1 /*
2 ******************************************************************************
3 *
4 * Copyright (C) 1999-2012, International Business Machines
5 * Corporation and others. All Rights Reserved.
6 *
7 ******************************************************************************
8 * file name: ubidi.c
9 * encoding: US-ASCII
10 * tab size: 8 (not used)
11 * indentation:4
12 *
13 * created on: 1999jul27
14 * created by: Markus W. Scherer, updated by Matitiahu Allouche
15 */
16
17 #include "cmemory.h"
18 #include "unicode/utypes.h"
19 #include "unicode/ustring.h"
20 #include "unicode/uchar.h"
21 #include "unicode/ubidi.h"
22 #include "unicode/utf16.h"
23 #include "ubidi_props.h"
24 #include "ubidiimp.h"
25 #include "uassert.h"
26
27 /*
28 * General implementation notes:
29 *
30 * Throughout the implementation, there are comments like (W2) that refer to
31 * rules of the BiDi algorithm in its version 5, in this example to the second
32 * rule of the resolution of weak types.
33 *
34 * For handling surrogate pairs, where two UChar's form one "abstract" (or UTF-32)
35 * character according to UTF-16, the second UChar gets the directional property of
36 * the entire character assigned, while the first one gets a BN, a boundary
37 * neutral, type, which is ignored by most of the algorithm according to
38 * rule (X9) and the implementation suggestions of the BiDi algorithm.
39 *
40 * Later, adjustWSLevels() will set the level for each BN to that of the
41 * following character (UChar), which results in surrogate pairs getting the
42 * same level on each of their surrogates.
43 *
44 * In a UTF-8 implementation, the same thing could be done: the last byte of
45 * a multi-byte sequence would get the "real" property, while all previous
46 * bytes of that sequence would get BN.
47 *
48 * It is not possible to assign all those parts of a character the same real
49 * property because this would fail in the resolution of weak types with rules
50 * that look at immediately surrounding types.
51 *
52 * As a related topic, this implementation does not remove Boundary Neutral
53 * types from the input, but ignores them wherever this is relevant.
54 * For example, the loop for the resolution of the weak types reads
55 * types until it finds a non-BN.
56 * Also, explicit embedding codes are neither changed into BN nor removed.
57 * They are only treated the same way real BNs are.
58 * As stated before, adjustWSLevels() takes care of them at the end.
59 * For the purpose of conformance, the levels of all these codes
60 * do not matter.
61 *
62 * Note that this implementation never modifies the dirProps
63 * after the initial setup.
64 *
65 *
66 * In this implementation, the resolution of weak types (Wn),
67 * neutrals (Nn), and the assignment of the resolved level (In)
68 * are all done in one single loop, in resolveImplicitLevels().
69 * Changes of dirProp values are done on the fly, without writing
70 * them back to the dirProps array.
71 *
72 *
73 * This implementation contains code that allows to bypass steps of the
74 * algorithm that are not needed on the specific paragraph
75 * in order to speed up the most common cases considerably,
76 * like text that is entirely LTR, or RTL text without numbers.
77 *
78 * Most of this is done by setting a bit for each directional property
79 * in a flags variable and later checking for whether there are
80 * any LTR characters or any RTL characters, or both, whether
81 * there are any explicit embedding codes, etc.
82 *
83 * If the (Xn) steps are performed, then the flags are re-evaluated,
84 * because they will then not contain the embedding codes any more
85 * and will be adjusted for override codes, so that subsequently
86 * more bypassing may be possible than what the initial flags suggested.
87 *
88 * If the text is not mixed-directional, then the
89 * algorithm steps for the weak type resolution are not performed,
90 * and all levels are set to the paragraph level.
91 *
92 * If there are no explicit embedding codes, then the (Xn) steps
93 * are not performed.
94 *
95 * If embedding levels are supplied as a parameter, then all
96 * explicit embedding codes are ignored, and the (Xn) steps
97 * are not performed.
98 *
99 * White Space types could get the level of the run they belong to,
100 * and are checked with a test of (flags&MASK_EMBEDDING) to
101 * consider if the paragraph direction should be considered in
102 * the flags variable.
103 *
104 * If there are no White Space types in the paragraph, then
105 * (L1) is not necessary in adjustWSLevels().
106 */
107
108 /* to avoid some conditional statements, use tiny constant arrays */
109 static const Flags flagLR[2]={ DIRPROP_FLAG(L), DIRPROP_FLAG(R) };
110 static const Flags flagE[2]={ DIRPROP_FLAG(LRE), DIRPROP_FLAG(RLE) };
111 static const Flags flagO[2]={ DIRPROP_FLAG(LRO), DIRPROP_FLAG(RLO) };
112
113 #define DIRPROP_FLAG_LR(level) flagLR[(level)&1]
114 #define DIRPROP_FLAG_E(level) flagE[(level)&1]
115 #define DIRPROP_FLAG_O(level) flagO[(level)&1]
116
117 /* UBiDi object management -------------------------------------------------- */
118
119 U_CAPI UBiDi * U_EXPORT2
ubidi_open(void)120 ubidi_open(void)
121 {
122 UErrorCode errorCode=U_ZERO_ERROR;
123 return ubidi_openSized(0, 0, &errorCode);
124 }
125
126 U_CAPI UBiDi * U_EXPORT2
ubidi_openSized(int32_t maxLength,int32_t maxRunCount,UErrorCode * pErrorCode)127 ubidi_openSized(int32_t maxLength, int32_t maxRunCount, UErrorCode *pErrorCode) {
128 UBiDi *pBiDi;
129
130 /* check the argument values */
131 if(pErrorCode==NULL || U_FAILURE(*pErrorCode)) {
132 return NULL;
133 } else if(maxLength<0 || maxRunCount<0) {
134 *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
135 return NULL; /* invalid arguments */
136 }
137
138 /* allocate memory for the object */
139 pBiDi=(UBiDi *)uprv_malloc(sizeof(UBiDi));
140 if(pBiDi==NULL) {
141 *pErrorCode=U_MEMORY_ALLOCATION_ERROR;
142 return NULL;
143 }
144
145 /* reset the object, all pointers NULL, all flags FALSE, all sizes 0 */
146 uprv_memset(pBiDi, 0, sizeof(UBiDi));
147
148 /* get BiDi properties */
149 pBiDi->bdp=ubidi_getSingleton();
150
151 /* allocate memory for arrays as requested */
152 if(maxLength>0) {
153 if( !getInitialDirPropsMemory(pBiDi, maxLength) ||
154 !getInitialLevelsMemory(pBiDi, maxLength)
155 ) {
156 *pErrorCode=U_MEMORY_ALLOCATION_ERROR;
157 }
158 } else {
159 pBiDi->mayAllocateText=TRUE;
160 }
161
162 if(maxRunCount>0) {
163 if(maxRunCount==1) {
164 /* use simpleRuns[] */
165 pBiDi->runsSize=sizeof(Run);
166 } else if(!getInitialRunsMemory(pBiDi, maxRunCount)) {
167 *pErrorCode=U_MEMORY_ALLOCATION_ERROR;
168 }
169 } else {
170 pBiDi->mayAllocateRuns=TRUE;
171 }
172
173 if(U_SUCCESS(*pErrorCode)) {
174 return pBiDi;
175 } else {
176 ubidi_close(pBiDi);
177 return NULL;
178 }
179 }
180
181 /*
182 * We are allowed to allocate memory if memory==NULL or
183 * mayAllocate==TRUE for each array that we need.
184 * We also try to grow memory as needed if we
185 * allocate it.
186 *
187 * Assume sizeNeeded>0.
188 * If *pMemory!=NULL, then assume *pSize>0.
189 *
190 * ### this realloc() may unnecessarily copy the old data,
191 * which we know we don't need any more;
192 * is this the best way to do this??
193 */
194 U_CFUNC UBool
ubidi_getMemory(BidiMemoryForAllocation * bidiMem,int32_t * pSize,UBool mayAllocate,int32_t sizeNeeded)195 ubidi_getMemory(BidiMemoryForAllocation *bidiMem, int32_t *pSize, UBool mayAllocate, int32_t sizeNeeded) {
196 void **pMemory = (void **)bidiMem;
197 /* check for existing memory */
198 if(*pMemory==NULL) {
199 /* we need to allocate memory */
200 if(mayAllocate && (*pMemory=uprv_malloc(sizeNeeded))!=NULL) {
201 *pSize=sizeNeeded;
202 return TRUE;
203 } else {
204 return FALSE;
205 }
206 } else {
207 if(sizeNeeded<=*pSize) {
208 /* there is already enough memory */
209 return TRUE;
210 }
211 else if(!mayAllocate) {
212 /* not enough memory, and we must not allocate */
213 return FALSE;
214 } else {
215 /* we try to grow */
216 void *memory;
217 /* in most cases, we do not need the copy-old-data part of
218 * realloc, but it is needed when adding runs using getRunsMemory()
219 * in setParaRunsOnly()
220 */
221 if((memory=uprv_realloc(*pMemory, sizeNeeded))!=NULL) {
222 *pMemory=memory;
223 *pSize=sizeNeeded;
224 return TRUE;
225 } else {
226 /* we failed to grow */
227 return FALSE;
228 }
229 }
230 }
231 }
232
233 U_CAPI void U_EXPORT2
ubidi_close(UBiDi * pBiDi)234 ubidi_close(UBiDi *pBiDi) {
235 if(pBiDi!=NULL) {
236 pBiDi->pParaBiDi=NULL; /* in case one tries to reuse this block */
237 if(pBiDi->dirPropsMemory!=NULL) {
238 uprv_free(pBiDi->dirPropsMemory);
239 }
240 if(pBiDi->levelsMemory!=NULL) {
241 uprv_free(pBiDi->levelsMemory);
242 }
243 if(pBiDi->runsMemory!=NULL) {
244 uprv_free(pBiDi->runsMemory);
245 }
246 if(pBiDi->parasMemory!=NULL) {
247 uprv_free(pBiDi->parasMemory);
248 }
249 if(pBiDi->insertPoints.points!=NULL) {
250 uprv_free(pBiDi->insertPoints.points);
251 }
252
253 uprv_free(pBiDi);
254 }
255 }
256
257 /* set to approximate "inverse BiDi" ---------------------------------------- */
258
259 U_CAPI void U_EXPORT2
ubidi_setInverse(UBiDi * pBiDi,UBool isInverse)260 ubidi_setInverse(UBiDi *pBiDi, UBool isInverse) {
261 if(pBiDi!=NULL) {
262 pBiDi->isInverse=isInverse;
263 pBiDi->reorderingMode = isInverse ? UBIDI_REORDER_INVERSE_NUMBERS_AS_L
264 : UBIDI_REORDER_DEFAULT;
265 }
266 }
267
268 U_CAPI UBool U_EXPORT2
ubidi_isInverse(UBiDi * pBiDi)269 ubidi_isInverse(UBiDi *pBiDi) {
270 if(pBiDi!=NULL) {
271 return pBiDi->isInverse;
272 } else {
273 return FALSE;
274 }
275 }
276
277 /* FOOD FOR THOUGHT: currently the reordering modes are a mixture of
278 * algorithm for direct BiDi, algorithm for inverse BiDi and the bizarre
279 * concept of RUNS_ONLY which is a double operation.
280 * It could be advantageous to divide this into 3 concepts:
281 * a) Operation: direct / inverse / RUNS_ONLY
282 * b) Direct algorithm: default / NUMBERS_SPECIAL / GROUP_NUMBERS_WITH_R
283 * c) Inverse algorithm: default / INVERSE_LIKE_DIRECT / NUMBERS_SPECIAL
284 * This would allow combinations not possible today like RUNS_ONLY with
285 * NUMBERS_SPECIAL.
286 * Also allow to set INSERT_MARKS for the direct step of RUNS_ONLY and
287 * REMOVE_CONTROLS for the inverse step.
288 * Not all combinations would be supported, and probably not all do make sense.
289 * This would need to document which ones are supported and what are the
290 * fallbacks for unsupported combinations.
291 */
292 U_CAPI void U_EXPORT2
ubidi_setReorderingMode(UBiDi * pBiDi,UBiDiReorderingMode reorderingMode)293 ubidi_setReorderingMode(UBiDi *pBiDi, UBiDiReorderingMode reorderingMode) {
294 if ((pBiDi!=NULL) && (reorderingMode >= UBIDI_REORDER_DEFAULT)
295 && (reorderingMode < UBIDI_REORDER_COUNT)) {
296 pBiDi->reorderingMode = reorderingMode;
297 pBiDi->isInverse = (UBool)(reorderingMode == UBIDI_REORDER_INVERSE_NUMBERS_AS_L);
298 }
299 }
300
301 U_CAPI UBiDiReorderingMode U_EXPORT2
ubidi_getReorderingMode(UBiDi * pBiDi)302 ubidi_getReorderingMode(UBiDi *pBiDi) {
303 if (pBiDi!=NULL) {
304 return pBiDi->reorderingMode;
305 } else {
306 return UBIDI_REORDER_DEFAULT;
307 }
308 }
309
310 U_CAPI void U_EXPORT2
ubidi_setReorderingOptions(UBiDi * pBiDi,uint32_t reorderingOptions)311 ubidi_setReorderingOptions(UBiDi *pBiDi, uint32_t reorderingOptions) {
312 if (reorderingOptions & UBIDI_OPTION_REMOVE_CONTROLS) {
313 reorderingOptions&=~UBIDI_OPTION_INSERT_MARKS;
314 }
315 if (pBiDi!=NULL) {
316 pBiDi->reorderingOptions=reorderingOptions;
317 }
318 }
319
320 U_CAPI uint32_t U_EXPORT2
ubidi_getReorderingOptions(UBiDi * pBiDi)321 ubidi_getReorderingOptions(UBiDi *pBiDi) {
322 if (pBiDi!=NULL) {
323 return pBiDi->reorderingOptions;
324 } else {
325 return 0;
326 }
327 }
328
329 U_CAPI UBiDiDirection U_EXPORT2
ubidi_getBaseDirection(const UChar * text,int32_t length)330 ubidi_getBaseDirection(const UChar *text,
331 int32_t length){
332
333 int32_t i;
334 UChar32 uchar;
335 UCharDirection dir;
336
337 if( text==NULL || length<-1 ){
338 return UBIDI_NEUTRAL;
339 }
340
341 if(length==-1) {
342 length=u_strlen(text);
343 }
344
345 for( i = 0 ; i < length; ) {
346 /* i is incremented by U16_NEXT */
347 U16_NEXT(text, i, length, uchar);
348 dir = u_charDirection(uchar);
349 if( dir == U_LEFT_TO_RIGHT )
350 return UBIDI_LTR;
351 if( dir == U_RIGHT_TO_LEFT || dir ==U_RIGHT_TO_LEFT_ARABIC )
352 return UBIDI_RTL;
353 }
354 return UBIDI_NEUTRAL;
355 }
356
357 /* perform (P2)..(P3) ------------------------------------------------------- */
358
359 static DirProp
firstL_R_AL(UBiDi * pBiDi)360 firstL_R_AL(UBiDi *pBiDi) {
361 /* return first strong char after the last B in prologue if any */
362 const UChar *text=pBiDi->prologue;
363 int32_t length=pBiDi->proLength;
364 int32_t i;
365 UChar32 uchar;
366 DirProp dirProp, result=ON;
367 for(i=0; i<length; ) {
368 /* i is incremented by U16_NEXT */
369 U16_NEXT(text, i, length, uchar);
370 dirProp=(DirProp)ubidi_getCustomizedClass(pBiDi, uchar);
371 if(result==ON) {
372 if(dirProp==L || dirProp==R || dirProp==AL) {
373 result=dirProp;
374 }
375 } else {
376 if(dirProp==B) {
377 result=ON;
378 }
379 }
380 }
381 return result;
382 }
383
384 /*
385 * Get the directional properties for the text,
386 * calculate the flags bit-set, and
387 * determine the paragraph level if necessary.
388 */
389 static void
getDirProps(UBiDi * pBiDi)390 getDirProps(UBiDi *pBiDi) {
391 const UChar *text=pBiDi->text;
392 DirProp *dirProps=pBiDi->dirPropsMemory; /* pBiDi->dirProps is const */
393
394 int32_t i=0, i1, length=pBiDi->originalLength;
395 Flags flags=0; /* collect all directionalities in the text */
396 UChar32 uchar;
397 DirProp dirProp=0, paraDirDefault=0;/* initialize to avoid compiler warnings */
398 UBool isDefaultLevel=IS_DEFAULT_LEVEL(pBiDi->paraLevel);
399 /* for inverse BiDi, the default para level is set to RTL if there is a
400 strong R or AL character at either end of the text */
401 UBool isDefaultLevelInverse=isDefaultLevel && (UBool)
402 (pBiDi->reorderingMode==UBIDI_REORDER_INVERSE_LIKE_DIRECT ||
403 pBiDi->reorderingMode==UBIDI_REORDER_INVERSE_FOR_NUMBERS_SPECIAL);
404 int32_t lastArabicPos=-1;
405 int32_t controlCount=0;
406 UBool removeBiDiControls = (UBool)(pBiDi->reorderingOptions &
407 UBIDI_OPTION_REMOVE_CONTROLS);
408
409 typedef enum {
410 NOT_CONTEXTUAL, /* 0: not contextual paraLevel */
411 LOOKING_FOR_STRONG, /* 1: looking for first strong char */
412 FOUND_STRONG_CHAR /* 2: found first strong char */
413 } State;
414 State state;
415 int32_t paraStart=0; /* index of first char in paragraph */
416 DirProp paraDir; /* == CONTEXT_RTL within paragraphs
417 starting with strong R char */
418 DirProp lastStrongDir=0; /* for default level & inverse BiDi */
419 int32_t lastStrongLTR=0; /* for STREAMING option */
420
421 if(pBiDi->reorderingOptions & UBIDI_OPTION_STREAMING) {
422 pBiDi->length=0;
423 lastStrongLTR=0;
424 }
425 if(isDefaultLevel) {
426 DirProp lastStrong;
427 paraDirDefault=pBiDi->paraLevel&1 ? CONTEXT_RTL : 0;
428 if(pBiDi->proLength>0 &&
429 (lastStrong=firstL_R_AL(pBiDi))!=ON) {
430 paraDir=(lastStrong==L) ? 0 : CONTEXT_RTL;
431 state=FOUND_STRONG_CHAR;
432 } else {
433 paraDir=paraDirDefault;
434 state=LOOKING_FOR_STRONG;
435 }
436 lastStrongDir=paraDir;
437 } else {
438 state=NOT_CONTEXTUAL;
439 paraDir=0;
440 }
441 /* count paragraphs and determine the paragraph level (P2..P3) */
442 /*
443 * see comment in ubidi.h:
444 * the DEFAULT_XXX values are designed so that
445 * their bit 0 alone yields the intended default
446 */
447 for( /* i=0 above */ ; i<length; ) {
448 /* i is incremented by U16_NEXT */
449 U16_NEXT(text, i, length, uchar);
450 flags|=DIRPROP_FLAG(dirProp=(DirProp)ubidi_getCustomizedClass(pBiDi, uchar));
451 dirProps[i-1]=dirProp|paraDir;
452 if(uchar>0xffff) { /* set the lead surrogate's property to BN */
453 flags|=DIRPROP_FLAG(BN);
454 dirProps[i-2]=(DirProp)(BN|paraDir);
455 }
456 if(state==LOOKING_FOR_STRONG) {
457 if(dirProp==L) {
458 state=FOUND_STRONG_CHAR;
459 if(paraDir) {
460 paraDir=0;
461 for(i1=paraStart; i1<i; i1++) {
462 dirProps[i1]&=~CONTEXT_RTL;
463 }
464 }
465 continue;
466 }
467 if(dirProp==R || dirProp==AL) {
468 state=FOUND_STRONG_CHAR;
469 if(paraDir==0) {
470 paraDir=CONTEXT_RTL;
471 for(i1=paraStart; i1<i; i1++) {
472 dirProps[i1]|=CONTEXT_RTL;
473 }
474 }
475 continue;
476 }
477 }
478 if(dirProp==L) {
479 lastStrongDir=0;
480 lastStrongLTR=i; /* i is index to next character */
481 }
482 else if(dirProp==R) {
483 lastStrongDir=CONTEXT_RTL;
484 }
485 else if(dirProp==AL) {
486 lastStrongDir=CONTEXT_RTL;
487 lastArabicPos=i-1;
488 }
489 else if(dirProp==B) {
490 if(pBiDi->reorderingOptions & UBIDI_OPTION_STREAMING) {
491 pBiDi->length=i; /* i is index to next character */
492 }
493 if(isDefaultLevelInverse && (lastStrongDir==CONTEXT_RTL) &&(paraDir!=lastStrongDir)) {
494 for( ; paraStart<i; paraStart++) {
495 dirProps[paraStart]|=CONTEXT_RTL;
496 }
497 }
498 if(i<length) { /* B not last char in text */
499 if(!((uchar==CR) && (text[i]==LF))) {
500 pBiDi->paraCount++;
501 }
502 if(isDefaultLevel) {
503 state=LOOKING_FOR_STRONG;
504 paraStart=i; /* i is index to next character */
505 paraDir=paraDirDefault;
506 lastStrongDir=paraDirDefault;
507 }
508 }
509 }
510 if(removeBiDiControls && IS_BIDI_CONTROL_CHAR(uchar)) {
511 controlCount++;
512 }
513 }
514 if(isDefaultLevelInverse && (lastStrongDir==CONTEXT_RTL) &&(paraDir!=lastStrongDir)) {
515 for(i1=paraStart; i1<length; i1++) {
516 dirProps[i1]|=CONTEXT_RTL;
517 }
518 }
519 if(isDefaultLevel) {
520 pBiDi->paraLevel=GET_PARALEVEL(pBiDi, 0);
521 }
522 if(pBiDi->reorderingOptions & UBIDI_OPTION_STREAMING) {
523 if((lastStrongLTR>pBiDi->length) &&
524 (GET_PARALEVEL(pBiDi, lastStrongLTR)==0)) {
525 pBiDi->length = lastStrongLTR;
526 }
527 if(pBiDi->length<pBiDi->originalLength) {
528 pBiDi->paraCount--;
529 }
530 }
531 /* The following line does nothing new for contextual paraLevel, but is
532 needed for absolute paraLevel. */
533 flags|=DIRPROP_FLAG_LR(pBiDi->paraLevel);
534
535 if(pBiDi->orderParagraphsLTR && (flags&DIRPROP_FLAG(B))) {
536 flags|=DIRPROP_FLAG(L);
537 }
538
539 pBiDi->controlCount = controlCount;
540 pBiDi->flags=flags;
541 pBiDi->lastArabicPos=lastArabicPos;
542 }
543
544 /* perform (X1)..(X9) ------------------------------------------------------- */
545
546 /* determine if the text is mixed-directional or single-directional */
547 static UBiDiDirection
directionFromFlags(UBiDi * pBiDi)548 directionFromFlags(UBiDi *pBiDi) {
549 Flags flags=pBiDi->flags;
550 /* if the text contains AN and neutrals, then some neutrals may become RTL */
551 if(!(flags&MASK_RTL || ((flags&DIRPROP_FLAG(AN)) && (flags&MASK_POSSIBLE_N)))) {
552 return UBIDI_LTR;
553 } else if(!(flags&MASK_LTR)) {
554 return UBIDI_RTL;
555 } else {
556 return UBIDI_MIXED;
557 }
558 }
559
560 /*
561 * Resolve the explicit levels as specified by explicit embedding codes.
562 * Recalculate the flags to have them reflect the real properties
563 * after taking the explicit embeddings into account.
564 *
565 * The BiDi algorithm is designed to result in the same behavior whether embedding
566 * levels are externally specified (from "styled text", supposedly the preferred
567 * method) or set by explicit embedding codes (LRx, RLx, PDF) in the plain text.
568 * That is why (X9) instructs to remove all explicit codes (and BN).
569 * However, in a real implementation, this removal of these codes and their index
570 * positions in the plain text is undesirable since it would result in
571 * reallocated, reindexed text.
572 * Instead, this implementation leaves the codes in there and just ignores them
573 * in the subsequent processing.
574 * In order to get the same reordering behavior, positions with a BN or an
575 * explicit embedding code just get the same level assigned as the last "real"
576 * character.
577 *
578 * Some implementations, not this one, then overwrite some of these
579 * directionality properties at "real" same-level-run boundaries by
580 * L or R codes so that the resolution of weak types can be performed on the
581 * entire paragraph at once instead of having to parse it once more and
582 * perform that resolution on same-level-runs.
583 * This limits the scope of the implicit rules in effectively
584 * the same way as the run limits.
585 *
586 * Instead, this implementation does not modify these codes.
587 * On one hand, the paragraph has to be scanned for same-level-runs, but
588 * on the other hand, this saves another loop to reset these codes,
589 * or saves making and modifying a copy of dirProps[].
590 *
591 *
592 * Note that (Pn) and (Xn) changed significantly from version 4 of the BiDi algorithm.
593 *
594 *
595 * Handling the stack of explicit levels (Xn):
596 *
597 * With the BiDi stack of explicit levels,
598 * as pushed with each LRE, RLE, LRO, and RLO and popped with each PDF,
599 * the explicit level must never exceed UBIDI_MAX_EXPLICIT_LEVEL==61.
600 *
601 * In order to have a correct push-pop semantics even in the case of overflows,
602 * there are two overflow counters:
603 * - countOver60 is incremented with each LRx at level 60
604 * - from level 60, one RLx increases the level to 61
605 * - countOver61 is incremented with each LRx and RLx at level 61
606 *
607 * Popping levels with PDF must work in the opposite order so that level 61
608 * is correct at the correct point. Underflows (too many PDFs) must be checked.
609 *
610 * This implementation assumes that UBIDI_MAX_EXPLICIT_LEVEL is odd.
611 */
612 static UBiDiDirection
resolveExplicitLevels(UBiDi * pBiDi)613 resolveExplicitLevels(UBiDi *pBiDi) {
614 const DirProp *dirProps=pBiDi->dirProps;
615 UBiDiLevel *levels=pBiDi->levels;
616 const UChar *text=pBiDi->text;
617
618 int32_t i=0, length=pBiDi->length;
619 Flags flags=pBiDi->flags; /* collect all directionalities in the text */
620 DirProp dirProp;
621 UBiDiLevel level=GET_PARALEVEL(pBiDi, 0);
622
623 UBiDiDirection direction;
624 int32_t paraIndex=0;
625
626 /* determine if the text is mixed-directional or single-directional */
627 direction=directionFromFlags(pBiDi);
628
629 /* we may not need to resolve any explicit levels, but for multiple
630 paragraphs we want to loop on all chars to set the para boundaries */
631 if((direction!=UBIDI_MIXED) && (pBiDi->paraCount==1)) {
632 /* not mixed directionality: levels don't matter - trailingWSStart will be 0 */
633 } else if((pBiDi->paraCount==1) &&
634 (!(flags&MASK_EXPLICIT) ||
635 (pBiDi->reorderingMode > UBIDI_REORDER_LAST_LOGICAL_TO_VISUAL))) {
636 /* mixed, but all characters are at the same embedding level */
637 /* or we are in "inverse BiDi" */
638 /* and we don't have contextual multiple paragraphs with some B char */
639 /* set all levels to the paragraph level */
640 for(i=0; i<length; ++i) {
641 levels[i]=level;
642 }
643 } else {
644 /* continue to perform (Xn) */
645
646 /* (X1) level is set for all codes, embeddingLevel keeps track of the push/pop operations */
647 /* both variables may carry the UBIDI_LEVEL_OVERRIDE flag to indicate the override status */
648 UBiDiLevel embeddingLevel=level, newLevel, stackTop=0;
649
650 UBiDiLevel stack[UBIDI_MAX_EXPLICIT_LEVEL]; /* we never push anything >=UBIDI_MAX_EXPLICIT_LEVEL */
651 uint32_t countOver60=0, countOver61=0; /* count overflows of explicit levels */
652
653 /* recalculate the flags */
654 flags=0;
655
656 for(i=0; i<length; ++i) {
657 dirProp=NO_CONTEXT_RTL(dirProps[i]);
658 switch(dirProp) {
659 case LRE:
660 case LRO:
661 /* (X3, X5) */
662 newLevel=(UBiDiLevel)((embeddingLevel+2)&~(UBIDI_LEVEL_OVERRIDE|1)); /* least greater even level */
663 if(newLevel<=UBIDI_MAX_EXPLICIT_LEVEL) {
664 stack[stackTop]=embeddingLevel;
665 ++stackTop;
666 embeddingLevel=newLevel;
667 if(dirProp==LRO) {
668 embeddingLevel|=UBIDI_LEVEL_OVERRIDE;
669 }
670 /* we don't need to set UBIDI_LEVEL_OVERRIDE off for LRE
671 since this has already been done for newLevel which is
672 the source for embeddingLevel.
673 */
674 } else if((embeddingLevel&~UBIDI_LEVEL_OVERRIDE)==UBIDI_MAX_EXPLICIT_LEVEL) {
675 ++countOver61;
676 } else /* (embeddingLevel&~UBIDI_LEVEL_OVERRIDE)==UBIDI_MAX_EXPLICIT_LEVEL-1 */ {
677 ++countOver60;
678 }
679 flags|=DIRPROP_FLAG(BN);
680 break;
681 case RLE:
682 case RLO:
683 /* (X2, X4) */
684 newLevel=(UBiDiLevel)(((embeddingLevel&~UBIDI_LEVEL_OVERRIDE)+1)|1); /* least greater odd level */
685 if(newLevel<=UBIDI_MAX_EXPLICIT_LEVEL) {
686 stack[stackTop]=embeddingLevel;
687 ++stackTop;
688 embeddingLevel=newLevel;
689 if(dirProp==RLO) {
690 embeddingLevel|=UBIDI_LEVEL_OVERRIDE;
691 }
692 /* we don't need to set UBIDI_LEVEL_OVERRIDE off for RLE
693 since this has already been done for newLevel which is
694 the source for embeddingLevel.
695 */
696 } else {
697 ++countOver61;
698 }
699 flags|=DIRPROP_FLAG(BN);
700 break;
701 case PDF:
702 /* (X7) */
703 /* handle all the overflow cases first */
704 if(countOver61>0) {
705 --countOver61;
706 } else if(countOver60>0 && (embeddingLevel&~UBIDI_LEVEL_OVERRIDE)!=UBIDI_MAX_EXPLICIT_LEVEL) {
707 /* handle LRx overflows from level 60 */
708 --countOver60;
709 } else if(stackTop>0) {
710 /* this is the pop operation; it also pops level 61 while countOver60>0 */
711 --stackTop;
712 embeddingLevel=stack[stackTop];
713 /* } else { (underflow) */
714 }
715 flags|=DIRPROP_FLAG(BN);
716 break;
717 case B:
718 stackTop=0;
719 countOver60=countOver61=0;
720 level=GET_PARALEVEL(pBiDi, i);
721 if((i+1)<length) {
722 embeddingLevel=GET_PARALEVEL(pBiDi, i+1);
723 if(!((text[i]==CR) && (text[i+1]==LF))) {
724 pBiDi->paras[paraIndex++]=i+1;
725 }
726 }
727 flags|=DIRPROP_FLAG(B);
728 break;
729 case BN:
730 /* BN, LRE, RLE, and PDF are supposed to be removed (X9) */
731 /* they will get their levels set correctly in adjustWSLevels() */
732 flags|=DIRPROP_FLAG(BN);
733 break;
734 default:
735 /* all other types get the "real" level */
736 if(level!=embeddingLevel) {
737 level=embeddingLevel;
738 if(level&UBIDI_LEVEL_OVERRIDE) {
739 flags|=DIRPROP_FLAG_O(level)|DIRPROP_FLAG_MULTI_RUNS;
740 } else {
741 flags|=DIRPROP_FLAG_E(level)|DIRPROP_FLAG_MULTI_RUNS;
742 }
743 }
744 if(!(level&UBIDI_LEVEL_OVERRIDE)) {
745 flags|=DIRPROP_FLAG(dirProp);
746 }
747 break;
748 }
749
750 /*
751 * We need to set reasonable levels even on BN codes and
752 * explicit codes because we will later look at same-level runs (X10).
753 */
754 levels[i]=level;
755 }
756 if(flags&MASK_EMBEDDING) {
757 flags|=DIRPROP_FLAG_LR(pBiDi->paraLevel);
758 }
759 if(pBiDi->orderParagraphsLTR && (flags&DIRPROP_FLAG(B))) {
760 flags|=DIRPROP_FLAG(L);
761 }
762
763 /* subsequently, ignore the explicit codes and BN (X9) */
764
765 /* again, determine if the text is mixed-directional or single-directional */
766 pBiDi->flags=flags;
767 direction=directionFromFlags(pBiDi);
768 }
769
770 return direction;
771 }
772
773 /*
774 * Use a pre-specified embedding levels array:
775 *
776 * Adjust the directional properties for overrides (->LEVEL_OVERRIDE),
777 * ignore all explicit codes (X9),
778 * and check all the preset levels.
779 *
780 * Recalculate the flags to have them reflect the real properties
781 * after taking the explicit embeddings into account.
782 */
783 static UBiDiDirection
checkExplicitLevels(UBiDi * pBiDi,UErrorCode * pErrorCode)784 checkExplicitLevels(UBiDi *pBiDi, UErrorCode *pErrorCode) {
785 const DirProp *dirProps=pBiDi->dirProps;
786 DirProp dirProp;
787 UBiDiLevel *levels=pBiDi->levels;
788 const UChar *text=pBiDi->text;
789
790 int32_t i, length=pBiDi->length;
791 Flags flags=0; /* collect all directionalities in the text */
792 UBiDiLevel level;
793 uint32_t paraIndex=0;
794
795 for(i=0; i<length; ++i) {
796 level=levels[i];
797 dirProp=NO_CONTEXT_RTL(dirProps[i]);
798 if(level&UBIDI_LEVEL_OVERRIDE) {
799 /* keep the override flag in levels[i] but adjust the flags */
800 level&=~UBIDI_LEVEL_OVERRIDE; /* make the range check below simpler */
801 flags|=DIRPROP_FLAG_O(level);
802 } else {
803 /* set the flags */
804 flags|=DIRPROP_FLAG_E(level)|DIRPROP_FLAG(dirProp);
805 }
806 if((level<GET_PARALEVEL(pBiDi, i) &&
807 !((0==level)&&(dirProp==B))) ||
808 (UBIDI_MAX_EXPLICIT_LEVEL<level)) {
809 /* level out of bounds */
810 *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
811 return UBIDI_LTR;
812 }
813 if((dirProp==B) && ((i+1)<length)) {
814 if(!((text[i]==CR) && (text[i+1]==LF))) {
815 pBiDi->paras[paraIndex++]=i+1;
816 }
817 }
818 }
819 if(flags&MASK_EMBEDDING) {
820 flags|=DIRPROP_FLAG_LR(pBiDi->paraLevel);
821 }
822
823 /* determine if the text is mixed-directional or single-directional */
824 pBiDi->flags=flags;
825 return directionFromFlags(pBiDi);
826 }
827
828 /******************************************************************
829 The Properties state machine table
830 *******************************************************************
831
832 All table cells are 8 bits:
833 bits 0..4: next state
834 bits 5..7: action to perform (if > 0)
835
836 Cells may be of format "n" where n represents the next state
837 (except for the rightmost column).
838 Cells may also be of format "s(x,y)" where x represents an action
839 to perform and y represents the next state.
840
841 *******************************************************************
842 Definitions and type for properties state table
843 *******************************************************************
844 */
845 #define IMPTABPROPS_COLUMNS 14
846 #define IMPTABPROPS_RES (IMPTABPROPS_COLUMNS - 1)
847 #define GET_STATEPROPS(cell) ((cell)&0x1f)
848 #define GET_ACTIONPROPS(cell) ((cell)>>5)
849 #define s(action, newState) ((uint8_t)(newState+(action<<5)))
850
851 static const uint8_t groupProp[] = /* dirProp regrouped */
852 {
853 /* L R EN ES ET AN CS B S WS ON LRE LRO AL RLE RLO PDF NSM BN */
854 0, 1, 2, 7, 8, 3, 9, 6, 5, 4, 4, 10, 10, 12, 10, 10, 10, 11, 10
855 };
856 enum { DirProp_L=0, DirProp_R=1, DirProp_EN=2, DirProp_AN=3, DirProp_ON=4, DirProp_S=5, DirProp_B=6 }; /* reduced dirProp */
857
858 /******************************************************************
859
860 PROPERTIES STATE TABLE
861
862 In table impTabProps,
863 - the ON column regroups ON and WS
864 - the BN column regroups BN, LRE, RLE, LRO, RLO, PDF
865 - the Res column is the reduced property assigned to a run
866
867 Action 1: process current run1, init new run1
868 2: init new run2
869 3: process run1, process run2, init new run1
870 4: process run1, set run1=run2, init new run2
871
872 Notes:
873 1) This table is used in resolveImplicitLevels().
874 2) This table triggers actions when there is a change in the Bidi
875 property of incoming characters (action 1).
876 3) Most such property sequences are processed immediately (in
877 fact, passed to processPropertySeq().
878 4) However, numbers are assembled as one sequence. This means
879 that undefined situations (like CS following digits, until
880 it is known if the next char will be a digit) are held until
881 following chars define them.
882 Example: digits followed by CS, then comes another CS or ON;
883 the digits will be processed, then the CS assigned
884 as the start of an ON sequence (action 3).
885 5) There are cases where more than one sequence must be
886 processed, for instance digits followed by CS followed by L:
887 the digits must be processed as one sequence, and the CS
888 must be processed as an ON sequence, all this before starting
889 assembling chars for the opening L sequence.
890
891
892 */
893 static const uint8_t impTabProps[][IMPTABPROPS_COLUMNS] =
894 {
895 /* L , R , EN , AN , ON , S , B , ES , ET , CS , BN , NSM , AL , Res */
896 /* 0 Init */ { 1 , 2 , 4 , 5 , 7 , 15 , 17 , 7 , 9 , 7 , 0 , 7 , 3 , DirProp_ON },
897 /* 1 L */ { 1 , s(1,2), s(1,4), s(1,5), s(1,7),s(1,15),s(1,17), s(1,7), s(1,9), s(1,7), 1 , 1 , s(1,3), DirProp_L },
898 /* 2 R */ { s(1,1), 2 , s(1,4), s(1,5), s(1,7),s(1,15),s(1,17), s(1,7), s(1,9), s(1,7), 2 , 2 , s(1,3), DirProp_R },
899 /* 3 AL */ { s(1,1), s(1,2), s(1,6), s(1,6), s(1,8),s(1,16),s(1,17), s(1,8), s(1,8), s(1,8), 3 , 3 , 3 , DirProp_R },
900 /* 4 EN */ { s(1,1), s(1,2), 4 , s(1,5), s(1,7),s(1,15),s(1,17),s(2,10), 11 ,s(2,10), 4 , 4 , s(1,3), DirProp_EN },
901 /* 5 AN */ { s(1,1), s(1,2), s(1,4), 5 , s(1,7),s(1,15),s(1,17), s(1,7), s(1,9),s(2,12), 5 , 5 , s(1,3), DirProp_AN },
902 /* 6 AL:EN/AN */ { s(1,1), s(1,2), 6 , 6 , s(1,8),s(1,16),s(1,17), s(1,8), s(1,8),s(2,13), 6 , 6 , s(1,3), DirProp_AN },
903 /* 7 ON */ { s(1,1), s(1,2), s(1,4), s(1,5), 7 ,s(1,15),s(1,17), 7 ,s(2,14), 7 , 7 , 7 , s(1,3), DirProp_ON },
904 /* 8 AL:ON */ { s(1,1), s(1,2), s(1,6), s(1,6), 8 ,s(1,16),s(1,17), 8 , 8 , 8 , 8 , 8 , s(1,3), DirProp_ON },
905 /* 9 ET */ { s(1,1), s(1,2), 4 , s(1,5), 7 ,s(1,15),s(1,17), 7 , 9 , 7 , 9 , 9 , s(1,3), DirProp_ON },
906 /*10 EN+ES/CS */ { s(3,1), s(3,2), 4 , s(3,5), s(4,7),s(3,15),s(3,17), s(4,7),s(4,14), s(4,7), 10 , s(4,7), s(3,3), DirProp_EN },
907 /*11 EN+ET */ { s(1,1), s(1,2), 4 , s(1,5), s(1,7),s(1,15),s(1,17), s(1,7), 11 , s(1,7), 11 , 11 , s(1,3), DirProp_EN },
908 /*12 AN+CS */ { s(3,1), s(3,2), s(3,4), 5 , s(4,7),s(3,15),s(3,17), s(4,7),s(4,14), s(4,7), 12 , s(4,7), s(3,3), DirProp_AN },
909 /*13 AL:EN/AN+CS */ { s(3,1), s(3,2), 6 , 6 , s(4,8),s(3,16),s(3,17), s(4,8), s(4,8), s(4,8), 13 , s(4,8), s(3,3), DirProp_AN },
910 /*14 ON+ET */ { s(1,1), s(1,2), s(4,4), s(1,5), 7 ,s(1,15),s(1,17), 7 , 14 , 7 , 14 , 14 , s(1,3), DirProp_ON },
911 /*15 S */ { s(1,1), s(1,2), s(1,4), s(1,5), s(1,7), 15 ,s(1,17), s(1,7), s(1,9), s(1,7), 15 , s(1,7), s(1,3), DirProp_S },
912 /*16 AL:S */ { s(1,1), s(1,2), s(1,6), s(1,6), s(1,8), 16 ,s(1,17), s(1,8), s(1,8), s(1,8), 16 , s(1,8), s(1,3), DirProp_S },
913 /*17 B */ { s(1,1), s(1,2), s(1,4), s(1,5), s(1,7),s(1,15), 17 , s(1,7), s(1,9), s(1,7), 17 , s(1,7), s(1,3), DirProp_B }
914 };
915
916 /* we must undef macro s because the levels table have a different
917 * structure (4 bits for action and 4 bits for next state.
918 */
919 #undef s
920
921 /******************************************************************
922 The levels state machine tables
923 *******************************************************************
924
925 All table cells are 8 bits:
926 bits 0..3: next state
927 bits 4..7: action to perform (if > 0)
928
929 Cells may be of format "n" where n represents the next state
930 (except for the rightmost column).
931 Cells may also be of format "s(x,y)" where x represents an action
932 to perform and y represents the next state.
933
934 This format limits each table to 16 states each and to 15 actions.
935
936 *******************************************************************
937 Definitions and type for levels state tables
938 *******************************************************************
939 */
940 #define IMPTABLEVELS_COLUMNS (DirProp_B + 2)
941 #define IMPTABLEVELS_RES (IMPTABLEVELS_COLUMNS - 1)
942 #define GET_STATE(cell) ((cell)&0x0f)
943 #define GET_ACTION(cell) ((cell)>>4)
944 #define s(action, newState) ((uint8_t)(newState+(action<<4)))
945
946 typedef uint8_t ImpTab[][IMPTABLEVELS_COLUMNS];
947 typedef uint8_t ImpAct[];
948
949 /* FOOD FOR THOUGHT: each ImpTab should have its associated ImpAct,
950 * instead of having a pair of ImpTab and a pair of ImpAct.
951 */
952 typedef struct ImpTabPair {
953 const void * pImpTab[2];
954 const void * pImpAct[2];
955 } ImpTabPair;
956
957 /******************************************************************
958
959 LEVELS STATE TABLES
960
961 In all levels state tables,
962 - state 0 is the initial state
963 - the Res column is the increment to add to the text level
964 for this property sequence.
965
966 The impAct arrays for each table of a pair map the local action
967 numbers of the table to the total list of actions. For instance,
968 action 2 in a given table corresponds to the action number which
969 appears in entry [2] of the impAct array for that table.
970 The first entry of all impAct arrays must be 0.
971
972 Action 1: init conditional sequence
973 2: prepend conditional sequence to current sequence
974 3: set ON sequence to new level - 1
975 4: init EN/AN/ON sequence
976 5: fix EN/AN/ON sequence followed by R
977 6: set previous level sequence to level 2
978
979 Notes:
980 1) These tables are used in processPropertySeq(). The input
981 is property sequences as determined by resolveImplicitLevels.
982 2) Most such property sequences are processed immediately
983 (levels are assigned).
984 3) However, some sequences cannot be assigned a final level till
985 one or more following sequences are received. For instance,
986 ON following an R sequence within an even-level paragraph.
987 If the following sequence is R, the ON sequence will be
988 assigned basic run level+1, and so will the R sequence.
989 4) S is generally handled like ON, since its level will be fixed
990 to paragraph level in adjustWSLevels().
991
992 */
993
994 static const ImpTab impTabL_DEFAULT = /* Even paragraph level */
995 /* In this table, conditional sequences receive the higher possible level
996 until proven otherwise.
997 */
998 {
999 /* L , R , EN , AN , ON , S , B , Res */
1000 /* 0 : init */ { 0 , 1 , 0 , 2 , 0 , 0 , 0 , 0 },
1001 /* 1 : R */ { 0 , 1 , 3 , 3 , s(1,4), s(1,4), 0 , 1 },
1002 /* 2 : AN */ { 0 , 1 , 0 , 2 , s(1,5), s(1,5), 0 , 2 },
1003 /* 3 : R+EN/AN */ { 0 , 1 , 3 , 3 , s(1,4), s(1,4), 0 , 2 },
1004 /* 4 : R+ON */ { s(2,0), 1 , 3 , 3 , 4 , 4 , s(2,0), 1 },
1005 /* 5 : AN+ON */ { s(2,0), 1 , s(2,0), 2 , 5 , 5 , s(2,0), 1 }
1006 };
1007 static const ImpTab impTabR_DEFAULT = /* Odd paragraph level */
1008 /* In this table, conditional sequences receive the lower possible level
1009 until proven otherwise.
1010 */
1011 {
1012 /* L , R , EN , AN , ON , S , B , Res */
1013 /* 0 : init */ { 1 , 0 , 2 , 2 , 0 , 0 , 0 , 0 },
1014 /* 1 : L */ { 1 , 0 , 1 , 3 , s(1,4), s(1,4), 0 , 1 },
1015 /* 2 : EN/AN */ { 1 , 0 , 2 , 2 , 0 , 0 , 0 , 1 },
1016 /* 3 : L+AN */ { 1 , 0 , 1 , 3 , 5 , 5 , 0 , 1 },
1017 /* 4 : L+ON */ { s(2,1), 0 , s(2,1), 3 , 4 , 4 , 0 , 0 },
1018 /* 5 : L+AN+ON */ { 1 , 0 , 1 , 3 , 5 , 5 , 0 , 0 }
1019 };
1020 static const ImpAct impAct0 = {0,1,2,3,4,5,6};
1021 static const ImpTabPair impTab_DEFAULT = {{&impTabL_DEFAULT,
1022 &impTabR_DEFAULT},
1023 {&impAct0, &impAct0}};
1024
1025 static const ImpTab impTabL_NUMBERS_SPECIAL = /* Even paragraph level */
1026 /* In this table, conditional sequences receive the higher possible level
1027 until proven otherwise.
1028 */
1029 {
1030 /* L , R , EN , AN , ON , S , B , Res */
1031 /* 0 : init */ { 0 , 2 , 1 , 1 , 0 , 0 , 0 , 0 },
1032 /* 1 : L+EN/AN */ { 0 , 2 , 1 , 1 , 0 , 0 , 0 , 2 },
1033 /* 2 : R */ { 0 , 2 , 4 , 4 , s(1,3), 0 , 0 , 1 },
1034 /* 3 : R+ON */ { s(2,0), 2 , 4 , 4 , 3 , 3 , s(2,0), 1 },
1035 /* 4 : R+EN/AN */ { 0 , 2 , 4 , 4 , s(1,3), s(1,3), 0 , 2 }
1036 };
1037 static const ImpTabPair impTab_NUMBERS_SPECIAL = {{&impTabL_NUMBERS_SPECIAL,
1038 &impTabR_DEFAULT},
1039 {&impAct0, &impAct0}};
1040
1041 static const ImpTab impTabL_GROUP_NUMBERS_WITH_R =
1042 /* In this table, EN/AN+ON sequences receive levels as if associated with R
1043 until proven that there is L or sor/eor on both sides. AN is handled like EN.
1044 */
1045 {
1046 /* L , R , EN , AN , ON , S , B , Res */
1047 /* 0 init */ { 0 , 3 , s(1,1), s(1,1), 0 , 0 , 0 , 0 },
1048 /* 1 EN/AN */ { s(2,0), 3 , 1 , 1 , 2 , s(2,0), s(2,0), 2 },
1049 /* 2 EN/AN+ON */ { s(2,0), 3 , 1 , 1 , 2 , s(2,0), s(2,0), 1 },
1050 /* 3 R */ { 0 , 3 , 5 , 5 , s(1,4), 0 , 0 , 1 },
1051 /* 4 R+ON */ { s(2,0), 3 , 5 , 5 , 4 , s(2,0), s(2,0), 1 },
1052 /* 5 R+EN/AN */ { 0 , 3 , 5 , 5 , s(1,4), 0 , 0 , 2 }
1053 };
1054 static const ImpTab impTabR_GROUP_NUMBERS_WITH_R =
1055 /* In this table, EN/AN+ON sequences receive levels as if associated with R
1056 until proven that there is L on both sides. AN is handled like EN.
1057 */
1058 {
1059 /* L , R , EN , AN , ON , S , B , Res */
1060 /* 0 init */ { 2 , 0 , 1 , 1 , 0 , 0 , 0 , 0 },
1061 /* 1 EN/AN */ { 2 , 0 , 1 , 1 , 0 , 0 , 0 , 1 },
1062 /* 2 L */ { 2 , 0 , s(1,4), s(1,4), s(1,3), 0 , 0 , 1 },
1063 /* 3 L+ON */ { s(2,2), 0 , 4 , 4 , 3 , 0 , 0 , 0 },
1064 /* 4 L+EN/AN */ { s(2,2), 0 , 4 , 4 , 3 , 0 , 0 , 1 }
1065 };
1066 static const ImpTabPair impTab_GROUP_NUMBERS_WITH_R = {
1067 {&impTabL_GROUP_NUMBERS_WITH_R,
1068 &impTabR_GROUP_NUMBERS_WITH_R},
1069 {&impAct0, &impAct0}};
1070
1071
1072 static const ImpTab impTabL_INVERSE_NUMBERS_AS_L =
1073 /* This table is identical to the Default LTR table except that EN and AN are
1074 handled like L.
1075 */
1076 {
1077 /* L , R , EN , AN , ON , S , B , Res */
1078 /* 0 : init */ { 0 , 1 , 0 , 0 , 0 , 0 , 0 , 0 },
1079 /* 1 : R */ { 0 , 1 , 0 , 0 , s(1,4), s(1,4), 0 , 1 },
1080 /* 2 : AN */ { 0 , 1 , 0 , 0 , s(1,5), s(1,5), 0 , 2 },
1081 /* 3 : R+EN/AN */ { 0 , 1 , 0 , 0 , s(1,4), s(1,4), 0 , 2 },
1082 /* 4 : R+ON */ { s(2,0), 1 , s(2,0), s(2,0), 4 , 4 , s(2,0), 1 },
1083 /* 5 : AN+ON */ { s(2,0), 1 , s(2,0), s(2,0), 5 , 5 , s(2,0), 1 }
1084 };
1085 static const ImpTab impTabR_INVERSE_NUMBERS_AS_L =
1086 /* This table is identical to the Default RTL table except that EN and AN are
1087 handled like L.
1088 */
1089 {
1090 /* L , R , EN , AN , ON , S , B , Res */
1091 /* 0 : init */ { 1 , 0 , 1 , 1 , 0 , 0 , 0 , 0 },
1092 /* 1 : L */ { 1 , 0 , 1 , 1 , s(1,4), s(1,4), 0 , 1 },
1093 /* 2 : EN/AN */ { 1 , 0 , 1 , 1 , 0 , 0 , 0 , 1 },
1094 /* 3 : L+AN */ { 1 , 0 , 1 , 1 , 5 , 5 , 0 , 1 },
1095 /* 4 : L+ON */ { s(2,1), 0 , s(2,1), s(2,1), 4 , 4 , 0 , 0 },
1096 /* 5 : L+AN+ON */ { 1 , 0 , 1 , 1 , 5 , 5 , 0 , 0 }
1097 };
1098 static const ImpTabPair impTab_INVERSE_NUMBERS_AS_L = {
1099 {&impTabL_INVERSE_NUMBERS_AS_L,
1100 &impTabR_INVERSE_NUMBERS_AS_L},
1101 {&impAct0, &impAct0}};
1102
1103 static const ImpTab impTabR_INVERSE_LIKE_DIRECT = /* Odd paragraph level */
1104 /* In this table, conditional sequences receive the lower possible level
1105 until proven otherwise.
1106 */
1107 {
1108 /* L , R , EN , AN , ON , S , B , Res */
1109 /* 0 : init */ { 1 , 0 , 2 , 2 , 0 , 0 , 0 , 0 },
1110 /* 1 : L */ { 1 , 0 , 1 , 2 , s(1,3), s(1,3), 0 , 1 },
1111 /* 2 : EN/AN */ { 1 , 0 , 2 , 2 , 0 , 0 , 0 , 1 },
1112 /* 3 : L+ON */ { s(2,1), s(3,0), 6 , 4 , 3 , 3 , s(3,0), 0 },
1113 /* 4 : L+ON+AN */ { s(2,1), s(3,0), 6 , 4 , 5 , 5 , s(3,0), 3 },
1114 /* 5 : L+AN+ON */ { s(2,1), s(3,0), 6 , 4 , 5 , 5 , s(3,0), 2 },
1115 /* 6 : L+ON+EN */ { s(2,1), s(3,0), 6 , 4 , 3 , 3 , s(3,0), 1 }
1116 };
1117 static const ImpAct impAct1 = {0,1,11,12};
1118 /* FOOD FOR THOUGHT: in LTR table below, check case "JKL 123abc"
1119 */
1120 static const ImpTabPair impTab_INVERSE_LIKE_DIRECT = {
1121 {&impTabL_DEFAULT,
1122 &impTabR_INVERSE_LIKE_DIRECT},
1123 {&impAct0, &impAct1}};
1124
1125 static const ImpTab impTabL_INVERSE_LIKE_DIRECT_WITH_MARKS =
1126 /* The case handled in this table is (visually): R EN L
1127 */
1128 {
1129 /* L , R , EN , AN , ON , S , B , Res */
1130 /* 0 : init */ { 0 , s(6,3), 0 , 1 , 0 , 0 , 0 , 0 },
1131 /* 1 : L+AN */ { 0 , s(6,3), 0 , 1 , s(1,2), s(3,0), 0 , 4 },
1132 /* 2 : L+AN+ON */ { s(2,0), s(6,3), s(2,0), 1 , 2 , s(3,0), s(2,0), 3 },
1133 /* 3 : R */ { 0 , s(6,3), s(5,5), s(5,6), s(1,4), s(3,0), 0 , 3 },
1134 /* 4 : R+ON */ { s(3,0), s(4,3), s(5,5), s(5,6), 4 , s(3,0), s(3,0), 3 },
1135 /* 5 : R+EN */ { s(3,0), s(4,3), 5 , s(5,6), s(1,4), s(3,0), s(3,0), 4 },
1136 /* 6 : R+AN */ { s(3,0), s(4,3), s(5,5), 6 , s(1,4), s(3,0), s(3,0), 4 }
1137 };
1138 static const ImpTab impTabR_INVERSE_LIKE_DIRECT_WITH_MARKS =
1139 /* The cases handled in this table are (visually): R EN L
1140 R L AN L
1141 */
1142 {
1143 /* L , R , EN , AN , ON , S , B , Res */
1144 /* 0 : init */ { s(1,3), 0 , 1 , 1 , 0 , 0 , 0 , 0 },
1145 /* 1 : R+EN/AN */ { s(2,3), 0 , 1 , 1 , 2 , s(4,0), 0 , 1 },
1146 /* 2 : R+EN/AN+ON */ { s(2,3), 0 , 1 , 1 , 2 , s(4,0), 0 , 0 },
1147 /* 3 : L */ { 3 , 0 , 3 , s(3,6), s(1,4), s(4,0), 0 , 1 },
1148 /* 4 : L+ON */ { s(5,3), s(4,0), 5 , s(3,6), 4 , s(4,0), s(4,0), 0 },
1149 /* 5 : L+ON+EN */ { s(5,3), s(4,0), 5 , s(3,6), 4 , s(4,0), s(4,0), 1 },
1150 /* 6 : L+AN */ { s(5,3), s(4,0), 6 , 6 , 4 , s(4,0), s(4,0), 3 }
1151 };
1152 static const ImpAct impAct2 = {0,1,7,8,9,10};
1153 static const ImpTabPair impTab_INVERSE_LIKE_DIRECT_WITH_MARKS = {
1154 {&impTabL_INVERSE_LIKE_DIRECT_WITH_MARKS,
1155 &impTabR_INVERSE_LIKE_DIRECT_WITH_MARKS},
1156 {&impAct0, &impAct2}};
1157
1158 static const ImpTabPair impTab_INVERSE_FOR_NUMBERS_SPECIAL = {
1159 {&impTabL_NUMBERS_SPECIAL,
1160 &impTabR_INVERSE_LIKE_DIRECT},
1161 {&impAct0, &impAct1}};
1162
1163 static const ImpTab impTabL_INVERSE_FOR_NUMBERS_SPECIAL_WITH_MARKS =
1164 /* The case handled in this table is (visually): R EN L
1165 */
1166 {
1167 /* L , R , EN , AN , ON , S , B , Res */
1168 /* 0 : init */ { 0 , s(6,2), 1 , 1 , 0 , 0 , 0 , 0 },
1169 /* 1 : L+EN/AN */ { 0 , s(6,2), 1 , 1 , 0 , s(3,0), 0 , 4 },
1170 /* 2 : R */ { 0 , s(6,2), s(5,4), s(5,4), s(1,3), s(3,0), 0 , 3 },
1171 /* 3 : R+ON */ { s(3,0), s(4,2), s(5,4), s(5,4), 3 , s(3,0), s(3,0), 3 },
1172 /* 4 : R+EN/AN */ { s(3,0), s(4,2), 4 , 4 , s(1,3), s(3,0), s(3,0), 4 }
1173 };
1174 static const ImpTabPair impTab_INVERSE_FOR_NUMBERS_SPECIAL_WITH_MARKS = {
1175 {&impTabL_INVERSE_FOR_NUMBERS_SPECIAL_WITH_MARKS,
1176 &impTabR_INVERSE_LIKE_DIRECT_WITH_MARKS},
1177 {&impAct0, &impAct2}};
1178
1179 #undef s
1180
1181 typedef struct {
1182 const ImpTab * pImpTab; /* level table pointer */
1183 const ImpAct * pImpAct; /* action map array */
1184 int32_t startON; /* start of ON sequence */
1185 int32_t startL2EN; /* start of level 2 sequence */
1186 int32_t lastStrongRTL; /* index of last found R or AL */
1187 int32_t state; /* current state */
1188 UBiDiLevel runLevel; /* run level before implicit solving */
1189 } LevState;
1190
1191 /*------------------------------------------------------------------------*/
1192
1193 static void
addPoint(UBiDi * pBiDi,int32_t pos,int32_t flag)1194 addPoint(UBiDi *pBiDi, int32_t pos, int32_t flag)
1195 /* param pos: position where to insert
1196 param flag: one of LRM_BEFORE, LRM_AFTER, RLM_BEFORE, RLM_AFTER
1197 */
1198 {
1199 #define FIRSTALLOC 10
1200 Point point;
1201 InsertPoints * pInsertPoints=&(pBiDi->insertPoints);
1202
1203 if (pInsertPoints->capacity == 0)
1204 {
1205 pInsertPoints->points=uprv_malloc(sizeof(Point)*FIRSTALLOC);
1206 if (pInsertPoints->points == NULL)
1207 {
1208 pInsertPoints->errorCode=U_MEMORY_ALLOCATION_ERROR;
1209 return;
1210 }
1211 pInsertPoints->capacity=FIRSTALLOC;
1212 }
1213 if (pInsertPoints->size >= pInsertPoints->capacity) /* no room for new point */
1214 {
1215 void * savePoints=pInsertPoints->points;
1216 pInsertPoints->points=uprv_realloc(pInsertPoints->points,
1217 pInsertPoints->capacity*2*sizeof(Point));
1218 if (pInsertPoints->points == NULL)
1219 {
1220 pInsertPoints->points=savePoints;
1221 pInsertPoints->errorCode=U_MEMORY_ALLOCATION_ERROR;
1222 return;
1223 }
1224 else pInsertPoints->capacity*=2;
1225 }
1226 point.pos=pos;
1227 point.flag=flag;
1228 pInsertPoints->points[pInsertPoints->size]=point;
1229 pInsertPoints->size++;
1230 #undef FIRSTALLOC
1231 }
1232
1233 /* perform rules (Wn), (Nn), and (In) on a run of the text ------------------ */
1234
1235 /*
1236 * This implementation of the (Wn) rules applies all rules in one pass.
1237 * In order to do so, it needs a look-ahead of typically 1 character
1238 * (except for W5: sequences of ET) and keeps track of changes
1239 * in a rule Wp that affect a later Wq (p<q).
1240 *
1241 * The (Nn) and (In) rules are also performed in that same single loop,
1242 * but effectively one iteration behind for white space.
1243 *
1244 * Since all implicit rules are performed in one step, it is not necessary
1245 * to actually store the intermediate directional properties in dirProps[].
1246 */
1247
1248 static void
processPropertySeq(UBiDi * pBiDi,LevState * pLevState,uint8_t _prop,int32_t start,int32_t limit)1249 processPropertySeq(UBiDi *pBiDi, LevState *pLevState, uint8_t _prop,
1250 int32_t start, int32_t limit) {
1251 uint8_t cell, oldStateSeq, actionSeq;
1252 const ImpTab * pImpTab=pLevState->pImpTab;
1253 const ImpAct * pImpAct=pLevState->pImpAct;
1254 UBiDiLevel * levels=pBiDi->levels;
1255 UBiDiLevel level, addLevel;
1256 InsertPoints * pInsertPoints;
1257 int32_t start0, k;
1258
1259 start0=start; /* save original start position */
1260 oldStateSeq=(uint8_t)pLevState->state;
1261 cell=(*pImpTab)[oldStateSeq][_prop];
1262 pLevState->state=GET_STATE(cell); /* isolate the new state */
1263 actionSeq=(*pImpAct)[GET_ACTION(cell)]; /* isolate the action */
1264 addLevel=(*pImpTab)[pLevState->state][IMPTABLEVELS_RES];
1265
1266 if(actionSeq) {
1267 switch(actionSeq) {
1268 case 1: /* init ON seq */
1269 pLevState->startON=start0;
1270 break;
1271
1272 case 2: /* prepend ON seq to current seq */
1273 start=pLevState->startON;
1274 break;
1275
1276 case 3: /* L or S after possible relevant EN/AN */
1277 /* check if we had EN after R/AL */
1278 if (pLevState->startL2EN >= 0) {
1279 addPoint(pBiDi, pLevState->startL2EN, LRM_BEFORE);
1280 }
1281 pLevState->startL2EN=-1; /* not within previous if since could also be -2 */
1282 /* check if we had any relevant EN/AN after R/AL */
1283 pInsertPoints=&(pBiDi->insertPoints);
1284 if ((pInsertPoints->capacity == 0) ||
1285 (pInsertPoints->size <= pInsertPoints->confirmed))
1286 {
1287 /* nothing, just clean up */
1288 pLevState->lastStrongRTL=-1;
1289 /* check if we have a pending conditional segment */
1290 level=(*pImpTab)[oldStateSeq][IMPTABLEVELS_RES];
1291 if ((level & 1) && (pLevState->startON > 0)) { /* after ON */
1292 start=pLevState->startON; /* reset to basic run level */
1293 }
1294 if (_prop == DirProp_S) /* add LRM before S */
1295 {
1296 addPoint(pBiDi, start0, LRM_BEFORE);
1297 pInsertPoints->confirmed=pInsertPoints->size;
1298 }
1299 break;
1300 }
1301 /* reset previous RTL cont to level for LTR text */
1302 for (k=pLevState->lastStrongRTL+1; k<start0; k++)
1303 {
1304 /* reset odd level, leave runLevel+2 as is */
1305 levels[k]=(levels[k] - 2) & ~1;
1306 }
1307 /* mark insert points as confirmed */
1308 pInsertPoints->confirmed=pInsertPoints->size;
1309 pLevState->lastStrongRTL=-1;
1310 if (_prop == DirProp_S) /* add LRM before S */
1311 {
1312 addPoint(pBiDi, start0, LRM_BEFORE);
1313 pInsertPoints->confirmed=pInsertPoints->size;
1314 }
1315 break;
1316
1317 case 4: /* R/AL after possible relevant EN/AN */
1318 /* just clean up */
1319 pInsertPoints=&(pBiDi->insertPoints);
1320 if (pInsertPoints->capacity > 0)
1321 /* remove all non confirmed insert points */
1322 pInsertPoints->size=pInsertPoints->confirmed;
1323 pLevState->startON=-1;
1324 pLevState->startL2EN=-1;
1325 pLevState->lastStrongRTL=limit - 1;
1326 break;
1327
1328 case 5: /* EN/AN after R/AL + possible cont */
1329 /* check for real AN */
1330 if ((_prop == DirProp_AN) && (NO_CONTEXT_RTL(pBiDi->dirProps[start0]) == AN) &&
1331 (pBiDi->reorderingMode!=UBIDI_REORDER_INVERSE_FOR_NUMBERS_SPECIAL))
1332 {
1333 /* real AN */
1334 if (pLevState->startL2EN == -1) /* if no relevant EN already found */
1335 {
1336 /* just note the righmost digit as a strong RTL */
1337 pLevState->lastStrongRTL=limit - 1;
1338 break;
1339 }
1340 if (pLevState->startL2EN >= 0) /* after EN, no AN */
1341 {
1342 addPoint(pBiDi, pLevState->startL2EN, LRM_BEFORE);
1343 pLevState->startL2EN=-2;
1344 }
1345 /* note AN */
1346 addPoint(pBiDi, start0, LRM_BEFORE);
1347 break;
1348 }
1349 /* if first EN/AN after R/AL */
1350 if (pLevState->startL2EN == -1) {
1351 pLevState->startL2EN=start0;
1352 }
1353 break;
1354
1355 case 6: /* note location of latest R/AL */
1356 pLevState->lastStrongRTL=limit - 1;
1357 pLevState->startON=-1;
1358 break;
1359
1360 case 7: /* L after R+ON/EN/AN */
1361 /* include possible adjacent number on the left */
1362 for (k=start0-1; k>=0 && !(levels[k]&1); k--);
1363 if(k>=0) {
1364 addPoint(pBiDi, k, RLM_BEFORE); /* add RLM before */
1365 pInsertPoints=&(pBiDi->insertPoints);
1366 pInsertPoints->confirmed=pInsertPoints->size; /* confirm it */
1367 }
1368 pLevState->startON=start0;
1369 break;
1370
1371 case 8: /* AN after L */
1372 /* AN numbers between L text on both sides may be trouble. */
1373 /* tentatively bracket with LRMs; will be confirmed if followed by L */
1374 addPoint(pBiDi, start0, LRM_BEFORE); /* add LRM before */
1375 addPoint(pBiDi, start0, LRM_AFTER); /* add LRM after */
1376 break;
1377
1378 case 9: /* R after L+ON/EN/AN */
1379 /* false alert, infirm LRMs around previous AN */
1380 pInsertPoints=&(pBiDi->insertPoints);
1381 pInsertPoints->size=pInsertPoints->confirmed;
1382 if (_prop == DirProp_S) /* add RLM before S */
1383 {
1384 addPoint(pBiDi, start0, RLM_BEFORE);
1385 pInsertPoints->confirmed=pInsertPoints->size;
1386 }
1387 break;
1388
1389 case 10: /* L after L+ON/AN */
1390 level=pLevState->runLevel + addLevel;
1391 for(k=pLevState->startON; k<start0; k++) {
1392 if (levels[k]<level)
1393 levels[k]=level;
1394 }
1395 pInsertPoints=&(pBiDi->insertPoints);
1396 pInsertPoints->confirmed=pInsertPoints->size; /* confirm inserts */
1397 pLevState->startON=start0;
1398 break;
1399
1400 case 11: /* L after L+ON+EN/AN/ON */
1401 level=pLevState->runLevel;
1402 for(k=start0-1; k>=pLevState->startON; k--) {
1403 if(levels[k]==level+3) {
1404 while(levels[k]==level+3) {
1405 levels[k--]-=2;
1406 }
1407 while(levels[k]==level) {
1408 k--;
1409 }
1410 }
1411 if(levels[k]==level+2) {
1412 levels[k]=level;
1413 continue;
1414 }
1415 levels[k]=level+1;
1416 }
1417 break;
1418
1419 case 12: /* R after L+ON+EN/AN/ON */
1420 level=pLevState->runLevel+1;
1421 for(k=start0-1; k>=pLevState->startON; k--) {
1422 if(levels[k]>level) {
1423 levels[k]-=2;
1424 }
1425 }
1426 break;
1427
1428 default: /* we should never get here */
1429 U_ASSERT(FALSE);
1430 break;
1431 }
1432 }
1433 if((addLevel) || (start < start0)) {
1434 level=pLevState->runLevel + addLevel;
1435 for(k=start; k<limit; k++) {
1436 levels[k]=level;
1437 }
1438 }
1439 }
1440
1441 static DirProp
lastL_R_AL(UBiDi * pBiDi)1442 lastL_R_AL(UBiDi *pBiDi) {
1443 /* return last strong char at the end of the prologue */
1444 const UChar *text=pBiDi->prologue;
1445 int32_t length=pBiDi->proLength;
1446 int32_t i;
1447 UChar32 uchar;
1448 DirProp dirProp;
1449 for(i=length; i>0; ) {
1450 /* i is decremented by U16_PREV */
1451 U16_PREV(text, 0, i, uchar);
1452 dirProp=(DirProp)ubidi_getCustomizedClass(pBiDi, uchar);
1453 if(dirProp==L) {
1454 return DirProp_L;
1455 }
1456 if(dirProp==R || dirProp==AL) {
1457 return DirProp_R;
1458 }
1459 if(dirProp==B) {
1460 return DirProp_ON;
1461 }
1462 }
1463 return DirProp_ON;
1464 }
1465
1466 static DirProp
firstL_R_AL_EN_AN(UBiDi * pBiDi)1467 firstL_R_AL_EN_AN(UBiDi *pBiDi) {
1468 /* return first strong char or digit in epilogue */
1469 const UChar *text=pBiDi->epilogue;
1470 int32_t length=pBiDi->epiLength;
1471 int32_t i;
1472 UChar32 uchar;
1473 DirProp dirProp;
1474 for(i=0; i<length; ) {
1475 /* i is incremented by U16_NEXT */
1476 U16_NEXT(text, i, length, uchar);
1477 dirProp=(DirProp)ubidi_getCustomizedClass(pBiDi, uchar);
1478 if(dirProp==L) {
1479 return DirProp_L;
1480 }
1481 if(dirProp==R || dirProp==AL) {
1482 return DirProp_R;
1483 }
1484 if(dirProp==EN) {
1485 return DirProp_EN;
1486 }
1487 if(dirProp==AN) {
1488 return DirProp_AN;
1489 }
1490 }
1491 return DirProp_ON;
1492 }
1493
1494 static void
resolveImplicitLevels(UBiDi * pBiDi,int32_t start,int32_t limit,DirProp sor,DirProp eor)1495 resolveImplicitLevels(UBiDi *pBiDi,
1496 int32_t start, int32_t limit,
1497 DirProp sor, DirProp eor) {
1498 const DirProp *dirProps=pBiDi->dirProps;
1499
1500 LevState levState;
1501 int32_t i, start1, start2;
1502 uint8_t oldStateImp, stateImp, actionImp;
1503 uint8_t gprop, resProp, cell;
1504 UBool inverseRTL;
1505 DirProp nextStrongProp=R;
1506 int32_t nextStrongPos=-1;
1507
1508 levState.startON = -1; /* silence gcc flow analysis */
1509
1510 /* check for RTL inverse BiDi mode */
1511 /* FOOD FOR THOUGHT: in case of RTL inverse BiDi, it would make sense to
1512 * loop on the text characters from end to start.
1513 * This would need a different properties state table (at least different
1514 * actions) and different levels state tables (maybe very similar to the
1515 * LTR corresponding ones.
1516 */
1517 inverseRTL=(UBool)
1518 ((start<pBiDi->lastArabicPos) && (GET_PARALEVEL(pBiDi, start) & 1) &&
1519 (pBiDi->reorderingMode==UBIDI_REORDER_INVERSE_LIKE_DIRECT ||
1520 pBiDi->reorderingMode==UBIDI_REORDER_INVERSE_FOR_NUMBERS_SPECIAL));
1521 /* initialize for levels state table */
1522 levState.startL2EN=-1; /* used for INVERSE_LIKE_DIRECT_WITH_MARKS */
1523 levState.lastStrongRTL=-1; /* used for INVERSE_LIKE_DIRECT_WITH_MARKS */
1524 levState.state=0;
1525 levState.runLevel=pBiDi->levels[start];
1526 levState.pImpTab=(const ImpTab*)((pBiDi->pImpTabPair)->pImpTab)[levState.runLevel&1];
1527 levState.pImpAct=(const ImpAct*)((pBiDi->pImpTabPair)->pImpAct)[levState.runLevel&1];
1528 if(start==0 && pBiDi->proLength>0) {
1529 DirProp lastStrong=lastL_R_AL(pBiDi);
1530 if(lastStrong!=DirProp_ON) {
1531 sor=lastStrong;
1532 }
1533 }
1534 processPropertySeq(pBiDi, &levState, sor, start, start);
1535 /* initialize for property state table */
1536 if(NO_CONTEXT_RTL(dirProps[start])==NSM) {
1537 stateImp = 1 + sor;
1538 } else {
1539 stateImp=0;
1540 }
1541 start1=start;
1542 start2=start;
1543
1544 for(i=start; i<=limit; i++) {
1545 if(i>=limit) {
1546 gprop=eor;
1547 } else {
1548 DirProp prop, prop1;
1549 prop=NO_CONTEXT_RTL(dirProps[i]);
1550 if(inverseRTL) {
1551 if(prop==AL) {
1552 /* AL before EN does not make it AN */
1553 prop=R;
1554 } else if(prop==EN) {
1555 if(nextStrongPos<=i) {
1556 /* look for next strong char (L/R/AL) */
1557 int32_t j;
1558 nextStrongProp=R; /* set default */
1559 nextStrongPos=limit;
1560 for(j=i+1; j<limit; j++) {
1561 prop1=NO_CONTEXT_RTL(dirProps[j]);
1562 if(prop1==L || prop1==R || prop1==AL) {
1563 nextStrongProp=prop1;
1564 nextStrongPos=j;
1565 break;
1566 }
1567 }
1568 }
1569 if(nextStrongProp==AL) {
1570 prop=AN;
1571 }
1572 }
1573 }
1574 gprop=groupProp[prop];
1575 }
1576 oldStateImp=stateImp;
1577 cell=impTabProps[oldStateImp][gprop];
1578 stateImp=GET_STATEPROPS(cell); /* isolate the new state */
1579 actionImp=GET_ACTIONPROPS(cell); /* isolate the action */
1580 if((i==limit) && (actionImp==0)) {
1581 /* there is an unprocessed sequence if its property == eor */
1582 actionImp=1; /* process the last sequence */
1583 }
1584 if(actionImp) {
1585 resProp=impTabProps[oldStateImp][IMPTABPROPS_RES];
1586 switch(actionImp) {
1587 case 1: /* process current seq1, init new seq1 */
1588 processPropertySeq(pBiDi, &levState, resProp, start1, i);
1589 start1=i;
1590 break;
1591 case 2: /* init new seq2 */
1592 start2=i;
1593 break;
1594 case 3: /* process seq1, process seq2, init new seq1 */
1595 processPropertySeq(pBiDi, &levState, resProp, start1, start2);
1596 processPropertySeq(pBiDi, &levState, DirProp_ON, start2, i);
1597 start1=i;
1598 break;
1599 case 4: /* process seq1, set seq1=seq2, init new seq2 */
1600 processPropertySeq(pBiDi, &levState, resProp, start1, start2);
1601 start1=start2;
1602 start2=i;
1603 break;
1604 default: /* we should never get here */
1605 U_ASSERT(FALSE);
1606 break;
1607 }
1608 }
1609 }
1610 /* flush possible pending sequence, e.g. ON */
1611 if(limit==pBiDi->length && pBiDi->epiLength>0) {
1612 DirProp firstStrong=firstL_R_AL_EN_AN(pBiDi);
1613 if(firstStrong!=DirProp_ON) {
1614 eor=firstStrong;
1615 }
1616 }
1617 processPropertySeq(pBiDi, &levState, eor, limit, limit);
1618 }
1619
1620 /* perform (L1) and (X9) ---------------------------------------------------- */
1621
1622 /*
1623 * Reset the embedding levels for some non-graphic characters (L1).
1624 * This function also sets appropriate levels for BN, and
1625 * explicit embedding types that are supposed to have been removed
1626 * from the paragraph in (X9).
1627 */
1628 static void
adjustWSLevels(UBiDi * pBiDi)1629 adjustWSLevels(UBiDi *pBiDi) {
1630 const DirProp *dirProps=pBiDi->dirProps;
1631 UBiDiLevel *levels=pBiDi->levels;
1632 int32_t i;
1633
1634 if(pBiDi->flags&MASK_WS) {
1635 UBool orderParagraphsLTR=pBiDi->orderParagraphsLTR;
1636 Flags flag;
1637
1638 i=pBiDi->trailingWSStart;
1639 while(i>0) {
1640 /* reset a sequence of WS/BN before eop and B/S to the paragraph paraLevel */
1641 while(i>0 && (flag=DIRPROP_FLAG_NC(dirProps[--i]))&MASK_WS) {
1642 if(orderParagraphsLTR&&(flag&DIRPROP_FLAG(B))) {
1643 levels[i]=0;
1644 } else {
1645 levels[i]=GET_PARALEVEL(pBiDi, i);
1646 }
1647 }
1648
1649 /* reset BN to the next character's paraLevel until B/S, which restarts above loop */
1650 /* here, i+1 is guaranteed to be <length */
1651 while(i>0) {
1652 flag=DIRPROP_FLAG_NC(dirProps[--i]);
1653 if(flag&MASK_BN_EXPLICIT) {
1654 levels[i]=levels[i+1];
1655 } else if(orderParagraphsLTR&&(flag&DIRPROP_FLAG(B))) {
1656 levels[i]=0;
1657 break;
1658 } else if(flag&MASK_B_S) {
1659 levels[i]=GET_PARALEVEL(pBiDi, i);
1660 break;
1661 }
1662 }
1663 }
1664 }
1665 }
1666
1667 U_CAPI void U_EXPORT2
ubidi_setContext(UBiDi * pBiDi,const UChar * prologue,int32_t proLength,const UChar * epilogue,int32_t epiLength,UErrorCode * pErrorCode)1668 ubidi_setContext(UBiDi *pBiDi,
1669 const UChar *prologue, int32_t proLength,
1670 const UChar *epilogue, int32_t epiLength,
1671 UErrorCode *pErrorCode) {
1672 /* check the argument values */
1673 RETURN_VOID_IF_NULL_OR_FAILING_ERRCODE(pErrorCode);
1674 if(pBiDi==NULL || proLength<-1 || epiLength<-1 ||
1675 (prologue==NULL && proLength!=0) || (epilogue==NULL && epiLength!=0)) {
1676 *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
1677 return;
1678 }
1679
1680 if(proLength==-1) {
1681 pBiDi->proLength=u_strlen(prologue);
1682 } else {
1683 pBiDi->proLength=proLength;
1684 }
1685 if(epiLength==-1) {
1686 pBiDi->epiLength=u_strlen(epilogue);
1687 } else {
1688 pBiDi->epiLength=epiLength;
1689 }
1690 pBiDi->prologue=prologue;
1691 pBiDi->epilogue=epilogue;
1692 }
1693
1694 static void
setParaSuccess(UBiDi * pBiDi)1695 setParaSuccess(UBiDi *pBiDi) {
1696 pBiDi->proLength=0; /* forget the last context */
1697 pBiDi->epiLength=0;
1698 pBiDi->pParaBiDi=pBiDi; /* mark successful setPara */
1699 }
1700
1701 #define BIDI_MIN(x, y) ((x)<(y) ? (x) : (y))
1702 #define BIDI_ABS(x) ((x)>=0 ? (x) : (-(x)))
1703 static void
setParaRunsOnly(UBiDi * pBiDi,const UChar * text,int32_t length,UBiDiLevel paraLevel,UErrorCode * pErrorCode)1704 setParaRunsOnly(UBiDi *pBiDi, const UChar *text, int32_t length,
1705 UBiDiLevel paraLevel, UErrorCode *pErrorCode) {
1706 void *runsOnlyMemory;
1707 int32_t *visualMap;
1708 UChar *visualText;
1709 int32_t saveLength, saveTrailingWSStart;
1710 const UBiDiLevel *levels;
1711 UBiDiLevel *saveLevels;
1712 UBiDiDirection saveDirection;
1713 UBool saveMayAllocateText;
1714 Run *runs;
1715 int32_t visualLength, i, j, visualStart, logicalStart,
1716 runCount, runLength, addedRuns, insertRemove,
1717 start, limit, step, indexOddBit, logicalPos,
1718 index0, index1;
1719 uint32_t saveOptions;
1720
1721 pBiDi->reorderingMode=UBIDI_REORDER_DEFAULT;
1722 if(length==0) {
1723 ubidi_setPara(pBiDi, text, length, paraLevel, NULL, pErrorCode);
1724 goto cleanup3;
1725 }
1726 /* obtain memory for mapping table and visual text */
1727 runsOnlyMemory=uprv_malloc(length*(sizeof(int32_t)+sizeof(UChar)+sizeof(UBiDiLevel)));
1728 if(runsOnlyMemory==NULL) {
1729 *pErrorCode=U_MEMORY_ALLOCATION_ERROR;
1730 goto cleanup3;
1731 }
1732 visualMap=runsOnlyMemory;
1733 visualText=(UChar *)&visualMap[length];
1734 saveLevels=(UBiDiLevel *)&visualText[length];
1735 saveOptions=pBiDi->reorderingOptions;
1736 if(saveOptions & UBIDI_OPTION_INSERT_MARKS) {
1737 pBiDi->reorderingOptions&=~UBIDI_OPTION_INSERT_MARKS;
1738 pBiDi->reorderingOptions|=UBIDI_OPTION_REMOVE_CONTROLS;
1739 }
1740 paraLevel&=1; /* accept only 0 or 1 */
1741 ubidi_setPara(pBiDi, text, length, paraLevel, NULL, pErrorCode);
1742 if(U_FAILURE(*pErrorCode)) {
1743 goto cleanup3;
1744 }
1745 /* we cannot access directly pBiDi->levels since it is not yet set if
1746 * direction is not MIXED
1747 */
1748 levels=ubidi_getLevels(pBiDi, pErrorCode);
1749 uprv_memcpy(saveLevels, levels, pBiDi->length*sizeof(UBiDiLevel));
1750 saveTrailingWSStart=pBiDi->trailingWSStart;
1751 saveLength=pBiDi->length;
1752 saveDirection=pBiDi->direction;
1753
1754 /* FOOD FOR THOUGHT: instead of writing the visual text, we could use
1755 * the visual map and the dirProps array to drive the second call
1756 * to ubidi_setPara (but must make provision for possible removal of
1757 * BiDi controls. Alternatively, only use the dirProps array via
1758 * customized classifier callback.
1759 */
1760 visualLength=ubidi_writeReordered(pBiDi, visualText, length,
1761 UBIDI_DO_MIRRORING, pErrorCode);
1762 ubidi_getVisualMap(pBiDi, visualMap, pErrorCode);
1763 if(U_FAILURE(*pErrorCode)) {
1764 goto cleanup2;
1765 }
1766 pBiDi->reorderingOptions=saveOptions;
1767
1768 pBiDi->reorderingMode=UBIDI_REORDER_INVERSE_LIKE_DIRECT;
1769 paraLevel^=1;
1770 /* Because what we did with reorderingOptions, visualText may be shorter
1771 * than the original text. But we don't want the levels memory to be
1772 * reallocated shorter than the original length, since we need to restore
1773 * the levels as after the first call to ubidi_setpara() before returning.
1774 * We will force mayAllocateText to FALSE before the second call to
1775 * ubidi_setpara(), and will restore it afterwards.
1776 */
1777 saveMayAllocateText=pBiDi->mayAllocateText;
1778 pBiDi->mayAllocateText=FALSE;
1779 ubidi_setPara(pBiDi, visualText, visualLength, paraLevel, NULL, pErrorCode);
1780 pBiDi->mayAllocateText=saveMayAllocateText;
1781 ubidi_getRuns(pBiDi, pErrorCode);
1782 if(U_FAILURE(*pErrorCode)) {
1783 goto cleanup1;
1784 }
1785 /* check if some runs must be split, count how many splits */
1786 addedRuns=0;
1787 runCount=pBiDi->runCount;
1788 runs=pBiDi->runs;
1789 visualStart=0;
1790 for(i=0; i<runCount; i++, visualStart+=runLength) {
1791 runLength=runs[i].visualLimit-visualStart;
1792 if(runLength<2) {
1793 continue;
1794 }
1795 logicalStart=GET_INDEX(runs[i].logicalStart);
1796 for(j=logicalStart+1; j<logicalStart+runLength; j++) {
1797 index0=visualMap[j];
1798 index1=visualMap[j-1];
1799 if((BIDI_ABS(index0-index1)!=1) || (saveLevels[index0]!=saveLevels[index1])) {
1800 addedRuns++;
1801 }
1802 }
1803 }
1804 if(addedRuns) {
1805 if(getRunsMemory(pBiDi, runCount+addedRuns)) {
1806 if(runCount==1) {
1807 /* because we switch from UBiDi.simpleRuns to UBiDi.runs */
1808 pBiDi->runsMemory[0]=runs[0];
1809 }
1810 runs=pBiDi->runs=pBiDi->runsMemory;
1811 pBiDi->runCount+=addedRuns;
1812 } else {
1813 goto cleanup1;
1814 }
1815 }
1816 /* split runs which are not consecutive in source text */
1817 for(i=runCount-1; i>=0; i--) {
1818 runLength= i==0 ? runs[0].visualLimit :
1819 runs[i].visualLimit-runs[i-1].visualLimit;
1820 logicalStart=runs[i].logicalStart;
1821 indexOddBit=GET_ODD_BIT(logicalStart);
1822 logicalStart=GET_INDEX(logicalStart);
1823 if(runLength<2) {
1824 if(addedRuns) {
1825 runs[i+addedRuns]=runs[i];
1826 }
1827 logicalPos=visualMap[logicalStart];
1828 runs[i+addedRuns].logicalStart=MAKE_INDEX_ODD_PAIR(logicalPos,
1829 saveLevels[logicalPos]^indexOddBit);
1830 continue;
1831 }
1832 if(indexOddBit) {
1833 start=logicalStart;
1834 limit=logicalStart+runLength-1;
1835 step=1;
1836 } else {
1837 start=logicalStart+runLength-1;
1838 limit=logicalStart;
1839 step=-1;
1840 }
1841 for(j=start; j!=limit; j+=step) {
1842 index0=visualMap[j];
1843 index1=visualMap[j+step];
1844 if((BIDI_ABS(index0-index1)!=1) || (saveLevels[index0]!=saveLevels[index1])) {
1845 logicalPos=BIDI_MIN(visualMap[start], index0);
1846 runs[i+addedRuns].logicalStart=MAKE_INDEX_ODD_PAIR(logicalPos,
1847 saveLevels[logicalPos]^indexOddBit);
1848 runs[i+addedRuns].visualLimit=runs[i].visualLimit;
1849 runs[i].visualLimit-=BIDI_ABS(j-start)+1;
1850 insertRemove=runs[i].insertRemove&(LRM_AFTER|RLM_AFTER);
1851 runs[i+addedRuns].insertRemove=insertRemove;
1852 runs[i].insertRemove&=~insertRemove;
1853 start=j+step;
1854 addedRuns--;
1855 }
1856 }
1857 if(addedRuns) {
1858 runs[i+addedRuns]=runs[i];
1859 }
1860 logicalPos=BIDI_MIN(visualMap[start], visualMap[limit]);
1861 runs[i+addedRuns].logicalStart=MAKE_INDEX_ODD_PAIR(logicalPos,
1862 saveLevels[logicalPos]^indexOddBit);
1863 }
1864
1865 cleanup1:
1866 /* restore initial paraLevel */
1867 pBiDi->paraLevel^=1;
1868 cleanup2:
1869 /* restore real text */
1870 pBiDi->text=text;
1871 pBiDi->length=saveLength;
1872 pBiDi->originalLength=length;
1873 pBiDi->direction=saveDirection;
1874 /* the saved levels should never excess levelsSize, but we check anyway */
1875 if(saveLength>pBiDi->levelsSize) {
1876 saveLength=pBiDi->levelsSize;
1877 }
1878 uprv_memcpy(pBiDi->levels, saveLevels, saveLength*sizeof(UBiDiLevel));
1879 pBiDi->trailingWSStart=saveTrailingWSStart;
1880 /* free memory for mapping table and visual text */
1881 uprv_free(runsOnlyMemory);
1882 if(pBiDi->runCount>1) {
1883 pBiDi->direction=UBIDI_MIXED;
1884 }
1885 cleanup3:
1886 pBiDi->reorderingMode=UBIDI_REORDER_RUNS_ONLY;
1887 }
1888
1889 /* ubidi_setPara ------------------------------------------------------------ */
1890
1891 U_CAPI void U_EXPORT2
ubidi_setPara(UBiDi * pBiDi,const UChar * text,int32_t length,UBiDiLevel paraLevel,UBiDiLevel * embeddingLevels,UErrorCode * pErrorCode)1892 ubidi_setPara(UBiDi *pBiDi, const UChar *text, int32_t length,
1893 UBiDiLevel paraLevel, UBiDiLevel *embeddingLevels,
1894 UErrorCode *pErrorCode) {
1895 UBiDiDirection direction;
1896
1897 /* check the argument values */
1898 RETURN_VOID_IF_NULL_OR_FAILING_ERRCODE(pErrorCode);
1899 if(pBiDi==NULL || text==NULL || length<-1 ||
1900 (paraLevel>UBIDI_MAX_EXPLICIT_LEVEL && paraLevel<UBIDI_DEFAULT_LTR)) {
1901 *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
1902 return;
1903 }
1904
1905 if(length==-1) {
1906 length=u_strlen(text);
1907 }
1908
1909 /* special treatment for RUNS_ONLY mode */
1910 if(pBiDi->reorderingMode==UBIDI_REORDER_RUNS_ONLY) {
1911 setParaRunsOnly(pBiDi, text, length, paraLevel, pErrorCode);
1912 return;
1913 }
1914
1915 /* initialize the UBiDi structure */
1916 pBiDi->pParaBiDi=NULL; /* mark unfinished setPara */
1917 pBiDi->text=text;
1918 pBiDi->length=pBiDi->originalLength=pBiDi->resultLength=length;
1919 pBiDi->paraLevel=paraLevel;
1920 pBiDi->direction=UBIDI_LTR;
1921 pBiDi->paraCount=1;
1922
1923 pBiDi->dirProps=NULL;
1924 pBiDi->levels=NULL;
1925 pBiDi->runs=NULL;
1926 pBiDi->insertPoints.size=0; /* clean up from last call */
1927 pBiDi->insertPoints.confirmed=0; /* clean up from last call */
1928
1929 /*
1930 * Save the original paraLevel if contextual; otherwise, set to 0.
1931 */
1932 if(IS_DEFAULT_LEVEL(paraLevel)) {
1933 pBiDi->defaultParaLevel=paraLevel;
1934 } else {
1935 pBiDi->defaultParaLevel=0;
1936 }
1937
1938 if(length==0) {
1939 /*
1940 * For an empty paragraph, create a UBiDi object with the paraLevel and
1941 * the flags and the direction set but without allocating zero-length arrays.
1942 * There is nothing more to do.
1943 */
1944 if(IS_DEFAULT_LEVEL(paraLevel)) {
1945 pBiDi->paraLevel&=1;
1946 pBiDi->defaultParaLevel=0;
1947 }
1948 if(paraLevel&1) {
1949 pBiDi->flags=DIRPROP_FLAG(R);
1950 pBiDi->direction=UBIDI_RTL;
1951 } else {
1952 pBiDi->flags=DIRPROP_FLAG(L);
1953 pBiDi->direction=UBIDI_LTR;
1954 }
1955
1956 pBiDi->runCount=0;
1957 pBiDi->paraCount=0;
1958 setParaSuccess(pBiDi); /* mark successful setPara */
1959 return;
1960 }
1961
1962 pBiDi->runCount=-1;
1963
1964 /*
1965 * Get the directional properties,
1966 * the flags bit-set, and
1967 * determine the paragraph level if necessary.
1968 */
1969 if(getDirPropsMemory(pBiDi, length)) {
1970 pBiDi->dirProps=pBiDi->dirPropsMemory;
1971 getDirProps(pBiDi);
1972 } else {
1973 *pErrorCode=U_MEMORY_ALLOCATION_ERROR;
1974 return;
1975 }
1976 /* the processed length may have changed if UBIDI_OPTION_STREAMING */
1977 length= pBiDi->length;
1978 pBiDi->trailingWSStart=length; /* the levels[] will reflect the WS run */
1979 /* allocate paras memory */
1980 if(pBiDi->paraCount>1) {
1981 if(getInitialParasMemory(pBiDi, pBiDi->paraCount)) {
1982 pBiDi->paras=pBiDi->parasMemory;
1983 pBiDi->paras[pBiDi->paraCount-1]=length;
1984 } else {
1985 *pErrorCode=U_MEMORY_ALLOCATION_ERROR;
1986 return;
1987 }
1988 } else {
1989 /* initialize paras for single paragraph */
1990 pBiDi->paras=pBiDi->simpleParas;
1991 pBiDi->simpleParas[0]=length;
1992 }
1993
1994 /* are explicit levels specified? */
1995 if(embeddingLevels==NULL) {
1996 /* no: determine explicit levels according to the (Xn) rules */\
1997 if(getLevelsMemory(pBiDi, length)) {
1998 pBiDi->levels=pBiDi->levelsMemory;
1999 direction=resolveExplicitLevels(pBiDi);
2000 } else {
2001 *pErrorCode=U_MEMORY_ALLOCATION_ERROR;
2002 return;
2003 }
2004 } else {
2005 /* set BN for all explicit codes, check that all levels are 0 or paraLevel..UBIDI_MAX_EXPLICIT_LEVEL */
2006 pBiDi->levels=embeddingLevels;
2007 direction=checkExplicitLevels(pBiDi, pErrorCode);
2008 if(U_FAILURE(*pErrorCode)) {
2009 return;
2010 }
2011 }
2012
2013 /*
2014 * The steps after (X9) in the UBiDi algorithm are performed only if
2015 * the paragraph text has mixed directionality!
2016 */
2017 pBiDi->direction=direction;
2018 switch(direction) {
2019 case UBIDI_LTR:
2020 /* make sure paraLevel is even */
2021 pBiDi->paraLevel=(UBiDiLevel)((pBiDi->paraLevel+1)&~1);
2022
2023 /* all levels are implicitly at paraLevel (important for ubidi_getLevels()) */
2024 pBiDi->trailingWSStart=0;
2025 break;
2026 case UBIDI_RTL:
2027 /* make sure paraLevel is odd */
2028 pBiDi->paraLevel|=1;
2029
2030 /* all levels are implicitly at paraLevel (important for ubidi_getLevels()) */
2031 pBiDi->trailingWSStart=0;
2032 break;
2033 default:
2034 /*
2035 * Choose the right implicit state table
2036 */
2037 switch(pBiDi->reorderingMode) {
2038 case UBIDI_REORDER_DEFAULT:
2039 pBiDi->pImpTabPair=&impTab_DEFAULT;
2040 break;
2041 case UBIDI_REORDER_NUMBERS_SPECIAL:
2042 pBiDi->pImpTabPair=&impTab_NUMBERS_SPECIAL;
2043 break;
2044 case UBIDI_REORDER_GROUP_NUMBERS_WITH_R:
2045 pBiDi->pImpTabPair=&impTab_GROUP_NUMBERS_WITH_R;
2046 break;
2047 case UBIDI_REORDER_INVERSE_NUMBERS_AS_L:
2048 pBiDi->pImpTabPair=&impTab_INVERSE_NUMBERS_AS_L;
2049 break;
2050 case UBIDI_REORDER_INVERSE_LIKE_DIRECT:
2051 if (pBiDi->reorderingOptions & UBIDI_OPTION_INSERT_MARKS) {
2052 pBiDi->pImpTabPair=&impTab_INVERSE_LIKE_DIRECT_WITH_MARKS;
2053 } else {
2054 pBiDi->pImpTabPair=&impTab_INVERSE_LIKE_DIRECT;
2055 }
2056 break;
2057 case UBIDI_REORDER_INVERSE_FOR_NUMBERS_SPECIAL:
2058 if (pBiDi->reorderingOptions & UBIDI_OPTION_INSERT_MARKS) {
2059 pBiDi->pImpTabPair=&impTab_INVERSE_FOR_NUMBERS_SPECIAL_WITH_MARKS;
2060 } else {
2061 pBiDi->pImpTabPair=&impTab_INVERSE_FOR_NUMBERS_SPECIAL;
2062 }
2063 break;
2064 default:
2065 /* we should never get here */
2066 U_ASSERT(FALSE);
2067 break;
2068 }
2069 /*
2070 * If there are no external levels specified and there
2071 * are no significant explicit level codes in the text,
2072 * then we can treat the entire paragraph as one run.
2073 * Otherwise, we need to perform the following rules on runs of
2074 * the text with the same embedding levels. (X10)
2075 * "Significant" explicit level codes are ones that actually
2076 * affect non-BN characters.
2077 * Examples for "insignificant" ones are empty embeddings
2078 * LRE-PDF, LRE-RLE-PDF-PDF, etc.
2079 */
2080 if(embeddingLevels==NULL && pBiDi->paraCount<=1 &&
2081 !(pBiDi->flags&DIRPROP_FLAG_MULTI_RUNS)) {
2082 resolveImplicitLevels(pBiDi, 0, length,
2083 GET_LR_FROM_LEVEL(GET_PARALEVEL(pBiDi, 0)),
2084 GET_LR_FROM_LEVEL(GET_PARALEVEL(pBiDi, length-1)));
2085 } else {
2086 /* sor, eor: start and end types of same-level-run */
2087 UBiDiLevel *levels=pBiDi->levels;
2088 int32_t start, limit=0;
2089 UBiDiLevel level, nextLevel;
2090 DirProp sor, eor;
2091
2092 /* determine the first sor and set eor to it because of the loop body (sor=eor there) */
2093 level=GET_PARALEVEL(pBiDi, 0);
2094 nextLevel=levels[0];
2095 if(level<nextLevel) {
2096 eor=GET_LR_FROM_LEVEL(nextLevel);
2097 } else {
2098 eor=GET_LR_FROM_LEVEL(level);
2099 }
2100
2101 do {
2102 /* determine start and limit of the run (end points just behind the run) */
2103
2104 /* the values for this run's start are the same as for the previous run's end */
2105 start=limit;
2106 level=nextLevel;
2107 if((start>0) && (NO_CONTEXT_RTL(pBiDi->dirProps[start-1])==B)) {
2108 /* except if this is a new paragraph, then set sor = para level */
2109 sor=GET_LR_FROM_LEVEL(GET_PARALEVEL(pBiDi, start));
2110 } else {
2111 sor=eor;
2112 }
2113
2114 /* search for the limit of this run */
2115 while(++limit<length && levels[limit]==level) {}
2116
2117 /* get the correct level of the next run */
2118 if(limit<length) {
2119 nextLevel=levels[limit];
2120 } else {
2121 nextLevel=GET_PARALEVEL(pBiDi, length-1);
2122 }
2123
2124 /* determine eor from max(level, nextLevel); sor is last run's eor */
2125 if((level&~UBIDI_LEVEL_OVERRIDE)<(nextLevel&~UBIDI_LEVEL_OVERRIDE)) {
2126 eor=GET_LR_FROM_LEVEL(nextLevel);
2127 } else {
2128 eor=GET_LR_FROM_LEVEL(level);
2129 }
2130
2131 /* if the run consists of overridden directional types, then there
2132 are no implicit types to be resolved */
2133 if(!(level&UBIDI_LEVEL_OVERRIDE)) {
2134 resolveImplicitLevels(pBiDi, start, limit, sor, eor);
2135 } else {
2136 /* remove the UBIDI_LEVEL_OVERRIDE flags */
2137 do {
2138 levels[start++]&=~UBIDI_LEVEL_OVERRIDE;
2139 } while(start<limit);
2140 }
2141 } while(limit<length);
2142 }
2143 /* check if we got any memory shortage while adding insert points */
2144 if (U_FAILURE(pBiDi->insertPoints.errorCode))
2145 {
2146 *pErrorCode=pBiDi->insertPoints.errorCode;
2147 return;
2148 }
2149 /* reset the embedding levels for some non-graphic characters (L1), (X9) */
2150 adjustWSLevels(pBiDi);
2151 break;
2152 }
2153 /* add RLM for inverse Bidi with contextual orientation resolving
2154 * to RTL which would not round-trip otherwise
2155 */
2156 if((pBiDi->defaultParaLevel>0) &&
2157 (pBiDi->reorderingOptions & UBIDI_OPTION_INSERT_MARKS) &&
2158 ((pBiDi->reorderingMode==UBIDI_REORDER_INVERSE_LIKE_DIRECT) ||
2159 (pBiDi->reorderingMode==UBIDI_REORDER_INVERSE_FOR_NUMBERS_SPECIAL))) {
2160 int32_t i, j, start, last;
2161 DirProp dirProp;
2162 for(i=0; i<pBiDi->paraCount; i++) {
2163 last=pBiDi->paras[i]-1;
2164 if((pBiDi->dirProps[last] & CONTEXT_RTL)==0) {
2165 continue; /* LTR paragraph */
2166 }
2167 start= i==0 ? 0 : pBiDi->paras[i - 1];
2168 for(j=last; j>=start; j--) {
2169 dirProp=NO_CONTEXT_RTL(pBiDi->dirProps[j]);
2170 if(dirProp==L) {
2171 if(j<last) {
2172 while(NO_CONTEXT_RTL(pBiDi->dirProps[last])==B) {
2173 last--;
2174 }
2175 }
2176 addPoint(pBiDi, last, RLM_BEFORE);
2177 break;
2178 }
2179 if(DIRPROP_FLAG(dirProp) & MASK_R_AL) {
2180 break;
2181 }
2182 }
2183 }
2184 }
2185
2186 if(pBiDi->reorderingOptions & UBIDI_OPTION_REMOVE_CONTROLS) {
2187 pBiDi->resultLength -= pBiDi->controlCount;
2188 } else {
2189 pBiDi->resultLength += pBiDi->insertPoints.size;
2190 }
2191 setParaSuccess(pBiDi); /* mark successful setPara */
2192 }
2193
2194 U_CAPI void U_EXPORT2
ubidi_orderParagraphsLTR(UBiDi * pBiDi,UBool orderParagraphsLTR)2195 ubidi_orderParagraphsLTR(UBiDi *pBiDi, UBool orderParagraphsLTR) {
2196 if(pBiDi!=NULL) {
2197 pBiDi->orderParagraphsLTR=orderParagraphsLTR;
2198 }
2199 }
2200
2201 U_CAPI UBool U_EXPORT2
ubidi_isOrderParagraphsLTR(UBiDi * pBiDi)2202 ubidi_isOrderParagraphsLTR(UBiDi *pBiDi) {
2203 if(pBiDi!=NULL) {
2204 return pBiDi->orderParagraphsLTR;
2205 } else {
2206 return FALSE;
2207 }
2208 }
2209
2210 U_CAPI UBiDiDirection U_EXPORT2
ubidi_getDirection(const UBiDi * pBiDi)2211 ubidi_getDirection(const UBiDi *pBiDi) {
2212 if(IS_VALID_PARA_OR_LINE(pBiDi)) {
2213 return pBiDi->direction;
2214 } else {
2215 return UBIDI_LTR;
2216 }
2217 }
2218
2219 U_CAPI const UChar * U_EXPORT2
ubidi_getText(const UBiDi * pBiDi)2220 ubidi_getText(const UBiDi *pBiDi) {
2221 if(IS_VALID_PARA_OR_LINE(pBiDi)) {
2222 return pBiDi->text;
2223 } else {
2224 return NULL;
2225 }
2226 }
2227
2228 U_CAPI int32_t U_EXPORT2
ubidi_getLength(const UBiDi * pBiDi)2229 ubidi_getLength(const UBiDi *pBiDi) {
2230 if(IS_VALID_PARA_OR_LINE(pBiDi)) {
2231 return pBiDi->originalLength;
2232 } else {
2233 return 0;
2234 }
2235 }
2236
2237 U_CAPI int32_t U_EXPORT2
ubidi_getProcessedLength(const UBiDi * pBiDi)2238 ubidi_getProcessedLength(const UBiDi *pBiDi) {
2239 if(IS_VALID_PARA_OR_LINE(pBiDi)) {
2240 return pBiDi->length;
2241 } else {
2242 return 0;
2243 }
2244 }
2245
2246 U_CAPI int32_t U_EXPORT2
ubidi_getResultLength(const UBiDi * pBiDi)2247 ubidi_getResultLength(const UBiDi *pBiDi) {
2248 if(IS_VALID_PARA_OR_LINE(pBiDi)) {
2249 return pBiDi->resultLength;
2250 } else {
2251 return 0;
2252 }
2253 }
2254
2255 /* paragraphs API functions ------------------------------------------------- */
2256
2257 U_CAPI UBiDiLevel U_EXPORT2
ubidi_getParaLevel(const UBiDi * pBiDi)2258 ubidi_getParaLevel(const UBiDi *pBiDi) {
2259 if(IS_VALID_PARA_OR_LINE(pBiDi)) {
2260 return pBiDi->paraLevel;
2261 } else {
2262 return 0;
2263 }
2264 }
2265
2266 U_CAPI int32_t U_EXPORT2
ubidi_countParagraphs(UBiDi * pBiDi)2267 ubidi_countParagraphs(UBiDi *pBiDi) {
2268 if(!IS_VALID_PARA_OR_LINE(pBiDi)) {
2269 return 0;
2270 } else {
2271 return pBiDi->paraCount;
2272 }
2273 }
2274
2275 U_CAPI void U_EXPORT2
ubidi_getParagraphByIndex(const UBiDi * pBiDi,int32_t paraIndex,int32_t * pParaStart,int32_t * pParaLimit,UBiDiLevel * pParaLevel,UErrorCode * pErrorCode)2276 ubidi_getParagraphByIndex(const UBiDi *pBiDi, int32_t paraIndex,
2277 int32_t *pParaStart, int32_t *pParaLimit,
2278 UBiDiLevel *pParaLevel, UErrorCode *pErrorCode) {
2279 int32_t paraStart;
2280
2281 /* check the argument values */
2282 RETURN_VOID_IF_NULL_OR_FAILING_ERRCODE(pErrorCode);
2283 RETURN_VOID_IF_NOT_VALID_PARA_OR_LINE(pBiDi, *pErrorCode);
2284 RETURN_VOID_IF_BAD_RANGE(paraIndex, 0, pBiDi->paraCount, *pErrorCode);
2285
2286 pBiDi=pBiDi->pParaBiDi; /* get Para object if Line object */
2287 if(paraIndex) {
2288 paraStart=pBiDi->paras[paraIndex-1];
2289 } else {
2290 paraStart=0;
2291 }
2292 if(pParaStart!=NULL) {
2293 *pParaStart=paraStart;
2294 }
2295 if(pParaLimit!=NULL) {
2296 *pParaLimit=pBiDi->paras[paraIndex];
2297 }
2298 if(pParaLevel!=NULL) {
2299 *pParaLevel=GET_PARALEVEL(pBiDi, paraStart);
2300 }
2301 }
2302
2303 U_CAPI int32_t U_EXPORT2
ubidi_getParagraph(const UBiDi * pBiDi,int32_t charIndex,int32_t * pParaStart,int32_t * pParaLimit,UBiDiLevel * pParaLevel,UErrorCode * pErrorCode)2304 ubidi_getParagraph(const UBiDi *pBiDi, int32_t charIndex,
2305 int32_t *pParaStart, int32_t *pParaLimit,
2306 UBiDiLevel *pParaLevel, UErrorCode *pErrorCode) {
2307 uint32_t paraIndex;
2308
2309 /* check the argument values */
2310 /* pErrorCode will be checked by the call to ubidi_getParagraphByIndex */
2311 RETURN_IF_NULL_OR_FAILING_ERRCODE(pErrorCode, -1);
2312 RETURN_IF_NOT_VALID_PARA_OR_LINE(pBiDi, *pErrorCode, -1);
2313 pBiDi=pBiDi->pParaBiDi; /* get Para object if Line object */
2314 RETURN_IF_BAD_RANGE(charIndex, 0, pBiDi->length, *pErrorCode, -1);
2315
2316 for(paraIndex=0; charIndex>=pBiDi->paras[paraIndex]; paraIndex++);
2317 ubidi_getParagraphByIndex(pBiDi, paraIndex, pParaStart, pParaLimit, pParaLevel, pErrorCode);
2318 return paraIndex;
2319 }
2320
2321 U_CAPI void U_EXPORT2
ubidi_setClassCallback(UBiDi * pBiDi,UBiDiClassCallback * newFn,const void * newContext,UBiDiClassCallback ** oldFn,const void ** oldContext,UErrorCode * pErrorCode)2322 ubidi_setClassCallback(UBiDi *pBiDi, UBiDiClassCallback *newFn,
2323 const void *newContext, UBiDiClassCallback **oldFn,
2324 const void **oldContext, UErrorCode *pErrorCode)
2325 {
2326 RETURN_VOID_IF_NULL_OR_FAILING_ERRCODE(pErrorCode);
2327 if(pBiDi==NULL) {
2328 *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
2329 return;
2330 }
2331 if( oldFn )
2332 {
2333 *oldFn = pBiDi->fnClassCallback;
2334 }
2335 if( oldContext )
2336 {
2337 *oldContext = pBiDi->coClassCallback;
2338 }
2339 pBiDi->fnClassCallback = newFn;
2340 pBiDi->coClassCallback = newContext;
2341 }
2342
2343 U_CAPI void U_EXPORT2
ubidi_getClassCallback(UBiDi * pBiDi,UBiDiClassCallback ** fn,const void ** context)2344 ubidi_getClassCallback(UBiDi *pBiDi, UBiDiClassCallback **fn, const void **context)
2345 {
2346 if(pBiDi==NULL) {
2347 return;
2348 }
2349 if( fn )
2350 {
2351 *fn = pBiDi->fnClassCallback;
2352 }
2353 if( context )
2354 {
2355 *context = pBiDi->coClassCallback;
2356 }
2357 }
2358
2359 U_CAPI UCharDirection U_EXPORT2
ubidi_getCustomizedClass(UBiDi * pBiDi,UChar32 c)2360 ubidi_getCustomizedClass(UBiDi *pBiDi, UChar32 c)
2361 {
2362 UCharDirection dir;
2363
2364 if( pBiDi->fnClassCallback == NULL ||
2365 (dir = (*pBiDi->fnClassCallback)(pBiDi->coClassCallback, c)) == U_BIDI_CLASS_DEFAULT )
2366 {
2367 return ubidi_getClass(pBiDi->bdp, c);
2368 } else {
2369 return dir;
2370 }
2371 }
2372
2373