1]] ==
A_LLETTER) {
if (m_wordLen) {
if (!doemit(false, it.getBpos()))
return false;
}
}
goto NORMALCHAR;
// CamelCase handling.
// If we get lowercase after uppercase and the current
// word length is bigger than one, it means we had a
// string of several upper-case letters: an
// acronym (readHTML) or a single letter article (ALittleHelp).
// Emit the uppercase word before proceeding
case A_LLETTER:
if (m_span.length() &&
charclasses[(unsigned char)m_span[m_span.length() - 1]] ==
A_ULETTER && m_wordLen > 1) {
// Multiple upper-case letters. Single letter word
// or acronym which we want to emit now
m_wordLen--;
if (!doemit(false, it.getBpos()))
return false;
// m_wordstart could be 0 here if the span was reset
// for excessive length
if (m_wordStart)
m_wordStart--;
m_wordLen++;
}
goto NORMALCHAR;
#endif /* CAMELCASE */
default:
NORMALCHAR:
nonalnumcnt = 0;
if (m_inNumber && c != 'e' && c != 'E') {
m_inNumber = false;
}
m_wordLen += it.appendchartostring(m_span);
STATS_INC_WORDCHARS;
break;
}
softhyphenpending = false;
}
if (m_wordLen || m_span.length()) {
if (!doemit(true, it.getBpos()))
return false;
}
return true;
}
// Using an utf8iter pointer just to avoid needing its definition in
// textsplit.h
//
// We output ngrams for exemple for char input a b c and ngramlen== 2,
// we generate: a ab b bc c as words
//
// This is very different from the normal behaviour, so we don't use
// the doemit() and emitterm() routines
//
// The routine is sort of a mess and goes to show that we'd probably
// be better off converting the whole buffer to utf32 on entry...
bool TextSplit::cjk_to_words(Utf8Iter *itp, unsigned int *cp)
{
LOGDEB1("cjk_to_words: m_wordpos " << m_wordpos << "\n");
Utf8Iter &it = *itp;
// We use an offset buffer to remember the starts of the utf-8
// characters which we still need to use.
assert(o_CJKNgramLen < o_CJKMaxNgramLen);
unsigned int boffs[o_CJKMaxNgramLen+1];
// Current number of valid offsets;
unsigned int nchars = 0;
unsigned int c = 0;
for (; !it.eof(); it++) {
c = *it;
if (!UNICODE_IS_CJK(c)) {
// Return to normal handler
break;
}
if (whatcc(c) == SPACE) {
// Flush the ngram buffer and go on
nchars = 0;
continue;
}
if (nchars == o_CJKNgramLen) {
// Offset buffer full, shift it. Might be more efficient
// to have a circular one, but things are complicated
// enough already...
for (unsigned int i = 0; i < nchars-1; i++) {
boffs[i] = boffs[i+1];
}
} else {
nchars++;
}
// Take note of byte offset for this character.
boffs[nchars-1] = int(it.getBpos());
// Output all new ngrams: they begin at each existing position
// and end after the new character. onlyspans->only output
// maximum words, nospans=> single chars
if (!(m_flags & TXTS_ONLYSPANS) || nchars == o_CJKNgramLen) {
int btend = int(it.getBpos() + it.getBlen());
int loopbeg = (m_flags & TXTS_NOSPANS) ? nchars-1 : 0;
int loopend = (m_flags & TXTS_ONLYSPANS) ? 1 : nchars;
for (int i = loopbeg; i < loopend; i++) {
if (!takeword(it.buffer().substr(boffs[i],
btend-boffs[i]),
m_wordpos - (nchars-i-1), boffs[i], btend)) {
return false;
}
}
if ((m_flags & TXTS_ONLYSPANS)) {
// Only spans: don't overlap: flush buffer
nchars = 0;
}
}
// Increase word position by one, other words are at an
// existing position. This could be subject to discussion...
m_wordpos++;
}
// If onlyspans is set, there may be things to flush in the buffer
// first
if ((m_flags & TXTS_ONLYSPANS) && nchars > 0 && nchars != o_CJKNgramLen) {
int btend = int(it.getBpos()); // Current char is out
if (!takeword(it.buffer().substr(boffs[0], btend-boffs[0]),
m_wordpos - nchars,
boffs[0], btend)) {
return false;
}
}
// Reset state, saving term position, and return the found non-cjk
// unicode character value. The current input byte offset is kept
// in the utf8Iter
int pos = m_wordpos;
clearsplitstate();
m_spanpos = m_wordpos = pos;
*cp = c;
return true;
}
// Specialization for countWords
class TextSplitCW : public TextSplit {
public:
int wcnt;
TextSplitCW(Flags flags) : TextSplit(flags), wcnt(0) {}
bool takeword(const string &, int, int, int) {
wcnt++;
return true;
}
};
int TextSplit::countWords(const string& s, TextSplit::Flags flgs)
{
TextSplitCW splitter(flgs);
splitter.text_to_words(s);
return splitter.wcnt;
}
bool TextSplit::hasVisibleWhite(const string &in)
{
Utf8Iter it(in);
for (; !it.eof(); it++) {
unsigned int c = (unsigned char)*it;
if (c == (unsigned int)-1) {
LOGERR("hasVisibleWhite: error while scanning UTF-8 string\n");
return false;
}
if (visiblewhite.find(c) != visiblewhite.end())
return true;
}
return false;
}
template <class T> bool u8stringToStrings(const string &s, T &tokens)
{
Utf8Iter it(s);
string current;
tokens.clear();
enum states {SPACE, TOKEN, INQUOTE, ESCAPE};
states state = SPACE;
for (; !it.eof(); it++) {
unsigned int c = *it;
if (visiblewhite.find(c) != visiblewhite.end())
c = ' ';
if (c == (unsigned int)-1) {
LOGERR("TextSplit::stringToStrings: error while scanning UTF-8 "
"string\n");
return false;
}
switch (c) {
case '"':
switch(state) {
case SPACE: state = INQUOTE; continue;
case TOKEN: goto push_char;
case ESCAPE: state = INQUOTE; goto push_char;
case INQUOTE: tokens.push_back(current);current.clear();
state = SPACE; continue;
}
break;
case '\\':
switch(state) {
case SPACE:
case TOKEN: state=TOKEN; goto push_char;
case INQUOTE: state = ESCAPE; continue;
case ESCAPE: state = INQUOTE; goto push_char;
}
break;
case ' ':
case '\t':
case '\n':
case '\r':
switch(state) {
case SPACE: continue;
case TOKEN: tokens.push_back(current); current.clear();
state = SPACE; continue;
case INQUOTE:
case ESCAPE: goto push_char;
}
break;
default:
switch(state) {
case ESCAPE: state = INQUOTE; break;
case SPACE: state = TOKEN; break;
case TOKEN:
case INQUOTE: break;
}
push_char:
it.appendchartostring(current);
}
}
// End of string. Process residue, and possible error (unfinished quote)
switch(state) {
case SPACE: break;
case TOKEN: tokens.push_back(current); break;
case INQUOTE:
case ESCAPE: return false;
}
return true;
}
bool TextSplit::stringToStrings(const string &s, vector<string> &tokens)
{
return u8stringToStrings<vector<string> >(s, tokens);
}
#else // TEST driver ->
#include <stdio.h>
#include <stdlib.h>
#include <unistd.h>
#include <errno.h>
#include <fcntl.h>
#include <string.h>
#include <math.h>
#include <iostream>
#include "textsplit.h"
#include "readfile.h"
#include "log.h"
#include "transcode.h"
#include "unacpp.h"
#include "termproc.h"
using namespace std;
class myTermProc : public Rcl::TermProc {
int first;
bool nooutput;
public:
myTermProc() : TermProc(0), first(1), nooutput(false) {}
void setNoOut(bool val) {nooutput = val;}
virtual bool takeword(const string &term, int pos, int bs, int be)
{
if (nooutput)
return true;
FILE *fp = stdout;
if (first) {
fprintf(fp, "%3s %-20s %4s %4s\n", "pos", "Term", "bs", "be");
first = 0;
}
fprintf(fp, "%3d %-20s %4d %4d\n", pos, term.c_str(), bs, be);
return true;
}
};
#define OPT_s 0x1
#define OPT_w 0x2
#define OPT_q 0x4
#define OPT_c 0x8
#define OPT_k 0x10
#define OPT_C 0x20
#define OPT_n 0x40
#define OPT_S 0x80
#define OPT_u 0x100
#define OPT_p 0x200
bool dosplit(const string& data, TextSplit::Flags flags, int op_flags)
{
myTermProc printproc;
Rcl::TermProc *nxt = &printproc;
// Rcl::TermProcCommongrams commonproc(nxt, stoplist);
// if (op_flags & OPT_S)
// nxt = &commonproc;
Rcl::TermProcPrep preproc(nxt);
if (op_flags & OPT_u)
nxt = &preproc;
Rcl::TextSplitP splitter(nxt, flags);
if (op_flags & OPT_q)
printproc.setNoOut(true);
splitter.text_to_words(data);
#ifdef TEXTSPLIT_STATS
TextSplit::Stats::Values v = splitter.getStats();
cout << "Average length: "
<< v.avglen
<< " Standard deviation: "
<< v.sigma
<< " Coef of variation "
<< v.sigma / v.avglen
<< endl;
#endif
return true;
}
static const char *teststrings[] = {
"Un bout de texte \nnormal. 2eme phrase.3eme;quatrieme.\n",
"\"Jean-Francois Dockes\" <jfd@okyz.com>\n",
"n@d @net .net net@ t@v@c c# c++ o'brien 'o'brien'",
"_network_ some_span",
"data123\n",
"134 +134 -14 0.1 .1 2. -1.5 +1.5 1,2 1.54e10 1,2e30 .1e10 1.e-8\n",
"@^#$(#$(*)\n",
"192.168.4.1 one\n\rtwo\r",
"[olala][ululu] (valeur) (23)\n",
"utf-8 ucs-4�� \\nodef\n",
"A b C 2 . +",
"','this\n",
" ,able,test-domain",
" -wl,--export-dynamic",
" ~/.xsession-errors",
"this_very_long_span_this_very_long_span_this_very_long_span",
"soft\xc2\xadhyphen",
"soft\xc2\xad\nhyphen",
"soft\xc2\xad\n\rhyphen",
"real\xe2\x80\x90hyphen",
"real\xe2\x80\x90\nhyphen",
"hyphen-\nminus",
};
const int teststrings_cnt = sizeof(teststrings)/sizeof(char *);
static string teststring1 = " nouvel-an ";
static string thisprog;
static string usage =
" textsplit [opts] [filename]\n"
" -q : no output\n"
" -s : only spans\n"
" -w : only words\n"
" -n : no numbers\n"
" -k : preserve wildcards (?*)\n"
" -c : just count words\n"
" -u : use unac\n"
" -C [charset] : input charset\n"
" -S [stopfile] : stopfile to use for commongrams\n"
" if filename is 'stdin', will read stdin for data (end with ^D)\n\n"
" textplit -p somephrase : display results from stringToStrings()\n"
" \n"
;
static void
Usage(void)
{
cerr << thisprog << ": usage:\n" << usage;
exit(1);
}
static int op_flags;
int main(int argc, char **argv)
{
string charset, stopfile;
thisprog = argv[0];
argc--; argv++;
while (argc > 0 && **argv == '-') {
(*argv)++;
if (!(**argv))
/* Cas du "adb - core" */
Usage();
while (**argv)
switch (*(*argv)++) {
case 'c': op_flags |= OPT_c; break;
case 'C': op_flags |= OPT_C; if (argc < 2) Usage();
charset = *(++argv); argc--;
goto b1;
case 'k': op_flags |= OPT_k; break;
case 'n': op_flags |= OPT_n; break;
case 'p': op_flags |= OPT_p; break;
case 'q': op_flags |= OPT_q; break;
case 's': op_flags |= OPT_s; break;
case 'S': op_flags |= OPT_S; if (argc < 2) Usage();
stopfile = *(++argv); argc--;
goto b1;
case 'u': op_flags |= OPT_u; break;
case 'w': op_flags |= OPT_w; break;
default: Usage(); break;
}
b1: argc--; argv++;
}
TextSplit::Flags flags = TextSplit::TXTS_NONE;
if (op_flags&OPT_s)
flags = TextSplit::TXTS_ONLYSPANS;
else if (op_flags&OPT_w)
flags = TextSplit::TXTS_NOSPANS;
if (op_flags & OPT_k)
flags = (TextSplit::Flags)(flags | TextSplit::TXTS_KEEPWILD);
if (op_flags & OPT_n)
TextSplit::noNumbers();
Rcl::StopList stoplist;
if (op_flags & OPT_S) {
if (!stoplist.setFile(stopfile)) {
cerr << "Can't read stopfile: " << stopfile << endl;
exit(1);
}
}
string odata, reason;
if (argc == 1) {
const char *filename = *argv++; argc--;
if (op_flags& OPT_p) {
vector<string> tokens;
TextSplit::stringToStrings(filename, tokens);
for (vector<string>::const_iterator it = tokens.begin();
it != tokens.end(); it++) {
cout << "[" << *it << "] ";
}
cout << endl;
exit(0);
}
if (!strcmp(filename, "stdin")) {
char buf[1024];
int nread;
while ((nread = read(0, buf, 1024)) > 0) {
odata.append(buf, nread);
}
} else if (!file_to_string(filename, odata, &reason)) {
cerr << "Failed: file_to_string(" << filename << ") failed: "
<< reason << endl;
exit(1);
}
} else {
if (op_flags & OPT_p)
Usage();
for (int i = 0; i < teststrings_cnt; i++) {
cout << endl << teststrings[i] << endl;
dosplit(teststrings[i], flags, op_flags);
}
exit(0);
}
string& data = odata;
string ndata;
if ((op_flags & OPT_C)) {
if (!transcode(odata, ndata, charset, "UTF-8")) {
cerr << "Failed: transcode error" << endl;
exit(1);
} else {
data = ndata;
}
}
if (op_flags & OPT_c) {
int n = TextSplit::countWords(data, flags);
cout << n << " words" << endl;
} else {
dosplit(data, flags, op_flags);
}
}
#endif // TEST