michael@0: /*
michael@0: *******************************************************************************
michael@0: *
michael@0: *   Copyright (C) 2005-2012, International Business Machines
michael@0: *   Corporation and others.  All Rights Reserved.
michael@0: *
michael@0: *******************************************************************************
michael@0: *   file name:  utext.cpp
michael@0: *   encoding:   US-ASCII
michael@0: *   tab size:   8 (not used)
michael@0: *   indentation:4
michael@0: *
michael@0: *   created on: 2005apr12
michael@0: *   created by: Markus W. Scherer
michael@0: */
michael@0: 
michael@0: #include "unicode/utypes.h"
michael@0: #include "unicode/ustring.h"
michael@0: #include "unicode/unistr.h"
michael@0: #include "unicode/chariter.h"
michael@0: #include "unicode/utext.h"
michael@0: #include "unicode/utf.h"
michael@0: #include "unicode/utf8.h"
michael@0: #include "unicode/utf16.h"
michael@0: #include "ustr_imp.h"
michael@0: #include "cmemory.h"
michael@0: #include "cstring.h"
michael@0: #include "uassert.h"
michael@0: #include "putilimp.h"
michael@0: 
michael@0: U_NAMESPACE_USE
michael@0: 
michael@0: #define I32_FLAG(bitIndex) ((int32_t)1<<(bitIndex))
michael@0: 
michael@0: 
michael@0: static UBool
michael@0: utext_access(UText *ut, int64_t index, UBool forward) {
michael@0:     return ut->pFuncs->access(ut, index, forward);
michael@0: }
michael@0: 
michael@0: 
michael@0: 
michael@0: U_CAPI UBool U_EXPORT2
michael@0: utext_moveIndex32(UText *ut, int32_t delta) {
michael@0:     UChar32  c;
michael@0:     if (delta > 0) {
michael@0:         do {
michael@0:             if(ut->chunkOffset>=ut->chunkLength && !utext_access(ut, ut->chunkNativeLimit, TRUE)) {
michael@0:                 return FALSE;
michael@0:             }
michael@0:             c = ut->chunkContents[ut->chunkOffset];
michael@0:             if (U16_IS_SURROGATE(c)) {
michael@0:                 c = utext_next32(ut);
michael@0:                 if (c == U_SENTINEL) {
michael@0:                     return FALSE;
michael@0:                 }
michael@0:             } else {
michael@0:                 ut->chunkOffset++;
michael@0:             }
michael@0:         } while(--delta>0);
michael@0: 
michael@0:     } else if (delta<0) {
michael@0:         do {
michael@0:             if(ut->chunkOffset<=0 && !utext_access(ut, ut->chunkNativeStart, FALSE)) {
michael@0:                 return FALSE;
michael@0:             }
michael@0:             c = ut->chunkContents[ut->chunkOffset-1];
michael@0:             if (U16_IS_SURROGATE(c)) {
michael@0:                 c = utext_previous32(ut);
michael@0:                 if (c == U_SENTINEL) {
michael@0:                     return FALSE;
michael@0:                 }
michael@0:             } else {
michael@0:                 ut->chunkOffset--;
michael@0:             }
michael@0:         } while(++delta<0);
michael@0:     }
michael@0: 
michael@0:     return TRUE;
michael@0: }
michael@0: 
michael@0: 
michael@0: U_CAPI int64_t U_EXPORT2
michael@0: utext_nativeLength(UText *ut) {
michael@0:     return ut->pFuncs->nativeLength(ut);
michael@0: }
michael@0: 
michael@0: 
michael@0: U_CAPI UBool U_EXPORT2
michael@0: utext_isLengthExpensive(const UText *ut) {
michael@0:     UBool r = (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE)) != 0;
michael@0:     return r;
michael@0: }
michael@0: 
michael@0: 
michael@0: U_CAPI int64_t U_EXPORT2
michael@0: utext_getNativeIndex(const UText *ut) {
michael@0:     if(ut->chunkOffset <= ut->nativeIndexingLimit) {
michael@0:         return ut->chunkNativeStart+ut->chunkOffset;
michael@0:     } else {
michael@0:         return ut->pFuncs->mapOffsetToNative(ut);
michael@0:     }
michael@0: }
michael@0: 
michael@0: 
michael@0: U_CAPI void U_EXPORT2
michael@0: utext_setNativeIndex(UText *ut, int64_t index) {
michael@0:     if(index<ut->chunkNativeStart || index>=ut->chunkNativeLimit) {
michael@0:         // The desired position is outside of the current chunk.
michael@0:         // Access the new position.  Assume a forward iteration from here,
michael@0:         // which will also be optimimum for a single random access.
michael@0:         // Reverse iterations may suffer slightly.
michael@0:         ut->pFuncs->access(ut, index, TRUE);
michael@0:     } else if((int32_t)(index - ut->chunkNativeStart) <= ut->nativeIndexingLimit) {
michael@0:         // utf-16 indexing.
michael@0:         ut->chunkOffset=(int32_t)(index-ut->chunkNativeStart);
michael@0:     } else {
michael@0:          ut->chunkOffset=ut->pFuncs->mapNativeIndexToUTF16(ut, index);
michael@0:     }
michael@0:     // The convention is that the index must always be on a code point boundary.
michael@0:     // Adjust the index position if it is in the middle of a surrogate pair.
michael@0:     if (ut->chunkOffset<ut->chunkLength) {
michael@0:         UChar c= ut->chunkContents[ut->chunkOffset];
michael@0:         if (U16_IS_TRAIL(c)) {
michael@0:             if (ut->chunkOffset==0) {
michael@0:                 ut->pFuncs->access(ut, ut->chunkNativeStart, FALSE);
michael@0:             }
michael@0:             if (ut->chunkOffset>0) {
michael@0:                 UChar lead = ut->chunkContents[ut->chunkOffset-1];
michael@0:                 if (U16_IS_LEAD(lead)) {
michael@0:                     ut->chunkOffset--;
michael@0:                 }
michael@0:             }
michael@0:         }
michael@0:     }
michael@0: }
michael@0: 
michael@0: 
michael@0: 
michael@0: U_CAPI int64_t U_EXPORT2
michael@0: utext_getPreviousNativeIndex(UText *ut) {
michael@0:     //
michael@0:     //  Fast-path the common case.
michael@0:     //     Common means current position is not at the beginning of a chunk
michael@0:     //     and the preceding character is not supplementary.
michael@0:     //
michael@0:     int32_t i = ut->chunkOffset - 1;
michael@0:     int64_t result;
michael@0:     if (i >= 0) {
michael@0:         UChar c = ut->chunkContents[i];
michael@0:         if (U16_IS_TRAIL(c) == FALSE) {
michael@0:             if (i <= ut->nativeIndexingLimit) {
michael@0:                 result = ut->chunkNativeStart + i;
michael@0:             } else {
michael@0:                 ut->chunkOffset = i;
michael@0:                 result = ut->pFuncs->mapOffsetToNative(ut);
michael@0:                 ut->chunkOffset++;
michael@0:             }
michael@0:             return result;
michael@0:         }
michael@0:     }
michael@0: 
michael@0:     // If at the start of text, simply return 0.
michael@0:     if (ut->chunkOffset==0 && ut->chunkNativeStart==0) {
michael@0:         return 0;
michael@0:     }
michael@0: 
michael@0:     // Harder, less common cases.  We are at a chunk boundary, or on a surrogate.
michael@0:     //    Keep it simple, use other functions to handle the edges.
michael@0:     //
michael@0:     utext_previous32(ut);
michael@0:     result = UTEXT_GETNATIVEINDEX(ut);
michael@0:     utext_next32(ut);
michael@0:     return result;
michael@0: }
michael@0: 
michael@0: 
michael@0: //
michael@0: //  utext_current32.  Get the UChar32 at the current position.
michael@0: //                    UText iteration position is always on a code point boundary,
michael@0: //                    never on the trail half of a surrogate pair.
michael@0: //
michael@0: U_CAPI UChar32 U_EXPORT2
michael@0: utext_current32(UText *ut) {
michael@0:     UChar32  c;
michael@0:     if (ut->chunkOffset==ut->chunkLength) {
michael@0:         // Current position is just off the end of the chunk.
michael@0:         if (ut->pFuncs->access(ut, ut->chunkNativeLimit, TRUE) == FALSE) {
michael@0:             // Off the end of the text.
michael@0:             return U_SENTINEL;
michael@0:         }
michael@0:     }
michael@0: 
michael@0:     c = ut->chunkContents[ut->chunkOffset];
michael@0:     if (U16_IS_LEAD(c) == FALSE) {
michael@0:         // Normal, non-supplementary case.
michael@0:         return c;
michael@0:     }
michael@0: 
michael@0:     //
michael@0:     //  Possible supplementary char.
michael@0:     //
michael@0:     UChar32   trail = 0;
michael@0:     UChar32   supplementaryC = c;
michael@0:     if ((ut->chunkOffset+1) < ut->chunkLength) {
michael@0:         // The trail surrogate is in the same chunk.
michael@0:         trail = ut->chunkContents[ut->chunkOffset+1];
michael@0:     } else {
michael@0:         //  The trail surrogate is in a different chunk.
michael@0:         //     Because we must maintain the iteration position, we need to switch forward
michael@0:         //     into the new chunk, get the trail surrogate, then revert the chunk back to the
michael@0:         //     original one.
michael@0:         //     An edge case to be careful of:  the entire text may end with an unpaired
michael@0:         //        leading surrogate.  The attempt to access the trail will fail, but
michael@0:         //        the original position before the unpaired lead still needs to be restored.
michael@0:         int64_t  nativePosition = ut->chunkNativeLimit;
michael@0:         int32_t  originalOffset = ut->chunkOffset;
michael@0:         if (ut->pFuncs->access(ut, nativePosition, TRUE)) {
michael@0:             trail = ut->chunkContents[ut->chunkOffset];
michael@0:         }
michael@0:         UBool r = ut->pFuncs->access(ut, nativePosition, FALSE);  // reverse iteration flag loads preceding chunk
michael@0:         U_ASSERT(r==TRUE);
michael@0:         ut->chunkOffset = originalOffset;
michael@0:         if(!r) {
michael@0:             return U_SENTINEL;
michael@0:         }
michael@0:     }
michael@0: 
michael@0:     if (U16_IS_TRAIL(trail)) {
michael@0:         supplementaryC = U16_GET_SUPPLEMENTARY(c, trail);
michael@0:     }
michael@0:     return supplementaryC;
michael@0: 
michael@0: }
michael@0: 
michael@0: 
michael@0: U_CAPI UChar32 U_EXPORT2
michael@0: utext_char32At(UText *ut, int64_t nativeIndex) {
michael@0:     UChar32 c = U_SENTINEL;
michael@0: 
michael@0:     // Fast path the common case.
michael@0:     if (nativeIndex>=ut->chunkNativeStart && nativeIndex < ut->chunkNativeStart + ut->nativeIndexingLimit) {
michael@0:         ut->chunkOffset = (int32_t)(nativeIndex - ut->chunkNativeStart);
michael@0:         c = ut->chunkContents[ut->chunkOffset];
michael@0:         if (U16_IS_SURROGATE(c) == FALSE) {
michael@0:             return c;
michael@0:         }
michael@0:     }
michael@0: 
michael@0: 
michael@0:     utext_setNativeIndex(ut, nativeIndex);
michael@0:     if (nativeIndex>=ut->chunkNativeStart && ut->chunkOffset<ut->chunkLength) {
michael@0:         c = ut->chunkContents[ut->chunkOffset];
michael@0:         if (U16_IS_SURROGATE(c)) {
michael@0:             // For surrogates, let current32() deal with the complications
michael@0:             //    of supplementaries that may span chunk boundaries.
michael@0:             c = utext_current32(ut);
michael@0:         }
michael@0:     }
michael@0:     return c;
michael@0: }
michael@0: 
michael@0: 
michael@0: U_CAPI UChar32 U_EXPORT2
michael@0: utext_next32(UText *ut) {
michael@0:     UChar32       c;
michael@0: 
michael@0:     if (ut->chunkOffset >= ut->chunkLength) {
michael@0:         if (ut->pFuncs->access(ut, ut->chunkNativeLimit, TRUE) == FALSE) {
michael@0:             return U_SENTINEL;
michael@0:         }
michael@0:     }
michael@0: 
michael@0:     c = ut->chunkContents[ut->chunkOffset++];
michael@0:     if (U16_IS_LEAD(c) == FALSE) {
michael@0:         // Normal case, not supplementary.
michael@0:         //   (A trail surrogate seen here is just returned as is, as a surrogate value.
michael@0:         //    It cannot be part of a pair.)
michael@0:         return c;
michael@0:     }
michael@0: 
michael@0:     if (ut->chunkOffset >= ut->chunkLength) {
michael@0:         if (ut->pFuncs->access(ut, ut->chunkNativeLimit, TRUE) == FALSE) {
michael@0:             // c is an unpaired lead surrogate at the end of the text.
michael@0:             // return it as it is.
michael@0:             return c;
michael@0:         }
michael@0:     }
michael@0:     UChar32 trail = ut->chunkContents[ut->chunkOffset];
michael@0:     if (U16_IS_TRAIL(trail) == FALSE) {
michael@0:         // c was an unpaired lead surrogate, not at the end of the text.
michael@0:         // return it as it is (unpaired).  Iteration position is on the
michael@0:         // following character, possibly in the next chunk, where the
michael@0:         //  trail surrogate would have been if it had existed.
michael@0:         return c;
michael@0:     }
michael@0: 
michael@0:     UChar32 supplementary = U16_GET_SUPPLEMENTARY(c, trail);
michael@0:     ut->chunkOffset++;   // move iteration position over the trail surrogate.
michael@0:     return supplementary;
michael@0:     }
michael@0: 
michael@0: 
michael@0: U_CAPI UChar32 U_EXPORT2
michael@0: utext_previous32(UText *ut) {
michael@0:     UChar32       c;
michael@0: 
michael@0:     if (ut->chunkOffset <= 0) {
michael@0:         if (ut->pFuncs->access(ut, ut->chunkNativeStart, FALSE) == FALSE) {
michael@0:             return U_SENTINEL;
michael@0:         }
michael@0:     }
michael@0:     ut->chunkOffset--;
michael@0:     c = ut->chunkContents[ut->chunkOffset];
michael@0:     if (U16_IS_TRAIL(c) == FALSE) {
michael@0:         // Normal case, not supplementary.
michael@0:         //   (A lead surrogate seen here is just returned as is, as a surrogate value.
michael@0:         //    It cannot be part of a pair.)
michael@0:         return c;
michael@0:     }
michael@0: 
michael@0:     if (ut->chunkOffset <= 0) {
michael@0:         if (ut->pFuncs->access(ut, ut->chunkNativeStart, FALSE) == FALSE) {
michael@0:             // c is an unpaired trail surrogate at the start of the text.
michael@0:             // return it as it is.
michael@0:             return c;
michael@0:         }
michael@0:     }
michael@0: 
michael@0:     UChar32 lead = ut->chunkContents[ut->chunkOffset-1];
michael@0:     if (U16_IS_LEAD(lead) == FALSE) {
michael@0:         // c was an unpaired trail surrogate, not at the end of the text.
michael@0:         // return it as it is (unpaired).  Iteration position is at c
michael@0:         return c;
michael@0:     }
michael@0: 
michael@0:     UChar32 supplementary = U16_GET_SUPPLEMENTARY(lead, c);
michael@0:     ut->chunkOffset--;   // move iteration position over the lead surrogate.
michael@0:     return supplementary;
michael@0: }
michael@0: 
michael@0: 
michael@0: 
michael@0: U_CAPI UChar32 U_EXPORT2
michael@0: utext_next32From(UText *ut, int64_t index) {
michael@0:     UChar32       c      = U_SENTINEL;
michael@0: 
michael@0:     if(index<ut->chunkNativeStart || index>=ut->chunkNativeLimit) {
michael@0:         // Desired position is outside of the current chunk.
michael@0:         if(!ut->pFuncs->access(ut, index, TRUE)) {
michael@0:             // no chunk available here
michael@0:             return U_SENTINEL;
michael@0:         }
michael@0:     } else if (index - ut->chunkNativeStart  <= (int64_t)ut->nativeIndexingLimit) {
michael@0:         // Desired position is in chunk, with direct 1:1 native to UTF16 indexing
michael@0:         ut->chunkOffset = (int32_t)(index - ut->chunkNativeStart);
michael@0:     } else {
michael@0:         // Desired position is in chunk, with non-UTF16 indexing.
michael@0:         ut->chunkOffset = ut->pFuncs->mapNativeIndexToUTF16(ut, index);
michael@0:     }
michael@0: 
michael@0:     c = ut->chunkContents[ut->chunkOffset++];
michael@0:     if (U16_IS_SURROGATE(c)) {
michael@0:         // Surrogates.  Many edge cases.  Use other functions that already
michael@0:         //              deal with the problems.
michael@0:         utext_setNativeIndex(ut, index);
michael@0:         c = utext_next32(ut);
michael@0:     }
michael@0:     return c;
michael@0: }
michael@0: 
michael@0: 
michael@0: U_CAPI UChar32 U_EXPORT2
michael@0: utext_previous32From(UText *ut, int64_t index) {
michael@0:     //
michael@0:     //  Return the character preceding the specified index.
michael@0:     //  Leave the iteration position at the start of the character that was returned.
michael@0:     //
michael@0:     UChar32     cPrev;    // The character preceding cCurr, which is what we will return.
michael@0: 
michael@0:     // Address the chunk containg the position preceding the incoming index
michael@0:     // A tricky edge case:
michael@0:     //   We try to test the requested native index against the chunkNativeStart to determine
michael@0:     //    whether the character preceding the one at the index is in the current chunk.
michael@0:     //    BUT, this test can fail with UTF-8 (or any other multibyte encoding), when the
michael@0:     //    requested index is on something other than the first position of the first char.
michael@0:     //
michael@0:     if(index<=ut->chunkNativeStart || index>ut->chunkNativeLimit) {
michael@0:         // Requested native index is outside of the current chunk.
michael@0:         if(!ut->pFuncs->access(ut, index, FALSE)) {
michael@0:             // no chunk available here
michael@0:             return U_SENTINEL;
michael@0:         }
michael@0:     } else if(index - ut->chunkNativeStart <= (int64_t)ut->nativeIndexingLimit) {
michael@0:         // Direct UTF-16 indexing.
michael@0:         ut->chunkOffset = (int32_t)(index - ut->chunkNativeStart);
michael@0:     } else {
michael@0:         ut->chunkOffset=ut->pFuncs->mapNativeIndexToUTF16(ut, index);
michael@0:         if (ut->chunkOffset==0 && !ut->pFuncs->access(ut, index, FALSE)) {
michael@0:             // no chunk available here
michael@0:             return U_SENTINEL;
michael@0:         }
michael@0:     }
michael@0: 
michael@0:     //
michael@0:     // Simple case with no surrogates.
michael@0:     //
michael@0:     ut->chunkOffset--;
michael@0:     cPrev = ut->chunkContents[ut->chunkOffset];
michael@0: 
michael@0:     if (U16_IS_SURROGATE(cPrev)) {
michael@0:         // Possible supplementary.  Many edge cases.
michael@0:         // Let other functions do the heavy lifting.
michael@0:         utext_setNativeIndex(ut, index);
michael@0:         cPrev = utext_previous32(ut);
michael@0:     }
michael@0:     return cPrev;
michael@0: }
michael@0: 
michael@0: 
michael@0: U_CAPI int32_t U_EXPORT2
michael@0: utext_extract(UText *ut,
michael@0:              int64_t start, int64_t limit,
michael@0:              UChar *dest, int32_t destCapacity,
michael@0:              UErrorCode *status) {
michael@0:                  return ut->pFuncs->extract(ut, start, limit, dest, destCapacity, status);
michael@0:              }
michael@0: 
michael@0: 
michael@0: 
michael@0: U_CAPI UBool U_EXPORT2
michael@0: utext_equals(const UText *a, const UText *b) {
michael@0:     if (a==NULL || b==NULL ||
michael@0:         a->magic != UTEXT_MAGIC ||
michael@0:         b->magic != UTEXT_MAGIC) {
michael@0:             // Null or invalid arguments don't compare equal to anything.
michael@0:             return FALSE;
michael@0:     }
michael@0: 
michael@0:     if (a->pFuncs != b->pFuncs) {
michael@0:         // Different types of text providers.
michael@0:         return FALSE;
michael@0:     }
michael@0: 
michael@0:     if (a->context != b->context) {
michael@0:         // Different sources (different strings)
michael@0:         return FALSE;
michael@0:     }
michael@0:     if (utext_getNativeIndex(a) != utext_getNativeIndex(b)) {
michael@0:         // Different current position in the string.
michael@0:         return FALSE;
michael@0:     }
michael@0: 
michael@0:     return TRUE;
michael@0: }
michael@0: 
michael@0: U_CAPI UBool U_EXPORT2
michael@0: utext_isWritable(const UText *ut)
michael@0: {
michael@0:     UBool b = (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_WRITABLE)) != 0;
michael@0:     return b;
michael@0: }
michael@0: 
michael@0: 
michael@0: U_CAPI void U_EXPORT2
michael@0: utext_freeze(UText *ut) {
michael@0:     // Zero out the WRITABLE flag.
michael@0:     ut->providerProperties &= ~(I32_FLAG(UTEXT_PROVIDER_WRITABLE));
michael@0: }
michael@0: 
michael@0: 
michael@0: U_CAPI UBool U_EXPORT2
michael@0: utext_hasMetaData(const UText *ut)
michael@0: {
michael@0:     UBool b = (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_HAS_META_DATA)) != 0;
michael@0:     return b;
michael@0: }
michael@0: 
michael@0: 
michael@0: 
michael@0: U_CAPI int32_t U_EXPORT2
michael@0: utext_replace(UText *ut,
michael@0:              int64_t nativeStart, int64_t nativeLimit,
michael@0:              const UChar *replacementText, int32_t replacementLength,
michael@0:              UErrorCode *status)
michael@0: {
michael@0:     if (U_FAILURE(*status)) {
michael@0:         return 0;
michael@0:     }
michael@0:     if ((ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_WRITABLE)) == 0) {
michael@0:         *status = U_NO_WRITE_PERMISSION;
michael@0:         return 0;
michael@0:     }
michael@0:     int32_t i = ut->pFuncs->replace(ut, nativeStart, nativeLimit, replacementText, replacementLength, status);
michael@0:     return i;
michael@0: }
michael@0: 
michael@0: U_CAPI void U_EXPORT2
michael@0: utext_copy(UText *ut,
michael@0:           int64_t nativeStart, int64_t nativeLimit,
michael@0:           int64_t destIndex,
michael@0:           UBool move,
michael@0:           UErrorCode *status)
michael@0: {
michael@0:     if (U_FAILURE(*status)) {
michael@0:         return;
michael@0:     }
michael@0:     if ((ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_WRITABLE)) == 0) {
michael@0:         *status = U_NO_WRITE_PERMISSION;
michael@0:         return;
michael@0:     }
michael@0:     ut->pFuncs->copy(ut, nativeStart, nativeLimit, destIndex, move, status);
michael@0: }
michael@0: 
michael@0: 
michael@0: 
michael@0: U_CAPI UText * U_EXPORT2
michael@0: utext_clone(UText *dest, const UText *src, UBool deep, UBool readOnly, UErrorCode *status) {
michael@0:     UText *result;
michael@0:     result = src->pFuncs->clone(dest, src, deep, status);
michael@0:     if (readOnly) {
michael@0:         utext_freeze(result);
michael@0:     }
michael@0:     return result;
michael@0: }
michael@0: 
michael@0: 
michael@0: 
michael@0: //------------------------------------------------------------------------------
michael@0: //
michael@0: //   UText common functions implementation
michael@0: //
michael@0: //------------------------------------------------------------------------------
michael@0: 
michael@0: //
michael@0: //  UText.flags bit definitions
michael@0: //
michael@0: enum {
michael@0:     UTEXT_HEAP_ALLOCATED  = 1,      //  1 if ICU has allocated this UText struct on the heap.
michael@0:                                     //  0 if caller provided storage for the UText.
michael@0: 
michael@0:     UTEXT_EXTRA_HEAP_ALLOCATED = 2, //  1 if ICU has allocated extra storage as a separate
michael@0:                                     //     heap block.
michael@0:                                     //  0 if there is no separate allocation.  Either no extra
michael@0:                                     //     storage was requested, or it is appended to the end
michael@0:                                     //     of the main UText storage.
michael@0: 
michael@0:     UTEXT_OPEN = 4                  //  1 if this UText is currently open
michael@0:                                     //  0 if this UText is not open.
michael@0: };
michael@0: 
michael@0: 
michael@0: //
michael@0: //  Extended form of a UText.  The purpose is to aid in computing the total size required
michael@0: //    when a provider asks for a UText to be allocated with extra storage.
michael@0: 
michael@0: struct ExtendedUText {
michael@0:     UText          ut;
michael@0:     UAlignedMemory extension;
michael@0: };
michael@0: 
michael@0: static const UText emptyText = UTEXT_INITIALIZER;
michael@0: 
michael@0: U_CAPI UText * U_EXPORT2
michael@0: utext_setup(UText *ut, int32_t extraSpace, UErrorCode *status) {
michael@0:     if (U_FAILURE(*status)) {
michael@0:         return ut;
michael@0:     }
michael@0: 
michael@0:     if (ut == NULL) {
michael@0:         // We need to heap-allocate storage for the new UText
michael@0:         int32_t spaceRequired = sizeof(UText);
michael@0:         if (extraSpace > 0) {
michael@0:             spaceRequired = sizeof(ExtendedUText) + extraSpace - sizeof(UAlignedMemory);
michael@0:         }
michael@0:         ut = (UText *)uprv_malloc(spaceRequired);
michael@0:         if (ut == NULL) {
michael@0:             *status = U_MEMORY_ALLOCATION_ERROR;
michael@0:             return NULL;
michael@0:         } else {
michael@0:             *ut = emptyText;
michael@0:             ut->flags |= UTEXT_HEAP_ALLOCATED;
michael@0:             if (spaceRequired>0) {
michael@0:                 ut->extraSize = extraSpace;
michael@0:                 ut->pExtra    = &((ExtendedUText *)ut)->extension;
michael@0:             }
michael@0:         }
michael@0:     } else {
michael@0:         // We have been supplied with an already existing UText.
michael@0:         // Verify that it really appears to be a UText.
michael@0:         if (ut->magic != UTEXT_MAGIC) {
michael@0:             *status = U_ILLEGAL_ARGUMENT_ERROR;
michael@0:             return ut;
michael@0:         }
michael@0:         // If the ut is already open and there's a provider supplied close
michael@0:         //   function, call it.
michael@0:         if ((ut->flags & UTEXT_OPEN) && ut->pFuncs->close != NULL)  {
michael@0:             ut->pFuncs->close(ut);
michael@0:         }
michael@0:         ut->flags &= ~UTEXT_OPEN;
michael@0: 
michael@0:         // If extra space was requested by our caller, check whether
michael@0:         //   sufficient already exists, and allocate new if needed.
michael@0:         if (extraSpace > ut->extraSize) {
michael@0:             // Need more space.  If there is existing separately allocated space,
michael@0:             //   delete it first, then allocate new space.
michael@0:             if (ut->flags & UTEXT_EXTRA_HEAP_ALLOCATED) {
michael@0:                 uprv_free(ut->pExtra);
michael@0:                 ut->extraSize = 0;
michael@0:             }
michael@0:             ut->pExtra = uprv_malloc(extraSpace);
michael@0:             if (ut->pExtra == NULL) {
michael@0:                 *status = U_MEMORY_ALLOCATION_ERROR;
michael@0:             } else {
michael@0:                 ut->extraSize = extraSpace;
michael@0:                 ut->flags |= UTEXT_EXTRA_HEAP_ALLOCATED;
michael@0:             }
michael@0:         }
michael@0:     }
michael@0:     if (U_SUCCESS(*status)) {
michael@0:         ut->flags |= UTEXT_OPEN;
michael@0: 
michael@0:         // Initialize all remaining fields of the UText.
michael@0:         //
michael@0:         ut->context             = NULL;
michael@0:         ut->chunkContents       = NULL;
michael@0:         ut->p                   = NULL;
michael@0:         ut->q                   = NULL;
michael@0:         ut->r                   = NULL;
michael@0:         ut->a                   = 0;
michael@0:         ut->b                   = 0;
michael@0:         ut->c                   = 0;
michael@0:         ut->chunkOffset         = 0;
michael@0:         ut->chunkLength         = 0;
michael@0:         ut->chunkNativeStart    = 0;
michael@0:         ut->chunkNativeLimit    = 0;
michael@0:         ut->nativeIndexingLimit = 0;
michael@0:         ut->providerProperties  = 0;
michael@0:         ut->privA               = 0;
michael@0:         ut->privB               = 0;
michael@0:         ut->privC               = 0;
michael@0:         ut->privP               = NULL;
michael@0:         if (ut->pExtra!=NULL && ut->extraSize>0)
michael@0:             uprv_memset(ut->pExtra, 0, ut->extraSize);
michael@0: 
michael@0:     }
michael@0:     return ut;
michael@0: }
michael@0: 
michael@0: 
michael@0: U_CAPI UText * U_EXPORT2
michael@0: utext_close(UText *ut) {
michael@0:     if (ut==NULL ||
michael@0:         ut->magic != UTEXT_MAGIC ||
michael@0:         (ut->flags & UTEXT_OPEN) == 0)
michael@0:     {
michael@0:         // The supplied ut is not an open UText.
michael@0:         // Do nothing.
michael@0:         return ut;
michael@0:     }
michael@0: 
michael@0:     // If the provider gave us a close function, call it now.
michael@0:     // This will clean up anything allocated specifically by the provider.
michael@0:     if (ut->pFuncs->close != NULL) {
michael@0:         ut->pFuncs->close(ut);
michael@0:     }
michael@0:     ut->flags &= ~UTEXT_OPEN;
michael@0: 
michael@0:     // If we (the framework) allocated the UText or subsidiary storage,
michael@0:     //   delete it.
michael@0:     if (ut->flags & UTEXT_EXTRA_HEAP_ALLOCATED) {
michael@0:         uprv_free(ut->pExtra);
michael@0:         ut->pExtra = NULL;
michael@0:         ut->flags &= ~UTEXT_EXTRA_HEAP_ALLOCATED;
michael@0:         ut->extraSize = 0;
michael@0:     }
michael@0: 
michael@0:     // Zero out function table of the closed UText.  This is a defensive move,
michael@0:     //   inteded to cause applications that inadvertantly use a closed
michael@0:     //   utext to crash with null pointer errors.
michael@0:     ut->pFuncs        = NULL;
michael@0: 
michael@0:     if (ut->flags & UTEXT_HEAP_ALLOCATED) {
michael@0:         // This UText was allocated by UText setup.  We need to free it.
michael@0:         // Clear magic, so we can detect if the user messes up and immediately
michael@0:         //  tries to reopen another UText using the deleted storage.
michael@0:         ut->magic = 0;
michael@0:         uprv_free(ut);
michael@0:         ut = NULL;
michael@0:     }
michael@0:     return ut;
michael@0: }
michael@0: 
michael@0: 
michael@0: 
michael@0: 
michael@0: //
michael@0: // invalidateChunk   Reset a chunk to have no contents, so that the next call
michael@0: //                   to access will cause new data to load.
michael@0: //                   This is needed when copy/move/replace operate directly on the
michael@0: //                   backing text, potentially putting it out of sync with the
michael@0: //                   contents in the chunk.
michael@0: //
michael@0: static void
michael@0: invalidateChunk(UText *ut) {
michael@0:     ut->chunkLength = 0;
michael@0:     ut->chunkNativeLimit = 0;
michael@0:     ut->chunkNativeStart = 0;
michael@0:     ut->chunkOffset = 0;
michael@0:     ut->nativeIndexingLimit = 0;
michael@0: }
michael@0: 
michael@0: //
michael@0: // pinIndex        Do range pinning on a native index parameter.
michael@0: //                 64 bit pinning is done in place.
michael@0: //                 32 bit truncated result is returned as a convenience for
michael@0: //                        use in providers that don't need 64 bits.
michael@0: static int32_t
michael@0: pinIndex(int64_t &index, int64_t limit) {
michael@0:     if (index<0) {
michael@0:         index = 0;
michael@0:     } else if (index > limit) {
michael@0:         index = limit;
michael@0:     }
michael@0:     return (int32_t)index;
michael@0: }
michael@0: 
michael@0: 
michael@0: U_CDECL_BEGIN
michael@0: 
michael@0: //
michael@0: // Pointer relocation function,
michael@0: //   a utility used by shallow clone.
michael@0: //   Adjust a pointer that refers to something within one UText (the source)
michael@0: //   to refer to the same relative offset within a another UText (the target)
michael@0: //
michael@0: static void adjustPointer(UText *dest, const void **destPtr, const UText *src) {
michael@0:     // convert all pointers to (char *) so that byte address arithmetic will work.
michael@0:     char  *dptr = (char *)*destPtr;
michael@0:     char  *dUText = (char *)dest;
michael@0:     char  *sUText = (char *)src;
michael@0: 
michael@0:     if (dptr >= (char *)src->pExtra && dptr < ((char*)src->pExtra)+src->extraSize) {
michael@0:         // target ptr was to something within the src UText's pExtra storage.
michael@0:         //   relocate it into the target UText's pExtra region.
michael@0:         *destPtr = ((char *)dest->pExtra) + (dptr - (char *)src->pExtra);
michael@0:     } else if (dptr>=sUText && dptr < sUText+src->sizeOfStruct) {
michael@0:         // target ptr was pointing to somewhere within the source UText itself.
michael@0:         //   Move it to the same offset within the target UText.
michael@0:         *destPtr = dUText + (dptr-sUText);
michael@0:     }
michael@0: }
michael@0: 
michael@0: 
michael@0: //
michael@0: //  Clone.  This is a generic copy-the-utext-by-value clone function that can be
michael@0: //          used as-is with some utext types, and as a helper by other clones.
michael@0: //
michael@0: static UText * U_CALLCONV
michael@0: shallowTextClone(UText * dest, const UText * src, UErrorCode * status) {
michael@0:     if (U_FAILURE(*status)) {
michael@0:         return NULL;
michael@0:     }
michael@0:     int32_t  srcExtraSize = src->extraSize;
michael@0: 
michael@0:     //
michael@0:     // Use the generic text_setup to allocate storage if required.
michael@0:     //
michael@0:     dest = utext_setup(dest, srcExtraSize, status);
michael@0:     if (U_FAILURE(*status)) {
michael@0:         return dest;
michael@0:     }
michael@0: 
michael@0:     //
michael@0:     //  flags (how the UText was allocated) and the pointer to the
michael@0:     //   extra storage must retain the values in the cloned utext that
michael@0:     //   were set up by utext_setup.  Save them separately before
michael@0:     //   copying the whole struct.
michael@0:     //
michael@0:     void *destExtra = dest->pExtra;
michael@0:     int32_t flags   = dest->flags;
michael@0: 
michael@0: 
michael@0:     //
michael@0:     //  Copy the whole UText struct by value.
michael@0:     //  Any "Extra" storage is copied also.
michael@0:     //
michael@0:     int sizeToCopy = src->sizeOfStruct;
michael@0:     if (sizeToCopy > dest->sizeOfStruct) {
michael@0:         sizeToCopy = dest->sizeOfStruct;
michael@0:     }
michael@0:     uprv_memcpy(dest, src, sizeToCopy);
michael@0:     dest->pExtra = destExtra;
michael@0:     dest->flags  = flags;
michael@0:     if (srcExtraSize > 0) {
michael@0:         uprv_memcpy(dest->pExtra, src->pExtra, srcExtraSize);
michael@0:     }
michael@0: 
michael@0:     //
michael@0:     // Relocate any pointers in the target that refer to the UText itself
michael@0:     //   to point to the cloned copy rather than the original source.
michael@0:     //
michael@0:     adjustPointer(dest, &dest->context, src);
michael@0:     adjustPointer(dest, &dest->p, src);
michael@0:     adjustPointer(dest, &dest->q, src);
michael@0:     adjustPointer(dest, &dest->r, src);
michael@0:     adjustPointer(dest, (const void **)&dest->chunkContents, src);
michael@0: 
michael@0:     return dest;
michael@0: }
michael@0: 
michael@0: 
michael@0: U_CDECL_END
michael@0: 
michael@0: 
michael@0: 
michael@0: //------------------------------------------------------------------------------
michael@0: //
michael@0: //     UText implementation for UTF-8 char * strings (read-only)
michael@0: //     Limitation:  string length must be <= 0x7fffffff in length.
michael@0: //                  (length must for in an int32_t variable)
michael@0: //
michael@0: //         Use of UText data members:
michael@0: //              context    pointer to UTF-8 string
michael@0: //              utext.b    is the input string length (bytes).
michael@0: //              utext.c    Length scanned so far in string
michael@0: //                           (for optimizing finding length of zero terminated strings.)
michael@0: //              utext.p    pointer to the current buffer
michael@0: //              utext.q    pointer to the other buffer.
michael@0: //
michael@0: //------------------------------------------------------------------------------
michael@0: 
michael@0: // Chunk size.
michael@0: //     Must be less than 85, because of byte mapping from UChar indexes to native indexes.
michael@0: //     Worst case is three native bytes to one UChar.  (Supplemenaries are 4 native bytes
michael@0: //     to two UChars.)
michael@0: //
michael@0: enum { UTF8_TEXT_CHUNK_SIZE=32 };
michael@0: 
michael@0: //
michael@0: // UTF8Buf  Two of these structs will be set up in the UText's extra allocated space.
michael@0: //          Each contains the UChar chunk buffer, the to and from native maps, and
michael@0: //          header info.
michael@0: //
michael@0: //     because backwards iteration fills the buffers starting at the end and
michael@0: //     working towards the front, the filled part of the buffers may not begin
michael@0: //     at the start of the available storage for the buffers.
michael@0: //
michael@0: //     Buffer size is one bigger than the specified UTF8_TEXT_CHUNK_SIZE to allow for
michael@0: //     the last character added being a supplementary, and thus requiring a surrogate
michael@0: //     pair.  Doing this is simpler than checking for the edge case.
michael@0: //
michael@0: 
michael@0: struct UTF8Buf {
michael@0:     int32_t   bufNativeStart;                        // Native index of first char in UChar buf
michael@0:     int32_t   bufNativeLimit;                        // Native index following last char in buf.
michael@0:     int32_t   bufStartIdx;                           // First filled position in buf.
michael@0:     int32_t   bufLimitIdx;                           // Limit of filled range in buf.
michael@0:     int32_t   bufNILimit;                            // Limit of native indexing part of buf
michael@0:     int32_t   toUCharsMapStart;                      // Native index corresponding to
michael@0:                                                      //   mapToUChars[0].
michael@0:                                                      //   Set to bufNativeStart when filling forwards.
michael@0:                                                      //   Set to computed value when filling backwards.
michael@0: 
michael@0:     UChar     buf[UTF8_TEXT_CHUNK_SIZE+4];           // The UChar buffer.  Requires one extra position beyond the
michael@0:                                                      //   the chunk size, to allow for surrogate at the end.
michael@0:                                                      //   Length must be identical to mapToNative array, below,
michael@0:                                                      //   because of the way indexing works when the array is
michael@0:                                                      //   filled backwards during a reverse iteration.  Thus,
michael@0:                                                      //   the additional extra size.
michael@0:     uint8_t   mapToNative[UTF8_TEXT_CHUNK_SIZE+4];   // map UChar index in buf to
michael@0:                                                      //  native offset from bufNativeStart.
michael@0:                                                      //  Requires two extra slots,
michael@0:                                                      //    one for a supplementary starting in the last normal position,
michael@0:                                                      //    and one for an entry for the buffer limit position.
michael@0:     uint8_t   mapToUChars[UTF8_TEXT_CHUNK_SIZE*3+6]; // Map native offset from bufNativeStart to
michael@0:                                                      //   correspoding offset in filled part of buf.
michael@0:     int32_t   align;
michael@0: };
michael@0: 
michael@0: U_CDECL_BEGIN
michael@0: 
michael@0: //
michael@0: //   utf8TextLength
michael@0: //
michael@0: //        Get the length of the string.  If we don't already know it,
michael@0: //              we'll need to scan for the trailing  nul.
michael@0: //
michael@0: static int64_t U_CALLCONV
michael@0: utf8TextLength(UText *ut) {
michael@0:     if (ut->b < 0) {
michael@0:         // Zero terminated string, and we haven't scanned to the end yet.
michael@0:         // Scan it now.
michael@0:         const char *r = (const char *)ut->context + ut->c;
michael@0:         while (*r != 0) {
michael@0:             r++;
michael@0:         }
michael@0:         if ((r - (const char *)ut->context) < 0x7fffffff) {
michael@0:             ut->b = (int32_t)(r - (const char *)ut->context);
michael@0:         } else {
michael@0:             // Actual string was bigger (more than 2 gig) than we
michael@0:             //   can handle.  Clip it to 2 GB.
michael@0:             ut->b = 0x7fffffff;
michael@0:         }
michael@0:         ut->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
michael@0:     }
michael@0:     return ut->b;
michael@0: }
michael@0: 
michael@0: 
michael@0: 
michael@0: 
michael@0: 
michael@0: 
michael@0: static UBool U_CALLCONV
michael@0: utf8TextAccess(UText *ut, int64_t index, UBool forward) {
michael@0:     //
michael@0:     //  Apologies to those who are allergic to goto statements.
michael@0:     //    Consider each goto to a labelled block to be the equivalent of
michael@0:     //         call the named block as if it were a function();
michael@0:     //         return;
michael@0:     //
michael@0:     const uint8_t *s8=(const uint8_t *)ut->context;
michael@0:     UTF8Buf *u8b = NULL;
michael@0:     int32_t  length = ut->b;         // Length of original utf-8
michael@0:     int32_t  ix= (int32_t)index;     // Requested index, trimmed to 32 bits.
michael@0:     int32_t  mapIndex = 0;
michael@0:     if (index<0) {
michael@0:         ix=0;
michael@0:     } else if (index > 0x7fffffff) {
michael@0:         // Strings with 64 bit lengths not supported by this UTF-8 provider.
michael@0:         ix = 0x7fffffff;
michael@0:     }
michael@0: 
michael@0:     // Pin requested index to the string length.
michael@0:     if (ix>length) {
michael@0:         if (length>=0) {
michael@0:             ix=length;
michael@0:         } else if (ix>=ut->c) {
michael@0:             // Zero terminated string, and requested index is beyond
michael@0:             //   the region that has already been scanned.
michael@0:             //   Scan up to either the end of the string or to the
michael@0:             //   requested position, whichever comes first.
michael@0:             while (ut->c<ix && s8[ut->c]!=0) {
michael@0:                 ut->c++;
michael@0:             }
michael@0:             //  TODO:  support for null terminated string length > 32 bits.
michael@0:             if (s8[ut->c] == 0) {
michael@0:                 // We just found the actual length of the string.
michael@0:                 //  Trim the requested index back to that.
michael@0:                 ix     = ut->c;
michael@0:                 ut->b  = ut->c;
michael@0:                 length = ut->c;
michael@0:                 ut->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
michael@0:             }
michael@0:         }
michael@0:     }
michael@0: 
michael@0:     //
michael@0:     // Dispatch to the appropriate action for a forward iteration request.
michael@0:     //
michael@0:     if (forward) {
michael@0:         if (ix==ut->chunkNativeLimit) {
michael@0:             // Check for normal sequential iteration cases first.
michael@0:             if (ix==length) {
michael@0:                 // Just reached end of string
michael@0:                 // Don't swap buffers, but do set the
michael@0:                 //   current buffer position.
michael@0:                 ut->chunkOffset = ut->chunkLength;
michael@0:                 return FALSE;
michael@0:             } else {
michael@0:                 // End of current buffer.
michael@0:                 //   check whether other buffer already has what we need.
michael@0:                 UTF8Buf *altB = (UTF8Buf *)ut->q;
michael@0:                 if (ix>=altB->bufNativeStart && ix<altB->bufNativeLimit) {
michael@0:                     goto swapBuffers;
michael@0:                 }
michael@0:             }
michael@0:         }
michael@0: 
michael@0:         // A random access.  Desired index could be in either or niether buf.
michael@0:         // For optimizing the order of testing, first check for the index
michael@0:         //    being in the other buffer.  This will be the case for uses that
michael@0:         //    move back and forth over a fairly limited range
michael@0:         {
michael@0:             u8b = (UTF8Buf *)ut->q;   // the alternate buffer
michael@0:             if (ix>=u8b->bufNativeStart && ix<u8b->bufNativeLimit) {
michael@0:                 // Requested index is in the other buffer.
michael@0:                 goto swapBuffers;
michael@0:             }
michael@0:             if (ix == length) {
michael@0:                 // Requested index is end-of-string.
michael@0:                 //   (this is the case of randomly seeking to the end.
michael@0:                 //    The case of iterating off the end is handled earlier.)
michael@0:                 if (ix == ut->chunkNativeLimit) {
michael@0:                     // Current buffer extends up to the end of the string.
michael@0:                     //   Leave it as the current buffer.
michael@0:                     ut->chunkOffset = ut->chunkLength;
michael@0:                     return FALSE;
michael@0:                 }
michael@0:                 if (ix == u8b->bufNativeLimit) {
michael@0:                     // Alternate buffer extends to the end of string.
michael@0:                     //   Swap it in as the current buffer.
michael@0:                     goto swapBuffersAndFail;
michael@0:                 }
michael@0: 
michael@0:                 // Neither existing buffer extends to the end of the string.
michael@0:                 goto makeStubBuffer;
michael@0:             }
michael@0: 
michael@0:             if (ix<ut->chunkNativeStart || ix>=ut->chunkNativeLimit) {
michael@0:                 // Requested index is in neither buffer.
michael@0:                 goto fillForward;
michael@0:             }
michael@0: 
michael@0:             // Requested index is in this buffer.
michael@0:             u8b = (UTF8Buf *)ut->p;   // the current buffer
michael@0:             mapIndex = ix - u8b->toUCharsMapStart;
michael@0:             ut->chunkOffset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;
michael@0:             return TRUE;
michael@0: 
michael@0:         }
michael@0:     }
michael@0: 
michael@0: 
michael@0:     //
michael@0:     // Dispatch to the appropriate action for a
michael@0:     //   Backwards Diretion iteration request.
michael@0:     //
michael@0:     if (ix==ut->chunkNativeStart) {
michael@0:         // Check for normal sequential iteration cases first.
michael@0:         if (ix==0) {
michael@0:             // Just reached the start of string
michael@0:             // Don't swap buffers, but do set the
michael@0:             //   current buffer position.
michael@0:             ut->chunkOffset = 0;
michael@0:             return FALSE;
michael@0:         } else {
michael@0:             // Start of current buffer.
michael@0:             //   check whether other buffer already has what we need.
michael@0:             UTF8Buf *altB = (UTF8Buf *)ut->q;
michael@0:             if (ix>altB->bufNativeStart && ix<=altB->bufNativeLimit) {
michael@0:                 goto swapBuffers;
michael@0:             }
michael@0:         }
michael@0:     }
michael@0: 
michael@0:     // A random access.  Desired index could be in either or niether buf.
michael@0:     // For optimizing the order of testing,
michael@0:     //    Most likely case:  in the other buffer.
michael@0:     //    Second most likely: in neither buffer.
michael@0:     //    Unlikely, but must work:  in the current buffer.
michael@0:     u8b = (UTF8Buf *)ut->q;   // the alternate buffer
michael@0:     if (ix>u8b->bufNativeStart && ix<=u8b->bufNativeLimit) {
michael@0:         // Requested index is in the other buffer.
michael@0:         goto swapBuffers;
michael@0:     }
michael@0:     // Requested index is start-of-string.
michael@0:     //   (this is the case of randomly seeking to the start.
michael@0:     //    The case of iterating off the start is handled earlier.)
michael@0:     if (ix==0) {
michael@0:         if (u8b->bufNativeStart==0) {
michael@0:             // Alternate buffer contains the data for the start string.
michael@0:             // Make it be the current buffer.
michael@0:             goto swapBuffersAndFail;
michael@0:         } else {
michael@0:             // Request for data before the start of string,
michael@0:             //   neither buffer is usable.
michael@0:             //   set up a zero-length buffer.
michael@0:             goto makeStubBuffer;
michael@0:         }
michael@0:     }
michael@0: 
michael@0:     if (ix<=ut->chunkNativeStart || ix>ut->chunkNativeLimit) {
michael@0:         // Requested index is in neither buffer.
michael@0:         goto fillReverse;
michael@0:     }
michael@0: 
michael@0:     // Requested index is in this buffer.
michael@0:     //   Set the utf16 buffer index.
michael@0:     u8b = (UTF8Buf *)ut->p;
michael@0:     mapIndex = ix - u8b->toUCharsMapStart;
michael@0:     ut->chunkOffset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;
michael@0:     if (ut->chunkOffset==0) {
michael@0:         // This occurs when the first character in the text is
michael@0:         //   a multi-byte UTF-8 char, and the requested index is to
michael@0:         //   one of the trailing bytes.  Because there is no preceding ,
michael@0:         //   character, this access fails.  We can't pick up on the
michael@0:         //   situation sooner because the requested index is not zero.
michael@0:         return FALSE;
michael@0:     } else {
michael@0:         return TRUE;
michael@0:     }
michael@0: 
michael@0: 
michael@0: 
michael@0: swapBuffers:
michael@0:     //  The alternate buffer (ut->q) has the string data that was requested.
michael@0:     //  Swap the primary and alternate buffers, and set the
michael@0:     //   chunk index into the new primary buffer.
michael@0:     {
michael@0:         u8b   = (UTF8Buf *)ut->q;
michael@0:         ut->q = ut->p;
michael@0:         ut->p = u8b;
michael@0:         ut->chunkContents       = &u8b->buf[u8b->bufStartIdx];
michael@0:         ut->chunkLength         = u8b->bufLimitIdx - u8b->bufStartIdx;
michael@0:         ut->chunkNativeStart    = u8b->bufNativeStart;
michael@0:         ut->chunkNativeLimit    = u8b->bufNativeLimit;
michael@0:         ut->nativeIndexingLimit = u8b->bufNILimit;
michael@0: 
michael@0:         // Index into the (now current) chunk
michael@0:         // Use the map to set the chunk index.  It's more trouble than it's worth
michael@0:         //    to check whether native indexing can be used.
michael@0:         U_ASSERT(ix>=u8b->bufNativeStart);
michael@0:         U_ASSERT(ix<=u8b->bufNativeLimit);
michael@0:         mapIndex = ix - u8b->toUCharsMapStart;
michael@0:         U_ASSERT(mapIndex>=0);
michael@0:         U_ASSERT(mapIndex<(int32_t)sizeof(u8b->mapToUChars));
michael@0:         ut->chunkOffset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;
michael@0: 
michael@0:         return TRUE;
michael@0:     }
michael@0: 
michael@0: 
michael@0:  swapBuffersAndFail:
michael@0:     // We got a request for either the start or end of the string,
michael@0:     //  with iteration continuing in the out-of-bounds direction.
michael@0:     // The alternate buffer already contains the data up to the
michael@0:     //  start/end.
michael@0:     // Swap the buffers, then return failure, indicating that we couldn't
michael@0:     //  make things correct for continuing the iteration in the requested
michael@0:     //  direction.  The position & buffer are correct should the
michael@0:     //  user decide to iterate in the opposite direction.
michael@0:     u8b   = (UTF8Buf *)ut->q;
michael@0:     ut->q = ut->p;
michael@0:     ut->p = u8b;
michael@0:     ut->chunkContents       = &u8b->buf[u8b->bufStartIdx];
michael@0:     ut->chunkLength         = u8b->bufLimitIdx - u8b->bufStartIdx;
michael@0:     ut->chunkNativeStart    = u8b->bufNativeStart;
michael@0:     ut->chunkNativeLimit    = u8b->bufNativeLimit;
michael@0:     ut->nativeIndexingLimit = u8b->bufNILimit;
michael@0: 
michael@0:     // Index into the (now current) chunk
michael@0:     //  For this function  (swapBuffersAndFail), the requested index
michael@0:     //    will always be at either the start or end of the chunk.
michael@0:     if (ix==u8b->bufNativeLimit) {
michael@0:         ut->chunkOffset = ut->chunkLength;
michael@0:     } else  {
michael@0:         ut->chunkOffset = 0;
michael@0:         U_ASSERT(ix == u8b->bufNativeStart);
michael@0:     }
michael@0:     return FALSE;
michael@0: 
michael@0: makeStubBuffer:
michael@0:     //   The user has done a seek/access past the start or end
michael@0:     //   of the string.  Rather than loading data that is likely
michael@0:     //   to never be used, just set up a zero-length buffer at
michael@0:     //   the position.
michael@0:     u8b = (UTF8Buf *)ut->q;
michael@0:     u8b->bufNativeStart   = ix;
michael@0:     u8b->bufNativeLimit   = ix;
michael@0:     u8b->bufStartIdx      = 0;
michael@0:     u8b->bufLimitIdx      = 0;
michael@0:     u8b->bufNILimit       = 0;
michael@0:     u8b->toUCharsMapStart = ix;
michael@0:     u8b->mapToNative[0]   = 0;
michael@0:     u8b->mapToUChars[0]   = 0;
michael@0:     goto swapBuffersAndFail;
michael@0: 
michael@0: 
michael@0: 
michael@0: fillForward:
michael@0:     {
michael@0:         // Move the incoming index to a code point boundary.
michael@0:         U8_SET_CP_START(s8, 0, ix);
michael@0: 
michael@0:         // Swap the UText buffers.
michael@0:         //  We want to fill what was previously the alternate buffer,
michael@0:         //  and make what was the current buffer be the new alternate.
michael@0:         UTF8Buf *u8b = (UTF8Buf *)ut->q;
michael@0:         ut->q = ut->p;
michael@0:         ut->p = u8b;
michael@0: 
michael@0:         int32_t strLen = ut->b;
michael@0:         UBool   nulTerminated = FALSE;
michael@0:         if (strLen < 0) {
michael@0:             strLen = 0x7fffffff;
michael@0:             nulTerminated = TRUE;
michael@0:         }
michael@0: 
michael@0:         UChar   *buf = u8b->buf;
michael@0:         uint8_t *mapToNative  = u8b->mapToNative;
michael@0:         uint8_t *mapToUChars  = u8b->mapToUChars;
michael@0:         int32_t  destIx       = 0;
michael@0:         int32_t  srcIx        = ix;
michael@0:         UBool    seenNonAscii = FALSE;
michael@0:         UChar32  c = 0;
michael@0: 
michael@0:         // Fill the chunk buffer and mapping arrays.
michael@0:         while (destIx<UTF8_TEXT_CHUNK_SIZE) {
michael@0:             c = s8[srcIx];
michael@0:             if (c>0 && c<0x80) {
michael@0:                 // Special case ASCII range for speed.
michael@0:                 //   zero is excluded to simplify bounds checking.
michael@0:                 buf[destIx] = (UChar)c;
michael@0:                 mapToNative[destIx]    = (uint8_t)(srcIx - ix);
michael@0:                 mapToUChars[srcIx-ix]  = (uint8_t)destIx;
michael@0:                 srcIx++;
michael@0:                 destIx++;
michael@0:             } else {
michael@0:                 // General case, handle everything.
michael@0:                 if (seenNonAscii == FALSE) {
michael@0:                     seenNonAscii = TRUE;
michael@0:                     u8b->bufNILimit = destIx;
michael@0:                 }
michael@0: 
michael@0:                 int32_t  cIx      = srcIx;
michael@0:                 int32_t  dIx      = destIx;
michael@0:                 int32_t  dIxSaved = destIx;
michael@0:                 U8_NEXT_OR_FFFD(s8, srcIx, strLen, c);
michael@0:                 if (c==0 && nulTerminated) {
michael@0:                     srcIx--;
michael@0:                     break;
michael@0:                 }
michael@0: 
michael@0:                 U16_APPEND_UNSAFE(buf, destIx, c);
michael@0:                 do {
michael@0:                     mapToNative[dIx++] = (uint8_t)(cIx - ix);
michael@0:                 } while (dIx < destIx);
michael@0: 
michael@0:                 do {
michael@0:                     mapToUChars[cIx++ - ix] = (uint8_t)dIxSaved;
michael@0:                 } while (cIx < srcIx);
michael@0:             }
michael@0:             if (srcIx>=strLen) {
michael@0:                 break;
michael@0:             }
michael@0: 
michael@0:         }
michael@0: 
michael@0:         //  store Native <--> Chunk Map entries for the end of the buffer.
michael@0:         //    There is no actual character here, but the index position is valid.
michael@0:         mapToNative[destIx]     = (uint8_t)(srcIx - ix);
michael@0:         mapToUChars[srcIx - ix] = (uint8_t)destIx;
michael@0: 
michael@0:         //  fill in Buffer descriptor
michael@0:         u8b->bufNativeStart     = ix;
michael@0:         u8b->bufNativeLimit     = srcIx;
michael@0:         u8b->bufStartIdx        = 0;
michael@0:         u8b->bufLimitIdx        = destIx;
michael@0:         if (seenNonAscii == FALSE) {
michael@0:             u8b->bufNILimit     = destIx;
michael@0:         }
michael@0:         u8b->toUCharsMapStart   = u8b->bufNativeStart;
michael@0: 
michael@0:         // Set UText chunk to refer to this buffer.
michael@0:         ut->chunkContents       = buf;
michael@0:         ut->chunkOffset         = 0;
michael@0:         ut->chunkLength         = u8b->bufLimitIdx;
michael@0:         ut->chunkNativeStart    = u8b->bufNativeStart;
michael@0:         ut->chunkNativeLimit    = u8b->bufNativeLimit;
michael@0:         ut->nativeIndexingLimit = u8b->bufNILimit;
michael@0: 
michael@0:         // For zero terminated strings, keep track of the maximum point
michael@0:         //   scanned so far.
michael@0:         if (nulTerminated && srcIx>ut->c) {
michael@0:             ut->c = srcIx;
michael@0:             if (c==0) {
michael@0:                 // We scanned to the end.
michael@0:                 //   Remember the actual length.
michael@0:                 ut->b = srcIx;
michael@0:                 ut->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
michael@0:             }
michael@0:         }
michael@0:         return TRUE;
michael@0:     }
michael@0: 
michael@0: 
michael@0: fillReverse:
michael@0:     {
michael@0:         // Move the incoming index to a code point boundary.
michael@0:         // Can only do this if the incoming index is somewhere in the interior of the string.
michael@0:         //   If index is at the end, there is no character there to look at.
michael@0:         if (ix != ut->b) {
michael@0:             U8_SET_CP_START(s8, 0, ix);
michael@0:         }
michael@0: 
michael@0:         // Swap the UText buffers.
michael@0:         //  We want to fill what was previously the alternate buffer,
michael@0:         //  and make what was the current buffer be the new alternate.
michael@0:         UTF8Buf *u8b = (UTF8Buf *)ut->q;
michael@0:         ut->q = ut->p;
michael@0:         ut->p = u8b;
michael@0: 
michael@0:         UChar   *buf = u8b->buf;
michael@0:         uint8_t *mapToNative = u8b->mapToNative;
michael@0:         uint8_t *mapToUChars = u8b->mapToUChars;
michael@0:         int32_t  toUCharsMapStart = ix - (UTF8_TEXT_CHUNK_SIZE*3 + 1);
michael@0:         int32_t  destIx = UTF8_TEXT_CHUNK_SIZE+2;   // Start in the overflow region
michael@0:                                                     //   at end of buffer to leave room
michael@0:                                                     //   for a surrogate pair at the
michael@0:                                                     //   buffer start.
michael@0:         int32_t  srcIx  = ix;
michael@0:         int32_t  bufNILimit = destIx;
michael@0:         UChar32   c;
michael@0: 
michael@0:         // Map to/from Native Indexes, fill in for the position at the end of
michael@0:         //   the buffer.
michael@0:         //
michael@0:         mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
michael@0:         mapToUChars[srcIx - toUCharsMapStart] = (uint8_t)destIx;
michael@0: 
michael@0:         // Fill the chunk buffer
michael@0:         // Work backwards, filling from the end of the buffer towards the front.
michael@0:         //
michael@0:         while (destIx>2 && (srcIx - toUCharsMapStart > 5) && (srcIx > 0)) {
michael@0:             srcIx--;
michael@0:             destIx--;
michael@0: 
michael@0:             // Get last byte of the UTF-8 character
michael@0:             c = s8[srcIx];
michael@0:             if (c<0x80) {
michael@0:                 // Special case ASCII range for speed.
michael@0:                 buf[destIx] = (UChar)c;
michael@0:                 mapToUChars[srcIx - toUCharsMapStart] = (uint8_t)destIx;
michael@0:                 mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
michael@0:             } else {
michael@0:                 // General case, handle everything non-ASCII.
michael@0: 
michael@0:                 int32_t  sIx      = srcIx;  // ix of last byte of multi-byte u8 char
michael@0: 
michael@0:                 // Get the full character from the UTF8 string.
michael@0:                 //   use code derived from tbe macros in utf8.h
michael@0:                 //   Leaves srcIx pointing at the first byte of the UTF-8 char.
michael@0:                 //
michael@0:                 c=utf8_prevCharSafeBody(s8, 0, &srcIx, c, -3);
michael@0:                 // leaves srcIx at first byte of the multi-byte char.
michael@0: 
michael@0:                 // Store the character in UTF-16 buffer.
michael@0:                 if (c<0x10000) {
michael@0:                     buf[destIx] = (UChar)c;
michael@0:                     mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
michael@0:                 } else {
michael@0:                     buf[destIx]         = U16_TRAIL(c);
michael@0:                     mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
michael@0:                     buf[--destIx]       = U16_LEAD(c);
michael@0:                     mapToNative[destIx] = (uint8_t)(srcIx - toUCharsMapStart);
michael@0:                 }
michael@0: 
michael@0:                 // Fill in the map from native indexes to UChars buf index.
michael@0:                 do {
michael@0:                     mapToUChars[sIx-- - toUCharsMapStart] = (uint8_t)destIx;
michael@0:                 } while (sIx >= srcIx);
michael@0: 
michael@0:                 // Set native indexing limit to be the current position.
michael@0:                 //   We are processing a non-ascii, non-native-indexing char now;
michael@0:                 //     the limit will be here if the rest of the chars to be
michael@0:                 //     added to this buffer are ascii.
michael@0:                 bufNILimit = destIx;
michael@0:             }
michael@0:         }
michael@0:         u8b->bufNativeStart     = srcIx;
michael@0:         u8b->bufNativeLimit     = ix;
michael@0:         u8b->bufStartIdx        = destIx;
michael@0:         u8b->bufLimitIdx        = UTF8_TEXT_CHUNK_SIZE+2;
michael@0:         u8b->bufNILimit         = bufNILimit - u8b->bufStartIdx;
michael@0:         u8b->toUCharsMapStart   = toUCharsMapStart;
michael@0: 
michael@0:         ut->chunkContents       = &buf[u8b->bufStartIdx];
michael@0:         ut->chunkLength         = u8b->bufLimitIdx - u8b->bufStartIdx;
michael@0:         ut->chunkOffset         = ut->chunkLength;
michael@0:         ut->chunkNativeStart    = u8b->bufNativeStart;
michael@0:         ut->chunkNativeLimit    = u8b->bufNativeLimit;
michael@0:         ut->nativeIndexingLimit = u8b->bufNILimit;
michael@0:         return TRUE;
michael@0:     }
michael@0: 
michael@0: }
michael@0: 
michael@0: 
michael@0: 
michael@0: //
michael@0: //  This is a slightly modified copy of u_strFromUTF8,
michael@0: //     Inserts a Replacement Char rather than failing on invalid UTF-8
michael@0: //     Removes unnecessary features.
michael@0: //
michael@0: static UChar*
michael@0: utext_strFromUTF8(UChar *dest,
michael@0:               int32_t destCapacity,
michael@0:               int32_t *pDestLength,
michael@0:               const char* src,
michael@0:               int32_t srcLength,        // required.  NUL terminated not supported.
michael@0:               UErrorCode *pErrorCode
michael@0:               )
michael@0: {
michael@0: 
michael@0:     UChar *pDest = dest;
michael@0:     UChar *pDestLimit = (dest!=NULL)?(dest+destCapacity):NULL;
michael@0:     UChar32 ch=0;
michael@0:     int32_t index = 0;
michael@0:     int32_t reqLength = 0;
michael@0:     uint8_t* pSrc = (uint8_t*) src;
michael@0: 
michael@0: 
michael@0:     while((index < srcLength)&&(pDest<pDestLimit)){
michael@0:         ch = pSrc[index++];
michael@0:         if(ch <=0x7f){
michael@0:             *pDest++=(UChar)ch;
michael@0:         }else{
michael@0:             ch=utf8_nextCharSafeBody(pSrc, &index, srcLength, ch, -3);
michael@0:             if(U_IS_BMP(ch)){
michael@0:                 *(pDest++)=(UChar)ch;
michael@0:             }else{
michael@0:                 *(pDest++)=U16_LEAD(ch);
michael@0:                 if(pDest<pDestLimit){
michael@0:                     *(pDest++)=U16_TRAIL(ch);
michael@0:                 }else{
michael@0:                     reqLength++;
michael@0:                     break;
michael@0:                 }
michael@0:             }
michael@0:         }
michael@0:     }
michael@0:     /* donot fill the dest buffer just count the UChars needed */
michael@0:     while(index < srcLength){
michael@0:         ch = pSrc[index++];
michael@0:         if(ch <= 0x7f){
michael@0:             reqLength++;
michael@0:         }else{
michael@0:             ch=utf8_nextCharSafeBody(pSrc, &index, srcLength, ch, -3);
michael@0:             reqLength+=U16_LENGTH(ch);
michael@0:         }
michael@0:     }
michael@0: 
michael@0:     reqLength+=(int32_t)(pDest - dest);
michael@0: 
michael@0:     if(pDestLength){
michael@0:         *pDestLength = reqLength;
michael@0:     }
michael@0: 
michael@0:     /* Terminate the buffer */
michael@0:     u_terminateUChars(dest,destCapacity,reqLength,pErrorCode);
michael@0: 
michael@0:     return dest;
michael@0: }
michael@0: 
michael@0: 
michael@0: 
michael@0: static int32_t U_CALLCONV
michael@0: utf8TextExtract(UText *ut,
michael@0:                 int64_t start, int64_t limit,
michael@0:                 UChar *dest, int32_t destCapacity,
michael@0:                 UErrorCode *pErrorCode) {
michael@0:     if(U_FAILURE(*pErrorCode)) {
michael@0:         return 0;
michael@0:     }
michael@0:     if(destCapacity<0 || (dest==NULL && destCapacity>0)) {
michael@0:         *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
michael@0:         return 0;
michael@0:     }
michael@0:     int32_t  length  = ut->b;
michael@0:     int32_t  start32 = pinIndex(start, length);
michael@0:     int32_t  limit32 = pinIndex(limit, length);
michael@0: 
michael@0:     if(start32>limit32) {
michael@0:         *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR;
michael@0:         return 0;
michael@0:     }
michael@0: 
michael@0: 
michael@0:     // adjust the incoming indexes to land on code point boundaries if needed.
michael@0:     //    adjust by no more than three, because that is the largest number of trail bytes
michael@0:     //    in a well formed UTF8 character.
michael@0:     const uint8_t *buf = (const uint8_t *)ut->context;
michael@0:     int i;
michael@0:     if (start32 < ut->chunkNativeLimit) {
michael@0:         for (i=0; i<3; i++) {
michael@0:             if (U8_IS_SINGLE(buf[start32]) || U8_IS_LEAD(buf[start32]) || start32==0) {
michael@0:                 break;
michael@0:             }
michael@0:             start32--;
michael@0:         }
michael@0:     }
michael@0: 
michael@0:     if (limit32 < ut->chunkNativeLimit) {
michael@0:         for (i=0; i<3; i++) {
michael@0:             if (U8_IS_SINGLE(buf[limit32]) || U8_IS_LEAD(buf[limit32]) || limit32==0) {
michael@0:                 break;
michael@0:             }
michael@0:             limit32--;
michael@0:         }
michael@0:     }
michael@0: 
michael@0:     // Do the actual extract.
michael@0:     int32_t destLength=0;
michael@0:     utext_strFromUTF8(dest, destCapacity, &destLength,
michael@0:                     (const char *)ut->context+start32, limit32-start32,
michael@0:                     pErrorCode);
michael@0:     utf8TextAccess(ut, limit32, TRUE);
michael@0:     return destLength;
michael@0: }
michael@0: 
michael@0: //
michael@0: // utf8TextMapOffsetToNative
michael@0: //
michael@0: // Map a chunk (UTF-16) offset to a native index.
michael@0: static int64_t U_CALLCONV
michael@0: utf8TextMapOffsetToNative(const UText *ut) {
michael@0:     //
michael@0:     UTF8Buf *u8b = (UTF8Buf *)ut->p;
michael@0:     U_ASSERT(ut->chunkOffset>ut->nativeIndexingLimit && ut->chunkOffset<=ut->chunkLength);
michael@0:     int32_t nativeOffset = u8b->mapToNative[ut->chunkOffset + u8b->bufStartIdx] + u8b->toUCharsMapStart;
michael@0:     U_ASSERT(nativeOffset >= ut->chunkNativeStart && nativeOffset <= ut->chunkNativeLimit);
michael@0:     return nativeOffset;
michael@0: }
michael@0: 
michael@0: //
michael@0: // Map a native index to the corrsponding chunk offset
michael@0: //
michael@0: static int32_t U_CALLCONV
michael@0: utf8TextMapIndexToUTF16(const UText *ut, int64_t index64) {
michael@0:     U_ASSERT(index64 <= 0x7fffffff);
michael@0:     int32_t index = (int32_t)index64;
michael@0:     UTF8Buf *u8b = (UTF8Buf *)ut->p;
michael@0:     U_ASSERT(index>=ut->chunkNativeStart+ut->nativeIndexingLimit);
michael@0:     U_ASSERT(index<=ut->chunkNativeLimit);
michael@0:     int32_t mapIndex = index - u8b->toUCharsMapStart;
michael@0:     int32_t offset = u8b->mapToUChars[mapIndex] - u8b->bufStartIdx;
michael@0:     U_ASSERT(offset>=0 && offset<=ut->chunkLength);
michael@0:     return offset;
michael@0: }
michael@0: 
michael@0: static UText * U_CALLCONV
michael@0: utf8TextClone(UText *dest, const UText *src, UBool deep, UErrorCode *status)
michael@0: {
michael@0:     // First do a generic shallow clone.  Does everything needed for the UText struct itself.
michael@0:     dest = shallowTextClone(dest, src, status);
michael@0: 
michael@0:     // For deep clones, make a copy of the string.
michael@0:     //  The copied storage is owned by the newly created clone.
michael@0:     //
michael@0:     // TODO:  There is an isssue with using utext_nativeLength().
michael@0:     //        That function is non-const in cases where the input was NUL terminated
michael@0:     //          and the length has not yet been determined.
michael@0:     //        This function (clone()) is const.
michael@0:     //        There potentially a thread safety issue lurking here.
michael@0:     //
michael@0:     if (deep && U_SUCCESS(*status)) {
michael@0:         int32_t  len = (int32_t)utext_nativeLength((UText *)src);
michael@0:         char *copyStr = (char *)uprv_malloc(len+1);
michael@0:         if (copyStr == NULL) {
michael@0:             *status = U_MEMORY_ALLOCATION_ERROR;
michael@0:         } else {
michael@0:             uprv_memcpy(copyStr, src->context, len+1);
michael@0:             dest->context = copyStr;
michael@0:             dest->providerProperties |= I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT);
michael@0:         }
michael@0:     }
michael@0:     return dest;
michael@0: }
michael@0: 
michael@0: 
michael@0: static void U_CALLCONV
michael@0: utf8TextClose(UText *ut) {
michael@0:     // Most of the work of close is done by the generic UText framework close.
michael@0:     // All that needs to be done here is to delete the UTF8 string if the UText
michael@0:     //  owns it.  This occurs if the UText was created by cloning.
michael@0:     if (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT)) {
michael@0:         char *s = (char *)ut->context;
michael@0:         uprv_free(s);
michael@0:         ut->context = NULL;
michael@0:     }
michael@0: }
michael@0: 
michael@0: U_CDECL_END
michael@0: 
michael@0: 
michael@0: static const struct UTextFuncs utf8Funcs = 
michael@0: {
michael@0:     sizeof(UTextFuncs),
michael@0:     0, 0, 0,             // Reserved alignment padding
michael@0:     utf8TextClone,
michael@0:     utf8TextLength,
michael@0:     utf8TextAccess,
michael@0:     utf8TextExtract,
michael@0:     NULL,                /* replace*/
michael@0:     NULL,                /* copy   */
michael@0:     utf8TextMapOffsetToNative,
michael@0:     utf8TextMapIndexToUTF16,
michael@0:     utf8TextClose,
michael@0:     NULL,                // spare 1
michael@0:     NULL,                // spare 2
michael@0:     NULL                 // spare 3
michael@0: };
michael@0: 
michael@0: 
michael@0: static const char gEmptyString[] = {0};
michael@0: 
michael@0: U_CAPI UText * U_EXPORT2
michael@0: utext_openUTF8(UText *ut, const char *s, int64_t length, UErrorCode *status) {
michael@0:     if(U_FAILURE(*status)) {
michael@0:         return NULL;
michael@0:     }
michael@0:     if(s==NULL && length==0) {
michael@0:         s = gEmptyString;
michael@0:     }
michael@0: 
michael@0:     if(s==NULL || length<-1 || length>INT32_MAX) {
michael@0:         *status=U_ILLEGAL_ARGUMENT_ERROR;
michael@0:         return NULL;
michael@0:     }
michael@0: 
michael@0:     ut = utext_setup(ut, sizeof(UTF8Buf) * 2, status);
michael@0:     if (U_FAILURE(*status)) {
michael@0:         return ut;
michael@0:     }
michael@0: 
michael@0:     ut->pFuncs  = &utf8Funcs;
michael@0:     ut->context = s;
michael@0:     ut->b       = (int32_t)length;
michael@0:     ut->c       = (int32_t)length;
michael@0:     if (ut->c < 0) {
michael@0:         ut->c = 0;
michael@0:         ut->providerProperties |= I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
michael@0:     }
michael@0:     ut->p = ut->pExtra;
michael@0:     ut->q = (char *)ut->pExtra + sizeof(UTF8Buf);
michael@0:     return ut;
michael@0: 
michael@0: }
michael@0: 
michael@0: 
michael@0: 
michael@0: 
michael@0: 
michael@0: 
michael@0: 
michael@0: 
michael@0: //------------------------------------------------------------------------------
michael@0: //
michael@0: //     UText implementation wrapper for Replaceable (read/write)
michael@0: //
michael@0: //         Use of UText data members:
michael@0: //            context    pointer to Replaceable.
michael@0: //            p          pointer to Replaceable if it is owned by the UText.
michael@0: //
michael@0: //------------------------------------------------------------------------------
michael@0: 
michael@0: 
michael@0: 
michael@0: // minimum chunk size for this implementation: 3
michael@0: // to allow for possible trimming for code point boundaries
michael@0: enum { REP_TEXT_CHUNK_SIZE=10 };
michael@0: 
michael@0: struct ReplExtra {
michael@0:     /*
michael@0:      * Chunk UChars.
michael@0:      * +1 to simplify filling with surrogate pair at the end.
michael@0:      */
michael@0:     UChar s[REP_TEXT_CHUNK_SIZE+1];
michael@0: };
michael@0: 
michael@0: 
michael@0: U_CDECL_BEGIN
michael@0: 
michael@0: static UText * U_CALLCONV
michael@0: repTextClone(UText *dest, const UText *src, UBool deep, UErrorCode *status) {
michael@0:     // First do a generic shallow clone.  Does everything needed for the UText struct itself.
michael@0:     dest = shallowTextClone(dest, src, status);
michael@0: 
michael@0:     // For deep clones, make a copy of the Replaceable.
michael@0:     //  The copied Replaceable storage is owned by the newly created UText clone.
michael@0:     //  A non-NULL pointer in UText.p is the signal to the close() function to delete
michael@0:     //    it.
michael@0:     //
michael@0:     if (deep && U_SUCCESS(*status)) {
michael@0:         const Replaceable *replSrc = (const Replaceable *)src->context;
michael@0:         dest->context = replSrc->clone();
michael@0:         dest->providerProperties |= I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT);
michael@0: 
michael@0:         // with deep clone, the copy is writable, even when the source is not.
michael@0:         dest->providerProperties |= I32_FLAG(UTEXT_PROVIDER_WRITABLE);
michael@0:     }
michael@0:     return dest;
michael@0: }
michael@0: 
michael@0: 
michael@0: static void U_CALLCONV
michael@0: repTextClose(UText *ut) {
michael@0:     // Most of the work of close is done by the generic UText framework close.
michael@0:     // All that needs to be done here is delete the Replaceable if the UText
michael@0:     //  owns it.  This occurs if the UText was created by cloning.
michael@0:     if (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT)) {
michael@0:         Replaceable *rep = (Replaceable *)ut->context;
michael@0:         delete rep;
michael@0:         ut->context = NULL;
michael@0:     }
michael@0: }
michael@0: 
michael@0: 
michael@0: static int64_t U_CALLCONV
michael@0: repTextLength(UText *ut) {
michael@0:     const Replaceable *replSrc = (const Replaceable *)ut->context;
michael@0:     int32_t  len = replSrc->length();
michael@0:     return len;
michael@0: }
michael@0: 
michael@0: 
michael@0: static UBool U_CALLCONV
michael@0: repTextAccess(UText *ut, int64_t index, UBool forward) {
michael@0:     const Replaceable *rep=(const Replaceable *)ut->context;
michael@0:     int32_t length=rep->length();   // Full length of the input text (bigger than a chunk)
michael@0: 
michael@0:     // clip the requested index to the limits of the text.
michael@0:     int32_t index32 = pinIndex(index, length);
michael@0:     U_ASSERT(index<=INT32_MAX);
michael@0: 
michael@0: 
michael@0:     /*
michael@0:      * Compute start/limit boundaries around index, for a segment of text
michael@0:      * to be extracted.
michael@0:      * To allow for the possibility that our user gave an index to the trailing
michael@0:      * half of a surrogate pair, we must request one extra preceding UChar when
michael@0:      * going in the forward direction.  This will ensure that the buffer has the
michael@0:      * entire code point at the specified index.
michael@0:      */
michael@0:     if(forward) {
michael@0: 
michael@0:         if (index32>=ut->chunkNativeStart && index32<ut->chunkNativeLimit) {
michael@0:             // Buffer already contains the requested position.
michael@0:             ut->chunkOffset = (int32_t)(index - ut->chunkNativeStart);
michael@0:             return TRUE;
michael@0:         }
michael@0:         if (index32>=length && ut->chunkNativeLimit==length) {
michael@0:             // Request for end of string, and buffer already extends up to it.
michael@0:             // Can't get the data, but don't change the buffer.
michael@0:             ut->chunkOffset = length - (int32_t)ut->chunkNativeStart;
michael@0:             return FALSE;
michael@0:         }
michael@0: 
michael@0:         ut->chunkNativeLimit = index + REP_TEXT_CHUNK_SIZE - 1;
michael@0:         // Going forward, so we want to have the buffer with stuff at and beyond
michael@0:         //   the requested index.  The -1 gets us one code point before the
michael@0:         //   requested index also, to handle the case of the index being on
michael@0:         //   a trail surrogate of a surrogate pair.
michael@0:         if(ut->chunkNativeLimit > length) {
michael@0:             ut->chunkNativeLimit = length;
michael@0:         }
michael@0:         // unless buffer ran off end, start is index-1.
michael@0:         ut->chunkNativeStart = ut->chunkNativeLimit - REP_TEXT_CHUNK_SIZE;
michael@0:         if(ut->chunkNativeStart < 0) {
michael@0:             ut->chunkNativeStart = 0;
michael@0:         }
michael@0:     } else {
michael@0:         // Reverse iteration.  Fill buffer with data preceding the requested index.
michael@0:         if (index32>ut->chunkNativeStart && index32<=ut->chunkNativeLimit) {
michael@0:             // Requested position already in buffer.
michael@0:             ut->chunkOffset = index32 - (int32_t)ut->chunkNativeStart;
michael@0:             return TRUE;
michael@0:         }
michael@0:         if (index32==0 && ut->chunkNativeStart==0) {
michael@0:             // Request for start, buffer already begins at start.
michael@0:             //  No data, but keep the buffer as is.
michael@0:             ut->chunkOffset = 0;
michael@0:             return FALSE;
michael@0:         }
michael@0: 
michael@0:         // Figure out the bounds of the chunk to extract for reverse iteration.
michael@0:         // Need to worry about chunk not splitting surrogate pairs, and while still
michael@0:         // containing the data we need.
michael@0:         // Fix by requesting a chunk that includes an extra UChar at the end.
michael@0:         // If this turns out to be a lead surrogate, we can lop it off and still have
michael@0:         //   the data we wanted.
michael@0:         ut->chunkNativeStart = index32 + 1 - REP_TEXT_CHUNK_SIZE;
michael@0:         if (ut->chunkNativeStart < 0) {
michael@0:             ut->chunkNativeStart = 0;
michael@0:         }
michael@0: 
michael@0:         ut->chunkNativeLimit = index32 + 1;
michael@0:         if (ut->chunkNativeLimit > length) {
michael@0:             ut->chunkNativeLimit = length;
michael@0:         }
michael@0:     }
michael@0: 
michael@0:     // Extract the new chunk of text from the Replaceable source.
michael@0:     ReplExtra *ex = (ReplExtra *)ut->pExtra;
michael@0:     // UnicodeString with its buffer a writable alias to the chunk buffer
michael@0:     UnicodeString buffer(ex->s, 0 /*buffer length*/, REP_TEXT_CHUNK_SIZE /*buffer capacity*/);
michael@0:     rep->extractBetween((int32_t)ut->chunkNativeStart, (int32_t)ut->chunkNativeLimit, buffer);
michael@0: 
michael@0:     ut->chunkContents  = ex->s;
michael@0:     ut->chunkLength    = (int32_t)(ut->chunkNativeLimit - ut->chunkNativeStart);
michael@0:     ut->chunkOffset    = (int32_t)(index32 - ut->chunkNativeStart);
michael@0: 
michael@0:     // Surrogate pairs from the input text must not span chunk boundaries.
michael@0:     // If end of chunk could be the start of a surrogate, trim it off.
michael@0:     if (ut->chunkNativeLimit < length &&
michael@0:         U16_IS_LEAD(ex->s[ut->chunkLength-1])) {
michael@0:             ut->chunkLength--;
michael@0:             ut->chunkNativeLimit--;
michael@0:             if (ut->chunkOffset > ut->chunkLength) {
michael@0:                 ut->chunkOffset = ut->chunkLength;
michael@0:             }
michael@0:         }
michael@0: 
michael@0:     // if the first UChar in the chunk could be the trailing half of a surrogate pair,
michael@0:     // trim it off.
michael@0:     if(ut->chunkNativeStart>0 && U16_IS_TRAIL(ex->s[0])) {
michael@0:         ++(ut->chunkContents);
michael@0:         ++(ut->chunkNativeStart);
michael@0:         --(ut->chunkLength);
michael@0:         --(ut->chunkOffset);
michael@0:     }
michael@0: 
michael@0:     // adjust the index/chunkOffset to a code point boundary
michael@0:     U16_SET_CP_START(ut->chunkContents, 0, ut->chunkOffset);
michael@0: 
michael@0:     // Use fast indexing for get/setNativeIndex()
michael@0:     ut->nativeIndexingLimit = ut->chunkLength;
michael@0: 
michael@0:     return TRUE;
michael@0: }
michael@0: 
michael@0: 
michael@0: 
michael@0: static int32_t U_CALLCONV
michael@0: repTextExtract(UText *ut,
michael@0:                int64_t start, int64_t limit,
michael@0:                UChar *dest, int32_t destCapacity,
michael@0:                UErrorCode *status) {
michael@0:     const Replaceable *rep=(const Replaceable *)ut->context;
michael@0:     int32_t  length=rep->length();
michael@0: 
michael@0:     if(U_FAILURE(*status)) {
michael@0:         return 0;
michael@0:     }
michael@0:     if(destCapacity<0 || (dest==NULL && destCapacity>0)) {
michael@0:         *status=U_ILLEGAL_ARGUMENT_ERROR;
michael@0:     }
michael@0:     if(start>limit) {
michael@0:         *status=U_INDEX_OUTOFBOUNDS_ERROR;
michael@0:         return 0;
michael@0:     }
michael@0: 
michael@0:     int32_t  start32 = pinIndex(start, length);
michael@0:     int32_t  limit32 = pinIndex(limit, length);
michael@0: 
michael@0:     // adjust start, limit if they point to trail half of surrogates
michael@0:     if (start32<length && U16_IS_TRAIL(rep->charAt(start32)) &&
michael@0:         U_IS_SUPPLEMENTARY(rep->char32At(start32))){
michael@0:             start32--;
michael@0:     }
michael@0:     if (limit32<length && U16_IS_TRAIL(rep->charAt(limit32)) &&
michael@0:         U_IS_SUPPLEMENTARY(rep->char32At(limit32))){
michael@0:             limit32--;
michael@0:     }
michael@0: 
michael@0:     length=limit32-start32;
michael@0:     if(length>destCapacity) {
michael@0:         limit32 = start32 + destCapacity;
michael@0:     }
michael@0:     UnicodeString buffer(dest, 0, destCapacity); // writable alias
michael@0:     rep->extractBetween(start32, limit32, buffer);
michael@0:     repTextAccess(ut, limit32, TRUE);
michael@0:     
michael@0:     return u_terminateUChars(dest, destCapacity, length, status);
michael@0: }
michael@0: 
michael@0: static int32_t U_CALLCONV
michael@0: repTextReplace(UText *ut,
michael@0:                int64_t start, int64_t limit,
michael@0:                const UChar *src, int32_t length,
michael@0:                UErrorCode *status) {
michael@0:     Replaceable *rep=(Replaceable *)ut->context;
michael@0:     int32_t oldLength;
michael@0: 
michael@0:     if(U_FAILURE(*status)) {
michael@0:         return 0;
michael@0:     }
michael@0:     if(src==NULL && length!=0) {
michael@0:         *status=U_ILLEGAL_ARGUMENT_ERROR;
michael@0:         return 0;
michael@0:     }
michael@0:     oldLength=rep->length(); // will subtract from new length
michael@0:     if(start>limit ) {
michael@0:         *status=U_INDEX_OUTOFBOUNDS_ERROR;
michael@0:         return 0;
michael@0:     }
michael@0: 
michael@0:     int32_t start32 = pinIndex(start, oldLength);
michael@0:     int32_t limit32 = pinIndex(limit, oldLength);
michael@0: 
michael@0:     // Snap start & limit to code point boundaries.
michael@0:     if (start32<oldLength && U16_IS_TRAIL(rep->charAt(start32)) &&
michael@0:         start32>0 && U16_IS_LEAD(rep->charAt(start32-1)))
michael@0:     {
michael@0:             start32--;
michael@0:     }
michael@0:     if (limit32<oldLength && U16_IS_LEAD(rep->charAt(limit32-1)) &&
michael@0:         U16_IS_TRAIL(rep->charAt(limit32)))
michael@0:     {
michael@0:             limit32++;
michael@0:     }
michael@0: 
michael@0:     // Do the actual replace operation using methods of the Replaceable class
michael@0:     UnicodeString replStr((UBool)(length<0), src, length); // read-only alias
michael@0:     rep->handleReplaceBetween(start32, limit32, replStr);
michael@0:     int32_t newLength = rep->length();
michael@0:     int32_t lengthDelta = newLength - oldLength;
michael@0: 
michael@0:     // Is the UText chunk buffer OK?
michael@0:     if (ut->chunkNativeLimit > start32) {
michael@0:         // this replace operation may have impacted the current chunk.
michael@0:         // invalidate it, which will force a reload on the next access.
michael@0:         invalidateChunk(ut);
michael@0:     }
michael@0: 
michael@0:     // set the iteration position to the end of the newly inserted replacement text.
michael@0:     int32_t newIndexPos = limit32 + lengthDelta;
michael@0:     repTextAccess(ut, newIndexPos, TRUE);
michael@0: 
michael@0:     return lengthDelta;
michael@0: }
michael@0: 
michael@0: 
michael@0: static void U_CALLCONV
michael@0: repTextCopy(UText *ut,
michael@0:                 int64_t start, int64_t limit,
michael@0:                 int64_t destIndex,
michael@0:                 UBool move,
michael@0:                 UErrorCode *status)
michael@0: {
michael@0:     Replaceable *rep=(Replaceable *)ut->context;
michael@0:     int32_t length=rep->length();
michael@0: 
michael@0:     if(U_FAILURE(*status)) {
michael@0:         return;
michael@0:     }
michael@0:     if (start>limit || (start<destIndex && destIndex<limit))
michael@0:     {
michael@0:         *status=U_INDEX_OUTOFBOUNDS_ERROR;
michael@0:         return;
michael@0:     }
michael@0: 
michael@0:     int32_t start32     = pinIndex(start, length);
michael@0:     int32_t limit32     = pinIndex(limit, length);
michael@0:     int32_t destIndex32 = pinIndex(destIndex, length);
michael@0: 
michael@0:     // TODO:  snap input parameters to code point boundaries.
michael@0: 
michael@0:     if(move) {
michael@0:         // move: copy to destIndex, then replace original with nothing
michael@0:         int32_t segLength=limit32-start32;
michael@0:         rep->copy(start32, limit32, destIndex32);
michael@0:         if(destIndex32<start32) {
michael@0:             start32+=segLength;
michael@0:             limit32+=segLength;
michael@0:         }
michael@0:         rep->handleReplaceBetween(start32, limit32, UnicodeString());
michael@0:     } else {
michael@0:         // copy
michael@0:         rep->copy(start32, limit32, destIndex32);
michael@0:     }
michael@0: 
michael@0:     // If the change to the text touched the region in the chunk buffer,
michael@0:     //  invalidate the buffer.
michael@0:     int32_t firstAffectedIndex = destIndex32;
michael@0:     if (move && start32<firstAffectedIndex) {
michael@0:         firstAffectedIndex = start32;
michael@0:     }
michael@0:     if (firstAffectedIndex < ut->chunkNativeLimit) {
michael@0:         // changes may have affected range covered by the chunk
michael@0:         invalidateChunk(ut);
michael@0:     }
michael@0: 
michael@0:     // Put iteration position at the newly inserted (moved) block,
michael@0:     int32_t  nativeIterIndex = destIndex32 + limit32 - start32;
michael@0:     if (move && destIndex32>start32) {
michael@0:         // moved a block of text towards the end of the string.
michael@0:         nativeIterIndex = destIndex32;
michael@0:     }
michael@0: 
michael@0:     // Set position, reload chunk if needed.
michael@0:     repTextAccess(ut, nativeIterIndex, TRUE);
michael@0: }
michael@0: 
michael@0: static const struct UTextFuncs repFuncs = 
michael@0: {
michael@0:     sizeof(UTextFuncs),
michael@0:     0, 0, 0,           // Reserved alignment padding
michael@0:     repTextClone,
michael@0:     repTextLength,
michael@0:     repTextAccess,
michael@0:     repTextExtract,
michael@0:     repTextReplace,   
michael@0:     repTextCopy,   
michael@0:     NULL,              // MapOffsetToNative,
michael@0:     NULL,              // MapIndexToUTF16,
michael@0:     repTextClose,
michael@0:     NULL,              // spare 1
michael@0:     NULL,              // spare 2
michael@0:     NULL               // spare 3
michael@0: };
michael@0: 
michael@0: 
michael@0: U_CAPI UText * U_EXPORT2
michael@0: utext_openReplaceable(UText *ut, Replaceable *rep, UErrorCode *status)
michael@0: {
michael@0:     if(U_FAILURE(*status)) {
michael@0:         return NULL;
michael@0:     }
michael@0:     if(rep==NULL) {
michael@0:         *status=U_ILLEGAL_ARGUMENT_ERROR;
michael@0:         return NULL;
michael@0:     }
michael@0:     ut = utext_setup(ut, sizeof(ReplExtra), status);
michael@0: 
michael@0:     ut->providerProperties = I32_FLAG(UTEXT_PROVIDER_WRITABLE);
michael@0:     if(rep->hasMetaData()) {
michael@0:         ut->providerProperties |=I32_FLAG(UTEXT_PROVIDER_HAS_META_DATA);
michael@0:     }
michael@0: 
michael@0:     ut->pFuncs  = &repFuncs;
michael@0:     ut->context =  rep;
michael@0:     return ut;
michael@0: }
michael@0: 
michael@0: U_CDECL_END
michael@0: 
michael@0: 
michael@0: 
michael@0: 
michael@0: 
michael@0: 
michael@0: 
michael@0: 
michael@0: //------------------------------------------------------------------------------
michael@0: //
michael@0: //     UText implementation for UnicodeString (read/write)  and
michael@0: //                    for const UnicodeString (read only)
michael@0: //             (same implementation, only the flags are different)
michael@0: //
michael@0: //         Use of UText data members:
michael@0: //            context    pointer to UnicodeString
michael@0: //            p          pointer to UnicodeString IF this UText owns the string
michael@0: //                       and it must be deleted on close().  NULL otherwise.
michael@0: //
michael@0: //------------------------------------------------------------------------------
michael@0: 
michael@0: U_CDECL_BEGIN
michael@0: 
michael@0: 
michael@0: static UText * U_CALLCONV
michael@0: unistrTextClone(UText *dest, const UText *src, UBool deep, UErrorCode *status) {
michael@0:     // First do a generic shallow clone.  Does everything needed for the UText struct itself.
michael@0:     dest = shallowTextClone(dest, src, status);
michael@0: 
michael@0:     // For deep clones, make a copy of the UnicodeSring.
michael@0:     //  The copied UnicodeString storage is owned by the newly created UText clone.
michael@0:     //  A non-NULL pointer in UText.p is the signal to the close() function to delete
michael@0:     //    the UText.
michael@0:     //
michael@0:     if (deep && U_SUCCESS(*status)) {
michael@0:         const UnicodeString *srcString = (const UnicodeString *)src->context;
michael@0:         dest->context = new UnicodeString(*srcString);
michael@0:         dest->providerProperties |= I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT);
michael@0: 
michael@0:         // with deep clone, the copy is writable, even when the source is not.
michael@0:         dest->providerProperties |= I32_FLAG(UTEXT_PROVIDER_WRITABLE);
michael@0:     }
michael@0:     return dest;
michael@0: }
michael@0: 
michael@0: static void U_CALLCONV
michael@0: unistrTextClose(UText *ut) {
michael@0:     // Most of the work of close is done by the generic UText framework close.
michael@0:     // All that needs to be done here is delete the UnicodeString if the UText
michael@0:     //  owns it.  This occurs if the UText was created by cloning.
michael@0:     if (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT)) {
michael@0:         UnicodeString *str = (UnicodeString *)ut->context;
michael@0:         delete str;
michael@0:         ut->context = NULL;
michael@0:     }
michael@0: }
michael@0: 
michael@0: 
michael@0: static int64_t U_CALLCONV
michael@0: unistrTextLength(UText *t) {
michael@0:     return ((const UnicodeString *)t->context)->length();
michael@0: }
michael@0: 
michael@0: 
michael@0: static UBool U_CALLCONV
michael@0: unistrTextAccess(UText *ut, int64_t index, UBool  forward) {
michael@0:     int32_t length  = ut->chunkLength;
michael@0:     ut->chunkOffset = pinIndex(index, length);
michael@0: 
michael@0:     // Check whether request is at the start or end
michael@0:     UBool retVal = (forward && index<length) || (!forward && index>0);
michael@0:     return retVal;
michael@0: }
michael@0: 
michael@0: 
michael@0: 
michael@0: static int32_t U_CALLCONV
michael@0: unistrTextExtract(UText *t,
michael@0:                   int64_t start, int64_t limit,
michael@0:                   UChar *dest, int32_t destCapacity,
michael@0:                   UErrorCode *pErrorCode) {
michael@0:     const UnicodeString *us=(const UnicodeString *)t->context;
michael@0:     int32_t length=us->length();
michael@0: 
michael@0:     if(U_FAILURE(*pErrorCode)) {
michael@0:         return 0;
michael@0:     }
michael@0:     if(destCapacity<0 || (dest==NULL && destCapacity>0)) {
michael@0:         *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
michael@0:     }
michael@0:     if(start<0 || start>limit) {
michael@0:         *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR;
michael@0:         return 0;
michael@0:     }
michael@0: 
michael@0:     int32_t start32 = start<length ? us->getChar32Start((int32_t)start) : length;
michael@0:     int32_t limit32 = limit<length ? us->getChar32Start((int32_t)limit) : length;
michael@0: 
michael@0:     length=limit32-start32;
michael@0:     if (destCapacity>0 && dest!=NULL) {
michael@0:         int32_t trimmedLength = length;
michael@0:         if(trimmedLength>destCapacity) {
michael@0:             trimmedLength=destCapacity;
michael@0:         }
michael@0:         us->extract(start32, trimmedLength, dest);
michael@0:         t->chunkOffset = start32+trimmedLength;
michael@0:     } else {
michael@0:         t->chunkOffset = start32;
michael@0:     }
michael@0:     u_terminateUChars(dest, destCapacity, length, pErrorCode);
michael@0:     return length;
michael@0: }
michael@0: 
michael@0: static int32_t U_CALLCONV
michael@0: unistrTextReplace(UText *ut,
michael@0:                   int64_t start, int64_t limit,
michael@0:                   const UChar *src, int32_t length,
michael@0:                   UErrorCode *pErrorCode) {
michael@0:     UnicodeString *us=(UnicodeString *)ut->context;
michael@0:     int32_t oldLength;
michael@0: 
michael@0:     if(U_FAILURE(*pErrorCode)) {
michael@0:         return 0;
michael@0:     }
michael@0:     if(src==NULL && length!=0) {
michael@0:         *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
michael@0:     }
michael@0:     if(start>limit) {
michael@0:         *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR;
michael@0:         return 0;
michael@0:     }
michael@0:     oldLength=us->length();
michael@0:     int32_t start32 = pinIndex(start, oldLength);
michael@0:     int32_t limit32 = pinIndex(limit, oldLength);
michael@0:     if (start32 < oldLength) {
michael@0:         start32 = us->getChar32Start(start32);
michael@0:     }
michael@0:     if (limit32 < oldLength) {
michael@0:         limit32 = us->getChar32Start(limit32);
michael@0:     }
michael@0: 
michael@0:     // replace
michael@0:     us->replace(start32, limit32-start32, src, length);
michael@0:     int32_t newLength = us->length();
michael@0: 
michael@0:     // Update the chunk description.
michael@0:     ut->chunkContents    = us->getBuffer();
michael@0:     ut->chunkLength      = newLength;
michael@0:     ut->chunkNativeLimit = newLength;
michael@0:     ut->nativeIndexingLimit = newLength;
michael@0: 
michael@0:     // Set iteration position to the point just following the newly inserted text.
michael@0:     int32_t lengthDelta = newLength - oldLength;
michael@0:     ut->chunkOffset = limit32 + lengthDelta;
michael@0: 
michael@0:     return lengthDelta;
michael@0: }
michael@0: 
michael@0: static void U_CALLCONV
michael@0: unistrTextCopy(UText *ut,
michael@0:                int64_t start, int64_t limit,
michael@0:                int64_t destIndex,
michael@0:                UBool move,
michael@0:                UErrorCode *pErrorCode) {
michael@0:     UnicodeString *us=(UnicodeString *)ut->context;
michael@0:     int32_t length=us->length();
michael@0: 
michael@0:     if(U_FAILURE(*pErrorCode)) {
michael@0:         return;
michael@0:     }
michael@0:     int32_t start32 = pinIndex(start, length);
michael@0:     int32_t limit32 = pinIndex(limit, length);
michael@0:     int32_t destIndex32 = pinIndex(destIndex, length);
michael@0: 
michael@0:     if( start32>limit32 || (start32<destIndex32 && destIndex32<limit32)) {
michael@0:         *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR;
michael@0:         return;
michael@0:     }
michael@0: 
michael@0:     if(move) {
michael@0:         // move: copy to destIndex, then replace original with nothing
michael@0:         int32_t segLength=limit32-start32;
michael@0:         us->copy(start32, limit32, destIndex32);
michael@0:         if(destIndex32<start32) {
michael@0:             start32+=segLength;
michael@0:         }
michael@0:         us->replace(start32, segLength, NULL, 0);
michael@0:     } else {
michael@0:         // copy
michael@0:         us->copy(start32, limit32, destIndex32);
michael@0:     }
michael@0: 
michael@0:     // update chunk description, set iteration position.
michael@0:     ut->chunkContents = us->getBuffer();
michael@0:     if (move==FALSE) {
michael@0:         // copy operation, string length grows
michael@0:         ut->chunkLength += limit32-start32;
michael@0:         ut->chunkNativeLimit = ut->chunkLength;
michael@0:         ut->nativeIndexingLimit = ut->chunkLength;
michael@0:     }
michael@0: 
michael@0:     // Iteration position to end of the newly inserted text.
michael@0:     ut->chunkOffset = destIndex32+limit32-start32;
michael@0:     if (move && destIndex32>start32) {
michael@0:         ut->chunkOffset = destIndex32;
michael@0:     }
michael@0: 
michael@0: }
michael@0: 
michael@0: static const struct UTextFuncs unistrFuncs = 
michael@0: {
michael@0:     sizeof(UTextFuncs),
michael@0:     0, 0, 0,             // Reserved alignment padding
michael@0:     unistrTextClone,
michael@0:     unistrTextLength,
michael@0:     unistrTextAccess,
michael@0:     unistrTextExtract,
michael@0:     unistrTextReplace,   
michael@0:     unistrTextCopy,   
michael@0:     NULL,                // MapOffsetToNative,
michael@0:     NULL,                // MapIndexToUTF16,
michael@0:     unistrTextClose,
michael@0:     NULL,                // spare 1
michael@0:     NULL,                // spare 2
michael@0:     NULL                 // spare 3
michael@0: };
michael@0: 
michael@0: 
michael@0: 
michael@0: U_CDECL_END
michael@0: 
michael@0: 
michael@0: U_CAPI UText * U_EXPORT2
michael@0: utext_openUnicodeString(UText *ut, UnicodeString *s, UErrorCode *status) {
michael@0:     ut = utext_openConstUnicodeString(ut, s, status);
michael@0:     if (U_SUCCESS(*status)) {
michael@0:         ut->providerProperties |= I32_FLAG(UTEXT_PROVIDER_WRITABLE);
michael@0:     }
michael@0:     return ut;
michael@0: }
michael@0: 
michael@0: 
michael@0: 
michael@0: U_CAPI UText * U_EXPORT2
michael@0: utext_openConstUnicodeString(UText *ut, const UnicodeString *s, UErrorCode *status) {
michael@0:     if (U_SUCCESS(*status) && s->isBogus()) {
michael@0:         // The UnicodeString is bogus, but we still need to detach the UText
michael@0:         //   from whatever it was hooked to before, if anything.
michael@0:         utext_openUChars(ut, NULL, 0, status);
michael@0:         *status = U_ILLEGAL_ARGUMENT_ERROR;
michael@0:         return ut;
michael@0:     }
michael@0:     ut = utext_setup(ut, 0, status);
michael@0:     //    note:  use the standard (writable) function table for UnicodeString.
michael@0:     //           The flag settings disable writing, so having the functions in
michael@0:     //           the table is harmless.
michael@0:     if (U_SUCCESS(*status)) {
michael@0:         ut->pFuncs              = &unistrFuncs;
michael@0:         ut->context             = s;
michael@0:         ut->providerProperties  = I32_FLAG(UTEXT_PROVIDER_STABLE_CHUNKS);
michael@0:         ut->chunkContents       = s->getBuffer();
michael@0:         ut->chunkLength         = s->length();
michael@0:         ut->chunkNativeStart    = 0;
michael@0:         ut->chunkNativeLimit    = ut->chunkLength;
michael@0:         ut->nativeIndexingLimit = ut->chunkLength;
michael@0:     }
michael@0:     return ut;
michael@0: }
michael@0: 
michael@0: //------------------------------------------------------------------------------
michael@0: //
michael@0: //     UText implementation for const UChar * strings
michael@0: //
michael@0: //         Use of UText data members:
michael@0: //            context    pointer to UnicodeString
michael@0: //            a          length.  -1 if not yet known.
michael@0: //
michael@0: //         TODO:  support 64 bit lengths.
michael@0: //
michael@0: //------------------------------------------------------------------------------
michael@0: 
michael@0: U_CDECL_BEGIN
michael@0: 
michael@0: 
michael@0: static UText * U_CALLCONV
michael@0: ucstrTextClone(UText *dest, const UText * src, UBool deep, UErrorCode * status) {
michael@0:     // First do a generic shallow clone.
michael@0:     dest = shallowTextClone(dest, src, status);
michael@0: 
michael@0:     // For deep clones, make a copy of the string.
michael@0:     //  The copied storage is owned by the newly created clone.
michael@0:     //  A non-NULL pointer in UText.p is the signal to the close() function to delete
michael@0:     //    it.
michael@0:     //
michael@0:     if (deep && U_SUCCESS(*status)) {
michael@0:         U_ASSERT(utext_nativeLength(dest) < INT32_MAX);
michael@0:         int32_t  len = (int32_t)utext_nativeLength(dest);
michael@0: 
michael@0:         // The cloned string IS going to be NUL terminated, whether or not the original was.
michael@0:         const UChar *srcStr = (const UChar *)src->context;
michael@0:         UChar *copyStr = (UChar *)uprv_malloc((len+1) * sizeof(UChar));
michael@0:         if (copyStr == NULL) {
michael@0:             *status = U_MEMORY_ALLOCATION_ERROR;
michael@0:         } else {
michael@0:             int64_t i;
michael@0:             for (i=0; i<len; i++) {
michael@0:                 copyStr[i] = srcStr[i];
michael@0:             }
michael@0:             copyStr[len] = 0;
michael@0:             dest->context = copyStr;
michael@0:             dest->providerProperties |= I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT);
michael@0:         }
michael@0:     }
michael@0:     return dest;
michael@0: }
michael@0: 
michael@0: 
michael@0: static void U_CALLCONV
michael@0: ucstrTextClose(UText *ut) {
michael@0:     // Most of the work of close is done by the generic UText framework close.
michael@0:     // All that needs to be done here is delete the string if the UText
michael@0:     //  owns it.  This occurs if the UText was created by cloning.
michael@0:     if (ut->providerProperties & I32_FLAG(UTEXT_PROVIDER_OWNS_TEXT)) {
michael@0:         UChar *s = (UChar *)ut->context;
michael@0:         uprv_free(s);
michael@0:         ut->context = NULL;
michael@0:     }
michael@0: }
michael@0: 
michael@0: 
michael@0: 
michael@0: static int64_t U_CALLCONV
michael@0: ucstrTextLength(UText *ut) {
michael@0:     if (ut->a < 0) {
michael@0:         // null terminated, we don't yet know the length.  Scan for it.
michael@0:         //    Access is not convenient for doing this
michael@0:         //    because the current interation postion can't be changed.
michael@0:         const UChar  *str = (const UChar *)ut->context;
michael@0:         for (;;) {
michael@0:             if (str[ut->chunkNativeLimit] == 0) {
michael@0:                 break;
michael@0:             }
michael@0:             ut->chunkNativeLimit++;
michael@0:         }
michael@0:         ut->a = ut->chunkNativeLimit;
michael@0:         ut->chunkLength = (int32_t)ut->chunkNativeLimit;
michael@0:         ut->nativeIndexingLimit = ut->chunkLength;
michael@0:         ut->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
michael@0:     }
michael@0:     return ut->a;
michael@0: }
michael@0: 
michael@0: 
michael@0: static UBool U_CALLCONV
michael@0: ucstrTextAccess(UText *ut, int64_t index, UBool  forward) {
michael@0:     const UChar *str   = (const UChar *)ut->context;
michael@0: 
michael@0:     // pin the requested index to the bounds of the string,
michael@0:     //  and set current iteration position.
michael@0:     if (index<0) {
michael@0:         index = 0;
michael@0:     } else if (index < ut->chunkNativeLimit) {
michael@0:         // The request data is within the chunk as it is known so far.
michael@0:         // Put index on a code point boundary.
michael@0:         U16_SET_CP_START(str, 0, index);
michael@0:     } else if (ut->a >= 0) {
michael@0:         // We know the length of this string, and the user is requesting something
michael@0:         // at or beyond the length.  Pin the requested index to the length.
michael@0:         index = ut->a;
michael@0:     } else {
michael@0:         // Null terminated string, length not yet known, and the requested index
michael@0:         //  is beyond where we have scanned so far.
michael@0:         //  Scan to 32 UChars beyond the requested index.  The strategy here is
michael@0:         //  to avoid fully scanning a long string when the caller only wants to
michael@0:         //  see a few characters at its beginning.
michael@0:         int32_t scanLimit = (int32_t)index + 32;
michael@0:         if ((index + 32)>INT32_MAX || (index + 32)<0 ) {   // note: int64 expression
michael@0:             scanLimit = INT32_MAX;
michael@0:         }
michael@0: 
michael@0:         int32_t chunkLimit = (int32_t)ut->chunkNativeLimit;
michael@0:         for (; chunkLimit<scanLimit; chunkLimit++) {
michael@0:             if (str[chunkLimit] == 0) {
michael@0:                 // We found the end of the string.  Remember it, pin the requested index to it,
michael@0:                 //  and bail out of here.
michael@0:                 ut->a = chunkLimit;
michael@0:                 ut->chunkLength = chunkLimit;
michael@0:                 ut->nativeIndexingLimit = chunkLimit;
michael@0:                 if (index >= chunkLimit) {
michael@0:                     index = chunkLimit;
michael@0:                 } else {
michael@0:                     U16_SET_CP_START(str, 0, index);
michael@0:                 }
michael@0: 
michael@0:                 ut->chunkNativeLimit = chunkLimit;
michael@0:                 ut->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
michael@0:                 goto breakout;
michael@0:             }
michael@0:         }
michael@0:         // We scanned through the next batch of UChars without finding the end.
michael@0:         U16_SET_CP_START(str, 0, index);
michael@0:         if (chunkLimit == INT32_MAX) {
michael@0:             // Scanned to the limit of a 32 bit length.
michael@0:             // Forceably trim the overlength string back so length fits in int32
michael@0:             //  TODO:  add support for 64 bit strings.
michael@0:             ut->a = chunkLimit;
michael@0:             ut->chunkLength = chunkLimit;
michael@0:             ut->nativeIndexingLimit = chunkLimit;
michael@0:             if (index > chunkLimit) {
michael@0:                 index = chunkLimit;
michael@0:             }
michael@0:             ut->chunkNativeLimit = chunkLimit;
michael@0:             ut->providerProperties &= ~I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
michael@0:         } else {
michael@0:             // The endpoint of a chunk must not be left in the middle of a surrogate pair.
michael@0:             // If the current end is on a lead surrogate, back the end up by one.
michael@0:             // It doesn't matter if the end char happens to be an unpaired surrogate,
michael@0:             //    and it's simpler not to worry about it.
michael@0:             if (U16_IS_LEAD(str[chunkLimit-1])) {
michael@0:                 --chunkLimit;
michael@0:             }
michael@0:             // Null-terminated chunk with end still unknown.
michael@0:             // Update the chunk length to reflect what has been scanned thus far.
michael@0:             // That the full length is still unknown is (still) flagged by
michael@0:             //    ut->a being < 0.
michael@0:             ut->chunkNativeLimit = chunkLimit;
michael@0:             ut->nativeIndexingLimit = chunkLimit;
michael@0:             ut->chunkLength = chunkLimit;
michael@0:         }
michael@0: 
michael@0:     }
michael@0: breakout:
michael@0:     U_ASSERT(index<=INT32_MAX);
michael@0:     ut->chunkOffset = (int32_t)index;
michael@0: 
michael@0:     // Check whether request is at the start or end
michael@0:     UBool retVal = (forward && index<ut->chunkNativeLimit) || (!forward && index>0);
michael@0:     return retVal;
michael@0: }
michael@0: 
michael@0: 
michael@0: 
michael@0: static int32_t U_CALLCONV
michael@0: ucstrTextExtract(UText *ut,
michael@0:                   int64_t start, int64_t limit,
michael@0:                   UChar *dest, int32_t destCapacity,
michael@0:                   UErrorCode *pErrorCode)
michael@0: {
michael@0:     if(U_FAILURE(*pErrorCode)) {
michael@0:         return 0;
michael@0:     }
michael@0:     if(destCapacity<0 || (dest==NULL && destCapacity>0) || start>limit) {
michael@0:         *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR;
michael@0:         return 0;
michael@0:     }
michael@0: 
michael@0:     //const UChar *s=(const UChar *)ut->context;
michael@0:     int32_t si, di;
michael@0: 
michael@0:     int32_t start32;
michael@0:     int32_t limit32;
michael@0: 
michael@0:     // Access the start.  Does two things we need:
michael@0:     //   Pins 'start' to the length of the string, if it came in out-of-bounds.
michael@0:     //   Snaps 'start' to the beginning of a code point.
michael@0:     ucstrTextAccess(ut, start, TRUE);
michael@0:     const UChar *s=ut->chunkContents;
michael@0:     start32 = ut->chunkOffset;
michael@0: 
michael@0:     int32_t strLength=(int32_t)ut->a;
michael@0:     if (strLength >= 0) {
michael@0:         limit32 = pinIndex(limit, strLength);
michael@0:     } else {
michael@0:         limit32 = pinIndex(limit, INT32_MAX);
michael@0:     }
michael@0:     di = 0;
michael@0:     for (si=start32; si<limit32; si++) {
michael@0:         if (strLength<0 && s[si]==0) {
michael@0:             // Just hit the end of a null-terminated string.
michael@0:             ut->a = si;               // set string length for this UText
michael@0:             ut->chunkNativeLimit    = si;
michael@0:             ut->chunkLength         = si;
michael@0:             ut->nativeIndexingLimit = si;
michael@0:             strLength               = si;
michael@0:             break;
michael@0:         }
michael@0:         U_ASSERT(di>=0); /* to ensure di never exceeds INT32_MAX, which must not happen logically */
michael@0:         if (di<destCapacity) {
michael@0:             // only store if there is space.
michael@0:             dest[di] = s[si];
michael@0:         } else {
michael@0:             if (strLength>=0) {
michael@0:                 // We have filled the destination buffer, and the string length is known.
michael@0:                 //  Cut the loop short.  There is no need to scan string termination.
michael@0:                 di = limit32 - start32;
michael@0:                 si = limit32;
michael@0:                 break;
michael@0:             }
michael@0:         }
michael@0:         di++;
michael@0:     }
michael@0: 
michael@0:     // If the limit index points to a lead surrogate of a pair,
michael@0:     //   add the corresponding trail surrogate to the destination.
michael@0:     if (si>0 && U16_IS_LEAD(s[si-1]) &&
michael@0:         ((si<strLength || strLength<0)  && U16_IS_TRAIL(s[si])))
michael@0:     {
michael@0:         if (di<destCapacity) {
michael@0:             // store only if there is space in the output buffer.
michael@0:             dest[di++] = s[si++];
michael@0:         }
michael@0:     }
michael@0: 
michael@0:     // Put iteration position at the point just following the extracted text
michael@0:     ut->chunkOffset = uprv_min(strLength, start32 + destCapacity);
michael@0: 
michael@0:     // Add a terminating NUL if space in the buffer permits,
michael@0:     // and set the error status as required.
michael@0:     u_terminateUChars(dest, destCapacity, di, pErrorCode);
michael@0:     return di;
michael@0: }
michael@0: 
michael@0: static const struct UTextFuncs ucstrFuncs = 
michael@0: {
michael@0:     sizeof(UTextFuncs),
michael@0:     0, 0, 0,           // Reserved alignment padding
michael@0:     ucstrTextClone,
michael@0:     ucstrTextLength,
michael@0:     ucstrTextAccess,
michael@0:     ucstrTextExtract,
michael@0:     NULL,              // Replace
michael@0:     NULL,              // Copy
michael@0:     NULL,              // MapOffsetToNative,
michael@0:     NULL,              // MapIndexToUTF16,
michael@0:     ucstrTextClose,
michael@0:     NULL,              // spare 1
michael@0:     NULL,              // spare 2
michael@0:     NULL,              // spare 3
michael@0: };
michael@0: 
michael@0: U_CDECL_END
michael@0: 
michael@0: static const UChar gEmptyUString[] = {0};
michael@0: 
michael@0: U_CAPI UText * U_EXPORT2
michael@0: utext_openUChars(UText *ut, const UChar *s, int64_t length, UErrorCode *status) {
michael@0:     if (U_FAILURE(*status)) {
michael@0:         return NULL;
michael@0:     }
michael@0:     if(s==NULL && length==0) {
michael@0:         s = gEmptyUString;
michael@0:     }
michael@0:     if (s==NULL || length < -1 || length>INT32_MAX) {
michael@0:         *status = U_ILLEGAL_ARGUMENT_ERROR;
michael@0:         return NULL;
michael@0:     }
michael@0:     ut = utext_setup(ut, 0, status);
michael@0:     if (U_SUCCESS(*status)) {
michael@0:         ut->pFuncs               = &ucstrFuncs;
michael@0:         ut->context              = s;
michael@0:         ut->providerProperties   = I32_FLAG(UTEXT_PROVIDER_STABLE_CHUNKS);
michael@0:         if (length==-1) {
michael@0:             ut->providerProperties |= I32_FLAG(UTEXT_PROVIDER_LENGTH_IS_EXPENSIVE);
michael@0:         }
michael@0:         ut->a                    = length;
michael@0:         ut->chunkContents        = s;
michael@0:         ut->chunkNativeStart     = 0;
michael@0:         ut->chunkNativeLimit     = length>=0? length : 0;
michael@0:         ut->chunkLength          = (int32_t)ut->chunkNativeLimit;
michael@0:         ut->chunkOffset          = 0;
michael@0:         ut->nativeIndexingLimit  = ut->chunkLength;
michael@0:     }
michael@0:     return ut;
michael@0: }
michael@0: 
michael@0: 
michael@0: //------------------------------------------------------------------------------
michael@0: //
michael@0: //     UText implementation for text from ICU CharacterIterators
michael@0: //
michael@0: //         Use of UText data members:
michael@0: //            context    pointer to the CharacterIterator
michael@0: //            a          length of the full text.
michael@0: //            p          pointer to  buffer 1
michael@0: //            b          start index of local buffer 1 contents
michael@0: //            q          pointer to buffer 2
michael@0: //            c          start index of local buffer 2 contents
michael@0: //            r          pointer to the character iterator if the UText owns it.
michael@0: //                       Null otherwise.
michael@0: //
michael@0: //------------------------------------------------------------------------------
michael@0: #define CIBufSize 16
michael@0: 
michael@0: U_CDECL_BEGIN
michael@0: static void U_CALLCONV
michael@0: charIterTextClose(UText *ut) {
michael@0:     // Most of the work of close is done by the generic UText framework close.
michael@0:     // All that needs to be done here is delete the CharacterIterator if the UText
michael@0:     //  owns it.  This occurs if the UText was created by cloning.
michael@0:     CharacterIterator *ci = (CharacterIterator *)ut->r;
michael@0:     delete ci;
michael@0:     ut->r = NULL;
michael@0: }
michael@0: 
michael@0: static int64_t U_CALLCONV
michael@0: charIterTextLength(UText *ut) {
michael@0:     return (int32_t)ut->a;
michael@0: }
michael@0: 
michael@0: static UBool U_CALLCONV
michael@0: charIterTextAccess(UText *ut, int64_t index, UBool  forward) {
michael@0:     CharacterIterator *ci   = (CharacterIterator *)ut->context;
michael@0: 
michael@0:     int32_t clippedIndex = (int32_t)index;
michael@0:     if (clippedIndex<0) {
michael@0:         clippedIndex=0;
michael@0:     } else if (clippedIndex>=ut->a) {
michael@0:         clippedIndex=(int32_t)ut->a;
michael@0:     }
michael@0:     int32_t neededIndex = clippedIndex;
michael@0:     if (!forward && neededIndex>0) {
michael@0:         // reverse iteration, want the position just before what was asked for.
michael@0:         neededIndex--;
michael@0:     } else if (forward && neededIndex==ut->a && neededIndex>0) {
michael@0:         // Forward iteration, don't ask for something past the end of the text.
michael@0:         neededIndex--;
michael@0:     }
michael@0: 
michael@0:     // Find the native index of the start of the buffer containing what we want.
michael@0:     neededIndex -= neededIndex % CIBufSize;
michael@0: 
michael@0:     UChar *buf = NULL;
michael@0:     UBool  needChunkSetup = TRUE;
michael@0:     int    i;
michael@0:     if (ut->chunkNativeStart == neededIndex) {
michael@0:         // The buffer we want is already the current chunk.
michael@0:         needChunkSetup = FALSE;
michael@0:     } else if (ut->b == neededIndex) {
michael@0:         // The first buffer (buffer p) has what we need.
michael@0:         buf = (UChar *)ut->p;
michael@0:     } else if (ut->c == neededIndex) {
michael@0:         // The second buffer (buffer q) has what we need.
michael@0:         buf = (UChar *)ut->q;
michael@0:     } else {
michael@0:         // Neither buffer already has what we need.
michael@0:         // Load new data from the character iterator.
michael@0:         // Use the buf that is not the current buffer.
michael@0:         buf = (UChar *)ut->p;
michael@0:         if (ut->p == ut->chunkContents) {
michael@0:             buf = (UChar *)ut->q;
michael@0:         }
michael@0:         ci->setIndex(neededIndex);
michael@0:         for (i=0; i<CIBufSize; i++) {
michael@0:             buf[i] = ci->nextPostInc();
michael@0:             if (i+neededIndex > ut->a) {
michael@0:                 break;
michael@0:             }
michael@0:         }
michael@0:     }
michael@0: 
michael@0:     // We have a buffer with the data we need.
michael@0:     // Set it up as the current chunk, if it wasn't already.
michael@0:     if (needChunkSetup) {
michael@0:         ut->chunkContents = buf;
michael@0:         ut->chunkLength   = CIBufSize;
michael@0:         ut->chunkNativeStart = neededIndex;
michael@0:         ut->chunkNativeLimit = neededIndex + CIBufSize;
michael@0:         if (ut->chunkNativeLimit > ut->a) {
michael@0:             ut->chunkNativeLimit = ut->a;
michael@0:             ut->chunkLength  = (int32_t)(ut->chunkNativeLimit)-(int32_t)(ut->chunkNativeStart);
michael@0:         }
michael@0:         ut->nativeIndexingLimit = ut->chunkLength;
michael@0:         U_ASSERT(ut->chunkOffset>=0 && ut->chunkOffset<=CIBufSize);
michael@0:     }
michael@0:     ut->chunkOffset = clippedIndex - (int32_t)ut->chunkNativeStart;
michael@0:     UBool success = (forward? ut->chunkOffset<ut->chunkLength : ut->chunkOffset>0);
michael@0:     return success;
michael@0: }
michael@0: 
michael@0: static UText * U_CALLCONV
michael@0: charIterTextClone(UText *dest, const UText *src, UBool deep, UErrorCode * status) {
michael@0:     if (U_FAILURE(*status)) {
michael@0:         return NULL;
michael@0:     }
michael@0: 
michael@0:     if (deep) {
michael@0:         // There is no CharacterIterator API for cloning the underlying text storage.
michael@0:         *status = U_UNSUPPORTED_ERROR;
michael@0:         return NULL;
michael@0:     } else {
michael@0:         CharacterIterator *srcCI =(CharacterIterator *)src->context;
michael@0:         srcCI = srcCI->clone();
michael@0:         dest = utext_openCharacterIterator(dest, srcCI, status);
michael@0:         // cast off const on getNativeIndex.
michael@0:         //   For CharacterIterator based UTexts, this is safe, the operation is const.
michael@0:         int64_t  ix = utext_getNativeIndex((UText *)src);
michael@0:         utext_setNativeIndex(dest, ix);
michael@0:         dest->r = srcCI;    // flags that this UText owns the CharacterIterator
michael@0:     }
michael@0:     return dest;
michael@0: }
michael@0: 
michael@0: static int32_t U_CALLCONV
michael@0: charIterTextExtract(UText *ut,
michael@0:                   int64_t start, int64_t limit,
michael@0:                   UChar *dest, int32_t destCapacity,
michael@0:                   UErrorCode *status)
michael@0: {
michael@0:     if(U_FAILURE(*status)) {
michael@0:         return 0;
michael@0:     }
michael@0:     if(destCapacity<0 || (dest==NULL && destCapacity>0) || start>limit) {
michael@0:         *status=U_ILLEGAL_ARGUMENT_ERROR;
michael@0:         return 0;
michael@0:     }
michael@0:     int32_t  length  = (int32_t)ut->a;
michael@0:     int32_t  start32 = pinIndex(start, length);
michael@0:     int32_t  limit32 = pinIndex(limit, length);
michael@0:     int32_t  desti   = 0;
michael@0:     int32_t  srci;
michael@0:     int32_t  copyLimit;
michael@0: 
michael@0:     CharacterIterator *ci = (CharacterIterator *)ut->context;
michael@0:     ci->setIndex32(start32);   // Moves ix to lead of surrogate pair, if needed.
michael@0:     srci = ci->getIndex();
michael@0:     copyLimit = srci;
michael@0:     while (srci<limit32) {
michael@0:         UChar32 c = ci->next32PostInc();
michael@0:         int32_t  len = U16_LENGTH(c);
michael@0:         U_ASSERT(desti+len>0); /* to ensure desti+len never exceeds MAX_INT32, which must not happen logically */
michael@0:         if (desti+len <= destCapacity) {
michael@0:             U16_APPEND_UNSAFE(dest, desti, c);
michael@0:             copyLimit = srci+len;
michael@0:         } else {
michael@0:             desti += len;
michael@0:             *status = U_BUFFER_OVERFLOW_ERROR;
michael@0:         }
michael@0:         srci += len;
michael@0:     }
michael@0:     
michael@0:     charIterTextAccess(ut, copyLimit, TRUE);
michael@0: 
michael@0:     u_terminateUChars(dest, destCapacity, desti, status);
michael@0:     return desti;
michael@0: }
michael@0: 
michael@0: static const struct UTextFuncs charIterFuncs = 
michael@0: {
michael@0:     sizeof(UTextFuncs),
michael@0:     0, 0, 0,             // Reserved alignment padding
michael@0:     charIterTextClone,
michael@0:     charIterTextLength,
michael@0:     charIterTextAccess,
michael@0:     charIterTextExtract,
michael@0:     NULL,                // Replace
michael@0:     NULL,                // Copy
michael@0:     NULL,                // MapOffsetToNative,
michael@0:     NULL,                // MapIndexToUTF16,
michael@0:     charIterTextClose,
michael@0:     NULL,                // spare 1
michael@0:     NULL,                // spare 2
michael@0:     NULL                 // spare 3
michael@0: };
michael@0: U_CDECL_END
michael@0: 
michael@0: 
michael@0: U_CAPI UText * U_EXPORT2
michael@0: utext_openCharacterIterator(UText *ut, CharacterIterator *ci, UErrorCode *status) {
michael@0:     if (U_FAILURE(*status)) {
michael@0:         return NULL;
michael@0:     }
michael@0: 
michael@0:     if (ci->startIndex() > 0) {
michael@0:         // No support for CharacterIterators that do not start indexing from zero.
michael@0:         *status = U_UNSUPPORTED_ERROR;
michael@0:         return NULL;
michael@0:     }
michael@0: 
michael@0:     // Extra space in UText for 2 buffers of CIBufSize UChars each.
michael@0:     int32_t  extraSpace = 2 * CIBufSize * sizeof(UChar);
michael@0:     ut = utext_setup(ut, extraSpace, status);
michael@0:     if (U_SUCCESS(*status)) {
michael@0:         ut->pFuncs                = &charIterFuncs;
michael@0:         ut->context              = ci;
michael@0:         ut->providerProperties   = 0;
michael@0:         ut->a                    = ci->endIndex();        // Length of text
michael@0:         ut->p                    = ut->pExtra;            // First buffer
michael@0:         ut->b                    = -1;                    // Native index of first buffer contents
michael@0:         ut->q                    = (UChar*)ut->pExtra+CIBufSize;  // Second buffer
michael@0:         ut->c                    = -1;                    // Native index of second buffer contents
michael@0: 
michael@0:         // Initialize current chunk contents to be empty.
michael@0:         //   First access will fault something in.
michael@0:         //   Note:  The initial nativeStart and chunkOffset must sum to zero
michael@0:         //          so that getNativeIndex() will correctly compute to zero
michael@0:         //          if no call to Access() has ever been made.  They can't be both
michael@0:         //          zero without Access() thinking that the chunk is valid.
michael@0:         ut->chunkContents        = (UChar *)ut->p;
michael@0:         ut->chunkNativeStart     = -1;
michael@0:         ut->chunkOffset          = 1;
michael@0:         ut->chunkNativeLimit     = 0;
michael@0:         ut->chunkLength          = 0;
michael@0:         ut->nativeIndexingLimit  = ut->chunkOffset;  // enables native indexing
michael@0:     }
michael@0:     return ut;
michael@0: }
michael@0: 
michael@0: 
michael@0: