+ * loop for conversion and error handling
+ *
+ * loop {
+ * convert
+ * loop {
+ * update offsets
+ * handle end of input
+ * handle errors/call callback
+ * }
+ * }
+ */
+ for(;;) {
+ /* convert */
+ fromUnicode(pArgs, err);
+
+ /*
+ * set a flag for whether the converter
+ * successfully processed the end of the input
+ *
+ * need not check cnv->preFromULength==0 because a replay (<0) will cause
+ * s<sourceLimit before converterSawEndOfInput is checked
+ */
+ converterSawEndOfInput=
+ (UBool)(U_SUCCESS(*err) &&
+ pArgs->flush && pArgs->source==pArgs->sourceLimit &&
+ cnv->fromUChar32==0);
+
+ /* no callback called yet for this iteration */
+ calledCallback=FALSE;
+
+ /* no sourceIndex adjustment for conversion, only for callback output */
+ errorInputLength=0;
+
+ /*
+ * loop for offsets and error handling
+ *
+ * iterates at most 3 times:
+ * 1. to clean up after the conversion function
+ * 2. after the callback
+ * 3. after the callback again if there was truncated input
+ */
+ for(;;) {
+ /* update offsets if we write any */
+ if(offsets!=NULL) {
+ int32_t length=(int32_t)(pArgs->target-t);
+ if(length>0) {
+ _updateOffsets(offsets, length, sourceIndex, errorInputLength);
+
+ /*
+ * if a converter handles offsets and updates the offsets
+ * pointer at the end, then pArgs->offset should not change
+ * here;
+ * however, some converters do not handle offsets at all
+ * (sourceIndex<0) or may not update the offsets pointer
+ */
+ pArgs->offsets=offsets+=length;
+ }
+
+ if(sourceIndex>=0) {
+ sourceIndex+=(int32_t)(pArgs->source-s);
+ }
+ }
+
+ if(cnv->preFromULength<0) {
+ /*
+ * switch the source to new replay units (cannot occur while replaying)
+ * after offset handling and before end-of-input and callback handling
+ */
+ if(realSource==NULL) {
+ realSource=pArgs->source;
+ realSourceLimit=pArgs->sourceLimit;
+ realFlush=pArgs->flush;
+ realSourceIndex=sourceIndex;
+
+ uprv_memcpy(replay, cnv->preFromU, -cnv->preFromULength*U_SIZEOF_UCHAR);
+ pArgs->source=replay;
+ pArgs->sourceLimit=replay-cnv->preFromULength;
+ pArgs->flush=FALSE;
+ if((sourceIndex+=cnv->preFromULength)<0) {
+ sourceIndex=-1;
+ }
+
+ cnv->preFromULength=0;
+ } else {
+ /* see implementation note before _fromUnicodeWithCallback() */
+ U_ASSERT(realSource==NULL);
+ *err=U_INTERNAL_PROGRAM_ERROR;
+ }
+ }
+
+ /* update pointers */
+ s=pArgs->source;
+ t=pArgs->target;
+
+ if(U_SUCCESS(*err)) {
+ if(s<pArgs->sourceLimit) {
+ /*
+ * continue with the conversion loop while there is still input left
+ * (continue converting by breaking out of only the inner loop)
+ */
+ break;
+ } else if(realSource!=NULL) {
+ /* switch back from replaying to the real source and continue */
+ pArgs->source=realSource;
+ pArgs->sourceLimit=realSourceLimit;
+ pArgs->flush=realFlush;
+ sourceIndex=realSourceIndex;
+
+ realSource=NULL;
+ break;
+ } else if(pArgs->flush && cnv->fromUChar32!=0) {
+ /*
+ * the entire input stream is consumed
+ * and there is a partial, truncated input sequence left
+ */
+
+ /* inject an error and continue with callback handling */
+ *err=U_TRUNCATED_CHAR_FOUND;
+ calledCallback=FALSE; /* new error condition */
+ } else {
+ /* input consumed */
+ if(pArgs->flush) {
+ /*
+ * return to the conversion loop once more if the flush
+ * flag is set and the conversion function has not
+ * successfully processed the end of the input yet
+ *
+ * (continue converting by breaking out of only the inner loop)
+ */
+ if(!converterSawEndOfInput) {
+ break;
+ }
+
+ /* reset the converter without calling the callback function */
+ _reset(cnv, UCNV_RESET_FROM_UNICODE, FALSE);
+ }
+
+ /* done successfully */
+ return;
+ }
+ }
+
+ /* U_FAILURE(*err) */
+ {
+ UErrorCode e;
+
+ if( calledCallback ||
+ (e=*err)==U_BUFFER_OVERFLOW_ERROR ||
+ (e!=U_INVALID_CHAR_FOUND &&
+ e!=U_ILLEGAL_CHAR_FOUND &&
+ e!=U_TRUNCATED_CHAR_FOUND)
+ ) {
+ /*
+ * the callback did not or cannot resolve the error:
+ * set output pointers and return
+ *
+ * the check for buffer overflow is redundant but it is
+ * a high-runner case and hopefully documents the intent
+ * well
+ *
+ * if we were replaying, then the replay buffer must be
+ * copied back into the UConverter
+ * and the real arguments must be restored
+ */
+ if(realSource!=NULL) {
+ int32_t length;
+
+ U_ASSERT(cnv->preFromULength==0);
+
+ length=(int32_t)(pArgs->sourceLimit-pArgs->source);
+ if(length>0) {
+ uprv_memcpy(cnv->preFromU, pArgs->source, length*U_SIZEOF_UCHAR);
+ cnv->preFromULength=(int8_t)-length;
+ }
+
+ pArgs->source=realSource;
+ pArgs->sourceLimit=realSourceLimit;
+ pArgs->flush=realFlush;
+ }
+
+ return;
+ }
+ }
+
+ /* callback handling */
+ {
+ UChar32 codePoint;
+
+ /* get and write the code point */
+ codePoint=cnv->fromUChar32;
+ errorInputLength=0;
+ U16_APPEND_UNSAFE(cnv->invalidUCharBuffer, errorInputLength, codePoint);
+ cnv->invalidUCharLength=(int8_t)errorInputLength;
+
+ /* set the converter state to deal with the next character */
+ cnv->fromUChar32=0;
+
+ /* call the callback function */
+ cnv->fromUCharErrorBehaviour(cnv->fromUContext, pArgs,
+ cnv->invalidUCharBuffer, errorInputLength, codePoint,
+ *err==U_INVALID_CHAR_FOUND ? UCNV_UNASSIGNED : UCNV_ILLEGAL,
+ err);
+ }
+
+ /*
+ * loop back to the offset handling
+ *
+ * this flag will indicate after offset handling
+ * that a callback was called;
+ * if the callback did not resolve the error, then we return
+ */
+ calledCallback=TRUE;
+ }
+ }
+}
+
+U_CAPI void U_EXPORT2
+ucnv_fromUnicode(UConverter *cnv,
+ char **target, const char *targetLimit,
+ const UChar **source, const UChar *sourceLimit,
+ int32_t *offsets,
+ UBool flush,
+ UErrorCode *err) {
+ UConverterFromUnicodeArgs args;
+ const UChar *s;
+ char *t;
+
+ /* check parameters */
+ if(err==NULL || U_FAILURE(*err)) {
projects / apple / icu.git / blobdiff