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 <cstddef>
20
21#include "unicode/utypes.h"
22#include "unicode/ustring.h"
23#include "unicode/unistr.h"
24#include "unicode/chariter.h"
25#include "unicode/utext.h"
26#include "unicode/utf.h"
27#include "unicode/utf8.h"
28#include "unicode/utf16.h"
29#include "ustr_imp.h"
30#include "cmemory.h"
31#include "cstring.h"
32#include "uassert.h"
33#include "putilimp.h"
34
35U_NAMESPACE_USE
36
37#define I32_FLAG(bitIndex) ((int32_t)1<<(bitIndex))
38
39
40static UBool
41utext_access(UText *ut, int64_t index, UBool forward) {
42 return ut->pFuncs->access(ut, index, forward);
43}
44
45
46
47U_CAPI UBool U_EXPORT2
48utext_moveIndex32(UText *ut, int32_t delta) {
49 UChar32 c;
50 if (delta > 0) {
51 do {
52 if(ut->chunkOffset>=ut->chunkLength && !utext_access(ut, ut->chunkNativeLimit, true)) {
53 return false;
54 }
55 c = ut->chunkContents[ut->chunkOffset];
56 if (U16_IS_SURROGATE(c)) {
57 c = utext_next32(ut);
58 if (c == U_SENTINEL) {
59 return false;
60 }
61 } else {
62 ut->chunkOffset++;
63 }
64 } while(--delta>0);
65
66 } else if (delta<0) {
67 do {
68 if(ut->chunkOffset<=0 && !utext_access(ut, ut->chunkNativeStart, false)) {
69 return false;
70 }
71 c = ut->chunkContents[ut->chunkOffset-1];
72 if (U16_IS_SURROGATE(c)) {
73 c = utext_previous32(ut);
74 if (c == U_SENTINEL) {
75 return false;
76 }
77 } else {
78 ut->chunkOffset--;
79 }
80 } while(++delta<0);
81 }
82
83 return true;
84}
85
86
87U_CAPI int64_t U_EXPORT2
88utext_nativeLength(UText *ut) {
89 return ut->pFuncs->nativeLength(ut);
90}
91
92
93U_CAPI UBool U_EXPORT2
94utext_isLengthExpensive(const UText *ut) {
95 UBool r = (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE)) != 0;
96 return r;
97}
98
99
100U_CAPI int64_t U_EXPORT2
101utext_getNativeIndex(const UText *ut) {
102 if(ut->chunkOffset <= ut->nativeIndexingLimit) {
103 return ut->chunkNativeStart+ut->chunkOffset;
104 } else {
105 return ut->pFuncs->mapOffsetToNative(ut);
106 }
107}
108
109
110U_CAPI void U_EXPORT2
111utext_setNativeIndex(UText *ut, int64_t index) {
112 if(index<ut->chunkNativeStart || index>=ut->chunkNativeLimit) {
113 // The desired position is outside of the current chunk.
114 // Access the new position. Assume a forward iteration from here,
115 // which will also be optimimum for a single random access.
116 // Reverse iterations may suffer slightly.
117 ut->pFuncs->access(ut, index, true);
118 } else if((int32_t)(index - ut->chunkNativeStart) <= ut->nativeIndexingLimit) {
119 // utf-16 indexing.
120 ut->chunkOffset=(int32_t)(index-ut->chunkNativeStart);
121 } else {
122 ut->chunkOffset=ut->pFuncs->mapNativeIndexToUTF16(ut, index);
123 }
124 // The convention is that the index must always be on a code point boundary.
125 // Adjust the index position if it is in the middle of a surrogate pair.
126 if (ut->chunkOffset<ut->chunkLength) {
127 char16_t c= ut->chunkContents[ut->chunkOffset];
128 if (U16_IS_TRAIL(c)) {
129 if (ut->chunkOffset==0) {
130 ut->pFuncs->access(ut, ut->chunkNativeStart, false);
131 }
132 if (ut->chunkOffset>0) {
133 char16_t lead = ut->chunkContents[ut->chunkOffset-1];
134 if (U16_IS_LEAD(lead)) {
135 ut->chunkOffset--;
136 }
137 }
138 }
139 }
140}
141
142
143
144U_CAPI int64_t U_EXPORT2
145utext_getPreviousNativeIndex(UText *ut) {
146 //
147 // Fast-path the common case.
148 // Common means current position is not at the beginning of a chunk
149 // and the preceding character is not supplementary.
150 //
151 int32_t i = ut->chunkOffset - 1;
152 int64_t result;
153 if (i >= 0) {
154 char16_t c = ut->chunkContents[i];
155 if (U16_IS_TRAIL(c) == false) {
156 if (i <= ut->nativeIndexingLimit) {
157 result = ut->chunkNativeStart + i;
158 } else {
159 ut->chunkOffset = i;
160 result = ut->pFuncs->mapOffsetToNative(ut);
161 ut->chunkOffset++;
162 }
163 return result;
164 }
165 }
166
167 // If at the start of text, simply return 0.
168 if (ut->chunkOffset==0 && ut->chunkNativeStart==0) {
169 return 0;
170 }
171
172 // Harder, less common cases. We are at a chunk boundary, or on a surrogate.
173 // Keep it simple, use other functions to handle the edges.
174 //
175 utext_previous32(ut);
176 result = UTEXT_GETNATIVEINDEX(ut);
177 utext_next32(ut);
178 return result;
179}
180
181
182//
183// utext_current32. Get the UChar32 at the current position.
184// UText iteration position is always on a code point boundary,
185// never on the trail half of a surrogate pair.
186//
187U_CAPI UChar32 U_EXPORT2
188utext_current32(UText *ut) {
189 UChar32 c;
190 if (ut->chunkOffset==ut->chunkLength) {
191 // Current position is just off the end of the chunk.
192 if (ut->pFuncs->access(ut, ut->chunkNativeLimit, true) == false) {
193 // Off the end of the text.
194 return U_SENTINEL;
195 }
196 }
197
198 c = ut->chunkContents[ut->chunkOffset];
199 if (U16_IS_LEAD(c) == false) {
200 // Normal, non-supplementary case.
201 return c;
202 }
203
204 //
205 // Possible supplementary char.
206 //
207 UChar32 trail = 0;
208 UChar32 supplementaryC = c;
209 if ((ut->chunkOffset+1) < ut->chunkLength) {
210 // The trail surrogate is in the same chunk.
211 trail = ut->chunkContents[ut->chunkOffset+1];
212 } else {
213 // The trail surrogate is in a different chunk.
214 // Because we must maintain the iteration position, we need to switch forward
215 // into the new chunk, get the trail surrogate, then revert the chunk back to the
216 // original one.
217 // An edge case to be careful of: the entire text may end with an unpaired
218 // leading surrogate. The attempt to access the trail will fail, but
219 // the original position before the unpaired lead still needs to be restored.
220 int64_t nativePosition = ut->chunkNativeLimit;
221 if (ut->pFuncs->access(ut, nativePosition, true)) {
222 trail = ut->chunkContents[ut->chunkOffset];
223 }
224 UBool r = ut->pFuncs->access(ut, nativePosition, false); // reverse iteration flag loads preceding chunk
225 U_ASSERT(r);
226 // Here we need to restore chunkOffset since the access functions were called with
227 // chunkNativeLimit but that is not where we were (we were 1 code unit before the
228 // limit). Restoring was originally added in ICU-4669 but did not support access
229 // functions that changed the chunk size, the following does.
230 ut->chunkOffset = ut->chunkLength - 1;
231 if(!r) {
232 return U_SENTINEL;
233 }
234 }
235
236 if (U16_IS_TRAIL(trail)) {
237 supplementaryC = U16_GET_SUPPLEMENTARY(c, trail);
238 }
239 return supplementaryC;
240
241}
242
243
244U_CAPI UChar32 U_EXPORT2
245utext_char32At(UText *ut, int64_t nativeIndex) {
246 UChar32 c = U_SENTINEL;
247
248 // Fast path the common case.
249 if (nativeIndex>=ut->chunkNativeStart && nativeIndex < ut->chunkNativeStart + ut->nativeIndexingLimit) {
250 ut->chunkOffset = (int32_t)(nativeIndex - ut->chunkNativeStart);
251 c = ut->chunkContents[ut->chunkOffset];
252 if (U16_IS_SURROGATE(c) == false) {
253 return c;
254 }
255 }
256
257
258 utext_setNativeIndex(ut, nativeIndex);
259 if (nativeIndex>=ut->chunkNativeStart && ut->chunkOffset<ut->chunkLength) {
260 c = ut->chunkContents[ut->chunkOffset];
261 if (U16_IS_SURROGATE(c)) {
262 // For surrogates, let current32() deal with the complications
263 // of supplementaries that may span chunk boundaries.
264 c = utext_current32(ut);
265 }
266 }
267 return c;
268}
269
270
271U_CAPI UChar32 U_EXPORT2
272utext_next32(UText *ut) {
273 UChar32 c;
274
275 if (ut->chunkOffset >= ut->chunkLength) {
276 if (ut->pFuncs->access(ut, ut->chunkNativeLimit, true) == false) {
277 return U_SENTINEL;
278 }
279 }
280
281 c = ut->chunkContents[ut->chunkOffset++];
282 if (U16_IS_LEAD(c) == false) {
283 // Normal case, not supplementary.
284 // (A trail surrogate seen here is just returned as is, as a surrogate value.
285 // It cannot be part of a pair.)
286 return c;
287 }
288
289 if (ut->chunkOffset >= ut->chunkLength) {
290 if (ut->pFuncs->access(ut, ut->chunkNativeLimit, true) == false) {
291 // c is an unpaired lead surrogate at the end of the text.
292 // return it as it is.
293 return c;
294 }
295 }
296 UChar32 trail = ut->chunkContents[ut->chunkOffset];
297 if (U16_IS_TRAIL(trail) == false) {
298 // c was an unpaired lead surrogate, not at the end of the text.
299 // return it as it is (unpaired). Iteration position is on the
300 // following character, possibly in the next chunk, where the
301 // trail surrogate would have been if it had existed.
302 return c;
303 }
304
305 UChar32 supplementary = U16_GET_SUPPLEMENTARY(c, trail);
306 ut->chunkOffset++; // move iteration position over the trail surrogate.
307 return supplementary;
308 }
309
310
311U_CAPI UChar32 U_EXPORT2
312utext_previous32(UText *ut) {
313 UChar32 c;
314
315 if (ut->chunkOffset <= 0) {
316 if (ut->pFuncs->access(ut, ut->chunkNativeStart, false) == false) {
317 return U_SENTINEL;
318 }
319 }
320 ut->chunkOffset--;
321 c = ut->chunkContents[ut->chunkOffset];
322 if (U16_IS_TRAIL(c) == false) {
323 // Normal case, not supplementary.
324 // (A lead surrogate seen here is just returned as is, as a surrogate value.
325 // It cannot be part of a pair.)
326 return c;
327 }
328
329 if (ut->chunkOffset <= 0) {
330 if (ut->pFuncs->access(ut, ut->chunkNativeStart, false) == false) {
331 // c is an unpaired trail surrogate at the start of the text.
332 // return it as it is.
333 return c;
334 }
335 }
336
337 UChar32 lead = ut->chunkContents[ut->chunkOffset-1];
338 if (U16_IS_LEAD(lead) == false) {
339 // c was an unpaired trail surrogate, not at the end of the text.
340 // return it as it is (unpaired). Iteration position is at c
341 return c;
342 }
343
344 UChar32 supplementary = U16_GET_SUPPLEMENTARY(lead, c);
345 ut->chunkOffset--; // move iteration position over the lead surrogate.
346 return supplementary;
347}
348
349
350
351U_CAPI UChar32 U_EXPORT2
352utext_next32From(UText *ut, int64_t index) {
353 UChar32 c = U_SENTINEL;
354
355 if(index<ut->chunkNativeStart || index>=ut->chunkNativeLimit) {
356 // Desired position is outside of the current chunk.
357 if(!ut->pFuncs->access(ut, index, true)) {
358 // no chunk available here
359 return U_SENTINEL;
360 }
361 } else if (index - ut->chunkNativeStart <= (int64_t)ut->nativeIndexingLimit) {
362 // Desired position is in chunk, with direct 1:1 native to UTF16 indexing
363 ut->chunkOffset = (int32_t)(index - ut->chunkNativeStart);
364 } else {
365 // Desired position is in chunk, with non-UTF16 indexing.
366 ut->chunkOffset = ut->pFuncs->mapNativeIndexToUTF16(ut, index);
367 }
368
369 c = ut->chunkContents[ut->chunkOffset++];
370 if (U16_IS_SURROGATE(c)) {
371 // Surrogates. Many edge cases. Use other functions that already
372 // deal with the problems.
373 utext_setNativeIndex(ut, index);
374 c = utext_next32(ut);
375 }
376 return c;
377}
378
379
380U_CAPI UChar32 U_EXPORT2
381utext_previous32From(UText *ut, int64_t index) {
382 //
383 // Return the character preceding the specified index.
384 // Leave the iteration position at the start of the character that was returned.
385 //
386 UChar32 cPrev; // The character preceding cCurr, which is what we will return.
387
388 // Address the chunk containing the position preceding the incoming index
389 // A tricky edge case:
390 // We try to test the requested native index against the chunkNativeStart to determine
391 // whether the character preceding the one at the index is in the current chunk.
392 // BUT, this test can fail with UTF-8 (or any other multibyte encoding), when the
393 // requested index is on something other than the first position of the first char.
394 //
395 if(index<=ut->chunkNativeStart || index>ut->chunkNativeLimit) {
396 // Requested native index is outside of the current chunk.
397 if(!ut->pFuncs->access(ut, index, false)) {
398 // no chunk available here
399 return U_SENTINEL;
400 }
401 } else if(index - ut->chunkNativeStart <= (int64_t)ut->nativeIndexingLimit) {
402 // Direct UTF-16 indexing.
403 ut->chunkOffset = (int32_t)(index - ut->chunkNativeStart);
404 } else {
405 ut->chunkOffset=ut->pFuncs->mapNativeIndexToUTF16(ut, index);
406 if (ut->chunkOffset==0 && !ut->pFuncs->access(ut, index, false)) {
407 // no chunk available here
408 return U_SENTINEL;
409 }
410 }
411
412 //
413 // Simple case with no surrogates.
414 //
415 ut->chunkOffset--;
416 cPrev = ut->chunkContents[ut->chunkOffset];
417
418 if (U16_IS_SURROGATE(cPrev)) {
419 // Possible supplementary. Many edge cases.
420 // Let other functions do the heavy lifting.
421 utext_setNativeIndex(ut, index);
422 cPrev = utext_previous32(ut);
423 }
424 return cPrev;
425}
426
427
428U_CAPI int32_t U_EXPORT2
429utext_extract(UText *ut,
430 int64_t start, int64_t limit,
431 char16_t *dest, int32_t destCapacity,
432 UErrorCode *status) {
433 return ut->pFuncs->extract(ut, start, limit, dest, destCapacity, status);
434 }
435
436
437
438U_CAPI UBool U_EXPORT2
439utext_equals(const UText *a, const UText *b) {
440 if (a==nullptr || b==nullptr ||
441 a->magic != UTEXT_MAGIC ||
442 b->magic != UTEXT_MAGIC) {
443 // Null or invalid arguments don't compare equal to anything.
444 return false;
445 }
446
447 if (a->pFuncs != b->pFuncs) {
448 // Different types of text providers.
449 return false;
450 }
451
452 if (a->context != b->context) {
453 // Different sources (different strings)
454 return false;
455 }
456 if (utext_getNativeIndex(a) != utext_getNativeIndex(b)) {
457 // Different current position in the string.
458 return false;
459 }
460
461 return true;
462}
463
464U_CAPI UBool U_EXPORT2
465utext_isWritable(const UText *ut)
466{
467 UBool b = (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_WRITABLE)) != 0;
468 return b;
469}
470
471
472U_CAPI void U_EXPORT2
473utext_freeze(UText *ut) {
474 // Zero out the WRITABLE flag.
475 ut->providerProperties &= ~(I32_FLAG(UTEXT_PROVIDER_WRITABLE));
476}
477
478
479U_CAPI UBool U_EXPORT2
480utext_hasMetaData(const UText *ut)
481{
482 UBool b = (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_HAS_META_DATA)) != 0;
483 return b;
484}
485
486
487
488U_CAPI int32_t U_EXPORT2
489utext_replace(UText *ut,
490 int64_t nativeStart, int64_t nativeLimit,
491 const char16_t *replacementText, int32_t replacementLength,
492 UErrorCode *status)
493{
494 if (U_FAILURE(*status)) {
495 return 0;
496 }
497 if ((ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_WRITABLE)) == 0) {
498 *status = U_NO_WRITE_PERMISSION;
499 return 0;
500 }
501 int32_t i = ut->pFuncs->replace(ut, nativeStart, nativeLimit, replacementText, replacementLength, status);
502 return i;
503}
504
505U_CAPI void U_EXPORT2
506utext_copy(UText *ut,
507 int64_t nativeStart, int64_t nativeLimit,
508 int64_t destIndex,
509 UBool move,
510 UErrorCode *status)
511{
512 if (U_FAILURE(*status)) {
513 return;
514 }
515 if ((ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_WRITABLE)) == 0) {
516 *status = U_NO_WRITE_PERMISSION;
517 return;
518 }
519 ut->pFuncs->copy(ut, nativeStart, nativeLimit, destIndex, move, status);
520}
521
522
523
524U_CAPI UText * U_EXPORT2
525utext_clone(UText *dest, const UText *src, UBool deep, UBool readOnly, UErrorCode *status) {
526 if (U_FAILURE(*status)) {
527 return dest;
528 }
529 UText *result = src->pFuncs->clone(dest, src, deep, status);
530 if (U_FAILURE(*status)) {
531 return result;
532 }
533 if (result == nullptr) {
534 *status = U_MEMORY_ALLOCATION_ERROR;
535 return result;
536 }
537 if (readOnly) {
538 utext_freeze(result);
539 }
540 return result;
541}
542
543
544
545//------------------------------------------------------------------------------
546//
547// UText common functions implementation
548//
549//------------------------------------------------------------------------------
550
551//
552// UText.flags bit definitions
553//
554enum {
555 UTEXT_HEAP_ALLOCATED = 1, // 1 if ICU has allocated this UText struct on the heap.
556 // 0 if caller provided storage for the UText.
557
558 UTEXT_EXTRA_HEAP_ALLOCATED = 2, // 1 if ICU has allocated extra storage as a separate
559 // heap block.
560 // 0 if there is no separate allocation. Either no extra
561 // storage was requested, or it is appended to the end
562 // of the main UText storage.
563
564 UTEXT_OPEN = 4 // 1 if this UText is currently open
565 // 0 if this UText is not open.
566};
567
568
569//
570// Extended form of a UText. The purpose is to aid in computing the total size required
571// when a provider asks for a UText to be allocated with extra storage.
572
573struct ExtendedUText {
574 UText ut;
575 std::max_align_t extension;
576};
577
578static const UText emptyText = UTEXT_INITIALIZER;
579
580U_CAPI UText * U_EXPORT2
581utext_setup(UText *ut, int32_t extraSpace, UErrorCode *status) {
582 if (U_FAILURE(*status)) {
583 return ut;
584 }
585
586 if (ut == nullptr) {
587 // We need to heap-allocate storage for the new UText
588 int32_t spaceRequired = sizeof(UText);
589 if (extraSpace > 0) {
590 spaceRequired = sizeof(ExtendedUText) + extraSpace - sizeof(std::max_align_t);
591 }
592 ut = (UText *)uprv_malloc(spaceRequired);
593 if (ut == nullptr) {
594 *status = U_MEMORY_ALLOCATION_ERROR;
595 return nullptr;
596 } else {
597 *ut = emptyText;
598 ut->flags |= UTEXT_HEAP_ALLOCATED;
599 if (spaceRequired>0) {
600 ut->extraSize = extraSpace;
601 ut->pExtra = &((ExtendedUText *)ut)->extension;
602 }
603 }
604 } else {
605 // We have been supplied with an already existing UText.
606 // Verify that it really appears to be a UText.
607 if (ut->magic != UTEXT_MAGIC) {
608 *status = U_ILLEGAL_ARGUMENT_ERROR;
609 return ut;
610 }
611 // If the ut is already open and there's a provider supplied close
612 // function, call it.
613 if ((ut->flags & UTEXT_OPEN) && ut->pFuncs->close != nullptr) {
614 ut->pFuncs->close(ut);
615 }
616 ut->flags &= ~UTEXT_OPEN;
617
618 // If extra space was requested by our caller, check whether
619 // sufficient already exists, and allocate new if needed.
620 if (extraSpace > ut->extraSize) {
621 // Need more space. If there is existing separately allocated space,
622 // delete it first, then allocate new space.
623 if (ut->flags & UTEXT_EXTRA_HEAP_ALLOCATED) {
624 uprv_free(ut->pExtra);
625 ut->extraSize = 0;
626 }
627 ut->pExtra = uprv_malloc(extraSpace);
628 if (ut->pExtra == nullptr) {
629 *status = U_MEMORY_ALLOCATION_ERROR;
630 } else {
631 ut->extraSize = extraSpace;
632 ut->flags |= UTEXT_EXTRA_HEAP_ALLOCATED;
633 }
634 }
635 }
636 if (U_SUCCESS(*status)) {
637 ut->flags |= UTEXT_OPEN;
638
639 // Initialize all remaining fields of the UText.
640 //
641 ut->context = nullptr;
642 ut->chunkContents = nullptr;
643 ut->p = nullptr;
644 ut->q = nullptr;
645 ut->r = nullptr;
646 ut->a = 0;
647 ut->b = 0;
648 ut->c = 0;
649 ut->chunkOffset = 0;
650 ut->chunkLength = 0;
651 ut->chunkNativeStart = 0;
652 ut->chunkNativeLimit = 0;
653 ut->nativeIndexingLimit = 0;
654 ut->providerProperties = 0;
655 ut->privA = 0;
656 ut->privB = 0;
657 ut->privC = 0;
658 ut->privP = nullptr;
659 if (ut->pExtra!=nullptr && ut->extraSize>0)
660 uprv_memset(ut->pExtra, 0, ut->extraSize);
661
662 }
663 return ut;
664}
665
666
667U_CAPI UText * U_EXPORT2
668utext_close(UText *ut) {
669 if (ut==nullptr ||
670 ut->magic != UTEXT_MAGIC ||
671 (ut->flags & UTEXT_OPEN) == 0)
672 {
673 // The supplied ut is not an open UText.
674 // Do nothing.
675 return ut;
676 }
677
678 // If the provider gave us a close function, call it now.
679 // This will clean up anything allocated specifically by the provider.
680 if (ut->pFuncs->close != nullptr) {
681 ut->pFuncs->close(ut);
682 }
683 ut->flags &= ~UTEXT_OPEN;
684
685 // If we (the framework) allocated the UText or subsidiary storage,
686 // delete it.
687 if (ut->flags & UTEXT_EXTRA_HEAP_ALLOCATED) {
688 uprv_free(ut->pExtra);
689 ut->pExtra = nullptr;
690 ut->flags &= ~UTEXT_EXTRA_HEAP_ALLOCATED;
691 ut->extraSize = 0;
692 }
693
694 // Zero out function table of the closed UText. This is a defensive move,
695 // intended to cause applications that inadvertently use a closed
696 // utext to crash with null pointer errors.
697 ut->pFuncs = nullptr;
698
699 if (ut->flags & UTEXT_HEAP_ALLOCATED) {
700 // This UText was allocated by UText setup. We need to free it.
701 // Clear magic, so we can detect if the user messes up and immediately
702 // tries to reopen another UText using the deleted storage.
703 ut->magic = 0;
704 uprv_free(ut);
705 ut = nullptr;
706 }
707 return ut;
708}
709
710
711
712
713//
714// invalidateChunk Reset a chunk to have no contents, so that the next call
715// to access will cause new data to load.
716// This is needed when copy/move/replace operate directly on the
717// backing text, potentially putting it out of sync with the
718// contents in the chunk.
719//
720static void
721invalidateChunk(UText *ut) {
722 ut->chunkLength = 0;
723 ut->chunkNativeLimit = 0;
724 ut->chunkNativeStart = 0;
725 ut->chunkOffset = 0;
726 ut->nativeIndexingLimit = 0;
727}
728
729//
730// pinIndex Do range pinning on a native index parameter.
731// 64 bit pinning is done in place.
732// 32 bit truncated result is returned as a convenience for
733// use in providers that don't need 64 bits.
734static int32_t
735pinIndex(int64_t &index, int64_t limit) {
736 if (index<0) {
737 index = 0;
738 } else if (index > limit) {
739 index = limit;
740 }
741 return (int32_t)index;
742}
743
744
745U_CDECL_BEGIN
746
747//
748// Pointer relocation function,
749// a utility used by shallow clone.
750// Adjust a pointer that refers to something within one UText (the source)
751// to refer to the same relative offset within a another UText (the target)
752//
753static void adjustPointer(UText *dest, const void **destPtr, const UText *src) {
754 // convert all pointers to (char *) so that byte address arithmetic will work.
755 char *dptr = (char *)*destPtr;
756 char *dUText = (char *)dest;
757 char *sUText = (char *)src;
758
759 if (dptr >= (char *)src->pExtra && dptr < ((char*)src->pExtra)+src->extraSize) {
760 // target ptr was to something within the src UText's pExtra storage.
761 // relocate it into the target UText's pExtra region.
762 *destPtr = ((char *)dest->pExtra) + (dptr - (char *)src->pExtra);
763 } else if (dptr>=sUText && dptr < sUText+src->sizeOfStruct) {
764 // target ptr was pointing to somewhere within the source UText itself.
765 // Move it to the same offset within the target UText.
766 *destPtr = dUText + (dptr-sUText);
767 }
768}
769
770
771//
772// Clone. This is a generic copy-the-utext-by-value clone function that can be
773// used as-is with some utext types, and as a helper by other clones.
774//
775static UText * U_CALLCONV
776shallowTextClone(UText * dest, const UText * src, UErrorCode * status) {
777 if (U_FAILURE(*status)) {
778 return nullptr;
779 }
780 int32_t srcExtraSize = src->extraSize;
781
782 //
783 // Use the generic text_setup to allocate storage if required.
784 //
785 dest = utext_setup(dest, srcExtraSize, status);
786 if (U_FAILURE(*status)) {
787 return dest;
788 }
789
790 //
791 // flags (how the UText was allocated) and the pointer to the
792 // extra storage must retain the values in the cloned utext that
793 // were set up by utext_setup. Save them separately before
794 // copying the whole struct.
795 //
796 void *destExtra = dest->pExtra;
797 int32_t flags = dest->flags;
798
799
800 //
801 // Copy the whole UText struct by value.
802 // Any "Extra" storage is copied also.
803 //
804 int sizeToCopy = src->sizeOfStruct;
805 if (sizeToCopy > dest->sizeOfStruct) {
806 sizeToCopy = dest->sizeOfStruct;
807 }
808 uprv_memcpy(dest, src, sizeToCopy);
809 dest->pExtra = destExtra;
810 dest->flags = flags;
811 if (srcExtraSize > 0) {
812 uprv_memcpy(dest->pExtra, src->pExtra, srcExtraSize);
813 }
814
815 //
816 // Relocate any pointers in the target that refer to the UText itself
817 // to point to the cloned copy rather than the original source.
818 //
819 adjustPointer(dest, &dest->context, src);
820 adjustPointer(dest, &dest->p, src);
821 adjustPointer(dest, &dest->q, src);
822 adjustPointer(dest, &dest->r, src);
823 adjustPointer(dest, (const void **)&dest->chunkContents, src);
824
825 // The newly shallow-cloned UText does _not_ own the underlying storage for the text.
826 // (The source for the clone may or may not have owned the text.)
827
828 dest->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT);
829
830 return dest;
831}
832
833
834U_CDECL_END
835
836
837
838//------------------------------------------------------------------------------
839//
840// UText implementation for UTF-8 char * strings (read-only)
841// Limitation: string length must be <= 0x7fffffff in length.
842// (length must for in an int32_t variable)
843//
844// Use of UText data members:
845// context pointer to UTF-8 string
846// utext.b is the input string length (bytes).
847// utext.c Length scanned so far in string
848// (for optimizing finding length of zero terminated strings.)
849// utext.p pointer to the current buffer
850// utext.q pointer to the other buffer.
851//
852//------------------------------------------------------------------------------
853
854// Chunk size.
855// Must be less than 85 (256/3), because of byte mapping from char16_t indexes to native indexes.
856// Worst case is three native bytes to one char16_t. (Supplemenaries are 4 native bytes
857// to two UChars.)
858// The longest illegal byte sequence treated as a single error (and converted to U+FFFD)
859// is a three-byte sequence (truncated four-byte sequence).
860//
861enum { UTF8_TEXT_CHUNK_SIZE=32 };
862
863//
864// UTF8Buf Two of these structs will be set up in the UText's extra allocated space.
865// Each contains the char16_t chunk buffer, the to and from native maps, and
866// header info.
867//
868// because backwards iteration fills the buffers starting at the end and
869// working towards the front, the filled part of the buffers may not begin
870// at the start of the available storage for the buffers.
871//
872// Buffer size is one bigger than the specified UTF8_TEXT_CHUNK_SIZE to allow for
873// the last character added being a supplementary, and thus requiring a surrogate
874// pair. Doing this is simpler than checking for the edge case.
875//
876
877struct UTF8Buf {
878 int32_t bufNativeStart; // Native index of first char in char16_t buf
879 int32_t bufNativeLimit; // Native index following last char in buf.
880 int32_t bufStartIdx; // First filled position in buf.
881 int32_t bufLimitIdx; // Limit of filled range in buf.
882 int32_t bufNILimit; // Limit of native indexing part of buf
883 int32_t toUCharsMapStart; // Native index corresponding to
884 // mapToUChars[0].
885 // Set to bufNativeStart when filling forwards.
886 // Set to computed value when filling backwards.
887
888 char16_t buf[UTF8_TEXT_CHUNK_SIZE+4]; // The char16_t buffer. Requires one extra position beyond the
889 // the chunk size, to allow for surrogate at the end.
890 // Length must be identical to mapToNative array, below,
891 // because of the way indexing works when the array is
892 // filled backwards during a reverse iteration. Thus,
893 // the additional extra size.
894 uint8_t mapToNative[UTF8_TEXT_CHUNK_SIZE+4]; // map char16_t index in buf to
895 // native offset from bufNativeStart.
896 // Requires two extra slots,
897 // one for a supplementary starting in the last normal position,
898 // and one for an entry for the buffer limit position.
899 uint8_t mapToUChars[UTF8_TEXT_CHUNK_SIZE*3+6]; // Map native offset from bufNativeStart to
900 // corresponding offset in filled part of buf.
901 int32_t align;
902};
903
904U_CDECL_BEGIN
905
906//
907// utf8TextLength
908//
909// Get the length of the string. If we don't already know it,
910// we'll need to scan for the trailing nul.
911//
912static int64_t U_CALLCONV
913utf8TextLength(UText *ut) {
914 if (ut->b < 0) {
915 // Zero terminated string, and we haven't scanned to the end yet.
916 // Scan it now.
917 const char *r = (const char *)ut->context + ut->c;
918 while (*r != 0) {
919 r++;
920 }
921 if ((r - (const char *)ut->context) < 0x7fffffff) {
922 ut->b = (int32_t)(r - (const char *)ut->context);
923 } else {
924 // Actual string was bigger (more than 2 gig) than we
925 // can handle. Clip it to 2 GB.
926 ut->b = 0x7fffffff;
927 }
928 ut->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
929 }
930 return ut->b;
931}
932
933
934
935
936
937
938static UBool U_CALLCONV
939utf8TextAccess(UText *ut, int64_t index, UBool forward) {
940 //
941 // Apologies to those who are allergic to goto statements.
942 // Consider each goto to a labelled block to be the equivalent of
943 // call the named block as if it were a function();
944 // return;
945 //
946 const uint8_t *s8=(const uint8_t *)ut->context;
947 UTF8Buf *u8b = nullptr;
948 int32_t length = ut->b; // Length of original utf-8
949 int32_t ix= (int32_t)index; // Requested index, trimmed to 32 bits.
950 int32_t mapIndex = 0;
951 if (index<0) {
952 ix=0;
953 } else if (index > 0x7fffffff) {
954 // Strings with 64 bit lengths not supported by this UTF-8 provider.
955 ix = 0x7fffffff;
956 }
957
958 // Pin requested index to the string length.
959 if (ix>length) {
960 if (length>=0) {
961 ix=length;
962 } else if (ix>=ut->c) {
963 // Zero terminated string, and requested index is beyond
964 // the region that has already been scanned.
965 // Scan up to either the end of the string or to the
966 // requested position, whichever comes first.
967 while (ut->c<ix && s8[ut->c]!=0) {
968 ut->c++;
969 }
970 // TODO: support for null terminated string length > 32 bits.
971 if (s8[ut->c] == 0) {
972 // We just found the actual length of the string.
973 // Trim the requested index back to that.
974 ix = ut->c;
975 ut->b = ut->c;
976 length = ut->c;
977 ut->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
978 }
979 }
980 }
981
982 //
983 // Dispatch to the appropriate action for a forward iteration request.
984 //
985 if (forward) {
986 if (ix==ut->chunkNativeLimit) {
987 // Check for normal sequential iteration cases first.
988 if (ix==length) {
989 // Just reached end of string
990 // Don't swap buffers, but do set the
991 // current buffer position.
992 ut->chunkOffset = ut->chunkLength;
993 return false;
994 } else {
995 // End of current buffer.
996 // check whether other buffer already has what we need.
997 UTF8Buf *altB = (UTF8Buf *)ut->q;
998 if (ix>=altB->bufNativeStart && ix<altB->bufNativeLimit) {
999 goto swapBuffers;
1000 }
1001 }
1002 }
1003
1004 // A random access. Desired index could be in either or niether buf.
1005 // For optimizing the order of testing, first check for the index
1006 // being in the other buffer. This will be the case for uses that
1007 // move back and forth over a fairly limited range
1008 {
1009 u8b = (UTF8Buf *)ut->q; // the alternate buffer
1010 if (ix>=u8b->bufNativeStart && ix<u8b->bufNativeLimit) {
1011 // Requested index is in the other buffer.
1012 goto swapBuffers;
1013 }
1014 if (ix == length) {
1015 // Requested index is end-of-string.
1016 // (this is the case of randomly seeking to the end.
1017 // The case of iterating off the end is handled earlier.)
1018 if (ix == ut->chunkNativeLimit) {
1019 // Current buffer extends up to the end of the string.
1020 // Leave it as the current buffer.
1021 ut->chunkOffset = ut->chunkLength;
1022 return false;
1023 }
1024 if (ix == u8b->bufNativeLimit) {
1025 // Alternate buffer extends to the end of string.
1026 // Swap it in as the current buffer.
1027 goto swapBuffersAndFail;
1028 }
1029
1030 // Neither existing buffer extends to the end of the string.
1031 goto makeStubBuffer;
1032 }
1033
1034 if (ix<ut->chunkNativeStart || ix>=ut->chunkNativeLimit) {
1035 // Requested index is in neither buffer.
1036 goto fillForward;
1037 }
1038
1039 // Requested index is in this buffer.
1040 u8b = (UTF8Buf *)ut->p; // the current buffer
1041 mapIndex = ix - u8b->toUCharsMapStart;
1042 U_ASSERT(mapIndex < (int32_t)sizeof(UTF8Buf::mapToUChars));
1043 ut->chunkOffset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;
1044 return true;
1045
1046 }
1047 }
1048
1049
1050 //
1051 // Dispatch to the appropriate action for a
1052 // Backwards Direction iteration request.
1053 //
1054 if (ix==ut->chunkNativeStart) {
1055 // Check for normal sequential iteration cases first.
1056 if (ix==0) {
1057 // Just reached the start of string
1058 // Don't swap buffers, but do set the
1059 // current buffer position.
1060 ut->chunkOffset = 0;
1061 return false;
1062 } else {
1063 // Start of current buffer.
1064 // check whether other buffer already has what we need.
1065 UTF8Buf *altB = (UTF8Buf *)ut->q;
1066 if (ix>altB->bufNativeStart && ix<=altB->bufNativeLimit) {
1067 goto swapBuffers;
1068 }
1069 }
1070 }
1071
1072 // A random access. Desired index could be in either or niether buf.
1073 // For optimizing the order of testing,
1074 // Most likely case: in the other buffer.
1075 // Second most likely: in neither buffer.
1076 // Unlikely, but must work: in the current buffer.
1077 u8b = (UTF8Buf *)ut->q; // the alternate buffer
1078 if (ix>u8b->bufNativeStart && ix<=u8b->bufNativeLimit) {
1079 // Requested index is in the other buffer.
1080 goto swapBuffers;
1081 }
1082 // Requested index is start-of-string.
1083 // (this is the case of randomly seeking to the start.
1084 // The case of iterating off the start is handled earlier.)
1085 if (ix==0) {
1086 if (u8b->bufNativeStart==0) {
1087 // Alternate buffer contains the data for the start string.
1088 // Make it be the current buffer.
1089 goto swapBuffersAndFail;
1090 } else {
1091 // Request for data before the start of string,
1092 // neither buffer is usable.
1093 // set up a zero-length buffer.
1094 goto makeStubBuffer;
1095 }
1096 }
1097
1098 if (ix<=ut->chunkNativeStart || ix>ut->chunkNativeLimit) {
1099 // Requested index is in neither buffer.
1100 goto fillReverse;
1101 }
1102
1103 // Requested index is in this buffer.
1104 // Set the utf16 buffer index.
1105 u8b = (UTF8Buf *)ut->p;
1106 mapIndex = ix - u8b->toUCharsMapStart;
1107 ut->chunkOffset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;
1108 if (ut->chunkOffset==0) {
1109 // This occurs when the first character in the text is
1110 // a multi-byte UTF-8 char, and the requested index is to
1111 // one of the trailing bytes. Because there is no preceding ,
1112 // character, this access fails. We can't pick up on the
1113 // situation sooner because the requested index is not zero.
1114 return false;
1115 } else {
1116 return true;
1117 }
1118
1119
1120
1121swapBuffers:
1122 // The alternate buffer (ut->q) has the string data that was requested.
1123 // Swap the primary and alternate buffers, and set the
1124 // chunk index into the new primary buffer.
1125 {
1126 u8b = (UTF8Buf *)ut->q;
1127 ut->q = ut->p;
1128 ut->p = u8b;
1129 ut->chunkContents = &u8b->buf[u8b->bufStartIdx];
1130 ut->chunkLength = u8b->bufLimitIdx - u8b->bufStartIdx;
1131 ut->chunkNativeStart = u8b->bufNativeStart;
1132 ut->chunkNativeLimit = u8b->bufNativeLimit;
1133 ut->nativeIndexingLimit = u8b->bufNILimit;
1134
1135 // Index into the (now current) chunk
1136 // Use the map to set the chunk index. It's more trouble than it's worth
1137 // to check whether native indexing can be used.
1138 U_ASSERT(ix>=u8b->bufNativeStart);
1139 U_ASSERT(ix<=u8b->bufNativeLimit);
1140 mapIndex = ix - u8b->toUCharsMapStart;
1141 U_ASSERT(mapIndex>=0);
1142 U_ASSERT(mapIndex<(int32_t)sizeof(u8b->mapToUChars));
1143 ut->chunkOffset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;
1144
1145 return true;
1146 }
1147
1148
1149 swapBuffersAndFail:
1150 // We got a request for either the start or end of the string,
1151 // with iteration continuing in the out-of-bounds direction.
1152 // The alternate buffer already contains the data up to the
1153 // start/end.
1154 // Swap the buffers, then return failure, indicating that we couldn't
1155 // make things correct for continuing the iteration in the requested
1156 // direction. The position & buffer are correct should the
1157 // user decide to iterate in the opposite direction.
1158 u8b = (UTF8Buf *)ut->q;
1159 ut->q = ut->p;
1160 ut->p = u8b;
1161 ut->chunkContents = &u8b->buf[u8b->bufStartIdx];
1162 ut->chunkLength = u8b->bufLimitIdx - u8b->bufStartIdx;
1163 ut->chunkNativeStart = u8b->bufNativeStart;
1164 ut->chunkNativeLimit = u8b->bufNativeLimit;
1165 ut->nativeIndexingLimit = u8b->bufNILimit;
1166
1167 // Index into the (now current) chunk
1168 // For this function (swapBuffersAndFail), the requested index
1169 // will always be at either the start or end of the chunk.
1170 if (ix==u8b->bufNativeLimit) {
1171 ut->chunkOffset = ut->chunkLength;
1172 } else {
1173 ut->chunkOffset = 0;
1174 U_ASSERT(ix == u8b->bufNativeStart);
1175 }
1176 return false;
1177
1178makeStubBuffer:
1179 // The user has done a seek/access past the start or end
1180 // of the string. Rather than loading data that is likely
1181 // to never be used, just set up a zero-length buffer at
1182 // the position.
1183 u8b = (UTF8Buf *)ut->q;
1184 u8b->bufNativeStart = ix;
1185 u8b->bufNativeLimit = ix;
1186 u8b->bufStartIdx = 0;
1187 u8b->bufLimitIdx = 0;
1188 u8b->bufNILimit = 0;
1189 u8b->toUCharsMapStart = ix;
1190 u8b->mapToNative[0] = 0;
1191 u8b->mapToUChars[0] = 0;
1192 goto swapBuffersAndFail;
1193
1194
1195
1196fillForward:
1197 {
1198 // Move the incoming index to a code point boundary.
1199 U8_SET_CP_START(s8, 0, ix);
1200
1201 // Swap the UText buffers.
1202 // We want to fill what was previously the alternate buffer,
1203 // and make what was the current buffer be the new alternate.
1204 UTF8Buf *u8b_swap = (UTF8Buf *)ut->q;
1205 ut->q = ut->p;
1206 ut->p = u8b_swap;
1207
1208 int32_t strLen = ut->b;
1209 UBool nulTerminated = false;
1210 if (strLen < 0) {
1211 strLen = 0x7fffffff;
1212 nulTerminated = true;
1213 }
1214
1215 char16_t *buf = u8b_swap->buf;
1216 uint8_t *mapToNative = u8b_swap->mapToNative;
1217 uint8_t *mapToUChars = u8b_swap->mapToUChars;
1218 int32_t destIx = 0;
1219 int32_t srcIx = ix;
1220 UBool seenNonAscii = false;
1221 UChar32 c = 0;
1222
1223 // Fill the chunk buffer and mapping arrays.
1224 while (destIx<UTF8_TEXT_CHUNK_SIZE) {
1225 c = s8[srcIx];
1226 if (c>0 && c<0x80) {
1227 // Special case ASCII range for speed.
1228 // zero is excluded to simplify bounds checking.
1229 buf[destIx] = (char16_t)c;
1230 mapToNative[destIx] = (uint8_t)(srcIx - ix);
1231 mapToUChars[srcIx-ix] = (uint8_t)destIx;
1232 srcIx++;
1233 destIx++;
1234 } else {
1235 // General case, handle everything.
1236 if (seenNonAscii == false) {
1237 seenNonAscii = true;
1238 u8b_swap->bufNILimit = destIx;
1239 }
1240
1241 int32_t cIx = srcIx;
1242 int32_t dIx = destIx;
1243 int32_t dIxSaved = destIx;
1244 U8_NEXT_OR_FFFD(s8, srcIx, strLen, c);
1245 if (c==0 && nulTerminated) {
1246 srcIx--;
1247 break;
1248 }
1249
1250 U16_APPEND_UNSAFE(buf, destIx, c);
1251 do {
1252 mapToNative[dIx++] = (uint8_t)(cIx - ix);
1253 } while (dIx < destIx);
1254
1255 do {
1256 mapToUChars[cIx++ - ix] = (uint8_t)dIxSaved;
1257 } while (cIx < srcIx);
1258 }
1259 if (srcIx>=strLen) {
1260 break;
1261 }
1262
1263 }
1264
1265 // store Native <--> Chunk Map entries for the end of the buffer.
1266 // There is no actual character here, but the index position is valid.
1267 mapToNative[destIx] = (uint8_t)(srcIx - ix);
1268 mapToUChars[srcIx - ix] = (uint8_t)destIx;
1269
1270 // fill in Buffer descriptor
1271 u8b_swap->bufNativeStart = ix;
1272 u8b_swap->bufNativeLimit = srcIx;
1273 u8b_swap->bufStartIdx = 0;
1274 u8b_swap->bufLimitIdx = destIx;
1275 if (seenNonAscii == false) {
1276 u8b_swap->bufNILimit = destIx;
1277 }
1278 u8b_swap->toUCharsMapStart = u8b_swap->bufNativeStart;
1279
1280 // Set UText chunk to refer to this buffer.
1281 ut->chunkContents = buf;
1282 ut->chunkOffset = 0;
1283 ut->chunkLength = u8b_swap->bufLimitIdx;
1284 ut->chunkNativeStart = u8b_swap->bufNativeStart;
1285 ut->chunkNativeLimit = u8b_swap->bufNativeLimit;
1286 ut->nativeIndexingLimit = u8b_swap->bufNILimit;
1287
1288 // For zero terminated strings, keep track of the maximum point
1289 // scanned so far.
1290 if (nulTerminated && srcIx>ut->c) {
1291 ut->c = srcIx;
1292 if (c==0) {
1293 // We scanned to the end.
1294 // Remember the actual length.
1295 ut->b = srcIx;
1296 ut->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
1297 }
1298 }
1299 return true;
1300 }
1301
1302
1303fillReverse:
1304 {
1305 // Move the incoming index to a code point boundary.
1306 // Can only do this if the incoming index is somewhere in the interior of the string.
1307 // If index is at the end, there is no character there to look at.
1308 if (ix != ut->b) {
1309 // Note: this function will only move the index back if it is on a trail byte
1310 // and there is a preceding lead byte and the sequence from the lead
1311 // through this trail could be part of a valid UTF-8 sequence
1312 // Otherwise the index remains unchanged.
1313 U8_SET_CP_START(s8, 0, ix);
1314 }
1315
1316 // Swap the UText buffers.
1317 // We want to fill what was previously the alternate buffer,
1318 // and make what was the current buffer be the new alternate.
1319 UTF8Buf *u8b_swap = (UTF8Buf *)ut->q;
1320 ut->q = ut->p;
1321 ut->p = u8b_swap;
1322
1323 char16_t *buf = u8b_swap->buf;
1324 uint8_t *mapToNative = u8b_swap->mapToNative;
1325 uint8_t *mapToUChars = u8b_swap->mapToUChars;
1326 int32_t toUCharsMapStart = ix - sizeof(UTF8Buf::mapToUChars) + 1;
1327 // Note that toUCharsMapStart can be negative. Happens when the remaining
1328 // text from current position to the beginning is less than the buffer size.
1329 // + 1 because mapToUChars must have a slot at the end for the bufNativeLimit entry.
1330 int32_t destIx = UTF8_TEXT_CHUNK_SIZE+2; // Start in the overflow region
1331 // at end of buffer to leave room
1332 // for a surrogate pair at the
1333 // buffer start.
1334 int32_t srcIx = ix;
1335 int32_t bufNILimit = destIx;
1336 UChar32 c;
1337
1338 // Map to/from Native Indexes, fill in for the position at the end of
1339 // the buffer.
1340 //
1341 mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
1342 mapToUChars[srcIx - toUCharsMapStart] = (uint8_t)destIx;
1343
1344 // Fill the chunk buffer
1345 // Work backwards, filling from the end of the buffer towards the front.
1346 //
1347 while (destIx>2 && (srcIx - toUCharsMapStart > 5) && (srcIx > 0)) {
1348 srcIx--;
1349 destIx--;
1350
1351 // Get last byte of the UTF-8 character
1352 c = s8[srcIx];
1353 if (c<0x80) {
1354 // Special case ASCII range for speed.
1355 buf[destIx] = (char16_t)c;
1356 U_ASSERT(toUCharsMapStart <= srcIx);
1357 mapToUChars[srcIx - toUCharsMapStart] = (uint8_t)destIx;
1358 mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
1359 } else {
1360 // General case, handle everything non-ASCII.
1361
1362 int32_t sIx = srcIx; // ix of last byte of multi-byte u8 char
1363
1364 // Get the full character from the UTF8 string.
1365 // use code derived from the macros in utf8.h
1366 // Leaves srcIx pointing at the first byte of the UTF-8 char.
1367 //
1368 c=utf8_prevCharSafeBody(s8, 0, &srcIx, c, -3);
1369 // leaves srcIx at first byte of the multi-byte char.
1370
1371 // Store the character in UTF-16 buffer.
1372 if (c<0x10000) {
1373 buf[destIx] = (char16_t)c;
1374 mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
1375 } else {
1376 buf[destIx] = U16_TRAIL(c);
1377 mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
1378 buf[--destIx] = U16_LEAD(c);
1379 mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
1380 }
1381
1382 // Fill in the map from native indexes to UChars buf index.
1383 do {
1384 mapToUChars[sIx-- - toUCharsMapStart] = (uint8_t)destIx;
1385 } while (sIx >= srcIx);
1386 U_ASSERT(toUCharsMapStart <= (srcIx+1));
1387
1388 // Set native indexing limit to be the current position.
1389 // We are processing a non-ascii, non-native-indexing char now;
1390 // the limit will be here if the rest of the chars to be
1391 // added to this buffer are ascii.
1392 bufNILimit = destIx;
1393 }
1394 }
1395 u8b_swap->bufNativeStart = srcIx;
1396 u8b_swap->bufNativeLimit = ix;
1397 u8b_swap->bufStartIdx = destIx;
1398 u8b_swap->bufLimitIdx = UTF8_TEXT_CHUNK_SIZE+2;
1399 u8b_swap->bufNILimit = bufNILimit - u8b_swap->bufStartIdx;
1400 u8b_swap->toUCharsMapStart = toUCharsMapStart;
1401
1402 ut->chunkContents = &buf[u8b_swap->bufStartIdx];
1403 ut->chunkLength = u8b_swap->bufLimitIdx - u8b_swap->bufStartIdx;
1404 ut->chunkOffset = ut->chunkLength;
1405 ut->chunkNativeStart = u8b_swap->bufNativeStart;
1406 ut->chunkNativeLimit = u8b_swap->bufNativeLimit;
1407 ut->nativeIndexingLimit = u8b_swap->bufNILimit;
1408 return true;
1409 }
1410
1411}
1412
1413
1414
1415//
1416// This is a slightly modified copy of u_strFromUTF8,
1417// Inserts a Replacement Char rather than failing on invalid UTF-8
1418// Removes unnecessary features.
1419//
1420static char16_t*
1421utext_strFromUTF8(char16_t *dest,
1422 int32_t destCapacity,
1423 int32_t *pDestLength,
1424 const char* src,
1425 int32_t srcLength, // required. NUL terminated not supported.
1426 UErrorCode *pErrorCode
1427 )
1428{
1429
1430 char16_t *pDest = dest;
1431 char16_t *pDestLimit = (dest!=nullptr)?(dest+destCapacity):nullptr;
1432 UChar32 ch=0;
1433 int32_t index = 0;
1434 int32_t reqLength = 0;
1435 uint8_t* pSrc = (uint8_t*) src;
1436
1437
1438 while((index < srcLength)&&(pDest<pDestLimit)){
1439 ch = pSrc[index++];
1440 if(ch <=0x7f){
1441 *pDest++=(char16_t)ch;
1442 }else{
1443 ch=utf8_nextCharSafeBody(pSrc, &index, srcLength, ch, -3);
1444 if(U_IS_BMP(ch)){
1445 *(pDest++)=(char16_t)ch;
1446 }else{
1447 *(pDest++)=U16_LEAD(ch);
1448 if(pDest<pDestLimit){
1449 *(pDest++)=U16_TRAIL(ch);
1450 }else{
1451 reqLength++;
1452 break;
1453 }
1454 }
1455 }
1456 }
1457 /* donot fill the dest buffer just count the UChars needed */
1458 while(index < srcLength){
1459 ch = pSrc[index++];
1460 if(ch <= 0x7f){
1461 reqLength++;
1462 }else{
1463 ch=utf8_nextCharSafeBody(pSrc, &index, srcLength, ch, -3);
1464 reqLength+=U16_LENGTH(ch);
1465 }
1466 }
1467
1468 reqLength+=(int32_t)(pDest - dest);
1469
1470 if(pDestLength){
1471 *pDestLength = reqLength;
1472 }
1473
1474 /* Terminate the buffer */
1475 u_terminateUChars(dest,destCapacity,reqLength,pErrorCode);
1476
1477 return dest;
1478}
1479
1480
1481
1482static int32_t U_CALLCONV
1483utf8TextExtract(UText *ut,
1484 int64_t start, int64_t limit,
1485 char16_t *dest, int32_t destCapacity,
1486 UErrorCode *pErrorCode) {
1487 if(U_FAILURE(*pErrorCode)) {
1488 return 0;
1489 }
1490 if(destCapacity<0 || (dest==nullptr && destCapacity>0)) {
1491 *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
1492 return 0;
1493 }
1494 int32_t length = ut->b;
1495 int32_t start32 = pinIndex(start, length);
1496 int32_t limit32 = pinIndex(limit, length);
1497
1498 if(start32>limit32) {
1499 *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR;
1500 return 0;
1501 }
1502
1503
1504 // adjust the incoming indexes to land on code point boundaries if needed.
1505 // adjust by no more than three, because that is the largest number of trail bytes
1506 // in a well formed UTF8 character.
1507 const uint8_t *buf = (const uint8_t *)ut->context;
1508 int i;
1509 if (start32 < ut->chunkNativeLimit) {
1510 for (i=0; i<3; i++) {
1511 if (U8_IS_SINGLE(buf[start32]) || U8_IS_LEAD(buf[start32]) || start32==0) {
1512 break;
1513 }
1514 start32--;
1515 }
1516 }
1517
1518 if (limit32 < ut->chunkNativeLimit) {
1519 for (i=0; i<3; i++) {
1520 if (U8_IS_SINGLE(buf[limit32]) || U8_IS_LEAD(buf[limit32]) || limit32==0) {
1521 break;
1522 }
1523 limit32--;
1524 }
1525 }
1526
1527 // Do the actual extract.
1528 int32_t destLength=0;
1529 utext_strFromUTF8(dest, destCapacity, &destLength,
1530 (const char *)ut->context+start32, limit32-start32,
1531 pErrorCode);
1532 utf8TextAccess(ut, limit32, true);
1533 return destLength;
1534}
1535
1536//
1537// utf8TextMapOffsetToNative
1538//
1539// Map a chunk (UTF-16) offset to a native index.
1540static int64_t U_CALLCONV
1541utf8TextMapOffsetToNative(const UText *ut) {
1542 //
1543 UTF8Buf *u8b = (UTF8Buf *)ut->p;
1544 U_ASSERT(ut->chunkOffset>ut->nativeIndexingLimit && ut->chunkOffset<=ut->chunkLength);
1545 int32_t nativeOffset = u8b->mapToNative[ut->chunkOffset + u8b->bufStartIdx] + u8b->toUCharsMapStart;
1546 U_ASSERT(nativeOffset >= ut->chunkNativeStart && nativeOffset <= ut->chunkNativeLimit);
1547 return nativeOffset;
1548}
1549
1550//
1551// Map a native index to the corresponding chunk offset
1552//
1553static int32_t U_CALLCONV
1554utf8TextMapIndexToUTF16(const UText *ut, int64_t index64) {
1555 U_ASSERT(index64 <= 0x7fffffff);
1556 int32_t index = (int32_t)index64;
1557 UTF8Buf *u8b = (UTF8Buf *)ut->p;
1558 U_ASSERT(index>=ut->chunkNativeStart+ut->nativeIndexingLimit);
1559 U_ASSERT(index<=ut->chunkNativeLimit);
1560 int32_t mapIndex = index - u8b->toUCharsMapStart;
1561 U_ASSERT(mapIndex < (int32_t)sizeof(UTF8Buf::mapToUChars));
1562 int32_t offset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;
1563 U_ASSERT(offset>=0 && offset<=ut->chunkLength);
1564 return offset;
1565}
1566
1567static UText * U_CALLCONV
1568utf8TextClone(UText *dest, const UText *src, UBool deep, UErrorCode *status)
1569{
1570 // First do a generic shallow clone. Does everything needed for the UText struct itself.
1571 dest = shallowTextClone(dest, src, status);
1572
1573 // For deep clones, make a copy of the string.
1574 // The copied storage is owned by the newly created clone.
1575 //
1576 // TODO: There is an issue with using utext_nativeLength().
1577 // That function is non-const in cases where the input was NUL terminated
1578 // and the length has not yet been determined.
1579 // This function (clone()) is const.
1580 // There potentially a thread safety issue lurking here.
1581 //
1582 if (deep && U_SUCCESS(*status)) {
1583 int32_t len = (int32_t)utext_nativeLength((UText *)src);
1584 char *copyStr = (char *)uprv_malloc(len+1);
1585 if (copyStr == nullptr) {
1586 *status = U_MEMORY_ALLOCATION_ERROR;
1587 } else {
1588 uprv_memcpy(copyStr, src->context, len+1);
1589 dest->context = copyStr;
1590 dest->providerProperties |= I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT);
1591 }
1592 }
1593 return dest;
1594}
1595
1596
1597static void U_CALLCONV
1598utf8TextClose(UText *ut) {
1599 // Most of the work of close is done by the generic UText framework close.
1600 // All that needs to be done here is to delete the UTF8 string if the UText
1601 // owns it. This occurs if the UText was created by cloning.
1602 if (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT)) {
1603 char *s = (char *)ut->context;
1604 uprv_free(s);
1605 ut->context = nullptr;
1606 }
1607}
1608
1609U_CDECL_END
1610
1611
1612static const struct UTextFuncs utf8Funcs =
1613{
1614 sizeof(UTextFuncs),
1615 0, 0, 0, // Reserved alignment padding
1616 utf8TextClone,
1617 utf8TextLength,
1618 utf8TextAccess,
1619 utf8TextExtract,
1620 nullptr, /* replace*/
1621 nullptr, /* copy */
1622 utf8TextMapOffsetToNative,
1623 utf8TextMapIndexToUTF16,
1624 utf8TextClose,
1625 nullptr, // spare 1
1626 nullptr, // spare 2
1627 nullptr // spare 3
1628};
1629
1630
1631static const char gEmptyString[] = {0};
1632
1633U_CAPI UText * U_EXPORT2
1634utext_openUTF8(UText *ut, const char *s, int64_t length, UErrorCode *status) {
1635 if(U_FAILURE(*status)) {
1636 return nullptr;
1637 }
1638 if(s==nullptr && length==0) {
1639 s = gEmptyString;
1640 }
1641
1642 if(s==nullptr || length<-1 || length>INT32_MAX) {
1643 *status=U_ILLEGAL_ARGUMENT_ERROR;
1644 return nullptr;
1645 }
1646
1647 ut = utext_setup(ut, sizeof(UTF8Buf) * 2, status);
1648 if (U_FAILURE(*status)) {
1649 return ut;
1650 }
1651
1652 ut->pFuncs = &utf8Funcs;
1653 ut->context = s;
1654 ut->b = (int32_t)length;
1655 ut->c = (int32_t)length;
1656 if (ut->c < 0) {
1657 ut->c = 0;
1658 ut->providerProperties |= I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
1659 }
1660 ut->p = ut->pExtra;
1661 ut->q = (char *)ut->pExtra + sizeof(UTF8Buf);
1662 return ut;
1663
1664}
1665
1666
1667
1668
1669
1670
1671
1672
1673//------------------------------------------------------------------------------
1674//
1675// UText implementation wrapper for Replaceable (read/write)
1676//
1677// Use of UText data members:
1678// context pointer to Replaceable.
1679// p pointer to Replaceable if it is owned by the UText.
1680//
1681//------------------------------------------------------------------------------
1682
1683
1684
1685// minimum chunk size for this implementation: 3
1686// to allow for possible trimming for code point boundaries
1687enum { REP_TEXT_CHUNK_SIZE=10 };
1688
1689struct ReplExtra {
1690 /*
1691 * Chunk UChars.
1692 * +1 to simplify filling with surrogate pair at the end.
1693 */
1694 char16_t s[REP_TEXT_CHUNK_SIZE+1];
1695};
1696
1697
1698U_CDECL_BEGIN
1699
1700static UText * U_CALLCONV
1701repTextClone(UText *dest, const UText *src, UBool deep, UErrorCode *status) {
1702 // First do a generic shallow clone. Does everything needed for the UText struct itself.
1703 dest = shallowTextClone(dest, src, status);
1704
1705 // For deep clones, make a copy of the Replaceable.
1706 // The copied Replaceable storage is owned by the newly created UText clone.
1707 // A non-nullptr pointer in UText.p is the signal to the close() function to delete
1708 // it.
1709 //
1710 if (deep && U_SUCCESS(*status)) {
1711 const Replaceable *replSrc = (const Replaceable *)src->context;
1712 dest->context = replSrc->clone();
1713 dest->providerProperties |= I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT);
1714
1715 // with deep clone, the copy is writable, even when the source is not.
1716 dest->providerProperties |= I32_FLAG(UTEXT_PROVIDER_WRITABLE);
1717 }
1718 return dest;
1719}
1720
1721
1722static void U_CALLCONV
1723repTextClose(UText *ut) {
1724 // Most of the work of close is done by the generic UText framework close.
1725 // All that needs to be done here is delete the Replaceable if the UText
1726 // owns it. This occurs if the UText was created by cloning.
1727 if (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT)) {
1728 Replaceable *rep = (Replaceable *)ut->context;
1729 delete rep;
1730 ut->context = nullptr;
1731 }
1732}
1733
1734
1735static int64_t U_CALLCONV
1736repTextLength(UText *ut) {
1737 const Replaceable *replSrc = (const Replaceable *)ut->context;
1738 int32_t len = replSrc->length();
1739 return len;
1740}
1741
1742
1743static UBool U_CALLCONV
1744repTextAccess(UText *ut, int64_t index, UBool forward) {
1745 const Replaceable *rep=(const Replaceable *)ut->context;
1746 int32_t length=rep->length(); // Full length of the input text (bigger than a chunk)
1747
1748 // clip the requested index to the limits of the text.
1749 int32_t index32 = pinIndex(index, length);
1750 U_ASSERT(index<=INT32_MAX);
1751
1752
1753 /*
1754 * Compute start/limit boundaries around index, for a segment of text
1755 * to be extracted.
1756 * To allow for the possibility that our user gave an index to the trailing
1757 * half of a surrogate pair, we must request one extra preceding char16_t when
1758 * going in the forward direction. This will ensure that the buffer has the
1759 * entire code point at the specified index.
1760 */
1761 if(forward) {
1762
1763 if (index32>=ut->chunkNativeStart && index32<ut->chunkNativeLimit) {
1764 // Buffer already contains the requested position.
1765 ut->chunkOffset = (int32_t)(index - ut->chunkNativeStart);
1766 return true;
1767 }
1768 if (index32>=length && ut->chunkNativeLimit==length) {
1769 // Request for end of string, and buffer already extends up to it.
1770 // Can't get the data, but don't change the buffer.
1771 ut->chunkOffset = length - (int32_t)ut->chunkNativeStart;
1772 return false;
1773 }
1774
1775 ut->chunkNativeLimit = index + REP_TEXT_CHUNK_SIZE - 1;
1776 // Going forward, so we want to have the buffer with stuff at and beyond
1777 // the requested index. The -1 gets us one code point before the
1778 // requested index also, to handle the case of the index being on
1779 // a trail surrogate of a surrogate pair.
1780 if(ut->chunkNativeLimit > length) {
1781 ut->chunkNativeLimit = length;
1782 }
1783 // unless buffer ran off end, start is index-1.
1784 ut->chunkNativeStart = ut->chunkNativeLimit - REP_TEXT_CHUNK_SIZE;
1785 if(ut->chunkNativeStart < 0) {
1786 ut->chunkNativeStart = 0;
1787 }
1788 } else {
1789 // Reverse iteration. Fill buffer with data preceding the requested index.
1790 if (index32>ut->chunkNativeStart && index32<=ut->chunkNativeLimit) {
1791 // Requested position already in buffer.
1792 ut->chunkOffset = index32 - (int32_t)ut->chunkNativeStart;
1793 return true;
1794 }
1795 if (index32==0 && ut->chunkNativeStart==0) {
1796 // Request for start, buffer already begins at start.
1797 // No data, but keep the buffer as is.
1798 ut->chunkOffset = 0;
1799 return false;
1800 }
1801
1802 // Figure out the bounds of the chunk to extract for reverse iteration.
1803 // Need to worry about chunk not splitting surrogate pairs, and while still
1804 // containing the data we need.
1805 // Fix by requesting a chunk that includes an extra char16_t at the end.
1806 // If this turns out to be a lead surrogate, we can lop it off and still have
1807 // the data we wanted.
1808 ut->chunkNativeStart = index32 + 1 - REP_TEXT_CHUNK_SIZE;
1809 if (ut->chunkNativeStart < 0) {
1810 ut->chunkNativeStart = 0;
1811 }
1812
1813 ut->chunkNativeLimit = index32 + 1;
1814 if (ut->chunkNativeLimit > length) {
1815 ut->chunkNativeLimit = length;
1816 }
1817 }
1818
1819 // Extract the new chunk of text from the Replaceable source.
1820 ReplExtra *ex = (ReplExtra *)ut->pExtra;
1821 // UnicodeString with its buffer a writable alias to the chunk buffer
1822 UnicodeString buffer(ex->s, 0 /*buffer length*/, REP_TEXT_CHUNK_SIZE /*buffer capacity*/);
1823 rep->extractBetween((int32_t)ut->chunkNativeStart, (int32_t)ut->chunkNativeLimit, buffer);
1824
1825 ut->chunkContents = ex->s;
1826 ut->chunkLength = (int32_t)(ut->chunkNativeLimit - ut->chunkNativeStart);
1827 ut->chunkOffset = (int32_t)(index32 - ut->chunkNativeStart);
1828
1829 // Surrogate pairs from the input text must not span chunk boundaries.
1830 // If end of chunk could be the start of a surrogate, trim it off.
1831 if (ut->chunkNativeLimit < length &&
1832 U16_IS_LEAD(ex->s[ut->chunkLength-1])) {
1833 ut->chunkLength--;
1834 ut->chunkNativeLimit--;
1835 if (ut->chunkOffset > ut->chunkLength) {
1836 ut->chunkOffset = ut->chunkLength;
1837 }
1838 }
1839
1840 // if the first char16_t in the chunk could be the trailing half of a surrogate pair,
1841 // trim it off.
1842 if(ut->chunkNativeStart>0 && U16_IS_TRAIL(ex->s[0])) {
1843 ++(ut->chunkContents);
1844 ++(ut->chunkNativeStart);
1845 --(ut->chunkLength);
1846 --(ut->chunkOffset);
1847 }
1848
1849 // adjust the index/chunkOffset to a code point boundary
1850 U16_SET_CP_START(ut->chunkContents, 0, ut->chunkOffset);
1851
1852 // Use fast indexing for get/setNativeIndex()
1853 ut->nativeIndexingLimit = ut->chunkLength;
1854
1855 return true;
1856}
1857
1858
1859
1860static int32_t U_CALLCONV
1861repTextExtract(UText *ut,
1862 int64_t start, int64_t limit,
1863 char16_t *dest, int32_t destCapacity,
1864 UErrorCode *status) {
1865 const Replaceable *rep=(const Replaceable *)ut->context;
1866 int32_t length=rep->length();
1867
1868 if(U_FAILURE(*status)) {
1869 return 0;
1870 }
1871 if(destCapacity<0 || (dest==nullptr && destCapacity>0)) {
1872 *status=U_ILLEGAL_ARGUMENT_ERROR;
1873 }
1874 if(start>limit) {
1875 *status=U_INDEX_OUTOFBOUNDS_ERROR;
1876 return 0;
1877 }
1878
1879 int32_t start32 = pinIndex(start, length);
1880 int32_t limit32 = pinIndex(limit, length);
1881
1882 // adjust start, limit if they point to trail half of surrogates
1883 if (start32<length && U16_IS_TRAIL(rep->charAt(start32)) &&
1884 U_IS_SUPPLEMENTARY(rep->char32At(start32))){
1885 start32--;
1886 }
1887 if (limit32<length && U16_IS_TRAIL(rep->charAt(limit32)) &&
1888 U_IS_SUPPLEMENTARY(rep->char32At(limit32))){
1889 limit32--;
1890 }
1891
1892 length=limit32-start32;
1893 if(length>destCapacity) {
1894 limit32 = start32 + destCapacity;
1895 }
1896 UnicodeString buffer(dest, 0, destCapacity); // writable alias
1897 rep->extractBetween(start32, limit32, buffer);
1898 repTextAccess(ut, limit32, true);
1899
1900 return u_terminateUChars(dest, destCapacity, length, status);
1901}
1902
1903static int32_t U_CALLCONV
1904repTextReplace(UText *ut,
1905 int64_t start, int64_t limit,
1906 const char16_t *src, int32_t length,
1907 UErrorCode *status) {
1908 Replaceable *rep=(Replaceable *)ut->context;
1909 int32_t oldLength;
1910
1911 if(U_FAILURE(*status)) {
1912 return 0;
1913 }
1914 if(src==nullptr && length!=0) {
1915 *status=U_ILLEGAL_ARGUMENT_ERROR;
1916 return 0;
1917 }
1918 oldLength=rep->length(); // will subtract from new length
1919 if(start>limit ) {
1920 *status=U_INDEX_OUTOFBOUNDS_ERROR;
1921 return 0;
1922 }
1923
1924 int32_t start32 = pinIndex(start, oldLength);
1925 int32_t limit32 = pinIndex(limit, oldLength);
1926
1927 // Snap start & limit to code point boundaries.
1928 if (start32<oldLength && U16_IS_TRAIL(rep->charAt(start32)) &&
1929 start32>0 && U16_IS_LEAD(rep->charAt(start32-1)))
1930 {
1931 start32--;
1932 }
1933 if (limit32<oldLength && U16_IS_LEAD(rep->charAt(limit32-1)) &&
1934 U16_IS_TRAIL(rep->charAt(limit32)))
1935 {
1936 limit32++;
1937 }
1938
1939 // Do the actual replace operation using methods of the Replaceable class
1940 UnicodeString replStr((UBool)(length<0), src, length); // read-only alias
1941 rep->handleReplaceBetween(start32, limit32, replStr);
1942 int32_t newLength = rep->length();
1943 int32_t lengthDelta = newLength - oldLength;
1944
1945 // Is the UText chunk buffer OK?
1946 if (ut->chunkNativeLimit > start32) {
1947 // this replace operation may have impacted the current chunk.
1948 // invalidate it, which will force a reload on the next access.
1949 invalidateChunk(ut);
1950 }
1951
1952 // set the iteration position to the end of the newly inserted replacement text.
1953 int32_t newIndexPos = limit32 + lengthDelta;
1954 repTextAccess(ut, newIndexPos, true);
1955
1956 return lengthDelta;
1957}
1958
1959
1960static void U_CALLCONV
1961repTextCopy(UText *ut,
1962 int64_t start, int64_t limit,
1963 int64_t destIndex,
1964 UBool move,
1965 UErrorCode *status)
1966{
1967 Replaceable *rep=(Replaceable *)ut->context;
1968 int32_t length=rep->length();
1969
1970 if(U_FAILURE(*status)) {
1971 return;
1972 }
1973 if (start>limit || (start<destIndex && destIndex<limit))
1974 {
1975 *status=U_INDEX_OUTOFBOUNDS_ERROR;
1976 return;
1977 }
1978
1979 int32_t start32 = pinIndex(start, length);
1980 int32_t limit32 = pinIndex(limit, length);
1981 int32_t destIndex32 = pinIndex(destIndex, length);
1982
1983 // TODO: snap input parameters to code point boundaries.
1984
1985 if(move) {
1986 // move: copy to destIndex, then replace original with nothing
1987 int32_t segLength=limit32-start32;
1988 rep->copy(start32, limit32, destIndex32);
1989 if(destIndex32<start32) {
1990 start32+=segLength;
1991 limit32+=segLength;
1992 }
1993 rep->handleReplaceBetween(start32, limit32, UnicodeString());
1994 } else {
1995 // copy
1996 rep->copy(start32, limit32, destIndex32);
1997 }
1998
1999 // If the change to the text touched the region in the chunk buffer,
2000 // invalidate the buffer.
2001 int32_t firstAffectedIndex = destIndex32;
2002 if (move && start32<firstAffectedIndex) {
2003 firstAffectedIndex = start32;
2004 }
2005 if (firstAffectedIndex < ut->chunkNativeLimit) {
2006 // changes may have affected range covered by the chunk
2007 invalidateChunk(ut);
2008 }
2009
2010 // Put iteration position at the newly inserted (moved) block,
2011 int32_t nativeIterIndex = destIndex32 + limit32 - start32;
2012 if (move && destIndex32>start32) {
2013 // moved a block of text towards the end of the string.
2014 nativeIterIndex = destIndex32;
2015 }
2016
2017 // Set position, reload chunk if needed.
2018 repTextAccess(ut, nativeIterIndex, true);
2019}
2020
2021static const struct UTextFuncs repFuncs =
2022{
2023 sizeof(UTextFuncs),
2024 0, 0, 0, // Reserved alignment padding
2025 repTextClone,
2026 repTextLength,
2027 repTextAccess,
2028 repTextExtract,
2029 repTextReplace,
2030 repTextCopy,
2031 nullptr, // MapOffsetToNative,
2032 nullptr, // MapIndexToUTF16,
2033 repTextClose,
2034 nullptr, // spare 1
2035 nullptr, // spare 2
2036 nullptr // spare 3
2037};
2038
2039
2040U_CAPI UText * U_EXPORT2
2041utext_openReplaceable(UText *ut, Replaceable *rep, UErrorCode *status)
2042{
2043 if(U_FAILURE(*status)) {
2044 return nullptr;
2045 }
2046 if(rep==nullptr) {
2047 *status=U_ILLEGAL_ARGUMENT_ERROR;
2048 return nullptr;
2049 }
2050 ut = utext_setup(ut, sizeof(ReplExtra), status);
2051 if(U_FAILURE(*status)) {
2052 return ut;
2053 }
2054
2055 ut->providerProperties = I32_FLAG(UTEXT_PROVIDER_WRITABLE);
2056 if(rep->hasMetaData()) {
2057 ut->providerProperties |=I32_FLAG(UTEXT_PROVIDER_HAS_META_DATA);
2058 }
2059
2060 ut->pFuncs = &repFuncs;
2061 ut->context = rep;
2062 return ut;
2063}
2064
2065U_CDECL_END
2066
2067
2068
2069
2070
2071
2072
2073
2074//------------------------------------------------------------------------------
2075//
2076// UText implementation for UnicodeString (read/write) and
2077// for const UnicodeString (read only)
2078// (same implementation, only the flags are different)
2079//
2080// Use of UText data members:
2081// context pointer to UnicodeString
2082// p pointer to UnicodeString IF this UText owns the string
2083// and it must be deleted on close(). nullptr otherwise.
2084//
2085//------------------------------------------------------------------------------
2086
2087U_CDECL_BEGIN
2088
2089
2090static UText * U_CALLCONV
2091unistrTextClone(UText *dest, const UText *src, UBool deep, UErrorCode *status) {
2092 // First do a generic shallow clone. Does everything needed for the UText struct itself.
2093 dest = shallowTextClone(dest, src, status);
2094
2095 // For deep clones, make a copy of the UnicodeSring.
2096 // The copied UnicodeString storage is owned by the newly created UText clone.
2097 // A non-nullptr pointer in UText.p is the signal to the close() function to delete
2098 // the UText.
2099 //
2100 if (deep && U_SUCCESS(*status)) {
2101 const UnicodeString *srcString = (const UnicodeString *)src->context;
2102 dest->context = new UnicodeString(*srcString);
2103 dest->providerProperties |= I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT);
2104
2105 // with deep clone, the copy is writable, even when the source is not.
2106 dest->providerProperties |= I32_FLAG(UTEXT_PROVIDER_WRITABLE);
2107 }
2108 return dest;
2109}
2110
2111static void U_CALLCONV
2112unistrTextClose(UText *ut) {
2113 // Most of the work of close is done by the generic UText framework close.
2114 // All that needs to be done here is delete the UnicodeString if the UText
2115 // owns it. This occurs if the UText was created by cloning.
2116 if (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT)) {
2117 UnicodeString *str = (UnicodeString *)ut->context;
2118 delete str;
2119 ut->context = nullptr;
2120 }
2121}
2122
2123
2124static int64_t U_CALLCONV
2125unistrTextLength(UText *t) {
2126 return ((const UnicodeString *)t->context)->length();
2127}
2128
2129
2130static UBool U_CALLCONV
2131unistrTextAccess(UText *ut, int64_t index, UBool forward) {
2132 int32_t length = ut->chunkLength;
2133 ut->chunkOffset = pinIndex(index, length);
2134
2135 // Check whether request is at the start or end
2136 UBool retVal = (forward && index<length) || (!forward && index>0);
2137 return retVal;
2138}
2139
2140
2141
2142static int32_t U_CALLCONV
2143unistrTextExtract(UText *t,
2144 int64_t start, int64_t limit,
2145 char16_t *dest, int32_t destCapacity,
2146 UErrorCode *pErrorCode) {
2147 const UnicodeString *us=(const UnicodeString *)t->context;
2148 int32_t length=us->length();
2149
2150 if(U_FAILURE(*pErrorCode)) {
2151 return 0;
2152 }
2153 if(destCapacity<0 || (dest==nullptr && destCapacity>0)) {
2154 *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
2155 }
2156 if(start<0 || start>limit) {
2157 *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR;
2158 return 0;
2159 }
2160
2161 int32_t start32 = start<length ? us->getChar32Start((int32_t)start) : length;
2162 int32_t limit32 = limit<length ? us->getChar32Start((int32_t)limit) : length;
2163
2164 length=limit32-start32;
2165 if (destCapacity>0 && dest!=nullptr) {
2166 int32_t trimmedLength = length;
2167 if(trimmedLength>destCapacity) {
2168 trimmedLength=destCapacity;
2169 }
2170 us->extract(start32, trimmedLength, dest);
2171 t->chunkOffset = start32+trimmedLength;
2172 } else {
2173 t->chunkOffset = start32;
2174 }
2175 u_terminateUChars(dest, destCapacity, length, pErrorCode);
2176 return length;
2177}
2178
2179static int32_t U_CALLCONV
2180unistrTextReplace(UText *ut,
2181 int64_t start, int64_t limit,
2182 const char16_t *src, int32_t length,
2183 UErrorCode *pErrorCode) {
2184 UnicodeString *us=(UnicodeString *)ut->context;
2185 int32_t oldLength;
2186
2187 if(U_FAILURE(*pErrorCode)) {
2188 return 0;
2189 }
2190 if(src==nullptr && length!=0) {
2191 *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
2192 }
2193 if(start>limit) {
2194 *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR;
2195 return 0;
2196 }
2197 oldLength=us->length();
2198 int32_t start32 = pinIndex(start, oldLength);
2199 int32_t limit32 = pinIndex(limit, oldLength);
2200 if (start32 < oldLength) {
2201 start32 = us->getChar32Start(start32);
2202 }
2203 if (limit32 < oldLength) {
2204 limit32 = us->getChar32Start(limit32);
2205 }
2206
2207 // replace
2208 us->replace(start32, limit32-start32, src, length);
2209 int32_t newLength = us->length();
2210
2211 // Update the chunk description.
2212 ut->chunkContents = us->getBuffer();
2213 ut->chunkLength = newLength;
2214 ut->chunkNativeLimit = newLength;
2215 ut->nativeIndexingLimit = newLength;
2216
2217 // Set iteration position to the point just following the newly inserted text.
2218 int32_t lengthDelta = newLength - oldLength;
2219 ut->chunkOffset = limit32 + lengthDelta;
2220
2221 return lengthDelta;
2222}
2223
2224static void U_CALLCONV
2225unistrTextCopy(UText *ut,
2226 int64_t start, int64_t limit,
2227 int64_t destIndex,
2228 UBool move,
2229 UErrorCode *pErrorCode) {
2230 UnicodeString *us=(UnicodeString *)ut->context;
2231 int32_t length=us->length();
2232
2233 if(U_FAILURE(*pErrorCode)) {
2234 return;
2235 }
2236 int32_t start32 = pinIndex(start, length);
2237 int32_t limit32 = pinIndex(limit, length);
2238 int32_t destIndex32 = pinIndex(destIndex, length);
2239
2240 if( start32>limit32 || (start32<destIndex32 && destIndex32<limit32)) {
2241 *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR;
2242 return;
2243 }
2244
2245 if(move) {
2246 // move: copy to destIndex, then remove original
2247 int32_t segLength=limit32-start32;
2248 us->copy(start32, limit32, destIndex32);
2249 if(destIndex32<start32) {
2250 start32+=segLength;
2251 }
2252 us->remove(start32, segLength);
2253 } else {
2254 // copy
2255 us->copy(start32, limit32, destIndex32);
2256 }
2257
2258 // update chunk description, set iteration position.
2259 ut->chunkContents = us->getBuffer();
2260 if (move==false) {
2261 // copy operation, string length grows
2262 ut->chunkLength += limit32-start32;
2263 ut->chunkNativeLimit = ut->chunkLength;
2264 ut->nativeIndexingLimit = ut->chunkLength;
2265 }
2266
2267 // Iteration position to end of the newly inserted text.
2268 ut->chunkOffset = destIndex32+limit32-start32;
2269 if (move && destIndex32>start32) {
2270 ut->chunkOffset = destIndex32;
2271 }
2272
2273}
2274
2275static const struct UTextFuncs unistrFuncs =
2276{
2277 sizeof(UTextFuncs),
2278 0, 0, 0, // Reserved alignment padding
2279 unistrTextClone,
2280 unistrTextLength,
2281 unistrTextAccess,
2282 unistrTextExtract,
2283 unistrTextReplace,
2284 unistrTextCopy,
2285 nullptr, // MapOffsetToNative,
2286 nullptr, // MapIndexToUTF16,
2287 unistrTextClose,
2288 nullptr, // spare 1
2289 nullptr, // spare 2
2290 nullptr // spare 3
2291};
2292
2293
2294
2295U_CDECL_END
2296
2297
2298U_CAPI UText * U_EXPORT2
2299utext_openUnicodeString(UText *ut, UnicodeString *s, UErrorCode *status) {
2300 ut = utext_openConstUnicodeString(ut, s, status);
2301 if (U_SUCCESS(*status)) {
2302 ut->providerProperties |= I32_FLAG(UTEXT_PROVIDER_WRITABLE);
2303 }
2304 return ut;
2305}
2306
2307
2308
2309U_CAPI UText * U_EXPORT2
2310utext_openConstUnicodeString(UText *ut, const UnicodeString *s, UErrorCode *status) {
2311 if (U_SUCCESS(*status) && s->isBogus()) {
2312 // The UnicodeString is bogus, but we still need to detach the UText
2313 // from whatever it was hooked to before, if anything.
2314 utext_openUChars(ut, nullptr, 0, status);
2315 *status = U_ILLEGAL_ARGUMENT_ERROR;
2316 return ut;
2317 }
2318 ut = utext_setup(ut, 0, status);
2319 // note: use the standard (writable) function table for UnicodeString.
2320 // The flag settings disable writing, so having the functions in
2321 // the table is harmless.
2322 if (U_SUCCESS(*status)) {
2323 ut->pFuncs = &unistrFuncs;
2324 ut->context = s;
2325 ut->providerProperties = I32_FLAG(UTEXT_PROVIDER_STABLE_CHUNKS);
2326 ut->chunkContents = s->getBuffer();
2327 ut->chunkLength = s->length();
2328 ut->chunkNativeStart = 0;
2329 ut->chunkNativeLimit = ut->chunkLength;
2330 ut->nativeIndexingLimit = ut->chunkLength;
2331 }
2332 return ut;
2333}
2334
2335//------------------------------------------------------------------------------
2336//
2337// UText implementation for const char16_t * strings
2338//
2339// Use of UText data members:
2340// context pointer to UnicodeString
2341// a length. -1 if not yet known.
2342//
2343// TODO: support 64 bit lengths.
2344//
2345//------------------------------------------------------------------------------
2346
2347U_CDECL_BEGIN
2348
2349
2350static UText * U_CALLCONV
2351ucstrTextClone(UText *dest, const UText * src, UBool deep, UErrorCode * status) {
2352 // First do a generic shallow clone.
2353 dest = shallowTextClone(dest, src, status);
2354
2355 // For deep clones, make a copy of the string.
2356 // The copied storage is owned by the newly created clone.
2357 // A non-nullptr pointer in UText.p is the signal to the close() function to delete
2358 // it.
2359 //
2360 if (deep && U_SUCCESS(*status)) {
2361 U_ASSERT(utext_nativeLength(dest) < INT32_MAX);
2362 int32_t len = (int32_t)utext_nativeLength(dest);
2363
2364 // The cloned string IS going to be NUL terminated, whether or not the original was.
2365 const char16_t *srcStr = (const char16_t *)src->context;
2366 char16_t *copyStr = (char16_t *)uprv_malloc((len+1) * sizeof(char16_t));
2367 if (copyStr == nullptr) {
2368 *status = U_MEMORY_ALLOCATION_ERROR;
2369 } else {
2370 int64_t i;
2371 for (i=0; i<len; i++) {
2372 copyStr[i] = srcStr[i];
2373 }
2374 copyStr[len] = 0;
2375 dest->context = copyStr;
2376 dest->providerProperties |= I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT);
2377 }
2378 }
2379 return dest;
2380}
2381
2382
2383static void U_CALLCONV
2384ucstrTextClose(UText *ut) {
2385 // Most of the work of close is done by the generic UText framework close.
2386 // All that needs to be done here is delete the string if the UText
2387 // owns it. This occurs if the UText was created by cloning.
2388 if (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT)) {
2389 char16_t *s = (char16_t *)ut->context;
2390 uprv_free(s);
2391 ut->context = nullptr;
2392 }
2393}
2394
2395
2396
2397static int64_t U_CALLCONV
2398ucstrTextLength(UText *ut) {
2399 if (ut->a < 0) {
2400 // null terminated, we don't yet know the length. Scan for it.
2401 // Access is not convenient for doing this
2402 // because the current iteration position can't be changed.
2403 const char16_t *str = (const char16_t *)ut->context;
2404 for (;;) {
2405 if (str[ut->chunkNativeLimit] == 0) {
2406 break;
2407 }
2408 ut->chunkNativeLimit++;
2409 }
2410 ut->a = ut->chunkNativeLimit;
2411 ut->chunkLength = (int32_t)ut->chunkNativeLimit;
2412 ut->nativeIndexingLimit = ut->chunkLength;
2413 ut->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
2414 }
2415 return ut->a;
2416}
2417
2418
2419static UBool U_CALLCONV
2420ucstrTextAccess(UText *ut, int64_t index, UBool forward) {
2421 const char16_t *str = (const char16_t *)ut->context;
2422
2423 // pin the requested index to the bounds of the string,
2424 // and set current iteration position.
2425 if (index<0) {
2426 index = 0;
2427 } else if (index < ut->chunkNativeLimit) {
2428 // The request data is within the chunk as it is known so far.
2429 // Put index on a code point boundary.
2430 U16_SET_CP_START(str, 0, index);
2431 } else if (ut->a >= 0) {
2432 // We know the length of this string, and the user is requesting something
2433 // at or beyond the length. Pin the requested index to the length.
2434 index = ut->a;
2435 } else {
2436 // Null terminated string, length not yet known, and the requested index
2437 // is beyond where we have scanned so far.
2438 // Scan to 32 UChars beyond the requested index. The strategy here is
2439 // to avoid fully scanning a long string when the caller only wants to
2440 // see a few characters at its beginning.
2441 int32_t scanLimit = (int32_t)index + 32;
2442 if ((index + 32)>INT32_MAX || (index + 32)<0 ) { // note: int64 expression
2443 scanLimit = INT32_MAX;
2444 }
2445
2446 int32_t chunkLimit = (int32_t)ut->chunkNativeLimit;
2447 for (; chunkLimit<scanLimit; chunkLimit++) {
2448 if (str[chunkLimit] == 0) {
2449 // We found the end of the string. Remember it, pin the requested index to it,
2450 // and bail out of here.
2451 ut->a = chunkLimit;
2452 ut->chunkLength = chunkLimit;
2453 ut->nativeIndexingLimit = chunkLimit;
2454 if (index >= chunkLimit) {
2455 index = chunkLimit;
2456 } else {
2457 U16_SET_CP_START(str, 0, index);
2458 }
2459
2460 ut->chunkNativeLimit = chunkLimit;
2461 ut->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
2462 goto breakout;
2463 }
2464 }
2465 // We scanned through the next batch of UChars without finding the end.
2466 U16_SET_CP_START(str, 0, index);
2467 if (chunkLimit == INT32_MAX) {
2468 // Scanned to the limit of a 32 bit length.
2469 // Forceably trim the overlength string back so length fits in int32
2470 // TODO: add support for 64 bit strings.
2471 ut->a = chunkLimit;
2472 ut->chunkLength = chunkLimit;
2473 ut->nativeIndexingLimit = chunkLimit;
2474 if (index > chunkLimit) {
2475 index = chunkLimit;
2476 }
2477 ut->chunkNativeLimit = chunkLimit;
2478 ut->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
2479 } else {
2480 // The endpoint of a chunk must not be left in the middle of a surrogate pair.
2481 // If the current end is on a lead surrogate, back the end up by one.
2482 // It doesn't matter if the end char happens to be an unpaired surrogate,
2483 // and it's simpler not to worry about it.
2484 if (U16_IS_LEAD(str[chunkLimit-1])) {
2485 --chunkLimit;
2486 }
2487 // Null-terminated chunk with end still unknown.
2488 // Update the chunk length to reflect what has been scanned thus far.
2489 // That the full length is still unknown is (still) flagged by
2490 // ut->a being < 0.
2491 ut->chunkNativeLimit = chunkLimit;
2492 ut->nativeIndexingLimit = chunkLimit;
2493 ut->chunkLength = chunkLimit;
2494 }
2495
2496 }
2497breakout:
2498 U_ASSERT(index<=INT32_MAX);
2499 ut->chunkOffset = (int32_t)index;
2500
2501 // Check whether request is at the start or end
2502 UBool retVal = (forward && index<ut->chunkNativeLimit) || (!forward && index>0);
2503 return retVal;
2504}
2505
2506
2507
2508static int32_t U_CALLCONV
2509ucstrTextExtract(UText *ut,
2510 int64_t start, int64_t limit,
2511 char16_t *dest, int32_t destCapacity,
2512 UErrorCode *pErrorCode)
2513{
2514 if(U_FAILURE(*pErrorCode)) {
2515 return 0;
2516 }
2517 if(destCapacity<0 || (dest==nullptr && destCapacity>0) || start>limit) {
2518 *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
2519 return 0;
2520 }
2521
2522 //const char16_t *s=(const char16_t *)ut->context;
2523 int32_t si, di;
2524
2525 int32_t start32;
2526 int32_t limit32;
2527
2528 // Access the start. Does two things we need:
2529 // Pins 'start' to the length of the string, if it came in out-of-bounds.
2530 // Snaps 'start' to the beginning of a code point.
2531 ucstrTextAccess(ut, start, true);
2532 const char16_t *s=ut->chunkContents;
2533 start32 = ut->chunkOffset;
2534
2535 int32_t strLength=(int32_t)ut->a;
2536 if (strLength >= 0) {
2537 limit32 = pinIndex(limit, strLength);
2538 } else {
2539 limit32 = pinIndex(limit, INT32_MAX);
2540 }
2541 di = 0;
2542 for (si=start32; si<limit32; si++) {
2543 if (strLength<0 && s[si]==0) {
2544 // Just hit the end of a null-terminated string.
2545 ut->a = si; // set string length for this UText
2546 ut->chunkNativeLimit = si;
2547 ut->chunkLength = si;
2548 ut->nativeIndexingLimit = si;
2549 strLength = si;
2550 limit32 = si;
2551 break;
2552 }
2553 U_ASSERT(di>=0); /* to ensure di never exceeds INT32_MAX, which must not happen logically */
2554 if (di<destCapacity) {
2555 // only store if there is space.
2556 dest[di] = s[si];
2557 } else {
2558 if (strLength>=0) {
2559 // We have filled the destination buffer, and the string length is known.
2560 // Cut the loop short. There is no need to scan string termination.
2561 di = limit32 - start32;
2562 si = limit32;
2563 break;
2564 }
2565 }
2566 di++;
2567 }
2568
2569 // If the limit index points to a lead surrogate of a pair,
2570 // add the corresponding trail surrogate to the destination.
2571 if (si>0 && U16_IS_LEAD(s[si-1]) &&
2572 ((si<strLength || strLength<0) && U16_IS_TRAIL(s[si])))
2573 {
2574 if (di<destCapacity) {
2575 // store only if there is space in the output buffer.
2576 dest[di++] = s[si];
2577 }
2578 si++;
2579 }
2580
2581 // Put iteration position at the point just following the extracted text
2582 if (si <= ut->chunkNativeLimit) {
2583 ut->chunkOffset = si;
2584 } else {
2585 ucstrTextAccess(ut, si, true);
2586 }
2587
2588 // Add a terminating NUL if space in the buffer permits,
2589 // and set the error status as required.
2590 u_terminateUChars(dest, destCapacity, di, pErrorCode);
2591 return di;
2592}
2593
2594static const struct UTextFuncs ucstrFuncs =
2595{
2596 sizeof(UTextFuncs),
2597 0, 0, 0, // Reserved alignment padding
2598 ucstrTextClone,
2599 ucstrTextLength,
2600 ucstrTextAccess,
2601 ucstrTextExtract,
2602 nullptr, // Replace
2603 nullptr, // Copy
2604 nullptr, // MapOffsetToNative,
2605 nullptr, // MapIndexToUTF16,
2606 ucstrTextClose,
2607 nullptr, // spare 1
2608 nullptr, // spare 2
2609 nullptr, // spare 3
2610};
2611
2612U_CDECL_END
2613
2614static const char16_t gEmptyUString[] = {0};
2615
2616U_CAPI UText * U_EXPORT2
2617utext_openUChars(UText *ut, const char16_t *s, int64_t length, UErrorCode *status) {
2618 if (U_FAILURE(*status)) {
2619 return nullptr;
2620 }
2621 if(s==nullptr && length==0) {
2622 s = gEmptyUString;
2623 }
2624 if (s==nullptr || length < -1 || length>INT32_MAX) {
2625 *status = U_ILLEGAL_ARGUMENT_ERROR;
2626 return nullptr;
2627 }
2628 ut = utext_setup(ut, 0, status);
2629 if (U_SUCCESS(*status)) {
2630 ut->pFuncs = &ucstrFuncs;
2631 ut->context = s;
2632 ut->providerProperties = I32_FLAG(UTEXT_PROVIDER_STABLE_CHUNKS);
2633 if (length==-1) {
2634 ut->providerProperties |= I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
2635 }
2636 ut->a = length;
2637 ut->chunkContents = s;
2638 ut->chunkNativeStart = 0;
2639 ut->chunkNativeLimit = length>=0? length : 0;
2640 ut->chunkLength = (int32_t)ut->chunkNativeLimit;
2641 ut->chunkOffset = 0;
2642 ut->nativeIndexingLimit = ut->chunkLength;
2643 }
2644 return ut;
2645}
2646
2647
2648//------------------------------------------------------------------------------
2649//
2650// UText implementation for text from ICU CharacterIterators
2651//
2652// Use of UText data members:
2653// context pointer to the CharacterIterator
2654// a length of the full text.
2655// p pointer to buffer 1
2656// b start index of local buffer 1 contents
2657// q pointer to buffer 2
2658// c start index of local buffer 2 contents
2659// r pointer to the character iterator if the UText owns it.
2660// Null otherwise.
2661//
2662//------------------------------------------------------------------------------
2663#define CIBufSize 16
2664
2665U_CDECL_BEGIN
2666static void U_CALLCONV
2667charIterTextClose(UText *ut) {
2668 // Most of the work of close is done by the generic UText framework close.
2669 // All that needs to be done here is delete the CharacterIterator if the UText
2670 // owns it. This occurs if the UText was created by cloning.
2671 CharacterIterator *ci = (CharacterIterator *)ut->r;
2672 delete ci;
2673 ut->r = nullptr;
2674}
2675
2676static int64_t U_CALLCONV
2677charIterTextLength(UText *ut) {
2678 return (int32_t)ut->a;
2679}
2680
2681static UBool U_CALLCONV
2682charIterTextAccess(UText *ut, int64_t index, UBool forward) {
2683 CharacterIterator *ci = (CharacterIterator *)ut->context;
2684
2685 int32_t clippedIndex = (int32_t)index;
2686 if (clippedIndex<0) {
2687 clippedIndex=0;
2688 } else if (clippedIndex>=ut->a) {
2689 clippedIndex=(int32_t)ut->a;
2690 }
2691 int32_t neededIndex = clippedIndex;
2692 if (!forward && neededIndex>0) {
2693 // reverse iteration, want the position just before what was asked for.
2694 neededIndex--;
2695 } else if (forward && neededIndex==ut->a && neededIndex>0) {
2696 // Forward iteration, don't ask for something past the end of the text.
2697 neededIndex--;
2698 }
2699
2700 // Find the native index of the start of the buffer containing what we want.
2701 neededIndex -= neededIndex % CIBufSize;
2702
2703 char16_t *buf = nullptr;
2704 UBool needChunkSetup = true;
2705 int i;
2706 if (ut->chunkNativeStart == neededIndex) {
2707 // The buffer we want is already the current chunk.
2708 needChunkSetup = false;
2709 } else if (ut->b == neededIndex) {
2710 // The first buffer (buffer p) has what we need.
2711 buf = (char16_t *)ut->p;
2712 } else if (ut->c == neededIndex) {
2713 // The second buffer (buffer q) has what we need.
2714 buf = (char16_t *)ut->q;
2715 } else {
2716 // Neither buffer already has what we need.
2717 // Load new data from the character iterator.
2718 // Use the buf that is not the current buffer.
2719 buf = (char16_t *)ut->p;
2720 if (ut->p == ut->chunkContents) {
2721 buf = (char16_t *)ut->q;
2722 }
2723 ci->setIndex(neededIndex);
2724 for (i=0; i<CIBufSize; i++) {
2725 buf[i] = ci->nextPostInc();
2726 if (i+neededIndex > ut->a) {
2727 break;
2728 }
2729 }
2730 }
2731
2732 // We have a buffer with the data we need.
2733 // Set it up as the current chunk, if it wasn't already.
2734 if (needChunkSetup) {
2735 ut->chunkContents = buf;
2736 ut->chunkLength = CIBufSize;
2737 ut->chunkNativeStart = neededIndex;
2738 ut->chunkNativeLimit = neededIndex + CIBufSize;
2739 if (ut->chunkNativeLimit > ut->a) {
2740 ut->chunkNativeLimit = ut->a;
2741 ut->chunkLength = (int32_t)(ut->chunkNativeLimit)-(int32_t)(ut->chunkNativeStart);
2742 }
2743 ut->nativeIndexingLimit = ut->chunkLength;
2744 U_ASSERT(ut->chunkOffset>=0 && ut->chunkOffset<=CIBufSize);
2745 }
2746 ut->chunkOffset = clippedIndex - (int32_t)ut->chunkNativeStart;
2747 UBool success = (forward? ut->chunkOffset<ut->chunkLength : ut->chunkOffset>0);
2748 return success;
2749}
2750
2751static UText * U_CALLCONV
2752charIterTextClone(UText *dest, const UText *src, UBool deep, UErrorCode * status) {
2753 if (U_FAILURE(*status)) {
2754 return nullptr;
2755 }
2756
2757 if (deep) {
2758 // There is no CharacterIterator API for cloning the underlying text storage.
2759 *status = U_UNSUPPORTED_ERROR;
2760 return nullptr;
2761 } else {
2762 CharacterIterator *srcCI =(CharacterIterator *)src->context;
2763 srcCI = srcCI->clone();
2764 dest = utext_openCharacterIterator(dest, srcCI, status);
2765 if (U_FAILURE(*status)) {
2766 return dest;
2767 }
2768 // cast off const on getNativeIndex.
2769 // For CharacterIterator based UTexts, this is safe, the operation is const.
2770 int64_t ix = utext_getNativeIndex((UText *)src);
2771 utext_setNativeIndex(dest, ix);
2772 dest->r = srcCI; // flags that this UText owns the CharacterIterator
2773 }
2774 return dest;
2775}
2776
2777static int32_t U_CALLCONV
2778charIterTextExtract(UText *ut,
2779 int64_t start, int64_t limit,
2780 char16_t *dest, int32_t destCapacity,
2781 UErrorCode *status)
2782{
2783 if(U_FAILURE(*status)) {
2784 return 0;
2785 }
2786 if(destCapacity<0 || (dest==nullptr && destCapacity>0) || start>limit) {
2787 *status=U_ILLEGAL_ARGUMENT_ERROR;
2788 return 0;
2789 }
2790 int32_t length = (int32_t)ut->a;
2791 int32_t start32 = pinIndex(start, length);
2792 int32_t limit32 = pinIndex(limit, length);
2793 int32_t desti = 0;
2794 int32_t srci;
2795 int32_t copyLimit;
2796
2797 CharacterIterator *ci = (CharacterIterator *)ut->context;
2798 ci->setIndex32(start32); // Moves ix to lead of surrogate pair, if needed.
2799 srci = ci->getIndex();
2800 copyLimit = srci;
2801 while (srci<limit32) {
2802 UChar32 c = ci->next32PostInc();
2803 int32_t len = U16_LENGTH(c);
2804 U_ASSERT(desti+len>0); /* to ensure desti+len never exceeds MAX_INT32, which must not happen logically */
2805 if (desti+len <= destCapacity) {
2806 U16_APPEND_UNSAFE(dest, desti, c);
2807 copyLimit = srci+len;
2808 } else {
2809 desti += len;
2810 *status = U_BUFFER_OVERFLOW_ERROR;
2811 }
2812 srci += len;
2813 }
2814
2815 charIterTextAccess(ut, copyLimit, true);
2816
2817 u_terminateUChars(dest, destCapacity, desti, status);
2818 return desti;
2819}
2820
2821static const struct UTextFuncs charIterFuncs =
2822{
2823 sizeof(UTextFuncs),
2824 0, 0, 0, // Reserved alignment padding
2825 charIterTextClone,
2826 charIterTextLength,
2827 charIterTextAccess,
2828 charIterTextExtract,
2829 nullptr, // Replace
2830 nullptr, // Copy
2831 nullptr, // MapOffsetToNative,
2832 nullptr, // MapIndexToUTF16,
2833 charIterTextClose,
2834 nullptr, // spare 1
2835 nullptr, // spare 2
2836 nullptr // spare 3
2837};
2838U_CDECL_END
2839
2840
2841U_CAPI UText * U_EXPORT2
2842utext_openCharacterIterator(UText *ut, CharacterIterator *ci, UErrorCode *status) {
2843 if (U_FAILURE(*status)) {
2844 return nullptr;
2845 }
2846
2847 if (ci->startIndex() > 0) {
2848 // No support for CharacterIterators that do not start indexing from zero.
2849 *status = U_UNSUPPORTED_ERROR;
2850 return nullptr;
2851 }
2852
2853 // Extra space in UText for 2 buffers of CIBufSize UChars each.
2854 int32_t extraSpace = 2 * CIBufSize * sizeof(char16_t);
2855 ut = utext_setup(ut, extraSpace, status);
2856 if (U_SUCCESS(*status)) {
2857 ut->pFuncs = &charIterFuncs;
2858 ut->context = ci;
2859 ut->providerProperties = 0;
2860 ut->a = ci->endIndex(); // Length of text
2861 ut->p = ut->pExtra; // First buffer
2862 ut->b = -1; // Native index of first buffer contents
2863 ut->q = (char16_t*)ut->pExtra+CIBufSize; // Second buffer
2864 ut->c = -1; // Native index of second buffer contents
2865
2866 // Initialize current chunk contents to be empty.
2867 // First access will fault something in.
2868 // Note: The initial nativeStart and chunkOffset must sum to zero
2869 // so that getNativeIndex() will correctly compute to zero
2870 // if no call to Access() has ever been made. They can't be both
2871 // zero without Access() thinking that the chunk is valid.
2872 ut->chunkContents = (char16_t *)ut->p;
2873 ut->chunkNativeStart = -1;
2874 ut->chunkOffset = 1;
2875 ut->chunkNativeLimit = 0;
2876 ut->chunkLength = 0;
2877 ut->nativeIndexingLimit = ut->chunkOffset; // enables native indexing
2878 }
2879 return ut;
2880}
2881