Line data Source code
1 : /*
2 : * fuzzystrmatch.c
3 : *
4 : * Functions for "fuzzy" comparison of strings
5 : *
6 : * Joe Conway <mail@joeconway.com>
7 : *
8 : * contrib/fuzzystrmatch/fuzzystrmatch.c
9 : * Copyright (c) 2001-2025, PostgreSQL Global Development Group
10 : * ALL RIGHTS RESERVED;
11 : *
12 : * metaphone()
13 : * -----------
14 : * Modified for PostgreSQL by Joe Conway.
15 : * Based on CPAN's "Text-Metaphone-1.96" by Michael G Schwern <schwern@pobox.com>
16 : * Code slightly modified for use as PostgreSQL function (palloc, elog, etc).
17 : * Metaphone was originally created by Lawrence Philips and presented in article
18 : * in "Computer Language" December 1990 issue.
19 : *
20 : * Permission to use, copy, modify, and distribute this software and its
21 : * documentation for any purpose, without fee, and without a written agreement
22 : * is hereby granted, provided that the above copyright notice and this
23 : * paragraph and the following two paragraphs appear in all copies.
24 : *
25 : * IN NO EVENT SHALL THE AUTHORS OR DISTRIBUTORS BE LIABLE TO ANY PARTY FOR
26 : * DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES, INCLUDING
27 : * LOST PROFITS, ARISING OUT OF THE USE OF THIS SOFTWARE AND ITS
28 : * DOCUMENTATION, EVEN IF THE AUTHOR OR DISTRIBUTORS HAVE BEEN ADVISED OF THE
29 : * POSSIBILITY OF SUCH DAMAGE.
30 : *
31 : * THE AUTHORS AND DISTRIBUTORS SPECIFICALLY DISCLAIM ANY WARRANTIES,
32 : * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY
33 : * AND FITNESS FOR A PARTICULAR PURPOSE. THE SOFTWARE PROVIDED HEREUNDER IS
34 : * ON AN "AS IS" BASIS, AND THE AUTHOR AND DISTRIBUTORS HAS NO OBLIGATIONS TO
35 : * PROVIDE MAINTENANCE, SUPPORT, UPDATES, ENHANCEMENTS, OR MODIFICATIONS.
36 : *
37 : */
38 :
39 : #include "postgres.h"
40 :
41 : #include <ctype.h>
42 :
43 : #include "utils/builtins.h"
44 : #include "utils/varlena.h"
45 : #include "varatt.h"
46 :
47 4 : PG_MODULE_MAGIC_EXT(
48 : .name = "fuzzystrmatch",
49 : .version = PG_VERSION
50 : );
51 :
52 : /*
53 : * Soundex
54 : */
55 : static void _soundex(const char *instr, char *outstr);
56 :
57 : #define SOUNDEX_LEN 4
58 :
59 : /* ABCDEFGHIJKLMNOPQRSTUVWXYZ */
60 : static const char *const soundex_table = "01230120022455012623010202";
61 :
62 : static char
63 254 : soundex_code(char letter)
64 : {
65 254 : letter = toupper((unsigned char) letter);
66 : /* Defend against non-ASCII letters */
67 254 : if (letter >= 'A' && letter <= 'Z')
68 252 : return soundex_table[letter - 'A'];
69 2 : return letter;
70 : }
71 :
72 : /*
73 : * Metaphone
74 : */
75 : #define MAX_METAPHONE_STRLEN 255
76 :
77 : /*
78 : * Original code by Michael G Schwern starts here.
79 : * Code slightly modified for use as PostgreSQL function.
80 : */
81 :
82 :
83 : /**************************************************************************
84 : metaphone -- Breaks english phrases down into their phonemes.
85 :
86 : Input
87 : word -- An english word to be phonized
88 : max_phonemes -- How many phonemes to calculate. If 0, then it
89 : will phonize the entire phrase.
90 : phoned_word -- The final phonized word. (We'll allocate the
91 : memory.)
92 : Output
93 : error -- A simple error flag, returns true or false
94 :
95 : NOTES: ALL non-alpha characters are ignored, this includes whitespace,
96 : although non-alpha characters will break up phonemes.
97 : ****************************************************************************/
98 :
99 :
100 : /* I add modifications to the traditional metaphone algorithm that you
101 : might find in books. Define this if you want metaphone to behave
102 : traditionally */
103 : #undef USE_TRADITIONAL_METAPHONE
104 :
105 : /* Special encodings */
106 : #define SH 'X'
107 : #define TH '0'
108 :
109 : static char Lookahead(char *word, int how_far);
110 : static void _metaphone(char *word, int max_phonemes, char **phoned_word);
111 :
112 : /* Metachar.h ... little bits about characters for metaphone */
113 :
114 :
115 : /*-- Character encoding array & accessing macros --*/
116 : /* Stolen directly out of the book... */
117 : static const char _codes[26] = {
118 : 1, 16, 4, 16, 9, 2, 4, 16, 9, 2, 0, 2, 2, 2, 1, 4, 0, 2, 4, 4, 1, 0, 0, 0, 8, 0
119 : /* a b c d e f g h i j k l m n o p q r s t u v w x y z */
120 : };
121 :
122 : static int
123 2 : getcode(char c)
124 : {
125 2 : if (isalpha((unsigned char) c))
126 : {
127 2 : c = toupper((unsigned char) c);
128 : /* Defend against non-ASCII letters */
129 2 : if (c >= 'A' && c <= 'Z')
130 2 : return _codes[c - 'A'];
131 : }
132 0 : return 0;
133 : }
134 :
135 : #define isvowel(c) (getcode(c) & 1) /* AEIOU */
136 :
137 : /* These letters are passed through unchanged */
138 : #define NOCHANGE(c) (getcode(c) & 2) /* FJMNR */
139 :
140 : /* These form diphthongs when preceding H */
141 : #define AFFECTH(c) (getcode(c) & 4) /* CGPST */
142 :
143 : /* These make C and G soft */
144 : #define MAKESOFT(c) (getcode(c) & 8) /* EIY */
145 :
146 : /* These prevent GH from becoming F */
147 : #define NOGHTOF(c) (getcode(c) & 16) /* BDH */
148 :
149 4 : PG_FUNCTION_INFO_V1(levenshtein_with_costs);
150 : Datum
151 2 : levenshtein_with_costs(PG_FUNCTION_ARGS)
152 : {
153 2 : text *src = PG_GETARG_TEXT_PP(0);
154 2 : text *dst = PG_GETARG_TEXT_PP(1);
155 2 : int ins_c = PG_GETARG_INT32(2);
156 2 : int del_c = PG_GETARG_INT32(3);
157 2 : int sub_c = PG_GETARG_INT32(4);
158 : const char *s_data;
159 : const char *t_data;
160 : int s_bytes,
161 : t_bytes;
162 :
163 : /* Extract a pointer to the actual character data */
164 2 : s_data = VARDATA_ANY(src);
165 2 : t_data = VARDATA_ANY(dst);
166 : /* Determine length of each string in bytes */
167 2 : s_bytes = VARSIZE_ANY_EXHDR(src);
168 2 : t_bytes = VARSIZE_ANY_EXHDR(dst);
169 :
170 2 : PG_RETURN_INT32(varstr_levenshtein(s_data, s_bytes, t_data, t_bytes,
171 : ins_c, del_c, sub_c, false));
172 : }
173 :
174 :
175 4 : PG_FUNCTION_INFO_V1(levenshtein);
176 : Datum
177 2 : levenshtein(PG_FUNCTION_ARGS)
178 : {
179 2 : text *src = PG_GETARG_TEXT_PP(0);
180 2 : text *dst = PG_GETARG_TEXT_PP(1);
181 : const char *s_data;
182 : const char *t_data;
183 : int s_bytes,
184 : t_bytes;
185 :
186 : /* Extract a pointer to the actual character data */
187 2 : s_data = VARDATA_ANY(src);
188 2 : t_data = VARDATA_ANY(dst);
189 : /* Determine length of each string in bytes */
190 2 : s_bytes = VARSIZE_ANY_EXHDR(src);
191 2 : t_bytes = VARSIZE_ANY_EXHDR(dst);
192 :
193 2 : PG_RETURN_INT32(varstr_levenshtein(s_data, s_bytes, t_data, t_bytes,
194 : 1, 1, 1, false));
195 : }
196 :
197 :
198 2 : PG_FUNCTION_INFO_V1(levenshtein_less_equal_with_costs);
199 : Datum
200 0 : levenshtein_less_equal_with_costs(PG_FUNCTION_ARGS)
201 : {
202 0 : text *src = PG_GETARG_TEXT_PP(0);
203 0 : text *dst = PG_GETARG_TEXT_PP(1);
204 0 : int ins_c = PG_GETARG_INT32(2);
205 0 : int del_c = PG_GETARG_INT32(3);
206 0 : int sub_c = PG_GETARG_INT32(4);
207 0 : int max_d = PG_GETARG_INT32(5);
208 : const char *s_data;
209 : const char *t_data;
210 : int s_bytes,
211 : t_bytes;
212 :
213 : /* Extract a pointer to the actual character data */
214 0 : s_data = VARDATA_ANY(src);
215 0 : t_data = VARDATA_ANY(dst);
216 : /* Determine length of each string in bytes */
217 0 : s_bytes = VARSIZE_ANY_EXHDR(src);
218 0 : t_bytes = VARSIZE_ANY_EXHDR(dst);
219 :
220 0 : PG_RETURN_INT32(varstr_levenshtein_less_equal(s_data, s_bytes,
221 : t_data, t_bytes,
222 : ins_c, del_c, sub_c,
223 : max_d, false));
224 : }
225 :
226 :
227 4 : PG_FUNCTION_INFO_V1(levenshtein_less_equal);
228 : Datum
229 4 : levenshtein_less_equal(PG_FUNCTION_ARGS)
230 : {
231 4 : text *src = PG_GETARG_TEXT_PP(0);
232 4 : text *dst = PG_GETARG_TEXT_PP(1);
233 4 : int max_d = PG_GETARG_INT32(2);
234 : const char *s_data;
235 : const char *t_data;
236 : int s_bytes,
237 : t_bytes;
238 :
239 : /* Extract a pointer to the actual character data */
240 4 : s_data = VARDATA_ANY(src);
241 4 : t_data = VARDATA_ANY(dst);
242 : /* Determine length of each string in bytes */
243 4 : s_bytes = VARSIZE_ANY_EXHDR(src);
244 4 : t_bytes = VARSIZE_ANY_EXHDR(dst);
245 :
246 4 : PG_RETURN_INT32(varstr_levenshtein_less_equal(s_data, s_bytes,
247 : t_data, t_bytes,
248 : 1, 1, 1,
249 : max_d, false));
250 : }
251 :
252 :
253 : /*
254 : * Calculates the metaphone of an input string.
255 : * Returns number of characters requested
256 : * (suggested value is 4)
257 : */
258 4 : PG_FUNCTION_INFO_V1(metaphone);
259 : Datum
260 2 : metaphone(PG_FUNCTION_ARGS)
261 : {
262 2 : char *str_i = TextDatumGetCString(PG_GETARG_DATUM(0));
263 2 : size_t str_i_len = strlen(str_i);
264 : int reqlen;
265 : char *metaph;
266 :
267 : /* return an empty string if we receive one */
268 2 : if (!(str_i_len > 0))
269 0 : PG_RETURN_TEXT_P(cstring_to_text(""));
270 :
271 2 : if (str_i_len > MAX_METAPHONE_STRLEN)
272 0 : ereport(ERROR,
273 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
274 : errmsg("argument exceeds the maximum length of %d bytes",
275 : MAX_METAPHONE_STRLEN)));
276 :
277 2 : reqlen = PG_GETARG_INT32(1);
278 2 : if (reqlen > MAX_METAPHONE_STRLEN)
279 0 : ereport(ERROR,
280 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
281 : errmsg("output exceeds the maximum length of %d bytes",
282 : MAX_METAPHONE_STRLEN)));
283 :
284 2 : if (!(reqlen > 0))
285 0 : ereport(ERROR,
286 : (errcode(ERRCODE_ZERO_LENGTH_CHARACTER_STRING),
287 : errmsg("output cannot be empty string")));
288 :
289 2 : _metaphone(str_i, reqlen, &metaph);
290 2 : PG_RETURN_TEXT_P(cstring_to_text(metaph));
291 : }
292 :
293 :
294 : /*
295 : * Original code by Michael G Schwern starts here.
296 : * Code slightly modified for use as PostgreSQL
297 : * function (palloc, etc).
298 : */
299 :
300 : /* I suppose I could have been using a character pointer instead of
301 : * accessing the array directly... */
302 :
303 : /* Look at the next letter in the word */
304 : #define Next_Letter (toupper((unsigned char) word[w_idx+1]))
305 : /* Look at the current letter in the word */
306 : #define Curr_Letter (toupper((unsigned char) word[w_idx]))
307 : /* Go N letters back. */
308 : #define Look_Back_Letter(n) \
309 : (w_idx >= (n) ? toupper((unsigned char) word[w_idx-(n)]) : '\0')
310 : /* Previous letter. I dunno, should this return null on failure? */
311 : #define Prev_Letter (Look_Back_Letter(1))
312 : /* Look two letters down. It makes sure you don't walk off the string. */
313 : #define After_Next_Letter \
314 : (Next_Letter != '\0' ? toupper((unsigned char) word[w_idx+2]) : '\0')
315 : #define Look_Ahead_Letter(n) toupper((unsigned char) Lookahead(word+w_idx, n))
316 :
317 :
318 : /* Allows us to safely look ahead an arbitrary # of letters */
319 : /* I probably could have just used strlen... */
320 : static char
321 0 : Lookahead(char *word, int how_far)
322 : {
323 0 : char letter_ahead = '\0'; /* null by default */
324 : int idx;
325 :
326 0 : for (idx = 0; word[idx] != '\0' && idx < how_far; idx++);
327 : /* Edge forward in the string... */
328 :
329 0 : letter_ahead = word[idx]; /* idx will be either == to how_far or at the
330 : * end of the string */
331 0 : return letter_ahead;
332 : }
333 :
334 :
335 : /* phonize one letter */
336 : #define Phonize(c) do {(*phoned_word)[p_idx++] = c;} while (0)
337 : /* Slap a null character on the end of the phoned word */
338 : #define End_Phoned_Word do {(*phoned_word)[p_idx] = '\0';} while (0)
339 : /* How long is the phoned word? */
340 : #define Phone_Len (p_idx)
341 :
342 : /* Note is a letter is a 'break' in the word */
343 : #define Isbreak(c) (!isalpha((unsigned char) (c)))
344 :
345 :
346 : static void
347 2 : _metaphone(char *word, /* IN */
348 : int max_phonemes,
349 : char **phoned_word) /* OUT */
350 : {
351 2 : int w_idx = 0; /* point in the phonization we're at. */
352 2 : int p_idx = 0; /* end of the phoned phrase */
353 :
354 : /*-- Parameter checks --*/
355 :
356 : /*
357 : * Shouldn't be necessary, but left these here anyway jec Aug 3, 2001
358 : */
359 :
360 : /* Negative phoneme length is meaningless */
361 2 : if (!(max_phonemes > 0))
362 : /* internal error */
363 0 : elog(ERROR, "metaphone: Requested output length must be > 0");
364 :
365 : /* Empty/null string is meaningless */
366 2 : if ((word == NULL) || !(strlen(word) > 0))
367 : /* internal error */
368 0 : elog(ERROR, "metaphone: Input string length must be > 0");
369 :
370 : /*-- Allocate memory for our phoned_phrase --*/
371 2 : if (max_phonemes == 0)
372 : { /* Assume largest possible */
373 0 : *phoned_word = palloc(sizeof(char) * strlen(word) + 1);
374 : }
375 : else
376 : {
377 2 : *phoned_word = palloc(sizeof(char) * max_phonemes + 1);
378 : }
379 :
380 : /*-- The first phoneme has to be processed specially. --*/
381 : /* Find our first letter */
382 2 : for (; !isalpha((unsigned char) (Curr_Letter)); w_idx++)
383 : {
384 : /* On the off chance we were given nothing but crap... */
385 0 : if (Curr_Letter == '\0')
386 : {
387 0 : End_Phoned_Word;
388 0 : return;
389 : }
390 : }
391 :
392 2 : switch (Curr_Letter)
393 : {
394 : /* AE becomes E */
395 0 : case 'A':
396 0 : if (Next_Letter == 'E')
397 : {
398 0 : Phonize('E');
399 0 : w_idx += 2;
400 : }
401 : /* Remember, preserve vowels at the beginning */
402 : else
403 : {
404 0 : Phonize('A');
405 0 : w_idx++;
406 : }
407 0 : break;
408 : /* [GKP]N becomes N */
409 2 : case 'G':
410 : case 'K':
411 : case 'P':
412 2 : if (Next_Letter == 'N')
413 : {
414 0 : Phonize('N');
415 0 : w_idx += 2;
416 : }
417 2 : break;
418 :
419 : /*
420 : * WH becomes H, WR becomes R W if followed by a vowel
421 : */
422 0 : case 'W':
423 0 : if (Next_Letter == 'H' ||
424 0 : Next_Letter == 'R')
425 : {
426 0 : Phonize(Next_Letter);
427 0 : w_idx += 2;
428 : }
429 0 : else if (isvowel(Next_Letter))
430 : {
431 0 : Phonize('W');
432 0 : w_idx += 2;
433 : }
434 : /* else ignore */
435 0 : break;
436 : /* X becomes S */
437 0 : case 'X':
438 0 : Phonize('S');
439 0 : w_idx++;
440 0 : break;
441 : /* Vowels are kept */
442 :
443 : /*
444 : * We did A already case 'A': case 'a':
445 : */
446 0 : case 'E':
447 : case 'I':
448 : case 'O':
449 : case 'U':
450 0 : Phonize(Curr_Letter);
451 0 : w_idx++;
452 0 : break;
453 0 : default:
454 : /* do nothing */
455 0 : break;
456 : }
457 :
458 :
459 :
460 : /* On to the metaphoning */
461 12 : for (; Curr_Letter != '\0' &&
462 10 : (max_phonemes == 0 || Phone_Len < max_phonemes);
463 10 : w_idx++)
464 : {
465 : /*
466 : * How many letters to skip because an earlier encoding handled
467 : * multiple letters
468 : */
469 10 : unsigned short int skip_letter = 0;
470 :
471 :
472 : /*
473 : * THOUGHT: It would be nice if, rather than having things like...
474 : * well, SCI. For SCI you encode the S, then have to remember to skip
475 : * the C. So the phonome SCI invades both S and C. It would be
476 : * better, IMHO, to skip the C from the S part of the encoding. Hell,
477 : * I'm trying it.
478 : */
479 :
480 : /* Ignore non-alphas */
481 10 : if (!isalpha((unsigned char) (Curr_Letter)))
482 0 : continue;
483 :
484 : /* Drop duplicates, except CC */
485 10 : if (Curr_Letter == Prev_Letter &&
486 0 : Curr_Letter != 'C')
487 0 : continue;
488 :
489 10 : switch (Curr_Letter)
490 : {
491 : /* B -> B unless in MB */
492 2 : case 'B':
493 2 : if (Prev_Letter != 'M')
494 0 : Phonize('B');
495 2 : break;
496 :
497 : /*
498 : * 'sh' if -CIA- or -CH, but not SCH, except SCHW. (SCHW is
499 : * handled in S) S if -CI-, -CE- or -CY- dropped if -SCI-,
500 : * SCE-, -SCY- (handed in S) else K
501 : */
502 0 : case 'C':
503 0 : if (MAKESOFT(Next_Letter))
504 : { /* C[IEY] */
505 0 : if (After_Next_Letter == 'A' &&
506 0 : Next_Letter == 'I')
507 : { /* CIA */
508 0 : Phonize(SH);
509 : }
510 : /* SC[IEY] */
511 0 : else if (Prev_Letter == 'S')
512 : {
513 : /* Dropped */
514 : }
515 : else
516 0 : Phonize('S');
517 : }
518 0 : else if (Next_Letter == 'H')
519 : {
520 : #ifndef USE_TRADITIONAL_METAPHONE
521 0 : if (After_Next_Letter == 'R' ||
522 0 : Prev_Letter == 'S')
523 : { /* Christ, School */
524 0 : Phonize('K');
525 : }
526 : else
527 0 : Phonize(SH);
528 : #else
529 : Phonize(SH);
530 : #endif
531 0 : skip_letter++;
532 : }
533 : else
534 0 : Phonize('K');
535 0 : break;
536 :
537 : /*
538 : * J if in -DGE-, -DGI- or -DGY- else T
539 : */
540 0 : case 'D':
541 0 : if (Next_Letter == 'G' &&
542 0 : MAKESOFT(After_Next_Letter))
543 : {
544 0 : Phonize('J');
545 0 : skip_letter++;
546 : }
547 : else
548 0 : Phonize('T');
549 0 : break;
550 :
551 : /*
552 : * F if in -GH and not B--GH, D--GH, -H--GH, -H---GH else
553 : * dropped if -GNED, -GN, else dropped if -DGE-, -DGI- or
554 : * -DGY- (handled in D) else J if in -GE-, -GI, -GY and not GG
555 : * else K
556 : */
557 2 : case 'G':
558 2 : if (Next_Letter == 'H')
559 : {
560 0 : if (!(NOGHTOF(Look_Back_Letter(3)) ||
561 0 : Look_Back_Letter(4) == 'H'))
562 : {
563 0 : Phonize('F');
564 0 : skip_letter++;
565 : }
566 : else
567 : {
568 : /* silent */
569 : }
570 : }
571 2 : else if (Next_Letter == 'N')
572 : {
573 0 : if (Isbreak(After_Next_Letter) ||
574 0 : (After_Next_Letter == 'E' &&
575 0 : Look_Ahead_Letter(3) == 'D'))
576 : {
577 : /* dropped */
578 : }
579 : else
580 0 : Phonize('K');
581 : }
582 2 : else if (MAKESOFT(Next_Letter) &&
583 0 : Prev_Letter != 'G')
584 0 : Phonize('J');
585 : else
586 2 : Phonize('K');
587 2 : break;
588 : /* H if before a vowel and not after C,G,P,S,T */
589 0 : case 'H':
590 0 : if (isvowel(Next_Letter) &&
591 0 : !AFFECTH(Prev_Letter))
592 0 : Phonize('H');
593 0 : break;
594 :
595 : /*
596 : * dropped if after C else K
597 : */
598 0 : case 'K':
599 0 : if (Prev_Letter != 'C')
600 0 : Phonize('K');
601 0 : break;
602 :
603 : /*
604 : * F if before H else P
605 : */
606 0 : case 'P':
607 0 : if (Next_Letter == 'H')
608 0 : Phonize('F');
609 : else
610 0 : Phonize('P');
611 0 : break;
612 :
613 : /*
614 : * K
615 : */
616 0 : case 'Q':
617 0 : Phonize('K');
618 0 : break;
619 :
620 : /*
621 : * 'sh' in -SH-, -SIO- or -SIA- or -SCHW- else S
622 : */
623 0 : case 'S':
624 0 : if (Next_Letter == 'I' &&
625 0 : (After_Next_Letter == 'O' ||
626 0 : After_Next_Letter == 'A'))
627 0 : Phonize(SH);
628 0 : else if (Next_Letter == 'H')
629 : {
630 0 : Phonize(SH);
631 0 : skip_letter++;
632 : }
633 : #ifndef USE_TRADITIONAL_METAPHONE
634 0 : else if (Next_Letter == 'C' &&
635 0 : Look_Ahead_Letter(2) == 'H' &&
636 0 : Look_Ahead_Letter(3) == 'W')
637 : {
638 0 : Phonize(SH);
639 0 : skip_letter += 2;
640 : }
641 : #endif
642 : else
643 0 : Phonize('S');
644 0 : break;
645 :
646 : /*
647 : * 'sh' in -TIA- or -TIO- else 'th' before H else T
648 : */
649 0 : case 'T':
650 0 : if (Next_Letter == 'I' &&
651 0 : (After_Next_Letter == 'O' ||
652 0 : After_Next_Letter == 'A'))
653 0 : Phonize(SH);
654 0 : else if (Next_Letter == 'H')
655 : {
656 0 : Phonize(TH);
657 0 : skip_letter++;
658 : }
659 : else
660 0 : Phonize('T');
661 0 : break;
662 : /* F */
663 0 : case 'V':
664 0 : Phonize('F');
665 0 : break;
666 : /* W before a vowel, else dropped */
667 0 : case 'W':
668 0 : if (isvowel(Next_Letter))
669 0 : Phonize('W');
670 0 : break;
671 : /* KS */
672 0 : case 'X':
673 0 : Phonize('K');
674 0 : if (max_phonemes == 0 || Phone_Len < max_phonemes)
675 0 : Phonize('S');
676 0 : break;
677 : /* Y if followed by a vowel */
678 0 : case 'Y':
679 0 : if (isvowel(Next_Letter))
680 0 : Phonize('Y');
681 0 : break;
682 : /* S */
683 0 : case 'Z':
684 0 : Phonize('S');
685 0 : break;
686 : /* No transformation */
687 2 : case 'F':
688 : case 'J':
689 : case 'L':
690 : case 'M':
691 : case 'N':
692 : case 'R':
693 2 : Phonize(Curr_Letter);
694 2 : break;
695 4 : default:
696 : /* nothing */
697 4 : break;
698 : } /* END SWITCH */
699 :
700 10 : w_idx += skip_letter;
701 : } /* END FOR */
702 :
703 2 : End_Phoned_Word;
704 : } /* END metaphone */
705 :
706 :
707 : /*
708 : * SQL function: soundex(text) returns text
709 : */
710 6 : PG_FUNCTION_INFO_V1(soundex);
711 :
712 : Datum
713 16 : soundex(PG_FUNCTION_ARGS)
714 : {
715 : char outstr[SOUNDEX_LEN + 1];
716 : char *arg;
717 :
718 16 : arg = text_to_cstring(PG_GETARG_TEXT_PP(0));
719 :
720 16 : _soundex(arg, outstr);
721 :
722 16 : PG_RETURN_TEXT_P(cstring_to_text(outstr));
723 : }
724 :
725 : static void
726 32 : _soundex(const char *instr, char *outstr)
727 : {
728 : int count;
729 :
730 : Assert(instr);
731 : Assert(outstr);
732 :
733 : /* Skip leading non-alphabetic characters */
734 32 : while (*instr && !isalpha((unsigned char) *instr))
735 0 : ++instr;
736 :
737 : /* If no string left, return all-zeroes buffer */
738 32 : if (!*instr)
739 : {
740 6 : memset(outstr, '\0', SOUNDEX_LEN + 1);
741 6 : return;
742 : }
743 :
744 : /* Take the first letter as is */
745 26 : *outstr++ = (char) toupper((unsigned char) *instr++);
746 :
747 26 : count = 1;
748 120 : while (*instr && count < SOUNDEX_LEN)
749 : {
750 186 : if (isalpha((unsigned char) *instr) &&
751 92 : soundex_code(*instr) != soundex_code(*(instr - 1)))
752 : {
753 70 : *outstr = soundex_code(*instr);
754 70 : if (*outstr != '0')
755 : {
756 46 : ++outstr;
757 46 : ++count;
758 : }
759 : }
760 94 : ++instr;
761 : }
762 :
763 : /* Fill with 0's */
764 58 : while (count < SOUNDEX_LEN)
765 : {
766 32 : *outstr = '0';
767 32 : ++outstr;
768 32 : ++count;
769 : }
770 :
771 : /* And null-terminate */
772 26 : *outstr = '\0';
773 : }
774 :
775 4 : PG_FUNCTION_INFO_V1(difference);
776 :
777 : Datum
778 8 : difference(PG_FUNCTION_ARGS)
779 : {
780 : char sndx1[SOUNDEX_LEN + 1],
781 : sndx2[SOUNDEX_LEN + 1];
782 : int i,
783 : result;
784 :
785 8 : _soundex(text_to_cstring(PG_GETARG_TEXT_PP(0)), sndx1);
786 8 : _soundex(text_to_cstring(PG_GETARG_TEXT_PP(1)), sndx2);
787 :
788 8 : result = 0;
789 40 : for (i = 0; i < SOUNDEX_LEN; i++)
790 : {
791 32 : if (sndx1[i] == sndx2[i])
792 20 : result++;
793 : }
794 :
795 8 : PG_RETURN_INT32(result);
796 : }
|
