+// © 2016 and later: Unicode, Inc. and others.
+// License & terms of use: http://www.unicode.org/copyright.html
/*
*******************************************************************************
*
-* Copyright (C) 2005-2014, International Business Machines
+* Copyright (C) 2005-2016, International Business Machines
* Corporation and others. All Rights Reserved.
*
*******************************************************************************
* file name: utext.cpp
-* encoding: US-ASCII
+* encoding: UTF-8
* tab size: 8 (not used)
* indentation:4
*
//------------------------------------------------------------------------------
// Chunk size.
-// Must be less than 85, because of byte mapping from UChar indexes to native indexes.
+// Must be less than 85 (256/3), because of byte mapping from UChar indexes to native indexes.
// Worst case is three native bytes to one UChar. (Supplemenaries are 4 native bytes
// to two UChars.)
+// The longest illegal byte sequence treated as a single error (and converted to U+FFFD)
+// is a three-byte sequence (truncated four-byte sequence).
//
enum { UTF8_TEXT_CHUNK_SIZE=32 };
// Requested index is in this buffer.
u8b = (UTF8Buf *)ut->p; // the current buffer
mapIndex = ix - u8b->toUCharsMapStart;
+ U_ASSERT(mapIndex < (int32_t)sizeof(UTF8Buf::mapToUChars));
ut->chunkOffset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;
return TRUE;
// Can only do this if the incoming index is somewhere in the interior of the string.
// If index is at the end, there is no character there to look at.
if (ix != ut->b) {
+ // Note: this function will only move the index back if it is on a trail byte
+ // and there is a preceding lead byte and the sequence from the lead
+ // through this trail could be part of a valid UTF-8 sequence
+ // Otherwise the index remains unchanged.
U8_SET_CP_START(s8, 0, ix);
}
UChar *buf = u8b->buf;
uint8_t *mapToNative = u8b->mapToNative;
uint8_t *mapToUChars = u8b->mapToUChars;
- int32_t toUCharsMapStart = ix - (UTF8_TEXT_CHUNK_SIZE*3 + 1);
+ int32_t toUCharsMapStart = ix - sizeof(UTF8Buf::mapToUChars) + 1;
+ // Note that toUCharsMapStart can be negative. Happens when the remaining
+ // text from current position to the beginning is less than the buffer size.
+ // + 1 because mapToUChars must have a slot at the end for the bufNativeLimit entry.
int32_t destIx = UTF8_TEXT_CHUNK_SIZE+2; // Start in the overflow region
// at end of buffer to leave room
// for a surrogate pair at the
if (c<0x80) {
// Special case ASCII range for speed.
buf[destIx] = (UChar)c;
+ U_ASSERT(toUCharsMapStart <= srcIx);
mapToUChars[srcIx - toUCharsMapStart] = (uint8_t)destIx;
mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
} else {
do {
mapToUChars[sIx-- - toUCharsMapStart] = (uint8_t)destIx;
} while (sIx >= srcIx);
+ U_ASSERT(toUCharsMapStart <= (srcIx+1));
// Set native indexing limit to be the current position.
// We are processing a non-ascii, non-native-indexing char now;
U_ASSERT(index>=ut->chunkNativeStart+ut->nativeIndexingLimit);
U_ASSERT(index<=ut->chunkNativeLimit);
int32_t mapIndex = index - u8b->toUCharsMapStart;
+ U_ASSERT(mapIndex < (int32_t)sizeof(UTF8Buf::mapToUChars));
int32_t offset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;
U_ASSERT(offset>=0 && offset<=ut->chunkLength);
return offset;
return NULL;
}
ut = utext_setup(ut, sizeof(ReplExtra), status);
+ if(U_FAILURE(*status)) {
+ return ut;
+ }
ut->providerProperties = I32_FLAG(UTEXT_PROVIDER_WRITABLE);
if(rep->hasMetaData()) {
}
if(move) {
- // move: copy to destIndex, then replace original with nothing
+ // move: copy to destIndex, then remove original
int32_t segLength=limit32-start32;
us->copy(start32, limit32, destIndex32);
if(destIndex32<start32) {
start32+=segLength;
}
- us->replace(start32, segLength, NULL, 0);
+ us->remove(start32, segLength);
} else {
// copy
us->copy(start32, limit32, destIndex32);
ut->chunkLength = si;
ut->nativeIndexingLimit = si;
strLength = si;
+ limit32 = si;
break;
}
U_ASSERT(di>=0); /* to ensure di never exceeds INT32_MAX, which must not happen logically */
// If the limit index points to a lead surrogate of a pair,
// add the corresponding trail surrogate to the destination.
if (si>0 && U16_IS_LEAD(s[si-1]) &&
- ((si<strLength || strLength<0) && U16_IS_TRAIL(s[si])))
+ ((si<strLength || strLength<0) && U16_IS_TRAIL(s[si])))
{
if (di<destCapacity) {
// store only if there is space in the output buffer.
- dest[di++] = s[si++];
+ dest[di++] = s[si];
}
+ si++;
}
// Put iteration position at the point just following the extracted text
- ut->chunkOffset = uprv_min(strLength, start32 + destCapacity);
+ if (si <= ut->chunkNativeLimit) {
+ ut->chunkOffset = si;
+ } else {
+ ucstrTextAccess(ut, si, TRUE);
+ }
// Add a terminating NUL if space in the buffer permits,
// and set the error status as required.
CharacterIterator *srcCI =(CharacterIterator *)src->context;
srcCI = srcCI->clone();
dest = utext_openCharacterIterator(dest, srcCI, status);
+ if (U_FAILURE(*status)) {
+ return dest;
+ }
// cast off const on getNativeIndex.
// For CharacterIterator based UTexts, this is safe, the operation is const.
int64_t ix = utext_getNativeIndex((UText *)src);
projects / apple / icu.git / blobdiff