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, size_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
50 325 : unicode_lowercase_simple(char32_t code)
51 : {
52 325 : char32_t cp = find_case_map(code, case_map_lower);
53 :
54 325 : 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
66 325 : unicode_uppercase_simple(char32_t code)
67 : {
68 325 : char32_t cp = find_case_map(code, case_map_upper);
69 :
70 325 : 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.
88 : *
89 : * Result string is stored in dst, truncating if larger than dstsize. If
90 : * dstsize is greater than the result length, dst will be NUL-terminated;
91 : * otherwise not.
92 : *
93 : * If dstsize is zero, dst may be NULL. This is useful for calculating the
94 : * required buffer size before allocating.
95 : *
96 : * If full is true, use special case mappings if available and if the
97 : * conditions are satisfied.
98 : */
99 : size_t
100 6312 : unicode_strlower(char *dst, size_t dstsize, const char *src, size_t srclen,
101 : bool full)
102 : {
103 6312 : return convert_case(dst, dstsize, src, srclen, CaseLower, full, NULL,
104 : NULL);
105 : }
106 :
107 : /*
108 : * unicode_strtitle()
109 : *
110 : * Convert src to titlecase, and return the result length (not including
111 : * terminating NUL).
112 : *
113 : * String src must be encoded in UTF-8.
114 : *
115 : * Result string is stored in dst, truncating if larger than dstsize. If
116 : * dstsize is greater than the result length, dst will be NUL-terminated;
117 : * otherwise not.
118 : *
119 : * If dstsize is zero, dst may be NULL. This is useful for calculating the
120 : * required buffer size before allocating.
121 : *
122 : * If full is true, use special case mappings if available and if the
123 : * conditions are satisfied. Otherwise, use only simple mappings and use
124 : * uppercase instead of titlecase.
125 : *
126 : * Titlecasing requires knowledge about word boundaries, which is provided by
127 : * the callback wbnext. A word boundary is the offset of the start of a word
128 : * or the offset of the character immediately following a word.
129 : *
130 : * The caller is expected to initialize and free the callback state
131 : * wbstate. The callback should first return offset 0 for the first boundary;
132 : * then the offset of each subsequent word boundary; then the total length of
133 : * the string to indicate the final boundary.
134 : */
135 : size_t
136 133 : unicode_strtitle(char *dst, size_t dstsize, const char *src, size_t srclen,
137 : bool full, WordBoundaryNext wbnext, void *wbstate)
138 : {
139 133 : return convert_case(dst, dstsize, src, srclen, CaseTitle, full, wbnext,
140 : wbstate);
141 : }
142 :
143 : /*
144 : * unicode_strupper()
145 : *
146 : * Convert src to uppercase, and return the result length (not including
147 : * terminating NUL).
148 : *
149 : * String src must be encoded in UTF-8.
150 : *
151 : * Result string is stored in dst, truncating if larger than dstsize. If
152 : * dstsize is greater than the result length, dst will be NUL-terminated;
153 : * otherwise not.
154 : *
155 : * If dstsize is zero, dst may be NULL. This is useful for calculating the
156 : * required buffer size before allocating.
157 : *
158 : * If full is true, use special case mappings if available and if the
159 : * conditions are satisfied.
160 : */
161 : size_t
162 158561 : unicode_strupper(char *dst, size_t dstsize, const char *src, size_t srclen,
163 : bool full)
164 : {
165 158561 : return convert_case(dst, dstsize, src, srclen, CaseUpper, full, NULL,
166 : NULL);
167 : }
168 :
169 : /*
170 : * unicode_strfold()
171 : *
172 : * Case fold src, and return the result length (not including terminating
173 : * NUL).
174 : *
175 : * String src must be encoded in UTF-8.
176 : *
177 : * Result string is stored in dst, truncating if larger than dstsize. If
178 : * dstsize is greater than the result length, dst will be NUL-terminated;
179 : * otherwise not.
180 : *
181 : * If dstsize is zero, dst may be NULL. This is useful for calculating the
182 : * required buffer size before allocating.
183 : */
184 : size_t
185 10 : unicode_strfold(char *dst, size_t dstsize, const char *src, size_t srclen,
186 : bool full)
187 : {
188 10 : return convert_case(dst, dstsize, src, srclen, CaseFold, full, NULL,
189 : NULL);
190 : }
191 :
192 : /*
193 : * Implement Unicode Default Case Conversion algorithm.
194 : *
195 : * If str_casekind is CaseLower or CaseUpper, map each character in the string
196 : * for which a mapping is available.
197 : *
198 : * If str_casekind is CaseTitle, maps characters found on a word boundary to
199 : * titlecase (or uppercase if full is false) and other characters to
200 : * lowercase. NB: does not currently implement the Unicode behavior in which
201 : * the word boundary is adjusted to the next Cased character. That behavior
202 : * could be implemented as an option, but it doesn't match the default
203 : * behavior of ICU, nor does it match the documented behavior of INITCAP().
204 : *
205 : * If full is true, use special mappings for relevant characters, which can
206 : * map a single codepoint to multiple codepoints, or depend on conditions.
207 : */
208 : static size_t
209 165016 : convert_case(char *dst, size_t dstsize, const char *src, size_t srclen,
210 : CaseKind str_casekind, bool full, WordBoundaryNext wbnext,
211 : void *wbstate)
212 : {
213 : /* character CaseKind varies while titlecasing */
214 165016 : CaseKind chr_casekind = str_casekind;
215 165016 : size_t srcoff = 0;
216 165016 : size_t result_len = 0;
217 165016 : size_t boundary = 0;
218 :
219 : Assert((str_casekind == CaseTitle && wbnext && wbstate) ||
220 : (str_casekind != CaseTitle && !wbnext && !wbstate));
221 :
222 165016 : if (str_casekind == CaseTitle)
223 : {
224 133 : boundary = wbnext(wbstate);
225 : Assert(boundary == 0); /* start of text is always a boundary */
226 : }
227 :
228 494087 : while (srcoff < srclen)
229 : {
230 329071 : char32_t u1 = utf8_to_unicode((const unsigned char *) src + srcoff);
231 329071 : int u1len = unicode_utf8len(u1);
232 329071 : char32_t simple = 0;
233 329071 : const char32_t *special = NULL;
234 : enum CaseMapResult casemap_result;
235 :
236 329071 : if (str_casekind == CaseTitle)
237 : {
238 1032 : if (srcoff == boundary)
239 : {
240 431 : chr_casekind = full ? CaseTitle : CaseUpper;
241 431 : boundary = wbnext(wbstate);
242 : }
243 : else
244 601 : chr_casekind = CaseLower;
245 : }
246 :
247 329071 : casemap_result = casemap(u1, chr_casekind, full, src, srclen, srcoff,
248 : &simple, &special);
249 :
250 329071 : switch (casemap_result)
251 : {
252 88 : case CASEMAP_SELF:
253 : /* no mapping; copy bytes from src */
254 : Assert(simple == 0);
255 : Assert(special == NULL);
256 88 : if (result_len + u1len <= dstsize)
257 88 : memcpy(dst + result_len, src + srcoff, u1len);
258 :
259 88 : result_len += u1len;
260 88 : break;
261 328900 : case CASEMAP_SIMPLE:
262 : {
263 : /* replace with single character */
264 328900 : char32_t u2 = simple;
265 328900 : char32_t u2len = unicode_utf8len(u2);
266 :
267 : Assert(special == NULL);
268 328900 : if (result_len + u2len <= dstsize)
269 328868 : unicode_to_utf8(u2, (unsigned char *) dst + result_len);
270 :
271 328900 : result_len += u2len;
272 : }
273 328900 : break;
274 83 : case CASEMAP_SPECIAL:
275 : /* replace with up to MAX_CASE_EXPANSION characters */
276 : Assert(simple == 0);
277 205 : for (int i = 0; i < MAX_CASE_EXPANSION && special[i]; i++)
278 : {
279 122 : char32_t u2 = special[i];
280 122 : size_t u2len = unicode_utf8len(u2);
281 :
282 122 : if (result_len + u2len <= dstsize)
283 122 : unicode_to_utf8(u2, (unsigned char *) dst + result_len);
284 :
285 122 : result_len += u2len;
286 : }
287 83 : break;
288 : }
289 :
290 329071 : srcoff += u1len;
291 : }
292 :
293 165016 : if (result_len < dstsize)
294 164968 : dst[result_len] = '\0';
295 :
296 165016 : return result_len;
297 : }
298 :
299 : /*
300 : * Check that the condition matches Final_Sigma, described in Unicode Table
301 : * 3-17. The character at the given offset must be directly preceded by a
302 : * Cased character, and must not be directly followed by a Cased character.
303 : *
304 : * Case_Ignorable characters are ignored. NB: some characters may be both
305 : * Cased and Case_Ignorable, in which case they are ignored.
306 : */
307 : static bool
308 50 : check_final_sigma(const unsigned char *str, size_t len, size_t offset)
309 : {
310 : /* the start of the string is not preceded by a Cased character */
311 50 : if (offset == 0)
312 5 : return false;
313 :
314 : /* iterate backwards, looking for Cased character */
315 120 : for (int i = offset - 1; i >= 0; i--)
316 : {
317 120 : if ((str[i] & 0x80) == 0 || (str[i] & 0xC0) == 0xC0)
318 : {
319 65 : char32_t curr = utf8_to_unicode(str + i);
320 :
321 65 : if (pg_u_prop_case_ignorable(curr))
322 20 : continue;
323 45 : else if (pg_u_prop_cased(curr))
324 35 : break;
325 : else
326 10 : return false;
327 : }
328 55 : else if ((str[i] & 0xC0) == 0x80)
329 55 : continue;
330 :
331 : Assert(false); /* invalid UTF-8 */
332 : }
333 :
334 : /* end of string is not followed by a Cased character */
335 35 : if (offset == len)
336 0 : return true;
337 :
338 : /* iterate forwards, looking for Cased character */
339 110 : for (int i = offset + 1; i < len && str[i] != '\0'; i++)
340 : {
341 95 : if ((str[i] & 0x80) == 0 || (str[i] & 0xC0) == 0xC0)
342 : {
343 40 : char32_t curr = utf8_to_unicode(str + i);
344 :
345 40 : if (pg_u_prop_case_ignorable(curr))
346 20 : continue;
347 20 : else if (pg_u_prop_cased(curr))
348 15 : return false;
349 : else
350 5 : break;
351 : }
352 55 : else if ((str[i] & 0xC0) == 0x80)
353 55 : continue;
354 :
355 : Assert(false); /* invalid UTF-8 */
356 : }
357 :
358 20 : return true;
359 : }
360 :
361 : /*
362 : * Unicode allows for special casing to be applied only under certain
363 : * circumstances. The only currently-supported condition is Final_Sigma.
364 : */
365 : static bool
366 113 : check_special_conditions(int conditions, const char *str, size_t len,
367 : size_t offset)
368 : {
369 113 : if (conditions == 0)
370 63 : return true;
371 50 : else if (conditions == PG_U_FINAL_SIGMA)
372 50 : return check_final_sigma((const unsigned char *) str, len, offset);
373 :
374 : /* no other conditions supported */
375 : Assert(false);
376 0 : return false;
377 : }
378 :
379 : /*
380 : * Map the given character to the requested case.
381 : *
382 : * If full is true, and a special case mapping is found and the conditions are
383 : * met, 'special' is set to the mapping result (which is an array of up to
384 : * MAX_CASE_EXPANSION characters) and CASEMAP_SPECIAL is returned.
385 : *
386 : * Otherwise, search for a simple mapping, and if found, set 'simple' to the
387 : * result and return CASEMAP_SIMPLE.
388 : *
389 : * If no mapping is found, return CASEMAP_SELF, and the caller should copy the
390 : * character without modification.
391 : */
392 : static enum CaseMapResult
393 329071 : casemap(char32_t u1, CaseKind casekind, bool full,
394 : const char *src, size_t srclen, size_t srcoff,
395 : char32_t *simple, const char32_t **special)
396 : {
397 : uint16 idx;
398 :
399 : /* Fast path for codepoints < 0x80 */
400 329071 : if (u1 < 0x80)
401 : {
402 : /*
403 : * The first elements in all tables are reserved as 0 (as NULL). The
404 : * data starts at index 1, not 0.
405 : */
406 327650 : *simple = casekind_map[casekind][u1 + 1];
407 :
408 327650 : return CASEMAP_SIMPLE;
409 : }
410 :
411 1421 : idx = case_index(u1);
412 :
413 1421 : if (idx == 0)
414 88 : return CASEMAP_SELF;
415 :
416 1446 : if (full && case_map_special[idx] &&
417 113 : check_special_conditions(special_case[case_map_special[idx]].conditions,
418 : src, srclen, srcoff))
419 : {
420 83 : *special = special_case[case_map_special[idx]].map[casekind];
421 83 : return CASEMAP_SPECIAL;
422 : }
423 :
424 1250 : *simple = casekind_map[casekind][idx];
425 :
426 1250 : return CASEMAP_SIMPLE;
427 : }
428 :
429 : /*
430 : * Find entry in simple case map.
431 : * If the entry does not exist, 0 will be returned.
432 : */
433 : static char32_t
434 650 : find_case_map(char32_t ucs, const char32_t *map)
435 : {
436 : /* Fast path for codepoints < 0x80 */
437 650 : if (ucs < 0x80)
438 : /* The first elements in all tables are reserved as 0 (as NULL). */
439 362 : return map[ucs + 1];
440 288 : return map[case_index(ucs)];
441 : }
|
