1.1 --- /dev/null Thu Jan 01 00:00:00 1970 +0000
1.2 +++ b/intl/icu/source/common/utext.cpp Wed Dec 31 06:09:35 2014 +0100
1.3 @@ -0,0 +1,2837 @@
1.4 +/*
1.5 +*******************************************************************************
1.6 +*
1.7 +* Copyright (C) 2005-2012, International Business Machines
1.8 +* Corporation and others. All Rights Reserved.
1.9 +*
1.10 +*******************************************************************************
1.11 +* file name: utext.cpp
1.12 +* encoding: US-ASCII
1.13 +* tab size: 8 (not used)
1.14 +* indentation:4
1.15 +*
1.16 +* created on: 2005apr12
1.17 +* created by: Markus W. Scherer
1.18 +*/
1.19 +
1.20 +#include "unicode/utypes.h"
1.21 +#include "unicode/ustring.h"
1.22 +#include "unicode/unistr.h"
1.23 +#include "unicode/chariter.h"
1.24 +#include "unicode/utext.h"
1.25 +#include "unicode/utf.h"
1.26 +#include "unicode/utf8.h"
1.27 +#include "unicode/utf16.h"
1.28 +#include "ustr_imp.h"
1.29 +#include "cmemory.h"
1.30 +#include "cstring.h"
1.31 +#include "uassert.h"
1.32 +#include "putilimp.h"
1.33 +
1.34 +U_NAMESPACE_USE
1.35 +
1.36 +#define I32_FLAG(bitIndex) ((int32_t)1<<(bitIndex))
1.37 +
1.38 +
1.39 +static UBool
1.40 +utext_access(UText *ut, int64_t index, UBool forward) {
1.41 + return ut->pFuncs->access(ut, index, forward);
1.42 +}
1.43 +
1.44 +
1.45 +
1.46 +U_CAPI UBool U_EXPORT2
1.47 +utext_moveIndex32(UText *ut, int32_t delta) {
1.48 + UChar32 c;
1.49 + if (delta > 0) {
1.50 + do {
1.51 + if(ut->chunkOffset>=ut->chunkLength && !utext_access(ut, ut->chunkNativeLimit, TRUE)) {
1.52 + return FALSE;
1.53 + }
1.54 + c = ut->chunkContents[ut->chunkOffset];
1.55 + if (U16_IS_SURROGATE(c)) {
1.56 + c = utext_next32(ut);
1.57 + if (c == U_SENTINEL) {
1.58 + return FALSE;
1.59 + }
1.60 + } else {
1.61 + ut->chunkOffset++;
1.62 + }
1.63 + } while(--delta>0);
1.64 +
1.65 + } else if (delta<0) {
1.66 + do {
1.67 + if(ut->chunkOffset<=0 && !utext_access(ut, ut->chunkNativeStart, FALSE)) {
1.68 + return FALSE;
1.69 + }
1.70 + c = ut->chunkContents[ut->chunkOffset-1];
1.71 + if (U16_IS_SURROGATE(c)) {
1.72 + c = utext_previous32(ut);
1.73 + if (c == U_SENTINEL) {
1.74 + return FALSE;
1.75 + }
1.76 + } else {
1.77 + ut->chunkOffset--;
1.78 + }
1.79 + } while(++delta<0);
1.80 + }
1.81 +
1.82 + return TRUE;
1.83 +}
1.84 +
1.85 +
1.86 +U_CAPI int64_t U_EXPORT2
1.87 +utext_nativeLength(UText *ut) {
1.88 + return ut->pFuncs->nativeLength(ut);
1.89 +}
1.90 +
1.91 +
1.92 +U_CAPI UBool U_EXPORT2
1.93 +utext_isLengthExpensive(const UText *ut) {
1.94 + UBool r = (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE)) != 0;
1.95 + return r;
1.96 +}
1.97 +
1.98 +
1.99 +U_CAPI int64_t U_EXPORT2
1.100 +utext_getNativeIndex(const UText *ut) {
1.101 + if(ut->chunkOffset <= ut->nativeIndexingLimit) {
1.102 + return ut->chunkNativeStart+ut->chunkOffset;
1.103 + } else {
1.104 + return ut->pFuncs->mapOffsetToNative(ut);
1.105 + }
1.106 +}
1.107 +
1.108 +
1.109 +U_CAPI void U_EXPORT2
1.110 +utext_setNativeIndex(UText *ut, int64_t index) {
1.111 + if(index<ut->chunkNativeStart || index>=ut->chunkNativeLimit) {
1.112 + // The desired position is outside of the current chunk.
1.113 + // Access the new position. Assume a forward iteration from here,
1.114 + // which will also be optimimum for a single random access.
1.115 + // Reverse iterations may suffer slightly.
1.116 + ut->pFuncs->access(ut, index, TRUE);
1.117 + } else if((int32_t)(index - ut->chunkNativeStart) <= ut->nativeIndexingLimit) {
1.118 + // utf-16 indexing.
1.119 + ut->chunkOffset=(int32_t)(index-ut->chunkNativeStart);
1.120 + } else {
1.121 + ut->chunkOffset=ut->pFuncs->mapNativeIndexToUTF16(ut, index);
1.122 + }
1.123 + // The convention is that the index must always be on a code point boundary.
1.124 + // Adjust the index position if it is in the middle of a surrogate pair.
1.125 + if (ut->chunkOffset<ut->chunkLength) {
1.126 + UChar c= ut->chunkContents[ut->chunkOffset];
1.127 + if (U16_IS_TRAIL(c)) {
1.128 + if (ut->chunkOffset==0) {
1.129 + ut->pFuncs->access(ut, ut->chunkNativeStart, FALSE);
1.130 + }
1.131 + if (ut->chunkOffset>0) {
1.132 + UChar lead = ut->chunkContents[ut->chunkOffset-1];
1.133 + if (U16_IS_LEAD(lead)) {
1.134 + ut->chunkOffset--;
1.135 + }
1.136 + }
1.137 + }
1.138 + }
1.139 +}
1.140 +
1.141 +
1.142 +
1.143 +U_CAPI int64_t U_EXPORT2
1.144 +utext_getPreviousNativeIndex(UText *ut) {
1.145 + //
1.146 + // Fast-path the common case.
1.147 + // Common means current position is not at the beginning of a chunk
1.148 + // and the preceding character is not supplementary.
1.149 + //
1.150 + int32_t i = ut->chunkOffset - 1;
1.151 + int64_t result;
1.152 + if (i >= 0) {
1.153 + UChar c = ut->chunkContents[i];
1.154 + if (U16_IS_TRAIL(c) == FALSE) {
1.155 + if (i <= ut->nativeIndexingLimit) {
1.156 + result = ut->chunkNativeStart + i;
1.157 + } else {
1.158 + ut->chunkOffset = i;
1.159 + result = ut->pFuncs->mapOffsetToNative(ut);
1.160 + ut->chunkOffset++;
1.161 + }
1.162 + return result;
1.163 + }
1.164 + }
1.165 +
1.166 + // If at the start of text, simply return 0.
1.167 + if (ut->chunkOffset==0 && ut->chunkNativeStart==0) {
1.168 + return 0;
1.169 + }
1.170 +
1.171 + // Harder, less common cases. We are at a chunk boundary, or on a surrogate.
1.172 + // Keep it simple, use other functions to handle the edges.
1.173 + //
1.174 + utext_previous32(ut);
1.175 + result = UTEXT_GETNATIVEINDEX(ut);
1.176 + utext_next32(ut);
1.177 + return result;
1.178 +}
1.179 +
1.180 +
1.181 +//
1.182 +// utext_current32. Get the UChar32 at the current position.
1.183 +// UText iteration position is always on a code point boundary,
1.184 +// never on the trail half of a surrogate pair.
1.185 +//
1.186 +U_CAPI UChar32 U_EXPORT2
1.187 +utext_current32(UText *ut) {
1.188 + UChar32 c;
1.189 + if (ut->chunkOffset==ut->chunkLength) {
1.190 + // Current position is just off the end of the chunk.
1.191 + if (ut->pFuncs->access(ut, ut->chunkNativeLimit, TRUE) == FALSE) {
1.192 + // Off the end of the text.
1.193 + return U_SENTINEL;
1.194 + }
1.195 + }
1.196 +
1.197 + c = ut->chunkContents[ut->chunkOffset];
1.198 + if (U16_IS_LEAD(c) == FALSE) {
1.199 + // Normal, non-supplementary case.
1.200 + return c;
1.201 + }
1.202 +
1.203 + //
1.204 + // Possible supplementary char.
1.205 + //
1.206 + UChar32 trail = 0;
1.207 + UChar32 supplementaryC = c;
1.208 + if ((ut->chunkOffset+1) < ut->chunkLength) {
1.209 + // The trail surrogate is in the same chunk.
1.210 + trail = ut->chunkContents[ut->chunkOffset+1];
1.211 + } else {
1.212 + // The trail surrogate is in a different chunk.
1.213 + // Because we must maintain the iteration position, we need to switch forward
1.214 + // into the new chunk, get the trail surrogate, then revert the chunk back to the
1.215 + // original one.
1.216 + // An edge case to be careful of: the entire text may end with an unpaired
1.217 + // leading surrogate. The attempt to access the trail will fail, but
1.218 + // the original position before the unpaired lead still needs to be restored.
1.219 + int64_t nativePosition = ut->chunkNativeLimit;
1.220 + int32_t originalOffset = ut->chunkOffset;
1.221 + if (ut->pFuncs->access(ut, nativePosition, TRUE)) {
1.222 + trail = ut->chunkContents[ut->chunkOffset];
1.223 + }
1.224 + UBool r = ut->pFuncs->access(ut, nativePosition, FALSE); // reverse iteration flag loads preceding chunk
1.225 + U_ASSERT(r==TRUE);
1.226 + ut->chunkOffset = originalOffset;
1.227 + if(!r) {
1.228 + return U_SENTINEL;
1.229 + }
1.230 + }
1.231 +
1.232 + if (U16_IS_TRAIL(trail)) {
1.233 + supplementaryC = U16_GET_SUPPLEMENTARY(c, trail);
1.234 + }
1.235 + return supplementaryC;
1.236 +
1.237 +}
1.238 +
1.239 +
1.240 +U_CAPI UChar32 U_EXPORT2
1.241 +utext_char32At(UText *ut, int64_t nativeIndex) {
1.242 + UChar32 c = U_SENTINEL;
1.243 +
1.244 + // Fast path the common case.
1.245 + if (nativeIndex>=ut->chunkNativeStart && nativeIndex < ut->chunkNativeStart + ut->nativeIndexingLimit) {
1.246 + ut->chunkOffset = (int32_t)(nativeIndex - ut->chunkNativeStart);
1.247 + c = ut->chunkContents[ut->chunkOffset];
1.248 + if (U16_IS_SURROGATE(c) == FALSE) {
1.249 + return c;
1.250 + }
1.251 + }
1.252 +
1.253 +
1.254 + utext_setNativeIndex(ut, nativeIndex);
1.255 + if (nativeIndex>=ut->chunkNativeStart && ut->chunkOffset<ut->chunkLength) {
1.256 + c = ut->chunkContents[ut->chunkOffset];
1.257 + if (U16_IS_SURROGATE(c)) {
1.258 + // For surrogates, let current32() deal with the complications
1.259 + // of supplementaries that may span chunk boundaries.
1.260 + c = utext_current32(ut);
1.261 + }
1.262 + }
1.263 + return c;
1.264 +}
1.265 +
1.266 +
1.267 +U_CAPI UChar32 U_EXPORT2
1.268 +utext_next32(UText *ut) {
1.269 + UChar32 c;
1.270 +
1.271 + if (ut->chunkOffset >= ut->chunkLength) {
1.272 + if (ut->pFuncs->access(ut, ut->chunkNativeLimit, TRUE) == FALSE) {
1.273 + return U_SENTINEL;
1.274 + }
1.275 + }
1.276 +
1.277 + c = ut->chunkContents[ut->chunkOffset++];
1.278 + if (U16_IS_LEAD(c) == FALSE) {
1.279 + // Normal case, not supplementary.
1.280 + // (A trail surrogate seen here is just returned as is, as a surrogate value.
1.281 + // It cannot be part of a pair.)
1.282 + return c;
1.283 + }
1.284 +
1.285 + if (ut->chunkOffset >= ut->chunkLength) {
1.286 + if (ut->pFuncs->access(ut, ut->chunkNativeLimit, TRUE) == FALSE) {
1.287 + // c is an unpaired lead surrogate at the end of the text.
1.288 + // return it as it is.
1.289 + return c;
1.290 + }
1.291 + }
1.292 + UChar32 trail = ut->chunkContents[ut->chunkOffset];
1.293 + if (U16_IS_TRAIL(trail) == FALSE) {
1.294 + // c was an unpaired lead surrogate, not at the end of the text.
1.295 + // return it as it is (unpaired). Iteration position is on the
1.296 + // following character, possibly in the next chunk, where the
1.297 + // trail surrogate would have been if it had existed.
1.298 + return c;
1.299 + }
1.300 +
1.301 + UChar32 supplementary = U16_GET_SUPPLEMENTARY(c, trail);
1.302 + ut->chunkOffset++; // move iteration position over the trail surrogate.
1.303 + return supplementary;
1.304 + }
1.305 +
1.306 +
1.307 +U_CAPI UChar32 U_EXPORT2
1.308 +utext_previous32(UText *ut) {
1.309 + UChar32 c;
1.310 +
1.311 + if (ut->chunkOffset <= 0) {
1.312 + if (ut->pFuncs->access(ut, ut->chunkNativeStart, FALSE) == FALSE) {
1.313 + return U_SENTINEL;
1.314 + }
1.315 + }
1.316 + ut->chunkOffset--;
1.317 + c = ut->chunkContents[ut->chunkOffset];
1.318 + if (U16_IS_TRAIL(c) == FALSE) {
1.319 + // Normal case, not supplementary.
1.320 + // (A lead surrogate seen here is just returned as is, as a surrogate value.
1.321 + // It cannot be part of a pair.)
1.322 + return c;
1.323 + }
1.324 +
1.325 + if (ut->chunkOffset <= 0) {
1.326 + if (ut->pFuncs->access(ut, ut->chunkNativeStart, FALSE) == FALSE) {
1.327 + // c is an unpaired trail surrogate at the start of the text.
1.328 + // return it as it is.
1.329 + return c;
1.330 + }
1.331 + }
1.332 +
1.333 + UChar32 lead = ut->chunkContents[ut->chunkOffset-1];
1.334 + if (U16_IS_LEAD(lead) == FALSE) {
1.335 + // c was an unpaired trail surrogate, not at the end of the text.
1.336 + // return it as it is (unpaired). Iteration position is at c
1.337 + return c;
1.338 + }
1.339 +
1.340 + UChar32 supplementary = U16_GET_SUPPLEMENTARY(lead, c);
1.341 + ut->chunkOffset--; // move iteration position over the lead surrogate.
1.342 + return supplementary;
1.343 +}
1.344 +
1.345 +
1.346 +
1.347 +U_CAPI UChar32 U_EXPORT2
1.348 +utext_next32From(UText *ut, int64_t index) {
1.349 + UChar32 c = U_SENTINEL;
1.350 +
1.351 + if(index<ut->chunkNativeStart || index>=ut->chunkNativeLimit) {
1.352 + // Desired position is outside of the current chunk.
1.353 + if(!ut->pFuncs->access(ut, index, TRUE)) {
1.354 + // no chunk available here
1.355 + return U_SENTINEL;
1.356 + }
1.357 + } else if (index - ut->chunkNativeStart <= (int64_t)ut->nativeIndexingLimit) {
1.358 + // Desired position is in chunk, with direct 1:1 native to UTF16 indexing
1.359 + ut->chunkOffset = (int32_t)(index - ut->chunkNativeStart);
1.360 + } else {
1.361 + // Desired position is in chunk, with non-UTF16 indexing.
1.362 + ut->chunkOffset = ut->pFuncs->mapNativeIndexToUTF16(ut, index);
1.363 + }
1.364 +
1.365 + c = ut->chunkContents[ut->chunkOffset++];
1.366 + if (U16_IS_SURROGATE(c)) {
1.367 + // Surrogates. Many edge cases. Use other functions that already
1.368 + // deal with the problems.
1.369 + utext_setNativeIndex(ut, index);
1.370 + c = utext_next32(ut);
1.371 + }
1.372 + return c;
1.373 +}
1.374 +
1.375 +
1.376 +U_CAPI UChar32 U_EXPORT2
1.377 +utext_previous32From(UText *ut, int64_t index) {
1.378 + //
1.379 + // Return the character preceding the specified index.
1.380 + // Leave the iteration position at the start of the character that was returned.
1.381 + //
1.382 + UChar32 cPrev; // The character preceding cCurr, which is what we will return.
1.383 +
1.384 + // Address the chunk containg the position preceding the incoming index
1.385 + // A tricky edge case:
1.386 + // We try to test the requested native index against the chunkNativeStart to determine
1.387 + // whether the character preceding the one at the index is in the current chunk.
1.388 + // BUT, this test can fail with UTF-8 (or any other multibyte encoding), when the
1.389 + // requested index is on something other than the first position of the first char.
1.390 + //
1.391 + if(index<=ut->chunkNativeStart || index>ut->chunkNativeLimit) {
1.392 + // Requested native index is outside of the current chunk.
1.393 + if(!ut->pFuncs->access(ut, index, FALSE)) {
1.394 + // no chunk available here
1.395 + return U_SENTINEL;
1.396 + }
1.397 + } else if(index - ut->chunkNativeStart <= (int64_t)ut->nativeIndexingLimit) {
1.398 + // Direct UTF-16 indexing.
1.399 + ut->chunkOffset = (int32_t)(index - ut->chunkNativeStart);
1.400 + } else {
1.401 + ut->chunkOffset=ut->pFuncs->mapNativeIndexToUTF16(ut, index);
1.402 + if (ut->chunkOffset==0 && !ut->pFuncs->access(ut, index, FALSE)) {
1.403 + // no chunk available here
1.404 + return U_SENTINEL;
1.405 + }
1.406 + }
1.407 +
1.408 + //
1.409 + // Simple case with no surrogates.
1.410 + //
1.411 + ut->chunkOffset--;
1.412 + cPrev = ut->chunkContents[ut->chunkOffset];
1.413 +
1.414 + if (U16_IS_SURROGATE(cPrev)) {
1.415 + // Possible supplementary. Many edge cases.
1.416 + // Let other functions do the heavy lifting.
1.417 + utext_setNativeIndex(ut, index);
1.418 + cPrev = utext_previous32(ut);
1.419 + }
1.420 + return cPrev;
1.421 +}
1.422 +
1.423 +
1.424 +U_CAPI int32_t U_EXPORT2
1.425 +utext_extract(UText *ut,
1.426 + int64_t start, int64_t limit,
1.427 + UChar *dest, int32_t destCapacity,
1.428 + UErrorCode *status) {
1.429 + return ut->pFuncs->extract(ut, start, limit, dest, destCapacity, status);
1.430 + }
1.431 +
1.432 +
1.433 +
1.434 +U_CAPI UBool U_EXPORT2
1.435 +utext_equals(const UText *a, const UText *b) {
1.436 + if (a==NULL || b==NULL ||
1.437 + a->magic != UTEXT_MAGIC ||
1.438 + b->magic != UTEXT_MAGIC) {
1.439 + // Null or invalid arguments don't compare equal to anything.
1.440 + return FALSE;
1.441 + }
1.442 +
1.443 + if (a->pFuncs != b->pFuncs) {
1.444 + // Different types of text providers.
1.445 + return FALSE;
1.446 + }
1.447 +
1.448 + if (a->context != b->context) {
1.449 + // Different sources (different strings)
1.450 + return FALSE;
1.451 + }
1.452 + if (utext_getNativeIndex(a) != utext_getNativeIndex(b)) {
1.453 + // Different current position in the string.
1.454 + return FALSE;
1.455 + }
1.456 +
1.457 + return TRUE;
1.458 +}
1.459 +
1.460 +U_CAPI UBool U_EXPORT2
1.461 +utext_isWritable(const UText *ut)
1.462 +{
1.463 + UBool b = (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_WRITABLE)) != 0;
1.464 + return b;
1.465 +}
1.466 +
1.467 +
1.468 +U_CAPI void U_EXPORT2
1.469 +utext_freeze(UText *ut) {
1.470 + // Zero out the WRITABLE flag.
1.471 + ut->providerProperties &= ~(I32_FLAG(UTEXT_PROVIDER_WRITABLE));
1.472 +}
1.473 +
1.474 +
1.475 +U_CAPI UBool U_EXPORT2
1.476 +utext_hasMetaData(const UText *ut)
1.477 +{
1.478 + UBool b = (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_HAS_META_DATA)) != 0;
1.479 + return b;
1.480 +}
1.481 +
1.482 +
1.483 +
1.484 +U_CAPI int32_t U_EXPORT2
1.485 +utext_replace(UText *ut,
1.486 + int64_t nativeStart, int64_t nativeLimit,
1.487 + const UChar *replacementText, int32_t replacementLength,
1.488 + UErrorCode *status)
1.489 +{
1.490 + if (U_FAILURE(*status)) {
1.491 + return 0;
1.492 + }
1.493 + if ((ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_WRITABLE)) == 0) {
1.494 + *status = U_NO_WRITE_PERMISSION;
1.495 + return 0;
1.496 + }
1.497 + int32_t i = ut->pFuncs->replace(ut, nativeStart, nativeLimit, replacementText, replacementLength, status);
1.498 + return i;
1.499 +}
1.500 +
1.501 +U_CAPI void U_EXPORT2
1.502 +utext_copy(UText *ut,
1.503 + int64_t nativeStart, int64_t nativeLimit,
1.504 + int64_t destIndex,
1.505 + UBool move,
1.506 + UErrorCode *status)
1.507 +{
1.508 + if (U_FAILURE(*status)) {
1.509 + return;
1.510 + }
1.511 + if ((ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_WRITABLE)) == 0) {
1.512 + *status = U_NO_WRITE_PERMISSION;
1.513 + return;
1.514 + }
1.515 + ut->pFuncs->copy(ut, nativeStart, nativeLimit, destIndex, move, status);
1.516 +}
1.517 +
1.518 +
1.519 +
1.520 +U_CAPI UText * U_EXPORT2
1.521 +utext_clone(UText *dest, const UText *src, UBool deep, UBool readOnly, UErrorCode *status) {
1.522 + UText *result;
1.523 + result = src->pFuncs->clone(dest, src, deep, status);
1.524 + if (readOnly) {
1.525 + utext_freeze(result);
1.526 + }
1.527 + return result;
1.528 +}
1.529 +
1.530 +
1.531 +
1.532 +//------------------------------------------------------------------------------
1.533 +//
1.534 +// UText common functions implementation
1.535 +//
1.536 +//------------------------------------------------------------------------------
1.537 +
1.538 +//
1.539 +// UText.flags bit definitions
1.540 +//
1.541 +enum {
1.542 + UTEXT_HEAP_ALLOCATED = 1, // 1 if ICU has allocated this UText struct on the heap.
1.543 + // 0 if caller provided storage for the UText.
1.544 +
1.545 + UTEXT_EXTRA_HEAP_ALLOCATED = 2, // 1 if ICU has allocated extra storage as a separate
1.546 + // heap block.
1.547 + // 0 if there is no separate allocation. Either no extra
1.548 + // storage was requested, or it is appended to the end
1.549 + // of the main UText storage.
1.550 +
1.551 + UTEXT_OPEN = 4 // 1 if this UText is currently open
1.552 + // 0 if this UText is not open.
1.553 +};
1.554 +
1.555 +
1.556 +//
1.557 +// Extended form of a UText. The purpose is to aid in computing the total size required
1.558 +// when a provider asks for a UText to be allocated with extra storage.
1.559 +
1.560 +struct ExtendedUText {
1.561 + UText ut;
1.562 + UAlignedMemory extension;
1.563 +};
1.564 +
1.565 +static const UText emptyText = UTEXT_INITIALIZER;
1.566 +
1.567 +U_CAPI UText * U_EXPORT2
1.568 +utext_setup(UText *ut, int32_t extraSpace, UErrorCode *status) {
1.569 + if (U_FAILURE(*status)) {
1.570 + return ut;
1.571 + }
1.572 +
1.573 + if (ut == NULL) {
1.574 + // We need to heap-allocate storage for the new UText
1.575 + int32_t spaceRequired = sizeof(UText);
1.576 + if (extraSpace > 0) {
1.577 + spaceRequired = sizeof(ExtendedUText) + extraSpace - sizeof(UAlignedMemory);
1.578 + }
1.579 + ut = (UText *)uprv_malloc(spaceRequired);
1.580 + if (ut == NULL) {
1.581 + *status = U_MEMORY_ALLOCATION_ERROR;
1.582 + return NULL;
1.583 + } else {
1.584 + *ut = emptyText;
1.585 + ut->flags |= UTEXT_HEAP_ALLOCATED;
1.586 + if (spaceRequired>0) {
1.587 + ut->extraSize = extraSpace;
1.588 + ut->pExtra = &((ExtendedUText *)ut)->extension;
1.589 + }
1.590 + }
1.591 + } else {
1.592 + // We have been supplied with an already existing UText.
1.593 + // Verify that it really appears to be a UText.
1.594 + if (ut->magic != UTEXT_MAGIC) {
1.595 + *status = U_ILLEGAL_ARGUMENT_ERROR;
1.596 + return ut;
1.597 + }
1.598 + // If the ut is already open and there's a provider supplied close
1.599 + // function, call it.
1.600 + if ((ut->flags & UTEXT_OPEN) && ut->pFuncs->close != NULL) {
1.601 + ut->pFuncs->close(ut);
1.602 + }
1.603 + ut->flags &= ~UTEXT_OPEN;
1.604 +
1.605 + // If extra space was requested by our caller, check whether
1.606 + // sufficient already exists, and allocate new if needed.
1.607 + if (extraSpace > ut->extraSize) {
1.608 + // Need more space. If there is existing separately allocated space,
1.609 + // delete it first, then allocate new space.
1.610 + if (ut->flags & UTEXT_EXTRA_HEAP_ALLOCATED) {
1.611 + uprv_free(ut->pExtra);
1.612 + ut->extraSize = 0;
1.613 + }
1.614 + ut->pExtra = uprv_malloc(extraSpace);
1.615 + if (ut->pExtra == NULL) {
1.616 + *status = U_MEMORY_ALLOCATION_ERROR;
1.617 + } else {
1.618 + ut->extraSize = extraSpace;
1.619 + ut->flags |= UTEXT_EXTRA_HEAP_ALLOCATED;
1.620 + }
1.621 + }
1.622 + }
1.623 + if (U_SUCCESS(*status)) {
1.624 + ut->flags |= UTEXT_OPEN;
1.625 +
1.626 + // Initialize all remaining fields of the UText.
1.627 + //
1.628 + ut->context = NULL;
1.629 + ut->chunkContents = NULL;
1.630 + ut->p = NULL;
1.631 + ut->q = NULL;
1.632 + ut->r = NULL;
1.633 + ut->a = 0;
1.634 + ut->b = 0;
1.635 + ut->c = 0;
1.636 + ut->chunkOffset = 0;
1.637 + ut->chunkLength = 0;
1.638 + ut->chunkNativeStart = 0;
1.639 + ut->chunkNativeLimit = 0;
1.640 + ut->nativeIndexingLimit = 0;
1.641 + ut->providerProperties = 0;
1.642 + ut->privA = 0;
1.643 + ut->privB = 0;
1.644 + ut->privC = 0;
1.645 + ut->privP = NULL;
1.646 + if (ut->pExtra!=NULL && ut->extraSize>0)
1.647 + uprv_memset(ut->pExtra, 0, ut->extraSize);
1.648 +
1.649 + }
1.650 + return ut;
1.651 +}
1.652 +
1.653 +
1.654 +U_CAPI UText * U_EXPORT2
1.655 +utext_close(UText *ut) {
1.656 + if (ut==NULL ||
1.657 + ut->magic != UTEXT_MAGIC ||
1.658 + (ut->flags & UTEXT_OPEN) == 0)
1.659 + {
1.660 + // The supplied ut is not an open UText.
1.661 + // Do nothing.
1.662 + return ut;
1.663 + }
1.664 +
1.665 + // If the provider gave us a close function, call it now.
1.666 + // This will clean up anything allocated specifically by the provider.
1.667 + if (ut->pFuncs->close != NULL) {
1.668 + ut->pFuncs->close(ut);
1.669 + }
1.670 + ut->flags &= ~UTEXT_OPEN;
1.671 +
1.672 + // If we (the framework) allocated the UText or subsidiary storage,
1.673 + // delete it.
1.674 + if (ut->flags & UTEXT_EXTRA_HEAP_ALLOCATED) {
1.675 + uprv_free(ut->pExtra);
1.676 + ut->pExtra = NULL;
1.677 + ut->flags &= ~UTEXT_EXTRA_HEAP_ALLOCATED;
1.678 + ut->extraSize = 0;
1.679 + }
1.680 +
1.681 + // Zero out function table of the closed UText. This is a defensive move,
1.682 + // inteded to cause applications that inadvertantly use a closed
1.683 + // utext to crash with null pointer errors.
1.684 + ut->pFuncs = NULL;
1.685 +
1.686 + if (ut->flags & UTEXT_HEAP_ALLOCATED) {
1.687 + // This UText was allocated by UText setup. We need to free it.
1.688 + // Clear magic, so we can detect if the user messes up and immediately
1.689 + // tries to reopen another UText using the deleted storage.
1.690 + ut->magic = 0;
1.691 + uprv_free(ut);
1.692 + ut = NULL;
1.693 + }
1.694 + return ut;
1.695 +}
1.696 +
1.697 +
1.698 +
1.699 +
1.700 +//
1.701 +// invalidateChunk Reset a chunk to have no contents, so that the next call
1.702 +// to access will cause new data to load.
1.703 +// This is needed when copy/move/replace operate directly on the
1.704 +// backing text, potentially putting it out of sync with the
1.705 +// contents in the chunk.
1.706 +//
1.707 +static void
1.708 +invalidateChunk(UText *ut) {
1.709 + ut->chunkLength = 0;
1.710 + ut->chunkNativeLimit = 0;
1.711 + ut->chunkNativeStart = 0;
1.712 + ut->chunkOffset = 0;
1.713 + ut->nativeIndexingLimit = 0;
1.714 +}
1.715 +
1.716 +//
1.717 +// pinIndex Do range pinning on a native index parameter.
1.718 +// 64 bit pinning is done in place.
1.719 +// 32 bit truncated result is returned as a convenience for
1.720 +// use in providers that don't need 64 bits.
1.721 +static int32_t
1.722 +pinIndex(int64_t &index, int64_t limit) {
1.723 + if (index<0) {
1.724 + index = 0;
1.725 + } else if (index > limit) {
1.726 + index = limit;
1.727 + }
1.728 + return (int32_t)index;
1.729 +}
1.730 +
1.731 +
1.732 +U_CDECL_BEGIN
1.733 +
1.734 +//
1.735 +// Pointer relocation function,
1.736 +// a utility used by shallow clone.
1.737 +// Adjust a pointer that refers to something within one UText (the source)
1.738 +// to refer to the same relative offset within a another UText (the target)
1.739 +//
1.740 +static void adjustPointer(UText *dest, const void **destPtr, const UText *src) {
1.741 + // convert all pointers to (char *) so that byte address arithmetic will work.
1.742 + char *dptr = (char *)*destPtr;
1.743 + char *dUText = (char *)dest;
1.744 + char *sUText = (char *)src;
1.745 +
1.746 + if (dptr >= (char *)src->pExtra && dptr < ((char*)src->pExtra)+src->extraSize) {
1.747 + // target ptr was to something within the src UText's pExtra storage.
1.748 + // relocate it into the target UText's pExtra region.
1.749 + *destPtr = ((char *)dest->pExtra) + (dptr - (char *)src->pExtra);
1.750 + } else if (dptr>=sUText && dptr < sUText+src->sizeOfStruct) {
1.751 + // target ptr was pointing to somewhere within the source UText itself.
1.752 + // Move it to the same offset within the target UText.
1.753 + *destPtr = dUText + (dptr-sUText);
1.754 + }
1.755 +}
1.756 +
1.757 +
1.758 +//
1.759 +// Clone. This is a generic copy-the-utext-by-value clone function that can be
1.760 +// used as-is with some utext types, and as a helper by other clones.
1.761 +//
1.762 +static UText * U_CALLCONV
1.763 +shallowTextClone(UText * dest, const UText * src, UErrorCode * status) {
1.764 + if (U_FAILURE(*status)) {
1.765 + return NULL;
1.766 + }
1.767 + int32_t srcExtraSize = src->extraSize;
1.768 +
1.769 + //
1.770 + // Use the generic text_setup to allocate storage if required.
1.771 + //
1.772 + dest = utext_setup(dest, srcExtraSize, status);
1.773 + if (U_FAILURE(*status)) {
1.774 + return dest;
1.775 + }
1.776 +
1.777 + //
1.778 + // flags (how the UText was allocated) and the pointer to the
1.779 + // extra storage must retain the values in the cloned utext that
1.780 + // were set up by utext_setup. Save them separately before
1.781 + // copying the whole struct.
1.782 + //
1.783 + void *destExtra = dest->pExtra;
1.784 + int32_t flags = dest->flags;
1.785 +
1.786 +
1.787 + //
1.788 + // Copy the whole UText struct by value.
1.789 + // Any "Extra" storage is copied also.
1.790 + //
1.791 + int sizeToCopy = src->sizeOfStruct;
1.792 + if (sizeToCopy > dest->sizeOfStruct) {
1.793 + sizeToCopy = dest->sizeOfStruct;
1.794 + }
1.795 + uprv_memcpy(dest, src, sizeToCopy);
1.796 + dest->pExtra = destExtra;
1.797 + dest->flags = flags;
1.798 + if (srcExtraSize > 0) {
1.799 + uprv_memcpy(dest->pExtra, src->pExtra, srcExtraSize);
1.800 + }
1.801 +
1.802 + //
1.803 + // Relocate any pointers in the target that refer to the UText itself
1.804 + // to point to the cloned copy rather than the original source.
1.805 + //
1.806 + adjustPointer(dest, &dest->context, src);
1.807 + adjustPointer(dest, &dest->p, src);
1.808 + adjustPointer(dest, &dest->q, src);
1.809 + adjustPointer(dest, &dest->r, src);
1.810 + adjustPointer(dest, (const void **)&dest->chunkContents, src);
1.811 +
1.812 + return dest;
1.813 +}
1.814 +
1.815 +
1.816 +U_CDECL_END
1.817 +
1.818 +
1.819 +
1.820 +//------------------------------------------------------------------------------
1.821 +//
1.822 +// UText implementation for UTF-8 char * strings (read-only)
1.823 +// Limitation: string length must be <= 0x7fffffff in length.
1.824 +// (length must for in an int32_t variable)
1.825 +//
1.826 +// Use of UText data members:
1.827 +// context pointer to UTF-8 string
1.828 +// utext.b is the input string length (bytes).
1.829 +// utext.c Length scanned so far in string
1.830 +// (for optimizing finding length of zero terminated strings.)
1.831 +// utext.p pointer to the current buffer
1.832 +// utext.q pointer to the other buffer.
1.833 +//
1.834 +//------------------------------------------------------------------------------
1.835 +
1.836 +// Chunk size.
1.837 +// Must be less than 85, because of byte mapping from UChar indexes to native indexes.
1.838 +// Worst case is three native bytes to one UChar. (Supplemenaries are 4 native bytes
1.839 +// to two UChars.)
1.840 +//
1.841 +enum { UTF8_TEXT_CHUNK_SIZE=32 };
1.842 +
1.843 +//
1.844 +// UTF8Buf Two of these structs will be set up in the UText's extra allocated space.
1.845 +// Each contains the UChar chunk buffer, the to and from native maps, and
1.846 +// header info.
1.847 +//
1.848 +// because backwards iteration fills the buffers starting at the end and
1.849 +// working towards the front, the filled part of the buffers may not begin
1.850 +// at the start of the available storage for the buffers.
1.851 +//
1.852 +// Buffer size is one bigger than the specified UTF8_TEXT_CHUNK_SIZE to allow for
1.853 +// the last character added being a supplementary, and thus requiring a surrogate
1.854 +// pair. Doing this is simpler than checking for the edge case.
1.855 +//
1.856 +
1.857 +struct UTF8Buf {
1.858 + int32_t bufNativeStart; // Native index of first char in UChar buf
1.859 + int32_t bufNativeLimit; // Native index following last char in buf.
1.860 + int32_t bufStartIdx; // First filled position in buf.
1.861 + int32_t bufLimitIdx; // Limit of filled range in buf.
1.862 + int32_t bufNILimit; // Limit of native indexing part of buf
1.863 + int32_t toUCharsMapStart; // Native index corresponding to
1.864 + // mapToUChars[0].
1.865 + // Set to bufNativeStart when filling forwards.
1.866 + // Set to computed value when filling backwards.
1.867 +
1.868 + UChar buf[UTF8_TEXT_CHUNK_SIZE+4]; // The UChar buffer. Requires one extra position beyond the
1.869 + // the chunk size, to allow for surrogate at the end.
1.870 + // Length must be identical to mapToNative array, below,
1.871 + // because of the way indexing works when the array is
1.872 + // filled backwards during a reverse iteration. Thus,
1.873 + // the additional extra size.
1.874 + uint8_t mapToNative[UTF8_TEXT_CHUNK_SIZE+4]; // map UChar index in buf to
1.875 + // native offset from bufNativeStart.
1.876 + // Requires two extra slots,
1.877 + // one for a supplementary starting in the last normal position,
1.878 + // and one for an entry for the buffer limit position.
1.879 + uint8_t mapToUChars[UTF8_TEXT_CHUNK_SIZE*3+6]; // Map native offset from bufNativeStart to
1.880 + // correspoding offset in filled part of buf.
1.881 + int32_t align;
1.882 +};
1.883 +
1.884 +U_CDECL_BEGIN
1.885 +
1.886 +//
1.887 +// utf8TextLength
1.888 +//
1.889 +// Get the length of the string. If we don't already know it,
1.890 +// we'll need to scan for the trailing nul.
1.891 +//
1.892 +static int64_t U_CALLCONV
1.893 +utf8TextLength(UText *ut) {
1.894 + if (ut->b < 0) {
1.895 + // Zero terminated string, and we haven't scanned to the end yet.
1.896 + // Scan it now.
1.897 + const char *r = (const char *)ut->context + ut->c;
1.898 + while (*r != 0) {
1.899 + r++;
1.900 + }
1.901 + if ((r - (const char *)ut->context) < 0x7fffffff) {
1.902 + ut->b = (int32_t)(r - (const char *)ut->context);
1.903 + } else {
1.904 + // Actual string was bigger (more than 2 gig) than we
1.905 + // can handle. Clip it to 2 GB.
1.906 + ut->b = 0x7fffffff;
1.907 + }
1.908 + ut->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
1.909 + }
1.910 + return ut->b;
1.911 +}
1.912 +
1.913 +
1.914 +
1.915 +
1.916 +
1.917 +
1.918 +static UBool U_CALLCONV
1.919 +utf8TextAccess(UText *ut, int64_t index, UBool forward) {
1.920 + //
1.921 + // Apologies to those who are allergic to goto statements.
1.922 + // Consider each goto to a labelled block to be the equivalent of
1.923 + // call the named block as if it were a function();
1.924 + // return;
1.925 + //
1.926 + const uint8_t *s8=(const uint8_t *)ut->context;
1.927 + UTF8Buf *u8b = NULL;
1.928 + int32_t length = ut->b; // Length of original utf-8
1.929 + int32_t ix= (int32_t)index; // Requested index, trimmed to 32 bits.
1.930 + int32_t mapIndex = 0;
1.931 + if (index<0) {
1.932 + ix=0;
1.933 + } else if (index > 0x7fffffff) {
1.934 + // Strings with 64 bit lengths not supported by this UTF-8 provider.
1.935 + ix = 0x7fffffff;
1.936 + }
1.937 +
1.938 + // Pin requested index to the string length.
1.939 + if (ix>length) {
1.940 + if (length>=0) {
1.941 + ix=length;
1.942 + } else if (ix>=ut->c) {
1.943 + // Zero terminated string, and requested index is beyond
1.944 + // the region that has already been scanned.
1.945 + // Scan up to either the end of the string or to the
1.946 + // requested position, whichever comes first.
1.947 + while (ut->c<ix && s8[ut->c]!=0) {
1.948 + ut->c++;
1.949 + }
1.950 + // TODO: support for null terminated string length > 32 bits.
1.951 + if (s8[ut->c] == 0) {
1.952 + // We just found the actual length of the string.
1.953 + // Trim the requested index back to that.
1.954 + ix = ut->c;
1.955 + ut->b = ut->c;
1.956 + length = ut->c;
1.957 + ut->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
1.958 + }
1.959 + }
1.960 + }
1.961 +
1.962 + //
1.963 + // Dispatch to the appropriate action for a forward iteration request.
1.964 + //
1.965 + if (forward) {
1.966 + if (ix==ut->chunkNativeLimit) {
1.967 + // Check for normal sequential iteration cases first.
1.968 + if (ix==length) {
1.969 + // Just reached end of string
1.970 + // Don't swap buffers, but do set the
1.971 + // current buffer position.
1.972 + ut->chunkOffset = ut->chunkLength;
1.973 + return FALSE;
1.974 + } else {
1.975 + // End of current buffer.
1.976 + // check whether other buffer already has what we need.
1.977 + UTF8Buf *altB = (UTF8Buf *)ut->q;
1.978 + if (ix>=altB->bufNativeStart && ix<altB->bufNativeLimit) {
1.979 + goto swapBuffers;
1.980 + }
1.981 + }
1.982 + }
1.983 +
1.984 + // A random access. Desired index could be in either or niether buf.
1.985 + // For optimizing the order of testing, first check for the index
1.986 + // being in the other buffer. This will be the case for uses that
1.987 + // move back and forth over a fairly limited range
1.988 + {
1.989 + u8b = (UTF8Buf *)ut->q; // the alternate buffer
1.990 + if (ix>=u8b->bufNativeStart && ix<u8b->bufNativeLimit) {
1.991 + // Requested index is in the other buffer.
1.992 + goto swapBuffers;
1.993 + }
1.994 + if (ix == length) {
1.995 + // Requested index is end-of-string.
1.996 + // (this is the case of randomly seeking to the end.
1.997 + // The case of iterating off the end is handled earlier.)
1.998 + if (ix == ut->chunkNativeLimit) {
1.999 + // Current buffer extends up to the end of the string.
1.1000 + // Leave it as the current buffer.
1.1001 + ut->chunkOffset = ut->chunkLength;
1.1002 + return FALSE;
1.1003 + }
1.1004 + if (ix == u8b->bufNativeLimit) {
1.1005 + // Alternate buffer extends to the end of string.
1.1006 + // Swap it in as the current buffer.
1.1007 + goto swapBuffersAndFail;
1.1008 + }
1.1009 +
1.1010 + // Neither existing buffer extends to the end of the string.
1.1011 + goto makeStubBuffer;
1.1012 + }
1.1013 +
1.1014 + if (ix<ut->chunkNativeStart || ix>=ut->chunkNativeLimit) {
1.1015 + // Requested index is in neither buffer.
1.1016 + goto fillForward;
1.1017 + }
1.1018 +
1.1019 + // Requested index is in this buffer.
1.1020 + u8b = (UTF8Buf *)ut->p; // the current buffer
1.1021 + mapIndex = ix - u8b->toUCharsMapStart;
1.1022 + ut->chunkOffset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;
1.1023 + return TRUE;
1.1024 +
1.1025 + }
1.1026 + }
1.1027 +
1.1028 +
1.1029 + //
1.1030 + // Dispatch to the appropriate action for a
1.1031 + // Backwards Diretion iteration request.
1.1032 + //
1.1033 + if (ix==ut->chunkNativeStart) {
1.1034 + // Check for normal sequential iteration cases first.
1.1035 + if (ix==0) {
1.1036 + // Just reached the start of string
1.1037 + // Don't swap buffers, but do set the
1.1038 + // current buffer position.
1.1039 + ut->chunkOffset = 0;
1.1040 + return FALSE;
1.1041 + } else {
1.1042 + // Start of current buffer.
1.1043 + // check whether other buffer already has what we need.
1.1044 + UTF8Buf *altB = (UTF8Buf *)ut->q;
1.1045 + if (ix>altB->bufNativeStart && ix<=altB->bufNativeLimit) {
1.1046 + goto swapBuffers;
1.1047 + }
1.1048 + }
1.1049 + }
1.1050 +
1.1051 + // A random access. Desired index could be in either or niether buf.
1.1052 + // For optimizing the order of testing,
1.1053 + // Most likely case: in the other buffer.
1.1054 + // Second most likely: in neither buffer.
1.1055 + // Unlikely, but must work: in the current buffer.
1.1056 + u8b = (UTF8Buf *)ut->q; // the alternate buffer
1.1057 + if (ix>u8b->bufNativeStart && ix<=u8b->bufNativeLimit) {
1.1058 + // Requested index is in the other buffer.
1.1059 + goto swapBuffers;
1.1060 + }
1.1061 + // Requested index is start-of-string.
1.1062 + // (this is the case of randomly seeking to the start.
1.1063 + // The case of iterating off the start is handled earlier.)
1.1064 + if (ix==0) {
1.1065 + if (u8b->bufNativeStart==0) {
1.1066 + // Alternate buffer contains the data for the start string.
1.1067 + // Make it be the current buffer.
1.1068 + goto swapBuffersAndFail;
1.1069 + } else {
1.1070 + // Request for data before the start of string,
1.1071 + // neither buffer is usable.
1.1072 + // set up a zero-length buffer.
1.1073 + goto makeStubBuffer;
1.1074 + }
1.1075 + }
1.1076 +
1.1077 + if (ix<=ut->chunkNativeStart || ix>ut->chunkNativeLimit) {
1.1078 + // Requested index is in neither buffer.
1.1079 + goto fillReverse;
1.1080 + }
1.1081 +
1.1082 + // Requested index is in this buffer.
1.1083 + // Set the utf16 buffer index.
1.1084 + u8b = (UTF8Buf *)ut->p;
1.1085 + mapIndex = ix - u8b->toUCharsMapStart;
1.1086 + ut->chunkOffset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;
1.1087 + if (ut->chunkOffset==0) {
1.1088 + // This occurs when the first character in the text is
1.1089 + // a multi-byte UTF-8 char, and the requested index is to
1.1090 + // one of the trailing bytes. Because there is no preceding ,
1.1091 + // character, this access fails. We can't pick up on the
1.1092 + // situation sooner because the requested index is not zero.
1.1093 + return FALSE;
1.1094 + } else {
1.1095 + return TRUE;
1.1096 + }
1.1097 +
1.1098 +
1.1099 +
1.1100 +swapBuffers:
1.1101 + // The alternate buffer (ut->q) has the string data that was requested.
1.1102 + // Swap the primary and alternate buffers, and set the
1.1103 + // chunk index into the new primary buffer.
1.1104 + {
1.1105 + u8b = (UTF8Buf *)ut->q;
1.1106 + ut->q = ut->p;
1.1107 + ut->p = u8b;
1.1108 + ut->chunkContents = &u8b->buf[u8b->bufStartIdx];
1.1109 + ut->chunkLength = u8b->bufLimitIdx - u8b->bufStartIdx;
1.1110 + ut->chunkNativeStart = u8b->bufNativeStart;
1.1111 + ut->chunkNativeLimit = u8b->bufNativeLimit;
1.1112 + ut->nativeIndexingLimit = u8b->bufNILimit;
1.1113 +
1.1114 + // Index into the (now current) chunk
1.1115 + // Use the map to set the chunk index. It's more trouble than it's worth
1.1116 + // to check whether native indexing can be used.
1.1117 + U_ASSERT(ix>=u8b->bufNativeStart);
1.1118 + U_ASSERT(ix<=u8b->bufNativeLimit);
1.1119 + mapIndex = ix - u8b->toUCharsMapStart;
1.1120 + U_ASSERT(mapIndex>=0);
1.1121 + U_ASSERT(mapIndex<(int32_t)sizeof(u8b->mapToUChars));
1.1122 + ut->chunkOffset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;
1.1123 +
1.1124 + return TRUE;
1.1125 + }
1.1126 +
1.1127 +
1.1128 + swapBuffersAndFail:
1.1129 + // We got a request for either the start or end of the string,
1.1130 + // with iteration continuing in the out-of-bounds direction.
1.1131 + // The alternate buffer already contains the data up to the
1.1132 + // start/end.
1.1133 + // Swap the buffers, then return failure, indicating that we couldn't
1.1134 + // make things correct for continuing the iteration in the requested
1.1135 + // direction. The position & buffer are correct should the
1.1136 + // user decide to iterate in the opposite direction.
1.1137 + u8b = (UTF8Buf *)ut->q;
1.1138 + ut->q = ut->p;
1.1139 + ut->p = u8b;
1.1140 + ut->chunkContents = &u8b->buf[u8b->bufStartIdx];
1.1141 + ut->chunkLength = u8b->bufLimitIdx - u8b->bufStartIdx;
1.1142 + ut->chunkNativeStart = u8b->bufNativeStart;
1.1143 + ut->chunkNativeLimit = u8b->bufNativeLimit;
1.1144 + ut->nativeIndexingLimit = u8b->bufNILimit;
1.1145 +
1.1146 + // Index into the (now current) chunk
1.1147 + // For this function (swapBuffersAndFail), the requested index
1.1148 + // will always be at either the start or end of the chunk.
1.1149 + if (ix==u8b->bufNativeLimit) {
1.1150 + ut->chunkOffset = ut->chunkLength;
1.1151 + } else {
1.1152 + ut->chunkOffset = 0;
1.1153 + U_ASSERT(ix == u8b->bufNativeStart);
1.1154 + }
1.1155 + return FALSE;
1.1156 +
1.1157 +makeStubBuffer:
1.1158 + // The user has done a seek/access past the start or end
1.1159 + // of the string. Rather than loading data that is likely
1.1160 + // to never be used, just set up a zero-length buffer at
1.1161 + // the position.
1.1162 + u8b = (UTF8Buf *)ut->q;
1.1163 + u8b->bufNativeStart = ix;
1.1164 + u8b->bufNativeLimit = ix;
1.1165 + u8b->bufStartIdx = 0;
1.1166 + u8b->bufLimitIdx = 0;
1.1167 + u8b->bufNILimit = 0;
1.1168 + u8b->toUCharsMapStart = ix;
1.1169 + u8b->mapToNative[0] = 0;
1.1170 + u8b->mapToUChars[0] = 0;
1.1171 + goto swapBuffersAndFail;
1.1172 +
1.1173 +
1.1174 +
1.1175 +fillForward:
1.1176 + {
1.1177 + // Move the incoming index to a code point boundary.
1.1178 + U8_SET_CP_START(s8, 0, ix);
1.1179 +
1.1180 + // Swap the UText buffers.
1.1181 + // We want to fill what was previously the alternate buffer,
1.1182 + // and make what was the current buffer be the new alternate.
1.1183 + UTF8Buf *u8b = (UTF8Buf *)ut->q;
1.1184 + ut->q = ut->p;
1.1185 + ut->p = u8b;
1.1186 +
1.1187 + int32_t strLen = ut->b;
1.1188 + UBool nulTerminated = FALSE;
1.1189 + if (strLen < 0) {
1.1190 + strLen = 0x7fffffff;
1.1191 + nulTerminated = TRUE;
1.1192 + }
1.1193 +
1.1194 + UChar *buf = u8b->buf;
1.1195 + uint8_t *mapToNative = u8b->mapToNative;
1.1196 + uint8_t *mapToUChars = u8b->mapToUChars;
1.1197 + int32_t destIx = 0;
1.1198 + int32_t srcIx = ix;
1.1199 + UBool seenNonAscii = FALSE;
1.1200 + UChar32 c = 0;
1.1201 +
1.1202 + // Fill the chunk buffer and mapping arrays.
1.1203 + while (destIx<UTF8_TEXT_CHUNK_SIZE) {
1.1204 + c = s8[srcIx];
1.1205 + if (c>0 && c<0x80) {
1.1206 + // Special case ASCII range for speed.
1.1207 + // zero is excluded to simplify bounds checking.
1.1208 + buf[destIx] = (UChar)c;
1.1209 + mapToNative[destIx] = (uint8_t)(srcIx - ix);
1.1210 + mapToUChars[srcIx-ix] = (uint8_t)destIx;
1.1211 + srcIx++;
1.1212 + destIx++;
1.1213 + } else {
1.1214 + // General case, handle everything.
1.1215 + if (seenNonAscii == FALSE) {
1.1216 + seenNonAscii = TRUE;
1.1217 + u8b->bufNILimit = destIx;
1.1218 + }
1.1219 +
1.1220 + int32_t cIx = srcIx;
1.1221 + int32_t dIx = destIx;
1.1222 + int32_t dIxSaved = destIx;
1.1223 + U8_NEXT_OR_FFFD(s8, srcIx, strLen, c);
1.1224 + if (c==0 && nulTerminated) {
1.1225 + srcIx--;
1.1226 + break;
1.1227 + }
1.1228 +
1.1229 + U16_APPEND_UNSAFE(buf, destIx, c);
1.1230 + do {
1.1231 + mapToNative[dIx++] = (uint8_t)(cIx - ix);
1.1232 + } while (dIx < destIx);
1.1233 +
1.1234 + do {
1.1235 + mapToUChars[cIx++ - ix] = (uint8_t)dIxSaved;
1.1236 + } while (cIx < srcIx);
1.1237 + }
1.1238 + if (srcIx>=strLen) {
1.1239 + break;
1.1240 + }
1.1241 +
1.1242 + }
1.1243 +
1.1244 + // store Native <--> Chunk Map entries for the end of the buffer.
1.1245 + // There is no actual character here, but the index position is valid.
1.1246 + mapToNative[destIx] = (uint8_t)(srcIx - ix);
1.1247 + mapToUChars[srcIx - ix] = (uint8_t)destIx;
1.1248 +
1.1249 + // fill in Buffer descriptor
1.1250 + u8b->bufNativeStart = ix;
1.1251 + u8b->bufNativeLimit = srcIx;
1.1252 + u8b->bufStartIdx = 0;
1.1253 + u8b->bufLimitIdx = destIx;
1.1254 + if (seenNonAscii == FALSE) {
1.1255 + u8b->bufNILimit = destIx;
1.1256 + }
1.1257 + u8b->toUCharsMapStart = u8b->bufNativeStart;
1.1258 +
1.1259 + // Set UText chunk to refer to this buffer.
1.1260 + ut->chunkContents = buf;
1.1261 + ut->chunkOffset = 0;
1.1262 + ut->chunkLength = u8b->bufLimitIdx;
1.1263 + ut->chunkNativeStart = u8b->bufNativeStart;
1.1264 + ut->chunkNativeLimit = u8b->bufNativeLimit;
1.1265 + ut->nativeIndexingLimit = u8b->bufNILimit;
1.1266 +
1.1267 + // For zero terminated strings, keep track of the maximum point
1.1268 + // scanned so far.
1.1269 + if (nulTerminated && srcIx>ut->c) {
1.1270 + ut->c = srcIx;
1.1271 + if (c==0) {
1.1272 + // We scanned to the end.
1.1273 + // Remember the actual length.
1.1274 + ut->b = srcIx;
1.1275 + ut->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
1.1276 + }
1.1277 + }
1.1278 + return TRUE;
1.1279 + }
1.1280 +
1.1281 +
1.1282 +fillReverse:
1.1283 + {
1.1284 + // Move the incoming index to a code point boundary.
1.1285 + // Can only do this if the incoming index is somewhere in the interior of the string.
1.1286 + // If index is at the end, there is no character there to look at.
1.1287 + if (ix != ut->b) {
1.1288 + U8_SET_CP_START(s8, 0, ix);
1.1289 + }
1.1290 +
1.1291 + // Swap the UText buffers.
1.1292 + // We want to fill what was previously the alternate buffer,
1.1293 + // and make what was the current buffer be the new alternate.
1.1294 + UTF8Buf *u8b = (UTF8Buf *)ut->q;
1.1295 + ut->q = ut->p;
1.1296 + ut->p = u8b;
1.1297 +
1.1298 + UChar *buf = u8b->buf;
1.1299 + uint8_t *mapToNative = u8b->mapToNative;
1.1300 + uint8_t *mapToUChars = u8b->mapToUChars;
1.1301 + int32_t toUCharsMapStart = ix - (UTF8_TEXT_CHUNK_SIZE*3 + 1);
1.1302 + int32_t destIx = UTF8_TEXT_CHUNK_SIZE+2; // Start in the overflow region
1.1303 + // at end of buffer to leave room
1.1304 + // for a surrogate pair at the
1.1305 + // buffer start.
1.1306 + int32_t srcIx = ix;
1.1307 + int32_t bufNILimit = destIx;
1.1308 + UChar32 c;
1.1309 +
1.1310 + // Map to/from Native Indexes, fill in for the position at the end of
1.1311 + // the buffer.
1.1312 + //
1.1313 + mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
1.1314 + mapToUChars[srcIx - toUCharsMapStart] = (uint8_t)destIx;
1.1315 +
1.1316 + // Fill the chunk buffer
1.1317 + // Work backwards, filling from the end of the buffer towards the front.
1.1318 + //
1.1319 + while (destIx>2 && (srcIx - toUCharsMapStart > 5) && (srcIx > 0)) {
1.1320 + srcIx--;
1.1321 + destIx--;
1.1322 +
1.1323 + // Get last byte of the UTF-8 character
1.1324 + c = s8[srcIx];
1.1325 + if (c<0x80) {
1.1326 + // Special case ASCII range for speed.
1.1327 + buf[destIx] = (UChar)c;
1.1328 + mapToUChars[srcIx - toUCharsMapStart] = (uint8_t)destIx;
1.1329 + mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
1.1330 + } else {
1.1331 + // General case, handle everything non-ASCII.
1.1332 +
1.1333 + int32_t sIx = srcIx; // ix of last byte of multi-byte u8 char
1.1334 +
1.1335 + // Get the full character from the UTF8 string.
1.1336 + // use code derived from tbe macros in utf8.h
1.1337 + // Leaves srcIx pointing at the first byte of the UTF-8 char.
1.1338 + //
1.1339 + c=utf8_prevCharSafeBody(s8, 0, &srcIx, c, -3);
1.1340 + // leaves srcIx at first byte of the multi-byte char.
1.1341 +
1.1342 + // Store the character in UTF-16 buffer.
1.1343 + if (c<0x10000) {
1.1344 + buf[destIx] = (UChar)c;
1.1345 + mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
1.1346 + } else {
1.1347 + buf[destIx] = U16_TRAIL(c);
1.1348 + mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
1.1349 + buf[--destIx] = U16_LEAD(c);
1.1350 + mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
1.1351 + }
1.1352 +
1.1353 + // Fill in the map from native indexes to UChars buf index.
1.1354 + do {
1.1355 + mapToUChars[sIx-- - toUCharsMapStart] = (uint8_t)destIx;
1.1356 + } while (sIx >= srcIx);
1.1357 +
1.1358 + // Set native indexing limit to be the current position.
1.1359 + // We are processing a non-ascii, non-native-indexing char now;
1.1360 + // the limit will be here if the rest of the chars to be
1.1361 + // added to this buffer are ascii.
1.1362 + bufNILimit = destIx;
1.1363 + }
1.1364 + }
1.1365 + u8b->bufNativeStart = srcIx;
1.1366 + u8b->bufNativeLimit = ix;
1.1367 + u8b->bufStartIdx = destIx;
1.1368 + u8b->bufLimitIdx = UTF8_TEXT_CHUNK_SIZE+2;
1.1369 + u8b->bufNILimit = bufNILimit - u8b->bufStartIdx;
1.1370 + u8b->toUCharsMapStart = toUCharsMapStart;
1.1371 +
1.1372 + ut->chunkContents = &buf[u8b->bufStartIdx];
1.1373 + ut->chunkLength = u8b->bufLimitIdx - u8b->bufStartIdx;
1.1374 + ut->chunkOffset = ut->chunkLength;
1.1375 + ut->chunkNativeStart = u8b->bufNativeStart;
1.1376 + ut->chunkNativeLimit = u8b->bufNativeLimit;
1.1377 + ut->nativeIndexingLimit = u8b->bufNILimit;
1.1378 + return TRUE;
1.1379 + }
1.1380 +
1.1381 +}
1.1382 +
1.1383 +
1.1384 +
1.1385 +//
1.1386 +// This is a slightly modified copy of u_strFromUTF8,
1.1387 +// Inserts a Replacement Char rather than failing on invalid UTF-8
1.1388 +// Removes unnecessary features.
1.1389 +//
1.1390 +static UChar*
1.1391 +utext_strFromUTF8(UChar *dest,
1.1392 + int32_t destCapacity,
1.1393 + int32_t *pDestLength,
1.1394 + const char* src,
1.1395 + int32_t srcLength, // required. NUL terminated not supported.
1.1396 + UErrorCode *pErrorCode
1.1397 + )
1.1398 +{
1.1399 +
1.1400 + UChar *pDest = dest;
1.1401 + UChar *pDestLimit = (dest!=NULL)?(dest+destCapacity):NULL;
1.1402 + UChar32 ch=0;
1.1403 + int32_t index = 0;
1.1404 + int32_t reqLength = 0;
1.1405 + uint8_t* pSrc = (uint8_t*) src;
1.1406 +
1.1407 +
1.1408 + while((index < srcLength)&&(pDest<pDestLimit)){
1.1409 + ch = pSrc[index++];
1.1410 + if(ch <=0x7f){
1.1411 + *pDest++=(UChar)ch;
1.1412 + }else{
1.1413 + ch=utf8_nextCharSafeBody(pSrc, &index, srcLength, ch, -3);
1.1414 + if(U_IS_BMP(ch)){
1.1415 + *(pDest++)=(UChar)ch;
1.1416 + }else{
1.1417 + *(pDest++)=U16_LEAD(ch);
1.1418 + if(pDest<pDestLimit){
1.1419 + *(pDest++)=U16_TRAIL(ch);
1.1420 + }else{
1.1421 + reqLength++;
1.1422 + break;
1.1423 + }
1.1424 + }
1.1425 + }
1.1426 + }
1.1427 + /* donot fill the dest buffer just count the UChars needed */
1.1428 + while(index < srcLength){
1.1429 + ch = pSrc[index++];
1.1430 + if(ch <= 0x7f){
1.1431 + reqLength++;
1.1432 + }else{
1.1433 + ch=utf8_nextCharSafeBody(pSrc, &index, srcLength, ch, -3);
1.1434 + reqLength+=U16_LENGTH(ch);
1.1435 + }
1.1436 + }
1.1437 +
1.1438 + reqLength+=(int32_t)(pDest - dest);
1.1439 +
1.1440 + if(pDestLength){
1.1441 + *pDestLength = reqLength;
1.1442 + }
1.1443 +
1.1444 + /* Terminate the buffer */
1.1445 + u_terminateUChars(dest,destCapacity,reqLength,pErrorCode);
1.1446 +
1.1447 + return dest;
1.1448 +}
1.1449 +
1.1450 +
1.1451 +
1.1452 +static int32_t U_CALLCONV
1.1453 +utf8TextExtract(UText *ut,
1.1454 + int64_t start, int64_t limit,
1.1455 + UChar *dest, int32_t destCapacity,
1.1456 + UErrorCode *pErrorCode) {
1.1457 + if(U_FAILURE(*pErrorCode)) {
1.1458 + return 0;
1.1459 + }
1.1460 + if(destCapacity<0 || (dest==NULL && destCapacity>0)) {
1.1461 + *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
1.1462 + return 0;
1.1463 + }
1.1464 + int32_t length = ut->b;
1.1465 + int32_t start32 = pinIndex(start, length);
1.1466 + int32_t limit32 = pinIndex(limit, length);
1.1467 +
1.1468 + if(start32>limit32) {
1.1469 + *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR;
1.1470 + return 0;
1.1471 + }
1.1472 +
1.1473 +
1.1474 + // adjust the incoming indexes to land on code point boundaries if needed.
1.1475 + // adjust by no more than three, because that is the largest number of trail bytes
1.1476 + // in a well formed UTF8 character.
1.1477 + const uint8_t *buf = (const uint8_t *)ut->context;
1.1478 + int i;
1.1479 + if (start32 < ut->chunkNativeLimit) {
1.1480 + for (i=0; i<3; i++) {
1.1481 + if (U8_IS_SINGLE(buf[start32]) || U8_IS_LEAD(buf[start32]) || start32==0) {
1.1482 + break;
1.1483 + }
1.1484 + start32--;
1.1485 + }
1.1486 + }
1.1487 +
1.1488 + if (limit32 < ut->chunkNativeLimit) {
1.1489 + for (i=0; i<3; i++) {
1.1490 + if (U8_IS_SINGLE(buf[limit32]) || U8_IS_LEAD(buf[limit32]) || limit32==0) {
1.1491 + break;
1.1492 + }
1.1493 + limit32--;
1.1494 + }
1.1495 + }
1.1496 +
1.1497 + // Do the actual extract.
1.1498 + int32_t destLength=0;
1.1499 + utext_strFromUTF8(dest, destCapacity, &destLength,
1.1500 + (const char *)ut->context+start32, limit32-start32,
1.1501 + pErrorCode);
1.1502 + utf8TextAccess(ut, limit32, TRUE);
1.1503 + return destLength;
1.1504 +}
1.1505 +
1.1506 +//
1.1507 +// utf8TextMapOffsetToNative
1.1508 +//
1.1509 +// Map a chunk (UTF-16) offset to a native index.
1.1510 +static int64_t U_CALLCONV
1.1511 +utf8TextMapOffsetToNative(const UText *ut) {
1.1512 + //
1.1513 + UTF8Buf *u8b = (UTF8Buf *)ut->p;
1.1514 + U_ASSERT(ut->chunkOffset>ut->nativeIndexingLimit && ut->chunkOffset<=ut->chunkLength);
1.1515 + int32_t nativeOffset = u8b->mapToNative[ut->chunkOffset + u8b->bufStartIdx] + u8b->toUCharsMapStart;
1.1516 + U_ASSERT(nativeOffset >= ut->chunkNativeStart && nativeOffset <= ut->chunkNativeLimit);
1.1517 + return nativeOffset;
1.1518 +}
1.1519 +
1.1520 +//
1.1521 +// Map a native index to the corrsponding chunk offset
1.1522 +//
1.1523 +static int32_t U_CALLCONV
1.1524 +utf8TextMapIndexToUTF16(const UText *ut, int64_t index64) {
1.1525 + U_ASSERT(index64 <= 0x7fffffff);
1.1526 + int32_t index = (int32_t)index64;
1.1527 + UTF8Buf *u8b = (UTF8Buf *)ut->p;
1.1528 + U_ASSERT(index>=ut->chunkNativeStart+ut->nativeIndexingLimit);
1.1529 + U_ASSERT(index<=ut->chunkNativeLimit);
1.1530 + int32_t mapIndex = index - u8b->toUCharsMapStart;
1.1531 + int32_t offset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;
1.1532 + U_ASSERT(offset>=0 && offset<=ut->chunkLength);
1.1533 + return offset;
1.1534 +}
1.1535 +
1.1536 +static UText * U_CALLCONV
1.1537 +utf8TextClone(UText *dest, const UText *src, UBool deep, UErrorCode *status)
1.1538 +{
1.1539 + // First do a generic shallow clone. Does everything needed for the UText struct itself.
1.1540 + dest = shallowTextClone(dest, src, status);
1.1541 +
1.1542 + // For deep clones, make a copy of the string.
1.1543 + // The copied storage is owned by the newly created clone.
1.1544 + //
1.1545 + // TODO: There is an isssue with using utext_nativeLength().
1.1546 + // That function is non-const in cases where the input was NUL terminated
1.1547 + // and the length has not yet been determined.
1.1548 + // This function (clone()) is const.
1.1549 + // There potentially a thread safety issue lurking here.
1.1550 + //
1.1551 + if (deep && U_SUCCESS(*status)) {
1.1552 + int32_t len = (int32_t)utext_nativeLength((UText *)src);
1.1553 + char *copyStr = (char *)uprv_malloc(len+1);
1.1554 + if (copyStr == NULL) {
1.1555 + *status = U_MEMORY_ALLOCATION_ERROR;
1.1556 + } else {
1.1557 + uprv_memcpy(copyStr, src->context, len+1);
1.1558 + dest->context = copyStr;
1.1559 + dest->providerProperties |= I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT);
1.1560 + }
1.1561 + }
1.1562 + return dest;
1.1563 +}
1.1564 +
1.1565 +
1.1566 +static void U_CALLCONV
1.1567 +utf8TextClose(UText *ut) {
1.1568 + // Most of the work of close is done by the generic UText framework close.
1.1569 + // All that needs to be done here is to delete the UTF8 string if the UText
1.1570 + // owns it. This occurs if the UText was created by cloning.
1.1571 + if (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT)) {
1.1572 + char *s = (char *)ut->context;
1.1573 + uprv_free(s);
1.1574 + ut->context = NULL;
1.1575 + }
1.1576 +}
1.1577 +
1.1578 +U_CDECL_END
1.1579 +
1.1580 +
1.1581 +static const struct UTextFuncs utf8Funcs =
1.1582 +{
1.1583 + sizeof(UTextFuncs),
1.1584 + 0, 0, 0, // Reserved alignment padding
1.1585 + utf8TextClone,
1.1586 + utf8TextLength,
1.1587 + utf8TextAccess,
1.1588 + utf8TextExtract,
1.1589 + NULL, /* replace*/
1.1590 + NULL, /* copy */
1.1591 + utf8TextMapOffsetToNative,
1.1592 + utf8TextMapIndexToUTF16,
1.1593 + utf8TextClose,
1.1594 + NULL, // spare 1
1.1595 + NULL, // spare 2
1.1596 + NULL // spare 3
1.1597 +};
1.1598 +
1.1599 +
1.1600 +static const char gEmptyString[] = {0};
1.1601 +
1.1602 +U_CAPI UText * U_EXPORT2
1.1603 +utext_openUTF8(UText *ut, const char *s, int64_t length, UErrorCode *status) {
1.1604 + if(U_FAILURE(*status)) {
1.1605 + return NULL;
1.1606 + }
1.1607 + if(s==NULL && length==0) {
1.1608 + s = gEmptyString;
1.1609 + }
1.1610 +
1.1611 + if(s==NULL || length<-1 || length>INT32_MAX) {
1.1612 + *status=U_ILLEGAL_ARGUMENT_ERROR;
1.1613 + return NULL;
1.1614 + }
1.1615 +
1.1616 + ut = utext_setup(ut, sizeof(UTF8Buf) * 2, status);
1.1617 + if (U_FAILURE(*status)) {
1.1618 + return ut;
1.1619 + }
1.1620 +
1.1621 + ut->pFuncs = &utf8Funcs;
1.1622 + ut->context = s;
1.1623 + ut->b = (int32_t)length;
1.1624 + ut->c = (int32_t)length;
1.1625 + if (ut->c < 0) {
1.1626 + ut->c = 0;
1.1627 + ut->providerProperties |= I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
1.1628 + }
1.1629 + ut->p = ut->pExtra;
1.1630 + ut->q = (char *)ut->pExtra + sizeof(UTF8Buf);
1.1631 + return ut;
1.1632 +
1.1633 +}
1.1634 +
1.1635 +
1.1636 +
1.1637 +
1.1638 +
1.1639 +
1.1640 +
1.1641 +
1.1642 +//------------------------------------------------------------------------------
1.1643 +//
1.1644 +// UText implementation wrapper for Replaceable (read/write)
1.1645 +//
1.1646 +// Use of UText data members:
1.1647 +// context pointer to Replaceable.
1.1648 +// p pointer to Replaceable if it is owned by the UText.
1.1649 +//
1.1650 +//------------------------------------------------------------------------------
1.1651 +
1.1652 +
1.1653 +
1.1654 +// minimum chunk size for this implementation: 3
1.1655 +// to allow for possible trimming for code point boundaries
1.1656 +enum { REP_TEXT_CHUNK_SIZE=10 };
1.1657 +
1.1658 +struct ReplExtra {
1.1659 + /*
1.1660 + * Chunk UChars.
1.1661 + * +1 to simplify filling with surrogate pair at the end.
1.1662 + */
1.1663 + UChar s[REP_TEXT_CHUNK_SIZE+1];
1.1664 +};
1.1665 +
1.1666 +
1.1667 +U_CDECL_BEGIN
1.1668 +
1.1669 +static UText * U_CALLCONV
1.1670 +repTextClone(UText *dest, const UText *src, UBool deep, UErrorCode *status) {
1.1671 + // First do a generic shallow clone. Does everything needed for the UText struct itself.
1.1672 + dest = shallowTextClone(dest, src, status);
1.1673 +
1.1674 + // For deep clones, make a copy of the Replaceable.
1.1675 + // The copied Replaceable storage is owned by the newly created UText clone.
1.1676 + // A non-NULL pointer in UText.p is the signal to the close() function to delete
1.1677 + // it.
1.1678 + //
1.1679 + if (deep && U_SUCCESS(*status)) {
1.1680 + const Replaceable *replSrc = (const Replaceable *)src->context;
1.1681 + dest->context = replSrc->clone();
1.1682 + dest->providerProperties |= I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT);
1.1683 +
1.1684 + // with deep clone, the copy is writable, even when the source is not.
1.1685 + dest->providerProperties |= I32_FLAG(UTEXT_PROVIDER_WRITABLE);
1.1686 + }
1.1687 + return dest;
1.1688 +}
1.1689 +
1.1690 +
1.1691 +static void U_CALLCONV
1.1692 +repTextClose(UText *ut) {
1.1693 + // Most of the work of close is done by the generic UText framework close.
1.1694 + // All that needs to be done here is delete the Replaceable if the UText
1.1695 + // owns it. This occurs if the UText was created by cloning.
1.1696 + if (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT)) {
1.1697 + Replaceable *rep = (Replaceable *)ut->context;
1.1698 + delete rep;
1.1699 + ut->context = NULL;
1.1700 + }
1.1701 +}
1.1702 +
1.1703 +
1.1704 +static int64_t U_CALLCONV
1.1705 +repTextLength(UText *ut) {
1.1706 + const Replaceable *replSrc = (const Replaceable *)ut->context;
1.1707 + int32_t len = replSrc->length();
1.1708 + return len;
1.1709 +}
1.1710 +
1.1711 +
1.1712 +static UBool U_CALLCONV
1.1713 +repTextAccess(UText *ut, int64_t index, UBool forward) {
1.1714 + const Replaceable *rep=(const Replaceable *)ut->context;
1.1715 + int32_t length=rep->length(); // Full length of the input text (bigger than a chunk)
1.1716 +
1.1717 + // clip the requested index to the limits of the text.
1.1718 + int32_t index32 = pinIndex(index, length);
1.1719 + U_ASSERT(index<=INT32_MAX);
1.1720 +
1.1721 +
1.1722 + /*
1.1723 + * Compute start/limit boundaries around index, for a segment of text
1.1724 + * to be extracted.
1.1725 + * To allow for the possibility that our user gave an index to the trailing
1.1726 + * half of a surrogate pair, we must request one extra preceding UChar when
1.1727 + * going in the forward direction. This will ensure that the buffer has the
1.1728 + * entire code point at the specified index.
1.1729 + */
1.1730 + if(forward) {
1.1731 +
1.1732 + if (index32>=ut->chunkNativeStart && index32<ut->chunkNativeLimit) {
1.1733 + // Buffer already contains the requested position.
1.1734 + ut->chunkOffset = (int32_t)(index - ut->chunkNativeStart);
1.1735 + return TRUE;
1.1736 + }
1.1737 + if (index32>=length && ut->chunkNativeLimit==length) {
1.1738 + // Request for end of string, and buffer already extends up to it.
1.1739 + // Can't get the data, but don't change the buffer.
1.1740 + ut->chunkOffset = length - (int32_t)ut->chunkNativeStart;
1.1741 + return FALSE;
1.1742 + }
1.1743 +
1.1744 + ut->chunkNativeLimit = index + REP_TEXT_CHUNK_SIZE - 1;
1.1745 + // Going forward, so we want to have the buffer with stuff at and beyond
1.1746 + // the requested index. The -1 gets us one code point before the
1.1747 + // requested index also, to handle the case of the index being on
1.1748 + // a trail surrogate of a surrogate pair.
1.1749 + if(ut->chunkNativeLimit > length) {
1.1750 + ut->chunkNativeLimit = length;
1.1751 + }
1.1752 + // unless buffer ran off end, start is index-1.
1.1753 + ut->chunkNativeStart = ut->chunkNativeLimit - REP_TEXT_CHUNK_SIZE;
1.1754 + if(ut->chunkNativeStart < 0) {
1.1755 + ut->chunkNativeStart = 0;
1.1756 + }
1.1757 + } else {
1.1758 + // Reverse iteration. Fill buffer with data preceding the requested index.
1.1759 + if (index32>ut->chunkNativeStart && index32<=ut->chunkNativeLimit) {
1.1760 + // Requested position already in buffer.
1.1761 + ut->chunkOffset = index32 - (int32_t)ut->chunkNativeStart;
1.1762 + return TRUE;
1.1763 + }
1.1764 + if (index32==0 && ut->chunkNativeStart==0) {
1.1765 + // Request for start, buffer already begins at start.
1.1766 + // No data, but keep the buffer as is.
1.1767 + ut->chunkOffset = 0;
1.1768 + return FALSE;
1.1769 + }
1.1770 +
1.1771 + // Figure out the bounds of the chunk to extract for reverse iteration.
1.1772 + // Need to worry about chunk not splitting surrogate pairs, and while still
1.1773 + // containing the data we need.
1.1774 + // Fix by requesting a chunk that includes an extra UChar at the end.
1.1775 + // If this turns out to be a lead surrogate, we can lop it off and still have
1.1776 + // the data we wanted.
1.1777 + ut->chunkNativeStart = index32 + 1 - REP_TEXT_CHUNK_SIZE;
1.1778 + if (ut->chunkNativeStart < 0) {
1.1779 + ut->chunkNativeStart = 0;
1.1780 + }
1.1781 +
1.1782 + ut->chunkNativeLimit = index32 + 1;
1.1783 + if (ut->chunkNativeLimit > length) {
1.1784 + ut->chunkNativeLimit = length;
1.1785 + }
1.1786 + }
1.1787 +
1.1788 + // Extract the new chunk of text from the Replaceable source.
1.1789 + ReplExtra *ex = (ReplExtra *)ut->pExtra;
1.1790 + // UnicodeString with its buffer a writable alias to the chunk buffer
1.1791 + UnicodeString buffer(ex->s, 0 /*buffer length*/, REP_TEXT_CHUNK_SIZE /*buffer capacity*/);
1.1792 + rep->extractBetween((int32_t)ut->chunkNativeStart, (int32_t)ut->chunkNativeLimit, buffer);
1.1793 +
1.1794 + ut->chunkContents = ex->s;
1.1795 + ut->chunkLength = (int32_t)(ut->chunkNativeLimit - ut->chunkNativeStart);
1.1796 + ut->chunkOffset = (int32_t)(index32 - ut->chunkNativeStart);
1.1797 +
1.1798 + // Surrogate pairs from the input text must not span chunk boundaries.
1.1799 + // If end of chunk could be the start of a surrogate, trim it off.
1.1800 + if (ut->chunkNativeLimit < length &&
1.1801 + U16_IS_LEAD(ex->s[ut->chunkLength-1])) {
1.1802 + ut->chunkLength--;
1.1803 + ut->chunkNativeLimit--;
1.1804 + if (ut->chunkOffset > ut->chunkLength) {
1.1805 + ut->chunkOffset = ut->chunkLength;
1.1806 + }
1.1807 + }
1.1808 +
1.1809 + // if the first UChar in the chunk could be the trailing half of a surrogate pair,
1.1810 + // trim it off.
1.1811 + if(ut->chunkNativeStart>0 && U16_IS_TRAIL(ex->s[0])) {
1.1812 + ++(ut->chunkContents);
1.1813 + ++(ut->chunkNativeStart);
1.1814 + --(ut->chunkLength);
1.1815 + --(ut->chunkOffset);
1.1816 + }
1.1817 +
1.1818 + // adjust the index/chunkOffset to a code point boundary
1.1819 + U16_SET_CP_START(ut->chunkContents, 0, ut->chunkOffset);
1.1820 +
1.1821 + // Use fast indexing for get/setNativeIndex()
1.1822 + ut->nativeIndexingLimit = ut->chunkLength;
1.1823 +
1.1824 + return TRUE;
1.1825 +}
1.1826 +
1.1827 +
1.1828 +
1.1829 +static int32_t U_CALLCONV
1.1830 +repTextExtract(UText *ut,
1.1831 + int64_t start, int64_t limit,
1.1832 + UChar *dest, int32_t destCapacity,
1.1833 + UErrorCode *status) {
1.1834 + const Replaceable *rep=(const Replaceable *)ut->context;
1.1835 + int32_t length=rep->length();
1.1836 +
1.1837 + if(U_FAILURE(*status)) {
1.1838 + return 0;
1.1839 + }
1.1840 + if(destCapacity<0 || (dest==NULL && destCapacity>0)) {
1.1841 + *status=U_ILLEGAL_ARGUMENT_ERROR;
1.1842 + }
1.1843 + if(start>limit) {
1.1844 + *status=U_INDEX_OUTOFBOUNDS_ERROR;
1.1845 + return 0;
1.1846 + }
1.1847 +
1.1848 + int32_t start32 = pinIndex(start, length);
1.1849 + int32_t limit32 = pinIndex(limit, length);
1.1850 +
1.1851 + // adjust start, limit if they point to trail half of surrogates
1.1852 + if (start32<length && U16_IS_TRAIL(rep->charAt(start32)) &&
1.1853 + U_IS_SUPPLEMENTARY(rep->char32At(start32))){
1.1854 + start32--;
1.1855 + }
1.1856 + if (limit32<length && U16_IS_TRAIL(rep->charAt(limit32)) &&
1.1857 + U_IS_SUPPLEMENTARY(rep->char32At(limit32))){
1.1858 + limit32--;
1.1859 + }
1.1860 +
1.1861 + length=limit32-start32;
1.1862 + if(length>destCapacity) {
1.1863 + limit32 = start32 + destCapacity;
1.1864 + }
1.1865 + UnicodeString buffer(dest, 0, destCapacity); // writable alias
1.1866 + rep->extractBetween(start32, limit32, buffer);
1.1867 + repTextAccess(ut, limit32, TRUE);
1.1868 +
1.1869 + return u_terminateUChars(dest, destCapacity, length, status);
1.1870 +}
1.1871 +
1.1872 +static int32_t U_CALLCONV
1.1873 +repTextReplace(UText *ut,
1.1874 + int64_t start, int64_t limit,
1.1875 + const UChar *src, int32_t length,
1.1876 + UErrorCode *status) {
1.1877 + Replaceable *rep=(Replaceable *)ut->context;
1.1878 + int32_t oldLength;
1.1879 +
1.1880 + if(U_FAILURE(*status)) {
1.1881 + return 0;
1.1882 + }
1.1883 + if(src==NULL && length!=0) {
1.1884 + *status=U_ILLEGAL_ARGUMENT_ERROR;
1.1885 + return 0;
1.1886 + }
1.1887 + oldLength=rep->length(); // will subtract from new length
1.1888 + if(start>limit ) {
1.1889 + *status=U_INDEX_OUTOFBOUNDS_ERROR;
1.1890 + return 0;
1.1891 + }
1.1892 +
1.1893 + int32_t start32 = pinIndex(start, oldLength);
1.1894 + int32_t limit32 = pinIndex(limit, oldLength);
1.1895 +
1.1896 + // Snap start & limit to code point boundaries.
1.1897 + if (start32<oldLength && U16_IS_TRAIL(rep->charAt(start32)) &&
1.1898 + start32>0 && U16_IS_LEAD(rep->charAt(start32-1)))
1.1899 + {
1.1900 + start32--;
1.1901 + }
1.1902 + if (limit32<oldLength && U16_IS_LEAD(rep->charAt(limit32-1)) &&
1.1903 + U16_IS_TRAIL(rep->charAt(limit32)))
1.1904 + {
1.1905 + limit32++;
1.1906 + }
1.1907 +
1.1908 + // Do the actual replace operation using methods of the Replaceable class
1.1909 + UnicodeString replStr((UBool)(length<0), src, length); // read-only alias
1.1910 + rep->handleReplaceBetween(start32, limit32, replStr);
1.1911 + int32_t newLength = rep->length();
1.1912 + int32_t lengthDelta = newLength - oldLength;
1.1913 +
1.1914 + // Is the UText chunk buffer OK?
1.1915 + if (ut->chunkNativeLimit > start32) {
1.1916 + // this replace operation may have impacted the current chunk.
1.1917 + // invalidate it, which will force a reload on the next access.
1.1918 + invalidateChunk(ut);
1.1919 + }
1.1920 +
1.1921 + // set the iteration position to the end of the newly inserted replacement text.
1.1922 + int32_t newIndexPos = limit32 + lengthDelta;
1.1923 + repTextAccess(ut, newIndexPos, TRUE);
1.1924 +
1.1925 + return lengthDelta;
1.1926 +}
1.1927 +
1.1928 +
1.1929 +static void U_CALLCONV
1.1930 +repTextCopy(UText *ut,
1.1931 + int64_t start, int64_t limit,
1.1932 + int64_t destIndex,
1.1933 + UBool move,
1.1934 + UErrorCode *status)
1.1935 +{
1.1936 + Replaceable *rep=(Replaceable *)ut->context;
1.1937 + int32_t length=rep->length();
1.1938 +
1.1939 + if(U_FAILURE(*status)) {
1.1940 + return;
1.1941 + }
1.1942 + if (start>limit || (start<destIndex && destIndex<limit))
1.1943 + {
1.1944 + *status=U_INDEX_OUTOFBOUNDS_ERROR;
1.1945 + return;
1.1946 + }
1.1947 +
1.1948 + int32_t start32 = pinIndex(start, length);
1.1949 + int32_t limit32 = pinIndex(limit, length);
1.1950 + int32_t destIndex32 = pinIndex(destIndex, length);
1.1951 +
1.1952 + // TODO: snap input parameters to code point boundaries.
1.1953 +
1.1954 + if(move) {
1.1955 + // move: copy to destIndex, then replace original with nothing
1.1956 + int32_t segLength=limit32-start32;
1.1957 + rep->copy(start32, limit32, destIndex32);
1.1958 + if(destIndex32<start32) {
1.1959 + start32+=segLength;
1.1960 + limit32+=segLength;
1.1961 + }
1.1962 + rep->handleReplaceBetween(start32, limit32, UnicodeString());
1.1963 + } else {
1.1964 + // copy
1.1965 + rep->copy(start32, limit32, destIndex32);
1.1966 + }
1.1967 +
1.1968 + // If the change to the text touched the region in the chunk buffer,
1.1969 + // invalidate the buffer.
1.1970 + int32_t firstAffectedIndex = destIndex32;
1.1971 + if (move && start32<firstAffectedIndex) {
1.1972 + firstAffectedIndex = start32;
1.1973 + }
1.1974 + if (firstAffectedIndex < ut->chunkNativeLimit) {
1.1975 + // changes may have affected range covered by the chunk
1.1976 + invalidateChunk(ut);
1.1977 + }
1.1978 +
1.1979 + // Put iteration position at the newly inserted (moved) block,
1.1980 + int32_t nativeIterIndex = destIndex32 + limit32 - start32;
1.1981 + if (move && destIndex32>start32) {
1.1982 + // moved a block of text towards the end of the string.
1.1983 + nativeIterIndex = destIndex32;
1.1984 + }
1.1985 +
1.1986 + // Set position, reload chunk if needed.
1.1987 + repTextAccess(ut, nativeIterIndex, TRUE);
1.1988 +}
1.1989 +
1.1990 +static const struct UTextFuncs repFuncs =
1.1991 +{
1.1992 + sizeof(UTextFuncs),
1.1993 + 0, 0, 0, // Reserved alignment padding
1.1994 + repTextClone,
1.1995 + repTextLength,
1.1996 + repTextAccess,
1.1997 + repTextExtract,
1.1998 + repTextReplace,
1.1999 + repTextCopy,
1.2000 + NULL, // MapOffsetToNative,
1.2001 + NULL, // MapIndexToUTF16,
1.2002 + repTextClose,
1.2003 + NULL, // spare 1
1.2004 + NULL, // spare 2
1.2005 + NULL // spare 3
1.2006 +};
1.2007 +
1.2008 +
1.2009 +U_CAPI UText * U_EXPORT2
1.2010 +utext_openReplaceable(UText *ut, Replaceable *rep, UErrorCode *status)
1.2011 +{
1.2012 + if(U_FAILURE(*status)) {
1.2013 + return NULL;
1.2014 + }
1.2015 + if(rep==NULL) {
1.2016 + *status=U_ILLEGAL_ARGUMENT_ERROR;
1.2017 + return NULL;
1.2018 + }
1.2019 + ut = utext_setup(ut, sizeof(ReplExtra), status);
1.2020 +
1.2021 + ut->providerProperties = I32_FLAG(UTEXT_PROVIDER_WRITABLE);
1.2022 + if(rep->hasMetaData()) {
1.2023 + ut->providerProperties |=I32_FLAG(UTEXT_PROVIDER_HAS_META_DATA);
1.2024 + }
1.2025 +
1.2026 + ut->pFuncs = &repFuncs;
1.2027 + ut->context = rep;
1.2028 + return ut;
1.2029 +}
1.2030 +
1.2031 +U_CDECL_END
1.2032 +
1.2033 +
1.2034 +
1.2035 +
1.2036 +
1.2037 +
1.2038 +
1.2039 +
1.2040 +//------------------------------------------------------------------------------
1.2041 +//
1.2042 +// UText implementation for UnicodeString (read/write) and
1.2043 +// for const UnicodeString (read only)
1.2044 +// (same implementation, only the flags are different)
1.2045 +//
1.2046 +// Use of UText data members:
1.2047 +// context pointer to UnicodeString
1.2048 +// p pointer to UnicodeString IF this UText owns the string
1.2049 +// and it must be deleted on close(). NULL otherwise.
1.2050 +//
1.2051 +//------------------------------------------------------------------------------
1.2052 +
1.2053 +U_CDECL_BEGIN
1.2054 +
1.2055 +
1.2056 +static UText * U_CALLCONV
1.2057 +unistrTextClone(UText *dest, const UText *src, UBool deep, UErrorCode *status) {
1.2058 + // First do a generic shallow clone. Does everything needed for the UText struct itself.
1.2059 + dest = shallowTextClone(dest, src, status);
1.2060 +
1.2061 + // For deep clones, make a copy of the UnicodeSring.
1.2062 + // The copied UnicodeString storage is owned by the newly created UText clone.
1.2063 + // A non-NULL pointer in UText.p is the signal to the close() function to delete
1.2064 + // the UText.
1.2065 + //
1.2066 + if (deep && U_SUCCESS(*status)) {
1.2067 + const UnicodeString *srcString = (const UnicodeString *)src->context;
1.2068 + dest->context = new UnicodeString(*srcString);
1.2069 + dest->providerProperties |= I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT);
1.2070 +
1.2071 + // with deep clone, the copy is writable, even when the source is not.
1.2072 + dest->providerProperties |= I32_FLAG(UTEXT_PROVIDER_WRITABLE);
1.2073 + }
1.2074 + return dest;
1.2075 +}
1.2076 +
1.2077 +static void U_CALLCONV
1.2078 +unistrTextClose(UText *ut) {
1.2079 + // Most of the work of close is done by the generic UText framework close.
1.2080 + // All that needs to be done here is delete the UnicodeString if the UText
1.2081 + // owns it. This occurs if the UText was created by cloning.
1.2082 + if (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT)) {
1.2083 + UnicodeString *str = (UnicodeString *)ut->context;
1.2084 + delete str;
1.2085 + ut->context = NULL;
1.2086 + }
1.2087 +}
1.2088 +
1.2089 +
1.2090 +static int64_t U_CALLCONV
1.2091 +unistrTextLength(UText *t) {
1.2092 + return ((const UnicodeString *)t->context)->length();
1.2093 +}
1.2094 +
1.2095 +
1.2096 +static UBool U_CALLCONV
1.2097 +unistrTextAccess(UText *ut, int64_t index, UBool forward) {
1.2098 + int32_t length = ut->chunkLength;
1.2099 + ut->chunkOffset = pinIndex(index, length);
1.2100 +
1.2101 + // Check whether request is at the start or end
1.2102 + UBool retVal = (forward && index<length) || (!forward && index>0);
1.2103 + return retVal;
1.2104 +}
1.2105 +
1.2106 +
1.2107 +
1.2108 +static int32_t U_CALLCONV
1.2109 +unistrTextExtract(UText *t,
1.2110 + int64_t start, int64_t limit,
1.2111 + UChar *dest, int32_t destCapacity,
1.2112 + UErrorCode *pErrorCode) {
1.2113 + const UnicodeString *us=(const UnicodeString *)t->context;
1.2114 + int32_t length=us->length();
1.2115 +
1.2116 + if(U_FAILURE(*pErrorCode)) {
1.2117 + return 0;
1.2118 + }
1.2119 + if(destCapacity<0 || (dest==NULL && destCapacity>0)) {
1.2120 + *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
1.2121 + }
1.2122 + if(start<0 || start>limit) {
1.2123 + *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR;
1.2124 + return 0;
1.2125 + }
1.2126 +
1.2127 + int32_t start32 = start<length ? us->getChar32Start((int32_t)start) : length;
1.2128 + int32_t limit32 = limit<length ? us->getChar32Start((int32_t)limit) : length;
1.2129 +
1.2130 + length=limit32-start32;
1.2131 + if (destCapacity>0 && dest!=NULL) {
1.2132 + int32_t trimmedLength = length;
1.2133 + if(trimmedLength>destCapacity) {
1.2134 + trimmedLength=destCapacity;
1.2135 + }
1.2136 + us->extract(start32, trimmedLength, dest);
1.2137 + t->chunkOffset = start32+trimmedLength;
1.2138 + } else {
1.2139 + t->chunkOffset = start32;
1.2140 + }
1.2141 + u_terminateUChars(dest, destCapacity, length, pErrorCode);
1.2142 + return length;
1.2143 +}
1.2144 +
1.2145 +static int32_t U_CALLCONV
1.2146 +unistrTextReplace(UText *ut,
1.2147 + int64_t start, int64_t limit,
1.2148 + const UChar *src, int32_t length,
1.2149 + UErrorCode *pErrorCode) {
1.2150 + UnicodeString *us=(UnicodeString *)ut->context;
1.2151 + int32_t oldLength;
1.2152 +
1.2153 + if(U_FAILURE(*pErrorCode)) {
1.2154 + return 0;
1.2155 + }
1.2156 + if(src==NULL && length!=0) {
1.2157 + *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
1.2158 + }
1.2159 + if(start>limit) {
1.2160 + *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR;
1.2161 + return 0;
1.2162 + }
1.2163 + oldLength=us->length();
1.2164 + int32_t start32 = pinIndex(start, oldLength);
1.2165 + int32_t limit32 = pinIndex(limit, oldLength);
1.2166 + if (start32 < oldLength) {
1.2167 + start32 = us->getChar32Start(start32);
1.2168 + }
1.2169 + if (limit32 < oldLength) {
1.2170 + limit32 = us->getChar32Start(limit32);
1.2171 + }
1.2172 +
1.2173 + // replace
1.2174 + us->replace(start32, limit32-start32, src, length);
1.2175 + int32_t newLength = us->length();
1.2176 +
1.2177 + // Update the chunk description.
1.2178 + ut->chunkContents = us->getBuffer();
1.2179 + ut->chunkLength = newLength;
1.2180 + ut->chunkNativeLimit = newLength;
1.2181 + ut->nativeIndexingLimit = newLength;
1.2182 +
1.2183 + // Set iteration position to the point just following the newly inserted text.
1.2184 + int32_t lengthDelta = newLength - oldLength;
1.2185 + ut->chunkOffset = limit32 + lengthDelta;
1.2186 +
1.2187 + return lengthDelta;
1.2188 +}
1.2189 +
1.2190 +static void U_CALLCONV
1.2191 +unistrTextCopy(UText *ut,
1.2192 + int64_t start, int64_t limit,
1.2193 + int64_t destIndex,
1.2194 + UBool move,
1.2195 + UErrorCode *pErrorCode) {
1.2196 + UnicodeString *us=(UnicodeString *)ut->context;
1.2197 + int32_t length=us->length();
1.2198 +
1.2199 + if(U_FAILURE(*pErrorCode)) {
1.2200 + return;
1.2201 + }
1.2202 + int32_t start32 = pinIndex(start, length);
1.2203 + int32_t limit32 = pinIndex(limit, length);
1.2204 + int32_t destIndex32 = pinIndex(destIndex, length);
1.2205 +
1.2206 + if( start32>limit32 || (start32<destIndex32 && destIndex32<limit32)) {
1.2207 + *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR;
1.2208 + return;
1.2209 + }
1.2210 +
1.2211 + if(move) {
1.2212 + // move: copy to destIndex, then replace original with nothing
1.2213 + int32_t segLength=limit32-start32;
1.2214 + us->copy(start32, limit32, destIndex32);
1.2215 + if(destIndex32<start32) {
1.2216 + start32+=segLength;
1.2217 + }
1.2218 + us->replace(start32, segLength, NULL, 0);
1.2219 + } else {
1.2220 + // copy
1.2221 + us->copy(start32, limit32, destIndex32);
1.2222 + }
1.2223 +
1.2224 + // update chunk description, set iteration position.
1.2225 + ut->chunkContents = us->getBuffer();
1.2226 + if (move==FALSE) {
1.2227 + // copy operation, string length grows
1.2228 + ut->chunkLength += limit32-start32;
1.2229 + ut->chunkNativeLimit = ut->chunkLength;
1.2230 + ut->nativeIndexingLimit = ut->chunkLength;
1.2231 + }
1.2232 +
1.2233 + // Iteration position to end of the newly inserted text.
1.2234 + ut->chunkOffset = destIndex32+limit32-start32;
1.2235 + if (move && destIndex32>start32) {
1.2236 + ut->chunkOffset = destIndex32;
1.2237 + }
1.2238 +
1.2239 +}
1.2240 +
1.2241 +static const struct UTextFuncs unistrFuncs =
1.2242 +{
1.2243 + sizeof(UTextFuncs),
1.2244 + 0, 0, 0, // Reserved alignment padding
1.2245 + unistrTextClone,
1.2246 + unistrTextLength,
1.2247 + unistrTextAccess,
1.2248 + unistrTextExtract,
1.2249 + unistrTextReplace,
1.2250 + unistrTextCopy,
1.2251 + NULL, // MapOffsetToNative,
1.2252 + NULL, // MapIndexToUTF16,
1.2253 + unistrTextClose,
1.2254 + NULL, // spare 1
1.2255 + NULL, // spare 2
1.2256 + NULL // spare 3
1.2257 +};
1.2258 +
1.2259 +
1.2260 +
1.2261 +U_CDECL_END
1.2262 +
1.2263 +
1.2264 +U_CAPI UText * U_EXPORT2
1.2265 +utext_openUnicodeString(UText *ut, UnicodeString *s, UErrorCode *status) {
1.2266 + ut = utext_openConstUnicodeString(ut, s, status);
1.2267 + if (U_SUCCESS(*status)) {
1.2268 + ut->providerProperties |= I32_FLAG(UTEXT_PROVIDER_WRITABLE);
1.2269 + }
1.2270 + return ut;
1.2271 +}
1.2272 +
1.2273 +
1.2274 +
1.2275 +U_CAPI UText * U_EXPORT2
1.2276 +utext_openConstUnicodeString(UText *ut, const UnicodeString *s, UErrorCode *status) {
1.2277 + if (U_SUCCESS(*status) && s->isBogus()) {
1.2278 + // The UnicodeString is bogus, but we still need to detach the UText
1.2279 + // from whatever it was hooked to before, if anything.
1.2280 + utext_openUChars(ut, NULL, 0, status);
1.2281 + *status = U_ILLEGAL_ARGUMENT_ERROR;
1.2282 + return ut;
1.2283 + }
1.2284 + ut = utext_setup(ut, 0, status);
1.2285 + // note: use the standard (writable) function table for UnicodeString.
1.2286 + // The flag settings disable writing, so having the functions in
1.2287 + // the table is harmless.
1.2288 + if (U_SUCCESS(*status)) {
1.2289 + ut->pFuncs = &unistrFuncs;
1.2290 + ut->context = s;
1.2291 + ut->providerProperties = I32_FLAG(UTEXT_PROVIDER_STABLE_CHUNKS);
1.2292 + ut->chunkContents = s->getBuffer();
1.2293 + ut->chunkLength = s->length();
1.2294 + ut->chunkNativeStart = 0;
1.2295 + ut->chunkNativeLimit = ut->chunkLength;
1.2296 + ut->nativeIndexingLimit = ut->chunkLength;
1.2297 + }
1.2298 + return ut;
1.2299 +}
1.2300 +
1.2301 +//------------------------------------------------------------------------------
1.2302 +//
1.2303 +// UText implementation for const UChar * strings
1.2304 +//
1.2305 +// Use of UText data members:
1.2306 +// context pointer to UnicodeString
1.2307 +// a length. -1 if not yet known.
1.2308 +//
1.2309 +// TODO: support 64 bit lengths.
1.2310 +//
1.2311 +//------------------------------------------------------------------------------
1.2312 +
1.2313 +U_CDECL_BEGIN
1.2314 +
1.2315 +
1.2316 +static UText * U_CALLCONV
1.2317 +ucstrTextClone(UText *dest, const UText * src, UBool deep, UErrorCode * status) {
1.2318 + // First do a generic shallow clone.
1.2319 + dest = shallowTextClone(dest, src, status);
1.2320 +
1.2321 + // For deep clones, make a copy of the string.
1.2322 + // The copied storage is owned by the newly created clone.
1.2323 + // A non-NULL pointer in UText.p is the signal to the close() function to delete
1.2324 + // it.
1.2325 + //
1.2326 + if (deep && U_SUCCESS(*status)) {
1.2327 + U_ASSERT(utext_nativeLength(dest) < INT32_MAX);
1.2328 + int32_t len = (int32_t)utext_nativeLength(dest);
1.2329 +
1.2330 + // The cloned string IS going to be NUL terminated, whether or not the original was.
1.2331 + const UChar *srcStr = (const UChar *)src->context;
1.2332 + UChar *copyStr = (UChar *)uprv_malloc((len+1) * sizeof(UChar));
1.2333 + if (copyStr == NULL) {
1.2334 + *status = U_MEMORY_ALLOCATION_ERROR;
1.2335 + } else {
1.2336 + int64_t i;
1.2337 + for (i=0; i<len; i++) {
1.2338 + copyStr[i] = srcStr[i];
1.2339 + }
1.2340 + copyStr[len] = 0;
1.2341 + dest->context = copyStr;
1.2342 + dest->providerProperties |= I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT);
1.2343 + }
1.2344 + }
1.2345 + return dest;
1.2346 +}
1.2347 +
1.2348 +
1.2349 +static void U_CALLCONV
1.2350 +ucstrTextClose(UText *ut) {
1.2351 + // Most of the work of close is done by the generic UText framework close.
1.2352 + // All that needs to be done here is delete the string if the UText
1.2353 + // owns it. This occurs if the UText was created by cloning.
1.2354 + if (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT)) {
1.2355 + UChar *s = (UChar *)ut->context;
1.2356 + uprv_free(s);
1.2357 + ut->context = NULL;
1.2358 + }
1.2359 +}
1.2360 +
1.2361 +
1.2362 +
1.2363 +static int64_t U_CALLCONV
1.2364 +ucstrTextLength(UText *ut) {
1.2365 + if (ut->a < 0) {
1.2366 + // null terminated, we don't yet know the length. Scan for it.
1.2367 + // Access is not convenient for doing this
1.2368 + // because the current interation postion can't be changed.
1.2369 + const UChar *str = (const UChar *)ut->context;
1.2370 + for (;;) {
1.2371 + if (str[ut->chunkNativeLimit] == 0) {
1.2372 + break;
1.2373 + }
1.2374 + ut->chunkNativeLimit++;
1.2375 + }
1.2376 + ut->a = ut->chunkNativeLimit;
1.2377 + ut->chunkLength = (int32_t)ut->chunkNativeLimit;
1.2378 + ut->nativeIndexingLimit = ut->chunkLength;
1.2379 + ut->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
1.2380 + }
1.2381 + return ut->a;
1.2382 +}
1.2383 +
1.2384 +
1.2385 +static UBool U_CALLCONV
1.2386 +ucstrTextAccess(UText *ut, int64_t index, UBool forward) {
1.2387 + const UChar *str = (const UChar *)ut->context;
1.2388 +
1.2389 + // pin the requested index to the bounds of the string,
1.2390 + // and set current iteration position.
1.2391 + if (index<0) {
1.2392 + index = 0;
1.2393 + } else if (index < ut->chunkNativeLimit) {
1.2394 + // The request data is within the chunk as it is known so far.
1.2395 + // Put index on a code point boundary.
1.2396 + U16_SET_CP_START(str, 0, index);
1.2397 + } else if (ut->a >= 0) {
1.2398 + // We know the length of this string, and the user is requesting something
1.2399 + // at or beyond the length. Pin the requested index to the length.
1.2400 + index = ut->a;
1.2401 + } else {
1.2402 + // Null terminated string, length not yet known, and the requested index
1.2403 + // is beyond where we have scanned so far.
1.2404 + // Scan to 32 UChars beyond the requested index. The strategy here is
1.2405 + // to avoid fully scanning a long string when the caller only wants to
1.2406 + // see a few characters at its beginning.
1.2407 + int32_t scanLimit = (int32_t)index + 32;
1.2408 + if ((index + 32)>INT32_MAX || (index + 32)<0 ) { // note: int64 expression
1.2409 + scanLimit = INT32_MAX;
1.2410 + }
1.2411 +
1.2412 + int32_t chunkLimit = (int32_t)ut->chunkNativeLimit;
1.2413 + for (; chunkLimit<scanLimit; chunkLimit++) {
1.2414 + if (str[chunkLimit] == 0) {
1.2415 + // We found the end of the string. Remember it, pin the requested index to it,
1.2416 + // and bail out of here.
1.2417 + ut->a = chunkLimit;
1.2418 + ut->chunkLength = chunkLimit;
1.2419 + ut->nativeIndexingLimit = chunkLimit;
1.2420 + if (index >= chunkLimit) {
1.2421 + index = chunkLimit;
1.2422 + } else {
1.2423 + U16_SET_CP_START(str, 0, index);
1.2424 + }
1.2425 +
1.2426 + ut->chunkNativeLimit = chunkLimit;
1.2427 + ut->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
1.2428 + goto breakout;
1.2429 + }
1.2430 + }
1.2431 + // We scanned through the next batch of UChars without finding the end.
1.2432 + U16_SET_CP_START(str, 0, index);
1.2433 + if (chunkLimit == INT32_MAX) {
1.2434 + // Scanned to the limit of a 32 bit length.
1.2435 + // Forceably trim the overlength string back so length fits in int32
1.2436 + // TODO: add support for 64 bit strings.
1.2437 + ut->a = chunkLimit;
1.2438 + ut->chunkLength = chunkLimit;
1.2439 + ut->nativeIndexingLimit = chunkLimit;
1.2440 + if (index > chunkLimit) {
1.2441 + index = chunkLimit;
1.2442 + }
1.2443 + ut->chunkNativeLimit = chunkLimit;
1.2444 + ut->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
1.2445 + } else {
1.2446 + // The endpoint of a chunk must not be left in the middle of a surrogate pair.
1.2447 + // If the current end is on a lead surrogate, back the end up by one.
1.2448 + // It doesn't matter if the end char happens to be an unpaired surrogate,
1.2449 + // and it's simpler not to worry about it.
1.2450 + if (U16_IS_LEAD(str[chunkLimit-1])) {
1.2451 + --chunkLimit;
1.2452 + }
1.2453 + // Null-terminated chunk with end still unknown.
1.2454 + // Update the chunk length to reflect what has been scanned thus far.
1.2455 + // That the full length is still unknown is (still) flagged by
1.2456 + // ut->a being < 0.
1.2457 + ut->chunkNativeLimit = chunkLimit;
1.2458 + ut->nativeIndexingLimit = chunkLimit;
1.2459 + ut->chunkLength = chunkLimit;
1.2460 + }
1.2461 +
1.2462 + }
1.2463 +breakout:
1.2464 + U_ASSERT(index<=INT32_MAX);
1.2465 + ut->chunkOffset = (int32_t)index;
1.2466 +
1.2467 + // Check whether request is at the start or end
1.2468 + UBool retVal = (forward && index<ut->chunkNativeLimit) || (!forward && index>0);
1.2469 + return retVal;
1.2470 +}
1.2471 +
1.2472 +
1.2473 +
1.2474 +static int32_t U_CALLCONV
1.2475 +ucstrTextExtract(UText *ut,
1.2476 + int64_t start, int64_t limit,
1.2477 + UChar *dest, int32_t destCapacity,
1.2478 + UErrorCode *pErrorCode)
1.2479 +{
1.2480 + if(U_FAILURE(*pErrorCode)) {
1.2481 + return 0;
1.2482 + }
1.2483 + if(destCapacity<0 || (dest==NULL && destCapacity>0) || start>limit) {
1.2484 + *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
1.2485 + return 0;
1.2486 + }
1.2487 +
1.2488 + //const UChar *s=(const UChar *)ut->context;
1.2489 + int32_t si, di;
1.2490 +
1.2491 + int32_t start32;
1.2492 + int32_t limit32;
1.2493 +
1.2494 + // Access the start. Does two things we need:
1.2495 + // Pins 'start' to the length of the string, if it came in out-of-bounds.
1.2496 + // Snaps 'start' to the beginning of a code point.
1.2497 + ucstrTextAccess(ut, start, TRUE);
1.2498 + const UChar *s=ut->chunkContents;
1.2499 + start32 = ut->chunkOffset;
1.2500 +
1.2501 + int32_t strLength=(int32_t)ut->a;
1.2502 + if (strLength >= 0) {
1.2503 + limit32 = pinIndex(limit, strLength);
1.2504 + } else {
1.2505 + limit32 = pinIndex(limit, INT32_MAX);
1.2506 + }
1.2507 + di = 0;
1.2508 + for (si=start32; si<limit32; si++) {
1.2509 + if (strLength<0 && s[si]==0) {
1.2510 + // Just hit the end of a null-terminated string.
1.2511 + ut->a = si; // set string length for this UText
1.2512 + ut->chunkNativeLimit = si;
1.2513 + ut->chunkLength = si;
1.2514 + ut->nativeIndexingLimit = si;
1.2515 + strLength = si;
1.2516 + break;
1.2517 + }
1.2518 + U_ASSERT(di>=0); /* to ensure di never exceeds INT32_MAX, which must not happen logically */
1.2519 + if (di<destCapacity) {
1.2520 + // only store if there is space.
1.2521 + dest[di] = s[si];
1.2522 + } else {
1.2523 + if (strLength>=0) {
1.2524 + // We have filled the destination buffer, and the string length is known.
1.2525 + // Cut the loop short. There is no need to scan string termination.
1.2526 + di = limit32 - start32;
1.2527 + si = limit32;
1.2528 + break;
1.2529 + }
1.2530 + }
1.2531 + di++;
1.2532 + }
1.2533 +
1.2534 + // If the limit index points to a lead surrogate of a pair,
1.2535 + // add the corresponding trail surrogate to the destination.
1.2536 + if (si>0 && U16_IS_LEAD(s[si-1]) &&
1.2537 + ((si<strLength || strLength<0) && U16_IS_TRAIL(s[si])))
1.2538 + {
1.2539 + if (di<destCapacity) {
1.2540 + // store only if there is space in the output buffer.
1.2541 + dest[di++] = s[si++];
1.2542 + }
1.2543 + }
1.2544 +
1.2545 + // Put iteration position at the point just following the extracted text
1.2546 + ut->chunkOffset = uprv_min(strLength, start32 + destCapacity);
1.2547 +
1.2548 + // Add a terminating NUL if space in the buffer permits,
1.2549 + // and set the error status as required.
1.2550 + u_terminateUChars(dest, destCapacity, di, pErrorCode);
1.2551 + return di;
1.2552 +}
1.2553 +
1.2554 +static const struct UTextFuncs ucstrFuncs =
1.2555 +{
1.2556 + sizeof(UTextFuncs),
1.2557 + 0, 0, 0, // Reserved alignment padding
1.2558 + ucstrTextClone,
1.2559 + ucstrTextLength,
1.2560 + ucstrTextAccess,
1.2561 + ucstrTextExtract,
1.2562 + NULL, // Replace
1.2563 + NULL, // Copy
1.2564 + NULL, // MapOffsetToNative,
1.2565 + NULL, // MapIndexToUTF16,
1.2566 + ucstrTextClose,
1.2567 + NULL, // spare 1
1.2568 + NULL, // spare 2
1.2569 + NULL, // spare 3
1.2570 +};
1.2571 +
1.2572 +U_CDECL_END
1.2573 +
1.2574 +static const UChar gEmptyUString[] = {0};
1.2575 +
1.2576 +U_CAPI UText * U_EXPORT2
1.2577 +utext_openUChars(UText *ut, const UChar *s, int64_t length, UErrorCode *status) {
1.2578 + if (U_FAILURE(*status)) {
1.2579 + return NULL;
1.2580 + }
1.2581 + if(s==NULL && length==0) {
1.2582 + s = gEmptyUString;
1.2583 + }
1.2584 + if (s==NULL || length < -1 || length>INT32_MAX) {
1.2585 + *status = U_ILLEGAL_ARGUMENT_ERROR;
1.2586 + return NULL;
1.2587 + }
1.2588 + ut = utext_setup(ut, 0, status);
1.2589 + if (U_SUCCESS(*status)) {
1.2590 + ut->pFuncs = &ucstrFuncs;
1.2591 + ut->context = s;
1.2592 + ut->providerProperties = I32_FLAG(UTEXT_PROVIDER_STABLE_CHUNKS);
1.2593 + if (length==-1) {
1.2594 + ut->providerProperties |= I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
1.2595 + }
1.2596 + ut->a = length;
1.2597 + ut->chunkContents = s;
1.2598 + ut->chunkNativeStart = 0;
1.2599 + ut->chunkNativeLimit = length>=0? length : 0;
1.2600 + ut->chunkLength = (int32_t)ut->chunkNativeLimit;
1.2601 + ut->chunkOffset = 0;
1.2602 + ut->nativeIndexingLimit = ut->chunkLength;
1.2603 + }
1.2604 + return ut;
1.2605 +}
1.2606 +
1.2607 +
1.2608 +//------------------------------------------------------------------------------
1.2609 +//
1.2610 +// UText implementation for text from ICU CharacterIterators
1.2611 +//
1.2612 +// Use of UText data members:
1.2613 +// context pointer to the CharacterIterator
1.2614 +// a length of the full text.
1.2615 +// p pointer to buffer 1
1.2616 +// b start index of local buffer 1 contents
1.2617 +// q pointer to buffer 2
1.2618 +// c start index of local buffer 2 contents
1.2619 +// r pointer to the character iterator if the UText owns it.
1.2620 +// Null otherwise.
1.2621 +//
1.2622 +//------------------------------------------------------------------------------
1.2623 +#define CIBufSize 16
1.2624 +
1.2625 +U_CDECL_BEGIN
1.2626 +static void U_CALLCONV
1.2627 +charIterTextClose(UText *ut) {
1.2628 + // Most of the work of close is done by the generic UText framework close.
1.2629 + // All that needs to be done here is delete the CharacterIterator if the UText
1.2630 + // owns it. This occurs if the UText was created by cloning.
1.2631 + CharacterIterator *ci = (CharacterIterator *)ut->r;
1.2632 + delete ci;
1.2633 + ut->r = NULL;
1.2634 +}
1.2635 +
1.2636 +static int64_t U_CALLCONV
1.2637 +charIterTextLength(UText *ut) {
1.2638 + return (int32_t)ut->a;
1.2639 +}
1.2640 +
1.2641 +static UBool U_CALLCONV
1.2642 +charIterTextAccess(UText *ut, int64_t index, UBool forward) {
1.2643 + CharacterIterator *ci = (CharacterIterator *)ut->context;
1.2644 +
1.2645 + int32_t clippedIndex = (int32_t)index;
1.2646 + if (clippedIndex<0) {
1.2647 + clippedIndex=0;
1.2648 + } else if (clippedIndex>=ut->a) {
1.2649 + clippedIndex=(int32_t)ut->a;
1.2650 + }
1.2651 + int32_t neededIndex = clippedIndex;
1.2652 + if (!forward && neededIndex>0) {
1.2653 + // reverse iteration, want the position just before what was asked for.
1.2654 + neededIndex--;
1.2655 + } else if (forward && neededIndex==ut->a && neededIndex>0) {
1.2656 + // Forward iteration, don't ask for something past the end of the text.
1.2657 + neededIndex--;
1.2658 + }
1.2659 +
1.2660 + // Find the native index of the start of the buffer containing what we want.
1.2661 + neededIndex -= neededIndex % CIBufSize;
1.2662 +
1.2663 + UChar *buf = NULL;
1.2664 + UBool needChunkSetup = TRUE;
1.2665 + int i;
1.2666 + if (ut->chunkNativeStart == neededIndex) {
1.2667 + // The buffer we want is already the current chunk.
1.2668 + needChunkSetup = FALSE;
1.2669 + } else if (ut->b == neededIndex) {
1.2670 + // The first buffer (buffer p) has what we need.
1.2671 + buf = (UChar *)ut->p;
1.2672 + } else if (ut->c == neededIndex) {
1.2673 + // The second buffer (buffer q) has what we need.
1.2674 + buf = (UChar *)ut->q;
1.2675 + } else {
1.2676 + // Neither buffer already has what we need.
1.2677 + // Load new data from the character iterator.
1.2678 + // Use the buf that is not the current buffer.
1.2679 + buf = (UChar *)ut->p;
1.2680 + if (ut->p == ut->chunkContents) {
1.2681 + buf = (UChar *)ut->q;
1.2682 + }
1.2683 + ci->setIndex(neededIndex);
1.2684 + for (i=0; i<CIBufSize; i++) {
1.2685 + buf[i] = ci->nextPostInc();
1.2686 + if (i+neededIndex > ut->a) {
1.2687 + break;
1.2688 + }
1.2689 + }
1.2690 + }
1.2691 +
1.2692 + // We have a buffer with the data we need.
1.2693 + // Set it up as the current chunk, if it wasn't already.
1.2694 + if (needChunkSetup) {
1.2695 + ut->chunkContents = buf;
1.2696 + ut->chunkLength = CIBufSize;
1.2697 + ut->chunkNativeStart = neededIndex;
1.2698 + ut->chunkNativeLimit = neededIndex + CIBufSize;
1.2699 + if (ut->chunkNativeLimit > ut->a) {
1.2700 + ut->chunkNativeLimit = ut->a;
1.2701 + ut->chunkLength = (int32_t)(ut->chunkNativeLimit)-(int32_t)(ut->chunkNativeStart);
1.2702 + }
1.2703 + ut->nativeIndexingLimit = ut->chunkLength;
1.2704 + U_ASSERT(ut->chunkOffset>=0 && ut->chunkOffset<=CIBufSize);
1.2705 + }
1.2706 + ut->chunkOffset = clippedIndex - (int32_t)ut->chunkNativeStart;
1.2707 + UBool success = (forward? ut->chunkOffset<ut->chunkLength : ut->chunkOffset>0);
1.2708 + return success;
1.2709 +}
1.2710 +
1.2711 +static UText * U_CALLCONV
1.2712 +charIterTextClone(UText *dest, const UText *src, UBool deep, UErrorCode * status) {
1.2713 + if (U_FAILURE(*status)) {
1.2714 + return NULL;
1.2715 + }
1.2716 +
1.2717 + if (deep) {
1.2718 + // There is no CharacterIterator API for cloning the underlying text storage.
1.2719 + *status = U_UNSUPPORTED_ERROR;
1.2720 + return NULL;
1.2721 + } else {
1.2722 + CharacterIterator *srcCI =(CharacterIterator *)src->context;
1.2723 + srcCI = srcCI->clone();
1.2724 + dest = utext_openCharacterIterator(dest, srcCI, status);
1.2725 + // cast off const on getNativeIndex.
1.2726 + // For CharacterIterator based UTexts, this is safe, the operation is const.
1.2727 + int64_t ix = utext_getNativeIndex((UText *)src);
1.2728 + utext_setNativeIndex(dest, ix);
1.2729 + dest->r = srcCI; // flags that this UText owns the CharacterIterator
1.2730 + }
1.2731 + return dest;
1.2732 +}
1.2733 +
1.2734 +static int32_t U_CALLCONV
1.2735 +charIterTextExtract(UText *ut,
1.2736 + int64_t start, int64_t limit,
1.2737 + UChar *dest, int32_t destCapacity,
1.2738 + UErrorCode *status)
1.2739 +{
1.2740 + if(U_FAILURE(*status)) {
1.2741 + return 0;
1.2742 + }
1.2743 + if(destCapacity<0 || (dest==NULL && destCapacity>0) || start>limit) {
1.2744 + *status=U_ILLEGAL_ARGUMENT_ERROR;
1.2745 + return 0;
1.2746 + }
1.2747 + int32_t length = (int32_t)ut->a;
1.2748 + int32_t start32 = pinIndex(start, length);
1.2749 + int32_t limit32 = pinIndex(limit, length);
1.2750 + int32_t desti = 0;
1.2751 + int32_t srci;
1.2752 + int32_t copyLimit;
1.2753 +
1.2754 + CharacterIterator *ci = (CharacterIterator *)ut->context;
1.2755 + ci->setIndex32(start32); // Moves ix to lead of surrogate pair, if needed.
1.2756 + srci = ci->getIndex();
1.2757 + copyLimit = srci;
1.2758 + while (srci<limit32) {
1.2759 + UChar32 c = ci->next32PostInc();
1.2760 + int32_t len = U16_LENGTH(c);
1.2761 + U_ASSERT(desti+len>0); /* to ensure desti+len never exceeds MAX_INT32, which must not happen logically */
1.2762 + if (desti+len <= destCapacity) {
1.2763 + U16_APPEND_UNSAFE(dest, desti, c);
1.2764 + copyLimit = srci+len;
1.2765 + } else {
1.2766 + desti += len;
1.2767 + *status = U_BUFFER_OVERFLOW_ERROR;
1.2768 + }
1.2769 + srci += len;
1.2770 + }
1.2771 +
1.2772 + charIterTextAccess(ut, copyLimit, TRUE);
1.2773 +
1.2774 + u_terminateUChars(dest, destCapacity, desti, status);
1.2775 + return desti;
1.2776 +}
1.2777 +
1.2778 +static const struct UTextFuncs charIterFuncs =
1.2779 +{
1.2780 + sizeof(UTextFuncs),
1.2781 + 0, 0, 0, // Reserved alignment padding
1.2782 + charIterTextClone,
1.2783 + charIterTextLength,
1.2784 + charIterTextAccess,
1.2785 + charIterTextExtract,
1.2786 + NULL, // Replace
1.2787 + NULL, // Copy
1.2788 + NULL, // MapOffsetToNative,
1.2789 + NULL, // MapIndexToUTF16,
1.2790 + charIterTextClose,
1.2791 + NULL, // spare 1
1.2792 + NULL, // spare 2
1.2793 + NULL // spare 3
1.2794 +};
1.2795 +U_CDECL_END
1.2796 +
1.2797 +
1.2798 +U_CAPI UText * U_EXPORT2
1.2799 +utext_openCharacterIterator(UText *ut, CharacterIterator *ci, UErrorCode *status) {
1.2800 + if (U_FAILURE(*status)) {
1.2801 + return NULL;
1.2802 + }
1.2803 +
1.2804 + if (ci->startIndex() > 0) {
1.2805 + // No support for CharacterIterators that do not start indexing from zero.
1.2806 + *status = U_UNSUPPORTED_ERROR;
1.2807 + return NULL;
1.2808 + }
1.2809 +
1.2810 + // Extra space in UText for 2 buffers of CIBufSize UChars each.
1.2811 + int32_t extraSpace = 2 * CIBufSize * sizeof(UChar);
1.2812 + ut = utext_setup(ut, extraSpace, status);
1.2813 + if (U_SUCCESS(*status)) {
1.2814 + ut->pFuncs = &charIterFuncs;
1.2815 + ut->context = ci;
1.2816 + ut->providerProperties = 0;
1.2817 + ut->a = ci->endIndex(); // Length of text
1.2818 + ut->p = ut->pExtra; // First buffer
1.2819 + ut->b = -1; // Native index of first buffer contents
1.2820 + ut->q = (UChar*)ut->pExtra+CIBufSize; // Second buffer
1.2821 + ut->c = -1; // Native index of second buffer contents
1.2822 +
1.2823 + // Initialize current chunk contents to be empty.
1.2824 + // First access will fault something in.
1.2825 + // Note: The initial nativeStart and chunkOffset must sum to zero
1.2826 + // so that getNativeIndex() will correctly compute to zero
1.2827 + // if no call to Access() has ever been made. They can't be both
1.2828 + // zero without Access() thinking that the chunk is valid.
1.2829 + ut->chunkContents = (UChar *)ut->p;
1.2830 + ut->chunkNativeStart = -1;
1.2831 + ut->chunkOffset = 1;
1.2832 + ut->chunkNativeLimit = 0;
1.2833 + ut->chunkLength = 0;
1.2834 + ut->nativeIndexingLimit = ut->chunkOffset; // enables native indexing
1.2835 + }
1.2836 + return ut;
1.2837 +}
1.2838 +
1.2839 +
1.2840 +
