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) 2001-2026, 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 2 : 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
63 127 : soundex_code(char letter)
64 : {
65 127 : letter = pg_ascii_toupper((unsigned char) letter);
66 : /* Defend against non-ASCII letters */
67 127 : if (letter >= 'A' && letter <= 'Z')
68 126 : return soundex_table[letter - 'A'];
69 1 : 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 : /*
101 : * I add modifications to the traditional metaphone algorithm that you
102 : * might find in books. Define this if you want metaphone to behave
103 : * traditionally
104 : */
105 : #undef USE_TRADITIONAL_METAPHONE
106 :
107 : /* Special encodings */
108 : #define SH 'X'
109 : #define TH '0'
110 :
111 : static char Lookahead(char *word, int how_far);
112 : static void _metaphone(char *word, int max_phonemes, char **phoned_word);
113 :
114 : /* Metachar.h ... little bits about characters for metaphone */
115 :
116 :
117 : /*-- Character encoding array & accessing macros --*/
118 : /* Stolen directly out of the book... */
119 : static const char _codes[26] = {
120 : 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
121 : /* 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 */
122 : };
123 :
124 : static int
125 1 : getcode(char c)
126 : {
127 1 : c = pg_ascii_toupper((unsigned char) c);
128 : /* Defend against non-ASCII letters */
129 1 : if (c >= 'A' && c <= 'Z')
130 1 : return _codes[c - 'A'];
131 :
132 0 : return 0;
133 : }
134 :
135 : static bool
136 66 : ascii_isalpha(char c)
137 : {
138 113 : return (c >= 'A' && c <= 'Z') ||
139 47 : (c >= 'a' && c <= 'z');
140 : }
141 :
142 : #define isvowel(c) (getcode(c) & 1) /* AEIOU */
143 :
144 : /* These letters are passed through unchanged */
145 : #define NOCHANGE(c) (getcode(c) & 2) /* FJMNR */
146 :
147 : /* These form diphthongs when preceding H */
148 : #define AFFECTH(c) (getcode(c) & 4) /* CGPST */
149 :
150 : /* These make C and G soft */
151 : #define MAKESOFT(c) (getcode(c) & 8) /* EIY */
152 :
153 : /* These prevent GH from becoming F */
154 : #define NOGHTOF(c) (getcode(c) & 16) /* BDH */
155 :
156 2 : PG_FUNCTION_INFO_V1(levenshtein_with_costs);
157 : Datum
158 1 : levenshtein_with_costs(PG_FUNCTION_ARGS)
159 : {
160 1 : text *src = PG_GETARG_TEXT_PP(0);
161 1 : text *dst = PG_GETARG_TEXT_PP(1);
162 1 : int ins_c = PG_GETARG_INT32(2);
163 1 : int del_c = PG_GETARG_INT32(3);
164 1 : int sub_c = PG_GETARG_INT32(4);
165 : const char *s_data;
166 : const char *t_data;
167 : int s_bytes,
168 : t_bytes;
169 :
170 : /* Extract a pointer to the actual character data */
171 1 : s_data = VARDATA_ANY(src);
172 1 : t_data = VARDATA_ANY(dst);
173 : /* Determine length of each string in bytes */
174 1 : s_bytes = VARSIZE_ANY_EXHDR(src);
175 1 : t_bytes = VARSIZE_ANY_EXHDR(dst);
176 :
177 1 : PG_RETURN_INT32(varstr_levenshtein(s_data, s_bytes, t_data, t_bytes,
178 : ins_c, del_c, sub_c, false));
179 : }
180 :
181 :
182 2 : PG_FUNCTION_INFO_V1(levenshtein);
183 : Datum
184 1 : levenshtein(PG_FUNCTION_ARGS)
185 : {
186 1 : text *src = PG_GETARG_TEXT_PP(0);
187 1 : text *dst = PG_GETARG_TEXT_PP(1);
188 : const char *s_data;
189 : const char *t_data;
190 : int s_bytes,
191 : t_bytes;
192 :
193 : /* Extract a pointer to the actual character data */
194 1 : s_data = VARDATA_ANY(src);
195 1 : t_data = VARDATA_ANY(dst);
196 : /* Determine length of each string in bytes */
197 1 : s_bytes = VARSIZE_ANY_EXHDR(src);
198 1 : t_bytes = VARSIZE_ANY_EXHDR(dst);
199 :
200 1 : PG_RETURN_INT32(varstr_levenshtein(s_data, s_bytes, t_data, t_bytes,
201 : 1, 1, 1, false));
202 : }
203 :
204 :
205 1 : PG_FUNCTION_INFO_V1(levenshtein_less_equal_with_costs);
206 : Datum
207 0 : levenshtein_less_equal_with_costs(PG_FUNCTION_ARGS)
208 : {
209 0 : text *src = PG_GETARG_TEXT_PP(0);
210 0 : text *dst = PG_GETARG_TEXT_PP(1);
211 0 : int ins_c = PG_GETARG_INT32(2);
212 0 : int del_c = PG_GETARG_INT32(3);
213 0 : int sub_c = PG_GETARG_INT32(4);
214 0 : int max_d = PG_GETARG_INT32(5);
215 : const char *s_data;
216 : const char *t_data;
217 : int s_bytes,
218 : t_bytes;
219 :
220 : /* Extract a pointer to the actual character data */
221 0 : s_data = VARDATA_ANY(src);
222 0 : t_data = VARDATA_ANY(dst);
223 : /* Determine length of each string in bytes */
224 0 : s_bytes = VARSIZE_ANY_EXHDR(src);
225 0 : t_bytes = VARSIZE_ANY_EXHDR(dst);
226 :
227 0 : PG_RETURN_INT32(varstr_levenshtein_less_equal(s_data, s_bytes,
228 : t_data, t_bytes,
229 : ins_c, del_c, sub_c,
230 : max_d, false));
231 : }
232 :
233 :
234 2 : PG_FUNCTION_INFO_V1(levenshtein_less_equal);
235 : Datum
236 2 : levenshtein_less_equal(PG_FUNCTION_ARGS)
237 : {
238 2 : text *src = PG_GETARG_TEXT_PP(0);
239 2 : text *dst = PG_GETARG_TEXT_PP(1);
240 2 : int max_d = PG_GETARG_INT32(2);
241 : const char *s_data;
242 : const char *t_data;
243 : int s_bytes,
244 : t_bytes;
245 :
246 : /* Extract a pointer to the actual character data */
247 2 : s_data = VARDATA_ANY(src);
248 2 : t_data = VARDATA_ANY(dst);
249 : /* Determine length of each string in bytes */
250 2 : s_bytes = VARSIZE_ANY_EXHDR(src);
251 2 : t_bytes = VARSIZE_ANY_EXHDR(dst);
252 :
253 2 : PG_RETURN_INT32(varstr_levenshtein_less_equal(s_data, s_bytes,
254 : t_data, t_bytes,
255 : 1, 1, 1,
256 : max_d, false));
257 : }
258 :
259 :
260 : /*
261 : * Calculates the metaphone of an input string.
262 : * Returns number of characters requested
263 : * (suggested value is 4)
264 : */
265 2 : PG_FUNCTION_INFO_V1(metaphone);
266 : Datum
267 1 : metaphone(PG_FUNCTION_ARGS)
268 : {
269 1 : char *str_i = TextDatumGetCString(PG_GETARG_DATUM(0));
270 1 : size_t str_i_len = strlen(str_i);
271 : int reqlen;
272 : char *metaph;
273 :
274 : /* return an empty string if we receive one */
275 1 : if (!(str_i_len > 0))
276 0 : PG_RETURN_TEXT_P(cstring_to_text(""));
277 :
278 1 : if (str_i_len > MAX_METAPHONE_STRLEN)
279 0 : ereport(ERROR,
280 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
281 : errmsg("argument exceeds the maximum length of %d bytes",
282 : MAX_METAPHONE_STRLEN)));
283 :
284 1 : reqlen = PG_GETARG_INT32(1);
285 1 : if (reqlen > MAX_METAPHONE_STRLEN)
286 0 : ereport(ERROR,
287 : (errcode(ERRCODE_INVALID_PARAMETER_VALUE),
288 : errmsg("output exceeds the maximum length of %d bytes",
289 : MAX_METAPHONE_STRLEN)));
290 :
291 1 : if (!(reqlen > 0))
292 0 : ereport(ERROR,
293 : (errcode(ERRCODE_ZERO_LENGTH_CHARACTER_STRING),
294 : errmsg("output cannot be empty string")));
295 :
296 1 : _metaphone(str_i, reqlen, &metaph);
297 1 : PG_RETURN_TEXT_P(cstring_to_text(metaph));
298 : }
299 :
300 :
301 : /*
302 : * Original code by Michael G Schwern starts here.
303 : * Code slightly modified for use as PostgreSQL
304 : * function (palloc, etc).
305 : */
306 :
307 : /*
308 : * I suppose I could have been using a character pointer instead of
309 : * accessing the array directly...
310 : */
311 :
312 : /* Look at the next letter in the word */
313 : #define Next_Letter (pg_ascii_toupper((unsigned char) word[w_idx+1]))
314 : /* Look at the current letter in the word */
315 : #define Curr_Letter (pg_ascii_toupper((unsigned char) word[w_idx]))
316 : /* Go N letters back. */
317 : #define Look_Back_Letter(n) \
318 : (w_idx >= (n) ? pg_ascii_toupper((unsigned char) word[w_idx-(n)]) : '\0')
319 : /* Previous letter. I dunno, should this return null on failure? */
320 : #define Prev_Letter (Look_Back_Letter(1))
321 : /* Look two letters down. It makes sure you don't walk off the string. */
322 : #define After_Next_Letter \
323 : (Next_Letter != '\0' ? pg_ascii_toupper((unsigned char) word[w_idx+2]) : '\0')
324 : #define Look_Ahead_Letter(n) pg_ascii_toupper((unsigned char) Lookahead(word+w_idx, n))
325 :
326 :
327 : /* Allows us to safely look ahead an arbitrary # of letters */
328 : /* I probably could have just used strlen... */
329 : static char
330 0 : Lookahead(char *word, int how_far)
331 : {
332 0 : char letter_ahead = '\0'; /* null by default */
333 : int idx;
334 :
335 0 : for (idx = 0; word[idx] != '\0' && idx < how_far; idx++);
336 : /* Edge forward in the string... */
337 :
338 0 : letter_ahead = word[idx]; /* idx will be either == to how_far or at the
339 : * end of the string */
340 0 : return letter_ahead;
341 : }
342 :
343 :
344 : /* phonize one letter */
345 : #define Phonize(c) do {(*phoned_word)[p_idx++] = c;} while (0)
346 : /* Slap a null character on the end of the phoned word */
347 : #define End_Phoned_Word do {(*phoned_word)[p_idx] = '\0';} while (0)
348 : /* How long is the phoned word? */
349 : #define Phone_Len (p_idx)
350 :
351 : /* Note is a letter is a 'break' in the word */
352 : #define Isbreak(c) (!ascii_isalpha((unsigned char) (c)))
353 :
354 :
355 : static void
356 1 : _metaphone(char *word, /* IN */
357 : int max_phonemes,
358 : char **phoned_word) /* OUT */
359 : {
360 1 : int w_idx = 0; /* point in the phonization we're at. */
361 1 : int p_idx = 0; /* end of the phoned phrase */
362 :
363 : /*-- Parameter checks --*/
364 :
365 : /*
366 : * Shouldn't be necessary, but left these here anyway jec Aug 3, 2001
367 : */
368 :
369 : /* Negative phoneme length is meaningless */
370 1 : if (!(max_phonemes > 0))
371 : /* internal error */
372 0 : elog(ERROR, "metaphone: Requested output length must be > 0");
373 :
374 : /* Empty/null string is meaningless */
375 1 : if ((word == NULL) || !(strlen(word) > 0))
376 : /* internal error */
377 0 : elog(ERROR, "metaphone: Input string length must be > 0");
378 :
379 : /*-- Allocate memory for our phoned_phrase --*/
380 1 : if (max_phonemes == 0)
381 : { /* Assume largest possible */
382 0 : *phoned_word = palloc(sizeof(char) * strlen(word) + 1);
383 : }
384 : else
385 : {
386 1 : *phoned_word = palloc(sizeof(char) * max_phonemes + 1);
387 : }
388 :
389 : /*-- The first phoneme has to be processed specially. --*/
390 : /* Find our first letter */
391 1 : for (; !ascii_isalpha((unsigned char) (Curr_Letter)); w_idx++)
392 : {
393 : /* On the off chance we were given nothing but crap... */
394 0 : if (Curr_Letter == '\0')
395 : {
396 0 : End_Phoned_Word;
397 0 : return;
398 : }
399 : }
400 :
401 1 : switch (Curr_Letter)
402 : {
403 : /* AE becomes E */
404 0 : case 'A':
405 0 : if (Next_Letter == 'E')
406 : {
407 0 : Phonize('E');
408 0 : w_idx += 2;
409 : }
410 : /* Remember, preserve vowels at the beginning */
411 : else
412 : {
413 0 : Phonize('A');
414 0 : w_idx++;
415 : }
416 0 : break;
417 : /* [GKP]N becomes N */
418 1 : case 'G':
419 : case 'K':
420 : case 'P':
421 1 : if (Next_Letter == 'N')
422 : {
423 0 : Phonize('N');
424 0 : w_idx += 2;
425 : }
426 1 : break;
427 :
428 : /*
429 : * WH becomes H, WR becomes R W if followed by a vowel
430 : */
431 0 : case 'W':
432 0 : if (Next_Letter == 'H' ||
433 0 : Next_Letter == 'R')
434 : {
435 0 : Phonize(Next_Letter);
436 0 : w_idx += 2;
437 : }
438 0 : else if (isvowel(Next_Letter))
439 : {
440 0 : Phonize('W');
441 0 : w_idx += 2;
442 : }
443 : /* else ignore */
444 0 : break;
445 : /* X becomes S */
446 0 : case 'X':
447 0 : Phonize('S');
448 0 : w_idx++;
449 0 : break;
450 : /* Vowels are kept */
451 :
452 : /*
453 : * We did A already case 'A': case 'a':
454 : */
455 0 : case 'E':
456 : case 'I':
457 : case 'O':
458 : case 'U':
459 0 : Phonize(Curr_Letter);
460 0 : w_idx++;
461 0 : break;
462 0 : default:
463 : /* do nothing */
464 0 : break;
465 : }
466 :
467 :
468 :
469 : /* On to the metaphoning */
470 6 : for (; Curr_Letter != '\0' &&
471 5 : (max_phonemes == 0 || Phone_Len < max_phonemes);
472 5 : w_idx++)
473 : {
474 : /*
475 : * How many letters to skip because an earlier encoding handled
476 : * multiple letters
477 : */
478 5 : unsigned short int skip_letter = 0;
479 :
480 :
481 : /*
482 : * THOUGHT: It would be nice if, rather than having things like...
483 : * well, SCI. For SCI you encode the S, then have to remember to skip
484 : * the C. So the phonome SCI invades both S and C. It would be
485 : * better, IMHO, to skip the C from the S part of the encoding. Hell,
486 : * I'm trying it.
487 : */
488 :
489 : /* Ignore non-alphas */
490 5 : if (!ascii_isalpha((unsigned char) (Curr_Letter)))
491 0 : continue;
492 :
493 : /* Drop duplicates, except CC */
494 5 : if (Curr_Letter == Prev_Letter &&
495 0 : Curr_Letter != 'C')
496 0 : continue;
497 :
498 5 : switch (Curr_Letter)
499 : {
500 : /* B -> B unless in MB */
501 1 : case 'B':
502 1 : if (Prev_Letter != 'M')
503 0 : Phonize('B');
504 1 : break;
505 :
506 : /*
507 : * 'sh' if -CIA- or -CH, but not SCH, except SCHW. (SCHW is
508 : * handled in S) S if -CI-, -CE- or -CY- dropped if -SCI-,
509 : * SCE-, -SCY- (handed in S) else K
510 : */
511 0 : case 'C':
512 0 : if (MAKESOFT(Next_Letter))
513 : { /* C[IEY] */
514 0 : if (After_Next_Letter == 'A' &&
515 0 : Next_Letter == 'I')
516 : { /* CIA */
517 0 : Phonize(SH);
518 : }
519 : /* SC[IEY] */
520 0 : else if (Prev_Letter == 'S')
521 : {
522 : /* Dropped */
523 : }
524 : else
525 0 : Phonize('S');
526 : }
527 0 : else if (Next_Letter == 'H')
528 : {
529 : #ifndef USE_TRADITIONAL_METAPHONE
530 0 : if (After_Next_Letter == 'R' ||
531 0 : Prev_Letter == 'S')
532 : { /* Christ, School */
533 0 : Phonize('K');
534 : }
535 : else
536 0 : Phonize(SH);
537 : #else
538 : Phonize(SH);
539 : #endif
540 0 : skip_letter++;
541 : }
542 : else
543 0 : Phonize('K');
544 0 : break;
545 :
546 : /*
547 : * J if in -DGE-, -DGI- or -DGY- else T
548 : */
549 0 : case 'D':
550 0 : if (Next_Letter == 'G' &&
551 0 : MAKESOFT(After_Next_Letter))
552 : {
553 0 : Phonize('J');
554 0 : skip_letter++;
555 : }
556 : else
557 0 : Phonize('T');
558 0 : break;
559 :
560 : /*
561 : * F if in -GH and not B--GH, D--GH, -H--GH, -H---GH else
562 : * dropped if -GNED, -GN, else dropped if -DGE-, -DGI- or
563 : * -DGY- (handled in D) else J if in -GE-, -GI, -GY and not GG
564 : * else K
565 : */
566 1 : case 'G':
567 1 : if (Next_Letter == 'H')
568 : {
569 0 : if (!(NOGHTOF(Look_Back_Letter(3)) ||
570 0 : Look_Back_Letter(4) == 'H'))
571 : {
572 0 : Phonize('F');
573 0 : skip_letter++;
574 : }
575 : else
576 : {
577 : /* silent */
578 : }
579 : }
580 1 : else if (Next_Letter == 'N')
581 : {
582 0 : if (Isbreak(After_Next_Letter) ||
583 0 : (After_Next_Letter == 'E' &&
584 0 : Look_Ahead_Letter(3) == 'D'))
585 : {
586 : /* dropped */
587 : }
588 : else
589 0 : Phonize('K');
590 : }
591 1 : else if (MAKESOFT(Next_Letter) &&
592 0 : Prev_Letter != 'G')
593 0 : Phonize('J');
594 : else
595 1 : Phonize('K');
596 1 : break;
597 : /* H if before a vowel and not after C,G,P,S,T */
598 0 : case 'H':
599 0 : if (isvowel(Next_Letter) &&
600 0 : !AFFECTH(Prev_Letter))
601 0 : Phonize('H');
602 0 : break;
603 :
604 : /*
605 : * dropped if after C else K
606 : */
607 0 : case 'K':
608 0 : if (Prev_Letter != 'C')
609 0 : Phonize('K');
610 0 : break;
611 :
612 : /*
613 : * F if before H else P
614 : */
615 0 : case 'P':
616 0 : if (Next_Letter == 'H')
617 0 : Phonize('F');
618 : else
619 0 : Phonize('P');
620 0 : break;
621 :
622 : /*
623 : * K
624 : */
625 0 : case 'Q':
626 0 : Phonize('K');
627 0 : break;
628 :
629 : /*
630 : * 'sh' in -SH-, -SIO- or -SIA- or -SCHW- else S
631 : */
632 0 : case 'S':
633 0 : if (Next_Letter == 'I' &&
634 0 : (After_Next_Letter == 'O' ||
635 0 : After_Next_Letter == 'A'))
636 0 : Phonize(SH);
637 0 : else if (Next_Letter == 'H')
638 : {
639 0 : Phonize(SH);
640 0 : skip_letter++;
641 : }
642 : #ifndef USE_TRADITIONAL_METAPHONE
643 0 : else if (Next_Letter == 'C' &&
644 0 : Look_Ahead_Letter(2) == 'H' &&
645 0 : Look_Ahead_Letter(3) == 'W')
646 : {
647 0 : Phonize(SH);
648 0 : skip_letter += 2;
649 : }
650 : #endif
651 : else
652 0 : Phonize('S');
653 0 : break;
654 :
655 : /*
656 : * 'sh' in -TIA- or -TIO- else 'th' before H else T
657 : */
658 0 : case 'T':
659 0 : if (Next_Letter == 'I' &&
660 0 : (After_Next_Letter == 'O' ||
661 0 : After_Next_Letter == 'A'))
662 0 : Phonize(SH);
663 0 : else if (Next_Letter == 'H')
664 : {
665 0 : Phonize(TH);
666 0 : skip_letter++;
667 : }
668 : else
669 0 : Phonize('T');
670 0 : break;
671 : /* F */
672 0 : case 'V':
673 0 : Phonize('F');
674 0 : break;
675 : /* W before a vowel, else dropped */
676 0 : case 'W':
677 0 : if (isvowel(Next_Letter))
678 0 : Phonize('W');
679 0 : break;
680 : /* KS */
681 0 : case 'X':
682 0 : Phonize('K');
683 0 : if (max_phonemes == 0 || Phone_Len < max_phonemes)
684 0 : Phonize('S');
685 0 : break;
686 : /* Y if followed by a vowel */
687 0 : case 'Y':
688 0 : if (isvowel(Next_Letter))
689 0 : Phonize('Y');
690 0 : break;
691 : /* S */
692 0 : case 'Z':
693 0 : Phonize('S');
694 0 : break;
695 : /* No transformation */
696 1 : case 'F':
697 : case 'J':
698 : case 'L':
699 : case 'M':
700 : case 'N':
701 : case 'R':
702 1 : Phonize(Curr_Letter);
703 1 : break;
704 2 : default:
705 : /* nothing */
706 2 : break;
707 : } /* END SWITCH */
708 :
709 5 : w_idx += skip_letter;
710 : } /* END FOR */
711 :
712 1 : End_Phoned_Word;
713 : } /* END metaphone */
714 :
715 :
716 : /*
717 : * SQL function: soundex(text) returns text
718 : */
719 3 : PG_FUNCTION_INFO_V1(soundex);
720 :
721 : Datum
722 8 : soundex(PG_FUNCTION_ARGS)
723 : {
724 : char outstr[SOUNDEX_LEN + 1];
725 : char *arg;
726 :
727 8 : arg = text_to_cstring(PG_GETARG_TEXT_PP(0));
728 :
729 8 : _soundex(arg, outstr);
730 :
731 8 : PG_RETURN_TEXT_P(cstring_to_text(outstr));
732 : }
733 :
734 : static void
735 16 : _soundex(const char *instr, char *outstr)
736 : {
737 : int count;
738 :
739 : Assert(instr);
740 : Assert(outstr);
741 :
742 : /* Skip leading non-alphabetic characters */
743 16 : while (*instr && !ascii_isalpha((unsigned char) *instr))
744 0 : ++instr;
745 :
746 : /* If no string left, return all-zeroes buffer */
747 16 : if (!*instr)
748 : {
749 3 : memset(outstr, '\0', SOUNDEX_LEN + 1);
750 3 : return;
751 : }
752 :
753 : /* Take the first letter as is */
754 13 : *outstr++ = (char) pg_ascii_toupper((unsigned char) *instr++);
755 :
756 13 : count = 1;
757 60 : while (*instr && count < SOUNDEX_LEN)
758 : {
759 93 : if (ascii_isalpha((unsigned char) *instr) &&
760 46 : soundex_code(*instr) != soundex_code(*(instr - 1)))
761 : {
762 35 : *outstr = soundex_code(*instr);
763 35 : if (*outstr != '0')
764 : {
765 23 : ++outstr;
766 23 : ++count;
767 : }
768 : }
769 47 : ++instr;
770 : }
771 :
772 : /* Fill with 0's */
773 29 : while (count < SOUNDEX_LEN)
774 : {
775 16 : *outstr = '0';
776 16 : ++outstr;
777 16 : ++count;
778 : }
779 :
780 : /* And null-terminate */
781 13 : *outstr = '\0';
782 : }
783 :
784 2 : PG_FUNCTION_INFO_V1(difference);
785 :
786 : Datum
787 4 : difference(PG_FUNCTION_ARGS)
788 : {
789 : char sndx1[SOUNDEX_LEN + 1],
790 : sndx2[SOUNDEX_LEN + 1];
791 : int i,
792 : result;
793 :
794 4 : _soundex(text_to_cstring(PG_GETARG_TEXT_PP(0)), sndx1);
795 4 : _soundex(text_to_cstring(PG_GETARG_TEXT_PP(1)), sndx2);
796 :
797 4 : result = 0;
798 20 : for (i = 0; i < SOUNDEX_LEN; i++)
799 : {
800 16 : if (sndx1[i] == sndx2[i])
801 10 : result++;
802 : }
803 :
804 4 : PG_RETURN_INT32(result);
805 : }
|
