1 | // © 2016 and later: Unicode, Inc. and others. |
2 | // License & terms of use: http://www.unicode.org/copyright.html |
3 | /* |
4 | ****************************************************************************** |
5 | * |
6 | * Copyright (C) 2002-2016, International Business Machines |
7 | * Corporation and others. All Rights Reserved. |
8 | * |
9 | ****************************************************************************** |
10 | * file name: ucnvbocu.cpp |
11 | * encoding: UTF-8 |
12 | * tab size: 8 (not used) |
13 | * indentation:4 |
14 | * |
15 | * created on: 2002mar27 |
16 | * created by: Markus W. Scherer |
17 | * |
18 | * This is an implementation of the Binary Ordered Compression for Unicode, |
19 | * in its MIME-friendly form as defined in http://www.unicode.org/notes/tn6/ |
20 | */ |
21 | |
22 | #include "unicode/utypes.h" |
23 | |
24 | #if !UCONFIG_NO_CONVERSION && !UCONFIG_ONLY_HTML_CONVERSION |
25 | |
26 | #include "unicode/ucnv.h" |
27 | #include "unicode/ucnv_cb.h" |
28 | #include "unicode/utf16.h" |
29 | #include "putilimp.h" |
30 | #include "ucnv_bld.h" |
31 | #include "ucnv_cnv.h" |
32 | #include "uassert.h" |
33 | |
34 | /* BOCU-1 constants and macros ---------------------------------------------- */ |
35 | |
36 | /* |
37 | * BOCU-1 encodes the code points of a Unicode string as |
38 | * a sequence of byte-encoded differences (slope detection), |
39 | * preserving lexical order. |
40 | * |
41 | * Optimize the difference-taking for runs of Unicode text within |
42 | * small scripts: |
43 | * |
44 | * Most small scripts are allocated within aligned 128-blocks of Unicode |
45 | * code points. Lexical order is preserved if the "previous code point" state |
46 | * is always moved into the middle of such a block. |
47 | * |
48 | * Additionally, "prev" is moved from anywhere in the Unihan and Hangul |
49 | * areas into the middle of those areas. |
50 | * |
51 | * C0 control codes and space are encoded with their US-ASCII bytes. |
52 | * "prev" is reset for C0 controls but not for space. |
53 | */ |
54 | |
55 | /* initial value for "prev": middle of the ASCII range */ |
56 | #define BOCU1_ASCII_PREV 0x40 |
57 | |
58 | /* bounding byte values for differences */ |
59 | #define BOCU1_MIN 0x21 |
60 | #define BOCU1_MIDDLE 0x90 |
61 | #define BOCU1_MAX_LEAD 0xfe |
62 | #define BOCU1_MAX_TRAIL 0xff |
63 | #define BOCU1_RESET 0xff |
64 | |
65 | /* number of lead bytes */ |
66 | #define BOCU1_COUNT (BOCU1_MAX_LEAD-BOCU1_MIN+1) |
67 | |
68 | /* adjust trail byte counts for the use of some C0 control byte values */ |
69 | #define BOCU1_TRAIL_CONTROLS_COUNT 20 |
70 | #define BOCU1_TRAIL_BYTE_OFFSET (BOCU1_MIN-BOCU1_TRAIL_CONTROLS_COUNT) |
71 | |
72 | /* number of trail bytes */ |
73 | #define BOCU1_TRAIL_COUNT ((BOCU1_MAX_TRAIL-BOCU1_MIN+1)+BOCU1_TRAIL_CONTROLS_COUNT) |
74 | |
75 | /* |
76 | * number of positive and negative single-byte codes |
77 | * (counting 0==BOCU1_MIDDLE among the positive ones) |
78 | */ |
79 | #define BOCU1_SINGLE 64 |
80 | |
81 | /* number of lead bytes for positive and negative 2/3/4-byte sequences */ |
82 | #define BOCU1_LEAD_2 43 |
83 | #define BOCU1_LEAD_3 3 |
84 | #define BOCU1_LEAD_4 1 |
85 | |
86 | /* The difference value range for single-byters. */ |
87 | #define BOCU1_REACH_POS_1 (BOCU1_SINGLE-1) |
88 | #define BOCU1_REACH_NEG_1 (-BOCU1_SINGLE) |
89 | |
90 | /* The difference value range for double-byters. */ |
91 | #define BOCU1_REACH_POS_2 (BOCU1_REACH_POS_1+BOCU1_LEAD_2*BOCU1_TRAIL_COUNT) |
92 | #define BOCU1_REACH_NEG_2 (BOCU1_REACH_NEG_1-BOCU1_LEAD_2*BOCU1_TRAIL_COUNT) |
93 | |
94 | /* The difference value range for 3-byters. */ |
95 | #define BOCU1_REACH_POS_3 \ |
96 | (BOCU1_REACH_POS_2+BOCU1_LEAD_3*BOCU1_TRAIL_COUNT*BOCU1_TRAIL_COUNT) |
97 | |
98 | #define BOCU1_REACH_NEG_3 (BOCU1_REACH_NEG_2-BOCU1_LEAD_3*BOCU1_TRAIL_COUNT*BOCU1_TRAIL_COUNT) |
99 | |
100 | /* The lead byte start values. */ |
101 | #define BOCU1_START_POS_2 (BOCU1_MIDDLE+BOCU1_REACH_POS_1+1) |
102 | #define BOCU1_START_POS_3 (BOCU1_START_POS_2+BOCU1_LEAD_2) |
103 | #define BOCU1_START_POS_4 (BOCU1_START_POS_3+BOCU1_LEAD_3) |
104 | /* ==BOCU1_MAX_LEAD */ |
105 | |
106 | #define BOCU1_START_NEG_2 (BOCU1_MIDDLE+BOCU1_REACH_NEG_1) |
107 | #define BOCU1_START_NEG_3 (BOCU1_START_NEG_2-BOCU1_LEAD_2) |
108 | #define BOCU1_START_NEG_4 (BOCU1_START_NEG_3-BOCU1_LEAD_3) |
109 | /* ==BOCU1_MIN+1 */ |
110 | |
111 | /* The length of a byte sequence, according to the lead byte (!=BOCU1_RESET). */ |
112 | #define BOCU1_LENGTH_FROM_LEAD(lead) \ |
113 | ((BOCU1_START_NEG_2<=(lead) && (lead)<BOCU1_START_POS_2) ? 1 : \ |
114 | (BOCU1_START_NEG_3<=(lead) && (lead)<BOCU1_START_POS_3) ? 2 : \ |
115 | (BOCU1_START_NEG_4<=(lead) && (lead)<BOCU1_START_POS_4) ? 3 : 4) |
116 | |
117 | /* The length of a byte sequence, according to its packed form. */ |
118 | #define BOCU1_LENGTH_FROM_PACKED(packed) \ |
119 | ((uint32_t)(packed)<0x04000000 ? (packed)>>24 : 4) |
120 | |
121 | /* |
122 | * 12 commonly used C0 control codes (and space) are only used to encode |
123 | * themselves directly, |
124 | * which makes BOCU-1 MIME-usable and reasonably safe for |
125 | * ASCII-oriented software. |
126 | * |
127 | * These controls are |
128 | * 0 NUL |
129 | * |
130 | * 7 BEL |
131 | * 8 BS |
132 | * |
133 | * 9 TAB |
134 | * a LF |
135 | * b VT |
136 | * c FF |
137 | * d CR |
138 | * |
139 | * e SO |
140 | * f SI |
141 | * |
142 | * 1a SUB |
143 | * 1b ESC |
144 | * |
145 | * The other 20 C0 controls are also encoded directly (to preserve order) |
146 | * but are also used as trail bytes in difference encoding |
147 | * (for better compression). |
148 | */ |
149 | #define BOCU1_TRAIL_TO_BYTE(t) ((t)>=BOCU1_TRAIL_CONTROLS_COUNT ? (t)+BOCU1_TRAIL_BYTE_OFFSET : bocu1TrailToByte[t]) |
150 | |
151 | /* |
152 | * Byte value map for control codes, |
153 | * from external byte values 0x00..0x20 |
154 | * to trail byte values 0..19 (0..0x13) as used in the difference calculation. |
155 | * External byte values that are illegal as trail bytes are mapped to -1. |
156 | */ |
157 | static const int8_t |
158 | bocu1ByteToTrail[BOCU1_MIN]={ |
159 | /* 0 1 2 3 4 5 6 7 */ |
160 | -1, 0x00, 0x01, 0x02, 0x03, 0x04, 0x05, -1, |
161 | |
162 | /* 8 9 a b c d e f */ |
163 | -1, -1, -1, -1, -1, -1, -1, -1, |
164 | |
165 | /* 10 11 12 13 14 15 16 17 */ |
166 | 0x06, 0x07, 0x08, 0x09, 0x0a, 0x0b, 0x0c, 0x0d, |
167 | |
168 | /* 18 19 1a 1b 1c 1d 1e 1f */ |
169 | 0x0e, 0x0f, -1, -1, 0x10, 0x11, 0x12, 0x13, |
170 | |
171 | /* 20 */ |
172 | -1 |
173 | }; |
174 | |
175 | /* |
176 | * Byte value map for control codes, |
177 | * from trail byte values 0..19 (0..0x13) as used in the difference calculation |
178 | * to external byte values 0x00..0x20. |
179 | */ |
180 | static const int8_t |
181 | bocu1TrailToByte[BOCU1_TRAIL_CONTROLS_COUNT]={ |
182 | /* 0 1 2 3 4 5 6 7 */ |
183 | 0x01, 0x02, 0x03, 0x04, 0x05, 0x06, 0x10, 0x11, |
184 | |
185 | /* 8 9 a b c d e f */ |
186 | 0x12, 0x13, 0x14, 0x15, 0x16, 0x17, 0x18, 0x19, |
187 | |
188 | /* 10 11 12 13 */ |
189 | 0x1c, 0x1d, 0x1e, 0x1f |
190 | }; |
191 | |
192 | /** |
193 | * Integer division and modulo with negative numerators |
194 | * yields negative modulo results and quotients that are one more than |
195 | * what we need here. |
196 | * This macro adjust the results so that the modulo-value m is always >=0. |
197 | * |
198 | * For positive n, the if() condition is always false. |
199 | * |
200 | * @param n Number to be split into quotient and rest. |
201 | * Will be modified to contain the quotient. |
202 | * @param d Divisor. |
203 | * @param m Output variable for the rest (modulo result). |
204 | */ |
205 | #define NEGDIVMOD(n, d, m) UPRV_BLOCK_MACRO_BEGIN { \ |
206 | (m)=(n)%(d); \ |
207 | (n)/=(d); \ |
208 | if((m)<0) { \ |
209 | --(n); \ |
210 | (m)+=(d); \ |
211 | } \ |
212 | } UPRV_BLOCK_MACRO_END |
213 | |
214 | /* Faster versions of packDiff() for single-byte-encoded diff values. */ |
215 | |
216 | /** Is a diff value encodable in a single byte? */ |
217 | #define DIFF_IS_SINGLE(diff) (BOCU1_REACH_NEG_1<=(diff) && (diff)<=BOCU1_REACH_POS_1) |
218 | |
219 | /** Encode a diff value in a single byte. */ |
220 | #define PACK_SINGLE_DIFF(diff) (BOCU1_MIDDLE+(diff)) |
221 | |
222 | /** Is a diff value encodable in two bytes? */ |
223 | #define DIFF_IS_DOUBLE(diff) (BOCU1_REACH_NEG_2<=(diff) && (diff)<=BOCU1_REACH_POS_2) |
224 | |
225 | /* BOCU-1 implementation functions ------------------------------------------ */ |
226 | |
227 | #define BOCU1_SIMPLE_PREV(c) (((c)&~0x7f)+BOCU1_ASCII_PREV) |
228 | |
229 | /** |
230 | * Compute the next "previous" value for differencing |
231 | * from the current code point. |
232 | * |
233 | * @param c current code point, 0x3040..0xd7a3 (rest handled by macro below) |
234 | * @return "previous code point" state value |
235 | */ |
236 | static inline int32_t |
237 | bocu1Prev(int32_t c) { |
238 | /* compute new prev */ |
239 | if(/* 0x3040<=c && */ c<=0x309f) { |
240 | /* Hiragana is not 128-aligned */ |
241 | return 0x3070; |
242 | } else if(0x4e00<=c && c<=0x9fa5) { |
243 | /* CJK Unihan */ |
244 | return 0x4e00-BOCU1_REACH_NEG_2; |
245 | } else if(0xac00<=c /* && c<=0xd7a3 */) { |
246 | /* Korean Hangul */ |
247 | return (0xd7a3+0xac00)/2; |
248 | } else { |
249 | /* mostly small scripts */ |
250 | return BOCU1_SIMPLE_PREV(c); |
251 | } |
252 | } |
253 | |
254 | /** Fast version of bocu1Prev() for most scripts. */ |
255 | #define BOCU1_PREV(c) ((c)<0x3040 || (c)>0xd7a3 ? BOCU1_SIMPLE_PREV(c) : bocu1Prev(c)) |
256 | |
257 | /* |
258 | * The BOCU-1 converter uses the standard setup code in ucnv.c/ucnv_bld.c. |
259 | * The UConverter fields are used as follows: |
260 | * |
261 | * fromUnicodeStatus encoder's prev (0 will be interpreted as BOCU1_ASCII_PREV) |
262 | * |
263 | * toUnicodeStatus decoder's prev (0 will be interpreted as BOCU1_ASCII_PREV) |
264 | * mode decoder's incomplete (diff<<2)|count (ignored when toULength==0) |
265 | */ |
266 | |
267 | /* BOCU-1-from-Unicode conversion functions --------------------------------- */ |
268 | |
269 | /** |
270 | * Encode a difference -0x10ffff..0x10ffff in 1..4 bytes |
271 | * and return a packed integer with them. |
272 | * |
273 | * The encoding favors small absolute differences with short encodings |
274 | * to compress runs of same-script characters. |
275 | * |
276 | * Optimized version with unrolled loops and fewer floating-point operations |
277 | * than the standard packDiff(). |
278 | * |
279 | * @param diff difference value -0x10ffff..0x10ffff |
280 | * @return |
281 | * 0x010000zz for 1-byte sequence zz |
282 | * 0x0200yyzz for 2-byte sequence yy zz |
283 | * 0x03xxyyzz for 3-byte sequence xx yy zz |
284 | * 0xwwxxyyzz for 4-byte sequence ww xx yy zz (ww>0x03) |
285 | */ |
286 | static int32_t |
287 | packDiff(int32_t diff) { |
288 | int32_t result, m; |
289 | |
290 | U_ASSERT(!DIFF_IS_SINGLE(diff)); /* assume we won't be called where diff==BOCU1_REACH_NEG_1=-64 */ |
291 | if(diff>=BOCU1_REACH_NEG_1) { |
292 | /* mostly positive differences, and single-byte negative ones */ |
293 | #if 0 /* single-byte case handled in macros, see below */ |
294 | if(diff<=BOCU1_REACH_POS_1) { |
295 | /* single byte */ |
296 | return 0x01000000|(BOCU1_MIDDLE+diff); |
297 | } else |
298 | #endif |
299 | if(diff<=BOCU1_REACH_POS_2) { |
300 | /* two bytes */ |
301 | diff-=BOCU1_REACH_POS_1+1; |
302 | result=0x02000000; |
303 | |
304 | m=diff%BOCU1_TRAIL_COUNT; |
305 | diff/=BOCU1_TRAIL_COUNT; |
306 | result|=BOCU1_TRAIL_TO_BYTE(m); |
307 | |
308 | result|=(BOCU1_START_POS_2+diff)<<8; |
309 | } else if(diff<=BOCU1_REACH_POS_3) { |
310 | /* three bytes */ |
311 | diff-=BOCU1_REACH_POS_2+1; |
312 | result=0x03000000; |
313 | |
314 | m=diff%BOCU1_TRAIL_COUNT; |
315 | diff/=BOCU1_TRAIL_COUNT; |
316 | result|=BOCU1_TRAIL_TO_BYTE(m); |
317 | |
318 | m=diff%BOCU1_TRAIL_COUNT; |
319 | diff/=BOCU1_TRAIL_COUNT; |
320 | result|=BOCU1_TRAIL_TO_BYTE(m)<<8; |
321 | |
322 | result|=(BOCU1_START_POS_3+diff)<<16; |
323 | } else { |
324 | /* four bytes */ |
325 | diff-=BOCU1_REACH_POS_3+1; |
326 | |
327 | m=diff%BOCU1_TRAIL_COUNT; |
328 | diff/=BOCU1_TRAIL_COUNT; |
329 | result=BOCU1_TRAIL_TO_BYTE(m); |
330 | |
331 | m=diff%BOCU1_TRAIL_COUNT; |
332 | diff/=BOCU1_TRAIL_COUNT; |
333 | result|=BOCU1_TRAIL_TO_BYTE(m)<<8; |
334 | |
335 | /* |
336 | * We know that / and % would deliver quotient 0 and rest=diff. |
337 | * Avoid division and modulo for performance. |
338 | */ |
339 | result|=BOCU1_TRAIL_TO_BYTE(diff)<<16; |
340 | |
341 | result|=((uint32_t)BOCU1_START_POS_4)<<24; |
342 | } |
343 | } else { |
344 | /* two- to four-byte negative differences */ |
345 | if(diff>=BOCU1_REACH_NEG_2) { |
346 | /* two bytes */ |
347 | diff-=BOCU1_REACH_NEG_1; |
348 | result=0x02000000; |
349 | |
350 | NEGDIVMOD(diff, BOCU1_TRAIL_COUNT, m); |
351 | result|=BOCU1_TRAIL_TO_BYTE(m); |
352 | |
353 | result|=(BOCU1_START_NEG_2+diff)<<8; |
354 | } else if(diff>=BOCU1_REACH_NEG_3) { |
355 | /* three bytes */ |
356 | diff-=BOCU1_REACH_NEG_2; |
357 | result=0x03000000; |
358 | |
359 | NEGDIVMOD(diff, BOCU1_TRAIL_COUNT, m); |
360 | result|=BOCU1_TRAIL_TO_BYTE(m); |
361 | |
362 | NEGDIVMOD(diff, BOCU1_TRAIL_COUNT, m); |
363 | result|=BOCU1_TRAIL_TO_BYTE(m)<<8; |
364 | |
365 | result|=(BOCU1_START_NEG_3+diff)<<16; |
366 | } else { |
367 | /* four bytes */ |
368 | diff-=BOCU1_REACH_NEG_3; |
369 | |
370 | NEGDIVMOD(diff, BOCU1_TRAIL_COUNT, m); |
371 | result=BOCU1_TRAIL_TO_BYTE(m); |
372 | |
373 | NEGDIVMOD(diff, BOCU1_TRAIL_COUNT, m); |
374 | result|=BOCU1_TRAIL_TO_BYTE(m)<<8; |
375 | |
376 | /* |
377 | * We know that NEGDIVMOD would deliver |
378 | * quotient -1 and rest=diff+BOCU1_TRAIL_COUNT. |
379 | * Avoid division and modulo for performance. |
380 | */ |
381 | m=diff+BOCU1_TRAIL_COUNT; |
382 | result|=BOCU1_TRAIL_TO_BYTE(m)<<16; |
383 | |
384 | result|=BOCU1_MIN<<24; |
385 | } |
386 | } |
387 | return result; |
388 | } |
389 | |
390 | |
391 | static void U_CALLCONV |
392 | _Bocu1FromUnicodeWithOffsets(UConverterFromUnicodeArgs *pArgs, |
393 | UErrorCode *pErrorCode) { |
394 | UConverter *cnv; |
395 | const char16_t *source, *sourceLimit; |
396 | uint8_t *target; |
397 | int32_t targetCapacity; |
398 | int32_t *offsets; |
399 | |
400 | int32_t prev, c, diff; |
401 | |
402 | int32_t sourceIndex, nextSourceIndex; |
403 | |
404 | /* set up the local pointers */ |
405 | cnv=pArgs->converter; |
406 | source=pArgs->source; |
407 | sourceLimit=pArgs->sourceLimit; |
408 | target=(uint8_t *)pArgs->target; |
409 | targetCapacity=(int32_t)(pArgs->targetLimit-pArgs->target); |
410 | offsets=pArgs->offsets; |
411 | |
412 | /* get the converter state from UConverter */ |
413 | c=cnv->fromUChar32; |
414 | prev=(int32_t)cnv->fromUnicodeStatus; |
415 | if(prev==0) { |
416 | prev=BOCU1_ASCII_PREV; |
417 | } |
418 | |
419 | /* sourceIndex=-1 if the current character began in the previous buffer */ |
420 | sourceIndex= c==0 ? 0 : -1; |
421 | nextSourceIndex=0; |
422 | |
423 | /* conversion loop */ |
424 | if(c!=0 && targetCapacity>0) { |
425 | goto getTrail; |
426 | } |
427 | |
428 | fastSingle: |
429 | /* fast loop for single-byte differences */ |
430 | /* use only one loop counter variable, targetCapacity, not also source */ |
431 | diff=(int32_t)(sourceLimit-source); |
432 | if(targetCapacity>diff) { |
433 | targetCapacity=diff; |
434 | } |
435 | while(targetCapacity>0 && (c=*source)<0x3000) { |
436 | if(c<=0x20) { |
437 | if(c!=0x20) { |
438 | prev=BOCU1_ASCII_PREV; |
439 | } |
440 | *target++=(uint8_t)c; |
441 | *offsets++=nextSourceIndex++; |
442 | ++source; |
443 | --targetCapacity; |
444 | } else { |
445 | diff=c-prev; |
446 | if(DIFF_IS_SINGLE(diff)) { |
447 | prev=BOCU1_SIMPLE_PREV(c); |
448 | *target++=(uint8_t)PACK_SINGLE_DIFF(diff); |
449 | *offsets++=nextSourceIndex++; |
450 | ++source; |
451 | --targetCapacity; |
452 | } else { |
453 | break; |
454 | } |
455 | } |
456 | } |
457 | /* restore real values */ |
458 | targetCapacity=(int32_t)((const uint8_t *)pArgs->targetLimit-target); |
459 | sourceIndex=nextSourceIndex; /* wrong if offsets==nullptr but does not matter */ |
460 | |
461 | /* regular loop for all cases */ |
462 | while(source<sourceLimit) { |
463 | if(targetCapacity>0) { |
464 | c=*source++; |
465 | ++nextSourceIndex; |
466 | |
467 | if(c<=0x20) { |
468 | /* |
469 | * ISO C0 control & space: |
470 | * Encode directly for MIME compatibility, |
471 | * and reset state except for space, to not disrupt compression. |
472 | */ |
473 | if(c!=0x20) { |
474 | prev=BOCU1_ASCII_PREV; |
475 | } |
476 | *target++=(uint8_t)c; |
477 | *offsets++=sourceIndex; |
478 | --targetCapacity; |
479 | |
480 | sourceIndex=nextSourceIndex; |
481 | continue; |
482 | } |
483 | |
484 | if(U16_IS_LEAD(c)) { |
485 | getTrail: |
486 | if(source<sourceLimit) { |
487 | /* test the following code unit */ |
488 | char16_t trail=*source; |
489 | if(U16_IS_TRAIL(trail)) { |
490 | ++source; |
491 | ++nextSourceIndex; |
492 | c=U16_GET_SUPPLEMENTARY(c, trail); |
493 | } |
494 | } else { |
495 | /* no more input */ |
496 | c=-c; /* negative lead surrogate as "incomplete" indicator to avoid c=0 everywhere else */ |
497 | break; |
498 | } |
499 | } |
500 | |
501 | /* |
502 | * all other Unicode code points c==U+0021..U+10ffff |
503 | * are encoded with the difference c-prev |
504 | * |
505 | * a new prev is computed from c, |
506 | * placed in the middle of a 0x80-block (for most small scripts) or |
507 | * in the middle of the Unihan and Hangul blocks |
508 | * to statistically minimize the following difference |
509 | */ |
510 | diff=c-prev; |
511 | prev=BOCU1_PREV(c); |
512 | if(DIFF_IS_SINGLE(diff)) { |
513 | *target++=(uint8_t)PACK_SINGLE_DIFF(diff); |
514 | *offsets++=sourceIndex; |
515 | --targetCapacity; |
516 | sourceIndex=nextSourceIndex; |
517 | if(c<0x3000) { |
518 | goto fastSingle; |
519 | } |
520 | } else if(DIFF_IS_DOUBLE(diff) && 2<=targetCapacity) { |
521 | /* optimize 2-byte case */ |
522 | int32_t m; |
523 | |
524 | if(diff>=0) { |
525 | diff-=BOCU1_REACH_POS_1+1; |
526 | m=diff%BOCU1_TRAIL_COUNT; |
527 | diff/=BOCU1_TRAIL_COUNT; |
528 | diff+=BOCU1_START_POS_2; |
529 | } else { |
530 | diff-=BOCU1_REACH_NEG_1; |
531 | NEGDIVMOD(diff, BOCU1_TRAIL_COUNT, m); |
532 | diff+=BOCU1_START_NEG_2; |
533 | } |
534 | *target++=(uint8_t)diff; |
535 | *target++=(uint8_t)BOCU1_TRAIL_TO_BYTE(m); |
536 | *offsets++=sourceIndex; |
537 | *offsets++=sourceIndex; |
538 | targetCapacity-=2; |
539 | sourceIndex=nextSourceIndex; |
540 | } else { |
541 | int32_t length; /* will be 2..4 */ |
542 | |
543 | diff=packDiff(diff); |
544 | length=BOCU1_LENGTH_FROM_PACKED(diff); |
545 | |
546 | /* write the output character bytes from diff and length */ |
547 | /* from the first if in the loop we know that targetCapacity>0 */ |
548 | if(length<=targetCapacity) { |
549 | switch(length) { |
550 | /* each branch falls through to the next one */ |
551 | case 4: |
552 | *target++=(uint8_t)(diff>>24); |
553 | *offsets++=sourceIndex; |
554 | U_FALLTHROUGH; |
555 | case 3: |
556 | *target++=(uint8_t)(diff>>16); |
557 | *offsets++=sourceIndex; |
558 | U_FALLTHROUGH; |
559 | case 2: |
560 | *target++=(uint8_t)(diff>>8); |
561 | *offsets++=sourceIndex; |
562 | /* case 1: handled above */ |
563 | *target++=(uint8_t)diff; |
564 | *offsets++=sourceIndex; |
565 | U_FALLTHROUGH; |
566 | default: |
567 | /* will never occur */ |
568 | break; |
569 | } |
570 | targetCapacity-=length; |
571 | sourceIndex=nextSourceIndex; |
572 | } else { |
573 | uint8_t *charErrorBuffer; |
574 | |
575 | /* |
576 | * We actually do this backwards here: |
577 | * In order to save an intermediate variable, we output |
578 | * first to the overflow buffer what does not fit into the |
579 | * regular target. |
580 | */ |
581 | /* we know that 1<=targetCapacity<length<=4 */ |
582 | length-=targetCapacity; |
583 | charErrorBuffer=(uint8_t *)cnv->charErrorBuffer; |
584 | switch(length) { |
585 | /* each branch falls through to the next one */ |
586 | case 3: |
587 | *charErrorBuffer++=(uint8_t)(diff>>16); |
588 | U_FALLTHROUGH; |
589 | case 2: |
590 | *charErrorBuffer++=(uint8_t)(diff>>8); |
591 | U_FALLTHROUGH; |
592 | case 1: |
593 | *charErrorBuffer=(uint8_t)diff; |
594 | U_FALLTHROUGH; |
595 | default: |
596 | /* will never occur */ |
597 | break; |
598 | } |
599 | cnv->charErrorBufferLength=(int8_t)length; |
600 | |
601 | /* now output what fits into the regular target */ |
602 | diff>>=8*length; /* length was reduced by targetCapacity */ |
603 | switch(targetCapacity) { |
604 | /* each branch falls through to the next one */ |
605 | case 3: |
606 | *target++=(uint8_t)(diff>>16); |
607 | *offsets++=sourceIndex; |
608 | U_FALLTHROUGH; |
609 | case 2: |
610 | *target++=(uint8_t)(diff>>8); |
611 | *offsets++=sourceIndex; |
612 | U_FALLTHROUGH; |
613 | case 1: |
614 | *target++=(uint8_t)diff; |
615 | *offsets++=sourceIndex; |
616 | U_FALLTHROUGH; |
617 | default: |
618 | /* will never occur */ |
619 | break; |
620 | } |
621 | |
622 | /* target overflow */ |
623 | targetCapacity=0; |
624 | *pErrorCode=U_BUFFER_OVERFLOW_ERROR; |
625 | break; |
626 | } |
627 | } |
628 | } else { |
629 | /* target is full */ |
630 | *pErrorCode=U_BUFFER_OVERFLOW_ERROR; |
631 | break; |
632 | } |
633 | } |
634 | |
635 | /* set the converter state back into UConverter */ |
636 | cnv->fromUChar32= c<0 ? -c : 0; |
637 | cnv->fromUnicodeStatus=(uint32_t)prev; |
638 | |
639 | /* write back the updated pointers */ |
640 | pArgs->source=source; |
641 | pArgs->target=(char *)target; |
642 | pArgs->offsets=offsets; |
643 | } |
644 | |
645 | /* |
646 | * Identical to _Bocu1FromUnicodeWithOffsets but without offset handling. |
647 | * If a change is made in the original function, then either |
648 | * change this function the same way or |
649 | * re-copy the original function and remove the variables |
650 | * offsets, sourceIndex, and nextSourceIndex. |
651 | */ |
652 | static void U_CALLCONV |
653 | _Bocu1FromUnicode(UConverterFromUnicodeArgs *pArgs, |
654 | UErrorCode *pErrorCode) { |
655 | UConverter *cnv; |
656 | const char16_t *source, *sourceLimit; |
657 | uint8_t *target; |
658 | int32_t targetCapacity; |
659 | |
660 | int32_t prev, c, diff; |
661 | |
662 | /* set up the local pointers */ |
663 | cnv=pArgs->converter; |
664 | source=pArgs->source; |
665 | sourceLimit=pArgs->sourceLimit; |
666 | target=(uint8_t *)pArgs->target; |
667 | targetCapacity=(int32_t)(pArgs->targetLimit-pArgs->target); |
668 | |
669 | /* get the converter state from UConverter */ |
670 | c=cnv->fromUChar32; |
671 | prev=(int32_t)cnv->fromUnicodeStatus; |
672 | if(prev==0) { |
673 | prev=BOCU1_ASCII_PREV; |
674 | } |
675 | |
676 | /* conversion loop */ |
677 | if(c!=0 && targetCapacity>0) { |
678 | goto getTrail; |
679 | } |
680 | |
681 | fastSingle: |
682 | /* fast loop for single-byte differences */ |
683 | /* use only one loop counter variable, targetCapacity, not also source */ |
684 | diff=(int32_t)(sourceLimit-source); |
685 | if(targetCapacity>diff) { |
686 | targetCapacity=diff; |
687 | } |
688 | while(targetCapacity>0 && (c=*source)<0x3000) { |
689 | if(c<=0x20) { |
690 | if(c!=0x20) { |
691 | prev=BOCU1_ASCII_PREV; |
692 | } |
693 | *target++=(uint8_t)c; |
694 | } else { |
695 | diff=c-prev; |
696 | if(DIFF_IS_SINGLE(diff)) { |
697 | prev=BOCU1_SIMPLE_PREV(c); |
698 | *target++=(uint8_t)PACK_SINGLE_DIFF(diff); |
699 | } else { |
700 | break; |
701 | } |
702 | } |
703 | ++source; |
704 | --targetCapacity; |
705 | } |
706 | /* restore real values */ |
707 | targetCapacity=(int32_t)((const uint8_t *)pArgs->targetLimit-target); |
708 | |
709 | /* regular loop for all cases */ |
710 | while(source<sourceLimit) { |
711 | if(targetCapacity>0) { |
712 | c=*source++; |
713 | |
714 | if(c<=0x20) { |
715 | /* |
716 | * ISO C0 control & space: |
717 | * Encode directly for MIME compatibility, |
718 | * and reset state except for space, to not disrupt compression. |
719 | */ |
720 | if(c!=0x20) { |
721 | prev=BOCU1_ASCII_PREV; |
722 | } |
723 | *target++=(uint8_t)c; |
724 | --targetCapacity; |
725 | continue; |
726 | } |
727 | |
728 | if(U16_IS_LEAD(c)) { |
729 | getTrail: |
730 | if(source<sourceLimit) { |
731 | /* test the following code unit */ |
732 | char16_t trail=*source; |
733 | if(U16_IS_TRAIL(trail)) { |
734 | ++source; |
735 | c=U16_GET_SUPPLEMENTARY(c, trail); |
736 | } |
737 | } else { |
738 | /* no more input */ |
739 | c=-c; /* negative lead surrogate as "incomplete" indicator to avoid c=0 everywhere else */ |
740 | break; |
741 | } |
742 | } |
743 | |
744 | /* |
745 | * all other Unicode code points c==U+0021..U+10ffff |
746 | * are encoded with the difference c-prev |
747 | * |
748 | * a new prev is computed from c, |
749 | * placed in the middle of a 0x80-block (for most small scripts) or |
750 | * in the middle of the Unihan and Hangul blocks |
751 | * to statistically minimize the following difference |
752 | */ |
753 | diff=c-prev; |
754 | prev=BOCU1_PREV(c); |
755 | if(DIFF_IS_SINGLE(diff)) { |
756 | *target++=(uint8_t)PACK_SINGLE_DIFF(diff); |
757 | --targetCapacity; |
758 | if(c<0x3000) { |
759 | goto fastSingle; |
760 | } |
761 | } else if(DIFF_IS_DOUBLE(diff) && 2<=targetCapacity) { |
762 | /* optimize 2-byte case */ |
763 | int32_t m; |
764 | |
765 | if(diff>=0) { |
766 | diff-=BOCU1_REACH_POS_1+1; |
767 | m=diff%BOCU1_TRAIL_COUNT; |
768 | diff/=BOCU1_TRAIL_COUNT; |
769 | diff+=BOCU1_START_POS_2; |
770 | } else { |
771 | diff-=BOCU1_REACH_NEG_1; |
772 | NEGDIVMOD(diff, BOCU1_TRAIL_COUNT, m); |
773 | diff+=BOCU1_START_NEG_2; |
774 | } |
775 | *target++=(uint8_t)diff; |
776 | *target++=(uint8_t)BOCU1_TRAIL_TO_BYTE(m); |
777 | targetCapacity-=2; |
778 | } else { |
779 | int32_t length; /* will be 2..4 */ |
780 | |
781 | diff=packDiff(diff); |
782 | length=BOCU1_LENGTH_FROM_PACKED(diff); |
783 | |
784 | /* write the output character bytes from diff and length */ |
785 | /* from the first if in the loop we know that targetCapacity>0 */ |
786 | if(length<=targetCapacity) { |
787 | switch(length) { |
788 | /* each branch falls through to the next one */ |
789 | case 4: |
790 | *target++=(uint8_t)(diff>>24); |
791 | U_FALLTHROUGH; |
792 | case 3: |
793 | *target++=(uint8_t)(diff>>16); |
794 | /* case 2: handled above */ |
795 | *target++=(uint8_t)(diff>>8); |
796 | /* case 1: handled above */ |
797 | *target++=(uint8_t)diff; |
798 | U_FALLTHROUGH; |
799 | default: |
800 | /* will never occur */ |
801 | break; |
802 | } |
803 | targetCapacity-=length; |
804 | } else { |
805 | uint8_t *charErrorBuffer; |
806 | |
807 | /* |
808 | * We actually do this backwards here: |
809 | * In order to save an intermediate variable, we output |
810 | * first to the overflow buffer what does not fit into the |
811 | * regular target. |
812 | */ |
813 | /* we know that 1<=targetCapacity<length<=4 */ |
814 | length-=targetCapacity; |
815 | charErrorBuffer=(uint8_t *)cnv->charErrorBuffer; |
816 | switch(length) { |
817 | /* each branch falls through to the next one */ |
818 | case 3: |
819 | *charErrorBuffer++=(uint8_t)(diff>>16); |
820 | U_FALLTHROUGH; |
821 | case 2: |
822 | *charErrorBuffer++=(uint8_t)(diff>>8); |
823 | U_FALLTHROUGH; |
824 | case 1: |
825 | *charErrorBuffer=(uint8_t)diff; |
826 | U_FALLTHROUGH; |
827 | default: |
828 | /* will never occur */ |
829 | break; |
830 | } |
831 | cnv->charErrorBufferLength=(int8_t)length; |
832 | |
833 | /* now output what fits into the regular target */ |
834 | diff>>=8*length; /* length was reduced by targetCapacity */ |
835 | switch(targetCapacity) { |
836 | /* each branch falls through to the next one */ |
837 | case 3: |
838 | *target++=(uint8_t)(diff>>16); |
839 | U_FALLTHROUGH; |
840 | case 2: |
841 | *target++=(uint8_t)(diff>>8); |
842 | U_FALLTHROUGH; |
843 | case 1: |
844 | *target++=(uint8_t)diff; |
845 | U_FALLTHROUGH; |
846 | default: |
847 | /* will never occur */ |
848 | break; |
849 | } |
850 | |
851 | /* target overflow */ |
852 | targetCapacity=0; |
853 | *pErrorCode=U_BUFFER_OVERFLOW_ERROR; |
854 | break; |
855 | } |
856 | } |
857 | } else { |
858 | /* target is full */ |
859 | *pErrorCode=U_BUFFER_OVERFLOW_ERROR; |
860 | break; |
861 | } |
862 | } |
863 | |
864 | /* set the converter state back into UConverter */ |
865 | cnv->fromUChar32= c<0 ? -c : 0; |
866 | cnv->fromUnicodeStatus=(uint32_t)prev; |
867 | |
868 | /* write back the updated pointers */ |
869 | pArgs->source=source; |
870 | pArgs->target=(char *)target; |
871 | } |
872 | |
873 | /* BOCU-1-to-Unicode conversion functions ----------------------------------- */ |
874 | |
875 | /** |
876 | * Function for BOCU-1 decoder; handles multi-byte lead bytes. |
877 | * |
878 | * @param b lead byte; |
879 | * BOCU1_MIN<=b<BOCU1_START_NEG_2 or BOCU1_START_POS_2<=b<BOCU1_MAX_LEAD |
880 | * @return (diff<<2)|count |
881 | */ |
882 | static inline int32_t |
883 | decodeBocu1LeadByte(int32_t b) { |
884 | int32_t diff, count; |
885 | |
886 | if(b>=BOCU1_START_NEG_2) { |
887 | /* positive difference */ |
888 | if(b<BOCU1_START_POS_3) { |
889 | /* two bytes */ |
890 | diff=((int32_t)b-BOCU1_START_POS_2)*BOCU1_TRAIL_COUNT+BOCU1_REACH_POS_1+1; |
891 | count=1; |
892 | } else if(b<BOCU1_START_POS_4) { |
893 | /* three bytes */ |
894 | diff=((int32_t)b-BOCU1_START_POS_3)*BOCU1_TRAIL_COUNT*BOCU1_TRAIL_COUNT+BOCU1_REACH_POS_2+1; |
895 | count=2; |
896 | } else { |
897 | /* four bytes */ |
898 | diff=BOCU1_REACH_POS_3+1; |
899 | count=3; |
900 | } |
901 | } else { |
902 | /* negative difference */ |
903 | if(b>=BOCU1_START_NEG_3) { |
904 | /* two bytes */ |
905 | diff=((int32_t)b-BOCU1_START_NEG_2)*BOCU1_TRAIL_COUNT+BOCU1_REACH_NEG_1; |
906 | count=1; |
907 | } else if(b>BOCU1_MIN) { |
908 | /* three bytes */ |
909 | diff=((int32_t)b-BOCU1_START_NEG_3)*BOCU1_TRAIL_COUNT*BOCU1_TRAIL_COUNT+BOCU1_REACH_NEG_2; |
910 | count=2; |
911 | } else { |
912 | /* four bytes */ |
913 | diff=-BOCU1_TRAIL_COUNT*BOCU1_TRAIL_COUNT*BOCU1_TRAIL_COUNT+BOCU1_REACH_NEG_3; |
914 | count=3; |
915 | } |
916 | } |
917 | |
918 | /* return the state for decoding the trail byte(s) */ |
919 | return ((uint32_t)diff<<2)|count; |
920 | } |
921 | |
922 | /** |
923 | * Function for BOCU-1 decoder; handles multi-byte trail bytes. |
924 | * |
925 | * @param count number of remaining trail bytes including this one |
926 | * @param b trail byte |
927 | * @return new delta for diff including b - <0 indicates an error |
928 | * |
929 | * @see decodeBocu1 |
930 | */ |
931 | static inline int32_t |
932 | decodeBocu1TrailByte(int32_t count, int32_t b) { |
933 | if(b<=0x20) { |
934 | /* skip some C0 controls and make the trail byte range contiguous */ |
935 | b=bocu1ByteToTrail[b]; |
936 | /* b<0 for an illegal trail byte value will result in return<0 below */ |
937 | #if BOCU1_MAX_TRAIL<0xff |
938 | } else if(b>BOCU1_MAX_TRAIL) { |
939 | return -99; |
940 | #endif |
941 | } else { |
942 | b-=BOCU1_TRAIL_BYTE_OFFSET; |
943 | } |
944 | |
945 | /* add trail byte into difference and decrement count */ |
946 | if(count==1) { |
947 | return b; |
948 | } else if(count==2) { |
949 | return b*BOCU1_TRAIL_COUNT; |
950 | } else /* count==3 */ { |
951 | return b*(BOCU1_TRAIL_COUNT*BOCU1_TRAIL_COUNT); |
952 | } |
953 | } |
954 | |
955 | static void U_CALLCONV |
956 | _Bocu1ToUnicodeWithOffsets(UConverterToUnicodeArgs *pArgs, |
957 | UErrorCode *pErrorCode) { |
958 | UConverter *cnv; |
959 | const uint8_t *source, *sourceLimit; |
960 | char16_t *target; |
961 | const char16_t *targetLimit; |
962 | int32_t *offsets; |
963 | |
964 | int32_t prev, count, diff, c; |
965 | |
966 | int8_t byteIndex; |
967 | uint8_t *bytes; |
968 | |
969 | int32_t sourceIndex, nextSourceIndex; |
970 | |
971 | /* set up the local pointers */ |
972 | cnv=pArgs->converter; |
973 | source=(const uint8_t *)pArgs->source; |
974 | sourceLimit=(const uint8_t *)pArgs->sourceLimit; |
975 | target=pArgs->target; |
976 | targetLimit=pArgs->targetLimit; |
977 | offsets=pArgs->offsets; |
978 | |
979 | /* get the converter state from UConverter */ |
980 | prev=(int32_t)cnv->toUnicodeStatus; |
981 | if(prev==0) { |
982 | prev=BOCU1_ASCII_PREV; |
983 | } |
984 | diff=cnv->mode; /* mode may be set to UCNV_SI by ucnv_bld.c but then toULength==0 */ |
985 | count=diff&3; |
986 | diff>>=2; |
987 | |
988 | byteIndex=cnv->toULength; |
989 | bytes=cnv->toUBytes; |
990 | |
991 | /* sourceIndex=-1 if the current character began in the previous buffer */ |
992 | sourceIndex=byteIndex==0 ? 0 : -1; |
993 | nextSourceIndex=0; |
994 | |
995 | /* conversion "loop" similar to _SCSUToUnicodeWithOffsets() */ |
996 | if(count>0 && byteIndex>0 && target<targetLimit) { |
997 | goto getTrail; |
998 | } |
999 | |
1000 | fastSingle: |
1001 | /* fast loop for single-byte differences */ |
1002 | /* use count as the only loop counter variable */ |
1003 | diff=(int32_t)(sourceLimit-source); |
1004 | count=(int32_t)(pArgs->targetLimit-target); |
1005 | if(count>diff) { |
1006 | count=diff; |
1007 | } |
1008 | while(count>0) { |
1009 | if(BOCU1_START_NEG_2<=(c=*source) && c<BOCU1_START_POS_2) { |
1010 | c=prev+(c-BOCU1_MIDDLE); |
1011 | if(c<0x3000) { |
1012 | *target++=(char16_t)c; |
1013 | *offsets++=nextSourceIndex++; |
1014 | prev=BOCU1_SIMPLE_PREV(c); |
1015 | } else { |
1016 | break; |
1017 | } |
1018 | } else if(c<=0x20) { |
1019 | if(c!=0x20) { |
1020 | prev=BOCU1_ASCII_PREV; |
1021 | } |
1022 | *target++=(char16_t)c; |
1023 | *offsets++=nextSourceIndex++; |
1024 | } else { |
1025 | break; |
1026 | } |
1027 | ++source; |
1028 | --count; |
1029 | } |
1030 | sourceIndex=nextSourceIndex; /* wrong if offsets==nullptr but does not matter */ |
1031 | |
1032 | /* decode a sequence of single and lead bytes */ |
1033 | while(source<sourceLimit) { |
1034 | if(target>=targetLimit) { |
1035 | /* target is full */ |
1036 | *pErrorCode=U_BUFFER_OVERFLOW_ERROR; |
1037 | break; |
1038 | } |
1039 | |
1040 | ++nextSourceIndex; |
1041 | c=*source++; |
1042 | if(BOCU1_START_NEG_2<=c && c<BOCU1_START_POS_2) { |
1043 | /* Write a code point directly from a single-byte difference. */ |
1044 | c=prev+(c-BOCU1_MIDDLE); |
1045 | if(c<0x3000) { |
1046 | *target++=(char16_t)c; |
1047 | *offsets++=sourceIndex; |
1048 | prev=BOCU1_SIMPLE_PREV(c); |
1049 | sourceIndex=nextSourceIndex; |
1050 | goto fastSingle; |
1051 | } |
1052 | } else if(c<=0x20) { |
1053 | /* |
1054 | * Direct-encoded C0 control code or space. |
1055 | * Reset prev for C0 control codes but not for space. |
1056 | */ |
1057 | if(c!=0x20) { |
1058 | prev=BOCU1_ASCII_PREV; |
1059 | } |
1060 | *target++=(char16_t)c; |
1061 | *offsets++=sourceIndex; |
1062 | sourceIndex=nextSourceIndex; |
1063 | continue; |
1064 | } else if(BOCU1_START_NEG_3<=c && c<BOCU1_START_POS_3 && source<sourceLimit) { |
1065 | /* Optimize two-byte case. */ |
1066 | if(c>=BOCU1_MIDDLE) { |
1067 | diff=((int32_t)c-BOCU1_START_POS_2)*BOCU1_TRAIL_COUNT+BOCU1_REACH_POS_1+1; |
1068 | } else { |
1069 | diff=((int32_t)c-BOCU1_START_NEG_2)*BOCU1_TRAIL_COUNT+BOCU1_REACH_NEG_1; |
1070 | } |
1071 | |
1072 | /* trail byte */ |
1073 | ++nextSourceIndex; |
1074 | c=decodeBocu1TrailByte(1, *source++); |
1075 | if(c<0 || (uint32_t)(c=prev+diff+c)>0x10ffff) { |
1076 | bytes[0]=source[-2]; |
1077 | bytes[1]=source[-1]; |
1078 | byteIndex=2; |
1079 | *pErrorCode=U_ILLEGAL_CHAR_FOUND; |
1080 | break; |
1081 | } |
1082 | } else if(c==BOCU1_RESET) { |
1083 | /* only reset the state, no code point */ |
1084 | prev=BOCU1_ASCII_PREV; |
1085 | sourceIndex=nextSourceIndex; |
1086 | continue; |
1087 | } else { |
1088 | /* |
1089 | * For multi-byte difference lead bytes, set the decoder state |
1090 | * with the partial difference value from the lead byte and |
1091 | * with the number of trail bytes. |
1092 | */ |
1093 | bytes[0]=(uint8_t)c; |
1094 | byteIndex=1; |
1095 | |
1096 | diff=decodeBocu1LeadByte(c); |
1097 | count=diff&3; |
1098 | diff>>=2; |
1099 | getTrail: |
1100 | for(;;) { |
1101 | if(source>=sourceLimit) { |
1102 | goto endloop; |
1103 | } |
1104 | ++nextSourceIndex; |
1105 | c=bytes[byteIndex++]=*source++; |
1106 | |
1107 | /* trail byte in any position */ |
1108 | c=decodeBocu1TrailByte(count, c); |
1109 | if(c<0) { |
1110 | *pErrorCode=U_ILLEGAL_CHAR_FOUND; |
1111 | goto endloop; |
1112 | } |
1113 | |
1114 | diff+=c; |
1115 | if(--count==0) { |
1116 | /* final trail byte, deliver a code point */ |
1117 | byteIndex=0; |
1118 | c=prev+diff; |
1119 | if((uint32_t)c>0x10ffff) { |
1120 | *pErrorCode=U_ILLEGAL_CHAR_FOUND; |
1121 | goto endloop; |
1122 | } |
1123 | break; |
1124 | } |
1125 | } |
1126 | } |
1127 | |
1128 | /* calculate the next prev and output c */ |
1129 | prev=BOCU1_PREV(c); |
1130 | if(c<=0xffff) { |
1131 | *target++=(char16_t)c; |
1132 | *offsets++=sourceIndex; |
1133 | } else { |
1134 | /* output surrogate pair */ |
1135 | *target++=U16_LEAD(c); |
1136 | if(target<targetLimit) { |
1137 | *target++=U16_TRAIL(c); |
1138 | *offsets++=sourceIndex; |
1139 | *offsets++=sourceIndex; |
1140 | } else { |
1141 | /* target overflow */ |
1142 | *offsets++=sourceIndex; |
1143 | cnv->UCharErrorBuffer[0]=U16_TRAIL(c); |
1144 | cnv->UCharErrorBufferLength=1; |
1145 | *pErrorCode=U_BUFFER_OVERFLOW_ERROR; |
1146 | break; |
1147 | } |
1148 | } |
1149 | sourceIndex=nextSourceIndex; |
1150 | } |
1151 | endloop: |
1152 | |
1153 | if(*pErrorCode==U_ILLEGAL_CHAR_FOUND) { |
1154 | /* set the converter state in UConverter to deal with the next character */ |
1155 | cnv->toUnicodeStatus=BOCU1_ASCII_PREV; |
1156 | cnv->mode=0; |
1157 | } else { |
1158 | /* set the converter state back into UConverter */ |
1159 | cnv->toUnicodeStatus=(uint32_t)prev; |
1160 | cnv->mode=(int32_t)((uint32_t)diff<<2)|count; |
1161 | } |
1162 | cnv->toULength=byteIndex; |
1163 | |
1164 | /* write back the updated pointers */ |
1165 | pArgs->source=(const char *)source; |
1166 | pArgs->target=target; |
1167 | pArgs->offsets=offsets; |
1168 | return; |
1169 | } |
1170 | |
1171 | /* |
1172 | * Identical to _Bocu1ToUnicodeWithOffsets but without offset handling. |
1173 | * If a change is made in the original function, then either |
1174 | * change this function the same way or |
1175 | * re-copy the original function and remove the variables |
1176 | * offsets, sourceIndex, and nextSourceIndex. |
1177 | */ |
1178 | static void U_CALLCONV |
1179 | _Bocu1ToUnicode(UConverterToUnicodeArgs *pArgs, |
1180 | UErrorCode *pErrorCode) { |
1181 | UConverter *cnv; |
1182 | const uint8_t *source, *sourceLimit; |
1183 | char16_t *target; |
1184 | const char16_t *targetLimit; |
1185 | |
1186 | int32_t prev, count, diff, c; |
1187 | |
1188 | int8_t byteIndex; |
1189 | uint8_t *bytes; |
1190 | |
1191 | /* set up the local pointers */ |
1192 | cnv=pArgs->converter; |
1193 | source=(const uint8_t *)pArgs->source; |
1194 | sourceLimit=(const uint8_t *)pArgs->sourceLimit; |
1195 | target=pArgs->target; |
1196 | targetLimit=pArgs->targetLimit; |
1197 | |
1198 | /* get the converter state from UConverter */ |
1199 | prev=(int32_t)cnv->toUnicodeStatus; |
1200 | if(prev==0) { |
1201 | prev=BOCU1_ASCII_PREV; |
1202 | } |
1203 | diff=cnv->mode; /* mode may be set to UCNV_SI by ucnv_bld.c but then toULength==0 */ |
1204 | count=diff&3; |
1205 | diff>>=2; |
1206 | |
1207 | byteIndex=cnv->toULength; |
1208 | bytes=cnv->toUBytes; |
1209 | |
1210 | /* conversion "loop" similar to _SCSUToUnicodeWithOffsets() */ |
1211 | if(count>0 && byteIndex>0 && target<targetLimit) { |
1212 | goto getTrail; |
1213 | } |
1214 | |
1215 | fastSingle: |
1216 | /* fast loop for single-byte differences */ |
1217 | /* use count as the only loop counter variable */ |
1218 | diff=(int32_t)(sourceLimit-source); |
1219 | count=(int32_t)(pArgs->targetLimit-target); |
1220 | if(count>diff) { |
1221 | count=diff; |
1222 | } |
1223 | while(count>0) { |
1224 | if(BOCU1_START_NEG_2<=(c=*source) && c<BOCU1_START_POS_2) { |
1225 | c=prev+(c-BOCU1_MIDDLE); |
1226 | if(c<0x3000) { |
1227 | *target++=(char16_t)c; |
1228 | prev=BOCU1_SIMPLE_PREV(c); |
1229 | } else { |
1230 | break; |
1231 | } |
1232 | } else if(c<=0x20) { |
1233 | if(c!=0x20) { |
1234 | prev=BOCU1_ASCII_PREV; |
1235 | } |
1236 | *target++=(char16_t)c; |
1237 | } else { |
1238 | break; |
1239 | } |
1240 | ++source; |
1241 | --count; |
1242 | } |
1243 | |
1244 | /* decode a sequence of single and lead bytes */ |
1245 | while(source<sourceLimit) { |
1246 | if(target>=targetLimit) { |
1247 | /* target is full */ |
1248 | *pErrorCode=U_BUFFER_OVERFLOW_ERROR; |
1249 | break; |
1250 | } |
1251 | |
1252 | c=*source++; |
1253 | if(BOCU1_START_NEG_2<=c && c<BOCU1_START_POS_2) { |
1254 | /* Write a code point directly from a single-byte difference. */ |
1255 | c=prev+(c-BOCU1_MIDDLE); |
1256 | if(c<0x3000) { |
1257 | *target++=(char16_t)c; |
1258 | prev=BOCU1_SIMPLE_PREV(c); |
1259 | goto fastSingle; |
1260 | } |
1261 | } else if(c<=0x20) { |
1262 | /* |
1263 | * Direct-encoded C0 control code or space. |
1264 | * Reset prev for C0 control codes but not for space. |
1265 | */ |
1266 | if(c!=0x20) { |
1267 | prev=BOCU1_ASCII_PREV; |
1268 | } |
1269 | *target++=(char16_t)c; |
1270 | continue; |
1271 | } else if(BOCU1_START_NEG_3<=c && c<BOCU1_START_POS_3 && source<sourceLimit) { |
1272 | /* Optimize two-byte case. */ |
1273 | if(c>=BOCU1_MIDDLE) { |
1274 | diff=((int32_t)c-BOCU1_START_POS_2)*BOCU1_TRAIL_COUNT+BOCU1_REACH_POS_1+1; |
1275 | } else { |
1276 | diff=((int32_t)c-BOCU1_START_NEG_2)*BOCU1_TRAIL_COUNT+BOCU1_REACH_NEG_1; |
1277 | } |
1278 | |
1279 | /* trail byte */ |
1280 | c=decodeBocu1TrailByte(1, *source++); |
1281 | if(c<0 || (uint32_t)(c=prev+diff+c)>0x10ffff) { |
1282 | bytes[0]=source[-2]; |
1283 | bytes[1]=source[-1]; |
1284 | byteIndex=2; |
1285 | *pErrorCode=U_ILLEGAL_CHAR_FOUND; |
1286 | break; |
1287 | } |
1288 | } else if(c==BOCU1_RESET) { |
1289 | /* only reset the state, no code point */ |
1290 | prev=BOCU1_ASCII_PREV; |
1291 | continue; |
1292 | } else { |
1293 | /* |
1294 | * For multi-byte difference lead bytes, set the decoder state |
1295 | * with the partial difference value from the lead byte and |
1296 | * with the number of trail bytes. |
1297 | */ |
1298 | bytes[0]=(uint8_t)c; |
1299 | byteIndex=1; |
1300 | |
1301 | diff=decodeBocu1LeadByte(c); |
1302 | count=diff&3; |
1303 | diff>>=2; |
1304 | getTrail: |
1305 | for(;;) { |
1306 | if(source>=sourceLimit) { |
1307 | goto endloop; |
1308 | } |
1309 | c=bytes[byteIndex++]=*source++; |
1310 | |
1311 | /* trail byte in any position */ |
1312 | c=decodeBocu1TrailByte(count, c); |
1313 | if(c<0) { |
1314 | *pErrorCode=U_ILLEGAL_CHAR_FOUND; |
1315 | goto endloop; |
1316 | } |
1317 | |
1318 | diff+=c; |
1319 | if(--count==0) { |
1320 | /* final trail byte, deliver a code point */ |
1321 | byteIndex=0; |
1322 | c=prev+diff; |
1323 | if((uint32_t)c>0x10ffff) { |
1324 | *pErrorCode=U_ILLEGAL_CHAR_FOUND; |
1325 | goto endloop; |
1326 | } |
1327 | break; |
1328 | } |
1329 | } |
1330 | } |
1331 | |
1332 | /* calculate the next prev and output c */ |
1333 | prev=BOCU1_PREV(c); |
1334 | if(c<=0xffff) { |
1335 | *target++=(char16_t)c; |
1336 | } else { |
1337 | /* output surrogate pair */ |
1338 | *target++=U16_LEAD(c); |
1339 | if(target<targetLimit) { |
1340 | *target++=U16_TRAIL(c); |
1341 | } else { |
1342 | /* target overflow */ |
1343 | cnv->UCharErrorBuffer[0]=U16_TRAIL(c); |
1344 | cnv->UCharErrorBufferLength=1; |
1345 | *pErrorCode=U_BUFFER_OVERFLOW_ERROR; |
1346 | break; |
1347 | } |
1348 | } |
1349 | } |
1350 | endloop: |
1351 | |
1352 | if(*pErrorCode==U_ILLEGAL_CHAR_FOUND) { |
1353 | /* set the converter state in UConverter to deal with the next character */ |
1354 | cnv->toUnicodeStatus=BOCU1_ASCII_PREV; |
1355 | cnv->mode=0; |
1356 | } else { |
1357 | /* set the converter state back into UConverter */ |
1358 | cnv->toUnicodeStatus=(uint32_t)prev; |
1359 | cnv->mode=((uint32_t)diff<<2)|count; |
1360 | } |
1361 | cnv->toULength=byteIndex; |
1362 | |
1363 | /* write back the updated pointers */ |
1364 | pArgs->source=(const char *)source; |
1365 | pArgs->target=target; |
1366 | return; |
1367 | } |
1368 | |
1369 | /* miscellaneous ------------------------------------------------------------ */ |
1370 | |
1371 | static const UConverterImpl _Bocu1Impl={ |
1372 | UCNV_BOCU1, |
1373 | |
1374 | nullptr, |
1375 | nullptr, |
1376 | |
1377 | nullptr, |
1378 | nullptr, |
1379 | nullptr, |
1380 | |
1381 | _Bocu1ToUnicode, |
1382 | _Bocu1ToUnicodeWithOffsets, |
1383 | _Bocu1FromUnicode, |
1384 | _Bocu1FromUnicodeWithOffsets, |
1385 | nullptr, |
1386 | |
1387 | nullptr, |
1388 | nullptr, |
1389 | nullptr, |
1390 | nullptr, |
1391 | ucnv_getCompleteUnicodeSet, |
1392 | |
1393 | nullptr, |
1394 | nullptr |
1395 | }; |
1396 | |
1397 | static const UConverterStaticData _Bocu1StaticData={ |
1398 | sizeof(UConverterStaticData), |
1399 | "BOCU-1" , |
1400 | 1214, /* CCSID for BOCU-1 */ |
1401 | UCNV_IBM, UCNV_BOCU1, |
1402 | 1, 4, /* one char16_t generates at least 1 byte and at most 4 bytes */ |
1403 | { 0x1a, 0, 0, 0 }, 1, /* BOCU-1 never needs to write a subchar */ |
1404 | false, false, |
1405 | 0, |
1406 | 0, |
1407 | { 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0 } /* reserved */ |
1408 | }; |
1409 | |
1410 | const UConverterSharedData _Bocu1Data= |
1411 | UCNV_IMMUTABLE_SHARED_DATA_INITIALIZER(&_Bocu1StaticData, &_Bocu1Impl); |
1412 | |
1413 | #endif |
1414 | |