intl/icu/source/common/utext.cpp@129ffea94266
intl/icu/source/common/utext.cpp
Sat, 03 Jan 2015 20:18:00 +0100
- author
- Michael Schloh von Bennewitz <michael@schloh.com>
- date
- Sat, 03 Jan 2015 20:18:00 +0100
- branch
- TOR_BUG_3246
- changeset 7
- 129ffea94266
- permissions
- -rw-r--r--
Conditionally enable double key logic according to:
private browsing mode or privacy.thirdparty.isolate preference and
implement in GetCookieStringCommon and FindCookie where it counts...
With some reservations of how to convince FindCookie users to test
condition and pass a nullptr when disabling double key logic.
1 /*
2 *******************************************************************************
3 *
4 * Copyright (C) 2005-2012, International Business Machines
5 * Corporation and others. All Rights Reserved.
6 *
7 *******************************************************************************
8 * file name: utext.cpp
9 * encoding: US-ASCII
10 * tab size: 8 (not used)
11 * indentation:4
12 *
13 * created on: 2005apr12
14 * created by: Markus W. Scherer
15 */
17 #include "unicode/utypes.h"
18 #include "unicode/ustring.h"
19 #include "unicode/unistr.h"
20 #include "unicode/chariter.h"
21 #include "unicode/utext.h"
22 #include "unicode/utf.h"
23 #include "unicode/utf8.h"
24 #include "unicode/utf16.h"
25 #include "ustr_imp.h"
26 #include "cmemory.h"
27 #include "cstring.h"
28 #include "uassert.h"
29 #include "putilimp.h"
31 U_NAMESPACE_USE
33 #define I32_FLAG(bitIndex) ((int32_t)1<<(bitIndex))
36 static UBool
37 utext_access(UText *ut, int64_t index, UBool forward) {
38 return ut->pFuncs->access(ut, index, forward);
39 }
43 U_CAPI UBool U_EXPORT2
44 utext_moveIndex32(UText *ut, int32_t delta) {
45 UChar32 c;
46 if (delta > 0) {
47 do {
48 if(ut->chunkOffset>=ut->chunkLength && !utext_access(ut, ut->chunkNativeLimit, TRUE)) {
49 return FALSE;
50 }
51 c = ut->chunkContents[ut->chunkOffset];
52 if (U16_IS_SURROGATE(c)) {
53 c = utext_next32(ut);
54 if (c == U_SENTINEL) {
55 return FALSE;
56 }
57 } else {
58 ut->chunkOffset++;
59 }
60 } while(--delta>0);
62 } else if (delta<0) {
63 do {
64 if(ut->chunkOffset<=0 && !utext_access(ut, ut->chunkNativeStart, FALSE)) {
65 return FALSE;
66 }
67 c = ut->chunkContents[ut->chunkOffset-1];
68 if (U16_IS_SURROGATE(c)) {
69 c = utext_previous32(ut);
70 if (c == U_SENTINEL) {
71 return FALSE;
72 }
73 } else {
74 ut->chunkOffset--;
75 }
76 } while(++delta<0);
77 }
79 return TRUE;
80 }
83 U_CAPI int64_t U_EXPORT2
84 utext_nativeLength(UText *ut) {
85 return ut->pFuncs->nativeLength(ut);
86 }
89 U_CAPI UBool U_EXPORT2
90 utext_isLengthExpensive(const UText *ut) {
91 UBool r = (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE)) != 0;
92 return r;
93 }
96 U_CAPI int64_t U_EXPORT2
97 utext_getNativeIndex(const UText *ut) {
98 if(ut->chunkOffset <= ut->nativeIndexingLimit) {
99 return ut->chunkNativeStart+ut->chunkOffset;
100 } else {
101 return ut->pFuncs->mapOffsetToNative(ut);
102 }
103 }
106 U_CAPI void U_EXPORT2
107 utext_setNativeIndex(UText *ut, int64_t index) {
108 if(index<ut->chunkNativeStart || index>=ut->chunkNativeLimit) {
109 // The desired position is outside of the current chunk.
110 // Access the new position. Assume a forward iteration from here,
111 // which will also be optimimum for a single random access.
112 // Reverse iterations may suffer slightly.
113 ut->pFuncs->access(ut, index, TRUE);
114 } else if((int32_t)(index - ut->chunkNativeStart) <= ut->nativeIndexingLimit) {
115 // utf-16 indexing.
116 ut->chunkOffset=(int32_t)(index-ut->chunkNativeStart);
117 } else {
118 ut->chunkOffset=ut->pFuncs->mapNativeIndexToUTF16(ut, index);
119 }
120 // The convention is that the index must always be on a code point boundary.
121 // Adjust the index position if it is in the middle of a surrogate pair.
122 if (ut->chunkOffset<ut->chunkLength) {
123 UChar c= ut->chunkContents[ut->chunkOffset];
124 if (U16_IS_TRAIL(c)) {
125 if (ut->chunkOffset==0) {
126 ut->pFuncs->access(ut, ut->chunkNativeStart, FALSE);
127 }
128 if (ut->chunkOffset>0) {
129 UChar lead = ut->chunkContents[ut->chunkOffset-1];
130 if (U16_IS_LEAD(lead)) {
131 ut->chunkOffset--;
132 }
133 }
134 }
135 }
136 }
140 U_CAPI int64_t U_EXPORT2
141 utext_getPreviousNativeIndex(UText *ut) {
142 //
143 // Fast-path the common case.
144 // Common means current position is not at the beginning of a chunk
145 // and the preceding character is not supplementary.
146 //
147 int32_t i = ut->chunkOffset - 1;
148 int64_t result;
149 if (i >= 0) {
150 UChar c = ut->chunkContents[i];
151 if (U16_IS_TRAIL(c) == FALSE) {
152 if (i <= ut->nativeIndexingLimit) {
153 result = ut->chunkNativeStart + i;
154 } else {
155 ut->chunkOffset = i;
156 result = ut->pFuncs->mapOffsetToNative(ut);
157 ut->chunkOffset++;
158 }
159 return result;
160 }
161 }
163 // If at the start of text, simply return 0.
164 if (ut->chunkOffset==0 && ut->chunkNativeStart==0) {
165 return 0;
166 }
168 // Harder, less common cases. We are at a chunk boundary, or on a surrogate.
169 // Keep it simple, use other functions to handle the edges.
170 //
171 utext_previous32(ut);
172 result = UTEXT_GETNATIVEINDEX(ut);
173 utext_next32(ut);
174 return result;
175 }
178 //
179 // utext_current32. Get the UChar32 at the current position.
180 // UText iteration position is always on a code point boundary,
181 // never on the trail half of a surrogate pair.
182 //
183 U_CAPI UChar32 U_EXPORT2
184 utext_current32(UText *ut) {
185 UChar32 c;
186 if (ut->chunkOffset==ut->chunkLength) {
187 // Current position is just off the end of the chunk.
188 if (ut->pFuncs->access(ut, ut->chunkNativeLimit, TRUE) == FALSE) {
189 // Off the end of the text.
190 return U_SENTINEL;
191 }
192 }
194 c = ut->chunkContents[ut->chunkOffset];
195 if (U16_IS_LEAD(c) == FALSE) {
196 // Normal, non-supplementary case.
197 return c;
198 }
200 //
201 // Possible supplementary char.
202 //
203 UChar32 trail = 0;
204 UChar32 supplementaryC = c;
205 if ((ut->chunkOffset+1) < ut->chunkLength) {
206 // The trail surrogate is in the same chunk.
207 trail = ut->chunkContents[ut->chunkOffset+1];
208 } else {
209 // The trail surrogate is in a different chunk.
210 // Because we must maintain the iteration position, we need to switch forward
211 // into the new chunk, get the trail surrogate, then revert the chunk back to the
212 // original one.
213 // An edge case to be careful of: the entire text may end with an unpaired
214 // leading surrogate. The attempt to access the trail will fail, but
215 // the original position before the unpaired lead still needs to be restored.
216 int64_t nativePosition = ut->chunkNativeLimit;
217 int32_t originalOffset = ut->chunkOffset;
218 if (ut->pFuncs->access(ut, nativePosition, TRUE)) {
219 trail = ut->chunkContents[ut->chunkOffset];
220 }
221 UBool r = ut->pFuncs->access(ut, nativePosition, FALSE); // reverse iteration flag loads preceding chunk
222 U_ASSERT(r==TRUE);
223 ut->chunkOffset = originalOffset;
224 if(!r) {
225 return U_SENTINEL;
226 }
227 }
229 if (U16_IS_TRAIL(trail)) {
230 supplementaryC = U16_GET_SUPPLEMENTARY(c, trail);
231 }
232 return supplementaryC;
234 }
237 U_CAPI UChar32 U_EXPORT2
238 utext_char32At(UText *ut, int64_t nativeIndex) {
239 UChar32 c = U_SENTINEL;
241 // Fast path the common case.
242 if (nativeIndex>=ut->chunkNativeStart && nativeIndex < ut->chunkNativeStart + ut->nativeIndexingLimit) {
243 ut->chunkOffset = (int32_t)(nativeIndex - ut->chunkNativeStart);
244 c = ut->chunkContents[ut->chunkOffset];
245 if (U16_IS_SURROGATE(c) == FALSE) {
246 return c;
247 }
248 }
251 utext_setNativeIndex(ut, nativeIndex);
252 if (nativeIndex>=ut->chunkNativeStart && ut->chunkOffset<ut->chunkLength) {
253 c = ut->chunkContents[ut->chunkOffset];
254 if (U16_IS_SURROGATE(c)) {
255 // For surrogates, let current32() deal with the complications
256 // of supplementaries that may span chunk boundaries.
257 c = utext_current32(ut);
258 }
259 }
260 return c;
261 }
264 U_CAPI UChar32 U_EXPORT2
265 utext_next32(UText *ut) {
266 UChar32 c;
268 if (ut->chunkOffset >= ut->chunkLength) {
269 if (ut->pFuncs->access(ut, ut->chunkNativeLimit, TRUE) == FALSE) {
270 return U_SENTINEL;
271 }
272 }
274 c = ut->chunkContents[ut->chunkOffset++];
275 if (U16_IS_LEAD(c) == FALSE) {
276 // Normal case, not supplementary.
277 // (A trail surrogate seen here is just returned as is, as a surrogate value.
278 // It cannot be part of a pair.)
279 return c;
280 }
282 if (ut->chunkOffset >= ut->chunkLength) {
283 if (ut->pFuncs->access(ut, ut->chunkNativeLimit, TRUE) == FALSE) {
284 // c is an unpaired lead surrogate at the end of the text.
285 // return it as it is.
286 return c;
287 }
288 }
289 UChar32 trail = ut->chunkContents[ut->chunkOffset];
290 if (U16_IS_TRAIL(trail) == FALSE) {
291 // c was an unpaired lead surrogate, not at the end of the text.
292 // return it as it is (unpaired). Iteration position is on the
293 // following character, possibly in the next chunk, where the
294 // trail surrogate would have been if it had existed.
295 return c;
296 }
298 UChar32 supplementary = U16_GET_SUPPLEMENTARY(c, trail);
299 ut->chunkOffset++; // move iteration position over the trail surrogate.
300 return supplementary;
301 }
304 U_CAPI UChar32 U_EXPORT2
305 utext_previous32(UText *ut) {
306 UChar32 c;
308 if (ut->chunkOffset <= 0) {
309 if (ut->pFuncs->access(ut, ut->chunkNativeStart, FALSE) == FALSE) {
310 return U_SENTINEL;
311 }
312 }
313 ut->chunkOffset--;
314 c = ut->chunkContents[ut->chunkOffset];
315 if (U16_IS_TRAIL(c) == FALSE) {
316 // Normal case, not supplementary.
317 // (A lead surrogate seen here is just returned as is, as a surrogate value.
318 // It cannot be part of a pair.)
319 return c;
320 }
322 if (ut->chunkOffset <= 0) {
323 if (ut->pFuncs->access(ut, ut->chunkNativeStart, FALSE) == FALSE) {
324 // c is an unpaired trail surrogate at the start of the text.
325 // return it as it is.
326 return c;
327 }
328 }
330 UChar32 lead = ut->chunkContents[ut->chunkOffset-1];
331 if (U16_IS_LEAD(lead) == FALSE) {
332 // c was an unpaired trail surrogate, not at the end of the text.
333 // return it as it is (unpaired). Iteration position is at c
334 return c;
335 }
337 UChar32 supplementary = U16_GET_SUPPLEMENTARY(lead, c);
338 ut->chunkOffset--; // move iteration position over the lead surrogate.
339 return supplementary;
340 }
344 U_CAPI UChar32 U_EXPORT2
345 utext_next32From(UText *ut, int64_t index) {
346 UChar32 c = U_SENTINEL;
348 if(index<ut->chunkNativeStart || index>=ut->chunkNativeLimit) {
349 // Desired position is outside of the current chunk.
350 if(!ut->pFuncs->access(ut, index, TRUE)) {
351 // no chunk available here
352 return U_SENTINEL;
353 }
354 } else if (index - ut->chunkNativeStart <= (int64_t)ut->nativeIndexingLimit) {
355 // Desired position is in chunk, with direct 1:1 native to UTF16 indexing
356 ut->chunkOffset = (int32_t)(index - ut->chunkNativeStart);
357 } else {
358 // Desired position is in chunk, with non-UTF16 indexing.
359 ut->chunkOffset = ut->pFuncs->mapNativeIndexToUTF16(ut, index);
360 }
362 c = ut->chunkContents[ut->chunkOffset++];
363 if (U16_IS_SURROGATE(c)) {
364 // Surrogates. Many edge cases. Use other functions that already
365 // deal with the problems.
366 utext_setNativeIndex(ut, index);
367 c = utext_next32(ut);
368 }
369 return c;
370 }
373 U_CAPI UChar32 U_EXPORT2
374 utext_previous32From(UText *ut, int64_t index) {
375 //
376 // Return the character preceding the specified index.
377 // Leave the iteration position at the start of the character that was returned.
378 //
379 UChar32 cPrev; // The character preceding cCurr, which is what we will return.
381 // Address the chunk containg the position preceding the incoming index
382 // A tricky edge case:
383 // We try to test the requested native index against the chunkNativeStart to determine
384 // whether the character preceding the one at the index is in the current chunk.
385 // BUT, this test can fail with UTF-8 (or any other multibyte encoding), when the
386 // requested index is on something other than the first position of the first char.
387 //
388 if(index<=ut->chunkNativeStart || index>ut->chunkNativeLimit) {
389 // Requested native index is outside of the current chunk.
390 if(!ut->pFuncs->access(ut, index, FALSE)) {
391 // no chunk available here
392 return U_SENTINEL;
393 }
394 } else if(index - ut->chunkNativeStart <= (int64_t)ut->nativeIndexingLimit) {
395 // Direct UTF-16 indexing.
396 ut->chunkOffset = (int32_t)(index - ut->chunkNativeStart);
397 } else {
398 ut->chunkOffset=ut->pFuncs->mapNativeIndexToUTF16(ut, index);
399 if (ut->chunkOffset==0 && !ut->pFuncs->access(ut, index, FALSE)) {
400 // no chunk available here
401 return U_SENTINEL;
402 }
403 }
405 //
406 // Simple case with no surrogates.
407 //
408 ut->chunkOffset--;
409 cPrev = ut->chunkContents[ut->chunkOffset];
411 if (U16_IS_SURROGATE(cPrev)) {
412 // Possible supplementary. Many edge cases.
413 // Let other functions do the heavy lifting.
414 utext_setNativeIndex(ut, index);
415 cPrev = utext_previous32(ut);
416 }
417 return cPrev;
418 }
421 U_CAPI int32_t U_EXPORT2
422 utext_extract(UText *ut,
423 int64_t start, int64_t limit,
424 UChar *dest, int32_t destCapacity,
425 UErrorCode *status) {
426 return ut->pFuncs->extract(ut, start, limit, dest, destCapacity, status);
427 }
431 U_CAPI UBool U_EXPORT2
432 utext_equals(const UText *a, const UText *b) {
433 if (a==NULL || b==NULL ||
434 a->magic != UTEXT_MAGIC ||
435 b->magic != UTEXT_MAGIC) {
436 // Null or invalid arguments don't compare equal to anything.
437 return FALSE;
438 }
440 if (a->pFuncs != b->pFuncs) {
441 // Different types of text providers.
442 return FALSE;
443 }
445 if (a->context != b->context) {
446 // Different sources (different strings)
447 return FALSE;
448 }
449 if (utext_getNativeIndex(a) != utext_getNativeIndex(b)) {
450 // Different current position in the string.
451 return FALSE;
452 }
454 return TRUE;
455 }
457 U_CAPI UBool U_EXPORT2
458 utext_isWritable(const UText *ut)
459 {
460 UBool b = (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_WRITABLE)) != 0;
461 return b;
462 }
465 U_CAPI void U_EXPORT2
466 utext_freeze(UText *ut) {
467 // Zero out the WRITABLE flag.
468 ut->providerProperties &= ~(I32_FLAG(UTEXT_PROVIDER_WRITABLE));
469 }
472 U_CAPI UBool U_EXPORT2
473 utext_hasMetaData(const UText *ut)
474 {
475 UBool b = (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_HAS_META_DATA)) != 0;
476 return b;
477 }
481 U_CAPI int32_t U_EXPORT2
482 utext_replace(UText *ut,
483 int64_t nativeStart, int64_t nativeLimit,
484 const UChar *replacementText, int32_t replacementLength,
485 UErrorCode *status)
486 {
487 if (U_FAILURE(*status)) {
488 return 0;
489 }
490 if ((ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_WRITABLE)) == 0) {
491 *status = U_NO_WRITE_PERMISSION;
492 return 0;
493 }
494 int32_t i = ut->pFuncs->replace(ut, nativeStart, nativeLimit, replacementText, replacementLength, status);
495 return i;
496 }
498 U_CAPI void U_EXPORT2
499 utext_copy(UText *ut,
500 int64_t nativeStart, int64_t nativeLimit,
501 int64_t destIndex,
502 UBool move,
503 UErrorCode *status)
504 {
505 if (U_FAILURE(*status)) {
506 return;
507 }
508 if ((ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_WRITABLE)) == 0) {
509 *status = U_NO_WRITE_PERMISSION;
510 return;
511 }
512 ut->pFuncs->copy(ut, nativeStart, nativeLimit, destIndex, move, status);
513 }
517 U_CAPI UText * U_EXPORT2
518 utext_clone(UText *dest, const UText *src, UBool deep, UBool readOnly, UErrorCode *status) {
519 UText *result;
520 result = src->pFuncs->clone(dest, src, deep, status);
521 if (readOnly) {
522 utext_freeze(result);
523 }
524 return result;
525 }
529 //------------------------------------------------------------------------------
530 //
531 // UText common functions implementation
532 //
533 //------------------------------------------------------------------------------
535 //
536 // UText.flags bit definitions
537 //
538 enum {
539 UTEXT_HEAP_ALLOCATED = 1, // 1 if ICU has allocated this UText struct on the heap.
540 // 0 if caller provided storage for the UText.
542 UTEXT_EXTRA_HEAP_ALLOCATED = 2, // 1 if ICU has allocated extra storage as a separate
543 // heap block.
544 // 0 if there is no separate allocation. Either no extra
545 // storage was requested, or it is appended to the end
546 // of the main UText storage.
548 UTEXT_OPEN = 4 // 1 if this UText is currently open
549 // 0 if this UText is not open.
550 };
553 //
554 // Extended form of a UText. The purpose is to aid in computing the total size required
555 // when a provider asks for a UText to be allocated with extra storage.
557 struct ExtendedUText {
558 UText ut;
559 UAlignedMemory extension;
560 };
562 static const UText emptyText = UTEXT_INITIALIZER;
564 U_CAPI UText * U_EXPORT2
565 utext_setup(UText *ut, int32_t extraSpace, UErrorCode *status) {
566 if (U_FAILURE(*status)) {
567 return ut;
568 }
570 if (ut == NULL) {
571 // We need to heap-allocate storage for the new UText
572 int32_t spaceRequired = sizeof(UText);
573 if (extraSpace > 0) {
574 spaceRequired = sizeof(ExtendedUText) + extraSpace - sizeof(UAlignedMemory);
575 }
576 ut = (UText *)uprv_malloc(spaceRequired);
577 if (ut == NULL) {
578 *status = U_MEMORY_ALLOCATION_ERROR;
579 return NULL;
580 } else {
581 *ut = emptyText;
582 ut->flags |= UTEXT_HEAP_ALLOCATED;
583 if (spaceRequired>0) {
584 ut->extraSize = extraSpace;
585 ut->pExtra = &((ExtendedUText *)ut)->extension;
586 }
587 }
588 } else {
589 // We have been supplied with an already existing UText.
590 // Verify that it really appears to be a UText.
591 if (ut->magic != UTEXT_MAGIC) {
592 *status = U_ILLEGAL_ARGUMENT_ERROR;
593 return ut;
594 }
595 // If the ut is already open and there's a provider supplied close
596 // function, call it.
597 if ((ut->flags & UTEXT_OPEN) && ut->pFuncs->close != NULL) {
598 ut->pFuncs->close(ut);
599 }
600 ut->flags &= ~UTEXT_OPEN;
602 // If extra space was requested by our caller, check whether
603 // sufficient already exists, and allocate new if needed.
604 if (extraSpace > ut->extraSize) {
605 // Need more space. If there is existing separately allocated space,
606 // delete it first, then allocate new space.
607 if (ut->flags & UTEXT_EXTRA_HEAP_ALLOCATED) {
608 uprv_free(ut->pExtra);
609 ut->extraSize = 0;
610 }
611 ut->pExtra = uprv_malloc(extraSpace);
612 if (ut->pExtra == NULL) {
613 *status = U_MEMORY_ALLOCATION_ERROR;
614 } else {
615 ut->extraSize = extraSpace;
616 ut->flags |= UTEXT_EXTRA_HEAP_ALLOCATED;
617 }
618 }
619 }
620 if (U_SUCCESS(*status)) {
621 ut->flags |= UTEXT_OPEN;
623 // Initialize all remaining fields of the UText.
624 //
625 ut->context = NULL;
626 ut->chunkContents = NULL;
627 ut->p = NULL;
628 ut->q = NULL;
629 ut->r = NULL;
630 ut->a = 0;
631 ut->b = 0;
632 ut->c = 0;
633 ut->chunkOffset = 0;
634 ut->chunkLength = 0;
635 ut->chunkNativeStart = 0;
636 ut->chunkNativeLimit = 0;
637 ut->nativeIndexingLimit = 0;
638 ut->providerProperties = 0;
639 ut->privA = 0;
640 ut->privB = 0;
641 ut->privC = 0;
642 ut->privP = NULL;
643 if (ut->pExtra!=NULL && ut->extraSize>0)
644 uprv_memset(ut->pExtra, 0, ut->extraSize);
646 }
647 return ut;
648 }
651 U_CAPI UText * U_EXPORT2
652 utext_close(UText *ut) {
653 if (ut==NULL ||
654 ut->magic != UTEXT_MAGIC ||
655 (ut->flags & UTEXT_OPEN) == 0)
656 {
657 // The supplied ut is not an open UText.
658 // Do nothing.
659 return ut;
660 }
662 // If the provider gave us a close function, call it now.
663 // This will clean up anything allocated specifically by the provider.
664 if (ut->pFuncs->close != NULL) {
665 ut->pFuncs->close(ut);
666 }
667 ut->flags &= ~UTEXT_OPEN;
669 // If we (the framework) allocated the UText or subsidiary storage,
670 // delete it.
671 if (ut->flags & UTEXT_EXTRA_HEAP_ALLOCATED) {
672 uprv_free(ut->pExtra);
673 ut->pExtra = NULL;
674 ut->flags &= ~UTEXT_EXTRA_HEAP_ALLOCATED;
675 ut->extraSize = 0;
676 }
678 // Zero out function table of the closed UText. This is a defensive move,
679 // inteded to cause applications that inadvertantly use a closed
680 // utext to crash with null pointer errors.
681 ut->pFuncs = NULL;
683 if (ut->flags & UTEXT_HEAP_ALLOCATED) {
684 // This UText was allocated by UText setup. We need to free it.
685 // Clear magic, so we can detect if the user messes up and immediately
686 // tries to reopen another UText using the deleted storage.
687 ut->magic = 0;
688 uprv_free(ut);
689 ut = NULL;
690 }
691 return ut;
692 }
697 //
698 // invalidateChunk Reset a chunk to have no contents, so that the next call
699 // to access will cause new data to load.
700 // This is needed when copy/move/replace operate directly on the
701 // backing text, potentially putting it out of sync with the
702 // contents in the chunk.
703 //
704 static void
705 invalidateChunk(UText *ut) {
706 ut->chunkLength = 0;
707 ut->chunkNativeLimit = 0;
708 ut->chunkNativeStart = 0;
709 ut->chunkOffset = 0;
710 ut->nativeIndexingLimit = 0;
711 }
713 //
714 // pinIndex Do range pinning on a native index parameter.
715 // 64 bit pinning is done in place.
716 // 32 bit truncated result is returned as a convenience for
717 // use in providers that don't need 64 bits.
718 static int32_t
719 pinIndex(int64_t &index, int64_t limit) {
720 if (index<0) {
721 index = 0;
722 } else if (index > limit) {
723 index = limit;
724 }
725 return (int32_t)index;
726 }
729 U_CDECL_BEGIN
731 //
732 // Pointer relocation function,
733 // a utility used by shallow clone.
734 // Adjust a pointer that refers to something within one UText (the source)
735 // to refer to the same relative offset within a another UText (the target)
736 //
737 static void adjustPointer(UText *dest, const void **destPtr, const UText *src) {
738 // convert all pointers to (char *) so that byte address arithmetic will work.
739 char *dptr = (char *)*destPtr;
740 char *dUText = (char *)dest;
741 char *sUText = (char *)src;
743 if (dptr >= (char *)src->pExtra && dptr < ((char*)src->pExtra)+src->extraSize) {
744 // target ptr was to something within the src UText's pExtra storage.
745 // relocate it into the target UText's pExtra region.
746 *destPtr = ((char *)dest->pExtra) + (dptr - (char *)src->pExtra);
747 } else if (dptr>=sUText && dptr < sUText+src->sizeOfStruct) {
748 // target ptr was pointing to somewhere within the source UText itself.
749 // Move it to the same offset within the target UText.
750 *destPtr = dUText + (dptr-sUText);
751 }
752 }
755 //
756 // Clone. This is a generic copy-the-utext-by-value clone function that can be
757 // used as-is with some utext types, and as a helper by other clones.
758 //
759 static UText * U_CALLCONV
760 shallowTextClone(UText * dest, const UText * src, UErrorCode * status) {
761 if (U_FAILURE(*status)) {
762 return NULL;
763 }
764 int32_t srcExtraSize = src->extraSize;
766 //
767 // Use the generic text_setup to allocate storage if required.
768 //
769 dest = utext_setup(dest, srcExtraSize, status);
770 if (U_FAILURE(*status)) {
771 return dest;
772 }
774 //
775 // flags (how the UText was allocated) and the pointer to the
776 // extra storage must retain the values in the cloned utext that
777 // were set up by utext_setup. Save them separately before
778 // copying the whole struct.
779 //
780 void *destExtra = dest->pExtra;
781 int32_t flags = dest->flags;
784 //
785 // Copy the whole UText struct by value.
786 // Any "Extra" storage is copied also.
787 //
788 int sizeToCopy = src->sizeOfStruct;
789 if (sizeToCopy > dest->sizeOfStruct) {
790 sizeToCopy = dest->sizeOfStruct;
791 }
792 uprv_memcpy(dest, src, sizeToCopy);
793 dest->pExtra = destExtra;
794 dest->flags = flags;
795 if (srcExtraSize > 0) {
796 uprv_memcpy(dest->pExtra, src->pExtra, srcExtraSize);
797 }
799 //
800 // Relocate any pointers in the target that refer to the UText itself
801 // to point to the cloned copy rather than the original source.
802 //
803 adjustPointer(dest, &dest->context, src);
804 adjustPointer(dest, &dest->p, src);
805 adjustPointer(dest, &dest->q, src);
806 adjustPointer(dest, &dest->r, src);
807 adjustPointer(dest, (const void **)&dest->chunkContents, src);
809 return dest;
810 }
813 U_CDECL_END
817 //------------------------------------------------------------------------------
818 //
819 // UText implementation for UTF-8 char * strings (read-only)
820 // Limitation: string length must be <= 0x7fffffff in length.
821 // (length must for in an int32_t variable)
822 //
823 // Use of UText data members:
824 // context pointer to UTF-8 string
825 // utext.b is the input string length (bytes).
826 // utext.c Length scanned so far in string
827 // (for optimizing finding length of zero terminated strings.)
828 // utext.p pointer to the current buffer
829 // utext.q pointer to the other buffer.
830 //
831 //------------------------------------------------------------------------------
833 // Chunk size.
834 // Must be less than 85, because of byte mapping from UChar indexes to native indexes.
835 // Worst case is three native bytes to one UChar. (Supplemenaries are 4 native bytes
836 // to two UChars.)
837 //
838 enum { UTF8_TEXT_CHUNK_SIZE=32 };
840 //
841 // UTF8Buf Two of these structs will be set up in the UText's extra allocated space.
842 // Each contains the UChar chunk buffer, the to and from native maps, and
843 // header info.
844 //
845 // because backwards iteration fills the buffers starting at the end and
846 // working towards the front, the filled part of the buffers may not begin
847 // at the start of the available storage for the buffers.
848 //
849 // Buffer size is one bigger than the specified UTF8_TEXT_CHUNK_SIZE to allow for
850 // the last character added being a supplementary, and thus requiring a surrogate
851 // pair. Doing this is simpler than checking for the edge case.
852 //
854 struct UTF8Buf {
855 int32_t bufNativeStart; // Native index of first char in UChar buf
856 int32_t bufNativeLimit; // Native index following last char in buf.
857 int32_t bufStartIdx; // First filled position in buf.
858 int32_t bufLimitIdx; // Limit of filled range in buf.
859 int32_t bufNILimit; // Limit of native indexing part of buf
860 int32_t toUCharsMapStart; // Native index corresponding to
861 // mapToUChars[0].
862 // Set to bufNativeStart when filling forwards.
863 // Set to computed value when filling backwards.
865 UChar buf[UTF8_TEXT_CHUNK_SIZE+4]; // The UChar buffer. Requires one extra position beyond the
866 // the chunk size, to allow for surrogate at the end.
867 // Length must be identical to mapToNative array, below,
868 // because of the way indexing works when the array is
869 // filled backwards during a reverse iteration. Thus,
870 // the additional extra size.
871 uint8_t mapToNative[UTF8_TEXT_CHUNK_SIZE+4]; // map UChar index in buf to
872 // native offset from bufNativeStart.
873 // Requires two extra slots,
874 // one for a supplementary starting in the last normal position,
875 // and one for an entry for the buffer limit position.
876 uint8_t mapToUChars[UTF8_TEXT_CHUNK_SIZE*3+6]; // Map native offset from bufNativeStart to
877 // correspoding offset in filled part of buf.
878 int32_t align;
879 };
881 U_CDECL_BEGIN
883 //
884 // utf8TextLength
885 //
886 // Get the length of the string. If we don't already know it,
887 // we'll need to scan for the trailing nul.
888 //
889 static int64_t U_CALLCONV
890 utf8TextLength(UText *ut) {
891 if (ut->b < 0) {
892 // Zero terminated string, and we haven't scanned to the end yet.
893 // Scan it now.
894 const char *r = (const char *)ut->context + ut->c;
895 while (*r != 0) {
896 r++;
897 }
898 if ((r - (const char *)ut->context) < 0x7fffffff) {
899 ut->b = (int32_t)(r - (const char *)ut->context);
900 } else {
901 // Actual string was bigger (more than 2 gig) than we
902 // can handle. Clip it to 2 GB.
903 ut->b = 0x7fffffff;
904 }
905 ut->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
906 }
907 return ut->b;
908 }
915 static UBool U_CALLCONV
916 utf8TextAccess(UText *ut, int64_t index, UBool forward) {
917 //
918 // Apologies to those who are allergic to goto statements.
919 // Consider each goto to a labelled block to be the equivalent of
920 // call the named block as if it were a function();
921 // return;
922 //
923 const uint8_t *s8=(const uint8_t *)ut->context;
924 UTF8Buf *u8b = NULL;
925 int32_t length = ut->b; // Length of original utf-8
926 int32_t ix= (int32_t)index; // Requested index, trimmed to 32 bits.
927 int32_t mapIndex = 0;
928 if (index<0) {
929 ix=0;
930 } else if (index > 0x7fffffff) {
931 // Strings with 64 bit lengths not supported by this UTF-8 provider.
932 ix = 0x7fffffff;
933 }
935 // Pin requested index to the string length.
936 if (ix>length) {
937 if (length>=0) {
938 ix=length;
939 } else if (ix>=ut->c) {
940 // Zero terminated string, and requested index is beyond
941 // the region that has already been scanned.
942 // Scan up to either the end of the string or to the
943 // requested position, whichever comes first.
944 while (ut->c<ix && s8[ut->c]!=0) {
945 ut->c++;
946 }
947 // TODO: support for null terminated string length > 32 bits.
948 if (s8[ut->c] == 0) {
949 // We just found the actual length of the string.
950 // Trim the requested index back to that.
951 ix = ut->c;
952 ut->b = ut->c;
953 length = ut->c;
954 ut->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
955 }
956 }
957 }
959 //
960 // Dispatch to the appropriate action for a forward iteration request.
961 //
962 if (forward) {
963 if (ix==ut->chunkNativeLimit) {
964 // Check for normal sequential iteration cases first.
965 if (ix==length) {
966 // Just reached end of string
967 // Don't swap buffers, but do set the
968 // current buffer position.
969 ut->chunkOffset = ut->chunkLength;
970 return FALSE;
971 } else {
972 // End of current buffer.
973 // check whether other buffer already has what we need.
974 UTF8Buf *altB = (UTF8Buf *)ut->q;
975 if (ix>=altB->bufNativeStart && ix<altB->bufNativeLimit) {
976 goto swapBuffers;
977 }
978 }
979 }
981 // A random access. Desired index could be in either or niether buf.
982 // For optimizing the order of testing, first check for the index
983 // being in the other buffer. This will be the case for uses that
984 // move back and forth over a fairly limited range
985 {
986 u8b = (UTF8Buf *)ut->q; // the alternate buffer
987 if (ix>=u8b->bufNativeStart && ix<u8b->bufNativeLimit) {
988 // Requested index is in the other buffer.
989 goto swapBuffers;
990 }
991 if (ix == length) {
992 // Requested index is end-of-string.
993 // (this is the case of randomly seeking to the end.
994 // The case of iterating off the end is handled earlier.)
995 if (ix == ut->chunkNativeLimit) {
996 // Current buffer extends up to the end of the string.
997 // Leave it as the current buffer.
998 ut->chunkOffset = ut->chunkLength;
999 return FALSE;
1000 }
1001 if (ix == u8b->bufNativeLimit) {
1002 // Alternate buffer extends to the end of string.
1003 // Swap it in as the current buffer.
1004 goto swapBuffersAndFail;
1005 }
1007 // Neither existing buffer extends to the end of the string.
1008 goto makeStubBuffer;
1009 }
1011 if (ix<ut->chunkNativeStart || ix>=ut->chunkNativeLimit) {
1012 // Requested index is in neither buffer.
1013 goto fillForward;
1014 }
1016 // Requested index is in this buffer.
1017 u8b = (UTF8Buf *)ut->p; // the current buffer
1018 mapIndex = ix - u8b->toUCharsMapStart;
1019 ut->chunkOffset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;
1020 return TRUE;
1022 }
1023 }
1026 //
1027 // Dispatch to the appropriate action for a
1028 // Backwards Diretion iteration request.
1029 //
1030 if (ix==ut->chunkNativeStart) {
1031 // Check for normal sequential iteration cases first.
1032 if (ix==0) {
1033 // Just reached the start of string
1034 // Don't swap buffers, but do set the
1035 // current buffer position.
1036 ut->chunkOffset = 0;
1037 return FALSE;
1038 } else {
1039 // Start of current buffer.
1040 // check whether other buffer already has what we need.
1041 UTF8Buf *altB = (UTF8Buf *)ut->q;
1042 if (ix>altB->bufNativeStart && ix<=altB->bufNativeLimit) {
1043 goto swapBuffers;
1044 }
1045 }
1046 }
1048 // A random access. Desired index could be in either or niether buf.
1049 // For optimizing the order of testing,
1050 // Most likely case: in the other buffer.
1051 // Second most likely: in neither buffer.
1052 // Unlikely, but must work: in the current buffer.
1053 u8b = (UTF8Buf *)ut->q; // the alternate buffer
1054 if (ix>u8b->bufNativeStart && ix<=u8b->bufNativeLimit) {
1055 // Requested index is in the other buffer.
1056 goto swapBuffers;
1057 }
1058 // Requested index is start-of-string.
1059 // (this is the case of randomly seeking to the start.
1060 // The case of iterating off the start is handled earlier.)
1061 if (ix==0) {
1062 if (u8b->bufNativeStart==0) {
1063 // Alternate buffer contains the data for the start string.
1064 // Make it be the current buffer.
1065 goto swapBuffersAndFail;
1066 } else {
1067 // Request for data before the start of string,
1068 // neither buffer is usable.
1069 // set up a zero-length buffer.
1070 goto makeStubBuffer;
1071 }
1072 }
1074 if (ix<=ut->chunkNativeStart || ix>ut->chunkNativeLimit) {
1075 // Requested index is in neither buffer.
1076 goto fillReverse;
1077 }
1079 // Requested index is in this buffer.
1080 // Set the utf16 buffer index.
1081 u8b = (UTF8Buf *)ut->p;
1082 mapIndex = ix - u8b->toUCharsMapStart;
1083 ut->chunkOffset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;
1084 if (ut->chunkOffset==0) {
1085 // This occurs when the first character in the text is
1086 // a multi-byte UTF-8 char, and the requested index is to
1087 // one of the trailing bytes. Because there is no preceding ,
1088 // character, this access fails. We can't pick up on the
1089 // situation sooner because the requested index is not zero.
1090 return FALSE;
1091 } else {
1092 return TRUE;
1093 }
1097 swapBuffers:
1098 // The alternate buffer (ut->q) has the string data that was requested.
1099 // Swap the primary and alternate buffers, and set the
1100 // chunk index into the new primary buffer.
1101 {
1102 u8b = (UTF8Buf *)ut->q;
1103 ut->q = ut->p;
1104 ut->p = u8b;
1105 ut->chunkContents = &u8b->buf[u8b->bufStartIdx];
1106 ut->chunkLength = u8b->bufLimitIdx - u8b->bufStartIdx;
1107 ut->chunkNativeStart = u8b->bufNativeStart;
1108 ut->chunkNativeLimit = u8b->bufNativeLimit;
1109 ut->nativeIndexingLimit = u8b->bufNILimit;
1111 // Index into the (now current) chunk
1112 // Use the map to set the chunk index. It's more trouble than it's worth
1113 // to check whether native indexing can be used.
1114 U_ASSERT(ix>=u8b->bufNativeStart);
1115 U_ASSERT(ix<=u8b->bufNativeLimit);
1116 mapIndex = ix - u8b->toUCharsMapStart;
1117 U_ASSERT(mapIndex>=0);
1118 U_ASSERT(mapIndex<(int32_t)sizeof(u8b->mapToUChars));
1119 ut->chunkOffset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;
1121 return TRUE;
1122 }
1125 swapBuffersAndFail:
1126 // We got a request for either the start or end of the string,
1127 // with iteration continuing in the out-of-bounds direction.
1128 // The alternate buffer already contains the data up to the
1129 // start/end.
1130 // Swap the buffers, then return failure, indicating that we couldn't
1131 // make things correct for continuing the iteration in the requested
1132 // direction. The position & buffer are correct should the
1133 // user decide to iterate in the opposite direction.
1134 u8b = (UTF8Buf *)ut->q;
1135 ut->q = ut->p;
1136 ut->p = u8b;
1137 ut->chunkContents = &u8b->buf[u8b->bufStartIdx];
1138 ut->chunkLength = u8b->bufLimitIdx - u8b->bufStartIdx;
1139 ut->chunkNativeStart = u8b->bufNativeStart;
1140 ut->chunkNativeLimit = u8b->bufNativeLimit;
1141 ut->nativeIndexingLimit = u8b->bufNILimit;
1143 // Index into the (now current) chunk
1144 // For this function (swapBuffersAndFail), the requested index
1145 // will always be at either the start or end of the chunk.
1146 if (ix==u8b->bufNativeLimit) {
1147 ut->chunkOffset = ut->chunkLength;
1148 } else {
1149 ut->chunkOffset = 0;
1150 U_ASSERT(ix == u8b->bufNativeStart);
1151 }
1152 return FALSE;
1154 makeStubBuffer:
1155 // The user has done a seek/access past the start or end
1156 // of the string. Rather than loading data that is likely
1157 // to never be used, just set up a zero-length buffer at
1158 // the position.
1159 u8b = (UTF8Buf *)ut->q;
1160 u8b->bufNativeStart = ix;
1161 u8b->bufNativeLimit = ix;
1162 u8b->bufStartIdx = 0;
1163 u8b->bufLimitIdx = 0;
1164 u8b->bufNILimit = 0;
1165 u8b->toUCharsMapStart = ix;
1166 u8b->mapToNative[0] = 0;
1167 u8b->mapToUChars[0] = 0;
1168 goto swapBuffersAndFail;
1172 fillForward:
1173 {
1174 // Move the incoming index to a code point boundary.
1175 U8_SET_CP_START(s8, 0, ix);
1177 // Swap the UText buffers.
1178 // We want to fill what was previously the alternate buffer,
1179 // and make what was the current buffer be the new alternate.
1180 UTF8Buf *u8b = (UTF8Buf *)ut->q;
1181 ut->q = ut->p;
1182 ut->p = u8b;
1184 int32_t strLen = ut->b;
1185 UBool nulTerminated = FALSE;
1186 if (strLen < 0) {
1187 strLen = 0x7fffffff;
1188 nulTerminated = TRUE;
1189 }
1191 UChar *buf = u8b->buf;
1192 uint8_t *mapToNative = u8b->mapToNative;
1193 uint8_t *mapToUChars = u8b->mapToUChars;
1194 int32_t destIx = 0;
1195 int32_t srcIx = ix;
1196 UBool seenNonAscii = FALSE;
1197 UChar32 c = 0;
1199 // Fill the chunk buffer and mapping arrays.
1200 while (destIx<UTF8_TEXT_CHUNK_SIZE) {
1201 c = s8[srcIx];
1202 if (c>0 && c<0x80) {
1203 // Special case ASCII range for speed.
1204 // zero is excluded to simplify bounds checking.
1205 buf[destIx] = (UChar)c;
1206 mapToNative[destIx] = (uint8_t)(srcIx - ix);
1207 mapToUChars[srcIx-ix] = (uint8_t)destIx;
1208 srcIx++;
1209 destIx++;
1210 } else {
1211 // General case, handle everything.
1212 if (seenNonAscii == FALSE) {
1213 seenNonAscii = TRUE;
1214 u8b->bufNILimit = destIx;
1215 }
1217 int32_t cIx = srcIx;
1218 int32_t dIx = destIx;
1219 int32_t dIxSaved = destIx;
1220 U8_NEXT_OR_FFFD(s8, srcIx, strLen, c);
1221 if (c==0 && nulTerminated) {
1222 srcIx--;
1223 break;
1224 }
1226 U16_APPEND_UNSAFE(buf, destIx, c);
1227 do {
1228 mapToNative[dIx++] = (uint8_t)(cIx - ix);
1229 } while (dIx < destIx);
1231 do {
1232 mapToUChars[cIx++ - ix] = (uint8_t)dIxSaved;
1233 } while (cIx < srcIx);
1234 }
1235 if (srcIx>=strLen) {
1236 break;
1237 }
1239 }
1241 // store Native <--> Chunk Map entries for the end of the buffer.
1242 // There is no actual character here, but the index position is valid.
1243 mapToNative[destIx] = (uint8_t)(srcIx - ix);
1244 mapToUChars[srcIx - ix] = (uint8_t)destIx;
1246 // fill in Buffer descriptor
1247 u8b->bufNativeStart = ix;
1248 u8b->bufNativeLimit = srcIx;
1249 u8b->bufStartIdx = 0;
1250 u8b->bufLimitIdx = destIx;
1251 if (seenNonAscii == FALSE) {
1252 u8b->bufNILimit = destIx;
1253 }
1254 u8b->toUCharsMapStart = u8b->bufNativeStart;
1256 // Set UText chunk to refer to this buffer.
1257 ut->chunkContents = buf;
1258 ut->chunkOffset = 0;
1259 ut->chunkLength = u8b->bufLimitIdx;
1260 ut->chunkNativeStart = u8b->bufNativeStart;
1261 ut->chunkNativeLimit = u8b->bufNativeLimit;
1262 ut->nativeIndexingLimit = u8b->bufNILimit;
1264 // For zero terminated strings, keep track of the maximum point
1265 // scanned so far.
1266 if (nulTerminated && srcIx>ut->c) {
1267 ut->c = srcIx;
1268 if (c==0) {
1269 // We scanned to the end.
1270 // Remember the actual length.
1271 ut->b = srcIx;
1272 ut->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
1273 }
1274 }
1275 return TRUE;
1276 }
1279 fillReverse:
1280 {
1281 // Move the incoming index to a code point boundary.
1282 // Can only do this if the incoming index is somewhere in the interior of the string.
1283 // If index is at the end, there is no character there to look at.
1284 if (ix != ut->b) {
1285 U8_SET_CP_START(s8, 0, ix);
1286 }
1288 // Swap the UText buffers.
1289 // We want to fill what was previously the alternate buffer,
1290 // and make what was the current buffer be the new alternate.
1291 UTF8Buf *u8b = (UTF8Buf *)ut->q;
1292 ut->q = ut->p;
1293 ut->p = u8b;
1295 UChar *buf = u8b->buf;
1296 uint8_t *mapToNative = u8b->mapToNative;
1297 uint8_t *mapToUChars = u8b->mapToUChars;
1298 int32_t toUCharsMapStart = ix - (UTF8_TEXT_CHUNK_SIZE*3 + 1);
1299 int32_t destIx = UTF8_TEXT_CHUNK_SIZE+2; // Start in the overflow region
1300 // at end of buffer to leave room
1301 // for a surrogate pair at the
1302 // buffer start.
1303 int32_t srcIx = ix;
1304 int32_t bufNILimit = destIx;
1305 UChar32 c;
1307 // Map to/from Native Indexes, fill in for the position at the end of
1308 // the buffer.
1309 //
1310 mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
1311 mapToUChars[srcIx - toUCharsMapStart] = (uint8_t)destIx;
1313 // Fill the chunk buffer
1314 // Work backwards, filling from the end of the buffer towards the front.
1315 //
1316 while (destIx>2 && (srcIx - toUCharsMapStart > 5) && (srcIx > 0)) {
1317 srcIx--;
1318 destIx--;
1320 // Get last byte of the UTF-8 character
1321 c = s8[srcIx];
1322 if (c<0x80) {
1323 // Special case ASCII range for speed.
1324 buf[destIx] = (UChar)c;
1325 mapToUChars[srcIx - toUCharsMapStart] = (uint8_t)destIx;
1326 mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
1327 } else {
1328 // General case, handle everything non-ASCII.
1330 int32_t sIx = srcIx; // ix of last byte of multi-byte u8 char
1332 // Get the full character from the UTF8 string.
1333 // use code derived from tbe macros in utf8.h
1334 // Leaves srcIx pointing at the first byte of the UTF-8 char.
1335 //
1336 c=utf8_prevCharSafeBody(s8, 0, &srcIx, c, -3);
1337 // leaves srcIx at first byte of the multi-byte char.
1339 // Store the character in UTF-16 buffer.
1340 if (c<0x10000) {
1341 buf[destIx] = (UChar)c;
1342 mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
1343 } else {
1344 buf[destIx] = U16_TRAIL(c);
1345 mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
1346 buf[--destIx] = U16_LEAD(c);
1347 mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
1348 }
1350 // Fill in the map from native indexes to UChars buf index.
1351 do {
1352 mapToUChars[sIx-- - toUCharsMapStart] = (uint8_t)destIx;
1353 } while (sIx >= srcIx);
1355 // Set native indexing limit to be the current position.
1356 // We are processing a non-ascii, non-native-indexing char now;
1357 // the limit will be here if the rest of the chars to be
1358 // added to this buffer are ascii.
1359 bufNILimit = destIx;
1360 }
1361 }
1362 u8b->bufNativeStart = srcIx;
1363 u8b->bufNativeLimit = ix;
1364 u8b->bufStartIdx = destIx;
1365 u8b->bufLimitIdx = UTF8_TEXT_CHUNK_SIZE+2;
1366 u8b->bufNILimit = bufNILimit - u8b->bufStartIdx;
1367 u8b->toUCharsMapStart = toUCharsMapStart;
1369 ut->chunkContents = &buf[u8b->bufStartIdx];
1370 ut->chunkLength = u8b->bufLimitIdx - u8b->bufStartIdx;
1371 ut->chunkOffset = ut->chunkLength;
1372 ut->chunkNativeStart = u8b->bufNativeStart;
1373 ut->chunkNativeLimit = u8b->bufNativeLimit;
1374 ut->nativeIndexingLimit = u8b->bufNILimit;
1375 return TRUE;
1376 }
1378 }
1382 //
1383 // This is a slightly modified copy of u_strFromUTF8,
1384 // Inserts a Replacement Char rather than failing on invalid UTF-8
1385 // Removes unnecessary features.
1386 //
1387 static UChar*
1388 utext_strFromUTF8(UChar *dest,
1389 int32_t destCapacity,
1390 int32_t *pDestLength,
1391 const char* src,
1392 int32_t srcLength, // required. NUL terminated not supported.
1393 UErrorCode *pErrorCode
1394 )
1395 {
1397 UChar *pDest = dest;
1398 UChar *pDestLimit = (dest!=NULL)?(dest+destCapacity):NULL;
1399 UChar32 ch=0;
1400 int32_t index = 0;
1401 int32_t reqLength = 0;
1402 uint8_t* pSrc = (uint8_t*) src;
1405 while((index < srcLength)&&(pDest<pDestLimit)){
1406 ch = pSrc[index++];
1407 if(ch <=0x7f){
1408 *pDest++=(UChar)ch;
1409 }else{
1410 ch=utf8_nextCharSafeBody(pSrc, &index, srcLength, ch, -3);
1411 if(U_IS_BMP(ch)){
1412 *(pDest++)=(UChar)ch;
1413 }else{
1414 *(pDest++)=U16_LEAD(ch);
1415 if(pDest<pDestLimit){
1416 *(pDest++)=U16_TRAIL(ch);
1417 }else{
1418 reqLength++;
1419 break;
1420 }
1421 }
1422 }
1423 }
1424 /* donot fill the dest buffer just count the UChars needed */
1425 while(index < srcLength){
1426 ch = pSrc[index++];
1427 if(ch <= 0x7f){
1428 reqLength++;
1429 }else{
1430 ch=utf8_nextCharSafeBody(pSrc, &index, srcLength, ch, -3);
1431 reqLength+=U16_LENGTH(ch);
1432 }
1433 }
1435 reqLength+=(int32_t)(pDest - dest);
1437 if(pDestLength){
1438 *pDestLength = reqLength;
1439 }
1441 /* Terminate the buffer */
1442 u_terminateUChars(dest,destCapacity,reqLength,pErrorCode);
1444 return dest;
1445 }
1449 static int32_t U_CALLCONV
1450 utf8TextExtract(UText *ut,
1451 int64_t start, int64_t limit,
1452 UChar *dest, int32_t destCapacity,
1453 UErrorCode *pErrorCode) {
1454 if(U_FAILURE(*pErrorCode)) {
1455 return 0;
1456 }
1457 if(destCapacity<0 || (dest==NULL && destCapacity>0)) {
1458 *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
1459 return 0;
1460 }
1461 int32_t length = ut->b;
1462 int32_t start32 = pinIndex(start, length);
1463 int32_t limit32 = pinIndex(limit, length);
1465 if(start32>limit32) {
1466 *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR;
1467 return 0;
1468 }
1471 // adjust the incoming indexes to land on code point boundaries if needed.
1472 // adjust by no more than three, because that is the largest number of trail bytes
1473 // in a well formed UTF8 character.
1474 const uint8_t *buf = (const uint8_t *)ut->context;
1475 int i;
1476 if (start32 < ut->chunkNativeLimit) {
1477 for (i=0; i<3; i++) {
1478 if (U8_IS_SINGLE(buf[start32]) || U8_IS_LEAD(buf[start32]) || start32==0) {
1479 break;
1480 }
1481 start32--;
1482 }
1483 }
1485 if (limit32 < ut->chunkNativeLimit) {
1486 for (i=0; i<3; i++) {
1487 if (U8_IS_SINGLE(buf[limit32]) || U8_IS_LEAD(buf[limit32]) || limit32==0) {
1488 break;
1489 }
1490 limit32--;
1491 }
1492 }
1494 // Do the actual extract.
1495 int32_t destLength=0;
1496 utext_strFromUTF8(dest, destCapacity, &destLength,
1497 (const char *)ut->context+start32, limit32-start32,
1498 pErrorCode);
1499 utf8TextAccess(ut, limit32, TRUE);
1500 return destLength;
1501 }
1503 //
1504 // utf8TextMapOffsetToNative
1505 //
1506 // Map a chunk (UTF-16) offset to a native index.
1507 static int64_t U_CALLCONV
1508 utf8TextMapOffsetToNative(const UText *ut) {
1509 //
1510 UTF8Buf *u8b = (UTF8Buf *)ut->p;
1511 U_ASSERT(ut->chunkOffset>ut->nativeIndexingLimit && ut->chunkOffset<=ut->chunkLength);
1512 int32_t nativeOffset = u8b->mapToNative[ut->chunkOffset + u8b->bufStartIdx] + u8b->toUCharsMapStart;
1513 U_ASSERT(nativeOffset >= ut->chunkNativeStart && nativeOffset <= ut->chunkNativeLimit);
1514 return nativeOffset;
1515 }
1517 //
1518 // Map a native index to the corrsponding chunk offset
1519 //
1520 static int32_t U_CALLCONV
1521 utf8TextMapIndexToUTF16(const UText *ut, int64_t index64) {
1522 U_ASSERT(index64 <= 0x7fffffff);
1523 int32_t index = (int32_t)index64;
1524 UTF8Buf *u8b = (UTF8Buf *)ut->p;
1525 U_ASSERT(index>=ut->chunkNativeStart+ut->nativeIndexingLimit);
1526 U_ASSERT(index<=ut->chunkNativeLimit);
1527 int32_t mapIndex = index - u8b->toUCharsMapStart;
1528 int32_t offset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;
1529 U_ASSERT(offset>=0 && offset<=ut->chunkLength);
1530 return offset;
1531 }
1533 static UText * U_CALLCONV
1534 utf8TextClone(UText *dest, const UText *src, UBool deep, UErrorCode *status)
1535 {
1536 // First do a generic shallow clone. Does everything needed for the UText struct itself.
1537 dest = shallowTextClone(dest, src, status);
1539 // For deep clones, make a copy of the string.
1540 // The copied storage is owned by the newly created clone.
1541 //
1542 // TODO: There is an isssue with using utext_nativeLength().
1543 // That function is non-const in cases where the input was NUL terminated
1544 // and the length has not yet been determined.
1545 // This function (clone()) is const.
1546 // There potentially a thread safety issue lurking here.
1547 //
1548 if (deep && U_SUCCESS(*status)) {
1549 int32_t len = (int32_t)utext_nativeLength((UText *)src);
1550 char *copyStr = (char *)uprv_malloc(len+1);
1551 if (copyStr == NULL) {
1552 *status = U_MEMORY_ALLOCATION_ERROR;
1553 } else {
1554 uprv_memcpy(copyStr, src->context, len+1);
1555 dest->context = copyStr;
1556 dest->providerProperties |= I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT);
1557 }
1558 }
1559 return dest;
1560 }
1563 static void U_CALLCONV
1564 utf8TextClose(UText *ut) {
1565 // Most of the work of close is done by the generic UText framework close.
1566 // All that needs to be done here is to delete the UTF8 string if the UText
1567 // owns it. This occurs if the UText was created by cloning.
1568 if (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT)) {
1569 char *s = (char *)ut->context;
1570 uprv_free(s);
1571 ut->context = NULL;
1572 }
1573 }
1575 U_CDECL_END
1578 static const struct UTextFuncs utf8Funcs =
1579 {
1580 sizeof(UTextFuncs),
1581 0, 0, 0, // Reserved alignment padding
1582 utf8TextClone,
1583 utf8TextLength,
1584 utf8TextAccess,
1585 utf8TextExtract,
1586 NULL, /* replace*/
1587 NULL, /* copy */
1588 utf8TextMapOffsetToNative,
1589 utf8TextMapIndexToUTF16,
1590 utf8TextClose,
1591 NULL, // spare 1
1592 NULL, // spare 2
1593 NULL // spare 3
1594 };
1597 static const char gEmptyString[] = {0};
1599 U_CAPI UText * U_EXPORT2
1600 utext_openUTF8(UText *ut, const char *s, int64_t length, UErrorCode *status) {
1601 if(U_FAILURE(*status)) {
1602 return NULL;
1603 }
1604 if(s==NULL && length==0) {
1605 s = gEmptyString;
1606 }
1608 if(s==NULL || length<-1 || length>INT32_MAX) {
1609 *status=U_ILLEGAL_ARGUMENT_ERROR;
1610 return NULL;
1611 }
1613 ut = utext_setup(ut, sizeof(UTF8Buf) * 2, status);
1614 if (U_FAILURE(*status)) {
1615 return ut;
1616 }
1618 ut->pFuncs = &utf8Funcs;
1619 ut->context = s;
1620 ut->b = (int32_t)length;
1621 ut->c = (int32_t)length;
1622 if (ut->c < 0) {
1623 ut->c = 0;
1624 ut->providerProperties |= I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
1625 }
1626 ut->p = ut->pExtra;
1627 ut->q = (char *)ut->pExtra + sizeof(UTF8Buf);
1628 return ut;
1630 }
1639 //------------------------------------------------------------------------------
1640 //
1641 // UText implementation wrapper for Replaceable (read/write)
1642 //
1643 // Use of UText data members:
1644 // context pointer to Replaceable.
1645 // p pointer to Replaceable if it is owned by the UText.
1646 //
1647 //------------------------------------------------------------------------------
1651 // minimum chunk size for this implementation: 3
1652 // to allow for possible trimming for code point boundaries
1653 enum { REP_TEXT_CHUNK_SIZE=10 };
1655 struct ReplExtra {
1656 /*
1657 * Chunk UChars.
1658 * +1 to simplify filling with surrogate pair at the end.
1659 */
1660 UChar s[REP_TEXT_CHUNK_SIZE+1];
1661 };
1664 U_CDECL_BEGIN
1666 static UText * U_CALLCONV
1667 repTextClone(UText *dest, const UText *src, UBool deep, UErrorCode *status) {
1668 // First do a generic shallow clone. Does everything needed for the UText struct itself.
1669 dest = shallowTextClone(dest, src, status);
1671 // For deep clones, make a copy of the Replaceable.
1672 // The copied Replaceable storage is owned by the newly created UText clone.
1673 // A non-NULL pointer in UText.p is the signal to the close() function to delete
1674 // it.
1675 //
1676 if (deep && U_SUCCESS(*status)) {
1677 const Replaceable *replSrc = (const Replaceable *)src->context;
1678 dest->context = replSrc->clone();
1679 dest->providerProperties |= I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT);
1681 // with deep clone, the copy is writable, even when the source is not.
1682 dest->providerProperties |= I32_FLAG(UTEXT_PROVIDER_WRITABLE);
1683 }
1684 return dest;
1685 }
1688 static void U_CALLCONV
1689 repTextClose(UText *ut) {
1690 // Most of the work of close is done by the generic UText framework close.
1691 // All that needs to be done here is delete the Replaceable if the UText
1692 // owns it. This occurs if the UText was created by cloning.
1693 if (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT)) {
1694 Replaceable *rep = (Replaceable *)ut->context;
1695 delete rep;
1696 ut->context = NULL;
1697 }
1698 }
1701 static int64_t U_CALLCONV
1702 repTextLength(UText *ut) {
1703 const Replaceable *replSrc = (const Replaceable *)ut->context;
1704 int32_t len = replSrc->length();
1705 return len;
1706 }
1709 static UBool U_CALLCONV
1710 repTextAccess(UText *ut, int64_t index, UBool forward) {
1711 const Replaceable *rep=(const Replaceable *)ut->context;
1712 int32_t length=rep->length(); // Full length of the input text (bigger than a chunk)
1714 // clip the requested index to the limits of the text.
1715 int32_t index32 = pinIndex(index, length);
1716 U_ASSERT(index<=INT32_MAX);
1719 /*
1720 * Compute start/limit boundaries around index, for a segment of text
1721 * to be extracted.
1722 * To allow for the possibility that our user gave an index to the trailing
1723 * half of a surrogate pair, we must request one extra preceding UChar when
1724 * going in the forward direction. This will ensure that the buffer has the
1725 * entire code point at the specified index.
1726 */
1727 if(forward) {
1729 if (index32>=ut->chunkNativeStart && index32<ut->chunkNativeLimit) {
1730 // Buffer already contains the requested position.
1731 ut->chunkOffset = (int32_t)(index - ut->chunkNativeStart);
1732 return TRUE;
1733 }
1734 if (index32>=length && ut->chunkNativeLimit==length) {
1735 // Request for end of string, and buffer already extends up to it.
1736 // Can't get the data, but don't change the buffer.
1737 ut->chunkOffset = length - (int32_t)ut->chunkNativeStart;
1738 return FALSE;
1739 }
1741 ut->chunkNativeLimit = index + REP_TEXT_CHUNK_SIZE - 1;
1742 // Going forward, so we want to have the buffer with stuff at and beyond
1743 // the requested index. The -1 gets us one code point before the
1744 // requested index also, to handle the case of the index being on
1745 // a trail surrogate of a surrogate pair.
1746 if(ut->chunkNativeLimit > length) {
1747 ut->chunkNativeLimit = length;
1748 }
1749 // unless buffer ran off end, start is index-1.
1750 ut->chunkNativeStart = ut->chunkNativeLimit - REP_TEXT_CHUNK_SIZE;
1751 if(ut->chunkNativeStart < 0) {
1752 ut->chunkNativeStart = 0;
1753 }
1754 } else {
1755 // Reverse iteration. Fill buffer with data preceding the requested index.
1756 if (index32>ut->chunkNativeStart && index32<=ut->chunkNativeLimit) {
1757 // Requested position already in buffer.
1758 ut->chunkOffset = index32 - (int32_t)ut->chunkNativeStart;
1759 return TRUE;
1760 }
1761 if (index32==0 && ut->chunkNativeStart==0) {
1762 // Request for start, buffer already begins at start.
1763 // No data, but keep the buffer as is.
1764 ut->chunkOffset = 0;
1765 return FALSE;
1766 }
1768 // Figure out the bounds of the chunk to extract for reverse iteration.
1769 // Need to worry about chunk not splitting surrogate pairs, and while still
1770 // containing the data we need.
1771 // Fix by requesting a chunk that includes an extra UChar at the end.
1772 // If this turns out to be a lead surrogate, we can lop it off and still have
1773 // the data we wanted.
1774 ut->chunkNativeStart = index32 + 1 - REP_TEXT_CHUNK_SIZE;
1775 if (ut->chunkNativeStart < 0) {
1776 ut->chunkNativeStart = 0;
1777 }
1779 ut->chunkNativeLimit = index32 + 1;
1780 if (ut->chunkNativeLimit > length) {
1781 ut->chunkNativeLimit = length;
1782 }
1783 }
1785 // Extract the new chunk of text from the Replaceable source.
1786 ReplExtra *ex = (ReplExtra *)ut->pExtra;
1787 // UnicodeString with its buffer a writable alias to the chunk buffer
1788 UnicodeString buffer(ex->s, 0 /*buffer length*/, REP_TEXT_CHUNK_SIZE /*buffer capacity*/);
1789 rep->extractBetween((int32_t)ut->chunkNativeStart, (int32_t)ut->chunkNativeLimit, buffer);
1791 ut->chunkContents = ex->s;
1792 ut->chunkLength = (int32_t)(ut->chunkNativeLimit - ut->chunkNativeStart);
1793 ut->chunkOffset = (int32_t)(index32 - ut->chunkNativeStart);
1795 // Surrogate pairs from the input text must not span chunk boundaries.
1796 // If end of chunk could be the start of a surrogate, trim it off.
1797 if (ut->chunkNativeLimit < length &&
1798 U16_IS_LEAD(ex->s[ut->chunkLength-1])) {
1799 ut->chunkLength--;
1800 ut->chunkNativeLimit--;
1801 if (ut->chunkOffset > ut->chunkLength) {
1802 ut->chunkOffset = ut->chunkLength;
1803 }
1804 }
1806 // if the first UChar in the chunk could be the trailing half of a surrogate pair,
1807 // trim it off.
1808 if(ut->chunkNativeStart>0 && U16_IS_TRAIL(ex->s[0])) {
1809 ++(ut->chunkContents);
1810 ++(ut->chunkNativeStart);
1811 --(ut->chunkLength);
1812 --(ut->chunkOffset);
1813 }
1815 // adjust the index/chunkOffset to a code point boundary
1816 U16_SET_CP_START(ut->chunkContents, 0, ut->chunkOffset);
1818 // Use fast indexing for get/setNativeIndex()
1819 ut->nativeIndexingLimit = ut->chunkLength;
1821 return TRUE;
1822 }
1826 static int32_t U_CALLCONV
1827 repTextExtract(UText *ut,
1828 int64_t start, int64_t limit,
1829 UChar *dest, int32_t destCapacity,
1830 UErrorCode *status) {
1831 const Replaceable *rep=(const Replaceable *)ut->context;
1832 int32_t length=rep->length();
1834 if(U_FAILURE(*status)) {
1835 return 0;
1836 }
1837 if(destCapacity<0 || (dest==NULL && destCapacity>0)) {
1838 *status=U_ILLEGAL_ARGUMENT_ERROR;
1839 }
1840 if(start>limit) {
1841 *status=U_INDEX_OUTOFBOUNDS_ERROR;
1842 return 0;
1843 }
1845 int32_t start32 = pinIndex(start, length);
1846 int32_t limit32 = pinIndex(limit, length);
1848 // adjust start, limit if they point to trail half of surrogates
1849 if (start32<length && U16_IS_TRAIL(rep->charAt(start32)) &&
1850 U_IS_SUPPLEMENTARY(rep->char32At(start32))){
1851 start32--;
1852 }
1853 if (limit32<length && U16_IS_TRAIL(rep->charAt(limit32)) &&
1854 U_IS_SUPPLEMENTARY(rep->char32At(limit32))){
1855 limit32--;
1856 }
1858 length=limit32-start32;
1859 if(length>destCapacity) {
1860 limit32 = start32 + destCapacity;
1861 }
1862 UnicodeString buffer(dest, 0, destCapacity); // writable alias
1863 rep->extractBetween(start32, limit32, buffer);
1864 repTextAccess(ut, limit32, TRUE);
1866 return u_terminateUChars(dest, destCapacity, length, status);
1867 }
1869 static int32_t U_CALLCONV
1870 repTextReplace(UText *ut,
1871 int64_t start, int64_t limit,
1872 const UChar *src, int32_t length,
1873 UErrorCode *status) {
1874 Replaceable *rep=(Replaceable *)ut->context;
1875 int32_t oldLength;
1877 if(U_FAILURE(*status)) {
1878 return 0;
1879 }
1880 if(src==NULL && length!=0) {
1881 *status=U_ILLEGAL_ARGUMENT_ERROR;
1882 return 0;
1883 }
1884 oldLength=rep->length(); // will subtract from new length
1885 if(start>limit ) {
1886 *status=U_INDEX_OUTOFBOUNDS_ERROR;
1887 return 0;
1888 }
1890 int32_t start32 = pinIndex(start, oldLength);
1891 int32_t limit32 = pinIndex(limit, oldLength);
1893 // Snap start & limit to code point boundaries.
1894 if (start32<oldLength && U16_IS_TRAIL(rep->charAt(start32)) &&
1895 start32>0 && U16_IS_LEAD(rep->charAt(start32-1)))
1896 {
1897 start32--;
1898 }
1899 if (limit32<oldLength && U16_IS_LEAD(rep->charAt(limit32-1)) &&
1900 U16_IS_TRAIL(rep->charAt(limit32)))
1901 {
1902 limit32++;
1903 }
1905 // Do the actual replace operation using methods of the Replaceable class
1906 UnicodeString replStr((UBool)(length<0), src, length); // read-only alias
1907 rep->handleReplaceBetween(start32, limit32, replStr);
1908 int32_t newLength = rep->length();
1909 int32_t lengthDelta = newLength - oldLength;
1911 // Is the UText chunk buffer OK?
1912 if (ut->chunkNativeLimit > start32) {
1913 // this replace operation may have impacted the current chunk.
1914 // invalidate it, which will force a reload on the next access.
1915 invalidateChunk(ut);
1916 }
1918 // set the iteration position to the end of the newly inserted replacement text.
1919 int32_t newIndexPos = limit32 + lengthDelta;
1920 repTextAccess(ut, newIndexPos, TRUE);
1922 return lengthDelta;
1923 }
1926 static void U_CALLCONV
1927 repTextCopy(UText *ut,
1928 int64_t start, int64_t limit,
1929 int64_t destIndex,
1930 UBool move,
1931 UErrorCode *status)
1932 {
1933 Replaceable *rep=(Replaceable *)ut->context;
1934 int32_t length=rep->length();
1936 if(U_FAILURE(*status)) {
1937 return;
1938 }
1939 if (start>limit || (start<destIndex && destIndex<limit))
1940 {
1941 *status=U_INDEX_OUTOFBOUNDS_ERROR;
1942 return;
1943 }
1945 int32_t start32 = pinIndex(start, length);
1946 int32_t limit32 = pinIndex(limit, length);
1947 int32_t destIndex32 = pinIndex(destIndex, length);
1949 // TODO: snap input parameters to code point boundaries.
1951 if(move) {
1952 // move: copy to destIndex, then replace original with nothing
1953 int32_t segLength=limit32-start32;
1954 rep->copy(start32, limit32, destIndex32);
1955 if(destIndex32<start32) {
1956 start32+=segLength;
1957 limit32+=segLength;
1958 }
1959 rep->handleReplaceBetween(start32, limit32, UnicodeString());
1960 } else {
1961 // copy
1962 rep->copy(start32, limit32, destIndex32);
1963 }
1965 // If the change to the text touched the region in the chunk buffer,
1966 // invalidate the buffer.
1967 int32_t firstAffectedIndex = destIndex32;
1968 if (move && start32<firstAffectedIndex) {
1969 firstAffectedIndex = start32;
1970 }
1971 if (firstAffectedIndex < ut->chunkNativeLimit) {
1972 // changes may have affected range covered by the chunk
1973 invalidateChunk(ut);
1974 }
1976 // Put iteration position at the newly inserted (moved) block,
1977 int32_t nativeIterIndex = destIndex32 + limit32 - start32;
1978 if (move && destIndex32>start32) {
1979 // moved a block of text towards the end of the string.
1980 nativeIterIndex = destIndex32;
1981 }
1983 // Set position, reload chunk if needed.
1984 repTextAccess(ut, nativeIterIndex, TRUE);
1985 }
1987 static const struct UTextFuncs repFuncs =
1988 {
1989 sizeof(UTextFuncs),
1990 0, 0, 0, // Reserved alignment padding
1991 repTextClone,
1992 repTextLength,
1993 repTextAccess,
1994 repTextExtract,
1995 repTextReplace,
1996 repTextCopy,
1997 NULL, // MapOffsetToNative,
1998 NULL, // MapIndexToUTF16,
1999 repTextClose,
2000 NULL, // spare 1
2001 NULL, // spare 2
2002 NULL // spare 3
2003 };
2006 U_CAPI UText * U_EXPORT2
2007 utext_openReplaceable(UText *ut, Replaceable *rep, UErrorCode *status)
2008 {
2009 if(U_FAILURE(*status)) {
2010 return NULL;
2011 }
2012 if(rep==NULL) {
2013 *status=U_ILLEGAL_ARGUMENT_ERROR;
2014 return NULL;
2015 }
2016 ut = utext_setup(ut, sizeof(ReplExtra), status);
2018 ut->providerProperties = I32_FLAG(UTEXT_PROVIDER_WRITABLE);
2019 if(rep->hasMetaData()) {
2020 ut->providerProperties |=I32_FLAG(UTEXT_PROVIDER_HAS_META_DATA);
2021 }
2023 ut->pFuncs = &repFuncs;
2024 ut->context = rep;
2025 return ut;
2026 }
2028 U_CDECL_END
2037 //------------------------------------------------------------------------------
2038 //
2039 // UText implementation for UnicodeString (read/write) and
2040 // for const UnicodeString (read only)
2041 // (same implementation, only the flags are different)
2042 //
2043 // Use of UText data members:
2044 // context pointer to UnicodeString
2045 // p pointer to UnicodeString IF this UText owns the string
2046 // and it must be deleted on close(). NULL otherwise.
2047 //
2048 //------------------------------------------------------------------------------
2050 U_CDECL_BEGIN
2053 static UText * U_CALLCONV
2054 unistrTextClone(UText *dest, const UText *src, UBool deep, UErrorCode *status) {
2055 // First do a generic shallow clone. Does everything needed for the UText struct itself.
2056 dest = shallowTextClone(dest, src, status);
2058 // For deep clones, make a copy of the UnicodeSring.
2059 // The copied UnicodeString storage is owned by the newly created UText clone.
2060 // A non-NULL pointer in UText.p is the signal to the close() function to delete
2061 // the UText.
2062 //
2063 if (deep && U_SUCCESS(*status)) {
2064 const UnicodeString *srcString = (const UnicodeString *)src->context;
2065 dest->context = new UnicodeString(*srcString);
2066 dest->providerProperties |= I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT);
2068 // with deep clone, the copy is writable, even when the source is not.
2069 dest->providerProperties |= I32_FLAG(UTEXT_PROVIDER_WRITABLE);
2070 }
2071 return dest;
2072 }
2074 static void U_CALLCONV
2075 unistrTextClose(UText *ut) {
2076 // Most of the work of close is done by the generic UText framework close.
2077 // All that needs to be done here is delete the UnicodeString if the UText
2078 // owns it. This occurs if the UText was created by cloning.
2079 if (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT)) {
2080 UnicodeString *str = (UnicodeString *)ut->context;
2081 delete str;
2082 ut->context = NULL;
2083 }
2084 }
2087 static int64_t U_CALLCONV
2088 unistrTextLength(UText *t) {
2089 return ((const UnicodeString *)t->context)->length();
2090 }
2093 static UBool U_CALLCONV
2094 unistrTextAccess(UText *ut, int64_t index, UBool forward) {
2095 int32_t length = ut->chunkLength;
2096 ut->chunkOffset = pinIndex(index, length);
2098 // Check whether request is at the start or end
2099 UBool retVal = (forward && index<length) || (!forward && index>0);
2100 return retVal;
2101 }
2105 static int32_t U_CALLCONV
2106 unistrTextExtract(UText *t,
2107 int64_t start, int64_t limit,
2108 UChar *dest, int32_t destCapacity,
2109 UErrorCode *pErrorCode) {
2110 const UnicodeString *us=(const UnicodeString *)t->context;
2111 int32_t length=us->length();
2113 if(U_FAILURE(*pErrorCode)) {
2114 return 0;
2115 }
2116 if(destCapacity<0 || (dest==NULL && destCapacity>0)) {
2117 *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
2118 }
2119 if(start<0 || start>limit) {
2120 *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR;
2121 return 0;
2122 }
2124 int32_t start32 = start<length ? us->getChar32Start((int32_t)start) : length;
2125 int32_t limit32 = limit<length ? us->getChar32Start((int32_t)limit) : length;
2127 length=limit32-start32;
2128 if (destCapacity>0 && dest!=NULL) {
2129 int32_t trimmedLength = length;
2130 if(trimmedLength>destCapacity) {
2131 trimmedLength=destCapacity;
2132 }
2133 us->extract(start32, trimmedLength, dest);
2134 t->chunkOffset = start32+trimmedLength;
2135 } else {
2136 t->chunkOffset = start32;
2137 }
2138 u_terminateUChars(dest, destCapacity, length, pErrorCode);
2139 return length;
2140 }
2142 static int32_t U_CALLCONV
2143 unistrTextReplace(UText *ut,
2144 int64_t start, int64_t limit,
2145 const UChar *src, int32_t length,
2146 UErrorCode *pErrorCode) {
2147 UnicodeString *us=(UnicodeString *)ut->context;
2148 int32_t oldLength;
2150 if(U_FAILURE(*pErrorCode)) {
2151 return 0;
2152 }
2153 if(src==NULL && length!=0) {
2154 *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
2155 }
2156 if(start>limit) {
2157 *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR;
2158 return 0;
2159 }
2160 oldLength=us->length();
2161 int32_t start32 = pinIndex(start, oldLength);
2162 int32_t limit32 = pinIndex(limit, oldLength);
2163 if (start32 < oldLength) {
2164 start32 = us->getChar32Start(start32);
2165 }
2166 if (limit32 < oldLength) {
2167 limit32 = us->getChar32Start(limit32);
2168 }
2170 // replace
2171 us->replace(start32, limit32-start32, src, length);
2172 int32_t newLength = us->length();
2174 // Update the chunk description.
2175 ut->chunkContents = us->getBuffer();
2176 ut->chunkLength = newLength;
2177 ut->chunkNativeLimit = newLength;
2178 ut->nativeIndexingLimit = newLength;
2180 // Set iteration position to the point just following the newly inserted text.
2181 int32_t lengthDelta = newLength - oldLength;
2182 ut->chunkOffset = limit32 + lengthDelta;
2184 return lengthDelta;
2185 }
2187 static void U_CALLCONV
2188 unistrTextCopy(UText *ut,
2189 int64_t start, int64_t limit,
2190 int64_t destIndex,
2191 UBool move,
2192 UErrorCode *pErrorCode) {
2193 UnicodeString *us=(UnicodeString *)ut->context;
2194 int32_t length=us->length();
2196 if(U_FAILURE(*pErrorCode)) {
2197 return;
2198 }
2199 int32_t start32 = pinIndex(start, length);
2200 int32_t limit32 = pinIndex(limit, length);
2201 int32_t destIndex32 = pinIndex(destIndex, length);
2203 if( start32>limit32 || (start32<destIndex32 && destIndex32<limit32)) {
2204 *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR;
2205 return;
2206 }
2208 if(move) {
2209 // move: copy to destIndex, then replace original with nothing
2210 int32_t segLength=limit32-start32;
2211 us->copy(start32, limit32, destIndex32);
2212 if(destIndex32<start32) {
2213 start32+=segLength;
2214 }
2215 us->replace(start32, segLength, NULL, 0);
2216 } else {
2217 // copy
2218 us->copy(start32, limit32, destIndex32);
2219 }
2221 // update chunk description, set iteration position.
2222 ut->chunkContents = us->getBuffer();
2223 if (move==FALSE) {
2224 // copy operation, string length grows
2225 ut->chunkLength += limit32-start32;
2226 ut->chunkNativeLimit = ut->chunkLength;
2227 ut->nativeIndexingLimit = ut->chunkLength;
2228 }
2230 // Iteration position to end of the newly inserted text.
2231 ut->chunkOffset = destIndex32+limit32-start32;
2232 if (move && destIndex32>start32) {
2233 ut->chunkOffset = destIndex32;
2234 }
2236 }
2238 static const struct UTextFuncs unistrFuncs =
2239 {
2240 sizeof(UTextFuncs),
2241 0, 0, 0, // Reserved alignment padding
2242 unistrTextClone,
2243 unistrTextLength,
2244 unistrTextAccess,
2245 unistrTextExtract,
2246 unistrTextReplace,
2247 unistrTextCopy,
2248 NULL, // MapOffsetToNative,
2249 NULL, // MapIndexToUTF16,
2250 unistrTextClose,
2251 NULL, // spare 1
2252 NULL, // spare 2
2253 NULL // spare 3
2254 };
2258 U_CDECL_END
2261 U_CAPI UText * U_EXPORT2
2262 utext_openUnicodeString(UText *ut, UnicodeString *s, UErrorCode *status) {
2263 ut = utext_openConstUnicodeString(ut, s, status);
2264 if (U_SUCCESS(*status)) {
2265 ut->providerProperties |= I32_FLAG(UTEXT_PROVIDER_WRITABLE);
2266 }
2267 return ut;
2268 }
2272 U_CAPI UText * U_EXPORT2
2273 utext_openConstUnicodeString(UText *ut, const UnicodeString *s, UErrorCode *status) {
2274 if (U_SUCCESS(*status) && s->isBogus()) {
2275 // The UnicodeString is bogus, but we still need to detach the UText
2276 // from whatever it was hooked to before, if anything.
2277 utext_openUChars(ut, NULL, 0, status);
2278 *status = U_ILLEGAL_ARGUMENT_ERROR;
2279 return ut;
2280 }
2281 ut = utext_setup(ut, 0, status);
2282 // note: use the standard (writable) function table for UnicodeString.
2283 // The flag settings disable writing, so having the functions in
2284 // the table is harmless.
2285 if (U_SUCCESS(*status)) {
2286 ut->pFuncs = &unistrFuncs;
2287 ut->context = s;
2288 ut->providerProperties = I32_FLAG(UTEXT_PROVIDER_STABLE_CHUNKS);
2289 ut->chunkContents = s->getBuffer();
2290 ut->chunkLength = s->length();
2291 ut->chunkNativeStart = 0;
2292 ut->chunkNativeLimit = ut->chunkLength;
2293 ut->nativeIndexingLimit = ut->chunkLength;
2294 }
2295 return ut;
2296 }
2298 //------------------------------------------------------------------------------
2299 //
2300 // UText implementation for const UChar * strings
2301 //
2302 // Use of UText data members:
2303 // context pointer to UnicodeString
2304 // a length. -1 if not yet known.
2305 //
2306 // TODO: support 64 bit lengths.
2307 //
2308 //------------------------------------------------------------------------------
2310 U_CDECL_BEGIN
2313 static UText * U_CALLCONV
2314 ucstrTextClone(UText *dest, const UText * src, UBool deep, UErrorCode * status) {
2315 // First do a generic shallow clone.
2316 dest = shallowTextClone(dest, src, status);
2318 // For deep clones, make a copy of the string.
2319 // The copied storage is owned by the newly created clone.
2320 // A non-NULL pointer in UText.p is the signal to the close() function to delete
2321 // it.
2322 //
2323 if (deep && U_SUCCESS(*status)) {
2324 U_ASSERT(utext_nativeLength(dest) < INT32_MAX);
2325 int32_t len = (int32_t)utext_nativeLength(dest);
2327 // The cloned string IS going to be NUL terminated, whether or not the original was.
2328 const UChar *srcStr = (const UChar *)src->context;
2329 UChar *copyStr = (UChar *)uprv_malloc((len+1) * sizeof(UChar));
2330 if (copyStr == NULL) {
2331 *status = U_MEMORY_ALLOCATION_ERROR;
2332 } else {
2333 int64_t i;
2334 for (i=0; i<len; i++) {
2335 copyStr[i] = srcStr[i];
2336 }
2337 copyStr[len] = 0;
2338 dest->context = copyStr;
2339 dest->providerProperties |= I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT);
2340 }
2341 }
2342 return dest;
2343 }
2346 static void U_CALLCONV
2347 ucstrTextClose(UText *ut) {
2348 // Most of the work of close is done by the generic UText framework close.
2349 // All that needs to be done here is delete the string if the UText
2350 // owns it. This occurs if the UText was created by cloning.
2351 if (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT)) {
2352 UChar *s = (UChar *)ut->context;
2353 uprv_free(s);
2354 ut->context = NULL;
2355 }
2356 }
2360 static int64_t U_CALLCONV
2361 ucstrTextLength(UText *ut) {
2362 if (ut->a < 0) {
2363 // null terminated, we don't yet know the length. Scan for it.
2364 // Access is not convenient for doing this
2365 // because the current interation postion can't be changed.
2366 const UChar *str = (const UChar *)ut->context;
2367 for (;;) {
2368 if (str[ut->chunkNativeLimit] == 0) {
2369 break;
2370 }
2371 ut->chunkNativeLimit++;
2372 }
2373 ut->a = ut->chunkNativeLimit;
2374 ut->chunkLength = (int32_t)ut->chunkNativeLimit;
2375 ut->nativeIndexingLimit = ut->chunkLength;
2376 ut->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
2377 }
2378 return ut->a;
2379 }
2382 static UBool U_CALLCONV
2383 ucstrTextAccess(UText *ut, int64_t index, UBool forward) {
2384 const UChar *str = (const UChar *)ut->context;
2386 // pin the requested index to the bounds of the string,
2387 // and set current iteration position.
2388 if (index<0) {
2389 index = 0;
2390 } else if (index < ut->chunkNativeLimit) {
2391 // The request data is within the chunk as it is known so far.
2392 // Put index on a code point boundary.
2393 U16_SET_CP_START(str, 0, index);
2394 } else if (ut->a >= 0) {
2395 // We know the length of this string, and the user is requesting something
2396 // at or beyond the length. Pin the requested index to the length.
2397 index = ut->a;
2398 } else {
2399 // Null terminated string, length not yet known, and the requested index
2400 // is beyond where we have scanned so far.
2401 // Scan to 32 UChars beyond the requested index. The strategy here is
2402 // to avoid fully scanning a long string when the caller only wants to
2403 // see a few characters at its beginning.
2404 int32_t scanLimit = (int32_t)index + 32;
2405 if ((index + 32)>INT32_MAX || (index + 32)<0 ) { // note: int64 expression
2406 scanLimit = INT32_MAX;
2407 }
2409 int32_t chunkLimit = (int32_t)ut->chunkNativeLimit;
2410 for (; chunkLimit<scanLimit; chunkLimit++) {
2411 if (str[chunkLimit] == 0) {
2412 // We found the end of the string. Remember it, pin the requested index to it,
2413 // and bail out of here.
2414 ut->a = chunkLimit;
2415 ut->chunkLength = chunkLimit;
2416 ut->nativeIndexingLimit = chunkLimit;
2417 if (index >= chunkLimit) {
2418 index = chunkLimit;
2419 } else {
2420 U16_SET_CP_START(str, 0, index);
2421 }
2423 ut->chunkNativeLimit = chunkLimit;
2424 ut->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
2425 goto breakout;
2426 }
2427 }
2428 // We scanned through the next batch of UChars without finding the end.
2429 U16_SET_CP_START(str, 0, index);
2430 if (chunkLimit == INT32_MAX) {
2431 // Scanned to the limit of a 32 bit length.
2432 // Forceably trim the overlength string back so length fits in int32
2433 // TODO: add support for 64 bit strings.
2434 ut->a = chunkLimit;
2435 ut->chunkLength = chunkLimit;
2436 ut->nativeIndexingLimit = chunkLimit;
2437 if (index > chunkLimit) {
2438 index = chunkLimit;
2439 }
2440 ut->chunkNativeLimit = chunkLimit;
2441 ut->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
2442 } else {
2443 // The endpoint of a chunk must not be left in the middle of a surrogate pair.
2444 // If the current end is on a lead surrogate, back the end up by one.
2445 // It doesn't matter if the end char happens to be an unpaired surrogate,
2446 // and it's simpler not to worry about it.
2447 if (U16_IS_LEAD(str[chunkLimit-1])) {
2448 --chunkLimit;
2449 }
2450 // Null-terminated chunk with end still unknown.
2451 // Update the chunk length to reflect what has been scanned thus far.
2452 // That the full length is still unknown is (still) flagged by
2453 // ut->a being < 0.
2454 ut->chunkNativeLimit = chunkLimit;
2455 ut->nativeIndexingLimit = chunkLimit;
2456 ut->chunkLength = chunkLimit;
2457 }
2459 }
2460 breakout:
2461 U_ASSERT(index<=INT32_MAX);
2462 ut->chunkOffset = (int32_t)index;
2464 // Check whether request is at the start or end
2465 UBool retVal = (forward && index<ut->chunkNativeLimit) || (!forward && index>0);
2466 return retVal;
2467 }
2471 static int32_t U_CALLCONV
2472 ucstrTextExtract(UText *ut,
2473 int64_t start, int64_t limit,
2474 UChar *dest, int32_t destCapacity,
2475 UErrorCode *pErrorCode)
2476 {
2477 if(U_FAILURE(*pErrorCode)) {
2478 return 0;
2479 }
2480 if(destCapacity<0 || (dest==NULL && destCapacity>0) || start>limit) {
2481 *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
2482 return 0;
2483 }
2485 //const UChar *s=(const UChar *)ut->context;
2486 int32_t si, di;
2488 int32_t start32;
2489 int32_t limit32;
2491 // Access the start. Does two things we need:
2492 // Pins 'start' to the length of the string, if it came in out-of-bounds.
2493 // Snaps 'start' to the beginning of a code point.
2494 ucstrTextAccess(ut, start, TRUE);
2495 const UChar *s=ut->chunkContents;
2496 start32 = ut->chunkOffset;
2498 int32_t strLength=(int32_t)ut->a;
2499 if (strLength >= 0) {
2500 limit32 = pinIndex(limit, strLength);
2501 } else {
2502 limit32 = pinIndex(limit, INT32_MAX);
2503 }
2504 di = 0;
2505 for (si=start32; si<limit32; si++) {
2506 if (strLength<0 && s[si]==0) {
2507 // Just hit the end of a null-terminated string.
2508 ut->a = si; // set string length for this UText
2509 ut->chunkNativeLimit = si;
2510 ut->chunkLength = si;
2511 ut->nativeIndexingLimit = si;
2512 strLength = si;
2513 break;
2514 }
2515 U_ASSERT(di>=0); /* to ensure di never exceeds INT32_MAX, which must not happen logically */
2516 if (di<destCapacity) {
2517 // only store if there is space.
2518 dest[di] = s[si];
2519 } else {
2520 if (strLength>=0) {
2521 // We have filled the destination buffer, and the string length is known.
2522 // Cut the loop short. There is no need to scan string termination.
2523 di = limit32 - start32;
2524 si = limit32;
2525 break;
2526 }
2527 }
2528 di++;
2529 }
2531 // If the limit index points to a lead surrogate of a pair,
2532 // add the corresponding trail surrogate to the destination.
2533 if (si>0 && U16_IS_LEAD(s[si-1]) &&
2534 ((si<strLength || strLength<0) && U16_IS_TRAIL(s[si])))
2535 {
2536 if (di<destCapacity) {
2537 // store only if there is space in the output buffer.
2538 dest[di++] = s[si++];
2539 }
2540 }
2542 // Put iteration position at the point just following the extracted text
2543 ut->chunkOffset = uprv_min(strLength, start32 + destCapacity);
2545 // Add a terminating NUL if space in the buffer permits,
2546 // and set the error status as required.
2547 u_terminateUChars(dest, destCapacity, di, pErrorCode);
2548 return di;
2549 }
2551 static const struct UTextFuncs ucstrFuncs =
2552 {
2553 sizeof(UTextFuncs),
2554 0, 0, 0, // Reserved alignment padding
2555 ucstrTextClone,
2556 ucstrTextLength,
2557 ucstrTextAccess,
2558 ucstrTextExtract,
2559 NULL, // Replace
2560 NULL, // Copy
2561 NULL, // MapOffsetToNative,
2562 NULL, // MapIndexToUTF16,
2563 ucstrTextClose,
2564 NULL, // spare 1
2565 NULL, // spare 2
2566 NULL, // spare 3
2567 };
2569 U_CDECL_END
2571 static const UChar gEmptyUString[] = {0};
2573 U_CAPI UText * U_EXPORT2
2574 utext_openUChars(UText *ut, const UChar *s, int64_t length, UErrorCode *status) {
2575 if (U_FAILURE(*status)) {
2576 return NULL;
2577 }
2578 if(s==NULL && length==0) {
2579 s = gEmptyUString;
2580 }
2581 if (s==NULL || length < -1 || length>INT32_MAX) {
2582 *status = U_ILLEGAL_ARGUMENT_ERROR;
2583 return NULL;
2584 }
2585 ut = utext_setup(ut, 0, status);
2586 if (U_SUCCESS(*status)) {
2587 ut->pFuncs = &ucstrFuncs;
2588 ut->context = s;
2589 ut->providerProperties = I32_FLAG(UTEXT_PROVIDER_STABLE_CHUNKS);
2590 if (length==-1) {
2591 ut->providerProperties |= I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
2592 }
2593 ut->a = length;
2594 ut->chunkContents = s;
2595 ut->chunkNativeStart = 0;
2596 ut->chunkNativeLimit = length>=0? length : 0;
2597 ut->chunkLength = (int32_t)ut->chunkNativeLimit;
2598 ut->chunkOffset = 0;
2599 ut->nativeIndexingLimit = ut->chunkLength;
2600 }
2601 return ut;
2602 }
2605 //------------------------------------------------------------------------------
2606 //
2607 // UText implementation for text from ICU CharacterIterators
2608 //
2609 // Use of UText data members:
2610 // context pointer to the CharacterIterator
2611 // a length of the full text.
2612 // p pointer to buffer 1
2613 // b start index of local buffer 1 contents
2614 // q pointer to buffer 2
2615 // c start index of local buffer 2 contents
2616 // r pointer to the character iterator if the UText owns it.
2617 // Null otherwise.
2618 //
2619 //------------------------------------------------------------------------------
2620 #define CIBufSize 16
2622 U_CDECL_BEGIN
2623 static void U_CALLCONV
2624 charIterTextClose(UText *ut) {
2625 // Most of the work of close is done by the generic UText framework close.
2626 // All that needs to be done here is delete the CharacterIterator if the UText
2627 // owns it. This occurs if the UText was created by cloning.
2628 CharacterIterator *ci = (CharacterIterator *)ut->r;
2629 delete ci;
2630 ut->r = NULL;
2631 }
2633 static int64_t U_CALLCONV
2634 charIterTextLength(UText *ut) {
2635 return (int32_t)ut->a;
2636 }
2638 static UBool U_CALLCONV
2639 charIterTextAccess(UText *ut, int64_t index, UBool forward) {
2640 CharacterIterator *ci = (CharacterIterator *)ut->context;
2642 int32_t clippedIndex = (int32_t)index;
2643 if (clippedIndex<0) {
2644 clippedIndex=0;
2645 } else if (clippedIndex>=ut->a) {
2646 clippedIndex=(int32_t)ut->a;
2647 }
2648 int32_t neededIndex = clippedIndex;
2649 if (!forward && neededIndex>0) {
2650 // reverse iteration, want the position just before what was asked for.
2651 neededIndex--;
2652 } else if (forward && neededIndex==ut->a && neededIndex>0) {
2653 // Forward iteration, don't ask for something past the end of the text.
2654 neededIndex--;
2655 }
2657 // Find the native index of the start of the buffer containing what we want.
2658 neededIndex -= neededIndex % CIBufSize;
2660 UChar *buf = NULL;
2661 UBool needChunkSetup = TRUE;
2662 int i;
2663 if (ut->chunkNativeStart == neededIndex) {
2664 // The buffer we want is already the current chunk.
2665 needChunkSetup = FALSE;
2666 } else if (ut->b == neededIndex) {
2667 // The first buffer (buffer p) has what we need.
2668 buf = (UChar *)ut->p;
2669 } else if (ut->c == neededIndex) {
2670 // The second buffer (buffer q) has what we need.
2671 buf = (UChar *)ut->q;
2672 } else {
2673 // Neither buffer already has what we need.
2674 // Load new data from the character iterator.
2675 // Use the buf that is not the current buffer.
2676 buf = (UChar *)ut->p;
2677 if (ut->p == ut->chunkContents) {
2678 buf = (UChar *)ut->q;
2679 }
2680 ci->setIndex(neededIndex);
2681 for (i=0; i<CIBufSize; i++) {
2682 buf[i] = ci->nextPostInc();
2683 if (i+neededIndex > ut->a) {
2684 break;
2685 }
2686 }
2687 }
2689 // We have a buffer with the data we need.
2690 // Set it up as the current chunk, if it wasn't already.
2691 if (needChunkSetup) {
2692 ut->chunkContents = buf;
2693 ut->chunkLength = CIBufSize;
2694 ut->chunkNativeStart = neededIndex;
2695 ut->chunkNativeLimit = neededIndex + CIBufSize;
2696 if (ut->chunkNativeLimit > ut->a) {
2697 ut->chunkNativeLimit = ut->a;
2698 ut->chunkLength = (int32_t)(ut->chunkNativeLimit)-(int32_t)(ut->chunkNativeStart);
2699 }
2700 ut->nativeIndexingLimit = ut->chunkLength;
2701 U_ASSERT(ut->chunkOffset>=0 && ut->chunkOffset<=CIBufSize);
2702 }
2703 ut->chunkOffset = clippedIndex - (int32_t)ut->chunkNativeStart;
2704 UBool success = (forward? ut->chunkOffset<ut->chunkLength : ut->chunkOffset>0);
2705 return success;
2706 }
2708 static UText * U_CALLCONV
2709 charIterTextClone(UText *dest, const UText *src, UBool deep, UErrorCode * status) {
2710 if (U_FAILURE(*status)) {
2711 return NULL;
2712 }
2714 if (deep) {
2715 // There is no CharacterIterator API for cloning the underlying text storage.
2716 *status = U_UNSUPPORTED_ERROR;
2717 return NULL;
2718 } else {
2719 CharacterIterator *srcCI =(CharacterIterator *)src->context;
2720 srcCI = srcCI->clone();
2721 dest = utext_openCharacterIterator(dest, srcCI, status);
2722 // cast off const on getNativeIndex.
2723 // For CharacterIterator based UTexts, this is safe, the operation is const.
2724 int64_t ix = utext_getNativeIndex((UText *)src);
2725 utext_setNativeIndex(dest, ix);
2726 dest->r = srcCI; // flags that this UText owns the CharacterIterator
2727 }
2728 return dest;
2729 }
2731 static int32_t U_CALLCONV
2732 charIterTextExtract(UText *ut,
2733 int64_t start, int64_t limit,
2734 UChar *dest, int32_t destCapacity,
2735 UErrorCode *status)
2736 {
2737 if(U_FAILURE(*status)) {
2738 return 0;
2739 }
2740 if(destCapacity<0 || (dest==NULL && destCapacity>0) || start>limit) {
2741 *status=U_ILLEGAL_ARGUMENT_ERROR;
2742 return 0;
2743 }
2744 int32_t length = (int32_t)ut->a;
2745 int32_t start32 = pinIndex(start, length);
2746 int32_t limit32 = pinIndex(limit, length);
2747 int32_t desti = 0;
2748 int32_t srci;
2749 int32_t copyLimit;
2751 CharacterIterator *ci = (CharacterIterator *)ut->context;
2752 ci->setIndex32(start32); // Moves ix to lead of surrogate pair, if needed.
2753 srci = ci->getIndex();
2754 copyLimit = srci;
2755 while (srci<limit32) {
2756 UChar32 c = ci->next32PostInc();
2757 int32_t len = U16_LENGTH(c);
2758 U_ASSERT(desti+len>0); /* to ensure desti+len never exceeds MAX_INT32, which must not happen logically */
2759 if (desti+len <= destCapacity) {
2760 U16_APPEND_UNSAFE(dest, desti, c);
2761 copyLimit = srci+len;
2762 } else {
2763 desti += len;
2764 *status = U_BUFFER_OVERFLOW_ERROR;
2765 }
2766 srci += len;
2767 }
2769 charIterTextAccess(ut, copyLimit, TRUE);
2771 u_terminateUChars(dest, destCapacity, desti, status);
2772 return desti;
2773 }
2775 static const struct UTextFuncs charIterFuncs =
2776 {
2777 sizeof(UTextFuncs),
2778 0, 0, 0, // Reserved alignment padding
2779 charIterTextClone,
2780 charIterTextLength,
2781 charIterTextAccess,
2782 charIterTextExtract,
2783 NULL, // Replace
2784 NULL, // Copy
2785 NULL, // MapOffsetToNative,
2786 NULL, // MapIndexToUTF16,
2787 charIterTextClose,
2788 NULL, // spare 1
2789 NULL, // spare 2
2790 NULL // spare 3
2791 };
2792 U_CDECL_END
2795 U_CAPI UText * U_EXPORT2
2796 utext_openCharacterIterator(UText *ut, CharacterIterator *ci, UErrorCode *status) {
2797 if (U_FAILURE(*status)) {
2798 return NULL;
2799 }
2801 if (ci->startIndex() > 0) {
2802 // No support for CharacterIterators that do not start indexing from zero.
2803 *status = U_UNSUPPORTED_ERROR;
2804 return NULL;
2805 }
2807 // Extra space in UText for 2 buffers of CIBufSize UChars each.
2808 int32_t extraSpace = 2 * CIBufSize * sizeof(UChar);
2809 ut = utext_setup(ut, extraSpace, status);
2810 if (U_SUCCESS(*status)) {
2811 ut->pFuncs = &charIterFuncs;
2812 ut->context = ci;
2813 ut->providerProperties = 0;
2814 ut->a = ci->endIndex(); // Length of text
2815 ut->p = ut->pExtra; // First buffer
2816 ut->b = -1; // Native index of first buffer contents
2817 ut->q = (UChar*)ut->pExtra+CIBufSize; // Second buffer
2818 ut->c = -1; // Native index of second buffer contents
2820 // Initialize current chunk contents to be empty.
2821 // First access will fault something in.
2822 // Note: The initial nativeStart and chunkOffset must sum to zero
2823 // so that getNativeIndex() will correctly compute to zero
2824 // if no call to Access() has ever been made. They can't be both
2825 // zero without Access() thinking that the chunk is valid.
2826 ut->chunkContents = (UChar *)ut->p;
2827 ut->chunkNativeStart = -1;
2828 ut->chunkOffset = 1;
2829 ut->chunkNativeLimit = 0;
2830 ut->chunkLength = 0;
2831 ut->nativeIndexingLimit = ut->chunkOffset; // enables native indexing
2832 }
2833 return ut;
2834 }
