summaryrefslogtreecommitdiff
path: root/src/search.cc
blob: 730a1ac11b31322c8cc9e01b13288ab0f77cb542 (plain)
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
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
/* Search algorithm.
   Copyright (C) 1989-1998, 2000, 2002 Free Software Foundation, Inc.
   Written by Douglas C. Schmidt <schmidt@ics.uci.edu>
   and Bruno Haible <bruno@clisp.org>.

   This file is part of GNU GPERF.

   GNU GPERF 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, or (at your option)
   any later version.

   GNU GPERF 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; see the file COPYING.
   If not, write to the Free Software Foundation, Inc.,
   59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.  */

/* Specification. */
#include "search.h"

#include <stdio.h>
#include <stdlib.h> /* declares exit(), rand(), srand() */
#include <string.h> /* declares memset(), memcmp() */
#include <time.h> /* declares time() */
#include <math.h> /* declares exp() */
#include <limits.h> /* defines INT_MIN, INT_MAX, UINT_MAX */
#include "options.h"
#include "hash-table.h"
/* Assume ISO C++ 'for' scoping rule.  */
#define for if (0) ; else for

/* The most general form of the hash function is

      hash (keyword) = sum (asso_values[keyword[i] + alpha_inc[i]] : i in Pos)
                       + len (keyword)

   where Pos is a set of byte positions,
   each alpha_inc[i] is a nonnegative integer,
   each asso_values[c] is a nonnegative integer,
   len (keyword) is the keyword's length if !option[NOLENGTH], or 0 otherwise.

   Theorem 1: If all keywords are different, there is a set Pos such that
   all tuples (keyword[i] : i in Pos) are different.

   Theorem 2: If all tuples (keyword[i] : i in Pos) are different, there
   are nonnegative integers alpha_inc[i] such that all multisets
   {keyword[i] + alpha_inc[i] : i in Pos} are different.

   Theorem 3: If all multisets selchars[keyword] are different, there are
   nonnegative integers asso_values[c] such that all hash values
   sum (asso_values[c] : c in selchars[keyword]) are different.

   Based on these three facts, we find the hash function in three steps:

   Step 1 (Finding good byte positions):
   Find a set Pos, as small as possible, such that all tuples
   (keyword[i] : i in Pos) are different.

   Step 2 (Finding good alpha increments):
   Find nonnegative integers alpha_inc[i], as many of them as possible being
   zero, and the others being as small as possible, such that all multisets
   {keyword[i] + alpha_inc[i] : i in Pos} are different.

   Step 3 (Finding good asso_values):
   Find asso_values[c] such that all hash (keyword) are different.

   In other words, each step finds a projection that is injective on the
   given finite set:
     proj1 : String --> Map (Pos --> N)
     proj2 : Map (Pos --> N) --> Map (Pos --> N) / S(Pos)
     proj3 : Map (Pos --> N) / S(Pos) --> N
   where N denotes the set of nonnegative integers, and S(Pos) is the
   symmetric group over Pos.

   This was the theory for option[NOLENGTH]; if !option[NOLENGTH], slight
   modifications apply:
     proj1 : String --> Map (Pos --> N) x N
     proj2 : Map (Pos --> N) x N --> Map (Pos --> N) / S(Pos) x N
     proj3 : Map (Pos --> N) / S(Pos) x N --> N
 */

/* ==================== Initialization and Preparation ===================== */

Search::Search (KeywordExt_List *list)
  : _head (list)
{
}

void
Search::preprepare ()
{
  KeywordExt_List *temp;

  /* Compute the total number of keywords.  */
  _total_keys = 0;
  for (temp = _head; temp; temp = temp->rest())
    _total_keys++;

  /* Compute the minimum and maximum keyword length.  */
  _max_key_len = INT_MIN;
  _min_key_len = INT_MAX;
  for (temp = _head; temp; temp = temp->rest())
    {
      KeywordExt *keyword = temp->first();

      if (_max_key_len < keyword->_allchars_length)
        _max_key_len = keyword->_allchars_length;
      if (_min_key_len > keyword->_allchars_length)
        _min_key_len = keyword->_allchars_length;
    }

  /* Exit program if an empty string is used as keyword, since the comparison
     expressions don't work correctly for looking up an empty string.  */
  if (_min_key_len == 0)
    {
      fprintf (stderr, "Empty input key is not allowed.\n"
                       "To recognize an empty input key, your code should check for\n"
                       "len == 0 before calling the gperf generated lookup function.\n");
      exit (1);
    }

  /* Exit program if the characters in the keywords are not in the required
     range.  */
  if (option[SEVENBIT])
    for (temp = _head; temp; temp = temp->rest())
      {
        KeywordExt *keyword = temp->first();

        const char *k = keyword->_allchars;
        for (int i = keyword->_allchars_length; i > 0; k++, i--)
          if (!(static_cast<unsigned char>(*k) < 128))
            {
              fprintf (stderr, "Option --seven-bit has been specified,\n"
                       "but keyword \"%.*s\" contains non-ASCII characters.\n"
                       "Try removing option --seven-bit.\n",
                       keyword->_allchars_length, keyword->_allchars);
              exit (1);
            }
      }
}

/* ====================== Finding good byte positions ====================== */

/* Initializes each keyword's _selchars array.  */
void
Search::init_selchars_tuple (bool use_all_chars, const Positions& positions) const
{
  for (KeywordExt_List *temp = _head; temp; temp = temp->rest())
    temp->first()->init_selchars_tuple(use_all_chars, positions);
}

/* Deletes each keyword's _selchars array.  */
void
Search::delete_selchars () const
{
  for (KeywordExt_List *temp = _head; temp; temp = temp->rest())
    temp->first()->delete_selchars();
}

/* Count the duplicate keywords that occur with a given set of positions.
   In other words, it returns the difference
     # K - # proj1 (K)
   where K is the multiset of given keywords.  */
unsigned int
Search::count_duplicates_tuple (const Positions& positions) const
{
  /* Run through the keyword list and count the duplicates incrementally.
     The result does not depend on the order of the keyword list, thanks to
     the formula above.  */
  init_selchars_tuple (option[ALLCHARS], positions);

  unsigned int count = 0;
  {
    Hash_Table representatives (_total_keys, option[NOLENGTH]);
    for (KeywordExt_List *temp = _head; temp; temp = temp->rest())
      {
        KeywordExt *keyword = temp->first();
        if (representatives.insert (keyword))
          count++;
      }
  }

  delete_selchars ();

  return count;
}

/* Find good key positions.  */

void
Search::find_positions ()
{
  /* If the user gave the key positions, we use them.  */
  if (option[POSITIONS])
    {
      _key_positions = option.get_key_positions();
      return;
    }

  /* 1. Find positions that must occur in order to distinguish duplicates.  */
  Positions mandatory;

  if (!option[DUP])
    {
      for (KeywordExt_List *l1 = _head; l1 && l1->rest(); l1 = l1->rest())
        {
          KeywordExt *keyword1 = l1->first();
          for (KeywordExt_List *l2 = l1->rest(); l2; l2 = l2->rest())
            {
              KeywordExt *keyword2 = l2->first();

              /* If keyword1 and keyword2 have the same length and differ
                 in just one position, and it is not the last character,
                 this position is mandatory.  */
              if (keyword1->_allchars_length == keyword2->_allchars_length)
                {
                  int n = keyword1->_allchars_length;
                  int i;
                  for (i = 1; i < n; i++)
                    if (keyword1->_allchars[i-1] != keyword2->_allchars[i-1])
                      break;
                  if (i < n
                      && memcmp (&keyword1->_allchars[i],
                                 &keyword2->_allchars[i],
                                 n - i)
                         == 0)
                    {
                      /* Position i is mandatory.  */
                      if (!mandatory.contains (i))
                        mandatory.add (i);
                    }
                }
            }
        }
    }

  /* 2. Add positions, as long as this decreases the duplicates count.  */
  int imax = (_max_key_len < Positions::MAX_KEY_POS
              ? _max_key_len : Positions::MAX_KEY_POS);
  Positions current = mandatory;
  unsigned int current_duplicates_count = count_duplicates_tuple (current);
  for (;;)
    {
      Positions best;
      unsigned int best_duplicates_count = UINT_MAX;

      for (int i = imax; i >= 0; i--)
        if (!current.contains (i))
          {
            Positions tryal = current;
            tryal.add (i);
            unsigned int try_duplicates_count = count_duplicates_tuple (tryal);

            /* We prefer 'try' to 'best' if it produces less duplicates,
               or if it produces the same number of duplicates but with
               a more efficient hash function.  */
            if (try_duplicates_count < best_duplicates_count
                || (try_duplicates_count == best_duplicates_count && i > 0))
              {
                best = tryal;
                best_duplicates_count = try_duplicates_count;
              }
          }

      /* Stop adding positions when it gives no improvement.  */
      if (best_duplicates_count >= current_duplicates_count)
        break;

      current = best;
      current_duplicates_count = best_duplicates_count;
    }

  /* 3. Remove positions, as long as this doesn't increase the duplicates
     count.  */
  for (;;)
    {
      Positions best;
      unsigned int best_duplicates_count = UINT_MAX;

      for (int i = imax; i >= 0; i--)
        if (current.contains (i) && !mandatory.contains (i))
          {
            Positions tryal = current;
            tryal.remove (i);
            unsigned int try_duplicates_count = count_duplicates_tuple (tryal);

            /* We prefer 'try' to 'best' if it produces less duplicates,
               or if it produces the same number of duplicates but with
               a more efficient hash function.  */
            if (try_duplicates_count < best_duplicates_count
                || (try_duplicates_count == best_duplicates_count && i == 0))
              {
                best = tryal;
                best_duplicates_count = try_duplicates_count;
              }
          }

      /* Stop removing positions when it gives no improvement.  */
      if (best_duplicates_count > current_duplicates_count)
        break;

      current = best;
      current_duplicates_count = best_duplicates_count;
    }

  /* 4. Replace two positions by one, as long as this doesn't increase the
     duplicates count.  */
  for (;;)
    {
      Positions best;
      unsigned int best_duplicates_count = UINT_MAX;

      for (int i1 = imax; i1 >= 0; i1--)
        if (current.contains (i1) && !mandatory.contains (i1))
          for (int i2 = imax; i2 >= 0; i2--)
            if (current.contains (i2) && !mandatory.contains (i2) && i2 != i1)
              for (int i3 = imax; i3 >= 0; i3--)
                if (!current.contains (i3))
                  {
                    Positions tryal = current;
                    tryal.remove (i1);
                    tryal.remove (i2);
                    tryal.add (i3);
                    unsigned int try_duplicates_count =
                      count_duplicates_tuple (tryal);

                    /* We prefer 'try' to 'best' if it produces less duplicates,
                       or if it produces the same number of duplicates but with
                       a more efficient hash function.  */
                    if (try_duplicates_count < best_duplicates_count
                        || (try_duplicates_count == best_duplicates_count
                            && (i1 == 0 || i2 == 0 || i3 > 0)))
                      {
                        best = tryal;
                        best_duplicates_count = try_duplicates_count;
                      }
                  }

      /* Stop removing positions when it gives no improvement.  */
      if (best_duplicates_count > current_duplicates_count)
        break;

      current = best;
      current_duplicates_count = best_duplicates_count;
    }

  /* That's it.  Hope it's good enough.  */
  _key_positions = current;

  if (option[DEBUG])
    {
      /* Print the result.  */
      fprintf (stderr, "\nComputed positions: ");
      PositionReverseIterator iter (_key_positions);
      bool seen_lastchar = false;
      bool first = true;
      for (int i; (i = iter.next ()) != PositionReverseIterator::EOS; )
        {
          if (!first)
            fprintf (stderr, ", ");
          if (i == Positions::LASTCHAR)
            seen_lastchar = true;
          else
            {
              fprintf (stderr, "%d", i);
              first = false;
            }
        }
      if (seen_lastchar)
        {
          if (!first)
            fprintf (stderr, ", ");
          fprintf (stderr, "$");
        }
      fprintf (stderr, "\n");
    }
}

/* ===================== Finding good alpha increments ===================== */

/* Initializes each keyword's _selchars array.  */
void
Search::init_selchars_multiset (bool use_all_chars, const Positions& positions, const unsigned int *alpha_inc) const
{
  for (KeywordExt_List *temp = _head; temp; temp = temp->rest())
    temp->first()->init_selchars_multiset(use_all_chars, positions, alpha_inc);
}

/* Count the duplicate keywords that occur with the given set of positions
   and a given alpha_inc[] array.
   In other words, it returns the difference
     # K - # proj2 (proj1 (K))
   where K is the multiset of given keywords.  */
unsigned int
Search::count_duplicates_multiset (const unsigned int *alpha_inc) const
{
  /* Run through the keyword list and count the duplicates incrementally.
     The result does not depend on the order of the keyword list, thanks to
     the formula above.  */
  init_selchars_multiset (option[ALLCHARS], _key_positions, alpha_inc);

  unsigned int count = 0;
  {
    Hash_Table representatives (_total_keys, option[NOLENGTH]);
    for (KeywordExt_List *temp = _head; temp; temp = temp->rest())
      {
        KeywordExt *keyword = temp->first();
        if (representatives.insert (keyword))
          count++;
      }
  }

  delete_selchars ();

  return count;
}

/* Find good _alpha_inc[].  */

void
Search::find_alpha_inc ()
{
  /* The goal is to choose _alpha_inc[] such that it doesn't introduce
     artificial duplicates.
     In other words, the goal is  # proj2 (proj1 (K)) = # proj1 (K).  */
  unsigned int duplicates_goal = count_duplicates_tuple (_key_positions);

  /* Start with zero increments.  This is sufficient in most cases.  */
  unsigned int *current = new unsigned int [_max_key_len];
  for (int i = 0; i < _max_key_len; i++)
    current[i] = 0;
  unsigned int current_duplicates_count = count_duplicates_multiset (current);

  if (current_duplicates_count > duplicates_goal)
    {
      /* Look which _alpha_inc[i] we are free to increment.  */
      unsigned int nindices;
      if (option[ALLCHARS])
        nindices = _max_key_len;
      else
        {
          /* Ignore Positions::LASTCHAR.  Remember that since Positions are
             sorted in decreasing order, Positions::LASTCHAR comes last.  */
          nindices = (_key_positions.get_size() == 0
                      || _key_positions[_key_positions.get_size() - 1]
                         != Positions::LASTCHAR
                      ? _key_positions.get_size()
                      : _key_positions.get_size() - 1);
        }

      unsigned int indices[nindices];
      if (option[ALLCHARS])
        for (unsigned int j = 0; j < nindices; j++)
          indices[j] = j;
      else
        {
          PositionIterator iter (_key_positions);
          for (unsigned int j = 0; j < nindices; j++)
            {
              int key_pos = iter.next ();
              if (key_pos == PositionIterator::EOS
                  || key_pos == Positions::LASTCHAR)
                abort ();
              indices[j] = key_pos - 1;
            }
        }

      /* Perform several rounds of searching for a good alpha increment.
         Each round reduces the number of artificial collisions by adding
         an increment in a single key position.  */
      unsigned int best[_max_key_len];
      unsigned int tryal[_max_key_len];
      do
        {
          /* An increment of 1 is not always enough.  Try higher increments
             also.  */
          for (unsigned int inc = 1; ; inc++)
            {
              unsigned int best_duplicates_count = UINT_MAX;

              for (unsigned int j = 0; j < nindices; j++)
                {
                  memcpy (tryal, current, _max_key_len * sizeof (unsigned int));
                  tryal[indices[j]] += inc;
                  unsigned int try_duplicates_count =
                    count_duplicates_multiset (tryal);

                  /* We prefer 'try' to 'best' if it produces less
                     duplicates.  */
                  if (try_duplicates_count < best_duplicates_count)
                    {
                      memcpy (best, tryal, _max_key_len * sizeof (unsigned int));
                      best_duplicates_count = try_duplicates_count;
                    }
                }

              /* Stop this round when we got an improvement.  */
              if (best_duplicates_count < current_duplicates_count)
                {
                  memcpy (current, best, _max_key_len * sizeof (unsigned int));
                  current_duplicates_count = best_duplicates_count;
                  break;
                }
            }
        }
      while (current_duplicates_count > duplicates_goal);

      if (option[DEBUG])
        {
          /* Print the result.  */
          fprintf (stderr, "\nComputed alpha increments: ");
          if (option[ALLCHARS])
            {
              bool first = true;
              for (unsigned int j = 0; j < nindices; j++)
                if (current[indices[j]] != 0)
                  {
                    if (!first)
                      fprintf (stderr, ", ");
                    fprintf (stderr, "%u:+%u",
                             indices[j] + 1, current[indices[j]]);
                    first = false;
                  }
            }
          else
            {
              bool first = true;
              for (unsigned int j = nindices; j-- > 0; )
                if (current[indices[j]] != 0)
                  {
                    if (!first)
                      fprintf (stderr, ", ");
                    fprintf (stderr, "%u:+%u",
                             indices[j] + 1, current[indices[j]]);
                    first = false;
                  }
            }
          fprintf (stderr, "\n");
        }
    }

  _alpha_inc = current;
}

/* ======================= Finding good asso_values ======================== */

void
Search::prepare ()
{
  KeywordExt_List *temp;

  /* Initialize each keyword's _selchars array.  */
  init_selchars_multiset(option[ALLCHARS], _key_positions, _alpha_inc);

  /* Check for duplicates, i.e. keywords with the same _selchars array
     (and - if !option[NOLENGTH] - also the same length).
     We deal with these by building an equivalence class, so that only
     1 keyword is representative of the entire collection.  Only this
     representative remains in the keyword list; the others are accessible
     through the _duplicate_link chain, starting at the representative.
     This *greatly* simplifies processing during later stages of the program.
     Set _total_duplicates and _list_len = _total_keys - _total_duplicates.  */
  {
    _list_len = _total_keys;
    _total_duplicates = 0;
    /* Make hash table for efficiency.  */
    Hash_Table representatives (_list_len, option[NOLENGTH]);

    KeywordExt_List *prev = NULL; /* list node before temp */
    for (temp = _head; temp; )
      {
        KeywordExt *keyword = temp->first();
        KeywordExt *other_keyword = representatives.insert (keyword);
        KeywordExt_List *garbage = NULL;

        if (other_keyword)
          {
            _total_duplicates++;
            _list_len--;
            /* Remove keyword from the main list.  */
            prev->rest() = temp->rest();
            garbage = temp;
            /* And insert it on other_keyword's duplicate list.  */
            keyword->_duplicate_link = other_keyword->_duplicate_link;
            other_keyword->_duplicate_link = keyword;

            /* Complain if user hasn't enabled the duplicate option. */
            if (!option[DUP] || option[DEBUG])
              {
                fprintf (stderr, "Key link: \"%.*s\" = \"%.*s\", with key set \"",
                         keyword->_allchars_length, keyword->_allchars,
                         other_keyword->_allchars_length, other_keyword->_allchars);
                for (int j = 0; j < keyword->_selchars_length; j++)
                  putc (keyword->_selchars[j], stderr);
                fprintf (stderr, "\".\n");
              }
          }
        else
          {
            keyword->_duplicate_link = NULL;
            prev = temp;
          }
        temp = temp->rest();
        if (garbage)
          delete garbage;
      }
    if (option[DEBUG])
      representatives.dump();
  }

  /* Exit program if duplicates exists and option[DUP] not set, since we
     don't want to continue in this case.  (We don't want to turn on
     option[DUP] implicitly, because the generated code is usually much
     slower.  */
  if (_total_duplicates)
    {
      if (option[DUP])
        fprintf (stderr, "%d input keys have identical hash values, examine output carefully...\n",
                         _total_duplicates);
      else
        {
          fprintf (stderr, "%d input keys have identical hash values,\n",
                           _total_duplicates);
          if (option[POSITIONS])
            fprintf (stderr, "try different key positions or use option -D.\n");
          else
            fprintf (stderr, "use option -D.\n");
          exit (1);
        }
    }

  /* Compute _alpha_size, the upper bound on the indices passed to
     asso_values[].  */
  unsigned int max_alpha_inc = 0;
  for (int i = 0; i < _max_key_len; i++)
    if (max_alpha_inc < _alpha_inc[i])
      max_alpha_inc = _alpha_inc[i];
  _alpha_size = (option[SEVENBIT] ? 128 : 256) + max_alpha_inc;

  /* Compute the occurrences of each character in the alphabet.  */
  _occurrences = new int[_alpha_size];
  memset (_occurrences, 0, _alpha_size * sizeof (_occurrences[0]));
  for (temp = _head; temp; temp = temp->rest())
    {
      KeywordExt *keyword = temp->first();
      const unsigned int *ptr = keyword->_selchars;
      for (int count = keyword->_selchars_length; count > 0; ptr++, count--)
        _occurrences[*ptr]++;
    }

  /* Memory allocation.  */
  _asso_values = new int[_alpha_size];
}

/* ------------------------------------------------------------------------- */

/* Returns the length of keyword list.  */

int
Search::keyword_list_length () const
{
  return _list_len;
}

/* Returns the maximum length of keywords.  */

int
Search::max_key_length () const
{
  return _max_key_len;
}

/* Returns the number of key positions.  */

int
Search::get_max_keysig_size () const
{
  return option[ALLCHARS] ? _max_key_len : _key_positions.get_size();
}

/* ---------------------- Finding good asso_values[] ----------------------- */

/* Initializes the asso_values[] related parameters.  */

void
Search::prepare_asso_values ()
{
  int non_linked_length = keyword_list_length ();
  unsigned int asso_value_max;

  asso_value_max =
    static_cast<unsigned int>(non_linked_length * option.get_size_multiple());
  /* Round up to the next power of two.  This makes it easy to ensure
     an _asso_value[c] is >= 0 and < asso_value_max.  Also, the jump value
     being odd, it guarantees that Search::try_asso_value() will iterate
     through different values for _asso_value[c].  */
  if (asso_value_max == 0)
    asso_value_max = 1;
  asso_value_max |= asso_value_max >> 1;
  asso_value_max |= asso_value_max >> 2;
  asso_value_max |= asso_value_max >> 4;
  asso_value_max |= asso_value_max >> 8;
  asso_value_max |= asso_value_max >> 16;
  asso_value_max++;
  _asso_value_max = asso_value_max;

  /* Given the bound for _asso_values[c], we have a bound for the possible
     hash values, as computed in compute_hash().  */
  _max_hash_value = (option[NOLENGTH] ? 0 : max_key_length ())
                    + (_asso_value_max - 1) * get_max_keysig_size ();
  /* Allocate a sparse bit vector for detection of collisions of hash
     values.  */
  _collision_detector = new Bool_Array (_max_hash_value + 1);

  if (option[DEBUG])
    {
      fprintf (stderr, "total non-linked keys = %d\nmaximum associated value is %d"
               "\nmaximum size of generated hash table is %d\n",
               non_linked_length, asso_value_max, _max_hash_value);

      int field_width;

      field_width = 0;
      {
        for (KeywordExt_List *temp = _head; temp; temp = temp->rest())
          {
            KeywordExt *keyword = temp->first();
            if (field_width < keyword->_selchars_length)
              field_width = keyword->_selchars_length;
          }
      }

      fprintf (stderr, "\ndumping the keyword list without duplicates\n");
      fprintf (stderr, "keyword #, %*s, keyword\n", field_width, "keysig");
      int i = 0;
      for (KeywordExt_List *temp = _head; temp; temp = temp->rest())
        {
          KeywordExt *keyword = temp->first();
          fprintf (stderr, "%9d, ", ++i);
          if (field_width > keyword->_selchars_length)
            fprintf (stderr, "%*s", field_width - keyword->_selchars_length, "");
          for (int j = 0; j < keyword->_selchars_length; j++)
            putc (keyword->_selchars[j], stderr);
          fprintf (stderr, ", %.*s\n",
                   keyword->_allchars_length, keyword->_allchars);
        }
      fprintf (stderr, "\nend of keyword list\n\n");
    }

  if (option[RANDOM] || option.get_jump () == 0)
    /* We will use rand(), so initialize the random number generator.  */
    srand (static_cast<long>(time (0)));

  _initial_asso_value = (option[RANDOM] ? -1 : option.get_initial_asso_value ());
  _jump = option.get_jump ();
}

/* Finds some _asso_values[] that fit.  */

/* The idea is to choose the _asso_values[] one by one, in a way that
   a choice that has been made never needs to be undone later.  This
   means that we split the work into several steps.  Each step chooses
   one or more _asso_values[c].  The result of choosing one or more
   _asso_values[c] is that the partitioning of the keyword set gets
   broader.
   Look at this partitioning:  After every step, the _asso_values[] of a
   certain set C of characters are undetermined.  (At the beginning, C
   is the set of characters c with _occurrences[c] > 0.  At the end, C
   is empty.)  To each keyword K, we associate the multiset of _selchars
   for which the _asso_values[] are undetermined:
                    K  -->  K->_selchars intersect C.
   Consider two keywords equivalent if their value under this mapping is
   the same.  This introduces an equivalence relation on the set of
   keywords.  The equivalence classes partition the keyword set.  (At the
   beginning, the partition is the finest possible: each K is an equivalence
   class by itself, because all K have a different _selchars.  At the end,
   all K have been merged into a single equivalence class.)
   The partition before a step is always a refinement of the partition
   after the step.
   We choose the steps in such a way that the partition really becomes
   broader at each step.  (A step that only chooses an _asso_values[c]
   without changing the partition is better merged with the previous step,
   to avoid useless backtracking.)  */

struct EquivalenceClass
{
  /* The keywords in this equivalence class.  */
  KeywordExt_List *     _keywords;
  KeywordExt_List *     _keywords_last;
  /* The number of keywords in this equivalence class.  */
  unsigned int          _cardinality;
  /* The undetermined selected characters for the keywords in this
     equivalence class, as a canonically reordered multiset.  */
  unsigned int *        _undetermined_chars;
  unsigned int          _undetermined_chars_length;

  EquivalenceClass *    _next;
};

struct Step
{
  /* The characters whose values are being determined in this step.  */
  unsigned int          _changing_count;
  unsigned int *        _changing;
  /* Exclusive upper bound for the _asso_values[c] of this step.
     A power of 2.  */
  unsigned int          _asso_value_max;
  /* The characters whose values will be determined after this step.  */
  bool *                _undetermined;
  /* The keyword set partition after this step.  */
  EquivalenceClass *    _partition;
  /* The expected number of iterations in this step.  */
  double                _expected_lower;
  double                _expected_upper;

  Step *                _next;
};

static inline bool
equals (const unsigned int *ptr1, const unsigned int *ptr2, unsigned int len)
{
  while (len > 0)
    {
      if (*ptr1 != *ptr2)
        return false;
      ptr1++;
      ptr2++;
      len--;
    }
  return true;
}

EquivalenceClass *
Search::compute_partition (bool *undetermined) const
{
  EquivalenceClass *partition = NULL;
  EquivalenceClass *partition_last = NULL;
  for (KeywordExt_List *temp = _head; temp; temp = temp->rest())
    {
      KeywordExt *keyword = temp->first();

      /* Compute the undetermined characters for this keyword.  */
      unsigned int *undetermined_chars =
        new unsigned int[keyword->_selchars_length];
      unsigned int undetermined_chars_length = 0;

      for (int i = 0; i < keyword->_selchars_length; i++)
        if (undetermined[keyword->_selchars[i]])
          undetermined_chars[undetermined_chars_length++] = keyword->_selchars[i];

      /* Look up the equivalence class to which this keyword belongs.  */
      EquivalenceClass *equclass;
      for (equclass = partition; equclass; equclass = equclass->_next)
        if (equclass->_undetermined_chars_length == undetermined_chars_length
            && equals (equclass->_undetermined_chars, undetermined_chars,
                       undetermined_chars_length))
          break;
      if (equclass == NULL)
        {
          equclass = new EquivalenceClass();
          equclass->_keywords = NULL;
          equclass->_keywords_last = NULL;
          equclass->_cardinality = 0;
          equclass->_undetermined_chars = undetermined_chars;
          equclass->_undetermined_chars_length = undetermined_chars_length;
          equclass->_next = NULL;
          if (partition)
            partition_last->_next = equclass;
          else
            partition = equclass;
          partition_last = equclass;
        }
      else
        delete[] undetermined_chars;

      /* Add the keyword to the equivalence class.  */
      KeywordExt_List *cons = new KeywordExt_List(keyword);
      if (equclass->_keywords)
        equclass->_keywords_last->rest() = cons;
      else
        equclass->_keywords = cons;
      equclass->_keywords_last = cons;
      equclass->_cardinality++;
    }

  /* Free some of the allocated memory.  The caller doesn't need it.  */
  for (EquivalenceClass *cls = partition; cls; cls = cls->_next)
    delete[] cls->_undetermined_chars;

  return partition;
}

static void
delete_partition (EquivalenceClass *partition)
{
  while (partition != NULL)
    {
      EquivalenceClass *equclass = partition;
      partition = equclass->_next;
      delete_list (equclass->_keywords);
      //delete[] equclass->_undetermined_chars; // already freed above
      delete equclass;
    }
}

/* Compute the possible number of collisions when _asso_values[c] is
   chosen, leading to the given partition.  */
unsigned int
Search::count_possible_collisions (EquivalenceClass *partition, unsigned int c) const
{
  /* Every equivalence class p is split according to the frequency of
     occurrence of c, leading to equivalence classes p1, p2, ...
     This leads to   (|p|^2 - |p1|^2 - |p2|^2 - ...)/2  possible collisions.
     Return the sum of this expression over all equivalence classes.  */
  unsigned int sum = 0;
  unsigned int m = get_max_keysig_size();
  unsigned int split_cardinalities[m+1];
  for (EquivalenceClass *cls = partition; cls; cls = cls->_next)
    {
      for (unsigned int i = 0; i <= m; i++)
        split_cardinalities[i] = 0;

      for (KeywordExt_List *temp = cls->_keywords; temp; temp = temp->rest())
        {
          KeywordExt *keyword = temp->first();

          unsigned int count = 0;
          for (int i = 0; i < keyword->_selchars_length; i++)
            if (keyword->_selchars[i] == c)
              count++;

          split_cardinalities[count]++;
        }

      sum += cls->_cardinality * cls->_cardinality;
      for (unsigned int i = 0; i <= m; i++)
        sum -= split_cardinalities[i] * split_cardinalities[i];
    }
  return sum;
}

/* Test whether adding c to the undetermined characters changes the given
   partition.  */
bool
Search::unchanged_partition (EquivalenceClass *partition, unsigned int c) const
{
  for (EquivalenceClass *cls = partition; cls; cls = cls->_next)
    {
      unsigned int first_count = UINT_MAX;

      for (KeywordExt_List *temp = cls->_keywords; temp; temp = temp->rest())
        {
          KeywordExt *keyword = temp->first();

          unsigned int count = 0;
          for (int i = 0; i < keyword->_selchars_length; i++)
            if (keyword->_selchars[i] == c)
              count++;

          if (temp == cls->_keywords)
            first_count = count;
          else if (count != first_count)
            /* c would split this equivalence class.  */
            return false;
        }
    }
  return true;
}

void
Search::find_asso_values ()
{
  Step *steps;

  /* Determine the steps, starting with the last one.  */
  {
    bool *undetermined;
    bool *determined;

    steps = NULL;

    undetermined = new bool[_alpha_size];
    for (unsigned int c = 0; c < _alpha_size; c++)
      undetermined[c] = false;

    determined = new bool[_alpha_size];
    for (unsigned int c = 0; c < _alpha_size; c++)
      determined[c] = true;

    for (;;)
      {
        /* Compute the partition that needs to be refined.  */
        EquivalenceClass *partition = compute_partition (undetermined);

        /* Determine the main character to be chosen in this step.
           Choosing such a character c has the effect of splitting every
           equivalence class (according the the frequency of occurrence of c).
           We choose the c with the minimum number of possible collisions,
           so that characters which lead to a large number of collisions get
           handled early during the search.  */
        unsigned int chosen_c;
        unsigned int chosen_possible_collisions;
        {
          unsigned int best_c = 0;
          unsigned int best_possible_collisions = UINT_MAX;
          for (unsigned int c = 0; c < _alpha_size; c++)
            if (_occurrences[c] > 0 && determined[c])
              {
                unsigned int possible_collisions =
                  count_possible_collisions (partition, c);
                if (possible_collisions < best_possible_collisions)
                  {
                    best_c = c;
                    best_possible_collisions = possible_collisions;
                  }
              }
          if (best_possible_collisions == UINT_MAX)
            {
              /* All c with _occurrences[c] > 0 are undetermined.  We are
                 are the starting situation and don't need any more step.  */
              delete_partition (partition);
              break;
            }
          chosen_c = best_c;
          chosen_possible_collisions = best_possible_collisions;
        }

        /* We need one more step.  */
        Step *step = new Step();

        step->_undetermined = new bool[_alpha_size];
        memcpy (step->_undetermined, undetermined, _alpha_size*sizeof(bool));

        step->_partition = partition;

        /* Now determine how the equivalence classes will be before this
           step.  */
        undetermined[chosen_c] = true;
        partition = compute_partition (undetermined);

        /* Now determine which other characters should be determined in this
           step, because they will not change the equivalence classes at
           this point.  It is the set of all c which, for all equivalence
           classes, have the same frequency of occurrence in every keyword
           of the equivalence class.  */
        for (unsigned int c = 0; c < _alpha_size; c++)
          if (_occurrences[c] > 0 && determined[c]
              && unchanged_partition (partition, c))
            {
              undetermined[c] = true;
              determined[c] = false;
            }

        /* main_c must be one of these.  */
        if (determined[chosen_c])
          abort ();

        /* Now the set of changing characters of this step.  */
        unsigned int changing_count;

        changing_count = 0;
        for (unsigned int c = 0; c < _alpha_size; c++)
          if (undetermined[c] && !step->_undetermined[c])
            changing_count++;

        unsigned int *changing = new unsigned int[changing_count];
        changing_count = 0;
        for (unsigned int c = 0; c < _alpha_size; c++)
          if (undetermined[c] && !step->_undetermined[c])
            changing[changing_count++] = c;

        step->_changing = changing;
        step->_changing_count = changing_count;

        step->_asso_value_max = _asso_value_max;

        step->_expected_lower =
          exp (static_cast<double>(chosen_possible_collisions)
               / static_cast<double>(_max_hash_value));
        step->_expected_upper =
          exp (static_cast<double>(chosen_possible_collisions)
               / static_cast<double>(_asso_value_max));

        delete_partition (partition);

        step->_next = steps;
        steps = step;
      }

    delete[] determined;
    delete[] undetermined;
  }

  if (option[DEBUG])
    {
      unsigned int stepno = 0;
      for (Step *step = steps; step; step = step->_next)
        {
          stepno++;
          fprintf (stderr, "Step %u chooses _asso_values[", stepno);
          for (unsigned int i = 0; i < step->_changing_count; i++)
            {
              if (i > 0)
                fprintf (stderr, ",");
              fprintf (stderr, "'%c'", step->_changing[i]);
            }
          fprintf (stderr, "], expected number of iterations between %g and %g.\n",
                   step->_expected_lower, step->_expected_upper);
          fprintf (stderr, "Keyword equivalence classes:\n");
          for (EquivalenceClass *cls = step->_partition; cls; cls = cls->_next)
            {
              fprintf (stderr, "\n");
              for (KeywordExt_List *temp = cls->_keywords; temp; temp = temp->rest())
                {
                  KeywordExt *keyword = temp->first();
                  fprintf (stderr, "  %.*s\n",
                           keyword->_allchars_length, keyword->_allchars);
                }
            }
          fprintf (stderr, "\n");
        }
    }

  /* Initialize _asso_values[].  (The value given here matters only
     for those c which occur in all keywords with equal multiplicity.)  */
  for (unsigned int c = 0; c < _alpha_size; c++)
    _asso_values[c] = 0;

  unsigned int stepno = 0;
  for (Step *step = steps; step; step = step->_next)
    {
      stepno++;

      /* Initialize the asso_values[].  */
      unsigned int k = step->_changing_count;
      for (unsigned int i = 0; i < k; i++)
        {
          unsigned int c = step->_changing[i];
          _asso_values[c] =
            (_initial_asso_value < 0 ? rand () : _initial_asso_value)
            & (step->_asso_value_max - 1);
        }

      unsigned int iterations = 0;
      unsigned int iter[k];
      for (unsigned int i = 0; i < k; i++)
        iter[i] = 0;
      unsigned int ii = (_jump != 0 ? k - 1 : 0);

      for (;;)
        {
          /* Test whether these asso_values[] lead to collisions among
             the equivalence classes that should be collision-free.  */
          bool has_collision = false;
          for (EquivalenceClass *cls = step->_partition; cls; cls = cls->_next)
            {
              /* Iteration Number array is a win, O(1) initialization time!  */
              _collision_detector->clear ();

              for (KeywordExt_List *ptr = cls->_keywords; ptr; ptr = ptr->rest())
                {
                  KeywordExt *keyword = ptr->first();

                  /* Compute the new hash code for the keyword, leaving apart
                     the yet undetermined asso_values[].  */
                  int hashcode;
                  {
                    int sum = option[NOLENGTH] ? 0 : keyword->_allchars_length;
                    const unsigned int *p = keyword->_selchars;
                    int i = keyword->_selchars_length;
                    for (; i > 0; p++, i--)
                      if (!step->_undetermined[*p])
                        sum += _asso_values[*p];
                    hashcode = sum;
                  }

                  /* See whether it collides with another keyword's hash code,
                     from the same equivalence class.  */
                  if (_collision_detector->set_bit (hashcode))
                    {
                      has_collision = true;
                      break;
                    }
                }

              /* Don't need to continue looking at the other equivalence
                 classes if we already have found a collision.  */
              if (has_collision)
                break;
            }

          iterations++;
          if (!has_collision)
            break;

          /* Try other asso_values[].  */
          if (_jump != 0)
            {
              /* The way we try various values for
                   asso_values[step->_changing[0],...step->_changing[k-1]]
                 is like this:
                 for (bound = 0,1,...)
                   for (ii = 0,...,k-1)
                     iter[ii] := bound
                     iter[0..ii-1] := values <= bound
                     iter[ii+1..k-1] := values < bound
                 and
                   asso_values[step->_changing[i]] =
                     _initial_asso_value + iter[i] * _jump.
                 This makes it more likely to find small asso_values[].
               */
              unsigned int bound = iter[ii];
              unsigned int i = 0;
              while (i < ii)
                {
                  unsigned int c = step->_changing[i];
                  iter[i]++;
                  _asso_values[c] =
                    (_asso_values[c] + _jump) & (step->_asso_value_max - 1);
                  if (iter[i] <= bound)
                    goto found_next;
                  _asso_values[c] =
                    (_asso_values[c] - iter[i] * _jump)
                    & (step->_asso_value_max - 1);
                  iter[i] = 0;
                  i++;
                }
              i = ii + 1;
              while (i < k)
                {
                  unsigned int c = step->_changing[i];
                  iter[i]++;
                  _asso_values[c] =
                    (_asso_values[c] + _jump) & (step->_asso_value_max - 1);
                  if (iter[i] < bound)
                    goto found_next;
                  _asso_values[c] =
                    (_asso_values[c] - iter[i] * _jump)
                    & (step->_asso_value_max - 1);
                  iter[i] = 0;
                  i++;
                }
              /* Switch from one ii to the next.  */
              {
                unsigned int c = step->_changing[ii];
                _asso_values[c] =
                  (_asso_values[c] - bound * _jump)
                  & (step->_asso_value_max - 1);
                iter[ii] = 0;
              }
              /* Here all iter[i] == 0.  */
              ii++;
              if (ii == k)
                {
                  ii = 0;
                  bound++;
                  if (bound == step->_asso_value_max)
                    {
                      /* Out of search space!  We can either backtrack, or
                         increase the available search space of this step.
                         It seems simpler to choose the latter solution.  */
                      step->_asso_value_max = 2 * step->_asso_value_max;
                      if (step->_asso_value_max > _asso_value_max)
                        {
                          _asso_value_max = step->_asso_value_max;
                          /* Reinitialize _max_hash_value.  */
                          _max_hash_value =
                            (option[NOLENGTH] ? 0 : max_key_length ())
                            + (_asso_value_max - 1) * get_max_keysig_size ();
                          /* Reinitialize _collision_detector.  */
                          delete _collision_detector;
                          _collision_detector =
                            new Bool_Array (_max_hash_value + 1);
                        }
                    }
                }
              {
                unsigned int c = step->_changing[ii];
                iter[ii] = bound;
                _asso_values[c] =
                  (_asso_values[c] + bound * _jump)
                  & (step->_asso_value_max - 1);
              }
             found_next: ;
            }
          else
            {
              /* Random.  */
              unsigned int c = step->_changing[ii];
              _asso_values[c] =
                (_asso_values[c] + rand ()) & (step->_asso_value_max - 1);
              /* Next time, change the next c.  */
              ii++;
              if (ii == k)
                ii = 0;
            }
        }
      if (option[DEBUG])
        {
          fprintf (stderr, "Step %u chose _asso_values[", stepno);
          for (unsigned int i = 0; i < step->_changing_count; i++)
            {
              if (i > 0)
                fprintf (stderr, ",");
              fprintf (stderr, "'%c'", step->_changing[i]);
            }
          fprintf (stderr, "] in %u iterations.\n", iterations);
        }
    }

  /* Free allocated memory.  */
  while (steps != NULL)
    {
      Step *step = steps;
      steps = step->_next;
      delete[] step->_changing;
      delete[] step->_undetermined;
      delete_partition (step->_partition);
      delete step;
    }
}

/* Computes a keyword's hash value, relative to the current _asso_values[],
   and stores it in keyword->_hash_value.  */

inline int
Search::compute_hash (KeywordExt *keyword) const
{
  int sum = option[NOLENGTH] ? 0 : keyword->_allchars_length;

  const unsigned int *p = keyword->_selchars;
  int i = keyword->_selchars_length;
  for (; i > 0; p++, i--)
    sum += _asso_values[*p];

  return keyword->_hash_value = sum;
}

/* Finds good _asso_values[].  */

void
Search::find_good_asso_values ()
{
  prepare ();
  prepare_asso_values ();

  /* Search for good _asso_values[].  */
  int asso_iteration;
  if ((asso_iteration = option.get_asso_iterations ()) == 0)
    /* Try only the given _initial_asso_value and _jump.  */
    find_asso_values ();
  else
    {
      /* Try different pairs of _initial_asso_value and _jump, in the
         following order:
           (0, 1)
           (1, 1)
           (2, 1) (0, 3)
           (3, 1) (1, 3)
           (4, 1) (2, 3) (0, 5)
           (5, 1) (3, 3) (1, 5)
           ..... */
      KeywordExt_List *saved_head = _head;
      int best_initial_asso_value = 0;
      int best_jump = 1;
      int *best_asso_values = new int[_alpha_size];
      int best_collisions = INT_MAX;
      int best_max_hash_value = INT_MAX;

      _initial_asso_value = 0; _jump = 1;
      for (;;)
        {
          /* Restore the keyword list in its original order.  */
          _head = copy_list (saved_head);
          /* Find good _asso_values[].  */
          find_asso_values ();
          /* Test whether it is the best solution so far.  */
          int collisions = 0;
          int max_hash_value = INT_MIN;
          _collision_detector->clear ();
          for (KeywordExt_List *ptr = _head; ptr; ptr = ptr->rest())
            {
              KeywordExt *keyword = ptr->first();
              int hashcode = compute_hash (keyword);
              if (max_hash_value < hashcode)
                max_hash_value = hashcode;
              if (_collision_detector->set_bit (hashcode))
                collisions++;
            }
          if (collisions < best_collisions
              || (collisions == best_collisions
                  && max_hash_value < best_max_hash_value))
            {
              memcpy (best_asso_values, _asso_values,
                      _alpha_size * sizeof (_asso_values[0]));
              best_collisions = collisions;
              best_max_hash_value = max_hash_value;
            }
          /* Delete the copied keyword list.  */
          delete_list (_head);

          if (--asso_iteration == 0)
            break;
          /* Prepare for next iteration.  */
          if (_initial_asso_value >= 2)
            _initial_asso_value -= 2, _jump += 2;
          else
            _initial_asso_value += _jump, _jump = 1;
        }
      _head = saved_head;
      /* Install the best found asso_values.  */
      _initial_asso_value = best_initial_asso_value;
      _jump = best_jump;
      memcpy (_asso_values, best_asso_values,
              _alpha_size * sizeof (_asso_values[0]));
      delete[] best_asso_values;
      /* The keywords' _hash_value fields are recomputed below.  */
    }
}

/* ========================================================================= */

/* Comparison function for sorting by increasing _hash_value.  */
static bool
less_by_hash_value (KeywordExt *keyword1, KeywordExt *keyword2)
{
  return keyword1->_hash_value < keyword2->_hash_value;
}

/* Sorts the keyword list by hash value.  */

void
Search::sort ()
{
  _head = mergesort_list (_head, less_by_hash_value);
}

void
Search::optimize ()
{
  /* Preparations.  */
  preprepare ();

  /* Step 1: Finding good byte positions.  */
  find_positions ();

  /* Step 2: Finding good alpha increments.  */
  find_alpha_inc ();

  /* Step 3: Finding good asso_values.  */
  find_good_asso_values ();

  /* Make one final check, just to make sure nothing weird happened.... */
  _collision_detector->clear ();
  for (KeywordExt_List *curr_ptr = _head; curr_ptr; curr_ptr = curr_ptr->rest())
    {
      KeywordExt *curr = curr_ptr->first();
      unsigned int hashcode = compute_hash (curr);
      if (_collision_detector->set_bit (hashcode))
        {
          /* This shouldn't happen.  proj1, proj2, proj3 must have been
             computed to be injective on the given keyword set.  */
          fprintf (stderr,
                   "\nInternal error, unexpected duplicate hash code\n");
          if (option[POSITIONS])
            fprintf (stderr, "try options -m or -r, or use new key positions.\n\n");
          else
            fprintf (stderr, "try options -m or -r.\n\n");
          exit (1);
        }
    }

  /* Sorts the keyword list by hash value.  */
  sort ();

  /* Set unused asso_values[c] to max_hash_value + 1.  This is not absolutely
     necessary, but speeds up the lookup function in many cases of lookup
     failure: no string comparison is needed once the hash value of a string
     is larger than the hash value of any keyword.  */
  int max_hash_value;
  {
    KeywordExt_List *temp;
    for (temp = _head; temp->rest(); temp = temp->rest())
      ;
    max_hash_value = temp->first()->_hash_value;
  }
  for (unsigned int c = 0; c < _alpha_size; c++)
    if (_occurrences[c] == 0)
      _asso_values[c] = max_hash_value + 1;
}

/* Prints out some diagnostics upon completion.  */

Search::~Search ()
{
  delete _collision_detector;
  if (option[DEBUG])
    {
      fprintf (stderr, "\ndumping occurrence and associated values tables\n");

      for (unsigned int i = 0; i < _alpha_size; i++)
        if (_occurrences[i])
          fprintf (stderr, "asso_values[%c] = %6d, occurrences[%c] = %6d\n",
                   i, _asso_values[i], i, _occurrences[i]);

      fprintf (stderr, "end table dumping\n");

      fprintf (stderr, "\nDumping key list information:\ntotal non-static linked keywords = %d"
               "\ntotal keywords = %d\ntotal duplicates = %d\nmaximum key length = %d\n",
               _list_len, _total_keys, _total_duplicates, _max_key_len);

      int field_width = get_max_keysig_size ();
      fprintf (stderr, "\nList contents are:\n(hash value, key length, index, %*s, keyword):\n",
               field_width, "selchars");
      for (KeywordExt_List *ptr = _head; ptr; ptr = ptr->rest())
        {
          fprintf (stderr, "%11d,%11d,%6d, ",
                   ptr->first()->_hash_value, ptr->first()->_allchars_length, ptr->first()->_final_index);
          if (field_width > ptr->first()->_selchars_length)
            fprintf (stderr, "%*s", field_width - ptr->first()->_selchars_length, "");
          for (int j = 0; j < ptr->first()->_selchars_length; j++)
            putc (ptr->first()->_selchars[j], stderr);
          fprintf (stderr, ", %.*s\n",
                   ptr->first()->_allchars_length, ptr->first()->_allchars);
        }

      fprintf (stderr, "End dumping list.\n\n");
    }
  delete[] _asso_values;
  delete[] _occurrences;
  delete[] _alpha_inc;
}