U_CAPI void U_EXPORT2
utext_setNativeIndex(UText *ut, int64_t index) { if(index<ut->chunkNativeStart || index>=ut->chunkNativeLimit) { // The desired position is outside of the current chunk. // Access the new position. Assume a forward iteration from here, // which will also be optimimum for a single random access. // Reverse iterations may suffer slightly.
ut->pFuncs->access(ut, index, true);
} elseif((int32_t)(index - ut->chunkNativeStart) <= ut->nativeIndexingLimit) { // utf-16 indexing.
ut->chunkOffset=(int32_t)(index-ut->chunkNativeStart);
} else {
ut->chunkOffset=ut->pFuncs->mapNativeIndexToUTF16(ut, index);
} // The convention is that the index must always be on a code point boundary. // Adjust the index position if it is in the middle of a surrogate pair. if (ut->chunkOffset<ut->chunkLength) {
char16_t c= ut->chunkContents[ut->chunkOffset]; if (U16_IS_TRAIL(c)) { if (ut->chunkOffset==0) {
ut->pFuncs->access(ut, ut->chunkNativeStart, false);
} if (ut->chunkOffset>0) {
char16_t lead = ut->chunkContents[ut->chunkOffset-1]; if (U16_IS_LEAD(lead)) {
ut->chunkOffset--;
}
}
}
}
}
U_CAPI int64_t U_EXPORT2
utext_getPreviousNativeIndex(UText *ut) { // // Fast-path the common case. // Common means current position is not at the beginning of a chunk // and the preceding character is not supplementary. //
int32_t i = ut->chunkOffset - 1;
int64_t result; if (i >= 0) {
char16_t c = ut->chunkContents[i]; if (U16_IS_TRAIL(c) == false) { if (i <= ut->nativeIndexingLimit) {
result = ut->chunkNativeStart + i;
} else {
ut->chunkOffset = i;
result = ut->pFuncs->mapOffsetToNative(ut);
ut->chunkOffset++;
} return result;
}
}
// If at the start of text, simply return 0. if (ut->chunkOffset==0 && ut->chunkNativeStart==0) { return 0;
}
// Harder, less common cases. We are at a chunk boundary, or on a surrogate. // Keep it simple, use other functions to handle the edges. //
utext_previous32(ut);
result = UTEXT_GETNATIVEINDEX(ut);
utext_next32(ut); return result;
}
// // utext_current32. Get the UChar32 at the current position. // UText iteration position is always on a code point boundary, // never on the trail half of a surrogate pair. //
U_CAPI UChar32 U_EXPORT2
utext_current32(UText *ut) {
UChar32 c; if (ut->chunkOffset==ut->chunkLength) { // Current position is just off the end of the chunk. if (ut->pFuncs->access(ut, ut->chunkNativeLimit, true) == false) { // Off the end of the text. return U_SENTINEL;
}
}
c = ut->chunkContents[ut->chunkOffset]; if (U16_IS_LEAD(c) == false) { // Normal, non-supplementary case. return c;
}
// // Possible supplementary char. //
UChar32 trail = 0;
UChar32 supplementaryC = c; if ((ut->chunkOffset+1) < ut->chunkLength) { // The trail surrogate is in the same chunk.
trail = ut->chunkContents[ut->chunkOffset+1];
} else { // The trail surrogate is in a different chunk. // Because we must maintain the iteration position, we need to switch forward // into the new chunk, get the trail surrogate, then revert the chunk back to the // original one. // An edge case to be careful of: the entire text may end with an unpaired // leading surrogate. The attempt to access the trail will fail, but // the original position before the unpaired lead still needs to be restored.
int64_t nativePosition = ut->chunkNativeLimit; if (ut->pFuncs->access(ut, nativePosition, true)) {
trail = ut->chunkContents[ut->chunkOffset];
}
UBool r = ut->pFuncs->access(ut, nativePosition, false); // reverse iteration flag loads preceding chunk
U_ASSERT(r); // Here we need to restore chunkOffset since the access functions were called with // chunkNativeLimit but that is not where we were (we were 1 code unit before the // limit). Restoring was originally added in ICU-4669 but did not support access // functions that changed the chunk size, the following does.
ut->chunkOffset = ut->chunkLength - 1; if(!r) { return U_SENTINEL;
}
}
if (U16_IS_TRAIL(trail)) {
supplementaryC = U16_GET_SUPPLEMENTARY(c, trail);
} return supplementaryC;
// Fast path the common case. if (nativeIndex>=ut->chunkNativeStart && nativeIndex < ut->chunkNativeStart + ut->nativeIndexingLimit) {
ut->chunkOffset = (int32_t)(nativeIndex - ut->chunkNativeStart);
c = ut->chunkContents[ut->chunkOffset]; if (U16_IS_SURROGATE(c) == false) { return c;
}
}
utext_setNativeIndex(ut, nativeIndex); if (nativeIndex>=ut->chunkNativeStart && ut->chunkOffset<ut->chunkLength) {
c = ut->chunkContents[ut->chunkOffset]; if (U16_IS_SURROGATE(c)) { // For surrogates, let current32() deal with the complications // of supplementaries that may span chunk boundaries.
c = utext_current32(ut);
}
} return c;
}
if (ut->chunkOffset >= ut->chunkLength) { if (ut->pFuncs->access(ut, ut->chunkNativeLimit, true) == false) { return U_SENTINEL;
}
}
c = ut->chunkContents[ut->chunkOffset++]; if (U16_IS_LEAD(c) == false) { // Normal case, not supplementary. // (A trail surrogate seen here is just returned as is, as a surrogate value. // It cannot be part of a pair.) return c;
}
if (ut->chunkOffset >= ut->chunkLength) { if (ut->pFuncs->access(ut, ut->chunkNativeLimit, true) == false) { // c is an unpaired lead surrogate at the end of the text. // return it as it is. return c;
}
}
UChar32 trail = ut->chunkContents[ut->chunkOffset]; if (U16_IS_TRAIL(trail) == false) { // c was an unpaired lead surrogate, not at the end of the text. // return it as it is (unpaired). Iteration position is on the // following character, possibly in the next chunk, where the // trail surrogate would have been if it had existed. return c;
}
UChar32 supplementary = U16_GET_SUPPLEMENTARY(c, trail);
ut->chunkOffset++; // move iteration position over the trail surrogate. return supplementary;
}
if (ut->chunkOffset <= 0) { if (ut->pFuncs->access(ut, ut->chunkNativeStart, false) == false) { return U_SENTINEL;
}
}
ut->chunkOffset--;
c = ut->chunkContents[ut->chunkOffset]; if (U16_IS_TRAIL(c) == false) { // Normal case, not supplementary. // (A lead surrogate seen here is just returned as is, as a surrogate value. // It cannot be part of a pair.) return c;
}
if (ut->chunkOffset <= 0) { if (ut->pFuncs->access(ut, ut->chunkNativeStart, false) == false) { // c is an unpaired trail surrogate at the start of the text. // return it as it is. return c;
}
}
UChar32 lead = ut->chunkContents[ut->chunkOffset-1]; if (U16_IS_LEAD(lead) == false) { // c was an unpaired trail surrogate, not at the end of the text. // return it as it is (unpaired). Iteration position is at c return c;
}
UChar32 supplementary = U16_GET_SUPPLEMENTARY(lead, c);
ut->chunkOffset--; // move iteration position over the lead surrogate. return supplementary;
}
if(index<ut->chunkNativeStart || index>=ut->chunkNativeLimit) { // Desired position is outside of the current chunk. if(!ut->pFuncs->access(ut, index, true)) { // no chunk available here return U_SENTINEL;
}
} elseif (index - ut->chunkNativeStart <= (int64_t)ut->nativeIndexingLimit) { // Desired position is in chunk, with direct 1:1 native to UTF16 indexing
ut->chunkOffset = (int32_t)(index - ut->chunkNativeStart);
} else { // Desired position is in chunk, with non-UTF16 indexing.
ut->chunkOffset = ut->pFuncs->mapNativeIndexToUTF16(ut, index);
}
c = ut->chunkContents[ut->chunkOffset++]; if (U16_IS_SURROGATE(c)) { // Surrogates. Many edge cases. Use other functions that already // deal with the problems.
utext_setNativeIndex(ut, index);
c = utext_next32(ut);
} return c;
}
U_CAPI UChar32 U_EXPORT2
utext_previous32From(UText *ut, int64_t index) { // // Return the character preceding the specified index. // Leave the iteration position at the start of the character that was returned. //
UChar32 cPrev; // The character preceding cCurr, which is what we will return.
// Address the chunk containing the position preceding the incoming index // A tricky edge case: // We try to test the requested native index against the chunkNativeStart to determine // whether the character preceding the one at the index is in the current chunk. // BUT, this test can fail with UTF-8 (or any other multibyte encoding), when the // requested index is on something other than the first position of the first char. // if(index<=ut->chunkNativeStart || index>ut->chunkNativeLimit) { // Requested native index is outside of the current chunk. if(!ut->pFuncs->access(ut, index, false)) { // no chunk available here return U_SENTINEL;
}
} elseif(index - ut->chunkNativeStart <= (int64_t)ut->nativeIndexingLimit) { // Direct UTF-16 indexing.
ut->chunkOffset = (int32_t)(index - ut->chunkNativeStart);
} else {
ut->chunkOffset=ut->pFuncs->mapNativeIndexToUTF16(ut, index); if (ut->chunkOffset==0 && !ut->pFuncs->access(ut, index, false)) { // no chunk available here return U_SENTINEL;
}
}
// // Simple case with no surrogates. //
ut->chunkOffset--;
cPrev = ut->chunkContents[ut->chunkOffset];
if (U16_IS_SURROGATE(cPrev)) { // Possible supplementary. Many edge cases. // Let other functions do the heavy lifting.
utext_setNativeIndex(ut, index);
cPrev = utext_previous32(ut);
} return cPrev;
}
if (a->pFuncs != b->pFuncs) { // Different types of text providers. returnfalse;
}
if (a->context != b->context) { // Different sources (different strings) returnfalse;
} if (utext_getNativeIndex(a) != utext_getNativeIndex(b)) { // Different current position in the string. returnfalse;
}
//------------------------------------------------------------------------------ // // UText common functions implementation // //------------------------------------------------------------------------------
// // UText.flags bit definitions // enum {
UTEXT_HEAP_ALLOCATED = 1, // 1 if ICU has allocated this UText struct on the heap. // 0 if caller provided storage for the UText.
UTEXT_EXTRA_HEAP_ALLOCATED = 2, // 1 if ICU has allocated extra storage as a separate // heap block. // 0 if there is no separate allocation. Either no extra // storage was requested, or it is appended to the end // of the main UText storage.
UTEXT_OPEN = 4 // 1 if this UText is currently open // 0 if this UText is not open.
};
// // Extended form of a UText. The purpose is to aid in computing the total size required // when a provider asks for a UText to be allocated with extra storage.
if (ut == nullptr) { // We need to heap-allocate storage for the new UText
int32_t spaceRequired = sizeof(UText); if (extraSpace > 0) {
spaceRequired = sizeof(ExtendedUText) + extraSpace - sizeof(std::max_align_t);
}
ut = (UText *)uprv_malloc(spaceRequired); if (ut == nullptr) {
*status = U_MEMORY_ALLOCATION_ERROR; return nullptr;
} else {
*ut = emptyText;
ut->flags |= UTEXT_HEAP_ALLOCATED; if (spaceRequired>0) {
ut->extraSize = extraSpace;
ut->pExtra = &((ExtendedUText *)ut)->extension;
}
}
} else { // We have been supplied with an already existing UText. // Verify that it really appears to be a UText. if (ut->magic != UTEXT_MAGIC) {
*status = U_ILLEGAL_ARGUMENT_ERROR; return ut;
} // If the ut is already open and there's a provider supplied close // function, call it. if ((ut->flags & UTEXT_OPEN) && ut->pFuncs->close != nullptr) {
ut->pFuncs->close(ut);
}
ut->flags &= ~UTEXT_OPEN;
// If extra space was requested by our caller, check whether // sufficient already exists, and allocate new if needed. if (extraSpace > ut->extraSize) { // Need more space. If there is existing separately allocated space, // delete it first, then allocate new space. if (ut->flags & UTEXT_EXTRA_HEAP_ALLOCATED) {
uprv_free(ut->pExtra);
ut->extraSize = 0;
}
ut->pExtra = uprv_malloc(extraSpace); if (ut->pExtra == nullptr) {
*status = U_MEMORY_ALLOCATION_ERROR;
} else {
ut->extraSize = extraSpace;
ut->flags |= UTEXT_EXTRA_HEAP_ALLOCATED;
}
}
} if (U_SUCCESS(*status)) {
ut->flags |= UTEXT_OPEN;
U_CAPI UText * U_EXPORT2
utext_close(UText *ut) { if (ut==nullptr ||
ut->magic != UTEXT_MAGIC ||
(ut->flags & UTEXT_OPEN) == 0)
{ // The supplied ut is not an open UText. // Do nothing. return ut;
}
// If the provider gave us a close function, call it now. // This will clean up anything allocated specifically by the provider. if (ut->pFuncs->close != nullptr) {
ut->pFuncs->close(ut);
}
ut->flags &= ~UTEXT_OPEN;
// If we (the framework) allocated the UText or subsidiary storage, // delete it. if (ut->flags & UTEXT_EXTRA_HEAP_ALLOCATED) {
uprv_free(ut->pExtra);
ut->pExtra = nullptr;
ut->flags &= ~UTEXT_EXTRA_HEAP_ALLOCATED;
ut->extraSize = 0;
}
// Zero out function table of the closed UText. This is a defensive move, // intended to cause applications that inadvertently use a closed // utext to crash with null pointer errors.
ut->pFuncs = nullptr;
if (ut->flags & UTEXT_HEAP_ALLOCATED) { // This UText was allocated by UText setup. We need to free it. // Clear magic, so we can detect if the user messes up and immediately // tries to reopen another UText using the deleted storage.
ut->magic = 0;
uprv_free(ut);
ut = nullptr;
} return ut;
}
// // invalidateChunk Reset a chunk to have no contents, so that the next call // to access will cause new data to load. // This is needed when copy/move/replace operate directly on the // backing text, potentially putting it out of sync with the // contents in the chunk. // staticvoid
invalidateChunk(UText *ut) {
ut->chunkLength = 0;
ut->chunkNativeLimit = 0;
ut->chunkNativeStart = 0;
ut->chunkOffset = 0;
ut->nativeIndexingLimit = 0;
}
// // pinIndex Do range pinning on a native index parameter. // 64 bit pinning is done in place. // 32 bit truncated result is returned as a convenience for // use in providers that don't need 64 bits. static int32_t
pinIndex(int64_t &index, int64_t limit) { if (index<0) {
index = 0;
} elseif (index > limit) {
index = limit;
} returnstatic_cast<int32_t>(index);
}
U_CDECL_BEGIN
// // Pointer relocation function, // a utility used by shallow clone. // Adjust a pointer that refers to something within one UText (the source) // to refer to the same relative offset within a another UText (the target) // staticvoid adjustPointer(UText *dest, constvoid **destPtr, const UText *src) { // convert all pointers to (char *) so that byte address arithmetic will work. char *dptr = (char *)*destPtr; char *dUText = (char *)dest; char *sUText = (char *)src;
if (dptr >= (char *)src->pExtra && dptr < ((char*)src->pExtra)+src->extraSize) { // target ptr was to something within the src UText's pExtra storage. // relocate it into the target UText's pExtra region.
*destPtr = ((char *)dest->pExtra) + (dptr - (char *)src->pExtra);
} elseif (dptr>=sUText && dptr < sUText+src->sizeOfStruct) { // target ptr was pointing to somewhere within the source UText itself. // Move it to the same offset within the target UText.
*destPtr = dUText + (dptr-sUText);
}
}
// // Clone. This is a generic copy-the-utext-by-value clone function that can be // used as-is with some utext types, and as a helper by other clones. // static UText * U_CALLCONV
shallowTextClone(UText * dest, const UText * src, UErrorCode * status) { if (U_FAILURE(*status)) { return nullptr;
}
int32_t srcExtraSize = src->extraSize;
// // Use the generic text_setup to allocate storage if required. //
dest = utext_setup(dest, srcExtraSize, status); if (U_FAILURE(*status)) { return dest;
}
// // flags (how the UText was allocated) and the pointer to the // extra storage must retain the values in the cloned utext that // were set up by utext_setup. Save them separately before // copying the whole struct. // void *destExtra = dest->pExtra;
int32_t flags = dest->flags;
// // Copy the whole UText struct by value. // Any "Extra" storage is copied also. // int sizeToCopy = src->sizeOfStruct; if (sizeToCopy > dest->sizeOfStruct) {
sizeToCopy = dest->sizeOfStruct;
}
uprv_memcpy(dest, src, sizeToCopy);
dest->pExtra = destExtra;
dest->flags = flags; if (srcExtraSize > 0) {
uprv_memcpy(dest->pExtra, src->pExtra, srcExtraSize);
}
// // Relocate any pointers in the target that refer to the UText itself // to point to the cloned copy rather than the original source. //
adjustPointer(dest, &dest->context, src);
adjustPointer(dest, &dest->p, src);
adjustPointer(dest, &dest->q, src);
adjustPointer(dest, &dest->r, src);
adjustPointer(dest, (constvoid **)&dest->chunkContents, src);
// The newly shallow-cloned UText does _not_ own the underlying storage for the text. // (The source for the clone may or may not have owned the text.)
//------------------------------------------------------------------------------ // // UText implementation for UTF-8 char * strings (read-only) // Limitation: string length must be <= 0x7fffffff in length. // (length must for in an int32_t variable) // // Use of UText data members: // context pointer to UTF-8 string // utext.b is the input string length (bytes). // utext.c Length scanned so far in string // (for optimizing finding length of zero terminated strings.) // utext.p pointer to the current buffer // utext.q pointer to the other buffer. // //------------------------------------------------------------------------------
// Chunk size. // Must be less than 85 (256/3), because of byte mapping from char16_t indexes to native indexes. // Worst case is three native bytes to one char16_t. (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 };
// // UTF8Buf Two of these structs will be set up in the UText's extra allocated space. // Each contains the char16_t chunk buffer, the to and from native maps, and // header info. // // because backwards iteration fills the buffers starting at the end and // working towards the front, the filled part of the buffers may not begin // at the start of the available storage for the buffers. // // Buffer size is one bigger than the specified UTF8_TEXT_CHUNK_SIZE to allow for // the last character added being a supplementary, and thus requiring a surrogate // pair. Doing this is simpler than checking for the edge case. //
struct UTF8Buf {
int32_t bufNativeStart; // Native index of first char in char16_t buf
int32_t bufNativeLimit; // Native index following last char in buf.
int32_t bufStartIdx; // First filled position in buf.
int32_t bufLimitIdx; // Limit of filled range in buf.
int32_t bufNILimit; // Limit of native indexing part of buf
int32_t toUCharsMapStart; // Native index corresponding to // mapToUChars[0]. // Set to bufNativeStart when filling forwards. // Set to computed value when filling backwards.
char16_t buf[UTF8_TEXT_CHUNK_SIZE+4]; // The char16_t buffer. Requires one extra position beyond the // the chunk size, to allow for surrogate at the end. // Length must be identical to mapToNative array, below, // because of the way indexing works when the array is // filled backwards during a reverse iteration. Thus, // the additional extra size.
uint8_t mapToNative[UTF8_TEXT_CHUNK_SIZE+4]; // map char16_t index in buf to // native offset from bufNativeStart. // Requires two extra slots, // one for a supplementary starting in the last normal position, // and one for an entry for the buffer limit position.
uint8_t mapToUChars[UTF8_TEXT_CHUNK_SIZE*3+6]; // Map native offset from bufNativeStart to // corresponding offset in filled part of buf.
int32_t align;
};
U_CDECL_BEGIN
// // utf8TextLength // // Get the length of the string. If we don't already know it, // we'll need to scan for the trailing nul. // static int64_t U_CALLCONV
utf8TextLength(UText *ut) { if (ut->b < 0) { // Zero terminated string, and we haven't scanned to the end yet. // Scan it now. constchar *r = (constchar *)ut->context + ut->c; while (*r != 0) {
r++;
} if ((r - (constchar *)ut->context) < 0x7fffffff) {
ut->b = (int32_t)(r - (constchar *)ut->context);
} else { // Actual string was bigger (more than 2 gig) than we // can handle. Clip it to 2 GB.
ut->b = 0x7fffffff;
}
ut->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
} return ut->b;
}
static UBool U_CALLCONV
utf8TextAccess(UText *ut, int64_t index, UBool forward) { // // Apologies to those who are allergic to goto statements. // Consider each goto to a labelled block to be the equivalent of // call the named block as if it were a function(); // return; // const uint8_t *s8=(const uint8_t *)ut->context;
UTF8Buf *u8b = nullptr;
int32_t length = ut->b; // Length of original utf-8
int32_t ix= (int32_t)index; // Requested index, trimmed to 32 bits.
int32_t mapIndex = 0; if (index<0) {
ix=0;
} elseif (index > 0x7fffffff) { // Strings with 64 bit lengths not supported by this UTF-8 provider.
ix = 0x7fffffff;
}
// Pin requested index to the string length. if (ix>length) { if (length>=0) {
ix=length;
} elseif (ix>=ut->c) { // Zero terminated string, and requested index is beyond // the region that has already been scanned. // Scan up to either the end of the string or to the // requested position, whichever comes first. while (ut->c<ix && s8[ut->c]!=0) {
ut->c++;
} // TODO: support for null terminated string length > 32 bits. if (s8[ut->c] == 0) { // We just found the actual length of the string. // Trim the requested index back to that.
ix = ut->c;
ut->b = ut->c;
length = ut->c;
ut->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
}
}
}
// // Dispatch to the appropriate action for a forward iteration request. // if (forward) { if (ix==ut->chunkNativeLimit) { // Check for normal sequential iteration cases first. if (ix==length) { // Just reached end of string // Don't swap buffers, but do set the // current buffer position.
ut->chunkOffset = ut->chunkLength; returnfalse;
} else { // End of current buffer. // check whether other buffer already has what we need.
UTF8Buf *altB = (UTF8Buf *)ut->q; if (ix>=altB->bufNativeStart && ix<altB->bufNativeLimit) { goto swapBuffers;
}
}
}
// A random access. Desired index could be in either or niether buf. // For optimizing the order of testing, first check for the index // being in the other buffer. This will be the case for uses that // move back and forth over a fairly limited range
{
u8b = (UTF8Buf *)ut->q; // the alternate buffer if (ix>=u8b->bufNativeStart && ix<u8b->bufNativeLimit) { // Requested index is in the other buffer. goto swapBuffers;
} if (ix == length) { // Requested index is end-of-string. // (this is the case of randomly seeking to the end. // The case of iterating off the end is handled earlier.) if (ix == ut->chunkNativeLimit) { // Current buffer extends up to the end of the string. // Leave it as the current buffer.
ut->chunkOffset = ut->chunkLength; returnfalse;
} if (ix == u8b->bufNativeLimit) { // Alternate buffer extends to the end of string. // Swap it in as the current buffer. goto swapBuffersAndFail;
}
// Neither existing buffer extends to the end of the string. goto makeStubBuffer;
}
if (ix<ut->chunkNativeStart || ix>=ut->chunkNativeLimit) { // Requested index is in neither buffer. goto fillForward;
}
// 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; returntrue;
}
}
// // Dispatch to the appropriate action for a // Backwards Direction iteration request. // if (ix==ut->chunkNativeStart) { // Check for normal sequential iteration cases first. if (ix==0) { // Just reached the start of string // Don't swap buffers, but do set the // current buffer position.
ut->chunkOffset = 0; returnfalse;
} else { // Start of current buffer. // check whether other buffer already has what we need.
UTF8Buf *altB = (UTF8Buf *)ut->q; if (ix>altB->bufNativeStart && ix<=altB->bufNativeLimit) { goto swapBuffers;
}
}
}
// A random access. Desired index could be in either or niether buf. // For optimizing the order of testing, // Most likely case: in the other buffer. // Second most likely: in neither buffer. // Unlikely, but must work: in the current buffer.
u8b = (UTF8Buf *)ut->q; // the alternate buffer if (ix>u8b->bufNativeStart && ix<=u8b->bufNativeLimit) { // Requested index is in the other buffer. goto swapBuffers;
} // Requested index is start-of-string. // (this is the case of randomly seeking to the start. // The case of iterating off the start is handled earlier.) if (ix==0) { if (u8b->bufNativeStart==0) { // Alternate buffer contains the data for the start string. // Make it be the current buffer. goto swapBuffersAndFail;
} else { // Request for data before the start of string, // neither buffer is usable. // set up a zero-length buffer. goto makeStubBuffer;
}
}
if (ix<=ut->chunkNativeStart || ix>ut->chunkNativeLimit) { // Requested index is in neither buffer. goto fillReverse;
}
// Requested index is in this buffer. // Set the utf16 buffer index.
u8b = (UTF8Buf *)ut->p;
mapIndex = ix - u8b->toUCharsMapStart;
ut->chunkOffset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx; if (ut->chunkOffset==0) { // This occurs when the first character in the text is // a multi-byte UTF-8 char, and the requested index is to // one of the trailing bytes. Because there is no preceding , // character, this access fails. We can't pick up on the // situation sooner because the requested index is not zero. returnfalse;
} else { returntrue;
}
swapBuffers: // The alternate buffer (ut->q) has the string data that was requested. // Swap the primary and alternate buffers, and set the // chunk index into the new primary buffer.
{
u8b = (UTF8Buf *)ut->q;
ut->q = ut->p;
ut->p = u8b;
ut->chunkContents = &u8b->buf[u8b->bufStartIdx];
ut->chunkLength = u8b->bufLimitIdx - u8b->bufStartIdx;
ut->chunkNativeStart = u8b->bufNativeStart;
ut->chunkNativeLimit = u8b->bufNativeLimit;
ut->nativeIndexingLimit = u8b->bufNILimit;
// Index into the (now current) chunk // Use the map to set the chunk index. It's more trouble than it's worth // to check whether native indexing can be used.
U_ASSERT(ix>=u8b->bufNativeStart);
U_ASSERT(ix<=u8b->bufNativeLimit);
mapIndex = ix - u8b->toUCharsMapStart;
U_ASSERT(mapIndex>=0);
U_ASSERT(mapIndex<(int32_t)sizeof(u8b->mapToUChars));
ut->chunkOffset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;
returntrue;
}
swapBuffersAndFail: // We got a request for either the start or end of the string, // with iteration continuing in the out-of-bounds direction. // The alternate buffer already contains the data up to the // start/end. // Swap the buffers, then return failure, indicating that we couldn't // make things correct for continuing the iteration in the requested // direction. The position & buffer are correct should the // user decide to iterate in the opposite direction.
u8b = (UTF8Buf *)ut->q;
ut->q = ut->p;
ut->p = u8b;
ut->chunkContents = &u8b->buf[u8b->bufStartIdx];
ut->chunkLength = u8b->bufLimitIdx - u8b->bufStartIdx;
ut->chunkNativeStart = u8b->bufNativeStart;
ut->chunkNativeLimit = u8b->bufNativeLimit;
ut->nativeIndexingLimit = u8b->bufNILimit;
// Index into the (now current) chunk // For this function (swapBuffersAndFail), the requested index // will always be at either the start or end of the chunk. if (ix==u8b->bufNativeLimit) {
ut->chunkOffset = ut->chunkLength;
} else {
ut->chunkOffset = 0;
U_ASSERT(ix == u8b->bufNativeStart);
} returnfalse;
makeStubBuffer: // The user has done a seek/access past the start or end // of the string. Rather than loading data that is likely // to never be used, just set up a zero-length buffer at // the position.
u8b = (UTF8Buf *)ut->q;
u8b->bufNativeStart = ix;
u8b->bufNativeLimit = ix;
u8b->bufStartIdx = 0;
u8b->bufLimitIdx = 0;
u8b->bufNILimit = 0;
u8b->toUCharsMapStart = ix;
u8b->mapToNative[0] = 0;
u8b->mapToUChars[0] = 0; goto swapBuffersAndFail;
fillForward:
{ // Move the incoming index to a code point boundary.
U8_SET_CP_START(s8, 0, ix);
// Swap the UText buffers. // We want to fill what was previously the alternate buffer, // and make what was the current buffer be the new alternate.
UTF8Buf *u8b_swap = (UTF8Buf *)ut->q;
ut->q = ut->p;
ut->p = u8b_swap;
// Fill the chunk buffer and mapping arrays. while (destIx<UTF8_TEXT_CHUNK_SIZE) {
c = s8[srcIx]; if (c>0 && c<0x80) { // Special case ASCII range for speed. // zero is excluded to simplify bounds checking.
buf[destIx] = (char16_t)c;
mapToNative[destIx] = (uint8_t)(srcIx - ix);
mapToUChars[srcIx-ix] = (uint8_t)destIx;
srcIx++;
destIx++;
} else { // General case, handle everything. if (seenNonAscii == false) {
seenNonAscii = true;
u8b_swap->bufNILimit = destIx;
}
U16_APPEND_UNSAFE(buf, destIx, c); do {
mapToNative[dIx++] = (uint8_t)(cIx - ix);
} while (dIx < destIx);
do {
mapToUChars[cIx++ - ix] = (uint8_t)dIxSaved;
} while (cIx < srcIx);
} if (srcIx>=strLen) { break;
}
}
// store Native <--> Chunk Map entries for the end of the buffer. // There is no actual character here, but the index position is valid.
mapToNative[destIx] = (uint8_t)(srcIx - ix);
mapToUChars[srcIx - ix] = (uint8_t)destIx;
// fill in Buffer descriptor
u8b_swap->bufNativeStart = ix;
u8b_swap->bufNativeLimit = srcIx;
u8b_swap->bufStartIdx = 0;
u8b_swap->bufLimitIdx = destIx; if (seenNonAscii == false) {
u8b_swap->bufNILimit = destIx;
}
u8b_swap->toUCharsMapStart = u8b_swap->bufNativeStart;
// Set UText chunk to refer to this buffer.
ut->chunkContents = buf;
ut->chunkOffset = 0;
ut->chunkLength = u8b_swap->bufLimitIdx;
ut->chunkNativeStart = u8b_swap->bufNativeStart;
ut->chunkNativeLimit = u8b_swap->bufNativeLimit;
ut->nativeIndexingLimit = u8b_swap->bufNILimit;
// For zero terminated strings, keep track of the maximum point // scanned so far. if (nulTerminated && srcIx>ut->c) {
ut->c = srcIx; if (c==0) { // We scanned to the end. // Remember the actual length.
ut->b = srcIx;
ut->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
}
} returntrue;
}
fillReverse:
{ // Move the incoming index to a code point boundary. // 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);
}
// Swap the UText buffers. // We want to fill what was previously the alternate buffer, // and make what was the current buffer be the new alternate.
UTF8Buf *u8b_swap = (UTF8Buf *)ut->q;
ut->q = ut->p;
ut->p = u8b_swap;
char16_t *buf = u8b_swap->buf;
uint8_t *mapToNative = u8b_swap->mapToNative;
uint8_t *mapToUChars = u8b_swap->mapToUChars;
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 // buffer start.
int32_t srcIx = ix;
int32_t bufNILimit = destIx;
UChar32 c;
// Map to/from Native Indexes, fill in for the position at the end of // the buffer. //
mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
mapToUChars[srcIx - toUCharsMapStart] = (uint8_t)destIx;
// Fill the chunk buffer // Work backwards, filling from the end of the buffer towards the front. // while (destIx>2 && (srcIx - toUCharsMapStart > 5) && (srcIx > 0)) {
srcIx--;
destIx--;
// Get last byte of the UTF-8 character
c = s8[srcIx]; if (c<0x80) { // Special case ASCII range for speed.
buf[destIx] = (char16_t)c;
U_ASSERT(toUCharsMapStart <= srcIx);
mapToUChars[srcIx - toUCharsMapStart] = (uint8_t)destIx;
mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
} else { // General case, handle everything non-ASCII.
int32_t sIx = srcIx; // ix of last byte of multi-byte u8 char
// Get the full character from the UTF8 string. // use code derived from the macros in utf8.h // Leaves srcIx pointing at the first byte of the UTF-8 char. //
c=utf8_prevCharSafeBody(s8, 0, &srcIx, c, -3); // leaves srcIx at first byte of the multi-byte char.
// Store the character in UTF-16 buffer. if (c<0x10000) {
buf[destIx] = (char16_t)c;
mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
} else {
buf[destIx] = U16_TRAIL(c);
mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
buf[--destIx] = U16_LEAD(c);
mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
}
// Fill in the map from native indexes to UChars buf index. 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; // the limit will be here if the rest of the chars to be // added to this buffer are ascii.
bufNILimit = destIx;
}
}
u8b_swap->bufNativeStart = srcIx;
u8b_swap->bufNativeLimit = ix;
u8b_swap->bufStartIdx = destIx;
u8b_swap->bufLimitIdx = UTF8_TEXT_CHUNK_SIZE+2;
u8b_swap->bufNILimit = bufNILimit - u8b_swap->bufStartIdx;
u8b_swap->toUCharsMapStart = toUCharsMapStart;
// adjust the incoming indexes to land on code point boundaries if needed. // adjust by no more than three, because that is the largest number of trail bytes // in a well formed UTF8 character. const uint8_t *buf = (const uint8_t *)ut->context; int i; if (start32 < ut->chunkNativeLimit) { for (i=0; i<3; i++) { if (U8_IS_SINGLE(buf[start32]) || U8_IS_LEAD(buf[start32]) || start32==0) { break;
}
start32--;
}
}
if (limit32 < ut->chunkNativeLimit) { for (i=0; i<3; i++) { if (U8_IS_SINGLE(buf[limit32]) || U8_IS_LEAD(buf[limit32]) || limit32==0) { break;
}
limit32--;
}
}
// Do the actual extract.
int32_t destLength=0;
utext_strFromUTF8(dest, destCapacity, &destLength,
(constchar *)ut->context+start32, limit32-start32,
pErrorCode);
utf8TextAccess(ut, limit32, true); return destLength;
}
// // Map a native index to the corresponding chunk offset // static int32_t U_CALLCONV
utf8TextMapIndexToUTF16(const UText *ut, int64_t index64) {
U_ASSERT(index64 <= 0x7fffffff);
int32_t index = (int32_t)index64;
UTF8Buf *u8b = (UTF8Buf *)ut->p;
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;
}
static UText * U_CALLCONV
utf8TextClone(UText *dest, const UText *src, UBool deep, UErrorCode *status)
{ // First do a generic shallow clone. Does everything needed for the UText struct itself.
dest = shallowTextClone(dest, src, status);
// For deep clones, make a copy of the string. // The copied storage is owned by the newly created clone. // // TODO: There is an issue with using utext_nativeLength(). // That function is non-const in cases where the input was NUL terminated // and the length has not yet been determined. // This function (clone()) is const. // There potentially a thread safety issue lurking here. // if (deep && U_SUCCESS(*status)) {
int32_t len = (int32_t)utext_nativeLength((UText *)src); char *copyStr = (char *)uprv_malloc(len+1); if (copyStr == nullptr) {
*status = U_MEMORY_ALLOCATION_ERROR;
} else {
uprv_memcpy(copyStr, src->context, len+1);
dest->context = copyStr;
dest->providerProperties |= I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT);
}
} return dest;
}
staticvoid U_CALLCONV
utf8TextClose(UText *ut) { // Most of the work of close is done by the generic UText framework close. // All that needs to be done here is to delete the UTF8 string if the UText // owns it. This occurs if the UText was created by cloning. if (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT)) { char *s = (char *)ut->context;
uprv_free(s);
ut->context = nullptr;
}
}
//------------------------------------------------------------------------------ // // UText implementation wrapper for Replaceable (read/write) // // Use of UText data members: // context pointer to Replaceable. // p pointer to Replaceable if it is owned by the UText. // //------------------------------------------------------------------------------
// minimum chunk size for this implementation: 3 // to allow for possible trimming for code point boundaries enum { REP_TEXT_CHUNK_SIZE=10 };
struct ReplExtra { /* * Chunk UChars. * +1 to simplify filling with surrogate pair at the end.
*/
char16_t s[REP_TEXT_CHUNK_SIZE+1];
};
U_CDECL_BEGIN
static UText * U_CALLCONV
repTextClone(UText *dest, const UText *src, UBool deep, UErrorCode *status) { // First do a generic shallow clone. Does everything needed for the UText struct itself.
dest = shallowTextClone(dest, src, status);
// For deep clones, make a copy of the Replaceable. // The copied Replaceable storage is owned by the newly created UText clone. // A non-nullptr pointer in UText.p is the signal to the close() function to delete // it. // if (deep && U_SUCCESS(*status)) { const Replaceable *replSrc = (const Replaceable *)src->context;
dest->context = replSrc->clone();
dest->providerProperties |= I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT);
// with deep clone, the copy is writable, even when the source is not.
dest->providerProperties |= I32_FLAG(UTEXT_PROVIDER_WRITABLE);
} return dest;
}
staticvoid U_CALLCONV
repTextClose(UText *ut) { // Most of the work of close is done by the generic UText framework close. // All that needs to be done here is delete the Replaceable if the UText // owns it. This occurs if the UText was created by cloning. if (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT)) {
Replaceable *rep = (Replaceable *)ut->context; delete rep;
ut->context = nullptr;
}
}
static UBool U_CALLCONV
repTextAccess(UText *ut, int64_t index, UBool forward) { const Replaceable *rep=(const Replaceable *)ut->context;
int32_t length=rep->length(); // Full length of the input text (bigger than a chunk)
// clip the requested index to the limits of the text.
int32_t index32 = pinIndex(index, length);
U_ASSERT(index<=INT32_MAX);
/* * Compute start/limit boundaries around index, for a segment of text * to be extracted. * To allow for the possibility that our user gave an index to the trailing * half of a surrogate pair, we must request one extra preceding char16_t when * going in the forward direction. This will ensure that the buffer has the * entire code point at the specified index.
*/ if(forward) {
if (index32>=ut->chunkNativeStart && index32<ut->chunkNativeLimit) { // Buffer already contains the requested position.
ut->chunkOffset = (int32_t)(index - ut->chunkNativeStart); returntrue;
} if (index32>=length && ut->chunkNativeLimit==length) { // Request for end of string, and buffer already extends up to it. // Can't get the data, but don't change the buffer.
ut->chunkOffset = length - (int32_t)ut->chunkNativeStart; returnfalse;
}
ut->chunkNativeLimit = index + REP_TEXT_CHUNK_SIZE - 1; // Going forward, so we want to have the buffer with stuff at and beyond // the requested index. The -1 gets us one code point before the // requested index also, to handle the case of the index being on // a trail surrogate of a surrogate pair. if(ut->chunkNativeLimit > length) {
ut->chunkNativeLimit = length;
} // unless buffer ran off end, start is index-1.
ut->chunkNativeStart = ut->chunkNativeLimit - REP_TEXT_CHUNK_SIZE; if(ut->chunkNativeStart < 0) {
ut->chunkNativeStart = 0;
}
} else { // Reverse iteration. Fill buffer with data preceding the requested index. if (index32>ut->chunkNativeStart && index32<=ut->chunkNativeLimit) { // Requested position already in buffer.
ut->chunkOffset = index32 - (int32_t)ut->chunkNativeStart; returntrue;
} if (index32==0 && ut->chunkNativeStart==0) { // Request for start, buffer already begins at start. // No data, but keep the buffer as is.
ut->chunkOffset = 0; returnfalse;
}
// Figure out the bounds of the chunk to extract for reverse iteration. // Need to worry about chunk not splitting surrogate pairs, and while still // containing the data we need. // Fix by requesting a chunk that includes an extra char16_t at the end. // If this turns out to be a lead surrogate, we can lop it off and still have // the data we wanted.
ut->chunkNativeStart = index32 + 1 - REP_TEXT_CHUNK_SIZE; if (ut->chunkNativeStart < 0) {
ut->chunkNativeStart = 0;
}
// Extract the new chunk of text from the Replaceable source.
ReplExtra *ex = (ReplExtra *)ut->pExtra; // UnicodeString with its buffer a writable alias to the chunk buffer
UnicodeString buffer(ex->s, 0 /*buffer length*/, REP_TEXT_CHUNK_SIZE /*buffer capacity*/);
rep->extractBetween((int32_t)ut->chunkNativeStart, (int32_t)ut->chunkNativeLimit, buffer);
// Surrogate pairs from the input text must not span chunk boundaries. // If end of chunk could be the start of a surrogate, trim it off. if (ut->chunkNativeLimit < length &&
U16_IS_LEAD(ex->s[ut->chunkLength-1])) {
ut->chunkLength--;
ut->chunkNativeLimit--; if (ut->chunkOffset > ut->chunkLength) {
ut->chunkOffset = ut->chunkLength;
}
}
// if the first char16_t in the chunk could be the trailing half of a surrogate pair, // trim it off. if(ut->chunkNativeStart>0 && U16_IS_TRAIL(ex->s[0])) {
++(ut->chunkContents);
++(ut->chunkNativeStart);
--(ut->chunkLength);
--(ut->chunkOffset);
}
// adjust the index/chunkOffset to a code point boundary
U16_SET_CP_START(ut->chunkContents, 0, ut->chunkOffset);
// Use fast indexing for get/setNativeIndex()
ut->nativeIndexingLimit = ut->chunkLength;
Die Informationen auf dieser Webseite wurden
nach bestem Wissen sorgfältig zusammengestellt. Es wird jedoch weder Vollständigkeit, noch Richtigkeit,
noch Qualität der bereit gestellten Informationen zugesichert.
Bemerkung:
Die farbliche Syntaxdarstellung und die Messung sind noch experimentell.