1 /*
2 ******************************************************************************
3 *
4 * Copyright (C) 2003-2013, International Business Machines
5 * Corporation and others. All Rights Reserved.
6 *
7 ******************************************************************************
8 * file name: ucnv_ext.cpp
9 * encoding: US-ASCII
10 * tab size: 8 (not used)
11 * indentation:4
12 *
13 * created on: 2003jun13
14 * created by: Markus W. Scherer
15 *
16 * Conversion extensions
17 */
18
19 #include "unicode/utypes.h"
20
21 #if !UCONFIG_NO_CONVERSION && !UCONFIG_NO_LEGACY_CONVERSION
22
23 #include "unicode/uset.h"
24 #include "ucnv_bld.h"
25 #include "ucnv_cnv.h"
26 #include "ucnv_ext.h"
27 #include "cmemory.h"
28 #include "uassert.h"
29
30 /* to Unicode --------------------------------------------------------------- */
31
32 /*
33 * @return lookup value for the byte, if found; else 0
34 */
35 static inline uint32_t
ucnv_extFindToU(const uint32_t * toUSection,int32_t length,uint8_t byte)36 ucnv_extFindToU(const uint32_t *toUSection, int32_t length, uint8_t byte) {
37 uint32_t word0, word;
38 int32_t i, start, limit;
39
40 /* check the input byte against the lowest and highest section bytes */
41 start=(int32_t)UCNV_EXT_TO_U_GET_BYTE(toUSection[0]);
42 limit=(int32_t)UCNV_EXT_TO_U_GET_BYTE(toUSection[length-1]);
43 if(byte<start || limit<byte) {
44 return 0; /* the byte is out of range */
45 }
46
47 if(length==((limit-start)+1)) {
48 /* direct access on a linear array */
49 return UCNV_EXT_TO_U_GET_VALUE(toUSection[byte-start]); /* could be 0 */
50 }
51
52 /* word0 is suitable for <=toUSection[] comparison, word for <toUSection[] */
53 word0=UCNV_EXT_TO_U_MAKE_WORD(byte, 0);
54
55 /*
56 * Shift byte once instead of each section word and add 0xffffff.
57 * We will compare the shifted/added byte (bbffffff) against
58 * section words which have byte values in the same bit position.
59 * If and only if byte bb < section byte ss then bbffffff<ssvvvvvv
60 * for all v=0..f
61 * so we need not mask off the lower 24 bits of each section word.
62 */
63 word=word0|UCNV_EXT_TO_U_VALUE_MASK;
64
65 /* binary search */
66 start=0;
67 limit=length;
68 for(;;) {
69 i=limit-start;
70 if(i<=1) {
71 break; /* done */
72 }
73 /* start<limit-1 */
74
75 if(i<=4) {
76 /* linear search for the last part */
77 if(word0<=toUSection[start]) {
78 break;
79 }
80 if(++start<limit && word0<=toUSection[start]) {
81 break;
82 }
83 if(++start<limit && word0<=toUSection[start]) {
84 break;
85 }
86 /* always break at start==limit-1 */
87 ++start;
88 break;
89 }
90
91 i=(start+limit)/2;
92 if(word<toUSection[i]) {
93 limit=i;
94 } else {
95 start=i;
96 }
97 }
98
99 /* did we really find it? */
100 if(start<limit && byte==UCNV_EXT_TO_U_GET_BYTE(word=toUSection[start])) {
101 return UCNV_EXT_TO_U_GET_VALUE(word); /* never 0 */
102 } else {
103 return 0; /* not found */
104 }
105 }
106
107 /*
108 * TRUE if not an SI/SO stateful converter,
109 * or if the match length fits with the current converter state
110 */
111 #define UCNV_EXT_TO_U_VERIFY_SISO_MATCH(sisoState, match) \
112 ((sisoState)<0 || ((sisoState)==0) == (match==1))
113
114 /*
115 * this works like ucnv_extMatchFromU() except
116 * - the first character is in pre
117 * - no trie is used
118 * - the returned matchLength is not offset by 2
119 */
120 static int32_t
ucnv_extMatchToU(const int32_t * cx,int8_t sisoState,const char * pre,int32_t preLength,const char * src,int32_t srcLength,uint32_t * pMatchValue,UBool,UBool flush)121 ucnv_extMatchToU(const int32_t *cx, int8_t sisoState,
122 const char *pre, int32_t preLength,
123 const char *src, int32_t srcLength,
124 uint32_t *pMatchValue,
125 UBool /*useFallback*/, UBool flush) {
126 const uint32_t *toUTable, *toUSection;
127
128 uint32_t value, matchValue;
129 int32_t i, j, idx, length, matchLength;
130 uint8_t b;
131
132 if(cx==NULL || cx[UCNV_EXT_TO_U_LENGTH]<=0) {
133 return 0; /* no extension data, no match */
134 }
135
136 /* initialize */
137 toUTable=UCNV_EXT_ARRAY(cx, UCNV_EXT_TO_U_INDEX, uint32_t);
138 idx=0;
139
140 matchValue=0;
141 i=j=matchLength=0;
142
143 if(sisoState==0) {
144 /* SBCS state of an SI/SO stateful converter, look at only exactly 1 byte */
145 if(preLength>1) {
146 return 0; /* no match of a DBCS sequence in SBCS mode */
147 } else if(preLength==1) {
148 srcLength=0;
149 } else /* preLength==0 */ {
150 if(srcLength>1) {
151 srcLength=1;
152 }
153 }
154 flush=TRUE;
155 }
156
157 /* we must not remember fallback matches when not using fallbacks */
158
159 /* match input units until there is a full match or the input is consumed */
160 for(;;) {
161 /* go to the next section */
162 toUSection=toUTable+idx;
163
164 /* read first pair of the section */
165 value=*toUSection++;
166 length=UCNV_EXT_TO_U_GET_BYTE(value);
167 value=UCNV_EXT_TO_U_GET_VALUE(value);
168 if( value!=0 &&
169 (UCNV_EXT_TO_U_IS_ROUNDTRIP(value) ||
170 TO_U_USE_FALLBACK(useFallback)) &&
171 UCNV_EXT_TO_U_VERIFY_SISO_MATCH(sisoState, i+j)
172 ) {
173 /* remember longest match so far */
174 matchValue=value;
175 matchLength=i+j;
176 }
177
178 /* match pre[] then src[] */
179 if(i<preLength) {
180 b=(uint8_t)pre[i++];
181 } else if(j<srcLength) {
182 b=(uint8_t)src[j++];
183 } else {
184 /* all input consumed, partial match */
185 if(flush || (length=(i+j))>UCNV_EXT_MAX_BYTES) {
186 /*
187 * end of the entire input stream, stop with the longest match so far
188 * or: partial match must not be longer than UCNV_EXT_MAX_BYTES
189 * because it must fit into state buffers
190 */
191 break;
192 } else {
193 /* continue with more input next time */
194 return -length;
195 }
196 }
197
198 /* search for the current UChar */
199 value=ucnv_extFindToU(toUSection, length, b);
200 if(value==0) {
201 /* no match here, stop with the longest match so far */
202 break;
203 } else {
204 if(UCNV_EXT_TO_U_IS_PARTIAL(value)) {
205 /* partial match, continue */
206 idx=(int32_t)UCNV_EXT_TO_U_GET_PARTIAL_INDEX(value);
207 } else {
208 if( (UCNV_EXT_TO_U_IS_ROUNDTRIP(value) ||
209 TO_U_USE_FALLBACK(useFallback)) &&
210 UCNV_EXT_TO_U_VERIFY_SISO_MATCH(sisoState, i+j)
211 ) {
212 /* full match, stop with result */
213 matchValue=value;
214 matchLength=i+j;
215 } else {
216 /* full match on fallback not taken, stop with the longest match so far */
217 }
218 break;
219 }
220 }
221 }
222
223 if(matchLength==0) {
224 /* no match at all */
225 return 0;
226 }
227
228 /* return result */
229 *pMatchValue=UCNV_EXT_TO_U_MASK_ROUNDTRIP(matchValue);
230 return matchLength;
231 }
232
233 static inline void
ucnv_extWriteToU(UConverter * cnv,const int32_t * cx,uint32_t value,UChar ** target,const UChar * targetLimit,int32_t ** offsets,int32_t srcIndex,UErrorCode * pErrorCode)234 ucnv_extWriteToU(UConverter *cnv, const int32_t *cx,
235 uint32_t value,
236 UChar **target, const UChar *targetLimit,
237 int32_t **offsets, int32_t srcIndex,
238 UErrorCode *pErrorCode) {
239 /* output the result */
240 if(UCNV_EXT_TO_U_IS_CODE_POINT(value)) {
241 /* output a single code point */
242 ucnv_toUWriteCodePoint(
243 cnv, UCNV_EXT_TO_U_GET_CODE_POINT(value),
244 target, targetLimit,
245 offsets, srcIndex,
246 pErrorCode);
247 } else {
248 /* output a string - with correct data we have resultLength>0 */
249 ucnv_toUWriteUChars(
250 cnv,
251 UCNV_EXT_ARRAY(cx, UCNV_EXT_TO_U_UCHARS_INDEX, UChar)+
252 UCNV_EXT_TO_U_GET_INDEX(value),
253 UCNV_EXT_TO_U_GET_LENGTH(value),
254 target, targetLimit,
255 offsets, srcIndex,
256 pErrorCode);
257 }
258 }
259
260 /*
261 * get the SI/SO toU state (state 0 is for SBCS, 1 for DBCS),
262 * or 1 for DBCS-only,
263 * or -1 if the converter is not SI/SO stateful
264 *
265 * Note: For SI/SO stateful converters getting here,
266 * cnv->mode==0 is equivalent to firstLength==1.
267 */
268 #define UCNV_SISO_STATE(cnv) \
269 ((cnv)->sharedData->mbcs.outputType==MBCS_OUTPUT_2_SISO ? (int8_t)(cnv)->mode : \
270 (cnv)->sharedData->mbcs.outputType==MBCS_OUTPUT_DBCS_ONLY ? 1 : -1)
271
272 /*
273 * target<targetLimit; set error code for overflow
274 */
275 U_CFUNC UBool
ucnv_extInitialMatchToU(UConverter * cnv,const int32_t * cx,int32_t firstLength,const char ** src,const char * srcLimit,UChar ** target,const UChar * targetLimit,int32_t ** offsets,int32_t srcIndex,UBool flush,UErrorCode * pErrorCode)276 ucnv_extInitialMatchToU(UConverter *cnv, const int32_t *cx,
277 int32_t firstLength,
278 const char **src, const char *srcLimit,
279 UChar **target, const UChar *targetLimit,
280 int32_t **offsets, int32_t srcIndex,
281 UBool flush,
282 UErrorCode *pErrorCode) {
283 uint32_t value = 0; /* initialize output-only param to 0 to silence gcc */
284 int32_t match;
285
286 /* try to match */
287 match=ucnv_extMatchToU(cx, (int8_t)UCNV_SISO_STATE(cnv),
288 (const char *)cnv->toUBytes, firstLength,
289 *src, (int32_t)(srcLimit-*src),
290 &value,
291 cnv->useFallback, flush);
292 if(match>0) {
293 /* advance src pointer for the consumed input */
294 *src+=match-firstLength;
295
296 /* write result to target */
297 ucnv_extWriteToU(cnv, cx,
298 value,
299 target, targetLimit,
300 offsets, srcIndex,
301 pErrorCode);
302 return TRUE;
303 } else if(match<0) {
304 /* save state for partial match */
305 const char *s;
306 int32_t j;
307
308 /* copy the first code point */
309 s=(const char *)cnv->toUBytes;
310 cnv->preToUFirstLength=(int8_t)firstLength;
311 for(j=0; j<firstLength; ++j) {
312 cnv->preToU[j]=*s++;
313 }
314
315 /* now copy the newly consumed input */
316 s=*src;
317 match=-match;
318 for(; j<match; ++j) {
319 cnv->preToU[j]=*s++;
320 }
321 *src=s; /* same as *src=srcLimit; because we reached the end of input */
322 cnv->preToULength=(int8_t)match;
323 return TRUE;
324 } else /* match==0 no match */ {
325 return FALSE;
326 }
327 }
328
329 U_CFUNC UChar32
ucnv_extSimpleMatchToU(const int32_t * cx,const char * source,int32_t length,UBool useFallback)330 ucnv_extSimpleMatchToU(const int32_t *cx,
331 const char *source, int32_t length,
332 UBool useFallback) {
333 uint32_t value = 0; /* initialize output-only param to 0 to silence gcc */
334 int32_t match;
335
336 if(length<=0) {
337 return 0xffff;
338 }
339
340 /* try to match */
341 match=ucnv_extMatchToU(cx, -1,
342 source, length,
343 NULL, 0,
344 &value,
345 useFallback, TRUE);
346 if(match==length) {
347 /* write result for simple, single-character conversion */
348 if(UCNV_EXT_TO_U_IS_CODE_POINT(value)) {
349 return UCNV_EXT_TO_U_GET_CODE_POINT(value);
350 }
351 }
352
353 /*
354 * return no match because
355 * - match>0 && value points to string: simple conversion cannot handle multiple code points
356 * - match>0 && match!=length: not all input consumed, forbidden for this function
357 * - match==0: no match found in the first place
358 * - match<0: partial match, not supported for simple conversion (and flush==TRUE)
359 */
360 return 0xfffe;
361 }
362
363 /*
364 * continue partial match with new input
365 * never called for simple, single-character conversion
366 */
367 U_CFUNC void
ucnv_extContinueMatchToU(UConverter * cnv,UConverterToUnicodeArgs * pArgs,int32_t srcIndex,UErrorCode * pErrorCode)368 ucnv_extContinueMatchToU(UConverter *cnv,
369 UConverterToUnicodeArgs *pArgs, int32_t srcIndex,
370 UErrorCode *pErrorCode) {
371 uint32_t value = 0; /* initialize output-only param to 0 to silence gcc */
372 int32_t match, length;
373
374 match=ucnv_extMatchToU(cnv->sharedData->mbcs.extIndexes, (int8_t)UCNV_SISO_STATE(cnv),
375 cnv->preToU, cnv->preToULength,
376 pArgs->source, (int32_t)(pArgs->sourceLimit-pArgs->source),
377 &value,
378 cnv->useFallback, pArgs->flush);
379 if(match>0) {
380 if(match>=cnv->preToULength) {
381 /* advance src pointer for the consumed input */
382 pArgs->source+=match-cnv->preToULength;
383 cnv->preToULength=0;
384 } else {
385 /* the match did not use all of preToU[] - keep the rest for replay */
386 length=cnv->preToULength-match;
387 uprv_memmove(cnv->preToU, cnv->preToU+match, length);
388 cnv->preToULength=(int8_t)-length;
389 }
390
391 /* write result */
392 ucnv_extWriteToU(cnv, cnv->sharedData->mbcs.extIndexes,
393 value,
394 &pArgs->target, pArgs->targetLimit,
395 &pArgs->offsets, srcIndex,
396 pErrorCode);
397 } else if(match<0) {
398 /* save state for partial match */
399 const char *s;
400 int32_t j;
401
402 /* just _append_ the newly consumed input to preToU[] */
403 s=pArgs->source;
404 match=-match;
405 for(j=cnv->preToULength; j<match; ++j) {
406 cnv->preToU[j]=*s++;
407 }
408 pArgs->source=s; /* same as *src=srcLimit; because we reached the end of input */
409 cnv->preToULength=(int8_t)match;
410 } else /* match==0 */ {
411 /*
412 * no match
413 *
414 * We need to split the previous input into two parts:
415 *
416 * 1. The first codepage character is unmappable - that's how we got into
417 * trying the extension data in the first place.
418 * We need to move it from the preToU buffer
419 * to the error buffer, set an error code,
420 * and prepare the rest of the previous input for 2.
421 *
422 * 2. The rest of the previous input must be converted once we
423 * come back from the callback for the first character.
424 * At that time, we have to try again from scratch to convert
425 * these input characters.
426 * The replay will be handled by the ucnv.c conversion code.
427 */
428
429 /* move the first codepage character to the error field */
430 uprv_memcpy(cnv->toUBytes, cnv->preToU, cnv->preToUFirstLength);
431 cnv->toULength=cnv->preToUFirstLength;
432
433 /* move the rest up inside the buffer */
434 length=cnv->preToULength-cnv->preToUFirstLength;
435 if(length>0) {
436 uprv_memmove(cnv->preToU, cnv->preToU+cnv->preToUFirstLength, length);
437 }
438
439 /* mark preToU for replay */
440 cnv->preToULength=(int8_t)-length;
441
442 /* set the error code for unassigned */
443 *pErrorCode=U_INVALID_CHAR_FOUND;
444 }
445 }
446
447 /* from Unicode ------------------------------------------------------------- */
448
449 // Use roundtrips, "good one-way" mappings, and some normal fallbacks.
450 static inline UBool
extFromUUseMapping(UBool useFallback,uint32_t value,UChar32 firstCP)451 extFromUUseMapping(UBool useFallback, uint32_t value, UChar32 firstCP) {
452 return
453 ((value&UCNV_EXT_FROM_U_STATUS_MASK)!=0 ||
454 FROM_U_USE_FALLBACK(useFallback, firstCP)) &&
455 (value&UCNV_EXT_FROM_U_RESERVED_MASK)==0;
456 }
457
458 /*
459 * @return index of the UChar, if found; else <0
460 */
461 static inline int32_t
ucnv_extFindFromU(const UChar * fromUSection,int32_t length,UChar u)462 ucnv_extFindFromU(const UChar *fromUSection, int32_t length, UChar u) {
463 int32_t i, start, limit;
464
465 /* binary search */
466 start=0;
467 limit=length;
468 for(;;) {
469 i=limit-start;
470 if(i<=1) {
471 break; /* done */
472 }
473 /* start<limit-1 */
474
475 if(i<=4) {
476 /* linear search for the last part */
477 if(u<=fromUSection[start]) {
478 break;
479 }
480 if(++start<limit && u<=fromUSection[start]) {
481 break;
482 }
483 if(++start<limit && u<=fromUSection[start]) {
484 break;
485 }
486 /* always break at start==limit-1 */
487 ++start;
488 break;
489 }
490
491 i=(start+limit)/2;
492 if(u<fromUSection[i]) {
493 limit=i;
494 } else {
495 start=i;
496 }
497 }
498
499 /* did we really find it? */
500 if(start<limit && u==fromUSection[start]) {
501 return start;
502 } else {
503 return -1; /* not found */
504 }
505 }
506
507 /*
508 * @param cx pointer to extension data; if NULL, returns 0
509 * @param firstCP the first code point before all the other UChars
510 * @param pre UChars that must match; !initialMatch: partial match with them
511 * @param preLength length of pre, >=0
512 * @param src UChars that can be used to complete a match
513 * @param srcLength length of src, >=0
514 * @param pMatchValue [out] output result value for the match from the data structure
515 * @param useFallback "use fallback" flag, usually from cnv->useFallback
516 * @param flush TRUE if the end of the input stream is reached
517 * @return >1: matched, return value=total match length (number of input units matched)
518 * 1: matched, no mapping but request for <subchar1>
519 * (only for the first code point)
520 * 0: no match
521 * <0: partial match, return value=negative total match length
522 * (partial matches are never returned for flush==TRUE)
523 * (partial matches are never returned as being longer than UCNV_EXT_MAX_UCHARS)
524 * the matchLength is 2 if only firstCP matched, and >2 if firstCP and
525 * further code units matched
526 */
527 static int32_t
ucnv_extMatchFromU(const int32_t * cx,UChar32 firstCP,const UChar * pre,int32_t preLength,const UChar * src,int32_t srcLength,uint32_t * pMatchValue,UBool useFallback,UBool flush)528 ucnv_extMatchFromU(const int32_t *cx,
529 UChar32 firstCP,
530 const UChar *pre, int32_t preLength,
531 const UChar *src, int32_t srcLength,
532 uint32_t *pMatchValue,
533 UBool useFallback, UBool flush) {
534 const uint16_t *stage12, *stage3;
535 const uint32_t *stage3b;
536
537 const UChar *fromUTableUChars, *fromUSectionUChars;
538 const uint32_t *fromUTableValues, *fromUSectionValues;
539
540 uint32_t value, matchValue;
541 int32_t i, j, idx, length, matchLength;
542 UChar c;
543
544 if(cx==NULL) {
545 return 0; /* no extension data, no match */
546 }
547
548 /* trie lookup of firstCP */
549 idx=firstCP>>10; /* stage 1 index */
550 if(idx>=cx[UCNV_EXT_FROM_U_STAGE_1_LENGTH]) {
551 return 0; /* the first code point is outside the trie */
552 }
553
554 stage12=UCNV_EXT_ARRAY(cx, UCNV_EXT_FROM_U_STAGE_12_INDEX, uint16_t);
555 stage3=UCNV_EXT_ARRAY(cx, UCNV_EXT_FROM_U_STAGE_3_INDEX, uint16_t);
556 idx=UCNV_EXT_FROM_U(stage12, stage3, idx, firstCP);
557
558 stage3b=UCNV_EXT_ARRAY(cx, UCNV_EXT_FROM_U_STAGE_3B_INDEX, uint32_t);
559 value=stage3b[idx];
560 if(value==0) {
561 return 0;
562 }
563
564 /*
565 * Tests for (value&UCNV_EXT_FROM_U_RESERVED_MASK)==0:
566 * Do not interpret values with reserved bits used, for forward compatibility,
567 * and do not even remember intermediate results with reserved bits used.
568 */
569
570 if(UCNV_EXT_TO_U_IS_PARTIAL(value)) {
571 /* partial match, enter the loop below */
572 idx=(int32_t)UCNV_EXT_FROM_U_GET_PARTIAL_INDEX(value);
573
574 /* initialize */
575 fromUTableUChars=UCNV_EXT_ARRAY(cx, UCNV_EXT_FROM_U_UCHARS_INDEX, UChar);
576 fromUTableValues=UCNV_EXT_ARRAY(cx, UCNV_EXT_FROM_U_VALUES_INDEX, uint32_t);
577
578 matchValue=0;
579 i=j=matchLength=0;
580
581 /* we must not remember fallback matches when not using fallbacks */
582
583 /* match input units until there is a full match or the input is consumed */
584 for(;;) {
585 /* go to the next section */
586 fromUSectionUChars=fromUTableUChars+idx;
587 fromUSectionValues=fromUTableValues+idx;
588
589 /* read first pair of the section */
590 length=*fromUSectionUChars++;
591 value=*fromUSectionValues++;
592 if(value!=0 && extFromUUseMapping(useFallback, value, firstCP)) {
593 /* remember longest match so far */
594 matchValue=value;
595 matchLength=2+i+j;
596 }
597
598 /* match pre[] then src[] */
599 if(i<preLength) {
600 c=pre[i++];
601 } else if(j<srcLength) {
602 c=src[j++];
603 } else {
604 /* all input consumed, partial match */
605 if(flush || (length=(i+j))>UCNV_EXT_MAX_UCHARS) {
606 /*
607 * end of the entire input stream, stop with the longest match so far
608 * or: partial match must not be longer than UCNV_EXT_MAX_UCHARS
609 * because it must fit into state buffers
610 */
611 break;
612 } else {
613 /* continue with more input next time */
614 return -(2+length);
615 }
616 }
617
618 /* search for the current UChar */
619 idx=ucnv_extFindFromU(fromUSectionUChars, length, c);
620 if(idx<0) {
621 /* no match here, stop with the longest match so far */
622 break;
623 } else {
624 value=fromUSectionValues[idx];
625 if(UCNV_EXT_FROM_U_IS_PARTIAL(value)) {
626 /* partial match, continue */
627 idx=(int32_t)UCNV_EXT_FROM_U_GET_PARTIAL_INDEX(value);
628 } else {
629 if(extFromUUseMapping(useFallback, value, firstCP)) {
630 /* full match, stop with result */
631 matchValue=value;
632 matchLength=2+i+j;
633 } else {
634 /* full match on fallback not taken, stop with the longest match so far */
635 }
636 break;
637 }
638 }
639 }
640
641 if(matchLength==0) {
642 /* no match at all */
643 return 0;
644 }
645 } else /* result from firstCP trie lookup */ {
646 if(extFromUUseMapping(useFallback, value, firstCP)) {
647 /* full match, stop with result */
648 matchValue=value;
649 matchLength=2;
650 } else {
651 /* fallback not taken */
652 return 0;
653 }
654 }
655
656 /* return result */
657 if(matchValue==UCNV_EXT_FROM_U_SUBCHAR1) {
658 return 1; /* assert matchLength==2 */
659 }
660
661 *pMatchValue=matchValue;
662 return matchLength;
663 }
664
665 /*
666 * @param value fromUnicode mapping table value; ignores roundtrip and reserved bits
667 */
668 static inline void
ucnv_extWriteFromU(UConverter * cnv,const int32_t * cx,uint32_t value,char ** target,const char * targetLimit,int32_t ** offsets,int32_t srcIndex,UErrorCode * pErrorCode)669 ucnv_extWriteFromU(UConverter *cnv, const int32_t *cx,
670 uint32_t value,
671 char **target, const char *targetLimit,
672 int32_t **offsets, int32_t srcIndex,
673 UErrorCode *pErrorCode) {
674 uint8_t buffer[1+UCNV_EXT_MAX_BYTES];
675 const uint8_t *result;
676 int32_t length, prevLength;
677
678 length=UCNV_EXT_FROM_U_GET_LENGTH(value);
679 value=(uint32_t)UCNV_EXT_FROM_U_GET_DATA(value);
680
681 /* output the result */
682 if(length<=UCNV_EXT_FROM_U_MAX_DIRECT_LENGTH) {
683 /*
684 * Generate a byte array and then write it below.
685 * This is not the fastest possible way, but it should be ok for
686 * extension mappings, and it is much simpler.
687 * Offset and overflow handling are only done once this way.
688 */
689 uint8_t *p=buffer+1; /* reserve buffer[0] for shiftByte below */
690 switch(length) {
691 case 3:
692 *p++=(uint8_t)(value>>16);
693 case 2: /*fall through*/
694 *p++=(uint8_t)(value>>8);
695 case 1: /*fall through*/
696 *p++=(uint8_t)value;
697 default:
698 break; /* will never occur */
699 }
700 result=buffer+1;
701 } else {
702 result=UCNV_EXT_ARRAY(cx, UCNV_EXT_FROM_U_BYTES_INDEX, uint8_t)+value;
703 }
704
705 /* with correct data we have length>0 */
706
707 if((prevLength=cnv->fromUnicodeStatus)!=0) {
708 /* handle SI/SO stateful output */
709 uint8_t shiftByte;
710
711 if(prevLength>1 && length==1) {
712 /* change from double-byte mode to single-byte */
713 shiftByte=(uint8_t)UCNV_SI;
714 cnv->fromUnicodeStatus=1;
715 } else if(prevLength==1 && length>1) {
716 /* change from single-byte mode to double-byte */
717 shiftByte=(uint8_t)UCNV_SO;
718 cnv->fromUnicodeStatus=2;
719 } else {
720 shiftByte=0;
721 }
722
723 if(shiftByte!=0) {
724 /* prepend the shift byte to the result bytes */
725 buffer[0]=shiftByte;
726 if(result!=buffer+1) {
727 uprv_memcpy(buffer+1, result, length);
728 }
729 result=buffer;
730 ++length;
731 }
732 }
733
734 ucnv_fromUWriteBytes(cnv, (const char *)result, length,
735 target, targetLimit,
736 offsets, srcIndex,
737 pErrorCode);
738 }
739
740 /*
741 * target<targetLimit; set error code for overflow
742 */
743 U_CFUNC UBool
ucnv_extInitialMatchFromU(UConverter * cnv,const int32_t * cx,UChar32 cp,const UChar ** src,const UChar * srcLimit,char ** target,const char * targetLimit,int32_t ** offsets,int32_t srcIndex,UBool flush,UErrorCode * pErrorCode)744 ucnv_extInitialMatchFromU(UConverter *cnv, const int32_t *cx,
745 UChar32 cp,
746 const UChar **src, const UChar *srcLimit,
747 char **target, const char *targetLimit,
748 int32_t **offsets, int32_t srcIndex,
749 UBool flush,
750 UErrorCode *pErrorCode) {
751 uint32_t value = 0; /* initialize output-only param to 0 to silence gcc */
752 int32_t match;
753
754 /* try to match */
755 match=ucnv_extMatchFromU(cx, cp,
756 NULL, 0,
757 *src, (int32_t)(srcLimit-*src),
758 &value,
759 cnv->useFallback, flush);
760
761 /* reject a match if the result is a single byte for DBCS-only */
762 if( match>=2 &&
763 !(UCNV_EXT_FROM_U_GET_LENGTH(value)==1 &&
764 cnv->sharedData->mbcs.outputType==MBCS_OUTPUT_DBCS_ONLY)
765 ) {
766 /* advance src pointer for the consumed input */
767 *src+=match-2; /* remove 2 for the initial code point */
768
769 /* write result to target */
770 ucnv_extWriteFromU(cnv, cx,
771 value,
772 target, targetLimit,
773 offsets, srcIndex,
774 pErrorCode);
775 return TRUE;
776 } else if(match<0) {
777 /* save state for partial match */
778 const UChar *s;
779 int32_t j;
780
781 /* copy the first code point */
782 cnv->preFromUFirstCP=cp;
783
784 /* now copy the newly consumed input */
785 s=*src;
786 match=-match-2; /* remove 2 for the initial code point */
787 for(j=0; j<match; ++j) {
788 cnv->preFromU[j]=*s++;
789 }
790 *src=s; /* same as *src=srcLimit; because we reached the end of input */
791 cnv->preFromULength=(int8_t)match;
792 return TRUE;
793 } else if(match==1) {
794 /* matched, no mapping but request for <subchar1> */
795 cnv->useSubChar1=TRUE;
796 return FALSE;
797 } else /* match==0 no match */ {
798 return FALSE;
799 }
800 }
801
802 /*
803 * Used by ISO 2022 implementation.
804 * @return number of bytes in *pValue; negative number if fallback; 0 for no mapping
805 */
806 U_CFUNC int32_t
ucnv_extSimpleMatchFromU(const int32_t * cx,UChar32 cp,uint32_t * pValue,UBool useFallback)807 ucnv_extSimpleMatchFromU(const int32_t *cx,
808 UChar32 cp, uint32_t *pValue,
809 UBool useFallback) {
810 uint32_t value;
811 int32_t match;
812
813 /* try to match */
814 match=ucnv_extMatchFromU(cx,
815 cp,
816 NULL, 0,
817 NULL, 0,
818 &value,
819 useFallback, TRUE);
820 if(match>=2) {
821 /* write result for simple, single-character conversion */
822 int32_t length;
823 int isRoundtrip;
824
825 isRoundtrip=UCNV_EXT_FROM_U_IS_ROUNDTRIP(value);
826 length=UCNV_EXT_FROM_U_GET_LENGTH(value);
827 value=(uint32_t)UCNV_EXT_FROM_U_GET_DATA(value);
828
829 if(length<=UCNV_EXT_FROM_U_MAX_DIRECT_LENGTH) {
830 *pValue=value;
831 return isRoundtrip ? length : -length;
832 #if 0 /* not currently used */
833 } else if(length==4) {
834 /* de-serialize a 4-byte result */
835 const uint8_t *result=UCNV_EXT_ARRAY(cx, UCNV_EXT_FROM_U_BYTES_INDEX, uint8_t)+value;
836 *pValue=
837 ((uint32_t)result[0]<<24)|
838 ((uint32_t)result[1]<<16)|
839 ((uint32_t)result[2]<<8)|
840 result[3];
841 return isRoundtrip ? 4 : -4;
842 #endif
843 }
844 }
845
846 /*
847 * return no match because
848 * - match>1 && resultLength>4: result too long for simple conversion
849 * - match==1: no match found, <subchar1> preferred
850 * - match==0: no match found in the first place
851 * - match<0: partial match, not supported for simple conversion (and flush==TRUE)
852 */
853 return 0;
854 }
855
856 /*
857 * continue partial match with new input, requires cnv->preFromUFirstCP>=0
858 * never called for simple, single-character conversion
859 */
860 U_CFUNC void
ucnv_extContinueMatchFromU(UConverter * cnv,UConverterFromUnicodeArgs * pArgs,int32_t srcIndex,UErrorCode * pErrorCode)861 ucnv_extContinueMatchFromU(UConverter *cnv,
862 UConverterFromUnicodeArgs *pArgs, int32_t srcIndex,
863 UErrorCode *pErrorCode) {
864 uint32_t value = 0; /* initialize output-only param to 0 to silence gcc */
865 int32_t match;
866
867 match=ucnv_extMatchFromU(cnv->sharedData->mbcs.extIndexes,
868 cnv->preFromUFirstCP,
869 cnv->preFromU, cnv->preFromULength,
870 pArgs->source, (int32_t)(pArgs->sourceLimit-pArgs->source),
871 &value,
872 cnv->useFallback, pArgs->flush);
873 if(match>=2) {
874 match-=2; /* remove 2 for the initial code point */
875
876 if(match>=cnv->preFromULength) {
877 /* advance src pointer for the consumed input */
878 pArgs->source+=match-cnv->preFromULength;
879 cnv->preFromULength=0;
880 } else {
881 /* the match did not use all of preFromU[] - keep the rest for replay */
882 int32_t length=cnv->preFromULength-match;
883 uprv_memmove(cnv->preFromU, cnv->preFromU+match, length*U_SIZEOF_UCHAR);
884 cnv->preFromULength=(int8_t)-length;
885 }
886
887 /* finish the partial match */
888 cnv->preFromUFirstCP=U_SENTINEL;
889
890 /* write result */
891 ucnv_extWriteFromU(cnv, cnv->sharedData->mbcs.extIndexes,
892 value,
893 &pArgs->target, pArgs->targetLimit,
894 &pArgs->offsets, srcIndex,
895 pErrorCode);
896 } else if(match<0) {
897 /* save state for partial match */
898 const UChar *s;
899 int32_t j;
900
901 /* just _append_ the newly consumed input to preFromU[] */
902 s=pArgs->source;
903 match=-match-2; /* remove 2 for the initial code point */
904 for(j=cnv->preFromULength; j<match; ++j) {
905 U_ASSERT(j>=0);
906 cnv->preFromU[j]=*s++;
907 }
908 pArgs->source=s; /* same as *src=srcLimit; because we reached the end of input */
909 cnv->preFromULength=(int8_t)match;
910 } else /* match==0 or 1 */ {
911 /*
912 * no match
913 *
914 * We need to split the previous input into two parts:
915 *
916 * 1. The first code point is unmappable - that's how we got into
917 * trying the extension data in the first place.
918 * We need to move it from the preFromU buffer
919 * to the error buffer, set an error code,
920 * and prepare the rest of the previous input for 2.
921 *
922 * 2. The rest of the previous input must be converted once we
923 * come back from the callback for the first code point.
924 * At that time, we have to try again from scratch to convert
925 * these input characters.
926 * The replay will be handled by the ucnv.c conversion code.
927 */
928
929 if(match==1) {
930 /* matched, no mapping but request for <subchar1> */
931 cnv->useSubChar1=TRUE;
932 }
933
934 /* move the first code point to the error field */
935 cnv->fromUChar32=cnv->preFromUFirstCP;
936 cnv->preFromUFirstCP=U_SENTINEL;
937
938 /* mark preFromU for replay */
939 cnv->preFromULength=-cnv->preFromULength;
940
941 /* set the error code for unassigned */
942 *pErrorCode=U_INVALID_CHAR_FOUND;
943 }
944 }
945
946 static UBool
extSetUseMapping(UConverterUnicodeSet which,int32_t minLength,uint32_t value)947 extSetUseMapping(UConverterUnicodeSet which, int32_t minLength, uint32_t value) {
948 if(which==UCNV_ROUNDTRIP_SET) {
949 // Add only code points for which the roundtrip flag is set.
950 // Do not add any fallbacks, even if ucnv_fromUnicode() would use them
951 // (fallbacks from PUA). See the API docs for ucnv_getUnicodeSet().
952 //
953 // By analogy, also do not add "good one-way" mappings.
954 //
955 // Do not add entries with reserved bits set.
956 if(((value&(UCNV_EXT_FROM_U_ROUNDTRIP_FLAG|UCNV_EXT_FROM_U_RESERVED_MASK))!=
957 UCNV_EXT_FROM_U_ROUNDTRIP_FLAG)) {
958 return FALSE;
959 }
960 } else /* UCNV_ROUNDTRIP_AND_FALLBACK_SET */ {
961 // Do not add entries with reserved bits set.
962 if((value&UCNV_EXT_FROM_U_RESERVED_MASK)!=0) {
963 return FALSE;
964 }
965 }
966 // Do not add <subchar1> entries or other (future?) pseudo-entries
967 // with an output length of 0.
968 return UCNV_EXT_FROM_U_GET_LENGTH(value)>=minLength;
969 }
970
971 static void
ucnv_extGetUnicodeSetString(const UConverterSharedData * sharedData,const int32_t * cx,const USetAdder * sa,UConverterUnicodeSet which,int32_t minLength,UChar32 firstCP,UChar s[UCNV_EXT_MAX_UCHARS],int32_t length,int32_t sectionIndex,UErrorCode * pErrorCode)972 ucnv_extGetUnicodeSetString(const UConverterSharedData *sharedData,
973 const int32_t *cx,
974 const USetAdder *sa,
975 UConverterUnicodeSet which,
976 int32_t minLength,
977 UChar32 firstCP,
978 UChar s[UCNV_EXT_MAX_UCHARS], int32_t length,
979 int32_t sectionIndex,
980 UErrorCode *pErrorCode) {
981 const UChar *fromUSectionUChars;
982 const uint32_t *fromUSectionValues;
983
984 uint32_t value;
985 int32_t i, count;
986
987 fromUSectionUChars=UCNV_EXT_ARRAY(cx, UCNV_EXT_FROM_U_UCHARS_INDEX, UChar)+sectionIndex;
988 fromUSectionValues=UCNV_EXT_ARRAY(cx, UCNV_EXT_FROM_U_VALUES_INDEX, uint32_t)+sectionIndex;
989
990 /* read first pair of the section */
991 count=*fromUSectionUChars++;
992 value=*fromUSectionValues++;
993
994 if(extSetUseMapping(which, minLength, value)) {
995 if(length==U16_LENGTH(firstCP)) {
996 /* add the initial code point */
997 sa->add(sa->set, firstCP);
998 } else {
999 /* add the string so far */
1000 sa->addString(sa->set, s, length);
1001 }
1002 }
1003
1004 for(i=0; i<count; ++i) {
1005 /* append this code unit and recurse or add the string */
1006 s[length]=fromUSectionUChars[i];
1007 value=fromUSectionValues[i];
1008
1009 if(value==0) {
1010 /* no mapping, do nothing */
1011 } else if(UCNV_EXT_FROM_U_IS_PARTIAL(value)) {
1012 ucnv_extGetUnicodeSetString(
1013 sharedData, cx, sa, which, minLength,
1014 firstCP, s, length+1,
1015 (int32_t)UCNV_EXT_FROM_U_GET_PARTIAL_INDEX(value),
1016 pErrorCode);
1017 } else if(extSetUseMapping(which, minLength, value)) {
1018 sa->addString(sa->set, s, length+1);
1019 }
1020 }
1021 }
1022
1023 U_CFUNC void
ucnv_extGetUnicodeSet(const UConverterSharedData * sharedData,const USetAdder * sa,UConverterUnicodeSet which,UConverterSetFilter filter,UErrorCode * pErrorCode)1024 ucnv_extGetUnicodeSet(const UConverterSharedData *sharedData,
1025 const USetAdder *sa,
1026 UConverterUnicodeSet which,
1027 UConverterSetFilter filter,
1028 UErrorCode *pErrorCode) {
1029 const int32_t *cx;
1030 const uint16_t *stage12, *stage3, *ps2, *ps3;
1031 const uint32_t *stage3b;
1032
1033 uint32_t value;
1034 int32_t st1, stage1Length, st2, st3, minLength;
1035
1036 UChar s[UCNV_EXT_MAX_UCHARS];
1037 UChar32 c;
1038 int32_t length;
1039
1040 cx=sharedData->mbcs.extIndexes;
1041 if(cx==NULL) {
1042 return;
1043 }
1044
1045 stage12=UCNV_EXT_ARRAY(cx, UCNV_EXT_FROM_U_STAGE_12_INDEX, uint16_t);
1046 stage3=UCNV_EXT_ARRAY(cx, UCNV_EXT_FROM_U_STAGE_3_INDEX, uint16_t);
1047 stage3b=UCNV_EXT_ARRAY(cx, UCNV_EXT_FROM_U_STAGE_3B_INDEX, uint32_t);
1048
1049 stage1Length=cx[UCNV_EXT_FROM_U_STAGE_1_LENGTH];
1050
1051 /* enumerate the from-Unicode trie table */
1052 c=0; /* keep track of the current code point while enumerating */
1053
1054 if(filter==UCNV_SET_FILTER_2022_CN) {
1055 minLength=3;
1056 } else if( sharedData->mbcs.outputType==MBCS_OUTPUT_DBCS_ONLY ||
1057 filter!=UCNV_SET_FILTER_NONE
1058 ) {
1059 /* DBCS-only, ignore single-byte results */
1060 minLength=2;
1061 } else {
1062 minLength=1;
1063 }
1064
1065 /*
1066 * the trie enumeration is almost the same as
1067 * in MBCSGetUnicodeSet() for MBCS_OUTPUT_1
1068 */
1069 for(st1=0; st1<stage1Length; ++st1) {
1070 st2=stage12[st1];
1071 if(st2>stage1Length) {
1072 ps2=stage12+st2;
1073 for(st2=0; st2<64; ++st2) {
1074 if((st3=(int32_t)ps2[st2]<<UCNV_EXT_STAGE_2_LEFT_SHIFT)!=0) {
1075 /* read the stage 3 block */
1076 ps3=stage3+st3;
1077
1078 do {
1079 value=stage3b[*ps3++];
1080 if(value==0) {
1081 /* no mapping, do nothing */
1082 } else if(UCNV_EXT_FROM_U_IS_PARTIAL(value)) {
1083 // Recurse for partial results.
1084 length=0;
1085 U16_APPEND_UNSAFE(s, length, c);
1086 ucnv_extGetUnicodeSetString(
1087 sharedData, cx, sa, which, minLength,
1088 c, s, length,
1089 (int32_t)UCNV_EXT_FROM_U_GET_PARTIAL_INDEX(value),
1090 pErrorCode);
1091 } else if(extSetUseMapping(which, minLength, value)) {
1092 switch(filter) {
1093 case UCNV_SET_FILTER_2022_CN:
1094 if(!(UCNV_EXT_FROM_U_GET_LENGTH(value)==3 && UCNV_EXT_FROM_U_GET_DATA(value)<=0x82ffff)) {
1095 continue;
1096 }
1097 break;
1098 case UCNV_SET_FILTER_SJIS:
1099 if(!(UCNV_EXT_FROM_U_GET_LENGTH(value)==2 && (value=UCNV_EXT_FROM_U_GET_DATA(value))>=0x8140 && value<=0xeffc)) {
1100 continue;
1101 }
1102 break;
1103 case UCNV_SET_FILTER_GR94DBCS:
1104 if(!(UCNV_EXT_FROM_U_GET_LENGTH(value)==2 &&
1105 (uint16_t)((value=UCNV_EXT_FROM_U_GET_DATA(value))-0xa1a1)<=(0xfefe - 0xa1a1) &&
1106 (uint8_t)(value-0xa1)<=(0xfe - 0xa1))) {
1107 continue;
1108 }
1109 break;
1110 case UCNV_SET_FILTER_HZ:
1111 if(!(UCNV_EXT_FROM_U_GET_LENGTH(value)==2 &&
1112 (uint16_t)((value=UCNV_EXT_FROM_U_GET_DATA(value))-0xa1a1)<=(0xfdfe - 0xa1a1) &&
1113 (uint8_t)(value-0xa1)<=(0xfe - 0xa1))) {
1114 continue;
1115 }
1116 break;
1117 default:
1118 /*
1119 * UCNV_SET_FILTER_NONE,
1120 * or UCNV_SET_FILTER_DBCS_ONLY which is handled via minLength
1121 */
1122 break;
1123 }
1124 sa->add(sa->set, c);
1125 }
1126 } while((++c&0xf)!=0);
1127 } else {
1128 c+=16; /* empty stage 3 block */
1129 }
1130 }
1131 } else {
1132 c+=1024; /* empty stage 2 block */
1133 }
1134 }
1135 }
1136
1137 #endif /* #if !UCONFIG_NO_LEGACY_CONVERSION */
1138