1// © 2016 and later: Unicode, Inc. and others.
2// License & terms of use: http://www.unicode.org/copyright.html
3/*
4*******************************************************************************
5*
6* Copyright (C) 2005-2016, International Business Machines
7* Corporation and others. All Rights Reserved.
8*
9*******************************************************************************
10* file name: utext.cpp
11* encoding: UTF-8
12* tab size: 8 (not used)
13* indentation:4
14*
15* created on: 2005apr12
16* created by: Markus W. Scherer
17*/
18
19#include "unicode/utypes.h"
20#include "unicode/ustring.h"
21#include "unicode/unistr.h"
22#include "unicode/chariter.h"
23#include "unicode/utext.h"
24#include "unicode/utf.h"
25#include "unicode/utf8.h"
26#include "unicode/utf16.h"
27#include "ustr_imp.h"
28#include "cmemory.h"
29#include "cstring.h"
30#include "uassert.h"
31#include "putilimp.h"
32
33U_NAMESPACE_USE
34
35#define I32_FLAG(bitIndex) ((int32_t)1<<(bitIndex))
36
37
38static UBool
39utext_access(UText *ut, int64_t index, UBool forward) {
40 return ut->pFuncs->access(ut, index, forward);
41}
42
43
44
45U_CAPI UBool U_EXPORT2
46utext_moveIndex32(UText *ut, int32_t delta) {
47 UChar32 c;
48 if (delta > 0) {
49 do {
50 if(ut->chunkOffset>=ut->chunkLength && !utext_access(ut, ut->chunkNativeLimit, TRUE)) {
51 return FALSE;
52 }
53 c = ut->chunkContents[ut->chunkOffset];
54 if (U16_IS_SURROGATE(c)) {
55 c = utext_next32(ut);
56 if (c == U_SENTINEL) {
57 return FALSE;
58 }
59 } else {
60 ut->chunkOffset++;
61 }
62 } while(--delta>0);
63
64 } else if (delta<0) {
65 do {
66 if(ut->chunkOffset<=0 && !utext_access(ut, ut->chunkNativeStart, FALSE)) {
67 return FALSE;
68 }
69 c = ut->chunkContents[ut->chunkOffset-1];
70 if (U16_IS_SURROGATE(c)) {
71 c = utext_previous32(ut);
72 if (c == U_SENTINEL) {
73 return FALSE;
74 }
75 } else {
76 ut->chunkOffset--;
77 }
78 } while(++delta<0);
79 }
80
81 return TRUE;
82}
83
84
85U_CAPI int64_t U_EXPORT2
86utext_nativeLength(UText *ut) {
87 return ut->pFuncs->nativeLength(ut);
88}
89
90
91U_CAPI UBool U_EXPORT2
92utext_isLengthExpensive(const UText *ut) {
93 UBool r = (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE)) != 0;
94 return r;
95}
96
97
98U_CAPI int64_t U_EXPORT2
99utext_getNativeIndex(const UText *ut) {
100 if(ut->chunkOffset <= ut->nativeIndexingLimit) {
101 return ut->chunkNativeStart+ut->chunkOffset;
102 } else {
103 return ut->pFuncs->mapOffsetToNative(ut);
104 }
105}
106
107
108U_CAPI void U_EXPORT2
109utext_setNativeIndex(UText *ut, int64_t index) {
110 if(index<ut->chunkNativeStart || index>=ut->chunkNativeLimit) {
111 // The desired position is outside of the current chunk.
112 // Access the new position. Assume a forward iteration from here,
113 // which will also be optimimum for a single random access.
114 // Reverse iterations may suffer slightly.
115 ut->pFuncs->access(ut, index, TRUE);
116 } else if((int32_t)(index - ut->chunkNativeStart) <= ut->nativeIndexingLimit) {
117 // utf-16 indexing.
118 ut->chunkOffset=(int32_t)(index-ut->chunkNativeStart);
119 } else {
120 ut->chunkOffset=ut->pFuncs->mapNativeIndexToUTF16(ut, index);
121 }
122 // The convention is that the index must always be on a code point boundary.
123 // Adjust the index position if it is in the middle of a surrogate pair.
124 if (ut->chunkOffset<ut->chunkLength) {
125 UChar c= ut->chunkContents[ut->chunkOffset];
126 if (U16_IS_TRAIL(c)) {
127 if (ut->chunkOffset==0) {
128 ut->pFuncs->access(ut, ut->chunkNativeStart, FALSE);
129 }
130 if (ut->chunkOffset>0) {
131 UChar lead = ut->chunkContents[ut->chunkOffset-1];
132 if (U16_IS_LEAD(lead)) {
133 ut->chunkOffset--;
134 }
135 }
136 }
137 }
138}
139
140
141
142U_CAPI int64_t U_EXPORT2
143utext_getPreviousNativeIndex(UText *ut) {
144 //
145 // Fast-path the common case.
146 // Common means current position is not at the beginning of a chunk
147 // and the preceding character is not supplementary.
148 //
149 int32_t i = ut->chunkOffset - 1;
150 int64_t result;
151 if (i >= 0) {
152 UChar c = ut->chunkContents[i];
153 if (U16_IS_TRAIL(c) == FALSE) {
154 if (i <= ut->nativeIndexingLimit) {
155 result = ut->chunkNativeStart + i;
156 } else {
157 ut->chunkOffset = i;
158 result = ut->pFuncs->mapOffsetToNative(ut);
159 ut->chunkOffset++;
160 }
161 return result;
162 }
163 }
164
165 // If at the start of text, simply return 0.
166 if (ut->chunkOffset==0 && ut->chunkNativeStart==0) {
167 return 0;
168 }
169
170 // Harder, less common cases. We are at a chunk boundary, or on a surrogate.
171 // Keep it simple, use other functions to handle the edges.
172 //
173 utext_previous32(ut);
174 result = UTEXT_GETNATIVEINDEX(ut);
175 utext_next32(ut);
176 return result;
177}
178
179
180//
181// utext_current32. Get the UChar32 at the current position.
182// UText iteration position is always on a code point boundary,
183// never on the trail half of a surrogate pair.
184//
185U_CAPI UChar32 U_EXPORT2
186utext_current32(UText *ut) {
187 UChar32 c;
188 if (ut->chunkOffset==ut->chunkLength) {
189 // Current position is just off the end of the chunk.
190 if (ut->pFuncs->access(ut, ut->chunkNativeLimit, TRUE) == FALSE) {
191 // Off the end of the text.
192 return U_SENTINEL;
193 }
194 }
195
196 c = ut->chunkContents[ut->chunkOffset];
197 if (U16_IS_LEAD(c) == FALSE) {
198 // Normal, non-supplementary case.
199 return c;
200 }
201
202 //
203 // Possible supplementary char.
204 //
205 UChar32 trail = 0;
206 UChar32 supplementaryC = c;
207 if ((ut->chunkOffset+1) < ut->chunkLength) {
208 // The trail surrogate is in the same chunk.
209 trail = ut->chunkContents[ut->chunkOffset+1];
210 } else {
211 // The trail surrogate is in a different chunk.
212 // Because we must maintain the iteration position, we need to switch forward
213 // into the new chunk, get the trail surrogate, then revert the chunk back to the
214 // original one.
215 // An edge case to be careful of: the entire text may end with an unpaired
216 // leading surrogate. The attempt to access the trail will fail, but
217 // the original position before the unpaired lead still needs to be restored.
218 int64_t nativePosition = ut->chunkNativeLimit;
219 int32_t originalOffset = ut->chunkOffset;
220 if (ut->pFuncs->access(ut, nativePosition, TRUE)) {
221 trail = ut->chunkContents[ut->chunkOffset];
222 }
223 UBool r = ut->pFuncs->access(ut, nativePosition, FALSE); // reverse iteration flag loads preceding chunk
224 U_ASSERT(r==TRUE);
225 ut->chunkOffset = originalOffset;
226 if(!r) {
227 return U_SENTINEL;
228 }
229 }
230
231 if (U16_IS_TRAIL(trail)) {
232 supplementaryC = U16_GET_SUPPLEMENTARY(c, trail);
233 }
234 return supplementaryC;
235
236}
237
238
239U_CAPI UChar32 U_EXPORT2
240utext_char32At(UText *ut, int64_t nativeIndex) {
241 UChar32 c = U_SENTINEL;
242
243 // Fast path the common case.
244 if (nativeIndex>=ut->chunkNativeStart && nativeIndex < ut->chunkNativeStart + ut->nativeIndexingLimit) {
245 ut->chunkOffset = (int32_t)(nativeIndex - ut->chunkNativeStart);
246 c = ut->chunkContents[ut->chunkOffset];
247 if (U16_IS_SURROGATE(c) == FALSE) {
248 return c;
249 }
250 }
251
252
253 utext_setNativeIndex(ut, nativeIndex);
254 if (nativeIndex>=ut->chunkNativeStart && ut->chunkOffset<ut->chunkLength) {
255 c = ut->chunkContents[ut->chunkOffset];
256 if (U16_IS_SURROGATE(c)) {
257 // For surrogates, let current32() deal with the complications
258 // of supplementaries that may span chunk boundaries.
259 c = utext_current32(ut);
260 }
261 }
262 return c;
263}
264
265
266U_CAPI UChar32 U_EXPORT2
267utext_next32(UText *ut) {
268 UChar32 c;
269
270 if (ut->chunkOffset >= ut->chunkLength) {
271 if (ut->pFuncs->access(ut, ut->chunkNativeLimit, TRUE) == FALSE) {
272 return U_SENTINEL;
273 }
274 }
275
276 c = ut->chunkContents[ut->chunkOffset++];
277 if (U16_IS_LEAD(c) == FALSE) {
278 // Normal case, not supplementary.
279 // (A trail surrogate seen here is just returned as is, as a surrogate value.
280 // It cannot be part of a pair.)
281 return c;
282 }
283
284 if (ut->chunkOffset >= ut->chunkLength) {
285 if (ut->pFuncs->access(ut, ut->chunkNativeLimit, TRUE) == FALSE) {
286 // c is an unpaired lead surrogate at the end of the text.
287 // return it as it is.
288 return c;
289 }
290 }
291 UChar32 trail = ut->chunkContents[ut->chunkOffset];
292 if (U16_IS_TRAIL(trail) == FALSE) {
293 // c was an unpaired lead surrogate, not at the end of the text.
294 // return it as it is (unpaired). Iteration position is on the
295 // following character, possibly in the next chunk, where the
296 // trail surrogate would have been if it had existed.
297 return c;
298 }
299
300 UChar32 supplementary = U16_GET_SUPPLEMENTARY(c, trail);
301 ut->chunkOffset++; // move iteration position over the trail surrogate.
302 return supplementary;
303 }
304
305
306U_CAPI UChar32 U_EXPORT2
307utext_previous32(UText *ut) {
308 UChar32 c;
309
310 if (ut->chunkOffset <= 0) {
311 if (ut->pFuncs->access(ut, ut->chunkNativeStart, FALSE) == FALSE) {
312 return U_SENTINEL;
313 }
314 }
315 ut->chunkOffset--;
316 c = ut->chunkContents[ut->chunkOffset];
317 if (U16_IS_TRAIL(c) == FALSE) {
318 // Normal case, not supplementary.
319 // (A lead surrogate seen here is just returned as is, as a surrogate value.
320 // It cannot be part of a pair.)
321 return c;
322 }
323
324 if (ut->chunkOffset <= 0) {
325 if (ut->pFuncs->access(ut, ut->chunkNativeStart, FALSE) == FALSE) {
326 // c is an unpaired trail surrogate at the start of the text.
327 // return it as it is.
328 return c;
329 }
330 }
331
332 UChar32 lead = ut->chunkContents[ut->chunkOffset-1];
333 if (U16_IS_LEAD(lead) == FALSE) {
334 // c was an unpaired trail surrogate, not at the end of the text.
335 // return it as it is (unpaired). Iteration position is at c
336 return c;
337 }
338
339 UChar32 supplementary = U16_GET_SUPPLEMENTARY(lead, c);
340 ut->chunkOffset--; // move iteration position over the lead surrogate.
341 return supplementary;
342}
343
344
345
346U_CAPI UChar32 U_EXPORT2
347utext_next32From(UText *ut, int64_t index) {
348 UChar32 c = U_SENTINEL;
349
350 if(index<ut->chunkNativeStart || index>=ut->chunkNativeLimit) {
351 // Desired position is outside of the current chunk.
352 if(!ut->pFuncs->access(ut, index, TRUE)) {
353 // no chunk available here
354 return U_SENTINEL;
355 }
356 } else if (index - ut->chunkNativeStart <= (int64_t)ut->nativeIndexingLimit) {
357 // Desired position is in chunk, with direct 1:1 native to UTF16 indexing
358 ut->chunkOffset = (int32_t)(index - ut->chunkNativeStart);
359 } else {
360 // Desired position is in chunk, with non-UTF16 indexing.
361 ut->chunkOffset = ut->pFuncs->mapNativeIndexToUTF16(ut, index);
362 }
363
364 c = ut->chunkContents[ut->chunkOffset++];
365 if (U16_IS_SURROGATE(c)) {
366 // Surrogates. Many edge cases. Use other functions that already
367 // deal with the problems.
368 utext_setNativeIndex(ut, index);
369 c = utext_next32(ut);
370 }
371 return c;
372}
373
374
375U_CAPI UChar32 U_EXPORT2
376utext_previous32From(UText *ut, int64_t index) {
377 //
378 // Return the character preceding the specified index.
379 // Leave the iteration position at the start of the character that was returned.
380 //
381 UChar32 cPrev; // The character preceding cCurr, which is what we will return.
382
383 // Address the chunk containg the position preceding the incoming index
384 // A tricky edge case:
385 // We try to test the requested native index against the chunkNativeStart to determine
386 // whether the character preceding the one at the index is in the current chunk.
387 // BUT, this test can fail with UTF-8 (or any other multibyte encoding), when the
388 // requested index is on something other than the first position of the first char.
389 //
390 if(index<=ut->chunkNativeStart || index>ut->chunkNativeLimit) {
391 // Requested native index is outside of the current chunk.
392 if(!ut->pFuncs->access(ut, index, FALSE)) {
393 // no chunk available here
394 return U_SENTINEL;
395 }
396 } else if(index - ut->chunkNativeStart <= (int64_t)ut->nativeIndexingLimit) {
397 // Direct UTF-16 indexing.
398 ut->chunkOffset = (int32_t)(index - ut->chunkNativeStart);
399 } else {
400 ut->chunkOffset=ut->pFuncs->mapNativeIndexToUTF16(ut, index);
401 if (ut->chunkOffset==0 && !ut->pFuncs->access(ut, index, FALSE)) {
402 // no chunk available here
403 return U_SENTINEL;
404 }
405 }
406
407 //
408 // Simple case with no surrogates.
409 //
410 ut->chunkOffset--;
411 cPrev = ut->chunkContents[ut->chunkOffset];
412
413 if (U16_IS_SURROGATE(cPrev)) {
414 // Possible supplementary. Many edge cases.
415 // Let other functions do the heavy lifting.
416 utext_setNativeIndex(ut, index);
417 cPrev = utext_previous32(ut);
418 }
419 return cPrev;
420}
421
422
423U_CAPI int32_t U_EXPORT2
424utext_extract(UText *ut,
425 int64_t start, int64_t limit,
426 UChar *dest, int32_t destCapacity,
427 UErrorCode *status) {
428 return ut->pFuncs->extract(ut, start, limit, dest, destCapacity, status);
429 }
430
431
432
433U_CAPI UBool U_EXPORT2
434utext_equals(const UText *a, const UText *b) {
435 if (a==NULL || b==NULL ||
436 a->magic != UTEXT_MAGIC ||
437 b->magic != UTEXT_MAGIC) {
438 // Null or invalid arguments don't compare equal to anything.
439 return FALSE;
440 }
441
442 if (a->pFuncs != b->pFuncs) {
443 // Different types of text providers.
444 return FALSE;
445 }
446
447 if (a->context != b->context) {
448 // Different sources (different strings)
449 return FALSE;
450 }
451 if (utext_getNativeIndex(a) != utext_getNativeIndex(b)) {
452 // Different current position in the string.
453 return FALSE;
454 }
455
456 return TRUE;
457}
458
459U_CAPI UBool U_EXPORT2
460utext_isWritable(const UText *ut)
461{
462 UBool b = (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_WRITABLE)) != 0;
463 return b;
464}
465
466
467U_CAPI void U_EXPORT2
468utext_freeze(UText *ut) {
469 // Zero out the WRITABLE flag.
470 ut->providerProperties &= ~(I32_FLAG(UTEXT_PROVIDER_WRITABLE));
471}
472
473
474U_CAPI UBool U_EXPORT2
475utext_hasMetaData(const UText *ut)
476{
477 UBool b = (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_HAS_META_DATA)) != 0;
478 return b;
479}
480
481
482
483U_CAPI int32_t U_EXPORT2
484utext_replace(UText *ut,
485 int64_t nativeStart, int64_t nativeLimit,
486 const UChar *replacementText, int32_t replacementLength,
487 UErrorCode *status)
488{
489 if (U_FAILURE(*status)) {
490 return 0;
491 }
492 if ((ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_WRITABLE)) == 0) {
493 *status = U_NO_WRITE_PERMISSION;
494 return 0;
495 }
496 int32_t i = ut->pFuncs->replace(ut, nativeStart, nativeLimit, replacementText, replacementLength, status);
497 return i;
498}
499
500U_CAPI void U_EXPORT2
501utext_copy(UText *ut,
502 int64_t nativeStart, int64_t nativeLimit,
503 int64_t destIndex,
504 UBool move,
505 UErrorCode *status)
506{
507 if (U_FAILURE(*status)) {
508 return;
509 }
510 if ((ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_WRITABLE)) == 0) {
511 *status = U_NO_WRITE_PERMISSION;
512 return;
513 }
514 ut->pFuncs->copy(ut, nativeStart, nativeLimit, destIndex, move, status);
515}
516
517
518
519U_CAPI UText * U_EXPORT2
520utext_clone(UText *dest, const UText *src, UBool deep, UBool readOnly, UErrorCode *status) {
521 if (U_FAILURE(*status)) {
522 return dest;
523 }
524 UText *result = src->pFuncs->clone(dest, src, deep, status);
525 if (U_FAILURE(*status)) {
526 return result;
527 }
528 if (result == NULL) {
529 *status = U_MEMORY_ALLOCATION_ERROR;
530 return result;
531 }
532 if (readOnly) {
533 utext_freeze(result);
534 }
535 return result;
536}
537
538
539
540//------------------------------------------------------------------------------
541//
542// UText common functions implementation
543//
544//------------------------------------------------------------------------------
545
546//
547// UText.flags bit definitions
548//
549enum {
550 UTEXT_HEAP_ALLOCATED = 1, // 1 if ICU has allocated this UText struct on the heap.
551 // 0 if caller provided storage for the UText.
552
553 UTEXT_EXTRA_HEAP_ALLOCATED = 2, // 1 if ICU has allocated extra storage as a separate
554 // heap block.
555 // 0 if there is no separate allocation. Either no extra
556 // storage was requested, or it is appended to the end
557 // of the main UText storage.
558
559 UTEXT_OPEN = 4 // 1 if this UText is currently open
560 // 0 if this UText is not open.
561};
562
563
564//
565// Extended form of a UText. The purpose is to aid in computing the total size required
566// when a provider asks for a UText to be allocated with extra storage.
567
568struct ExtendedUText {
569 UText ut;
570 max_align_t extension;
571};
572
573static const UText emptyText = UTEXT_INITIALIZER;
574
575U_CAPI UText * U_EXPORT2
576utext_setup(UText *ut, int32_t extraSpace, UErrorCode *status) {
577 if (U_FAILURE(*status)) {
578 return ut;
579 }
580
581 if (ut == NULL) {
582 // We need to heap-allocate storage for the new UText
583 int32_t spaceRequired = sizeof(UText);
584 if (extraSpace > 0) {
585 spaceRequired = sizeof(ExtendedUText) + extraSpace - sizeof(max_align_t);
586 }
587 ut = (UText *)uprv_malloc(spaceRequired);
588 if (ut == NULL) {
589 *status = U_MEMORY_ALLOCATION_ERROR;
590 return NULL;
591 } else {
592 *ut = emptyText;
593 ut->flags |= UTEXT_HEAP_ALLOCATED;
594 if (spaceRequired>0) {
595 ut->extraSize = extraSpace;
596 ut->pExtra = &((ExtendedUText *)ut)->extension;
597 }
598 }
599 } else {
600 // We have been supplied with an already existing UText.
601 // Verify that it really appears to be a UText.
602 if (ut->magic != UTEXT_MAGIC) {
603 *status = U_ILLEGAL_ARGUMENT_ERROR;
604 return ut;
605 }
606 // If the ut is already open and there's a provider supplied close
607 // function, call it.
608 if ((ut->flags & UTEXT_OPEN) && ut->pFuncs->close != NULL) {
609 ut->pFuncs->close(ut);
610 }
611 ut->flags &= ~UTEXT_OPEN;
612
613 // If extra space was requested by our caller, check whether
614 // sufficient already exists, and allocate new if needed.
615 if (extraSpace > ut->extraSize) {
616 // Need more space. If there is existing separately allocated space,
617 // delete it first, then allocate new space.
618 if (ut->flags & UTEXT_EXTRA_HEAP_ALLOCATED) {
619 uprv_free(ut->pExtra);
620 ut->extraSize = 0;
621 }
622 ut->pExtra = uprv_malloc(extraSpace);
623 if (ut->pExtra == NULL) {
624 *status = U_MEMORY_ALLOCATION_ERROR;
625 } else {
626 ut->extraSize = extraSpace;
627 ut->flags |= UTEXT_EXTRA_HEAP_ALLOCATED;
628 }
629 }
630 }
631 if (U_SUCCESS(*status)) {
632 ut->flags |= UTEXT_OPEN;
633
634 // Initialize all remaining fields of the UText.
635 //
636 ut->context = NULL;
637 ut->chunkContents = NULL;
638 ut->p = NULL;
639 ut->q = NULL;
640 ut->r = NULL;
641 ut->a = 0;
642 ut->b = 0;
643 ut->c = 0;
644 ut->chunkOffset = 0;
645 ut->chunkLength = 0;
646 ut->chunkNativeStart = 0;
647 ut->chunkNativeLimit = 0;
648 ut->nativeIndexingLimit = 0;
649 ut->providerProperties = 0;
650 ut->privA = 0;
651 ut->privB = 0;
652 ut->privC = 0;
653 ut->privP = NULL;
654 if (ut->pExtra!=NULL && ut->extraSize>0)
655 uprv_memset(ut->pExtra, 0, ut->extraSize);
656
657 }
658 return ut;
659}
660
661
662U_CAPI UText * U_EXPORT2
663utext_close(UText *ut) {
664 if (ut==NULL ||
665 ut->magic != UTEXT_MAGIC ||
666 (ut->flags & UTEXT_OPEN) == 0)
667 {
668 // The supplied ut is not an open UText.
669 // Do nothing.
670 return ut;
671 }
672
673 // If the provider gave us a close function, call it now.
674 // This will clean up anything allocated specifically by the provider.
675 if (ut->pFuncs->close != NULL) {
676 ut->pFuncs->close(ut);
677 }
678 ut->flags &= ~UTEXT_OPEN;
679
680 // If we (the framework) allocated the UText or subsidiary storage,
681 // delete it.
682 if (ut->flags & UTEXT_EXTRA_HEAP_ALLOCATED) {
683 uprv_free(ut->pExtra);
684 ut->pExtra = NULL;
685 ut->flags &= ~UTEXT_EXTRA_HEAP_ALLOCATED;
686 ut->extraSize = 0;
687 }
688
689 // Zero out function table of the closed UText. This is a defensive move,
690 // inteded to cause applications that inadvertantly use a closed
691 // utext to crash with null pointer errors.
692 ut->pFuncs = NULL;
693
694 if (ut->flags & UTEXT_HEAP_ALLOCATED) {
695 // This UText was allocated by UText setup. We need to free it.
696 // Clear magic, so we can detect if the user messes up and immediately
697 // tries to reopen another UText using the deleted storage.
698 ut->magic = 0;
699 uprv_free(ut);
700 ut = NULL;
701 }
702 return ut;
703}
704
705
706
707
708//
709// invalidateChunk Reset a chunk to have no contents, so that the next call
710// to access will cause new data to load.
711// This is needed when copy/move/replace operate directly on the
712// backing text, potentially putting it out of sync with the
713// contents in the chunk.
714//
715static void
716invalidateChunk(UText *ut) {
717 ut->chunkLength = 0;
718 ut->chunkNativeLimit = 0;
719 ut->chunkNativeStart = 0;
720 ut->chunkOffset = 0;
721 ut->nativeIndexingLimit = 0;
722}
723
724//
725// pinIndex Do range pinning on a native index parameter.
726// 64 bit pinning is done in place.
727// 32 bit truncated result is returned as a convenience for
728// use in providers that don't need 64 bits.
729static int32_t
730pinIndex(int64_t &index, int64_t limit) {
731 if (index<0) {
732 index = 0;
733 } else if (index > limit) {
734 index = limit;
735 }
736 return (int32_t)index;
737}
738
739
740U_CDECL_BEGIN
741
742//
743// Pointer relocation function,
744// a utility used by shallow clone.
745// Adjust a pointer that refers to something within one UText (the source)
746// to refer to the same relative offset within a another UText (the target)
747//
748static void adjustPointer(UText *dest, const void **destPtr, const UText *src) {
749 // convert all pointers to (char *) so that byte address arithmetic will work.
750 char *dptr = (char *)*destPtr;
751 char *dUText = (char *)dest;
752 char *sUText = (char *)src;
753
754 if (dptr >= (char *)src->pExtra && dptr < ((char*)src->pExtra)+src->extraSize) {
755 // target ptr was to something within the src UText's pExtra storage.
756 // relocate it into the target UText's pExtra region.
757 *destPtr = ((char *)dest->pExtra) + (dptr - (char *)src->pExtra);
758 } else if (dptr>=sUText && dptr < sUText+src->sizeOfStruct) {
759 // target ptr was pointing to somewhere within the source UText itself.
760 // Move it to the same offset within the target UText.
761 *destPtr = dUText + (dptr-sUText);
762 }
763}
764
765
766//
767// Clone. This is a generic copy-the-utext-by-value clone function that can be
768// used as-is with some utext types, and as a helper by other clones.
769//
770static UText * U_CALLCONV
771shallowTextClone(UText * dest, const UText * src, UErrorCode * status) {
772 if (U_FAILURE(*status)) {
773 return NULL;
774 }
775 int32_t srcExtraSize = src->extraSize;
776
777 //
778 // Use the generic text_setup to allocate storage if required.
779 //
780 dest = utext_setup(dest, srcExtraSize, status);
781 if (U_FAILURE(*status)) {
782 return dest;
783 }
784
785 //
786 // flags (how the UText was allocated) and the pointer to the
787 // extra storage must retain the values in the cloned utext that
788 // were set up by utext_setup. Save them separately before
789 // copying the whole struct.
790 //
791 void *destExtra = dest->pExtra;
792 int32_t flags = dest->flags;
793
794
795 //
796 // Copy the whole UText struct by value.
797 // Any "Extra" storage is copied also.
798 //
799 int sizeToCopy = src->sizeOfStruct;
800 if (sizeToCopy > dest->sizeOfStruct) {
801 sizeToCopy = dest->sizeOfStruct;
802 }
803 uprv_memcpy(dest, src, sizeToCopy);
804 dest->pExtra = destExtra;
805 dest->flags = flags;
806 if (srcExtraSize > 0) {
807 uprv_memcpy(dest->pExtra, src->pExtra, srcExtraSize);
808 }
809
810 //
811 // Relocate any pointers in the target that refer to the UText itself
812 // to point to the cloned copy rather than the original source.
813 //
814 adjustPointer(dest, &dest->context, src);
815 adjustPointer(dest, &dest->p, src);
816 adjustPointer(dest, &dest->q, src);
817 adjustPointer(dest, &dest->r, src);
818 adjustPointer(dest, (const void **)&dest->chunkContents, src);
819
820 // The newly shallow-cloned UText does _not_ own the underlying storage for the text.
821 // (The source for the clone may or may not have owned the text.)
822
823 dest->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT);
824
825 return dest;
826}
827
828
829U_CDECL_END
830
831
832
833//------------------------------------------------------------------------------
834//
835// UText implementation for UTF-8 char * strings (read-only)
836// Limitation: string length must be <= 0x7fffffff in length.
837// (length must for in an int32_t variable)
838//
839// Use of UText data members:
840// context pointer to UTF-8 string
841// utext.b is the input string length (bytes).
842// utext.c Length scanned so far in string
843// (for optimizing finding length of zero terminated strings.)
844// utext.p pointer to the current buffer
845// utext.q pointer to the other buffer.
846//
847//------------------------------------------------------------------------------
848
849// Chunk size.
850// Must be less than 85 (256/3), because of byte mapping from UChar indexes to native indexes.
851// Worst case is three native bytes to one UChar. (Supplemenaries are 4 native bytes
852// to two UChars.)
853// The longest illegal byte sequence treated as a single error (and converted to U+FFFD)
854// is a three-byte sequence (truncated four-byte sequence).
855//
856enum { UTF8_TEXT_CHUNK_SIZE=32 };
857
858//
859// UTF8Buf Two of these structs will be set up in the UText's extra allocated space.
860// Each contains the UChar chunk buffer, the to and from native maps, and
861// header info.
862//
863// because backwards iteration fills the buffers starting at the end and
864// working towards the front, the filled part of the buffers may not begin
865// at the start of the available storage for the buffers.
866//
867// Buffer size is one bigger than the specified UTF8_TEXT_CHUNK_SIZE to allow for
868// the last character added being a supplementary, and thus requiring a surrogate
869// pair. Doing this is simpler than checking for the edge case.
870//
871
872struct UTF8Buf {
873 int32_t bufNativeStart; // Native index of first char in UChar buf
874 int32_t bufNativeLimit; // Native index following last char in buf.
875 int32_t bufStartIdx; // First filled position in buf.
876 int32_t bufLimitIdx; // Limit of filled range in buf.
877 int32_t bufNILimit; // Limit of native indexing part of buf
878 int32_t toUCharsMapStart; // Native index corresponding to
879 // mapToUChars[0].
880 // Set to bufNativeStart when filling forwards.
881 // Set to computed value when filling backwards.
882
883 UChar buf[UTF8_TEXT_CHUNK_SIZE+4]; // The UChar buffer. Requires one extra position beyond the
884 // the chunk size, to allow for surrogate at the end.
885 // Length must be identical to mapToNative array, below,
886 // because of the way indexing works when the array is
887 // filled backwards during a reverse iteration. Thus,
888 // the additional extra size.
889 uint8_t mapToNative[UTF8_TEXT_CHUNK_SIZE+4]; // map UChar index in buf to
890 // native offset from bufNativeStart.
891 // Requires two extra slots,
892 // one for a supplementary starting in the last normal position,
893 // and one for an entry for the buffer limit position.
894 uint8_t mapToUChars[UTF8_TEXT_CHUNK_SIZE*3+6]; // Map native offset from bufNativeStart to
895 // correspoding offset in filled part of buf.
896 int32_t align;
897};
898
899U_CDECL_BEGIN
900
901//
902// utf8TextLength
903//
904// Get the length of the string. If we don't already know it,
905// we'll need to scan for the trailing nul.
906//
907static int64_t U_CALLCONV
908utf8TextLength(UText *ut) {
909 if (ut->b < 0) {
910 // Zero terminated string, and we haven't scanned to the end yet.
911 // Scan it now.
912 const char *r = (const char *)ut->context + ut->c;
913 while (*r != 0) {
914 r++;
915 }
916 if ((r - (const char *)ut->context) < 0x7fffffff) {
917 ut->b = (int32_t)(r - (const char *)ut->context);
918 } else {
919 // Actual string was bigger (more than 2 gig) than we
920 // can handle. Clip it to 2 GB.
921 ut->b = 0x7fffffff;
922 }
923 ut->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
924 }
925 return ut->b;
926}
927
928
929
930
931
932
933static UBool U_CALLCONV
934utf8TextAccess(UText *ut, int64_t index, UBool forward) {
935 //
936 // Apologies to those who are allergic to goto statements.
937 // Consider each goto to a labelled block to be the equivalent of
938 // call the named block as if it were a function();
939 // return;
940 //
941 const uint8_t *s8=(const uint8_t *)ut->context;
942 UTF8Buf *u8b = NULL;
943 int32_t length = ut->b; // Length of original utf-8
944 int32_t ix= (int32_t)index; // Requested index, trimmed to 32 bits.
945 int32_t mapIndex = 0;
946 if (index<0) {
947 ix=0;
948 } else if (index > 0x7fffffff) {
949 // Strings with 64 bit lengths not supported by this UTF-8 provider.
950 ix = 0x7fffffff;
951 }
952
953 // Pin requested index to the string length.
954 if (ix>length) {
955 if (length>=0) {
956 ix=length;
957 } else if (ix>=ut->c) {
958 // Zero terminated string, and requested index is beyond
959 // the region that has already been scanned.
960 // Scan up to either the end of the string or to the
961 // requested position, whichever comes first.
962 while (ut->c<ix && s8[ut->c]!=0) {
963 ut->c++;
964 }
965 // TODO: support for null terminated string length > 32 bits.
966 if (s8[ut->c] == 0) {
967 // We just found the actual length of the string.
968 // Trim the requested index back to that.
969 ix = ut->c;
970 ut->b = ut->c;
971 length = ut->c;
972 ut->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
973 }
974 }
975 }
976
977 //
978 // Dispatch to the appropriate action for a forward iteration request.
979 //
980 if (forward) {
981 if (ix==ut->chunkNativeLimit) {
982 // Check for normal sequential iteration cases first.
983 if (ix==length) {
984 // Just reached end of string
985 // Don't swap buffers, but do set the
986 // current buffer position.
987 ut->chunkOffset = ut->chunkLength;
988 return FALSE;
989 } else {
990 // End of current buffer.
991 // check whether other buffer already has what we need.
992 UTF8Buf *altB = (UTF8Buf *)ut->q;
993 if (ix>=altB->bufNativeStart && ix<altB->bufNativeLimit) {
994 goto swapBuffers;
995 }
996 }
997 }
998
999 // A random access. Desired index could be in either or niether buf.
1000 // For optimizing the order of testing, first check for the index
1001 // being in the other buffer. This will be the case for uses that
1002 // move back and forth over a fairly limited range
1003 {
1004 u8b = (UTF8Buf *)ut->q; // the alternate buffer
1005 if (ix>=u8b->bufNativeStart && ix<u8b->bufNativeLimit) {
1006 // Requested index is in the other buffer.
1007 goto swapBuffers;
1008 }
1009 if (ix == length) {
1010 // Requested index is end-of-string.
1011 // (this is the case of randomly seeking to the end.
1012 // The case of iterating off the end is handled earlier.)
1013 if (ix == ut->chunkNativeLimit) {
1014 // Current buffer extends up to the end of the string.
1015 // Leave it as the current buffer.
1016 ut->chunkOffset = ut->chunkLength;
1017 return FALSE;
1018 }
1019 if (ix == u8b->bufNativeLimit) {
1020 // Alternate buffer extends to the end of string.
1021 // Swap it in as the current buffer.
1022 goto swapBuffersAndFail;
1023 }
1024
1025 // Neither existing buffer extends to the end of the string.
1026 goto makeStubBuffer;
1027 }
1028
1029 if (ix<ut->chunkNativeStart || ix>=ut->chunkNativeLimit) {
1030 // Requested index is in neither buffer.
1031 goto fillForward;
1032 }
1033
1034 // Requested index is in this buffer.
1035 u8b = (UTF8Buf *)ut->p; // the current buffer
1036 mapIndex = ix - u8b->toUCharsMapStart;
1037 U_ASSERT(mapIndex < (int32_t)sizeof(UTF8Buf::mapToUChars));
1038 ut->chunkOffset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;
1039 return TRUE;
1040
1041 }
1042 }
1043
1044
1045 //
1046 // Dispatch to the appropriate action for a
1047 // Backwards Diretion iteration request.
1048 //
1049 if (ix==ut->chunkNativeStart) {
1050 // Check for normal sequential iteration cases first.
1051 if (ix==0) {
1052 // Just reached the start of string
1053 // Don't swap buffers, but do set the
1054 // current buffer position.
1055 ut->chunkOffset = 0;
1056 return FALSE;
1057 } else {
1058 // Start of current buffer.
1059 // check whether other buffer already has what we need.
1060 UTF8Buf *altB = (UTF8Buf *)ut->q;
1061 if (ix>altB->bufNativeStart && ix<=altB->bufNativeLimit) {
1062 goto swapBuffers;
1063 }
1064 }
1065 }
1066
1067 // A random access. Desired index could be in either or niether buf.
1068 // For optimizing the order of testing,
1069 // Most likely case: in the other buffer.
1070 // Second most likely: in neither buffer.
1071 // Unlikely, but must work: in the current buffer.
1072 u8b = (UTF8Buf *)ut->q; // the alternate buffer
1073 if (ix>u8b->bufNativeStart && ix<=u8b->bufNativeLimit) {
1074 // Requested index is in the other buffer.
1075 goto swapBuffers;
1076 }
1077 // Requested index is start-of-string.
1078 // (this is the case of randomly seeking to the start.
1079 // The case of iterating off the start is handled earlier.)
1080 if (ix==0) {
1081 if (u8b->bufNativeStart==0) {
1082 // Alternate buffer contains the data for the start string.
1083 // Make it be the current buffer.
1084 goto swapBuffersAndFail;
1085 } else {
1086 // Request for data before the start of string,
1087 // neither buffer is usable.
1088 // set up a zero-length buffer.
1089 goto makeStubBuffer;
1090 }
1091 }
1092
1093 if (ix<=ut->chunkNativeStart || ix>ut->chunkNativeLimit) {
1094 // Requested index is in neither buffer.
1095 goto fillReverse;
1096 }
1097
1098 // Requested index is in this buffer.
1099 // Set the utf16 buffer index.
1100 u8b = (UTF8Buf *)ut->p;
1101 mapIndex = ix - u8b->toUCharsMapStart;
1102 ut->chunkOffset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;
1103 if (ut->chunkOffset==0) {
1104 // This occurs when the first character in the text is
1105 // a multi-byte UTF-8 char, and the requested index is to
1106 // one of the trailing bytes. Because there is no preceding ,
1107 // character, this access fails. We can't pick up on the
1108 // situation sooner because the requested index is not zero.
1109 return FALSE;
1110 } else {
1111 return TRUE;
1112 }
1113
1114
1115
1116swapBuffers:
1117 // The alternate buffer (ut->q) has the string data that was requested.
1118 // Swap the primary and alternate buffers, and set the
1119 // chunk index into the new primary buffer.
1120 {
1121 u8b = (UTF8Buf *)ut->q;
1122 ut->q = ut->p;
1123 ut->p = u8b;
1124 ut->chunkContents = &u8b->buf[u8b->bufStartIdx];
1125 ut->chunkLength = u8b->bufLimitIdx - u8b->bufStartIdx;
1126 ut->chunkNativeStart = u8b->bufNativeStart;
1127 ut->chunkNativeLimit = u8b->bufNativeLimit;
1128 ut->nativeIndexingLimit = u8b->bufNILimit;
1129
1130 // Index into the (now current) chunk
1131 // Use the map to set the chunk index. It's more trouble than it's worth
1132 // to check whether native indexing can be used.
1133 U_ASSERT(ix>=u8b->bufNativeStart);
1134 U_ASSERT(ix<=u8b->bufNativeLimit);
1135 mapIndex = ix - u8b->toUCharsMapStart;
1136 U_ASSERT(mapIndex>=0);
1137 U_ASSERT(mapIndex<(int32_t)sizeof(u8b->mapToUChars));
1138 ut->chunkOffset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;
1139
1140 return TRUE;
1141 }
1142
1143
1144 swapBuffersAndFail:
1145 // We got a request for either the start or end of the string,
1146 // with iteration continuing in the out-of-bounds direction.
1147 // The alternate buffer already contains the data up to the
1148 // start/end.
1149 // Swap the buffers, then return failure, indicating that we couldn't
1150 // make things correct for continuing the iteration in the requested
1151 // direction. The position & buffer are correct should the
1152 // user decide to iterate in the opposite direction.
1153 u8b = (UTF8Buf *)ut->q;
1154 ut->q = ut->p;
1155 ut->p = u8b;
1156 ut->chunkContents = &u8b->buf[u8b->bufStartIdx];
1157 ut->chunkLength = u8b->bufLimitIdx - u8b->bufStartIdx;
1158 ut->chunkNativeStart = u8b->bufNativeStart;
1159 ut->chunkNativeLimit = u8b->bufNativeLimit;
1160 ut->nativeIndexingLimit = u8b->bufNILimit;
1161
1162 // Index into the (now current) chunk
1163 // For this function (swapBuffersAndFail), the requested index
1164 // will always be at either the start or end of the chunk.
1165 if (ix==u8b->bufNativeLimit) {
1166 ut->chunkOffset = ut->chunkLength;
1167 } else {
1168 ut->chunkOffset = 0;
1169 U_ASSERT(ix == u8b->bufNativeStart);
1170 }
1171 return FALSE;
1172
1173makeStubBuffer:
1174 // The user has done a seek/access past the start or end
1175 // of the string. Rather than loading data that is likely
1176 // to never be used, just set up a zero-length buffer at
1177 // the position.
1178 u8b = (UTF8Buf *)ut->q;
1179 u8b->bufNativeStart = ix;
1180 u8b->bufNativeLimit = ix;
1181 u8b->bufStartIdx = 0;
1182 u8b->bufLimitIdx = 0;
1183 u8b->bufNILimit = 0;
1184 u8b->toUCharsMapStart = ix;
1185 u8b->mapToNative[0] = 0;
1186 u8b->mapToUChars[0] = 0;
1187 goto swapBuffersAndFail;
1188
1189
1190
1191fillForward:
1192 {
1193 // Move the incoming index to a code point boundary.
1194 U8_SET_CP_START(s8, 0, ix);
1195
1196 // Swap the UText buffers.
1197 // We want to fill what was previously the alternate buffer,
1198 // and make what was the current buffer be the new alternate.
1199 UTF8Buf *u8b_swap = (UTF8Buf *)ut->q;
1200 ut->q = ut->p;
1201 ut->p = u8b_swap;
1202
1203 int32_t strLen = ut->b;
1204 UBool nulTerminated = FALSE;
1205 if (strLen < 0) {
1206 strLen = 0x7fffffff;
1207 nulTerminated = TRUE;
1208 }
1209
1210 UChar *buf = u8b_swap->buf;
1211 uint8_t *mapToNative = u8b_swap->mapToNative;
1212 uint8_t *mapToUChars = u8b_swap->mapToUChars;
1213 int32_t destIx = 0;
1214 int32_t srcIx = ix;
1215 UBool seenNonAscii = FALSE;
1216 UChar32 c = 0;
1217
1218 // Fill the chunk buffer and mapping arrays.
1219 while (destIx<UTF8_TEXT_CHUNK_SIZE) {
1220 c = s8[srcIx];
1221 if (c>0 && c<0x80) {
1222 // Special case ASCII range for speed.
1223 // zero is excluded to simplify bounds checking.
1224 buf[destIx] = (UChar)c;
1225 mapToNative[destIx] = (uint8_t)(srcIx - ix);
1226 mapToUChars[srcIx-ix] = (uint8_t)destIx;
1227 srcIx++;
1228 destIx++;
1229 } else {
1230 // General case, handle everything.
1231 if (seenNonAscii == FALSE) {
1232 seenNonAscii = TRUE;
1233 u8b_swap->bufNILimit = destIx;
1234 }
1235
1236 int32_t cIx = srcIx;
1237 int32_t dIx = destIx;
1238 int32_t dIxSaved = destIx;
1239 U8_NEXT_OR_FFFD(s8, srcIx, strLen, c);
1240 if (c==0 && nulTerminated) {
1241 srcIx--;
1242 break;
1243 }
1244
1245 U16_APPEND_UNSAFE(buf, destIx, c);
1246 do {
1247 mapToNative[dIx++] = (uint8_t)(cIx - ix);
1248 } while (dIx < destIx);
1249
1250 do {
1251 mapToUChars[cIx++ - ix] = (uint8_t)dIxSaved;
1252 } while (cIx < srcIx);
1253 }
1254 if (srcIx>=strLen) {
1255 break;
1256 }
1257
1258 }
1259
1260 // store Native <--> Chunk Map entries for the end of the buffer.
1261 // There is no actual character here, but the index position is valid.
1262 mapToNative[destIx] = (uint8_t)(srcIx - ix);
1263 mapToUChars[srcIx - ix] = (uint8_t)destIx;
1264
1265 // fill in Buffer descriptor
1266 u8b_swap->bufNativeStart = ix;
1267 u8b_swap->bufNativeLimit = srcIx;
1268 u8b_swap->bufStartIdx = 0;
1269 u8b_swap->bufLimitIdx = destIx;
1270 if (seenNonAscii == FALSE) {
1271 u8b_swap->bufNILimit = destIx;
1272 }
1273 u8b_swap->toUCharsMapStart = u8b_swap->bufNativeStart;
1274
1275 // Set UText chunk to refer to this buffer.
1276 ut->chunkContents = buf;
1277 ut->chunkOffset = 0;
1278 ut->chunkLength = u8b_swap->bufLimitIdx;
1279 ut->chunkNativeStart = u8b_swap->bufNativeStart;
1280 ut->chunkNativeLimit = u8b_swap->bufNativeLimit;
1281 ut->nativeIndexingLimit = u8b_swap->bufNILimit;
1282
1283 // For zero terminated strings, keep track of the maximum point
1284 // scanned so far.
1285 if (nulTerminated && srcIx>ut->c) {
1286 ut->c = srcIx;
1287 if (c==0) {
1288 // We scanned to the end.
1289 // Remember the actual length.
1290 ut->b = srcIx;
1291 ut->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
1292 }
1293 }
1294 return TRUE;
1295 }
1296
1297
1298fillReverse:
1299 {
1300 // Move the incoming index to a code point boundary.
1301 // Can only do this if the incoming index is somewhere in the interior of the string.
1302 // If index is at the end, there is no character there to look at.
1303 if (ix != ut->b) {
1304 // Note: this function will only move the index back if it is on a trail byte
1305 // and there is a preceding lead byte and the sequence from the lead
1306 // through this trail could be part of a valid UTF-8 sequence
1307 // Otherwise the index remains unchanged.
1308 U8_SET_CP_START(s8, 0, ix);
1309 }
1310
1311 // Swap the UText buffers.
1312 // We want to fill what was previously the alternate buffer,
1313 // and make what was the current buffer be the new alternate.
1314 UTF8Buf *u8b_swap = (UTF8Buf *)ut->q;
1315 ut->q = ut->p;
1316 ut->p = u8b_swap;
1317
1318 UChar *buf = u8b_swap->buf;
1319 uint8_t *mapToNative = u8b_swap->mapToNative;
1320 uint8_t *mapToUChars = u8b_swap->mapToUChars;
1321 int32_t toUCharsMapStart = ix - sizeof(UTF8Buf::mapToUChars) + 1;
1322 // Note that toUCharsMapStart can be negative. Happens when the remaining
1323 // text from current position to the beginning is less than the buffer size.
1324 // + 1 because mapToUChars must have a slot at the end for the bufNativeLimit entry.
1325 int32_t destIx = UTF8_TEXT_CHUNK_SIZE+2; // Start in the overflow region
1326 // at end of buffer to leave room
1327 // for a surrogate pair at the
1328 // buffer start.
1329 int32_t srcIx = ix;
1330 int32_t bufNILimit = destIx;
1331 UChar32 c;
1332
1333 // Map to/from Native Indexes, fill in for the position at the end of
1334 // the buffer.
1335 //
1336 mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
1337 mapToUChars[srcIx - toUCharsMapStart] = (uint8_t)destIx;
1338
1339 // Fill the chunk buffer
1340 // Work backwards, filling from the end of the buffer towards the front.
1341 //
1342 while (destIx>2 && (srcIx - toUCharsMapStart > 5) && (srcIx > 0)) {
1343 srcIx--;
1344 destIx--;
1345
1346 // Get last byte of the UTF-8 character
1347 c = s8[srcIx];
1348 if (c<0x80) {
1349 // Special case ASCII range for speed.
1350 buf[destIx] = (UChar)c;
1351 U_ASSERT(toUCharsMapStart <= srcIx);
1352 mapToUChars[srcIx - toUCharsMapStart] = (uint8_t)destIx;
1353 mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
1354 } else {
1355 // General case, handle everything non-ASCII.
1356
1357 int32_t sIx = srcIx; // ix of last byte of multi-byte u8 char
1358
1359 // Get the full character from the UTF8 string.
1360 // use code derived from tbe macros in utf8.h
1361 // Leaves srcIx pointing at the first byte of the UTF-8 char.
1362 //
1363 c=utf8_prevCharSafeBody(s8, 0, &srcIx, c, -3);
1364 // leaves srcIx at first byte of the multi-byte char.
1365
1366 // Store the character in UTF-16 buffer.
1367 if (c<0x10000) {
1368 buf[destIx] = (UChar)c;
1369 mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
1370 } else {
1371 buf[destIx] = U16_TRAIL(c);
1372 mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
1373 buf[--destIx] = U16_LEAD(c);
1374 mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
1375 }
1376
1377 // Fill in the map from native indexes to UChars buf index.
1378 do {
1379 mapToUChars[sIx-- - toUCharsMapStart] = (uint8_t)destIx;
1380 } while (sIx >= srcIx);
1381 U_ASSERT(toUCharsMapStart <= (srcIx+1));
1382
1383 // Set native indexing limit to be the current position.
1384 // We are processing a non-ascii, non-native-indexing char now;
1385 // the limit will be here if the rest of the chars to be
1386 // added to this buffer are ascii.
1387 bufNILimit = destIx;
1388 }
1389 }
1390 u8b_swap->bufNativeStart = srcIx;
1391 u8b_swap->bufNativeLimit = ix;
1392 u8b_swap->bufStartIdx = destIx;
1393 u8b_swap->bufLimitIdx = UTF8_TEXT_CHUNK_SIZE+2;
1394 u8b_swap->bufNILimit = bufNILimit - u8b_swap->bufStartIdx;
1395 u8b_swap->toUCharsMapStart = toUCharsMapStart;
1396
1397 ut->chunkContents = &buf[u8b_swap->bufStartIdx];
1398 ut->chunkLength = u8b_swap->bufLimitIdx - u8b_swap->bufStartIdx;
1399 ut->chunkOffset = ut->chunkLength;
1400 ut->chunkNativeStart = u8b_swap->bufNativeStart;
1401 ut->chunkNativeLimit = u8b_swap->bufNativeLimit;
1402 ut->nativeIndexingLimit = u8b_swap->bufNILimit;
1403 return TRUE;
1404 }
1405
1406}
1407
1408
1409
1410//
1411// This is a slightly modified copy of u_strFromUTF8,
1412// Inserts a Replacement Char rather than failing on invalid UTF-8
1413// Removes unnecessary features.
1414//
1415static UChar*
1416utext_strFromUTF8(UChar *dest,
1417 int32_t destCapacity,
1418 int32_t *pDestLength,
1419 const char* src,
1420 int32_t srcLength, // required. NUL terminated not supported.
1421 UErrorCode *pErrorCode
1422 )
1423{
1424
1425 UChar *pDest = dest;
1426 UChar *pDestLimit = (dest!=NULL)?(dest+destCapacity):NULL;
1427 UChar32 ch=0;
1428 int32_t index = 0;
1429 int32_t reqLength = 0;
1430 uint8_t* pSrc = (uint8_t*) src;
1431
1432
1433 while((index < srcLength)&&(pDest<pDestLimit)){
1434 ch = pSrc[index++];
1435 if(ch <=0x7f){
1436 *pDest++=(UChar)ch;
1437 }else{
1438 ch=utf8_nextCharSafeBody(pSrc, &index, srcLength, ch, -3);
1439 if(U_IS_BMP(ch)){
1440 *(pDest++)=(UChar)ch;
1441 }else{
1442 *(pDest++)=U16_LEAD(ch);
1443 if(pDest<pDestLimit){
1444 *(pDest++)=U16_TRAIL(ch);
1445 }else{
1446 reqLength++;
1447 break;
1448 }
1449 }
1450 }
1451 }
1452 /* donot fill the dest buffer just count the UChars needed */
1453 while(index < srcLength){
1454 ch = pSrc[index++];
1455 if(ch <= 0x7f){
1456 reqLength++;
1457 }else{
1458 ch=utf8_nextCharSafeBody(pSrc, &index, srcLength, ch, -3);
1459 reqLength+=U16_LENGTH(ch);
1460 }
1461 }
1462
1463 reqLength+=(int32_t)(pDest - dest);
1464
1465 if(pDestLength){
1466 *pDestLength = reqLength;
1467 }
1468
1469 /* Terminate the buffer */
1470 u_terminateUChars(dest,destCapacity,reqLength,pErrorCode);
1471
1472 return dest;
1473}
1474
1475
1476
1477static int32_t U_CALLCONV
1478utf8TextExtract(UText *ut,
1479 int64_t start, int64_t limit,
1480 UChar *dest, int32_t destCapacity,
1481 UErrorCode *pErrorCode) {
1482 if(U_FAILURE(*pErrorCode)) {
1483 return 0;
1484 }
1485 if(destCapacity<0 || (dest==NULL && destCapacity>0)) {
1486 *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
1487 return 0;
1488 }
1489 int32_t length = ut->b;
1490 int32_t start32 = pinIndex(start, length);
1491 int32_t limit32 = pinIndex(limit, length);
1492
1493 if(start32>limit32) {
1494 *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR;
1495 return 0;
1496 }
1497
1498
1499 // adjust the incoming indexes to land on code point boundaries if needed.
1500 // adjust by no more than three, because that is the largest number of trail bytes
1501 // in a well formed UTF8 character.
1502 const uint8_t *buf = (const uint8_t *)ut->context;
1503 int i;
1504 if (start32 < ut->chunkNativeLimit) {
1505 for (i=0; i<3; i++) {
1506 if (U8_IS_SINGLE(buf[start32]) || U8_IS_LEAD(buf[start32]) || start32==0) {
1507 break;
1508 }
1509 start32--;
1510 }
1511 }
1512
1513 if (limit32 < ut->chunkNativeLimit) {
1514 for (i=0; i<3; i++) {
1515 if (U8_IS_SINGLE(buf[limit32]) || U8_IS_LEAD(buf[limit32]) || limit32==0) {
1516 break;
1517 }
1518 limit32--;
1519 }
1520 }
1521
1522 // Do the actual extract.
1523 int32_t destLength=0;
1524 utext_strFromUTF8(dest, destCapacity, &destLength,
1525 (const char *)ut->context+start32, limit32-start32,
1526 pErrorCode);
1527 utf8TextAccess(ut, limit32, TRUE);
1528 return destLength;
1529}
1530
1531//
1532// utf8TextMapOffsetToNative
1533//
1534// Map a chunk (UTF-16) offset to a native index.
1535static int64_t U_CALLCONV
1536utf8TextMapOffsetToNative(const UText *ut) {
1537 //
1538 UTF8Buf *u8b = (UTF8Buf *)ut->p;
1539 U_ASSERT(ut->chunkOffset>ut->nativeIndexingLimit && ut->chunkOffset<=ut->chunkLength);
1540 int32_t nativeOffset = u8b->mapToNative[ut->chunkOffset + u8b->bufStartIdx] + u8b->toUCharsMapStart;
1541 U_ASSERT(nativeOffset >= ut->chunkNativeStart && nativeOffset <= ut->chunkNativeLimit);
1542 return nativeOffset;
1543}
1544
1545//
1546// Map a native index to the corrsponding chunk offset
1547//
1548static int32_t U_CALLCONV
1549utf8TextMapIndexToUTF16(const UText *ut, int64_t index64) {
1550 U_ASSERT(index64 <= 0x7fffffff);
1551 int32_t index = (int32_t)index64;
1552 UTF8Buf *u8b = (UTF8Buf *)ut->p;
1553 U_ASSERT(index>=ut->chunkNativeStart+ut->nativeIndexingLimit);
1554 U_ASSERT(index<=ut->chunkNativeLimit);
1555 int32_t mapIndex = index - u8b->toUCharsMapStart;
1556 U_ASSERT(mapIndex < (int32_t)sizeof(UTF8Buf::mapToUChars));
1557 int32_t offset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;
1558 U_ASSERT(offset>=0 && offset<=ut->chunkLength);
1559 return offset;
1560}
1561
1562static UText * U_CALLCONV
1563utf8TextClone(UText *dest, const UText *src, UBool deep, UErrorCode *status)
1564{
1565 // First do a generic shallow clone. Does everything needed for the UText struct itself.
1566 dest = shallowTextClone(dest, src, status);
1567
1568 // For deep clones, make a copy of the string.
1569 // The copied storage is owned by the newly created clone.
1570 //
1571 // TODO: There is an isssue with using utext_nativeLength().
1572 // That function is non-const in cases where the input was NUL terminated
1573 // and the length has not yet been determined.
1574 // This function (clone()) is const.
1575 // There potentially a thread safety issue lurking here.
1576 //
1577 if (deep && U_SUCCESS(*status)) {
1578 int32_t len = (int32_t)utext_nativeLength((UText *)src);
1579 char *copyStr = (char *)uprv_malloc(len+1);
1580 if (copyStr == NULL) {
1581 *status = U_MEMORY_ALLOCATION_ERROR;
1582 } else {
1583 uprv_memcpy(copyStr, src->context, len+1);
1584 dest->context = copyStr;
1585 dest->providerProperties |= I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT);
1586 }
1587 }
1588 return dest;
1589}
1590
1591
1592static void U_CALLCONV
1593utf8TextClose(UText *ut) {
1594 // Most of the work of close is done by the generic UText framework close.
1595 // All that needs to be done here is to delete the UTF8 string if the UText
1596 // owns it. This occurs if the UText was created by cloning.
1597 if (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT)) {
1598 char *s = (char *)ut->context;
1599 uprv_free(s);
1600 ut->context = NULL;
1601 }
1602}
1603
1604U_CDECL_END
1605
1606
1607static const struct UTextFuncs utf8Funcs =
1608{
1609 sizeof(UTextFuncs),
1610 0, 0, 0, // Reserved alignment padding
1611 utf8TextClone,
1612 utf8TextLength,
1613 utf8TextAccess,
1614 utf8TextExtract,
1615 NULL, /* replace*/
1616 NULL, /* copy */
1617 utf8TextMapOffsetToNative,
1618 utf8TextMapIndexToUTF16,
1619 utf8TextClose,
1620 NULL, // spare 1
1621 NULL, // spare 2
1622 NULL // spare 3
1623};
1624
1625
1626static const char gEmptyString[] = {0};
1627
1628U_CAPI UText * U_EXPORT2
1629utext_openUTF8(UText *ut, const char *s, int64_t length, UErrorCode *status) {
1630 if(U_FAILURE(*status)) {
1631 return NULL;
1632 }
1633 if(s==NULL && length==0) {
1634 s = gEmptyString;
1635 }
1636
1637 if(s==NULL || length<-1 || length>INT32_MAX) {
1638 *status=U_ILLEGAL_ARGUMENT_ERROR;
1639 return NULL;
1640 }
1641
1642 ut = utext_setup(ut, sizeof(UTF8Buf) * 2, status);
1643 if (U_FAILURE(*status)) {
1644 return ut;
1645 }
1646
1647 ut->pFuncs = &utf8Funcs;
1648 ut->context = s;
1649 ut->b = (int32_t)length;
1650 ut->c = (int32_t)length;
1651 if (ut->c < 0) {
1652 ut->c = 0;
1653 ut->providerProperties |= I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
1654 }
1655 ut->p = ut->pExtra;
1656 ut->q = (char *)ut->pExtra + sizeof(UTF8Buf);
1657 return ut;
1658
1659}
1660
1661
1662
1663
1664
1665
1666
1667
1668//------------------------------------------------------------------------------
1669//
1670// UText implementation wrapper for Replaceable (read/write)
1671//
1672// Use of UText data members:
1673// context pointer to Replaceable.
1674// p pointer to Replaceable if it is owned by the UText.
1675//
1676//------------------------------------------------------------------------------
1677
1678
1679
1680// minimum chunk size for this implementation: 3
1681// to allow for possible trimming for code point boundaries
1682enum { REP_TEXT_CHUNK_SIZE=10 };
1683
1684struct ReplExtra {
1685 /*
1686 * Chunk UChars.
1687 * +1 to simplify filling with surrogate pair at the end.
1688 */
1689 UChar s[REP_TEXT_CHUNK_SIZE+1];
1690};
1691
1692
1693U_CDECL_BEGIN
1694
1695static UText * U_CALLCONV
1696repTextClone(UText *dest, const UText *src, UBool deep, UErrorCode *status) {
1697 // First do a generic shallow clone. Does everything needed for the UText struct itself.
1698 dest = shallowTextClone(dest, src, status);
1699
1700 // For deep clones, make a copy of the Replaceable.
1701 // The copied Replaceable storage is owned by the newly created UText clone.
1702 // A non-NULL pointer in UText.p is the signal to the close() function to delete
1703 // it.
1704 //
1705 if (deep && U_SUCCESS(*status)) {
1706 const Replaceable *replSrc = (const Replaceable *)src->context;
1707 dest->context = replSrc->clone();
1708 dest->providerProperties |= I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT);
1709
1710 // with deep clone, the copy is writable, even when the source is not.
1711 dest->providerProperties |= I32_FLAG(UTEXT_PROVIDER_WRITABLE);
1712 }
1713 return dest;
1714}
1715
1716
1717static void U_CALLCONV
1718repTextClose(UText *ut) {
1719 // Most of the work of close is done by the generic UText framework close.
1720 // All that needs to be done here is delete the Replaceable if the UText
1721 // owns it. This occurs if the UText was created by cloning.
1722 if (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT)) {
1723 Replaceable *rep = (Replaceable *)ut->context;
1724 delete rep;
1725 ut->context = NULL;
1726 }
1727}
1728
1729
1730static int64_t U_CALLCONV
1731repTextLength(UText *ut) {
1732 const Replaceable *replSrc = (const Replaceable *)ut->context;
1733 int32_t len = replSrc->length();
1734 return len;
1735}
1736
1737
1738static UBool U_CALLCONV
1739repTextAccess(UText *ut, int64_t index, UBool forward) {
1740 const Replaceable *rep=(const Replaceable *)ut->context;
1741 int32_t length=rep->length(); // Full length of the input text (bigger than a chunk)
1742
1743 // clip the requested index to the limits of the text.
1744 int32_t index32 = pinIndex(index, length);
1745 U_ASSERT(index<=INT32_MAX);
1746
1747
1748 /*
1749 * Compute start/limit boundaries around index, for a segment of text
1750 * to be extracted.
1751 * To allow for the possibility that our user gave an index to the trailing
1752 * half of a surrogate pair, we must request one extra preceding UChar when
1753 * going in the forward direction. This will ensure that the buffer has the
1754 * entire code point at the specified index.
1755 */
1756 if(forward) {
1757
1758 if (index32>=ut->chunkNativeStart && index32<ut->chunkNativeLimit) {
1759 // Buffer already contains the requested position.
1760 ut->chunkOffset = (int32_t)(index - ut->chunkNativeStart);
1761 return TRUE;
1762 }
1763 if (index32>=length && ut->chunkNativeLimit==length) {
1764 // Request for end of string, and buffer already extends up to it.
1765 // Can't get the data, but don't change the buffer.
1766 ut->chunkOffset = length - (int32_t)ut->chunkNativeStart;
1767 return FALSE;
1768 }
1769
1770 ut->chunkNativeLimit = index + REP_TEXT_CHUNK_SIZE - 1;
1771 // Going forward, so we want to have the buffer with stuff at and beyond
1772 // the requested index. The -1 gets us one code point before the
1773 // requested index also, to handle the case of the index being on
1774 // a trail surrogate of a surrogate pair.
1775 if(ut->chunkNativeLimit > length) {
1776 ut->chunkNativeLimit = length;
1777 }
1778 // unless buffer ran off end, start is index-1.
1779 ut->chunkNativeStart = ut->chunkNativeLimit - REP_TEXT_CHUNK_SIZE;
1780 if(ut->chunkNativeStart < 0) {
1781 ut->chunkNativeStart = 0;
1782 }
1783 } else {
1784 // Reverse iteration. Fill buffer with data preceding the requested index.
1785 if (index32>ut->chunkNativeStart && index32<=ut->chunkNativeLimit) {
1786 // Requested position already in buffer.
1787 ut->chunkOffset = index32 - (int32_t)ut->chunkNativeStart;
1788 return TRUE;
1789 }
1790 if (index32==0 && ut->chunkNativeStart==0) {
1791 // Request for start, buffer already begins at start.
1792 // No data, but keep the buffer as is.
1793 ut->chunkOffset = 0;
1794 return FALSE;
1795 }
1796
1797 // Figure out the bounds of the chunk to extract for reverse iteration.
1798 // Need to worry about chunk not splitting surrogate pairs, and while still
1799 // containing the data we need.
1800 // Fix by requesting a chunk that includes an extra UChar at the end.
1801 // If this turns out to be a lead surrogate, we can lop it off and still have
1802 // the data we wanted.
1803 ut->chunkNativeStart = index32 + 1 - REP_TEXT_CHUNK_SIZE;
1804 if (ut->chunkNativeStart < 0) {
1805 ut->chunkNativeStart = 0;
1806 }
1807
1808 ut->chunkNativeLimit = index32 + 1;
1809 if (ut->chunkNativeLimit > length) {
1810 ut->chunkNativeLimit = length;
1811 }
1812 }
1813
1814 // Extract the new chunk of text from the Replaceable source.
1815 ReplExtra *ex = (ReplExtra *)ut->pExtra;
1816 // UnicodeString with its buffer a writable alias to the chunk buffer
1817 UnicodeString buffer(ex->s, 0 /*buffer length*/, REP_TEXT_CHUNK_SIZE /*buffer capacity*/);
1818 rep->extractBetween((int32_t)ut->chunkNativeStart, (int32_t)ut->chunkNativeLimit, buffer);
1819
1820 ut->chunkContents = ex->s;
1821 ut->chunkLength = (int32_t)(ut->chunkNativeLimit - ut->chunkNativeStart);
1822 ut->chunkOffset = (int32_t)(index32 - ut->chunkNativeStart);
1823
1824 // Surrogate pairs from the input text must not span chunk boundaries.
1825 // If end of chunk could be the start of a surrogate, trim it off.
1826 if (ut->chunkNativeLimit < length &&
1827 U16_IS_LEAD(ex->s[ut->chunkLength-1])) {
1828 ut->chunkLength--;
1829 ut->chunkNativeLimit--;
1830 if (ut->chunkOffset > ut->chunkLength) {
1831 ut->chunkOffset = ut->chunkLength;
1832 }
1833 }
1834
1835 // if the first UChar in the chunk could be the trailing half of a surrogate pair,
1836 // trim it off.
1837 if(ut->chunkNativeStart>0 && U16_IS_TRAIL(ex->s[0])) {
1838 ++(ut->chunkContents);
1839 ++(ut->chunkNativeStart);
1840 --(ut->chunkLength);
1841 --(ut->chunkOffset);
1842 }
1843
1844 // adjust the index/chunkOffset to a code point boundary
1845 U16_SET_CP_START(ut->chunkContents, 0, ut->chunkOffset);
1846
1847 // Use fast indexing for get/setNativeIndex()
1848 ut->nativeIndexingLimit = ut->chunkLength;
1849
1850 return TRUE;
1851}
1852
1853
1854
1855static int32_t U_CALLCONV
1856repTextExtract(UText *ut,
1857 int64_t start, int64_t limit,
1858 UChar *dest, int32_t destCapacity,
1859 UErrorCode *status) {
1860 const Replaceable *rep=(const Replaceable *)ut->context;
1861 int32_t length=rep->length();
1862
1863 if(U_FAILURE(*status)) {
1864 return 0;
1865 }
1866 if(destCapacity<0 || (dest==NULL && destCapacity>0)) {
1867 *status=U_ILLEGAL_ARGUMENT_ERROR;
1868 }
1869 if(start>limit) {
1870 *status=U_INDEX_OUTOFBOUNDS_ERROR;
1871 return 0;
1872 }
1873
1874 int32_t start32 = pinIndex(start, length);
1875 int32_t limit32 = pinIndex(limit, length);
1876
1877 // adjust start, limit if they point to trail half of surrogates
1878 if (start32<length && U16_IS_TRAIL(rep->charAt(start32)) &&
1879 U_IS_SUPPLEMENTARY(rep->char32At(start32))){
1880 start32--;
1881 }
1882 if (limit32<length && U16_IS_TRAIL(rep->charAt(limit32)) &&
1883 U_IS_SUPPLEMENTARY(rep->char32At(limit32))){
1884 limit32--;
1885 }
1886
1887 length=limit32-start32;
1888 if(length>destCapacity) {
1889 limit32 = start32 + destCapacity;
1890 }
1891 UnicodeString buffer(dest, 0, destCapacity); // writable alias
1892 rep->extractBetween(start32, limit32, buffer);
1893 repTextAccess(ut, limit32, TRUE);
1894
1895 return u_terminateUChars(dest, destCapacity, length, status);
1896}
1897
1898static int32_t U_CALLCONV
1899repTextReplace(UText *ut,
1900 int64_t start, int64_t limit,
1901 const UChar *src, int32_t length,
1902 UErrorCode *status) {
1903 Replaceable *rep=(Replaceable *)ut->context;
1904 int32_t oldLength;
1905
1906 if(U_FAILURE(*status)) {
1907 return 0;
1908 }
1909 if(src==NULL && length!=0) {
1910 *status=U_ILLEGAL_ARGUMENT_ERROR;
1911 return 0;
1912 }
1913 oldLength=rep->length(); // will subtract from new length
1914 if(start>limit ) {
1915 *status=U_INDEX_OUTOFBOUNDS_ERROR;
1916 return 0;
1917 }
1918
1919 int32_t start32 = pinIndex(start, oldLength);
1920 int32_t limit32 = pinIndex(limit, oldLength);
1921
1922 // Snap start & limit to code point boundaries.
1923 if (start32<oldLength && U16_IS_TRAIL(rep->charAt(start32)) &&
1924 start32>0 && U16_IS_LEAD(rep->charAt(start32-1)))
1925 {
1926 start32--;
1927 }
1928 if (limit32<oldLength && U16_IS_LEAD(rep->charAt(limit32-1)) &&
1929 U16_IS_TRAIL(rep->charAt(limit32)))
1930 {
1931 limit32++;
1932 }
1933
1934 // Do the actual replace operation using methods of the Replaceable class
1935 UnicodeString replStr((UBool)(length<0), src, length); // read-only alias
1936 rep->handleReplaceBetween(start32, limit32, replStr);
1937 int32_t newLength = rep->length();
1938 int32_t lengthDelta = newLength - oldLength;
1939
1940 // Is the UText chunk buffer OK?
1941 if (ut->chunkNativeLimit > start32) {
1942 // this replace operation may have impacted the current chunk.
1943 // invalidate it, which will force a reload on the next access.
1944 invalidateChunk(ut);
1945 }
1946
1947 // set the iteration position to the end of the newly inserted replacement text.
1948 int32_t newIndexPos = limit32 + lengthDelta;
1949 repTextAccess(ut, newIndexPos, TRUE);
1950
1951 return lengthDelta;
1952}
1953
1954
1955static void U_CALLCONV
1956repTextCopy(UText *ut,
1957 int64_t start, int64_t limit,
1958 int64_t destIndex,
1959 UBool move,
1960 UErrorCode *status)
1961{
1962 Replaceable *rep=(Replaceable *)ut->context;
1963 int32_t length=rep->length();
1964
1965 if(U_FAILURE(*status)) {
1966 return;
1967 }
1968 if (start>limit || (start<destIndex && destIndex<limit))
1969 {
1970 *status=U_INDEX_OUTOFBOUNDS_ERROR;
1971 return;
1972 }
1973
1974 int32_t start32 = pinIndex(start, length);
1975 int32_t limit32 = pinIndex(limit, length);
1976 int32_t destIndex32 = pinIndex(destIndex, length);
1977
1978 // TODO: snap input parameters to code point boundaries.
1979
1980 if(move) {
1981 // move: copy to destIndex, then replace original with nothing
1982 int32_t segLength=limit32-start32;
1983 rep->copy(start32, limit32, destIndex32);
1984 if(destIndex32<start32) {
1985 start32+=segLength;
1986 limit32+=segLength;
1987 }
1988 rep->handleReplaceBetween(start32, limit32, UnicodeString());
1989 } else {
1990 // copy
1991 rep->copy(start32, limit32, destIndex32);
1992 }
1993
1994 // If the change to the text touched the region in the chunk buffer,
1995 // invalidate the buffer.
1996 int32_t firstAffectedIndex = destIndex32;
1997 if (move && start32<firstAffectedIndex) {
1998 firstAffectedIndex = start32;
1999 }
2000 if (firstAffectedIndex < ut->chunkNativeLimit) {
2001 // changes may have affected range covered by the chunk
2002 invalidateChunk(ut);
2003 }
2004
2005 // Put iteration position at the newly inserted (moved) block,
2006 int32_t nativeIterIndex = destIndex32 + limit32 - start32;
2007 if (move && destIndex32>start32) {
2008 // moved a block of text towards the end of the string.
2009 nativeIterIndex = destIndex32;
2010 }
2011
2012 // Set position, reload chunk if needed.
2013 repTextAccess(ut, nativeIterIndex, TRUE);
2014}
2015
2016static const struct UTextFuncs repFuncs =
2017{
2018 sizeof(UTextFuncs),
2019 0, 0, 0, // Reserved alignment padding
2020 repTextClone,
2021 repTextLength,
2022 repTextAccess,
2023 repTextExtract,
2024 repTextReplace,
2025 repTextCopy,
2026 NULL, // MapOffsetToNative,
2027 NULL, // MapIndexToUTF16,
2028 repTextClose,
2029 NULL, // spare 1
2030 NULL, // spare 2
2031 NULL // spare 3
2032};
2033
2034
2035U_CAPI UText * U_EXPORT2
2036utext_openReplaceable(UText *ut, Replaceable *rep, UErrorCode *status)
2037{
2038 if(U_FAILURE(*status)) {
2039 return NULL;
2040 }
2041 if(rep==NULL) {
2042 *status=U_ILLEGAL_ARGUMENT_ERROR;
2043 return NULL;
2044 }
2045 ut = utext_setup(ut, sizeof(ReplExtra), status);
2046 if(U_FAILURE(*status)) {
2047 return ut;
2048 }
2049
2050 ut->providerProperties = I32_FLAG(UTEXT_PROVIDER_WRITABLE);
2051 if(rep->hasMetaData()) {
2052 ut->providerProperties |=I32_FLAG(UTEXT_PROVIDER_HAS_META_DATA);
2053 }
2054
2055 ut->pFuncs = &repFuncs;
2056 ut->context = rep;
2057 return ut;
2058}
2059
2060U_CDECL_END
2061
2062
2063
2064
2065
2066
2067
2068
2069//------------------------------------------------------------------------------
2070//
2071// UText implementation for UnicodeString (read/write) and
2072// for const UnicodeString (read only)
2073// (same implementation, only the flags are different)
2074//
2075// Use of UText data members:
2076// context pointer to UnicodeString
2077// p pointer to UnicodeString IF this UText owns the string
2078// and it must be deleted on close(). NULL otherwise.
2079//
2080//------------------------------------------------------------------------------
2081
2082U_CDECL_BEGIN
2083
2084
2085static UText * U_CALLCONV
2086unistrTextClone(UText *dest, const UText *src, UBool deep, UErrorCode *status) {
2087 // First do a generic shallow clone. Does everything needed for the UText struct itself.
2088 dest = shallowTextClone(dest, src, status);
2089
2090 // For deep clones, make a copy of the UnicodeSring.
2091 // The copied UnicodeString storage is owned by the newly created UText clone.
2092 // A non-NULL pointer in UText.p is the signal to the close() function to delete
2093 // the UText.
2094 //
2095 if (deep && U_SUCCESS(*status)) {
2096 const UnicodeString *srcString = (const UnicodeString *)src->context;
2097 dest->context = new UnicodeString(*srcString);
2098 dest->providerProperties |= I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT);
2099
2100 // with deep clone, the copy is writable, even when the source is not.
2101 dest->providerProperties |= I32_FLAG(UTEXT_PROVIDER_WRITABLE);
2102 }
2103 return dest;
2104}
2105
2106static void U_CALLCONV
2107unistrTextClose(UText *ut) {
2108 // Most of the work of close is done by the generic UText framework close.
2109 // All that needs to be done here is delete the UnicodeString if the UText
2110 // owns it. This occurs if the UText was created by cloning.
2111 if (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT)) {
2112 UnicodeString *str = (UnicodeString *)ut->context;
2113 delete str;
2114 ut->context = NULL;
2115 }
2116}
2117
2118
2119static int64_t U_CALLCONV
2120unistrTextLength(UText *t) {
2121 return ((const UnicodeString *)t->context)->length();
2122}
2123
2124
2125static UBool U_CALLCONV
2126unistrTextAccess(UText *ut, int64_t index, UBool forward) {
2127 int32_t length = ut->chunkLength;
2128 ut->chunkOffset = pinIndex(index, length);
2129
2130 // Check whether request is at the start or end
2131 UBool retVal = (forward && index<length) || (!forward && index>0);
2132 return retVal;
2133}
2134
2135
2136
2137static int32_t U_CALLCONV
2138unistrTextExtract(UText *t,
2139 int64_t start, int64_t limit,
2140 UChar *dest, int32_t destCapacity,
2141 UErrorCode *pErrorCode) {
2142 const UnicodeString *us=(const UnicodeString *)t->context;
2143 int32_t length=us->length();
2144
2145 if(U_FAILURE(*pErrorCode)) {
2146 return 0;
2147 }
2148 if(destCapacity<0 || (dest==NULL && destCapacity>0)) {
2149 *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
2150 }
2151 if(start<0 || start>limit) {
2152 *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR;
2153 return 0;
2154 }
2155
2156 int32_t start32 = start<length ? us->getChar32Start((int32_t)start) : length;
2157 int32_t limit32 = limit<length ? us->getChar32Start((int32_t)limit) : length;
2158
2159 length=limit32-start32;
2160 if (destCapacity>0 && dest!=NULL) {
2161 int32_t trimmedLength = length;
2162 if(trimmedLength>destCapacity) {
2163 trimmedLength=destCapacity;
2164 }
2165 us->extract(start32, trimmedLength, dest);
2166 t->chunkOffset = start32+trimmedLength;
2167 } else {
2168 t->chunkOffset = start32;
2169 }
2170 u_terminateUChars(dest, destCapacity, length, pErrorCode);
2171 return length;
2172}
2173
2174static int32_t U_CALLCONV
2175unistrTextReplace(UText *ut,
2176 int64_t start, int64_t limit,
2177 const UChar *src, int32_t length,
2178 UErrorCode *pErrorCode) {
2179 UnicodeString *us=(UnicodeString *)ut->context;
2180 int32_t oldLength;
2181
2182 if(U_FAILURE(*pErrorCode)) {
2183 return 0;
2184 }
2185 if(src==NULL && length!=0) {
2186 *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
2187 }
2188 if(start>limit) {
2189 *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR;
2190 return 0;
2191 }
2192 oldLength=us->length();
2193 int32_t start32 = pinIndex(start, oldLength);
2194 int32_t limit32 = pinIndex(limit, oldLength);
2195 if (start32 < oldLength) {
2196 start32 = us->getChar32Start(start32);
2197 }
2198 if (limit32 < oldLength) {
2199 limit32 = us->getChar32Start(limit32);
2200 }
2201
2202 // replace
2203 us->replace(start32, limit32-start32, src, length);
2204 int32_t newLength = us->length();
2205
2206 // Update the chunk description.
2207 ut->chunkContents = us->getBuffer();
2208 ut->chunkLength = newLength;
2209 ut->chunkNativeLimit = newLength;
2210 ut->nativeIndexingLimit = newLength;
2211
2212 // Set iteration position to the point just following the newly inserted text.
2213 int32_t lengthDelta = newLength - oldLength;
2214 ut->chunkOffset = limit32 + lengthDelta;
2215
2216 return lengthDelta;
2217}
2218
2219static void U_CALLCONV
2220unistrTextCopy(UText *ut,
2221 int64_t start, int64_t limit,
2222 int64_t destIndex,
2223 UBool move,
2224 UErrorCode *pErrorCode) {
2225 UnicodeString *us=(UnicodeString *)ut->context;
2226 int32_t length=us->length();
2227
2228 if(U_FAILURE(*pErrorCode)) {
2229 return;
2230 }
2231 int32_t start32 = pinIndex(start, length);
2232 int32_t limit32 = pinIndex(limit, length);
2233 int32_t destIndex32 = pinIndex(destIndex, length);
2234
2235 if( start32>limit32 || (start32<destIndex32 && destIndex32<limit32)) {
2236 *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR;
2237 return;
2238 }
2239
2240 if(move) {
2241 // move: copy to destIndex, then remove original
2242 int32_t segLength=limit32-start32;
2243 us->copy(start32, limit32, destIndex32);
2244 if(destIndex32<start32) {
2245 start32+=segLength;
2246 }
2247 us->remove(start32, segLength);
2248 } else {
2249 // copy
2250 us->copy(start32, limit32, destIndex32);
2251 }
2252
2253 // update chunk description, set iteration position.
2254 ut->chunkContents = us->getBuffer();
2255 if (move==FALSE) {
2256 // copy operation, string length grows
2257 ut->chunkLength += limit32-start32;
2258 ut->chunkNativeLimit = ut->chunkLength;
2259 ut->nativeIndexingLimit = ut->chunkLength;
2260 }
2261
2262 // Iteration position to end of the newly inserted text.
2263 ut->chunkOffset = destIndex32+limit32-start32;
2264 if (move && destIndex32>start32) {
2265 ut->chunkOffset = destIndex32;
2266 }
2267
2268}
2269
2270static const struct UTextFuncs unistrFuncs =
2271{
2272 sizeof(UTextFuncs),
2273 0, 0, 0, // Reserved alignment padding
2274 unistrTextClone,
2275 unistrTextLength,
2276 unistrTextAccess,
2277 unistrTextExtract,
2278 unistrTextReplace,
2279 unistrTextCopy,
2280 NULL, // MapOffsetToNative,
2281 NULL, // MapIndexToUTF16,
2282 unistrTextClose,
2283 NULL, // spare 1
2284 NULL, // spare 2
2285 NULL // spare 3
2286};
2287
2288
2289
2290U_CDECL_END
2291
2292
2293U_CAPI UText * U_EXPORT2
2294utext_openUnicodeString(UText *ut, UnicodeString *s, UErrorCode *status) {
2295 ut = utext_openConstUnicodeString(ut, s, status);
2296 if (U_SUCCESS(*status)) {
2297 ut->providerProperties |= I32_FLAG(UTEXT_PROVIDER_WRITABLE);
2298 }
2299 return ut;
2300}
2301
2302
2303
2304U_CAPI UText * U_EXPORT2
2305utext_openConstUnicodeString(UText *ut, const UnicodeString *s, UErrorCode *status) {
2306 if (U_SUCCESS(*status) && s->isBogus()) {
2307 // The UnicodeString is bogus, but we still need to detach the UText
2308 // from whatever it was hooked to before, if anything.
2309 utext_openUChars(ut, NULL, 0, status);
2310 *status = U_ILLEGAL_ARGUMENT_ERROR;
2311 return ut;
2312 }
2313 ut = utext_setup(ut, 0, status);
2314 // note: use the standard (writable) function table for UnicodeString.
2315 // The flag settings disable writing, so having the functions in
2316 // the table is harmless.
2317 if (U_SUCCESS(*status)) {
2318 ut->pFuncs = &unistrFuncs;
2319 ut->context = s;
2320 ut->providerProperties = I32_FLAG(UTEXT_PROVIDER_STABLE_CHUNKS);
2321 ut->chunkContents = s->getBuffer();
2322 ut->chunkLength = s->length();
2323 ut->chunkNativeStart = 0;
2324 ut->chunkNativeLimit = ut->chunkLength;
2325 ut->nativeIndexingLimit = ut->chunkLength;
2326 }
2327 return ut;
2328}
2329
2330//------------------------------------------------------------------------------
2331//
2332// UText implementation for const UChar * strings
2333//
2334// Use of UText data members:
2335// context pointer to UnicodeString
2336// a length. -1 if not yet known.
2337//
2338// TODO: support 64 bit lengths.
2339//
2340//------------------------------------------------------------------------------
2341
2342U_CDECL_BEGIN
2343
2344
2345static UText * U_CALLCONV
2346ucstrTextClone(UText *dest, const UText * src, UBool deep, UErrorCode * status) {
2347 // First do a generic shallow clone.
2348 dest = shallowTextClone(dest, src, status);
2349
2350 // For deep clones, make a copy of the string.
2351 // The copied storage is owned by the newly created clone.
2352 // A non-NULL pointer in UText.p is the signal to the close() function to delete
2353 // it.
2354 //
2355 if (deep && U_SUCCESS(*status)) {
2356 U_ASSERT(utext_nativeLength(dest) < INT32_MAX);
2357 int32_t len = (int32_t)utext_nativeLength(dest);
2358
2359 // The cloned string IS going to be NUL terminated, whether or not the original was.
2360 const UChar *srcStr = (const UChar *)src->context;
2361 UChar *copyStr = (UChar *)uprv_malloc((len+1) * sizeof(UChar));
2362 if (copyStr == NULL) {
2363 *status = U_MEMORY_ALLOCATION_ERROR;
2364 } else {
2365 int64_t i;
2366 for (i=0; i<len; i++) {
2367 copyStr[i] = srcStr[i];
2368 }
2369 copyStr[len] = 0;
2370 dest->context = copyStr;
2371 dest->providerProperties |= I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT);
2372 }
2373 }
2374 return dest;
2375}
2376
2377
2378static void U_CALLCONV
2379ucstrTextClose(UText *ut) {
2380 // Most of the work of close is done by the generic UText framework close.
2381 // All that needs to be done here is delete the string if the UText
2382 // owns it. This occurs if the UText was created by cloning.
2383 if (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT)) {
2384 UChar *s = (UChar *)ut->context;
2385 uprv_free(s);
2386 ut->context = NULL;
2387 }
2388}
2389
2390
2391
2392static int64_t U_CALLCONV
2393ucstrTextLength(UText *ut) {
2394 if (ut->a < 0) {
2395 // null terminated, we don't yet know the length. Scan for it.
2396 // Access is not convenient for doing this
2397 // because the current interation postion can't be changed.
2398 const UChar *str = (const UChar *)ut->context;
2399 for (;;) {
2400 if (str[ut->chunkNativeLimit] == 0) {
2401 break;
2402 }
2403 ut->chunkNativeLimit++;
2404 }
2405 ut->a = ut->chunkNativeLimit;
2406 ut->chunkLength = (int32_t)ut->chunkNativeLimit;
2407 ut->nativeIndexingLimit = ut->chunkLength;
2408 ut->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
2409 }
2410 return ut->a;
2411}
2412
2413
2414static UBool U_CALLCONV
2415ucstrTextAccess(UText *ut, int64_t index, UBool forward) {
2416 const UChar *str = (const UChar *)ut->context;
2417
2418 // pin the requested index to the bounds of the string,
2419 // and set current iteration position.
2420 if (index<0) {
2421 index = 0;
2422 } else if (index < ut->chunkNativeLimit) {
2423 // The request data is within the chunk as it is known so far.
2424 // Put index on a code point boundary.
2425 U16_SET_CP_START(str, 0, index);
2426 } else if (ut->a >= 0) {
2427 // We know the length of this string, and the user is requesting something
2428 // at or beyond the length. Pin the requested index to the length.
2429 index = ut->a;
2430 } else {
2431 // Null terminated string, length not yet known, and the requested index
2432 // is beyond where we have scanned so far.
2433 // Scan to 32 UChars beyond the requested index. The strategy here is
2434 // to avoid fully scanning a long string when the caller only wants to
2435 // see a few characters at its beginning.
2436 int32_t scanLimit = (int32_t)index + 32;
2437 if ((index + 32)>INT32_MAX || (index + 32)<0 ) { // note: int64 expression
2438 scanLimit = INT32_MAX;
2439 }
2440
2441 int32_t chunkLimit = (int32_t)ut->chunkNativeLimit;
2442 for (; chunkLimit<scanLimit; chunkLimit++) {
2443 if (str[chunkLimit] == 0) {
2444 // We found the end of the string. Remember it, pin the requested index to it,
2445 // and bail out of here.
2446 ut->a = chunkLimit;
2447 ut->chunkLength = chunkLimit;
2448 ut->nativeIndexingLimit = chunkLimit;
2449 if (index >= chunkLimit) {
2450 index = chunkLimit;
2451 } else {
2452 U16_SET_CP_START(str, 0, index);
2453 }
2454
2455 ut->chunkNativeLimit = chunkLimit;
2456 ut->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
2457 goto breakout;
2458 }
2459 }
2460 // We scanned through the next batch of UChars without finding the end.
2461 U16_SET_CP_START(str, 0, index);
2462 if (chunkLimit == INT32_MAX) {
2463 // Scanned to the limit of a 32 bit length.
2464 // Forceably trim the overlength string back so length fits in int32
2465 // TODO: add support for 64 bit strings.
2466 ut->a = chunkLimit;
2467 ut->chunkLength = chunkLimit;
2468 ut->nativeIndexingLimit = chunkLimit;
2469 if (index > chunkLimit) {
2470 index = chunkLimit;
2471 }
2472 ut->chunkNativeLimit = chunkLimit;
2473 ut->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
2474 } else {
2475 // The endpoint of a chunk must not be left in the middle of a surrogate pair.
2476 // If the current end is on a lead surrogate, back the end up by one.
2477 // It doesn't matter if the end char happens to be an unpaired surrogate,
2478 // and it's simpler not to worry about it.
2479 if (U16_IS_LEAD(str[chunkLimit-1])) {
2480 --chunkLimit;
2481 }
2482 // Null-terminated chunk with end still unknown.
2483 // Update the chunk length to reflect what has been scanned thus far.
2484 // That the full length is still unknown is (still) flagged by
2485 // ut->a being < 0.
2486 ut->chunkNativeLimit = chunkLimit;
2487 ut->nativeIndexingLimit = chunkLimit;
2488 ut->chunkLength = chunkLimit;
2489 }
2490
2491 }
2492breakout:
2493 U_ASSERT(index<=INT32_MAX);
2494 ut->chunkOffset = (int32_t)index;
2495
2496 // Check whether request is at the start or end
2497 UBool retVal = (forward && index<ut->chunkNativeLimit) || (!forward && index>0);
2498 return retVal;
2499}
2500
2501
2502
2503static int32_t U_CALLCONV
2504ucstrTextExtract(UText *ut,
2505 int64_t start, int64_t limit,
2506 UChar *dest, int32_t destCapacity,
2507 UErrorCode *pErrorCode)
2508{
2509 if(U_FAILURE(*pErrorCode)) {
2510 return 0;
2511 }
2512 if(destCapacity<0 || (dest==NULL && destCapacity>0) || start>limit) {
2513 *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
2514 return 0;
2515 }
2516
2517 //const UChar *s=(const UChar *)ut->context;
2518 int32_t si, di;
2519
2520 int32_t start32;
2521 int32_t limit32;
2522
2523 // Access the start. Does two things we need:
2524 // Pins 'start' to the length of the string, if it came in out-of-bounds.
2525 // Snaps 'start' to the beginning of a code point.
2526 ucstrTextAccess(ut, start, TRUE);
2527 const UChar *s=ut->chunkContents;
2528 start32 = ut->chunkOffset;
2529
2530 int32_t strLength=(int32_t)ut->a;
2531 if (strLength >= 0) {
2532 limit32 = pinIndex(limit, strLength);
2533 } else {
2534 limit32 = pinIndex(limit, INT32_MAX);
2535 }
2536 di = 0;
2537 for (si=start32; si<limit32; si++) {
2538 if (strLength<0 && s[si]==0) {
2539 // Just hit the end of a null-terminated string.
2540 ut->a = si; // set string length for this UText
2541 ut->chunkNativeLimit = si;
2542 ut->chunkLength = si;
2543 ut->nativeIndexingLimit = si;
2544 strLength = si;
2545 limit32 = si;
2546 break;
2547 }
2548 U_ASSERT(di>=0); /* to ensure di never exceeds INT32_MAX, which must not happen logically */
2549 if (di<destCapacity) {
2550 // only store if there is space.
2551 dest[di] = s[si];
2552 } else {
2553 if (strLength>=0) {
2554 // We have filled the destination buffer, and the string length is known.
2555 // Cut the loop short. There is no need to scan string termination.
2556 di = limit32 - start32;
2557 si = limit32;
2558 break;
2559 }
2560 }
2561 di++;
2562 }
2563
2564 // If the limit index points to a lead surrogate of a pair,
2565 // add the corresponding trail surrogate to the destination.
2566 if (si>0 && U16_IS_LEAD(s[si-1]) &&
2567 ((si<strLength || strLength<0) && U16_IS_TRAIL(s[si])))
2568 {
2569 if (di<destCapacity) {
2570 // store only if there is space in the output buffer.
2571 dest[di++] = s[si];
2572 }
2573 si++;
2574 }
2575
2576 // Put iteration position at the point just following the extracted text
2577 if (si <= ut->chunkNativeLimit) {
2578 ut->chunkOffset = si;
2579 } else {
2580 ucstrTextAccess(ut, si, TRUE);
2581 }
2582
2583 // Add a terminating NUL if space in the buffer permits,
2584 // and set the error status as required.
2585 u_terminateUChars(dest, destCapacity, di, pErrorCode);
2586 return di;
2587}
2588
2589static const struct UTextFuncs ucstrFuncs =
2590{
2591 sizeof(UTextFuncs),
2592 0, 0, 0, // Reserved alignment padding
2593 ucstrTextClone,
2594 ucstrTextLength,
2595 ucstrTextAccess,
2596 ucstrTextExtract,
2597 NULL, // Replace
2598 NULL, // Copy
2599 NULL, // MapOffsetToNative,
2600 NULL, // MapIndexToUTF16,
2601 ucstrTextClose,
2602 NULL, // spare 1
2603 NULL, // spare 2
2604 NULL, // spare 3
2605};
2606
2607U_CDECL_END
2608
2609static const UChar gEmptyUString[] = {0};
2610
2611U_CAPI UText * U_EXPORT2
2612utext_openUChars(UText *ut, const UChar *s, int64_t length, UErrorCode *status) {
2613 if (U_FAILURE(*status)) {
2614 return NULL;
2615 }
2616 if(s==NULL && length==0) {
2617 s = gEmptyUString;
2618 }
2619 if (s==NULL || length < -1 || length>INT32_MAX) {
2620 *status = U_ILLEGAL_ARGUMENT_ERROR;
2621 return NULL;
2622 }
2623 ut = utext_setup(ut, 0, status);
2624 if (U_SUCCESS(*status)) {
2625 ut->pFuncs = &ucstrFuncs;
2626 ut->context = s;
2627 ut->providerProperties = I32_FLAG(UTEXT_PROVIDER_STABLE_CHUNKS);
2628 if (length==-1) {
2629 ut->providerProperties |= I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
2630 }
2631 ut->a = length;
2632 ut->chunkContents = s;
2633 ut->chunkNativeStart = 0;
2634 ut->chunkNativeLimit = length>=0? length : 0;
2635 ut->chunkLength = (int32_t)ut->chunkNativeLimit;
2636 ut->chunkOffset = 0;
2637 ut->nativeIndexingLimit = ut->chunkLength;
2638 }
2639 return ut;
2640}
2641
2642
2643//------------------------------------------------------------------------------
2644//
2645// UText implementation for text from ICU CharacterIterators
2646//
2647// Use of UText data members:
2648// context pointer to the CharacterIterator
2649// a length of the full text.
2650// p pointer to buffer 1
2651// b start index of local buffer 1 contents
2652// q pointer to buffer 2
2653// c start index of local buffer 2 contents
2654// r pointer to the character iterator if the UText owns it.
2655// Null otherwise.
2656//
2657//------------------------------------------------------------------------------
2658#define CIBufSize 16
2659
2660U_CDECL_BEGIN
2661static void U_CALLCONV
2662charIterTextClose(UText *ut) {
2663 // Most of the work of close is done by the generic UText framework close.
2664 // All that needs to be done here is delete the CharacterIterator if the UText
2665 // owns it. This occurs if the UText was created by cloning.
2666 CharacterIterator *ci = (CharacterIterator *)ut->r;
2667 delete ci;
2668 ut->r = NULL;
2669}
2670
2671static int64_t U_CALLCONV
2672charIterTextLength(UText *ut) {
2673 return (int32_t)ut->a;
2674}
2675
2676static UBool U_CALLCONV
2677charIterTextAccess(UText *ut, int64_t index, UBool forward) {
2678 CharacterIterator *ci = (CharacterIterator *)ut->context;
2679
2680 int32_t clippedIndex = (int32_t)index;
2681 if (clippedIndex<0) {
2682 clippedIndex=0;
2683 } else if (clippedIndex>=ut->a) {
2684 clippedIndex=(int32_t)ut->a;
2685 }
2686 int32_t neededIndex = clippedIndex;
2687 if (!forward && neededIndex>0) {
2688 // reverse iteration, want the position just before what was asked for.
2689 neededIndex--;
2690 } else if (forward && neededIndex==ut->a && neededIndex>0) {
2691 // Forward iteration, don't ask for something past the end of the text.
2692 neededIndex--;
2693 }
2694
2695 // Find the native index of the start of the buffer containing what we want.
2696 neededIndex -= neededIndex % CIBufSize;
2697
2698 UChar *buf = NULL;
2699 UBool needChunkSetup = TRUE;
2700 int i;
2701 if (ut->chunkNativeStart == neededIndex) {
2702 // The buffer we want is already the current chunk.
2703 needChunkSetup = FALSE;
2704 } else if (ut->b == neededIndex) {
2705 // The first buffer (buffer p) has what we need.
2706 buf = (UChar *)ut->p;
2707 } else if (ut->c == neededIndex) {
2708 // The second buffer (buffer q) has what we need.
2709 buf = (UChar *)ut->q;
2710 } else {
2711 // Neither buffer already has what we need.
2712 // Load new data from the character iterator.
2713 // Use the buf that is not the current buffer.
2714 buf = (UChar *)ut->p;
2715 if (ut->p == ut->chunkContents) {
2716 buf = (UChar *)ut->q;
2717 }
2718 ci->setIndex(neededIndex);
2719 for (i=0; i<CIBufSize; i++) {
2720 buf[i] = ci->nextPostInc();
2721 if (i+neededIndex > ut->a) {
2722 break;
2723 }
2724 }
2725 }
2726
2727 // We have a buffer with the data we need.
2728 // Set it up as the current chunk, if it wasn't already.
2729 if (needChunkSetup) {
2730 ut->chunkContents = buf;
2731 ut->chunkLength = CIBufSize;
2732 ut->chunkNativeStart = neededIndex;
2733 ut->chunkNativeLimit = neededIndex + CIBufSize;
2734 if (ut->chunkNativeLimit > ut->a) {
2735 ut->chunkNativeLimit = ut->a;
2736 ut->chunkLength = (int32_t)(ut->chunkNativeLimit)-(int32_t)(ut->chunkNativeStart);
2737 }
2738 ut->nativeIndexingLimit = ut->chunkLength;
2739 U_ASSERT(ut->chunkOffset>=0 && ut->chunkOffset<=CIBufSize);
2740 }
2741 ut->chunkOffset = clippedIndex - (int32_t)ut->chunkNativeStart;
2742 UBool success = (forward? ut->chunkOffset<ut->chunkLength : ut->chunkOffset>0);
2743 return success;
2744}
2745
2746static UText * U_CALLCONV
2747charIterTextClone(UText *dest, const UText *src, UBool deep, UErrorCode * status) {
2748 if (U_FAILURE(*status)) {
2749 return NULL;
2750 }
2751
2752 if (deep) {
2753 // There is no CharacterIterator API for cloning the underlying text storage.
2754 *status = U_UNSUPPORTED_ERROR;
2755 return NULL;
2756 } else {
2757 CharacterIterator *srcCI =(CharacterIterator *)src->context;
2758 srcCI = srcCI->clone();
2759 dest = utext_openCharacterIterator(dest, srcCI, status);
2760 if (U_FAILURE(*status)) {
2761 return dest;
2762 }
2763 // cast off const on getNativeIndex.
2764 // For CharacterIterator based UTexts, this is safe, the operation is const.
2765 int64_t ix = utext_getNativeIndex((UText *)src);
2766 utext_setNativeIndex(dest, ix);
2767 dest->r = srcCI; // flags that this UText owns the CharacterIterator
2768 }
2769 return dest;
2770}
2771
2772static int32_t U_CALLCONV
2773charIterTextExtract(UText *ut,
2774 int64_t start, int64_t limit,
2775 UChar *dest, int32_t destCapacity,
2776 UErrorCode *status)
2777{
2778 if(U_FAILURE(*status)) {
2779 return 0;
2780 }
2781 if(destCapacity<0 || (dest==NULL && destCapacity>0) || start>limit) {
2782 *status=U_ILLEGAL_ARGUMENT_ERROR;
2783 return 0;
2784 }
2785 int32_t length = (int32_t)ut->a;
2786 int32_t start32 = pinIndex(start, length);
2787 int32_t limit32 = pinIndex(limit, length);
2788 int32_t desti = 0;
2789 int32_t srci;
2790 int32_t copyLimit;
2791
2792 CharacterIterator *ci = (CharacterIterator *)ut->context;
2793 ci->setIndex32(start32); // Moves ix to lead of surrogate pair, if needed.
2794 srci = ci->getIndex();
2795 copyLimit = srci;
2796 while (srci<limit32) {
2797 UChar32 c = ci->next32PostInc();
2798 int32_t len = U16_LENGTH(c);
2799 U_ASSERT(desti+len>0); /* to ensure desti+len never exceeds MAX_INT32, which must not happen logically */
2800 if (desti+len <= destCapacity) {
2801 U16_APPEND_UNSAFE(dest, desti, c);
2802 copyLimit = srci+len;
2803 } else {
2804 desti += len;
2805 *status = U_BUFFER_OVERFLOW_ERROR;
2806 }
2807 srci += len;
2808 }
2809
2810 charIterTextAccess(ut, copyLimit, TRUE);
2811
2812 u_terminateUChars(dest, destCapacity, desti, status);
2813 return desti;
2814}
2815
2816static const struct UTextFuncs charIterFuncs =
2817{
2818 sizeof(UTextFuncs),
2819 0, 0, 0, // Reserved alignment padding
2820 charIterTextClone,
2821 charIterTextLength,
2822 charIterTextAccess,
2823 charIterTextExtract,
2824 NULL, // Replace
2825 NULL, // Copy
2826 NULL, // MapOffsetToNative,
2827 NULL, // MapIndexToUTF16,
2828 charIterTextClose,
2829 NULL, // spare 1
2830 NULL, // spare 2
2831 NULL // spare 3
2832};
2833U_CDECL_END
2834
2835
2836U_CAPI UText * U_EXPORT2
2837utext_openCharacterIterator(UText *ut, CharacterIterator *ci, UErrorCode *status) {
2838 if (U_FAILURE(*status)) {
2839 return NULL;
2840 }
2841
2842 if (ci->startIndex() > 0) {
2843 // No support for CharacterIterators that do not start indexing from zero.
2844 *status = U_UNSUPPORTED_ERROR;
2845 return NULL;
2846 }
2847
2848 // Extra space in UText for 2 buffers of CIBufSize UChars each.
2849 int32_t extraSpace = 2 * CIBufSize * sizeof(UChar);
2850 ut = utext_setup(ut, extraSpace, status);
2851 if (U_SUCCESS(*status)) {
2852 ut->pFuncs = &charIterFuncs;
2853 ut->context = ci;
2854 ut->providerProperties = 0;
2855 ut->a = ci->endIndex(); // Length of text
2856 ut->p = ut->pExtra; // First buffer
2857 ut->b = -1; // Native index of first buffer contents
2858 ut->q = (UChar*)ut->pExtra+CIBufSize; // Second buffer
2859 ut->c = -1; // Native index of second buffer contents
2860
2861 // Initialize current chunk contents to be empty.
2862 // First access will fault something in.
2863 // Note: The initial nativeStart and chunkOffset must sum to zero
2864 // so that getNativeIndex() will correctly compute to zero
2865 // if no call to Access() has ever been made. They can't be both
2866 // zero without Access() thinking that the chunk is valid.
2867 ut->chunkContents = (UChar *)ut->p;
2868 ut->chunkNativeStart = -1;
2869 ut->chunkOffset = 1;
2870 ut->chunkNativeLimit = 0;
2871 ut->chunkLength = 0;
2872 ut->nativeIndexingLimit = ut->chunkOffset; // enables native indexing
2873 }
2874 return ut;
2875}
2876