Line data Source code
1 : /*-------------------------------------------------------------------------
2 : * unicode_case.c
3 : * Unicode case mapping and case conversion.
4 : *
5 : * Portions Copyright (c) 2017-2026, PostgreSQL Global Development Group
6 : *
7 : * IDENTIFICATION
8 : * src/common/unicode_case.c
9 : *
10 : *-------------------------------------------------------------------------
11 : */
12 : #ifndef FRONTEND
13 : #include "postgres.h"
14 : #else
15 : #include "postgres_fe.h"
16 : #endif
17 :
18 : #include "common/unicode_case.h"
19 : #include "common/unicode_case_table.h"
20 : #include "common/unicode_category.h"
21 : #include "mb/pg_wchar.h"
22 :
23 : enum CaseMapResult
24 : {
25 : CASEMAP_SELF,
26 : CASEMAP_SIMPLE,
27 : CASEMAP_SPECIAL,
28 : };
29 :
30 : /*
31 : * Map for each case kind.
32 : */
33 : static const char32_t *const casekind_map[NCaseKind] =
34 : {
35 : [CaseLower] = case_map_lower,
36 : [CaseTitle] = case_map_title,
37 : [CaseUpper] = case_map_upper,
38 : [CaseFold] = case_map_fold,
39 : };
40 :
41 : static char32_t find_case_map(char32_t ucs, const char32_t *map);
42 : static size_t convert_case(char *dst, size_t dstsize, const char *src, ssize_t srclen,
43 : CaseKind str_casekind, bool full, WordBoundaryNext wbnext,
44 : void *wbstate);
45 : static enum CaseMapResult casemap(char32_t u1, CaseKind casekind, bool full,
46 : const char *src, size_t srclen, size_t srcoff,
47 : char32_t *simple, const char32_t **special);
48 :
49 : char32_t
50 264 : unicode_lowercase_simple(char32_t code)
51 : {
52 264 : char32_t cp = find_case_map(code, case_map_lower);
53 :
54 264 : return cp != 0 ? cp : code;
55 : }
56 :
57 : char32_t
58 0 : unicode_titlecase_simple(char32_t code)
59 : {
60 0 : char32_t cp = find_case_map(code, case_map_title);
61 :
62 0 : return cp != 0 ? cp : code;
63 : }
64 :
65 : char32_t
66 264 : unicode_uppercase_simple(char32_t code)
67 : {
68 264 : char32_t cp = find_case_map(code, case_map_upper);
69 :
70 264 : return cp != 0 ? cp : code;
71 : }
72 :
73 : char32_t
74 0 : unicode_casefold_simple(char32_t code)
75 : {
76 0 : char32_t cp = find_case_map(code, case_map_fold);
77 :
78 0 : return cp != 0 ? cp : code;
79 : }
80 :
81 : /*
82 : * unicode_strlower()
83 : *
84 : * Convert src to lowercase, and return the result length (not including
85 : * terminating NUL).
86 : *
87 : * String src must be encoded in UTF-8. If srclen < 0, src must be
88 : * NUL-terminated.
89 : *
90 : * Result string is stored in dst, truncating if larger than dstsize. If
91 : * dstsize is greater than the result length, dst will be NUL-terminated;
92 : * otherwise not.
93 : *
94 : * If dstsize is zero, dst may be NULL. This is useful for calculating the
95 : * required buffer size before allocating.
96 : *
97 : * If full is true, use special case mappings if available and if the
98 : * conditions are satisfied.
99 : */
100 : size_t
101 6200 : unicode_strlower(char *dst, size_t dstsize, const char *src, ssize_t srclen,
102 : bool full)
103 : {
104 6200 : return convert_case(dst, dstsize, src, srclen, CaseLower, full, NULL,
105 : NULL);
106 : }
107 :
108 : /*
109 : * unicode_strtitle()
110 : *
111 : * Convert src to titlecase, and return the result length (not including
112 : * terminating NUL).
113 : *
114 : * String src must be encoded in UTF-8. If srclen < 0, src must be
115 : * NUL-terminated.
116 : *
117 : * Result string is stored in dst, truncating if larger than dstsize. If
118 : * dstsize is greater than the result length, dst will be NUL-terminated;
119 : * otherwise not.
120 : *
121 : * If dstsize is zero, dst may be NULL. This is useful for calculating the
122 : * required buffer size before allocating.
123 : *
124 : * If full is true, use special case mappings if available and if the
125 : * conditions are satisfied. Otherwise, use only simple mappings and use
126 : * uppercase instead of titlecase.
127 : *
128 : * Titlecasing requires knowledge about word boundaries, which is provided by
129 : * the callback wbnext. A word boundary is the offset of the start of a word
130 : * or the offset of the character immediately following a word.
131 : *
132 : * The caller is expected to initialize and free the callback state
133 : * wbstate. The callback should first return offset 0 for the first boundary;
134 : * then the offset of each subsequent word boundary; then the total length of
135 : * the string to indicate the final boundary.
136 : */
137 : size_t
138 97 : unicode_strtitle(char *dst, size_t dstsize, const char *src, ssize_t srclen,
139 : bool full, WordBoundaryNext wbnext, void *wbstate)
140 : {
141 97 : return convert_case(dst, dstsize, src, srclen, CaseTitle, full, wbnext,
142 : wbstate);
143 : }
144 :
145 : /*
146 : * unicode_strupper()
147 : *
148 : * Convert src to uppercase, and return the result length (not including
149 : * terminating NUL).
150 : *
151 : * String src must be encoded in UTF-8. If srclen < 0, src must be
152 : * NUL-terminated.
153 : *
154 : * Result string is stored in dst, truncating if larger than dstsize. If
155 : * dstsize is greater than the result length, dst will be NUL-terminated;
156 : * otherwise not.
157 : *
158 : * If dstsize is zero, dst may be NULL. This is useful for calculating the
159 : * required buffer size before allocating.
160 : *
161 : * If full is true, use special case mappings if available and if the
162 : * conditions are satisfied.
163 : */
164 : size_t
165 158517 : unicode_strupper(char *dst, size_t dstsize, const char *src, ssize_t srclen,
166 : bool full)
167 : {
168 158517 : return convert_case(dst, dstsize, src, srclen, CaseUpper, full, NULL,
169 : NULL);
170 : }
171 :
172 : /*
173 : * unicode_strfold()
174 : *
175 : * Case fold src, and return the result length (not including terminating
176 : * NUL).
177 : *
178 : * String src must be encoded in UTF-8. If srclen < 0, src must be
179 : * NUL-terminated.
180 : *
181 : * Result string is stored in dst, truncating if larger than dstsize. If
182 : * dstsize is greater than the result length, dst will be NUL-terminated;
183 : * otherwise not.
184 : *
185 : * If dstsize is zero, dst may be NULL. This is useful for calculating the
186 : * required buffer size before allocating.
187 : */
188 : size_t
189 6 : unicode_strfold(char *dst, size_t dstsize, const char *src, ssize_t srclen,
190 : bool full)
191 : {
192 6 : return convert_case(dst, dstsize, src, srclen, CaseFold, full, NULL,
193 : NULL);
194 : }
195 :
196 : /*
197 : * Implement Unicode Default Case Conversion algorithm.
198 : *
199 : * If str_casekind is CaseLower or CaseUpper, map each character in the string
200 : * for which a mapping is available.
201 : *
202 : * If str_casekind is CaseTitle, maps characters found on a word boundary to
203 : * titlecase (or uppercase if full is false) and other characters to
204 : * lowercase. NB: does not currently implement the Unicode behavior in which
205 : * the word boundary is adjusted to the next Cased character. That behavior
206 : * could be implemented as an option, but it doesn't match the default
207 : * behavior of ICU, nor does it match the documented behavior of INITCAP().
208 : *
209 : * If full is true, use special mappings for relevant characters, which can
210 : * map a single codepoint to multiple codepoints, or depend on conditions.
211 : */
212 : static size_t
213 164820 : convert_case(char *dst, size_t dstsize, const char *src, ssize_t srclen,
214 : CaseKind str_casekind, bool full, WordBoundaryNext wbnext,
215 : void *wbstate)
216 : {
217 : /* character CaseKind varies while titlecasing */
218 164820 : CaseKind chr_casekind = str_casekind;
219 164820 : size_t srcoff = 0;
220 164820 : size_t result_len = 0;
221 164820 : size_t boundary = 0;
222 :
223 : Assert((str_casekind == CaseTitle && wbnext && wbstate) ||
224 : (str_casekind != CaseTitle && !wbnext && !wbstate));
225 :
226 164820 : if (str_casekind == CaseTitle)
227 : {
228 97 : boundary = wbnext(wbstate);
229 : Assert(boundary == 0); /* start of text is always a boundary */
230 : }
231 :
232 491971 : while ((srclen < 0 || srcoff < srclen) && src[srcoff] != '\0')
233 : {
234 327151 : char32_t u1 = utf8_to_unicode((const unsigned char *) src + srcoff);
235 327151 : int u1len = unicode_utf8len(u1);
236 327151 : char32_t simple = 0;
237 327151 : const char32_t *special = NULL;
238 : enum CaseMapResult casemap_result;
239 :
240 327151 : if (str_casekind == CaseTitle)
241 : {
242 753 : if (srcoff == boundary)
243 : {
244 315 : chr_casekind = full ? CaseTitle : CaseUpper;
245 315 : boundary = wbnext(wbstate);
246 : }
247 : else
248 438 : chr_casekind = CaseLower;
249 : }
250 :
251 327151 : casemap_result = casemap(u1, chr_casekind, full, src, srclen, srcoff,
252 : &simple, &special);
253 :
254 327151 : switch (casemap_result)
255 : {
256 60 : case CASEMAP_SELF:
257 : /* no mapping; copy bytes from src */
258 : Assert(simple == 0);
259 : Assert(special == NULL);
260 60 : if (result_len + u1len <= dstsize)
261 60 : memcpy(dst + result_len, src + srcoff, u1len);
262 :
263 60 : result_len += u1len;
264 60 : break;
265 327034 : case CASEMAP_SIMPLE:
266 : {
267 : /* replace with single character */
268 327034 : char32_t u2 = simple;
269 327034 : char32_t u2len = unicode_utf8len(u2);
270 :
271 : Assert(special == NULL);
272 327034 : if (result_len + u2len <= dstsize)
273 327010 : unicode_to_utf8(u2, (unsigned char *) dst + result_len);
274 :
275 327034 : result_len += u2len;
276 : }
277 327034 : break;
278 57 : case CASEMAP_SPECIAL:
279 : /* replace with up to MAX_CASE_EXPANSION characters */
280 : Assert(simple == 0);
281 141 : for (int i = 0; i < MAX_CASE_EXPANSION && special[i]; i++)
282 : {
283 84 : char32_t u2 = special[i];
284 84 : size_t u2len = unicode_utf8len(u2);
285 :
286 84 : if (result_len + u2len <= dstsize)
287 84 : unicode_to_utf8(u2, (unsigned char *) dst + result_len);
288 :
289 84 : result_len += u2len;
290 : }
291 57 : break;
292 : }
293 :
294 327151 : srcoff += u1len;
295 : }
296 :
297 164820 : if (result_len < dstsize)
298 164784 : dst[result_len] = '\0';
299 :
300 164820 : return result_len;
301 : }
302 :
303 : /*
304 : * Check that the condition matches Final_Sigma, described in Unicode Table
305 : * 3-17. The character at the given offset must be directly preceded by a
306 : * Cased character, and must not be directly followed by a Cased character.
307 : *
308 : * Case_Ignorable characters are ignored. NB: some characters may be both
309 : * Cased and Case_Ignorable, in which case they are ignored.
310 : */
311 : static bool
312 30 : check_final_sigma(const unsigned char *str, size_t len, size_t offset)
313 : {
314 : /* the start of the string is not preceded by a Cased character */
315 30 : if (offset == 0)
316 3 : return false;
317 :
318 : /* iterate backwards, looking for Cased character */
319 72 : for (int i = offset - 1; i >= 0; i--)
320 : {
321 72 : if ((str[i] & 0x80) == 0 || (str[i] & 0xC0) == 0xC0)
322 : {
323 39 : char32_t curr = utf8_to_unicode(str + i);
324 :
325 39 : if (pg_u_prop_case_ignorable(curr))
326 12 : continue;
327 27 : else if (pg_u_prop_cased(curr))
328 21 : break;
329 : else
330 6 : return false;
331 : }
332 33 : else if ((str[i] & 0xC0) == 0x80)
333 33 : continue;
334 :
335 : Assert(false); /* invalid UTF-8 */
336 : }
337 :
338 : /* end of string is not followed by a Cased character */
339 21 : if (offset == len)
340 0 : return true;
341 :
342 : /* iterate forwards, looking for Cased character */
343 66 : for (int i = offset + 1; i < len && str[i] != '\0'; i++)
344 : {
345 57 : if ((str[i] & 0x80) == 0 || (str[i] & 0xC0) == 0xC0)
346 : {
347 24 : char32_t curr = utf8_to_unicode(str + i);
348 :
349 24 : if (pg_u_prop_case_ignorable(curr))
350 12 : continue;
351 12 : else if (pg_u_prop_cased(curr))
352 9 : return false;
353 : else
354 3 : break;
355 : }
356 33 : else if ((str[i] & 0xC0) == 0x80)
357 33 : continue;
358 :
359 : Assert(false); /* invalid UTF-8 */
360 : }
361 :
362 12 : return true;
363 : }
364 :
365 : /*
366 : * Unicode allows for special casing to be applied only under certain
367 : * circumstances. The only currently-supported condition is Final_Sigma.
368 : */
369 : static bool
370 75 : check_special_conditions(int conditions, const char *str, size_t len,
371 : size_t offset)
372 : {
373 75 : if (conditions == 0)
374 45 : return true;
375 30 : else if (conditions == PG_U_FINAL_SIGMA)
376 30 : return check_final_sigma((const unsigned char *) str, len, offset);
377 :
378 : /* no other conditions supported */
379 : Assert(false);
380 0 : return false;
381 : }
382 :
383 : /*
384 : * Map the given character to the requested case.
385 : *
386 : * If full is true, and a special case mapping is found and the conditions are
387 : * met, 'special' is set to the mapping result (which is an array of up to
388 : * MAX_CASE_EXPANSION characters) and CASEMAP_SPECIAL is returned.
389 : *
390 : * Otherwise, search for a simple mapping, and if found, set 'simple' to the
391 : * result and return CASEMAP_SIMPLE.
392 : *
393 : * If no mapping is found, return CASEMAP_SELF, and the caller should copy the
394 : * character without modification.
395 : */
396 : static enum CaseMapResult
397 327151 : casemap(char32_t u1, CaseKind casekind, bool full,
398 : const char *src, size_t srclen, size_t srcoff,
399 : char32_t *simple, const char32_t **special)
400 : {
401 : uint16 idx;
402 :
403 : /* Fast path for codepoints < 0x80 */
404 327151 : if (u1 < 0x80)
405 : {
406 : /*
407 : * The first elements in all tables are reserved as 0 (as NULL). The
408 : * data starts at index 1, not 0.
409 : */
410 326128 : *simple = casekind_map[casekind][u1 + 1];
411 :
412 326128 : return CASEMAP_SIMPLE;
413 : }
414 :
415 1023 : idx = case_index(u1);
416 :
417 1023 : if (idx == 0)
418 60 : return CASEMAP_SELF;
419 :
420 1038 : if (full && case_map_special[idx] &&
421 75 : check_special_conditions(special_case[case_map_special[idx]].conditions,
422 : src, srclen, srcoff))
423 : {
424 57 : *special = special_case[case_map_special[idx]].map[casekind];
425 57 : return CASEMAP_SPECIAL;
426 : }
427 :
428 906 : *simple = casekind_map[casekind][idx];
429 :
430 906 : return CASEMAP_SIMPLE;
431 : }
432 :
433 : /*
434 : * Find entry in simple case map.
435 : * If the entry does not exist, 0 will be returned.
436 : */
437 : static char32_t
438 528 : find_case_map(char32_t ucs, const char32_t *map)
439 : {
440 : /* Fast path for codepoints < 0x80 */
441 528 : if (ucs < 0x80)
442 : /* The first elements in all tables are reserved as 0 (as NULL). */
443 312 : return map[ucs + 1];
444 216 : return map[case_index(ucs)];
445 : }
|
