c6d25d8d |
1 | /* |
2 | * iso2022s.c - support for ISO-2022 subset encodings. |
3 | * |
4 | * (The `s' suffix on the filename is there to leave `iso2022.c' |
5 | * free for the unlikely event that I ever attempt to implement |
6 | * _full_ ISO-2022 in this library!) |
7 | */ |
8 | |
9 | #ifndef ENUM_CHARSETS |
10 | |
11 | #include <stdio.h> |
12 | #include <string.h> |
13 | #include <assert.h> |
14 | |
15 | #include "charset.h" |
16 | #include "internal.h" |
17 | |
18 | #define SO (0x0E) |
19 | #define SI (0x0F) |
20 | #define ESC (0x1B) |
21 | |
22 | /* Functional description of a single ISO 2022 escape sequence. */ |
23 | struct iso2022_escape { |
24 | char const *sequence; |
25 | unsigned long andbits, xorbits; |
26 | /* |
27 | * For output, these variables help us figure out which escape |
28 | * sequences we need to get where we want to be. |
29 | */ |
30 | int container, subcharset; |
31 | }; |
32 | |
33 | struct iso2022 { |
34 | /* |
35 | * List of escape sequences supported in this subset. Must be |
36 | * in ASCII order, so that we can narrow down the list as |
37 | * necessary. |
38 | */ |
39 | struct iso2022_escape *escapes; /* must be sorted in ASCII order! */ |
40 | int nescapes; |
41 | |
42 | /* |
43 | * We assign indices from 0 upwards to the sub-charsets of a |
44 | * given ISO 2022 subset. nbytes[i] tells us how many bytes per |
45 | * character are required by sub-charset i. (It's a string |
46 | * mainly because that makes it easier to declare in C syntax |
47 | * than an int array.) |
48 | */ |
49 | char const *nbytes; |
50 | |
51 | /* |
52 | * The characters in this string are indices-plus-one (so that |
53 | * NUL can still terminate) of escape sequences in `escapes'. |
54 | * These escapes are output in the given sequence to reset the |
55 | * encoding state, unless it turns out that a given escape |
56 | * would not change the state at all. |
57 | */ |
58 | char const *reset; |
59 | |
60 | /* |
61 | * Initial value of s1, in case the default container contents |
62 | * needs to be something other than charset 0 in all cases. |
63 | * (Note that this must have the top bit set!) |
64 | */ |
65 | unsigned long s1; |
66 | |
67 | /* |
68 | * For output, some ISO 2022 subsets _mandate_ an initial shift |
69 | * sequence. If so, here it is so we can output it. (For the |
70 | * sake of basic sanity we won't bother to _require_ it on |
71 | * input, although it should of course be listed under |
72 | * `escapes' above so that we ignore it when present.) |
73 | */ |
74 | char const *initial_sequence; |
75 | |
76 | /* |
77 | * Function calls to do the actual translation. |
78 | */ |
79 | long int (*to_ucs)(int subcharset, unsigned long bytes); |
80 | int (*from_ucs)(long int ucs, int *subcharset, unsigned long *bytes); |
81 | }; |
82 | |
83 | static void read_iso2022s(charset_spec const *charset, long int input_chr, |
84 | charset_state *state, |
85 | void (*emit)(void *ctx, long int output), |
86 | void *emitctx) |
87 | { |
88 | struct iso2022 const *iso = (struct iso2022 *)charset->data; |
89 | |
90 | /* |
91 | * For reading ISO-2022 subsets, we divide up our state |
92 | * variables as follows: |
93 | * |
94 | * - The top byte of s0 (bits 31:24) indicates, if nonzero, |
95 | * that we are part-way through a recognised ISO-2022 escape |
96 | * sequence. Five of those bits (31:27) give the index of |
97 | * the first member of the escapes list matching what we |
98 | * have so far; the remaining three (26:24) give the number |
99 | * of characters we have seen so far. |
100 | * |
101 | * - The top bit of s1 (bit 31) is non-zero at all times, to |
102 | * indicate that we have performed any necessary |
103 | * initialisation. When we start, we detect a zero s1 and |
104 | * respond to it by initialising the default container |
105 | * contents. |
106 | * |
107 | * - The next three bits of s1 (bits 30:28) indicate which |
108 | * _container_ is currently selected. This isn't quite as |
109 | * simple as it sounds, since we have to preserve memory of |
110 | * which of the SI/SO containers we came from when we're |
111 | * temporarily in SS2/SS3. Hence, what happens is: |
112 | * + bit 28 indicates SI/SO. |
113 | * + if we're in an SS2/SS3 container, that's indicated by |
114 | * the two bits above that being nonzero and holding |
115 | * either 2 or 3. |
116 | * + Hence: 0 is SI, 1 is SO, 4 is SS2-from-SI, 5 is |
117 | * SS2-from-SO, 6 is SS3-from-SI, 7 is SS3-from-SO. |
118 | * |
119 | * - The next nibble of s1 (27:24) indicates how many bytes |
120 | * have been accumulated in the current character. |
121 | * |
122 | * - The remaining three bytes of s1 are divided into four |
123 | * six-bit sections, and each section gives the current |
124 | * sub-charset selected in one of the possible containers. |
125 | * (Those containers are SI, SO, SS2 and SS3, respectively |
126 | * and in order from the bottom of s0 to the top.) |
127 | * |
128 | * - The bottom 24 bits of s0 give the accumulated character |
129 | * data so far. |
130 | * |
131 | * (Note that this means s1 contains all the parts of the state |
132 | * which might need to be operated on by escape sequences. |
133 | * Cunning, eh?) |
134 | */ |
135 | |
136 | if (!(state->s1 & 0x80000000)) { |
137 | state->s1 = iso->s1; |
138 | } |
139 | |
140 | /* |
141 | * So. Firstly, we process escape sequences, if we're in the |
142 | * middle of one or if we see a possible introducer (SI, SO, |
143 | * ESC). |
144 | */ |
145 | if ((state->s0 >> 24) || |
146 | (input_chr == SO || input_chr == SI || input_chr == ESC)) { |
147 | int n = (state->s0 >> 24) & 7, i = (state->s0 >> 27), oi = i, j; |
148 | |
149 | /* |
150 | * If this is the start of an escape sequence, we might be |
151 | * in mid-character. If so, clear the character state and |
152 | * emit an error token for the incomplete character. |
153 | */ |
154 | if (state->s1 & 0x0F000000) { |
155 | state->s1 &= ~0x0F000000; |
156 | state->s0 &= 0xFF000000; |
157 | /* |
158 | * If we were in the SS2 or SS3 container, we |
159 | * automatically exit it. |
160 | */ |
161 | if (state->s1 & 0x60000000) |
162 | state->s1 &= 0x9FFFFFFF; |
163 | emit(emitctx, ERROR); |
164 | } |
165 | |
166 | j = i; |
167 | while (j < iso->nescapes && |
168 | !memcmp(iso->escapes[j].sequence, |
169 | iso->escapes[oi].sequence, n)) { |
170 | if (iso->escapes[j].sequence[n] < input_chr) |
171 | i = ++j; |
172 | else |
173 | break; |
174 | } |
175 | if (i >= iso->nescapes || |
176 | memcmp(iso->escapes[i].sequence, |
177 | iso->escapes[oi].sequence, n) || |
178 | iso->escapes[i].sequence[n] != input_chr) { |
179 | /* |
180 | * This character does not appear in any valid escape |
181 | * sequence. Therefore, we must emit all the characters |
182 | * we had previously swallowed, plus this one, and |
183 | * return to non-escape-sequence state. |
184 | */ |
185 | for (j = 0; j < n; j++) |
186 | emit(emitctx, iso->escapes[oi].sequence[j]); |
187 | emit(emitctx, input_chr); |
188 | state->s0 = 0; |
189 | return; |
190 | } |
191 | |
192 | /* |
193 | * Otherwise, we have found an additional character in our |
194 | * escape sequence. See if we have reached the _end_ of our |
195 | * sequence (and therefore must process the sequence). |
196 | */ |
197 | n++; |
198 | if (!iso->escapes[i].sequence[n]) { |
199 | state->s0 = 0; |
200 | state->s1 &= iso->escapes[i].andbits; |
201 | state->s1 ^= iso->escapes[i].xorbits; |
202 | return; |
203 | } |
204 | |
205 | /* |
206 | * Failing _that_, we simply update our escape-sequence- |
207 | * tracking state. |
208 | */ |
209 | assert(i < 32 && n < 8); |
210 | state->s0 = (i << 27) | (n << 24); |
211 | return; |
212 | } |
213 | |
214 | /* |
215 | * If this isn't an escape sequence, it must be part of a |
216 | * character. One possibility is that it's a control character |
217 | * (outside the space 21-7E), in which case we output it verbatim. |
218 | */ |
219 | if (input_chr < 0x21 || input_chr > 0x7E) { |
220 | /* |
221 | * We might be in mid-multibyte-character. If so, clear the |
222 | * character state and emit an error token for the |
223 | * incomplete character. |
224 | */ |
225 | if (state->s1 & 0x0F000000) { |
226 | state->s1 &= ~0x0F000000; |
227 | state->s0 &= 0xFF000000; |
228 | emit(emitctx, ERROR); |
229 | /* |
230 | * If we were in the SS2 or SS3 container, we |
231 | * automatically exit it. |
232 | */ |
233 | if (state->s1 & 0x60000000) |
234 | state->s1 &= 0x9FFFFFFF; |
235 | } |
236 | |
237 | emit(emitctx, input_chr); |
238 | return; |
239 | } |
240 | |
241 | /* |
242 | * Otherwise, accumulate character data. |
243 | */ |
244 | { |
245 | unsigned long chr; |
246 | int chrlen, cont, subcharset, bytes; |
247 | |
248 | /* The current character and its length. */ |
249 | chr = ((state->s0 & 0x00FFFFFF) << 8) | input_chr; |
250 | chrlen = ((state->s1 >> 24) & 0xF) + 1; |
251 | /* The current sub-charset. */ |
252 | cont = (state->s1 >> 28) & 7; |
253 | if (cont > 1) cont >>= 1; |
254 | subcharset = (state->s1 >> (6*cont)) & 0x3F; |
255 | /* The number of bytes-per-character in that sub-charset. */ |
256 | bytes = iso->nbytes[subcharset]; |
257 | |
258 | /* |
259 | * If this character is now complete, we convert and emit |
260 | * it. Otherwise, we simply update the state and return. |
261 | */ |
262 | if (chrlen >= bytes) { |
263 | emit(emitctx, iso->to_ucs(subcharset, chr)); |
264 | chr = chrlen = 0; |
265 | /* |
266 | * If we were in the SS2 or SS3 container, we |
267 | * automatically exit it. |
268 | */ |
269 | if (state->s1 & 0x60000000) |
270 | state->s1 &= 0x9FFFFFFF; |
271 | } |
272 | state->s0 = (state->s0 & 0xFF000000) | chr; |
273 | state->s1 = (state->s1 & 0xF0FFFFFF) | (chrlen << 24); |
274 | } |
275 | } |
276 | |
277 | static int write_iso2022s(charset_spec const *charset, long int input_chr, |
278 | charset_state *state, |
279 | void (*emit)(void *ctx, long int output), |
280 | void *emitctx) |
281 | { |
282 | struct iso2022 const *iso = (struct iso2022 *)charset->data; |
283 | int subcharset, len, i, j, cont; |
284 | unsigned long bytes; |
285 | |
286 | /* |
287 | * For output, our s1 state variable contains most of the same |
288 | * stuff as it did for input - initial-state indicator bit, |
289 | * current container, and current subcharset selected in each |
290 | * container. |
291 | */ |
292 | |
293 | /* |
294 | * Analyse the character and find out what subcharset it needs |
295 | * to go in. |
296 | */ |
297 | if (input_chr >= 0 && !iso->from_ucs(input_chr, &subcharset, &bytes)) |
298 | return FALSE; |
299 | |
300 | if (!(state->s1 & 0x80000000)) { |
301 | state->s1 = iso->s1; |
302 | if (iso->initial_sequence) |
303 | for (i = 0; iso->initial_sequence[i]; i++) |
304 | emit(emitctx, iso->initial_sequence[i]); |
305 | } |
306 | |
307 | if (input_chr == -1) { |
308 | unsigned long oldstate; |
309 | int k; |
310 | |
311 | /* |
312 | * Special case: reset encoding state. |
313 | */ |
314 | for (i = 0; iso->reset[i]; i++) { |
315 | j = iso->reset[i] - 1; |
316 | oldstate = state->s1; |
317 | state->s1 &= iso->escapes[j].andbits; |
318 | state->s1 ^= iso->escapes[j].xorbits; |
319 | if (state->s1 != oldstate) { |
320 | /* We must actually emit this sequence. */ |
321 | for (k = 0; iso->escapes[j].sequence[k]; k++) |
322 | emit(emitctx, iso->escapes[j].sequence[k]); |
323 | } |
324 | } |
325 | |
326 | return TRUE; |
327 | } |
328 | |
329 | /* |
330 | * Now begins the fun. We now know what subcharset we want. So |
331 | * we must find out which container we should select it into, |
332 | * select it into it if necessary, select that _container_ if |
333 | * necessary, and then output the given bytes. |
334 | */ |
335 | for (i = 0; i < iso->nescapes; i++) |
336 | if (iso->escapes[i].subcharset == subcharset) |
337 | break; |
338 | assert(i < iso->nescapes); |
339 | |
340 | /* |
341 | * We've found the escape sequence which would select this |
342 | * subcharset into a container. However, that subcharset might |
343 | * already _be_ selected in that container! Check before we go |
344 | * to the effort of emitting the sequence. |
345 | */ |
346 | cont = iso->escapes[i].container; |
3cca0edf |
347 | if (((state->s1 >> (6*cont)) & 0x3F) != (unsigned)subcharset) { |
c6d25d8d |
348 | for (j = 0; iso->escapes[i].sequence[j]; j++) |
349 | emit(emitctx, iso->escapes[i].sequence[j]); |
350 | state->s1 &= iso->escapes[i].andbits; |
351 | state->s1 ^= iso->escapes[i].xorbits; |
352 | } |
353 | |
354 | /* |
355 | * Now we know what container our subcharset is in, so we want |
356 | * to select that container. |
357 | */ |
358 | if (cont > 1) { |
359 | /* SS2 or SS3; just output the sequence and be done. */ |
360 | emit(emitctx, ESC); |
361 | emit(emitctx, 'L' + cont); /* comes out to 'N' or 'O' */ |
362 | } else { |
363 | /* Emit SI or SO, but only if the current container isn't already |
364 | * the right one. */ |
3cca0edf |
365 | if (((state->s1 >> 28) & 7) != (unsigned)cont) { |
c6d25d8d |
366 | emit(emitctx, cont ? SO : SI); |
367 | state->s1 = (state->s1 & 0x8FFFFFFF) | (cont << 28); |
368 | } |
369 | } |
370 | |
371 | /* |
372 | * We're done. Subcharset is selected in container, container |
373 | * is selected. All we need now is to write out the bytes. |
374 | */ |
375 | len = iso->nbytes[subcharset]; |
376 | while (len--) |
377 | emit(emitctx, (bytes >> (8*len)) & 0xFF); |
378 | |
379 | return TRUE; |
380 | } |
381 | |
382 | /* |
383 | * ISO-2022-JP, defined in RFC 1468. |
384 | */ |
385 | static long int iso2022jp_to_ucs(int subcharset, unsigned long bytes) |
386 | { |
387 | switch (subcharset) { |
388 | case 0: return bytes; /* one-byte ASCII */ |
389 | case 1: /* JIS X 0201 half-width katakana */ |
390 | if (bytes >= 0x21 && bytes <= 0x5F) |
391 | return bytes + (0xFF61 - 0x21); |
392 | else |
393 | return ERROR; |
394 | /* (no break needed since all control paths have returned) */ |
395 | case 2: return jisx0208_to_unicode(((bytes >> 8) & 0xFF) - 0x21, |
396 | ((bytes ) & 0xFF) - 0x21); |
397 | default: return ERROR; |
398 | } |
399 | } |
400 | static int iso2022jp_from_ucs(long int ucs, int *subcharset, |
401 | unsigned long *bytes) |
402 | { |
403 | int r, c; |
404 | if (ucs < 0x80) { |
405 | *subcharset = 0; |
406 | *bytes = ucs; |
407 | return 1; |
408 | } else if (ucs >= 0xFF61 && ucs <= 0xFF9F) { |
409 | *subcharset = 1; |
410 | *bytes = ucs - (0xFF61 - 0x21); |
411 | return 1; |
412 | } else if (unicode_to_jisx0208(ucs, &r, &c)) { |
413 | *subcharset = 2; |
414 | *bytes = ((r+0x21) << 8) | (c+0x21); |
415 | return 1; |
416 | } else { |
417 | return 0; |
418 | } |
419 | } |
420 | static struct iso2022_escape iso2022jp_escapes[] = { |
421 | {"\033$@", 0xFFFFFFC0, 0x00000002, -1, -1}, /* we ignore this one */ |
422 | {"\033$B", 0xFFFFFFC0, 0x00000002, 0, 2}, |
423 | {"\033(B", 0xFFFFFFC0, 0x00000000, 0, 0}, |
424 | {"\033(J", 0xFFFFFFC0, 0x00000001, 0, 1}, |
425 | }; |
426 | static struct iso2022 iso2022jp = { |
427 | iso2022jp_escapes, lenof(iso2022jp_escapes), |
428 | "\1\1\2", "\3", 0x80000000, NULL, iso2022jp_to_ucs, iso2022jp_from_ucs |
429 | }; |
430 | const charset_spec charset_CS_ISO2022_JP = { |
431 | CS_ISO2022_JP, read_iso2022s, write_iso2022s, &iso2022jp |
432 | }; |
433 | |
434 | /* |
435 | * ISO-2022-KR, defined in RFC 1557. |
436 | */ |
437 | static long int iso2022kr_to_ucs(int subcharset, unsigned long bytes) |
438 | { |
439 | switch (subcharset) { |
440 | case 0: return bytes; /* one-byte ASCII */ |
441 | case 1: return ksx1001_to_unicode(((bytes >> 8) & 0xFF) - 0x21, |
442 | ((bytes ) & 0xFF) - 0x21); |
443 | default: return ERROR; |
444 | } |
445 | } |
446 | static int iso2022kr_from_ucs(long int ucs, int *subcharset, |
447 | unsigned long *bytes) |
448 | { |
449 | int r, c; |
450 | if (ucs < 0x80) { |
451 | *subcharset = 0; |
452 | *bytes = ucs; |
453 | return 1; |
454 | } else if (unicode_to_ksx1001(ucs, &r, &c)) { |
455 | *subcharset = 1; |
456 | *bytes = ((r+0x21) << 8) | (c+0x21); |
457 | return 1; |
458 | } else { |
459 | return 0; |
460 | } |
461 | } |
462 | static struct iso2022_escape iso2022kr_escapes[] = { |
463 | {"\016", 0x8FFFFFFF, 0x10000000, -1, -1}, |
464 | {"\017", 0x8FFFFFFF, 0x00000000, 0, 0}, |
465 | {"\033$)C", 0xFFFFF03F, 0x00000040, 1, 1}, /* bits[11:6] <- 1 */ |
466 | }; |
467 | static struct iso2022 iso2022kr = { |
468 | iso2022kr_escapes, lenof(iso2022kr_escapes), |
469 | "\1\2", "\2", 0x80000040, "\033$)C", iso2022kr_to_ucs, iso2022kr_from_ucs |
470 | }; |
471 | const charset_spec charset_CS_ISO2022_KR = { |
472 | CS_ISO2022_KR, read_iso2022s, write_iso2022s, &iso2022kr |
473 | }; |
474 | |
475 | #else /* ENUM_CHARSETS */ |
476 | |
477 | ENUM_CHARSET(CS_ISO2022_JP) |
478 | ENUM_CHARSET(CS_ISO2022_KR) |
479 | |
480 | #endif /* ENUM_CHARSETS */ |