Line data Source code
1 : /*-------------------------------------------------------------------------
2 : *
3 : * pg_wchar.h
4 : * multibyte-character support
5 : *
6 : * Portions Copyright (c) 1996-2025, PostgreSQL Global Development Group
7 : * Portions Copyright (c) 1994, Regents of the University of California
8 : *
9 : * src/include/mb/pg_wchar.h
10 : *
11 : * NOTES
12 : * This is used both by the backend and by frontends, but should not be
13 : * included by libpq client programs. In particular, a libpq client
14 : * should not assume that the encoding IDs used by the version of libpq
15 : * it's linked to match up with the IDs declared here.
16 : * To help prevent mistakes, relevant functions that are exported by
17 : * libpq have a physically different name when being referenced
18 : * statically.
19 : *
20 : *-------------------------------------------------------------------------
21 : */
22 : #ifndef PG_WCHAR_H
23 : #define PG_WCHAR_H
24 :
25 : /*
26 : * The pg_wchar type
27 : */
28 : typedef unsigned int pg_wchar;
29 :
30 : /*
31 : * Maximum byte length of multibyte characters in any backend encoding
32 : */
33 : #define MAX_MULTIBYTE_CHAR_LEN 4
34 :
35 : /*
36 : * various definitions for EUC
37 : */
38 : #define SS2 0x8e /* single shift 2 (JIS0201) */
39 : #define SS3 0x8f /* single shift 3 (JIS0212) */
40 :
41 : /*
42 : * SJIS validation macros
43 : */
44 : #define ISSJISHEAD(c) (((c) >= 0x81 && (c) <= 0x9f) || ((c) >= 0xe0 && (c) <= 0xfc))
45 : #define ISSJISTAIL(c) (((c) >= 0x40 && (c) <= 0x7e) || ((c) >= 0x80 && (c) <= 0xfc))
46 :
47 : /*----------------------------------------------------
48 : * MULE Internal Encoding (MIC)
49 : *
50 : * This encoding follows the design used within XEmacs; it is meant to
51 : * subsume many externally-defined character sets. Each character includes
52 : * identification of the character set it belongs to, so the encoding is
53 : * general but somewhat bulky.
54 : *
55 : * Currently PostgreSQL supports 5 types of MULE character sets:
56 : *
57 : * 1) 1-byte ASCII characters. Each byte is below 0x80.
58 : *
59 : * 2) "Official" single byte charsets such as ISO-8859-1 (Latin1).
60 : * Each MULE character consists of 2 bytes: LC1 + C1, where LC1 is
61 : * an identifier for the charset (in the range 0x81 to 0x8d) and C1
62 : * is the character code (in the range 0xa0 to 0xff).
63 : *
64 : * 3) "Private" single byte charsets such as SISHENG. Each MULE
65 : * character consists of 3 bytes: LCPRV1 + LC12 + C1, where LCPRV1
66 : * is a private-charset flag, LC12 is an identifier for the charset,
67 : * and C1 is the character code (in the range 0xa0 to 0xff).
68 : * LCPRV1 is either 0x9a (if LC12 is in the range 0xa0 to 0xdf)
69 : * or 0x9b (if LC12 is in the range 0xe0 to 0xef).
70 : *
71 : * 4) "Official" multibyte charsets such as JIS X0208. Each MULE
72 : * character consists of 3 bytes: LC2 + C1 + C2, where LC2 is
73 : * an identifier for the charset (in the range 0x90 to 0x99) and C1
74 : * and C2 form the character code (each in the range 0xa0 to 0xff).
75 : *
76 : * 5) "Private" multibyte charsets such as CNS 11643-1992 Plane 3.
77 : * Each MULE character consists of 4 bytes: LCPRV2 + LC22 + C1 + C2,
78 : * where LCPRV2 is a private-charset flag, LC22 is an identifier for
79 : * the charset, and C1 and C2 form the character code (each in the range
80 : * 0xa0 to 0xff). LCPRV2 is either 0x9c (if LC22 is in the range 0xf0
81 : * to 0xf4) or 0x9d (if LC22 is in the range 0xf5 to 0xfe).
82 : *
83 : * "Official" encodings are those that have been assigned code numbers by
84 : * the XEmacs project; "private" encodings have Postgres-specific charset
85 : * identifiers.
86 : *
87 : * See the "XEmacs Internals Manual", available at http://www.xemacs.org,
88 : * for more details. Note that for historical reasons, Postgres'
89 : * private-charset flag values do not match what XEmacs says they should be,
90 : * so this isn't really exactly MULE (not that private charsets would be
91 : * interoperable anyway).
92 : *
93 : * Note that XEmacs's implementation is different from what emacs does.
94 : * We follow emacs's implementation, rather than XEmacs's.
95 : *----------------------------------------------------
96 : */
97 :
98 : /*
99 : * Charset identifiers (also called "leading bytes" in the MULE documentation)
100 : */
101 :
102 : /*
103 : * Charset IDs for official single byte encodings (0x81-0x8e)
104 : */
105 : #define LC_ISO8859_1 0x81 /* ISO8859 Latin 1 */
106 : #define LC_ISO8859_2 0x82 /* ISO8859 Latin 2 */
107 : #define LC_ISO8859_3 0x83 /* ISO8859 Latin 3 */
108 : #define LC_ISO8859_4 0x84 /* ISO8859 Latin 4 */
109 : #define LC_TIS620 0x85 /* Thai (not supported yet) */
110 : #define LC_ISO8859_7 0x86 /* Greek (not supported yet) */
111 : #define LC_ISO8859_6 0x87 /* Arabic (not supported yet) */
112 : #define LC_ISO8859_8 0x88 /* Hebrew (not supported yet) */
113 : #define LC_JISX0201K 0x89 /* Japanese 1 byte kana */
114 : #define LC_JISX0201R 0x8a /* Japanese 1 byte Roman */
115 : /* Note that 0x8b seems to be unused as of Emacs 20.7.
116 : * However, there might be a chance that 0x8b could be used
117 : * in later versions of Emacs.
118 : */
119 : #define LC_KOI8_R 0x8b /* Cyrillic KOI8-R */
120 : #define LC_ISO8859_5 0x8c /* ISO8859 Cyrillic */
121 : #define LC_ISO8859_9 0x8d /* ISO8859 Latin 5 (not supported yet) */
122 : #define LC_ISO8859_15 0x8e /* ISO8859 Latin 15 (not supported yet) */
123 : /* #define CONTROL_1 0x8f control characters (unused) */
124 :
125 : /* Is a leading byte for "official" single byte encodings? */
126 : #define IS_LC1(c) ((unsigned char)(c) >= 0x81 && (unsigned char)(c) <= 0x8d)
127 :
128 : /*
129 : * Charset IDs for official multibyte encodings (0x90-0x99)
130 : * 0x9a-0x9d are free. 0x9e and 0x9f are reserved.
131 : */
132 : #define LC_JISX0208_1978 0x90 /* Japanese Kanji, old JIS (not supported) */
133 : #define LC_GB2312_80 0x91 /* Chinese */
134 : #define LC_JISX0208 0x92 /* Japanese Kanji (JIS X 0208) */
135 : #define LC_KS5601 0x93 /* Korean */
136 : #define LC_JISX0212 0x94 /* Japanese Kanji (JIS X 0212) */
137 : #define LC_CNS11643_1 0x95 /* CNS 11643-1992 Plane 1 */
138 : #define LC_CNS11643_2 0x96 /* CNS 11643-1992 Plane 2 */
139 : #define LC_JISX0213_1 0x97 /* Japanese Kanji (JIS X 0213 Plane 1)
140 : * (not supported) */
141 : #define LC_BIG5_1 0x98 /* Plane 1 Chinese traditional (not
142 : * supported) */
143 : #define LC_BIG5_2 0x99 /* Plane 1 Chinese traditional (not
144 : * supported) */
145 :
146 : /* Is a leading byte for "official" multibyte encodings? */
147 : #define IS_LC2(c) ((unsigned char)(c) >= 0x90 && (unsigned char)(c) <= 0x99)
148 :
149 : /*
150 : * Postgres-specific prefix bytes for "private" single byte encodings
151 : * (According to the MULE docs, we should be using 0x9e for this)
152 : */
153 : #define LCPRV1_A 0x9a
154 : #define LCPRV1_B 0x9b
155 : #define IS_LCPRV1(c) ((unsigned char)(c) == LCPRV1_A || (unsigned char)(c) == LCPRV1_B)
156 : #define IS_LCPRV1_A_RANGE(c) \
157 : ((unsigned char)(c) >= 0xa0 && (unsigned char)(c) <= 0xdf)
158 : #define IS_LCPRV1_B_RANGE(c) \
159 : ((unsigned char)(c) >= 0xe0 && (unsigned char)(c) <= 0xef)
160 :
161 : /*
162 : * Postgres-specific prefix bytes for "private" multibyte encodings
163 : * (According to the MULE docs, we should be using 0x9f for this)
164 : */
165 : #define LCPRV2_A 0x9c
166 : #define LCPRV2_B 0x9d
167 : #define IS_LCPRV2(c) ((unsigned char)(c) == LCPRV2_A || (unsigned char)(c) == LCPRV2_B)
168 : #define IS_LCPRV2_A_RANGE(c) \
169 : ((unsigned char)(c) >= 0xf0 && (unsigned char)(c) <= 0xf4)
170 : #define IS_LCPRV2_B_RANGE(c) \
171 : ((unsigned char)(c) >= 0xf5 && (unsigned char)(c) <= 0xfe)
172 :
173 : /*
174 : * Charset IDs for private single byte encodings (0xa0-0xef)
175 : */
176 : #define LC_SISHENG 0xa0 /* Chinese SiSheng characters for
177 : * PinYin/ZhuYin (not supported) */
178 : #define LC_IPA 0xa1 /* IPA (International Phonetic
179 : * Association) (not supported) */
180 : #define LC_VISCII_LOWER 0xa2 /* Vietnamese VISCII1.1 lower-case (not
181 : * supported) */
182 : #define LC_VISCII_UPPER 0xa3 /* Vietnamese VISCII1.1 upper-case (not
183 : * supported) */
184 : #define LC_ARABIC_DIGIT 0xa4 /* Arabic digit (not supported) */
185 : #define LC_ARABIC_1_COLUMN 0xa5 /* Arabic 1-column (not supported) */
186 : #define LC_ASCII_RIGHT_TO_LEFT 0xa6 /* ASCII (left half of ISO8859-1) with
187 : * right-to-left direction (not
188 : * supported) */
189 : #define LC_LAO 0xa7 /* Lao characters (ISO10646 0E80..0EDF)
190 : * (not supported) */
191 : #define LC_ARABIC_2_COLUMN 0xa8 /* Arabic 1-column (not supported) */
192 :
193 : /*
194 : * Charset IDs for private multibyte encodings (0xf0-0xff)
195 : */
196 : #define LC_INDIAN_1_COLUMN 0xf0 /* Indian charset for 1-column width
197 : * glyphs (not supported) */
198 : #define LC_TIBETAN_1_COLUMN 0xf1 /* Tibetan 1-column width glyphs (not
199 : * supported) */
200 : #define LC_UNICODE_SUBSET_2 0xf2 /* Unicode characters of the range
201 : * U+2500..U+33FF. (not supported) */
202 : #define LC_UNICODE_SUBSET_3 0xf3 /* Unicode characters of the range
203 : * U+E000..U+FFFF. (not supported) */
204 : #define LC_UNICODE_SUBSET 0xf4 /* Unicode characters of the range
205 : * U+0100..U+24FF. (not supported) */
206 : #define LC_ETHIOPIC 0xf5 /* Ethiopic characters (not supported) */
207 : #define LC_CNS11643_3 0xf6 /* CNS 11643-1992 Plane 3 */
208 : #define LC_CNS11643_4 0xf7 /* CNS 11643-1992 Plane 4 */
209 : #define LC_CNS11643_5 0xf8 /* CNS 11643-1992 Plane 5 */
210 : #define LC_CNS11643_6 0xf9 /* CNS 11643-1992 Plane 6 */
211 : #define LC_CNS11643_7 0xfa /* CNS 11643-1992 Plane 7 */
212 : #define LC_INDIAN_2_COLUMN 0xfb /* Indian charset for 2-column width
213 : * glyphs (not supported) */
214 : #define LC_TIBETAN 0xfc /* Tibetan (not supported) */
215 : /* #define FREE 0xfd free (unused) */
216 : /* #define FREE 0xfe free (unused) */
217 : /* #define FREE 0xff free (unused) */
218 :
219 : /*----------------------------------------------------
220 : * end of MULE stuff
221 : *----------------------------------------------------
222 : */
223 :
224 : /*
225 : * PostgreSQL encoding identifiers
226 : *
227 : * WARNING: If you add some encoding don't forget to update
228 : * the pg_enc2name_tbl[] array (in src/common/encnames.c),
229 : * the pg_enc2gettext_tbl[] array (in src/common/encnames.c) and
230 : * the pg_wchar_table[] array (in src/common/wchar.c) and to check
231 : * PG_ENCODING_BE_LAST macro.
232 : *
233 : * PG_SQL_ASCII is default encoding and must be = 0.
234 : *
235 : * XXX We must avoid renumbering any backend encoding until libpq's major
236 : * version number is increased beyond 5; it turns out that the backend
237 : * encoding IDs are effectively part of libpq's ABI as far as 8.2 initdb and
238 : * psql are concerned.
239 : */
240 : typedef enum pg_enc
241 : {
242 : PG_SQL_ASCII = 0, /* SQL/ASCII */
243 : PG_EUC_JP, /* EUC for Japanese */
244 : PG_EUC_CN, /* EUC for Chinese */
245 : PG_EUC_KR, /* EUC for Korean */
246 : PG_EUC_TW, /* EUC for Taiwan */
247 : PG_EUC_JIS_2004, /* EUC-JIS-2004 */
248 : PG_UTF8, /* Unicode UTF8 */
249 : PG_MULE_INTERNAL, /* Mule internal code */
250 : PG_LATIN1, /* ISO-8859-1 Latin 1 */
251 : PG_LATIN2, /* ISO-8859-2 Latin 2 */
252 : PG_LATIN3, /* ISO-8859-3 Latin 3 */
253 : PG_LATIN4, /* ISO-8859-4 Latin 4 */
254 : PG_LATIN5, /* ISO-8859-9 Latin 5 */
255 : PG_LATIN6, /* ISO-8859-10 Latin6 */
256 : PG_LATIN7, /* ISO-8859-13 Latin7 */
257 : PG_LATIN8, /* ISO-8859-14 Latin8 */
258 : PG_LATIN9, /* ISO-8859-15 Latin9 */
259 : PG_LATIN10, /* ISO-8859-16 Latin10 */
260 : PG_WIN1256, /* windows-1256 */
261 : PG_WIN1258, /* Windows-1258 */
262 : PG_WIN866, /* (MS-DOS CP866) */
263 : PG_WIN874, /* windows-874 */
264 : PG_KOI8R, /* KOI8-R */
265 : PG_WIN1251, /* windows-1251 */
266 : PG_WIN1252, /* windows-1252 */
267 : PG_ISO_8859_5, /* ISO-8859-5 */
268 : PG_ISO_8859_6, /* ISO-8859-6 */
269 : PG_ISO_8859_7, /* ISO-8859-7 */
270 : PG_ISO_8859_8, /* ISO-8859-8 */
271 : PG_WIN1250, /* windows-1250 */
272 : PG_WIN1253, /* windows-1253 */
273 : PG_WIN1254, /* windows-1254 */
274 : PG_WIN1255, /* windows-1255 */
275 : PG_WIN1257, /* windows-1257 */
276 : PG_KOI8U, /* KOI8-U */
277 : /* PG_ENCODING_BE_LAST points to the above entry */
278 :
279 : /* followings are for client encoding only */
280 : PG_SJIS, /* Shift JIS (Windows-932) */
281 : PG_BIG5, /* Big5 (Windows-950) */
282 : PG_GBK, /* GBK (Windows-936) */
283 : PG_UHC, /* UHC (Windows-949) */
284 : PG_GB18030, /* GB18030 */
285 : PG_JOHAB, /* EUC for Korean JOHAB */
286 : PG_SHIFT_JIS_2004, /* Shift-JIS-2004 */
287 : _PG_LAST_ENCODING_ /* mark only */
288 :
289 : } pg_enc;
290 :
291 : #define PG_ENCODING_BE_LAST PG_KOI8U
292 :
293 : /*
294 : * Please use these tests before access to pg_enc2name_tbl[]
295 : * or to other places...
296 : */
297 : #define PG_VALID_BE_ENCODING(_enc) \
298 : ((_enc) >= 0 && (_enc) <= PG_ENCODING_BE_LAST)
299 :
300 : #define PG_ENCODING_IS_CLIENT_ONLY(_enc) \
301 : ((_enc) > PG_ENCODING_BE_LAST && (_enc) < _PG_LAST_ENCODING_)
302 :
303 : #define PG_VALID_ENCODING(_enc) \
304 : ((_enc) >= 0 && (_enc) < _PG_LAST_ENCODING_)
305 :
306 : /* On FE are possible all encodings */
307 : #define PG_VALID_FE_ENCODING(_enc) PG_VALID_ENCODING(_enc)
308 :
309 : /*
310 : * When converting strings between different encodings, we assume that space
311 : * for converted result is 4-to-1 growth in the worst case. The rate for
312 : * currently supported encoding pairs are within 3 (SJIS JIS X0201 half width
313 : * kana -> UTF8 is the worst case). So "4" should be enough for the moment.
314 : *
315 : * Note that this is not the same as the maximum character width in any
316 : * particular encoding.
317 : */
318 : #define MAX_CONVERSION_GROWTH 4
319 :
320 : /*
321 : * Maximum byte length of a string that's required in any encoding to convert
322 : * at least one character to any other encoding. In other words, if you feed
323 : * MAX_CONVERSION_INPUT_LENGTH bytes to any encoding conversion function, it
324 : * is guaranteed to be able to convert something without needing more input
325 : * (assuming the input is valid).
326 : *
327 : * Currently, the maximum case is the conversion UTF8 -> SJIS JIS X0201 half
328 : * width kana, where a pair of UTF-8 characters is converted into a single
329 : * SHIFT_JIS_2004 character (the reverse of the worst case for
330 : * MAX_CONVERSION_GROWTH). It needs 6 bytes of input. In theory, a
331 : * user-defined conversion function might have more complicated cases, although
332 : * for the reverse mapping you would probably also need to bump up
333 : * MAX_CONVERSION_GROWTH. But there is no need to be stingy here, so make it
334 : * generous.
335 : */
336 : #define MAX_CONVERSION_INPUT_LENGTH 16
337 :
338 : /*
339 : * Maximum byte length of the string equivalent to any one Unicode code point,
340 : * in any backend encoding. The current value assumes that a 4-byte UTF-8
341 : * character might expand by MAX_CONVERSION_GROWTH, which is a huge
342 : * overestimate. But in current usage we don't allocate large multiples of
343 : * this, so there's little point in being stingy.
344 : */
345 : #define MAX_UNICODE_EQUIVALENT_STRING 16
346 :
347 : /*
348 : * Table for mapping an encoding number to official encoding name and
349 : * possibly other subsidiary data. Be careful to check encoding number
350 : * before accessing a table entry!
351 : *
352 : * if (PG_VALID_ENCODING(encoding))
353 : * pg_enc2name_tbl[ encoding ];
354 : */
355 : typedef struct pg_enc2name
356 : {
357 : const char *name;
358 : pg_enc encoding;
359 : #ifdef WIN32
360 : unsigned codepage; /* codepage for WIN32 */
361 : #endif
362 : } pg_enc2name;
363 :
364 : extern PGDLLIMPORT const pg_enc2name pg_enc2name_tbl[];
365 :
366 : /*
367 : * Encoding names for gettext
368 : */
369 : extern PGDLLIMPORT const char *pg_enc2gettext_tbl[];
370 :
371 : /*
372 : * pg_wchar stuff
373 : */
374 : typedef int (*mb2wchar_with_len_converter) (const unsigned char *from,
375 : pg_wchar *to,
376 : int len);
377 :
378 : typedef int (*wchar2mb_with_len_converter) (const pg_wchar *from,
379 : unsigned char *to,
380 : int len);
381 :
382 : typedef int (*mblen_converter) (const unsigned char *mbstr);
383 :
384 : typedef int (*mbdisplaylen_converter) (const unsigned char *mbstr);
385 :
386 : typedef bool (*mbcharacter_incrementer) (unsigned char *mbstr, int len);
387 :
388 : typedef int (*mbchar_verifier) (const unsigned char *mbstr, int len);
389 :
390 : typedef int (*mbstr_verifier) (const unsigned char *mbstr, int len);
391 :
392 : typedef struct
393 : {
394 : mb2wchar_with_len_converter mb2wchar_with_len; /* convert a multibyte
395 : * string to a wchar */
396 : wchar2mb_with_len_converter wchar2mb_with_len; /* convert a wchar string
397 : * to a multibyte */
398 : mblen_converter mblen; /* get byte length of a char */
399 : mbdisplaylen_converter dsplen; /* get display width of a char */
400 : mbchar_verifier mbverifychar; /* verify multibyte character */
401 : mbstr_verifier mbverifystr; /* verify multibyte string */
402 : int maxmblen; /* max bytes for a char in this encoding */
403 : } pg_wchar_tbl;
404 :
405 : extern PGDLLIMPORT const pg_wchar_tbl pg_wchar_table[];
406 :
407 : /*
408 : * Data structures for conversions between UTF-8 and other encodings
409 : * (UtfToLocal() and LocalToUtf()). In these data structures, characters of
410 : * either encoding are represented by uint32 words; hence we can only support
411 : * characters up to 4 bytes long. For example, the byte sequence 0xC2 0x89
412 : * would be represented by 0x0000C289, and 0xE8 0xA2 0xB4 by 0x00E8A2B4.
413 : *
414 : * There are three possible ways a character can be mapped:
415 : *
416 : * 1. Using a radix tree, from source to destination code.
417 : * 2. Using a sorted array of source -> destination code pairs. This
418 : * method is used for "combining" characters. There are so few of
419 : * them that building a radix tree would be wasteful.
420 : * 3. Using a conversion function.
421 : */
422 :
423 : /*
424 : * Radix tree for character conversion.
425 : *
426 : * Logically, this is actually four different radix trees, for 1-byte,
427 : * 2-byte, 3-byte and 4-byte inputs. The 1-byte tree is a simple lookup
428 : * table from source to target code. The 2-byte tree consists of two levels:
429 : * one lookup table for the first byte, where the value in the lookup table
430 : * points to a lookup table for the second byte. And so on.
431 : *
432 : * Physically, all the trees are stored in one big array, in 'chars16' or
433 : * 'chars32', depending on the maximum value that needs to be represented. For
434 : * each level in each tree, we also store lower and upper bound of allowed
435 : * values - values outside those bounds are considered invalid, and are left
436 : * out of the tables.
437 : *
438 : * In the intermediate levels of the trees, the values stored are offsets
439 : * into the chars[16|32] array.
440 : *
441 : * In the beginning of the chars[16|32] array, there is always a number of
442 : * zeros, so that you safely follow an index from an intermediate table
443 : * without explicitly checking for a zero. Following a zero any number of
444 : * times will always bring you to the dummy, all-zeros table in the
445 : * beginning. This helps to shave some cycles when looking up values.
446 : */
447 : typedef struct
448 : {
449 : /*
450 : * Array containing all the values. Only one of chars16 or chars32 is
451 : * used, depending on how wide the values we need to represent are.
452 : */
453 : const uint16 *chars16;
454 : const uint32 *chars32;
455 :
456 : /* Radix tree for 1-byte inputs */
457 : uint32 b1root; /* offset of table in the chars[16|32] array */
458 : uint8 b1_lower; /* min allowed value for a single byte input */
459 : uint8 b1_upper; /* max allowed value for a single byte input */
460 :
461 : /* Radix tree for 2-byte inputs */
462 : uint32 b2root; /* offset of 1st byte's table */
463 : uint8 b2_1_lower; /* min/max allowed value for 1st input byte */
464 : uint8 b2_1_upper;
465 : uint8 b2_2_lower; /* min/max allowed value for 2nd input byte */
466 : uint8 b2_2_upper;
467 :
468 : /* Radix tree for 3-byte inputs */
469 : uint32 b3root; /* offset of 1st byte's table */
470 : uint8 b3_1_lower; /* min/max allowed value for 1st input byte */
471 : uint8 b3_1_upper;
472 : uint8 b3_2_lower; /* min/max allowed value for 2nd input byte */
473 : uint8 b3_2_upper;
474 : uint8 b3_3_lower; /* min/max allowed value for 3rd input byte */
475 : uint8 b3_3_upper;
476 :
477 : /* Radix tree for 4-byte inputs */
478 : uint32 b4root; /* offset of 1st byte's table */
479 : uint8 b4_1_lower; /* min/max allowed value for 1st input byte */
480 : uint8 b4_1_upper;
481 : uint8 b4_2_lower; /* min/max allowed value for 2nd input byte */
482 : uint8 b4_2_upper;
483 : uint8 b4_3_lower; /* min/max allowed value for 3rd input byte */
484 : uint8 b4_3_upper;
485 : uint8 b4_4_lower; /* min/max allowed value for 4th input byte */
486 : uint8 b4_4_upper;
487 :
488 : } pg_mb_radix_tree;
489 :
490 : /*
491 : * UTF-8 to local code conversion map (for combined characters)
492 : */
493 : typedef struct
494 : {
495 : uint32 utf1; /* UTF-8 code 1 */
496 : uint32 utf2; /* UTF-8 code 2 */
497 : uint32 code; /* local code */
498 : } pg_utf_to_local_combined;
499 :
500 : /*
501 : * local code to UTF-8 conversion map (for combined characters)
502 : */
503 : typedef struct
504 : {
505 : uint32 code; /* local code */
506 : uint32 utf1; /* UTF-8 code 1 */
507 : uint32 utf2; /* UTF-8 code 2 */
508 : } pg_local_to_utf_combined;
509 :
510 : /*
511 : * callback function for algorithmic encoding conversions (in either direction)
512 : *
513 : * if function returns zero, it does not know how to convert the code
514 : */
515 : typedef uint32 (*utf_local_conversion_func) (uint32 code);
516 :
517 : /*
518 : * Support macro for encoding conversion functions to validate their
519 : * arguments. (This could be made more compact if we included fmgr.h
520 : * here, but we don't want to do that because this header file is also
521 : * used by frontends.)
522 : */
523 : #define CHECK_ENCODING_CONVERSION_ARGS(srcencoding,destencoding) \
524 : check_encoding_conversion_args(PG_GETARG_INT32(0), \
525 : PG_GETARG_INT32(1), \
526 : PG_GETARG_INT32(4), \
527 : (srcencoding), \
528 : (destencoding))
529 :
530 :
531 : /*
532 : * Some handy functions for Unicode-specific tests.
533 : */
534 : static inline bool
535 2094 : is_valid_unicode_codepoint(pg_wchar c)
536 : {
537 2094 : return (c > 0 && c <= 0x10FFFF);
538 : }
539 :
540 : static inline bool
541 1446 : is_utf16_surrogate_first(pg_wchar c)
542 : {
543 1446 : return (c >= 0xD800 && c <= 0xDBFF);
544 : }
545 :
546 : static inline bool
547 1332 : is_utf16_surrogate_second(pg_wchar c)
548 : {
549 1332 : return (c >= 0xDC00 && c <= 0xDFFF);
550 : }
551 :
552 : static inline pg_wchar
553 60 : surrogate_pair_to_codepoint(pg_wchar first, pg_wchar second)
554 : {
555 60 : return ((first & 0x3FF) << 10) + 0x10000 + (second & 0x3FF);
556 : }
557 :
558 : /*
559 : * Convert a UTF-8 character to a Unicode code point.
560 : * This is a one-character version of pg_utf2wchar_with_len.
561 : *
562 : * No error checks here, c must point to a long-enough string.
563 : */
564 : static inline pg_wchar
565 54683444 : utf8_to_unicode(const unsigned char *c)
566 : {
567 54683444 : if ((*c & 0x80) == 0)
568 54670492 : return (pg_wchar) c[0];
569 12952 : else if ((*c & 0xe0) == 0xc0)
570 11018 : return (pg_wchar) (((c[0] & 0x1f) << 6) |
571 11018 : (c[1] & 0x3f));
572 1934 : else if ((*c & 0xf0) == 0xe0)
573 1832 : return (pg_wchar) (((c[0] & 0x0f) << 12) |
574 1832 : ((c[1] & 0x3f) << 6) |
575 1832 : (c[2] & 0x3f));
576 102 : else if ((*c & 0xf8) == 0xf0)
577 102 : return (pg_wchar) (((c[0] & 0x07) << 18) |
578 102 : ((c[1] & 0x3f) << 12) |
579 102 : ((c[2] & 0x3f) << 6) |
580 102 : (c[3] & 0x3f));
581 : else
582 : /* that is an invalid code on purpose */
583 0 : return 0xffffffff;
584 : }
585 :
586 : /*
587 : * Map a Unicode code point to UTF-8. utf8string must have at least
588 : * unicode_utf8len(c) bytes available.
589 : */
590 : static inline unsigned char *
591 16335984 : unicode_to_utf8(pg_wchar c, unsigned char *utf8string)
592 : {
593 16335984 : if (c <= 0x7F)
594 : {
595 16333198 : utf8string[0] = c;
596 : }
597 2786 : else if (c <= 0x7FF)
598 : {
599 2008 : utf8string[0] = 0xC0 | ((c >> 6) & 0x1F);
600 2008 : utf8string[1] = 0x80 | (c & 0x3F);
601 : }
602 778 : else if (c <= 0xFFFF)
603 : {
604 712 : utf8string[0] = 0xE0 | ((c >> 12) & 0x0F);
605 712 : utf8string[1] = 0x80 | ((c >> 6) & 0x3F);
606 712 : utf8string[2] = 0x80 | (c & 0x3F);
607 : }
608 : else
609 : {
610 66 : utf8string[0] = 0xF0 | ((c >> 18) & 0x07);
611 66 : utf8string[1] = 0x80 | ((c >> 12) & 0x3F);
612 66 : utf8string[2] = 0x80 | ((c >> 6) & 0x3F);
613 66 : utf8string[3] = 0x80 | (c & 0x3F);
614 : }
615 :
616 16335984 : return utf8string;
617 : }
618 :
619 : /*
620 : * Number of bytes needed to represent the given char in UTF8.
621 : */
622 : static inline int
623 1317500 : unicode_utf8len(pg_wchar c)
624 : {
625 1317500 : if (c <= 0x7F)
626 1313000 : return 1;
627 4500 : else if (c <= 0x7FF)
628 3504 : return 2;
629 996 : else if (c <= 0xFFFF)
630 996 : return 3;
631 : else
632 0 : return 4;
633 : }
634 :
635 : /*
636 : * The functions in this list are exported by libpq, and we need to be sure
637 : * that we know which calls are satisfied by libpq and which are satisfied
638 : * by static linkage to libpgcommon. (This is because we might be using a
639 : * libpq.so that's of a different major version and has encoding IDs that
640 : * differ from the current version's.) The nominal function names are what
641 : * are actually used in and exported by libpq, while the names exported by
642 : * libpgcommon.a and libpgcommon_srv.a end in "_private".
643 : */
644 : #if defined(USE_PRIVATE_ENCODING_FUNCS) || !defined(FRONTEND)
645 : #define pg_char_to_encoding pg_char_to_encoding_private
646 : #define pg_encoding_to_char pg_encoding_to_char_private
647 : #define pg_valid_server_encoding pg_valid_server_encoding_private
648 : #define pg_valid_server_encoding_id pg_valid_server_encoding_id_private
649 : #define pg_utf_mblen pg_utf_mblen_private
650 : #endif
651 :
652 : /*
653 : * These functions are considered part of libpq's exported API and
654 : * are also declared in libpq-fe.h.
655 : */
656 : extern int pg_char_to_encoding(const char *name);
657 : extern const char *pg_encoding_to_char(int encoding);
658 : extern int pg_valid_server_encoding_id(int encoding);
659 :
660 : /*
661 : * These functions are available to frontend code that links with libpgcommon
662 : * (in addition to the ones just above). The constant tables declared
663 : * earlier in this file are also available from libpgcommon.
664 : */
665 : extern void pg_encoding_set_invalid(int encoding, char *dst);
666 : extern int pg_encoding_mblen(int encoding, const char *mbstr);
667 : extern int pg_encoding_mblen_or_incomplete(int encoding, const char *mbstr,
668 : size_t remaining);
669 : extern int pg_encoding_mblen_bounded(int encoding, const char *mbstr);
670 : extern int pg_encoding_dsplen(int encoding, const char *mbstr);
671 : extern int pg_encoding_verifymbchar(int encoding, const char *mbstr, int len);
672 : extern int pg_encoding_verifymbstr(int encoding, const char *mbstr, int len);
673 : extern int pg_encoding_max_length(int encoding);
674 : extern int pg_valid_client_encoding(const char *name);
675 : extern int pg_valid_server_encoding(const char *name);
676 : extern bool is_encoding_supported_by_icu(int encoding);
677 : extern const char *get_encoding_name_for_icu(int encoding);
678 :
679 : extern unsigned char *unicode_to_utf8(pg_wchar c, unsigned char *utf8string);
680 : extern pg_wchar utf8_to_unicode(const unsigned char *c);
681 : extern bool pg_utf8_islegal(const unsigned char *source, int length);
682 : extern int pg_utf_mblen(const unsigned char *s);
683 : extern int pg_mule_mblen(const unsigned char *s);
684 :
685 : /*
686 : * The remaining functions are backend-only.
687 : */
688 : extern int pg_mb2wchar(const char *from, pg_wchar *to);
689 : extern int pg_mb2wchar_with_len(const char *from, pg_wchar *to, int len);
690 : extern int pg_encoding_mb2wchar_with_len(int encoding,
691 : const char *from, pg_wchar *to, int len);
692 : extern int pg_wchar2mb(const pg_wchar *from, char *to);
693 : extern int pg_wchar2mb_with_len(const pg_wchar *from, char *to, int len);
694 : extern int pg_encoding_wchar2mb_with_len(int encoding,
695 : const pg_wchar *from, char *to, int len);
696 : extern int pg_char_and_wchar_strcmp(const char *s1, const pg_wchar *s2);
697 : extern int pg_wchar_strncmp(const pg_wchar *s1, const pg_wchar *s2, size_t n);
698 : extern int pg_char_and_wchar_strncmp(const char *s1, const pg_wchar *s2, size_t n);
699 : extern size_t pg_wchar_strlen(const pg_wchar *str);
700 : extern int pg_mblen(const char *mbstr);
701 : extern int pg_dsplen(const char *mbstr);
702 : extern int pg_mbstrlen(const char *mbstr);
703 : extern int pg_mbstrlen_with_len(const char *mbstr, int limit);
704 : extern int pg_mbcliplen(const char *mbstr, int len, int limit);
705 : extern int pg_encoding_mbcliplen(int encoding, const char *mbstr,
706 : int len, int limit);
707 : extern int pg_mbcharcliplen(const char *mbstr, int len, int limit);
708 : extern int pg_database_encoding_max_length(void);
709 : extern mbcharacter_incrementer pg_database_encoding_character_incrementer(void);
710 :
711 : extern int PrepareClientEncoding(int encoding);
712 : extern int SetClientEncoding(int encoding);
713 : extern void InitializeClientEncoding(void);
714 : extern int pg_get_client_encoding(void);
715 : extern const char *pg_get_client_encoding_name(void);
716 :
717 : extern void SetDatabaseEncoding(int encoding);
718 : extern int GetDatabaseEncoding(void);
719 : extern const char *GetDatabaseEncodingName(void);
720 : extern void SetMessageEncoding(int encoding);
721 : extern int GetMessageEncoding(void);
722 :
723 : #ifdef ENABLE_NLS
724 : extern int pg_bind_textdomain_codeset(const char *domainname);
725 : #endif
726 :
727 : extern unsigned char *pg_do_encoding_conversion(unsigned char *src, int len,
728 : int src_encoding,
729 : int dest_encoding);
730 : extern int pg_do_encoding_conversion_buf(Oid proc,
731 : int src_encoding,
732 : int dest_encoding,
733 : unsigned char *src, int srclen,
734 : unsigned char *dest, int destlen,
735 : bool noError);
736 :
737 : extern char *pg_client_to_server(const char *s, int len);
738 : extern char *pg_server_to_client(const char *s, int len);
739 : extern char *pg_any_to_server(const char *s, int len, int encoding);
740 : extern char *pg_server_to_any(const char *s, int len, int encoding);
741 :
742 : extern void pg_unicode_to_server(pg_wchar c, unsigned char *s);
743 : extern bool pg_unicode_to_server_noerror(pg_wchar c, unsigned char *s);
744 :
745 : extern unsigned short BIG5toCNS(unsigned short big5, unsigned char *lc);
746 : extern unsigned short CNStoBIG5(unsigned short cns, unsigned char lc);
747 :
748 : extern int UtfToLocal(const unsigned char *utf, int len,
749 : unsigned char *iso,
750 : const pg_mb_radix_tree *map,
751 : const pg_utf_to_local_combined *cmap, int cmapsize,
752 : utf_local_conversion_func conv_func,
753 : int encoding, bool noError);
754 : extern int LocalToUtf(const unsigned char *iso, int len,
755 : unsigned char *utf,
756 : const pg_mb_radix_tree *map,
757 : const pg_local_to_utf_combined *cmap, int cmapsize,
758 : utf_local_conversion_func conv_func,
759 : int encoding, bool noError);
760 :
761 : extern bool pg_verifymbstr(const char *mbstr, int len, bool noError);
762 : extern bool pg_verify_mbstr(int encoding, const char *mbstr, int len,
763 : bool noError);
764 : extern int pg_verify_mbstr_len(int encoding, const char *mbstr, int len,
765 : bool noError);
766 :
767 : extern void check_encoding_conversion_args(int src_encoding,
768 : int dest_encoding,
769 : int len,
770 : int expected_src_encoding,
771 : int expected_dest_encoding);
772 :
773 : pg_noreturn extern void report_invalid_encoding(int encoding, const char *mbstr, int len);
774 : pg_noreturn extern void report_untranslatable_char(int src_encoding, int dest_encoding,
775 : const char *mbstr, int len);
776 :
777 : extern int local2local(const unsigned char *l, unsigned char *p, int len,
778 : int src_encoding, int dest_encoding,
779 : const unsigned char *tab, bool noError);
780 : extern int latin2mic(const unsigned char *l, unsigned char *p, int len,
781 : int lc, int encoding, bool noError);
782 : extern int mic2latin(const unsigned char *mic, unsigned char *p, int len,
783 : int lc, int encoding, bool noError);
784 : extern int latin2mic_with_table(const unsigned char *l, unsigned char *p,
785 : int len, int lc, int encoding,
786 : const unsigned char *tab, bool noError);
787 : extern int mic2latin_with_table(const unsigned char *mic, unsigned char *p,
788 : int len, int lc, int encoding,
789 : const unsigned char *tab, bool noError);
790 :
791 : #ifdef WIN32
792 : extern WCHAR *pgwin32_message_to_UTF16(const char *str, int len, int *utf16len);
793 : #endif
794 :
795 : #endif /* PG_WCHAR_H */
|
