X-Git-Url: https://git.saurik.com/apple/icu.git/blobdiff_plain/b75a7d8f3b4adbae880cab104ce2c6a50eee4db2..b25be06635768807f8f693286fa73bb2297bb06c:/icuSources/common/ucnvmbcs.c?ds=inline diff --git a/icuSources/common/ucnvmbcs.c b/icuSources/common/ucnvmbcs.c index 0c283ecc..538a18c5 100644 --- a/icuSources/common/ucnvmbcs.c +++ b/icuSources/common/ucnvmbcs.c @@ -1,7 +1,7 @@ /* ****************************************************************************** * -* Copyright (C) 2000-2003, International Business Machines +* Copyright (C) 2000-2010, International Business Machines * Corporation and others. All Rights Reserved. * ****************************************************************************** @@ -21,7 +21,7 @@ * - efficient distinction of unassigned vs. illegal byte sequences * - it is possible in fromUnicode() to directly deal with simple * stateful encodings (used for EBCDIC_STATEFUL) -* - it is possible to convert Unicode code points other than U+0000 +* - it is possible to convert Unicode code points * to a single zero byte (but not as a fallback except for SBCS) * * Remaining limitations in fromUnicode: @@ -29,6 +29,10 @@ * - except for SBCS codepages: no fallback mapping from Unicode to a zero byte * - limitation to up to 4 bytes per character * +* ICU 2.8 (late 2003) adds a secondary data structure which lifts some of these +* limitations and adds m:n character mappings and other features. +* See ucnv_ext.h for details. +* * Change history: * * 5/6/2001 Ram Moved MBCS_SINGLE_RESULT_FROM_U,MBCS_STAGE_2_FROM_U, @@ -38,7 +42,7 @@ #include "unicode/utypes.h" -#if !UCONFIG_NO_LEGACY_CONVERSION +#if !UCONFIG_NO_CONVERSION && !UCONFIG_NO_LEGACY_CONVERSION #include "unicode/ucnv.h" #include "unicode/ucnv_cb.h" @@ -46,6 +50,7 @@ #include "unicode/uset.h" #include "ucnv_bld.h" #include "ucnvmbcs.h" +#include "ucnv_ext.h" #include "ucnv_cnv.h" #include "umutex.h" #include "cmemory.h" @@ -56,9 +61,64 @@ #define MBCS_UNROLL_SINGLE_FROM_BMP 0 /* - * _MBCSHeader versions 4.1 + * _MBCSHeader versions 5.3 & 4.3 * (Note that the _MBCSHeader version is in addition to the converter formatVersion.) * + * This version is optional. Version 5 is used for incompatible data format changes. + * makeconv will continue to generate version 4 files if possible. + * + * Changes from version 4: + * + * The main difference is an additional _MBCSHeader field with + * - the length (number of uint32_t) of the _MBCSHeader + * - flags for further incompatible data format changes + * - flags for further, backward compatible data format changes + * + * The MBCS_OPT_FROM_U flag indicates that most of the fromUnicode data is omitted from + * the file and needs to be reconstituted at load time. + * This requires a utf8Friendly format with an additional mbcsIndex table for fast + * (and UTF-8-friendly) fromUnicode conversion for Unicode code points up to maxFastUChar. + * (For details about these structures see below, and see ucnvmbcs.h.) + * + * utf8Friendly also implies that the fromUnicode mappings are stored in ascending order + * of the Unicode code points. (This requires that the .ucm file has the |0 etc. + * precision markers for all mappings.) + * + * All fallbacks have been moved to the extension table, leaving only roundtrips in the + * omitted data that can be reconstituted from the toUnicode data. + * + * Of the stage 2 table, the part corresponding to maxFastUChar and below is omitted. + * With only roundtrip mappings in the base fromUnicode data, this part is fully + * redundant with the mbcsIndex and will be reconstituted from that (also using the + * stage 1 table which contains the information about how stage 2 was compacted). + * + * The rest of the stage 2 table, the part for code points above maxFastUChar, + * is stored in the file and will be appended to the reconstituted part. + * + * The entire fromUBytes array is omitted from the file and will be reconstitued. + * This is done by enumerating all toUnicode roundtrip mappings, performing + * each mapping (using the stage 1 and reconstituted stage 2 tables) and + * writing instead of reading the byte values. + * + * _MBCSHeader version 4.3 + * + * Change from version 4.2: + * - Optional utf8Friendly data structures, with 64-entry stage 3 block + * allocation for parts of the BMP, and an additional mbcsIndex in non-SBCS + * files which can be used instead of stages 1 & 2. + * Faster lookups for roundtrips from most commonly used characters, + * and lookups from UTF-8 byte sequences with a natural bit distribution. + * See ucnvmbcs.h for more details. + * + * Change from version 4.1: + * - Added an optional extension table structure at the end of the .cnv file. + * It is present if the upper bits of the header flags field contains a non-zero + * byte offset to it. + * Files that contain only a conversion table and no base table + * use the special outputType MBCS_OUTPUT_EXT_ONLY. + * These contain the base table name between the MBCS header and the extension + * data. + * * Change from version 4.0: * - Replace header.reserved with header.fromUBytesLength so that all * fields in the data have length. @@ -218,7 +278,7 @@ * One trail byte state that results in code points, and one that only * has "unassigned" and "illegal" terminal states. * - * Note: partly by accident, this data structure supports simple stateless + * Note: partly by accident, this data structure supports simple stateful * encodings without any additional logic. * Currently, only simple Shift-In/Shift-Out schemes are handled with * appropriate state tables (especially EBCDIC_STATEFUL!). @@ -255,6 +315,12 @@ * 0 unassigned * Bits 7..0 contain the codepage byte. A zero byte is always possible. * + * In version 4.3, the runtime code can build an sbcsIndex for a utf8Friendly + * file. For 2-byte UTF-8 byte sequences and some 3-byte sequences the lookup + * becomes a 2-stage (single-index) trie lookup with 6 bits for stage 3. + * ASCII code points can be looked up with a linear array access into stage 3. + * See maxFastUChar and other details in ucnvmbcs.h. + * * Multi-byte lookup: * * Stage 2 contains a 32-bit word for each 16-block in stage 3: @@ -270,11 +336,17 @@ * Leading zero bytes are ignored, and the number of bytes is counted. * A zero byte mapping result is possible as a roundtrip result. * For some output types, the actual result is processed from this; - * see _MBCSFromUnicodeWithOffsets(). + * see ucnv_MBCSFromUnicodeWithOffsets(). * * Note that stage 1 always contains 0x440=1088 entries (0x440==0x110000>>10), * or (version 3 and up) for BMP-only codepages, it contains 64 entries. * + * In version 4.3, a utf8Friendly file contains an mbcsIndex table. + * For 2-byte UTF-8 byte sequences and most 3-byte sequences the lookup + * becomes a 2-stage (single-index) trie lookup with 6 bits for stage 3. + * ASCII code points can be looked up with a linear array access into stage 3. + * See maxFastUChar, mbcsIndex and other details in ucnvmbcs.h. + * * In version 3, stage 2 blocks may overlap by multiples of the multiplier * for compaction. * In version 4, stage 2 blocks (and for single-byte codepages, stage 3 blocks) @@ -285,47 +357,8 @@ * adding new ones without crashing an unaware converter */ -/* prototypes --------------------------------------------------------------- */ - -static void -_MBCSSingleToUnicodeWithOffsets(UConverterToUnicodeArgs *pArgs, - UErrorCode *pErrorCode); - -static void -_MBCSSingleToBMPWithOffsets(UConverterToUnicodeArgs *pArgs, - UErrorCode *pErrorCode); - -static UChar32 -_MBCSGetNextUChar(UConverterToUnicodeArgs *pArgs, - UErrorCode *pErrorCode); - -static UChar32 -_MBCSSingleGetNextUChar(UConverterToUnicodeArgs *pArgs, - UErrorCode *pErrorCode); - -static void -_MBCSDoubleFromUnicodeWithOffsets(UConverterFromUnicodeArgs *pArgs, - UErrorCode *pErrorCode); - -static void -_MBCSSingleFromUnicodeWithOffsets(UConverterFromUnicodeArgs *pArgs, - UErrorCode *pErrorCode); - -static void -_MBCSSingleFromBMPWithOffsets(UConverterFromUnicodeArgs *pArgs, - UErrorCode *pErrorCode); - -static void -fromUCallback(UConverter *cnv, - const void *context, UConverterFromUnicodeArgs *pArgs, - UChar32 codePoint, - UConverterCallbackReason reason, UErrorCode *pErrorCode); - -static void -toUCallback(UConverter *cnv, - const void *context, UConverterToUnicodeArgs *pArgs, - const char *codeUnits, int32_t length, - UConverterCallbackReason reason, UErrorCode *pErrorCode); +static const UConverterImpl _SBCSUTF8Impl; +static const UConverterImpl _DBCSUTF8Impl; /* GB 18030 data ------------------------------------------------------------ */ @@ -365,100 +398,320 @@ gb18030Ranges[13][4]={ /* bit flag for UConverter.options indicating GB 18030 special handling */ #define _MBCS_OPTION_GB18030 0x8000 -/* Miscellaneous ------------------------------------------------------------ */ +/* bit flag for UConverter.options indicating KEIS,JEF,JIF special handling */ +#define _MBCS_OPTION_KEIS 0x01000 +#define _MBCS_OPTION_JEF 0x02000 +#define _MBCS_OPTION_JIPS 0x04000 -static uint32_t -_MBCSSizeofFromUBytes(UConverterMBCSTable *mbcsTable) { - const uint16_t *table; +#define KEIS_SO_CHAR_1 0x0A +#define KEIS_SO_CHAR_2 0x42 +#define KEIS_SI_CHAR_1 0x0A +#define KEIS_SI_CHAR_2 0x41 - uint32_t st3, maxStage3; - uint16_t st1, maxStage1, st2; +#define JEF_SO_CHAR 0x28 +#define JEF_SI_CHAR 0x29 - if(mbcsTable->fromUBytesLength>0) { - /* - * We _know_ the number of bytes in the fromUnicodeBytes array - * starting with header.version 4.1. - * Otherwise, below, we need to enumerate the fromUnicode - * trie and find the highest entry. - */ - return mbcsTable->fromUBytesLength; - } +#define JIPS_SO_CHAR_1 0x1A +#define JIPS_SO_CHAR_2 0x70 +#define JIPS_SI_CHAR_1 0x1A +#define JIPS_SI_CHAR_2 0x71 - /* Enumerate the from-Unicode trie table to find the highest stage 3 index. */ - table=mbcsTable->fromUnicodeTable; - maxStage3=0; - if(mbcsTable->unicodeMask&UCNV_HAS_SUPPLEMENTARY) { - maxStage1=0x440; - } else { - maxStage1=0x40; +enum SISO_Option { + SI, + SO +}; +typedef enum SISO_Option SISO_Option; + +static int32_t getSISOBytes(SISO_Option option, uint32_t cnvOption, uint8_t *value) { + int32_t SISOLength = 0; + + switch (option) { + case SI: + if ((cnvOption&_MBCS_OPTION_KEIS)!=0) { + value[0] = KEIS_SI_CHAR_1; + value[1] = KEIS_SI_CHAR_2; + SISOLength = 2; + } else if ((cnvOption&_MBCS_OPTION_JEF)!=0) { + value[0] = JEF_SI_CHAR; + SISOLength = 1; + } else if ((cnvOption&_MBCS_OPTION_JIPS)!=0) { + value[0] = JIPS_SI_CHAR_1; + value[1] = JIPS_SI_CHAR_2; + SISOLength = 2; + } else { + value[0] = UCNV_SI; + SISOLength = 1; + } + break; + case SO: + if ((cnvOption&_MBCS_OPTION_KEIS)!=0) { + value[0] = KEIS_SO_CHAR_1; + value[1] = KEIS_SO_CHAR_2; + SISOLength = 2; + } else if ((cnvOption&_MBCS_OPTION_JEF)!=0) { + value[0] = JEF_SO_CHAR; + SISOLength = 1; + } else if ((cnvOption&_MBCS_OPTION_JIPS)!=0) { + value[0] = JIPS_SO_CHAR_1; + value[1] = JIPS_SO_CHAR_2; + SISOLength = 2; + } else { + value[0] = UCNV_SO; + SISOLength = 1; + } + break; + default: + /* Should never happen. */ + break; } + return SISOLength; +} - if(mbcsTable->outputType==MBCS_OUTPUT_1) { - const uint16_t *stage2; +/* Miscellaneous ------------------------------------------------------------ */ - for(st1=0; st1maxStage1) { - stage2=table+st2; - for(st2=0; st2<64; ++st2) { - st3=stage2[st2]; - if(st3>maxStage3) { - maxStage3=st3; - } +/** + * Callback from ucnv_MBCSEnumToUnicode(), takes 32 mappings from + * consecutive sequences of bytes, starting from the one encoded in value, + * to Unicode code points. (Multiple mappings to reduce per-function call overhead.) + * Does not currently support m:n mappings or reverse fallbacks. + * This function will not be called for sequences of bytes with leading zeros. + * + * @param context an opaque pointer, as passed into ucnv_MBCSEnumToUnicode() + * @param value contains 1..4 bytes of the first byte sequence, right-aligned + * @param codePoints resulting Unicode code points, or negative if a byte sequence does + * not map to anything + * @return TRUE to continue enumeration, FALSE to stop + */ +typedef UBool U_CALLCONV +UConverterEnumToUCallback(const void *context, uint32_t value, UChar32 codePoints[32]); + +/* similar to ucnv_MBCSGetNextUChar() but recursive */ +static UBool +enumToU(UConverterMBCSTable *mbcsTable, int8_t stateProps[], + int32_t state, uint32_t offset, + uint32_t value, + UConverterEnumToUCallback *callback, const void *context, + UErrorCode *pErrorCode) { + UChar32 codePoints[32]; + const int32_t *row; + const uint16_t *unicodeCodeUnits; + UChar32 anyCodePoints; + int32_t b, limit; + + row=mbcsTable->stateTable[state]; + unicodeCodeUnits=mbcsTable->unicodeCodeUnits; + + value<<=8; + anyCodePoints=-1; /* becomes non-negative if there is a mapping */ + + b=(stateProps[state]&0x38)<<2; + if(b==0 && stateProps[state]>=0x40) { + /* skip byte sequences with leading zeros because they are not stored in the fromUnicode table */ + codePoints[0]=U_SENTINEL; + b=1; + } + limit=((stateProps[state]&7)+1)<<5; + while(b=0) { + /* recurse to a state with non-ignorable actions */ + if(!enumToU( + mbcsTable, stateProps, nextState, + offset+MBCS_ENTRY_TRANSITION_OFFSET(entry), + value|(uint32_t)b, + callback, context, + pErrorCode)) { + return FALSE; } } - } + codePoints[b&0x1f]=U_SENTINEL; + } else { + UChar32 c; + int32_t action; - /* - * add 16 to get the limit not start index of the last stage 3 block, - * times 2 for number of bytes - */ - return (maxStage3+16)*2; - } else { - const uint32_t *stage2; + /* + * An if-else-if chain provides more reliable performance for + * the most common cases compared to a switch. + */ + action=MBCS_ENTRY_FINAL_ACTION(entry); + if(action==MBCS_STATE_VALID_DIRECT_16) { + /* output BMP code point */ + c=(UChar)MBCS_ENTRY_FINAL_VALUE_16(entry); + } else if(action==MBCS_STATE_VALID_16) { + int32_t finalOffset=offset+MBCS_ENTRY_FINAL_VALUE_16(entry); + c=unicodeCodeUnits[finalOffset]; + if(c<0xfffe) { + /* output BMP code point */ + } else { + c=U_SENTINEL; + } + } else if(action==MBCS_STATE_VALID_16_PAIR) { + int32_t finalOffset=offset+MBCS_ENTRY_FINAL_VALUE_16(entry); + c=unicodeCodeUnits[finalOffset++]; + if(c<0xd800) { + /* output BMP code point below 0xd800 */ + } else if(c<=0xdbff) { + /* output roundtrip or fallback supplementary code point */ + c=((c&0x3ff)<<10)+unicodeCodeUnits[finalOffset]+(0x10000-0xdc00); + } else if(c==0xe000) { + /* output roundtrip BMP code point above 0xd800 or fallback BMP code point */ + c=unicodeCodeUnits[finalOffset]; + } else { + c=U_SENTINEL; + } + } else if(action==MBCS_STATE_VALID_DIRECT_20) { + /* output supplementary code point */ + c=(UChar32)(MBCS_ENTRY_FINAL_VALUE(entry)+0x10000); + } else { + c=U_SENTINEL; + } - for(st1=0; st1(maxStage1>>1)) { - stage2=(const uint32_t *)table+st2; - for(st2=0; st2<64; ++st2) { - st3=stage2[st2]&0xffff; - if(st3>maxStage3) { - maxStage3=st3; - } + codePoints[b&0x1f]=c; + anyCodePoints&=c; + } + if(((++b)&0x1f)==0) { + if(anyCodePoints>=0) { + if(!callback(context, value|(uint32_t)(b-0x20), codePoints)) { + return FALSE; } + anyCodePoints=-1; } } + } + return TRUE; +} - /* - * add 16 to get the limit not start index of the last stage 3 block, - * times 2..4 for number of bytes - */ - maxStage3=16*maxStage3+16; - switch(mbcsTable->outputType) { - case MBCS_OUTPUT_3: - case MBCS_OUTPUT_4_EUC: - maxStage3*=3; - break; - case MBCS_OUTPUT_4: - maxStage3*=4; +/* + * Only called if stateProps[state]==-1. + * A recursive call may do stateProps[state]|=0x40 if this state is the target of an + * MBCS_STATE_CHANGE_ONLY. + */ +static int8_t +getStateProp(const int32_t (*stateTable)[256], int8_t stateProps[], int state) { + const int32_t *row; + int32_t min, max, entry, nextState; + + row=stateTable[state]; + stateProps[state]=0; + + /* find first non-ignorable state */ + for(min=0;; ++min) { + entry=row[min]; + nextState=MBCS_ENTRY_STATE(entry); + if(stateProps[nextState]==-1) { + getStateProp(stateTable, stateProps, nextState); + } + if(MBCS_ENTRY_IS_TRANSITION(entry)) { + if(stateProps[nextState]>=0) { + break; + } + } else if(MBCS_ENTRY_FINAL_ACTION(entry)>5)<<3); + + /* find last non-ignorable state */ + for(max=0xff; min=0) { + break; + } + } else if(MBCS_ENTRY_FINAL_ACTION(entry)>5); + + /* recurse further and collect direct-state information */ + while(min<=max) { + entry=row[min]; + nextState=MBCS_ENTRY_STATE(entry); + if(stateProps[nextState]==-1) { + getStateProp(stateTable, stateProps, nextState); + } + if(MBCS_ENTRY_IS_FINAL(entry)) { + stateProps[nextState]|=0x40; + if(MBCS_ENTRY_FINAL_ACTION(entry)<=MBCS_STATE_FALLBACK_DIRECT_20) { + stateProps[state]|=0x40; + } + } + ++min; + } + return stateProps[state]; } +/* + * Internal function enumerating the toUnicode data of an MBCS converter. + * Currently only used for reconstituting data for a MBCS_OPT_NO_FROM_U + * table, but could also be used for a future ucnv_getUnicodeSet() option + * that includes reverse fallbacks (after updating this function's implementation). + * Currently only handles roundtrip mappings. + * Does not currently handle extensions. + */ static void -_MBCSGetUnicodeSet(const UConverter *cnv, - USet *set, - UConverterUnicodeSet which, - UErrorCode *pErrorCode) { - UConverterMBCSTable *mbcsTable; +ucnv_MBCSEnumToUnicode(UConverterMBCSTable *mbcsTable, + UConverterEnumToUCallback *callback, const void *context, + UErrorCode *pErrorCode) { + /* + * Properties for each state, to speed up the enumeration. + * Ignorable actions are unassigned/illegal/state-change-only: + * They do not lead to mappings. + * + * Bits 7..6: + * 1 direct/initial state (stateful converters have multiple) + * 0 non-initial state with transitions or with non-ignorable result actions + * -1 final state with only ignorable actions + * + * Bits 5..3: + * The lowest byte value with non-ignorable actions is + * value<<5 (rounded down). + * + * Bits 2..0: + * The highest byte value with non-ignorable actions is + * (value<<5)&0x1f (rounded up). + */ + int8_t stateProps[MBCS_MAX_STATE_COUNT]; + int32_t state; + + uprv_memset(stateProps, -1, sizeof(stateProps)); + + /* recurse from state 0 and set all stateProps */ + getStateProp(mbcsTable->stateTable, stateProps, 0); + + for(state=0; statecountStates; ++state) { + /*if(stateProps[state]==-1) { + printf("unused/unreachable %d\n", state); + }*/ + if(stateProps[state]>=0x40) { + /* start from each direct state */ + enumToU( + mbcsTable, stateProps, state, 0, 0, + callback, context, + pErrorCode); + } + } +} + +U_CFUNC void +ucnv_MBCSGetFilteredUnicodeSetForUnicode(const UConverterSharedData *sharedData, + const USetAdder *sa, + UConverterUnicodeSet which, + UConverterSetFilter filter, + UErrorCode *pErrorCode) { + const UConverterMBCSTable *mbcsTable; const uint16_t *table; uint32_t st3; @@ -466,14 +719,8 @@ _MBCSGetUnicodeSet(const UConverter *cnv, UChar32 c; - if(cnv->options&_MBCS_OPTION_GB18030) { - uset_addRange(set, 0, 0xd7ff); - uset_addRange(set, 0xe000, 0x10ffff); - return; - } - /* enumerate the from-Unicode trie table */ - mbcsTable=&cnv->sharedData->table->mbcs; + mbcsTable=&sharedData->mbcs; table=mbcsTable->fromUnicodeTable; if(mbcsTable->unicodeMask&UCNV_HAS_SUPPLEMENTARY) { maxStage1=0x440; @@ -485,9 +732,23 @@ _MBCSGetUnicodeSet(const UConverter *cnv, if(mbcsTable->outputType==MBCS_OUTPUT_1) { const uint16_t *stage2, *stage3, *results; + uint16_t minValue; results=(const uint16_t *)mbcsTable->fromUnicodeBytes; + /* + * Set a threshold variable for selecting which mappings to use. + * See ucnv_MBCSSingleFromBMPWithOffsets() and + * MBCS_SINGLE_RESULT_FROM_U() for details. + */ + if(which==UCNV_ROUNDTRIP_SET) { + /* use only roundtrips */ + minValue=0xf00; + } else /* UCNV_ROUNDTRIP_AND_FALLBACK_SET */ { + /* use all roundtrip and fallback results */ + minValue=0x800; + } + for(st1=0; st1maxStage1) { @@ -497,16 +758,9 @@ _MBCSGetUnicodeSet(const UConverter *cnv, /* read the stage 3 block */ stage3=results+st3; - /* - * Add code points for which the roundtrip flag is set. - * Once we get a set for fallback mappings, we have to use - * a threshold variable with a value of 0x800. - * See _MBCSSingleFromBMPWithOffsets() and - * MBCS_SINGLE_RESULT_FROM_U() for details. - */ do { - if(*stage3++>=0xf00) { - uset_add(set, c); + if(*stage3++>=minValue) { + sa->add(sa->set, c); } } while((++c&0xf)!=0); } else { @@ -519,6 +773,27 @@ _MBCSGetUnicodeSet(const UConverter *cnv, } } else { const uint32_t *stage2; + const uint8_t *stage3, *bytes; + uint32_t st3Multiplier; + uint32_t value; + UBool useFallback; + + bytes=mbcsTable->fromUnicodeBytes; + + useFallback=(UBool)(which==UCNV_ROUNDTRIP_AND_FALLBACK_SET); + + switch(mbcsTable->outputType) { + case MBCS_OUTPUT_3: + case MBCS_OUTPUT_4_EUC: + st3Multiplier=3; + break; + case MBCS_OUTPUT_4: + st3Multiplier=4; + break; + default: + st3Multiplier=2; + break; + } for(st1=0; st1>=16; /* - * Add code points for which the roundtrip flag is set. - * Once we get a set for fallback mappings, we have to check - * non-roundtrip stage 3 results for whether they are 0. - * See _MBCSFromUnicodeWithOffsets() for details. + * Add code points for which the roundtrip flag is set, + * or which map to non-zero bytes if we use fallbacks. + * See ucnv_MBCSFromUnicodeWithOffsets() for details. */ - do { - if(st3&1) { - uset_add(set, c); - } - st3>>=1; - } while((++c&0xf)!=0); + switch(filter) { + case UCNV_SET_FILTER_NONE: + do { + if(st3&1) { + sa->add(sa->set, c); + stage3+=st3Multiplier; + } else if(useFallback) { + uint8_t b=0; + switch(st3Multiplier) { + case 4: + b|=*stage3++; + case 3: + b|=*stage3++; + case 2: + b|=stage3[0]|stage3[1]; + stage3+=2; + default: + break; + } + if(b!=0) { + sa->add(sa->set, c); + } + } + st3>>=1; + } while((++c&0xf)!=0); + break; + case UCNV_SET_FILTER_DBCS_ONLY: + /* Ignore single-byte results (<0x100). */ + do { + if(((st3&1)!=0 || useFallback) && *((const uint16_t *)stage3)>=0x100) { + sa->add(sa->set, c); + } + st3>>=1; + stage3+=2; /* +=st3Multiplier */ + } while((++c&0xf)!=0); + break; + case UCNV_SET_FILTER_2022_CN: + /* Only add code points that map to CNS 11643 planes 1 & 2 for non-EXT ISO-2022-CN. */ + do { + if(((st3&1)!=0 || useFallback) && ((value=*stage3)==0x81 || value==0x82)) { + sa->add(sa->set, c); + } + st3>>=1; + stage3+=3; /* +=st3Multiplier */ + } while((++c&0xf)!=0); + break; + case UCNV_SET_FILTER_SJIS: + /* Only add code points that map to Shift-JIS codes corresponding to JIS X 0208. */ + do { + if(((st3&1)!=0 || useFallback) && (value=*((const uint16_t *)stage3))>=0x8140 && value<=0xeffc) { + sa->add(sa->set, c); + } + st3>>=1; + stage3+=2; /* +=st3Multiplier */ + } while((++c&0xf)!=0); + break; + case UCNV_SET_FILTER_GR94DBCS: + /* Only add code points that map to ISO 2022 GR 94 DBCS codes (each byte A1..FE). */ + do { + if( ((st3&1)!=0 || useFallback) && + (uint16_t)((value=*((const uint16_t *)stage3)) - 0xa1a1)<=(0xfefe - 0xa1a1) && + (uint8_t)(value-0xa1)<=(0xfe - 0xa1) + ) { + sa->add(sa->set, c); + } + st3>>=1; + stage3+=2; /* +=st3Multiplier */ + } while((++c&0xf)!=0); + break; + case UCNV_SET_FILTER_HZ: + /* Only add code points that are suitable for HZ DBCS (lead byte A1..FD). */ + do { + if( ((st3&1)!=0 || useFallback) && + (uint16_t)((value=*((const uint16_t *)stage3))-0xa1a1)<=(0xfdfe - 0xa1a1) && + (uint8_t)(value-0xa1)<=(0xfe - 0xa1) + ) { + sa->add(sa->set, c); + } + st3>>=1; + stage3+=2; /* +=st3Multiplier */ + } while((++c&0xf)!=0); + break; + default: + *pErrorCode=U_INTERNAL_PROGRAM_ERROR; + return; + } } else { c+=16; /* empty stage 3 block */ } @@ -550,59 +907,222 @@ _MBCSGetUnicodeSet(const UConverter *cnv, } } } + + ucnv_extGetUnicodeSet(sharedData, sa, which, filter, pErrorCode); } -/* EBCDIC swap LF<->NL ------------------------------------------------------ */ +U_CFUNC void +ucnv_MBCSGetUnicodeSetForUnicode(const UConverterSharedData *sharedData, + const USetAdder *sa, + UConverterUnicodeSet which, + UErrorCode *pErrorCode) { + ucnv_MBCSGetFilteredUnicodeSetForUnicode( + sharedData, sa, which, + sharedData->mbcs.outputType==MBCS_OUTPUT_DBCS_ONLY ? + UCNV_SET_FILTER_DBCS_ONLY : + UCNV_SET_FILTER_NONE, + pErrorCode); +} + +static void +ucnv_MBCSGetUnicodeSet(const UConverter *cnv, + const USetAdder *sa, + UConverterUnicodeSet which, + UErrorCode *pErrorCode) { + if(cnv->options&_MBCS_OPTION_GB18030) { + sa->addRange(sa->set, 0, 0xd7ff); + sa->addRange(sa->set, 0xe000, 0x10ffff); + } else { + ucnv_MBCSGetUnicodeSetForUnicode(cnv->sharedData, sa, which, pErrorCode); + } +} + +/* conversion extensions for input not in the main table -------------------- */ /* - * This code modifies a standard EBCDIC<->Unicode mapping table for - * OS/390 (z/OS) Unix System Services (Open Edition). - * The difference is in the mapping of Line Feed and New Line control codes: - * Standard EBCDIC maps - * - * \x25 |0 - * \x15 |0 + * Hardcoded extension handling for GB 18030. + * Definition of LINEAR macros and gb18030Ranges see near the beginning of the file. * - * but OS/390 USS EBCDIC swaps the control codes for LF and NL, - * mapping - * - * \x15 |0 - * \x25 |0 + * In the future, conversion extensions may handle m:n mappings and delta tables, + * see http://source.icu-project.org/repos/icu/icuhtml/trunk/design/conversion/conversion_extensions.html * - * This code modifies a loaded standard EBCDIC<->Unicode mapping table - * by copying it into allocated memory and swapping the LF and NL values. - * It allows to support the same EBCDIC charset in both versions without - * duplicating the entire installed table. + * If an input character cannot be mapped, then these functions set an error + * code. The framework will then call the callback function. */ -/* standard EBCDIC codes */ -#define EBCDIC_LF 0x25 -#define EBCDIC_NL 0x15 +/* + * @return if(U_FAILURE) return the code point for cnv->fromUChar32 + * else return 0 after output has been written to the target + */ +static UChar32 +_extFromU(UConverter *cnv, const UConverterSharedData *sharedData, + UChar32 cp, + const UChar **source, const UChar *sourceLimit, + uint8_t **target, const uint8_t *targetLimit, + int32_t **offsets, int32_t sourceIndex, + UBool flush, + UErrorCode *pErrorCode) { + const int32_t *cx; + + cnv->useSubChar1=FALSE; + + if( (cx=sharedData->mbcs.extIndexes)!=NULL && + ucnv_extInitialMatchFromU( + cnv, cx, + cp, source, sourceLimit, + (char **)target, (char *)targetLimit, + offsets, sourceIndex, + flush, + pErrorCode) + ) { + return 0; /* an extension mapping handled the input */ + } -/* standard EBCDIC codes with roundtrip flag as stored in Unicode-to-single-byte tables */ -#define EBCDIC_RT_LF 0xf25 -#define EBCDIC_RT_NL 0xf15 + /* GB 18030 */ + if((cnv->options&_MBCS_OPTION_GB18030)!=0) { + const uint32_t *range; + int32_t i; -/* Unicode code points */ -#define U_LF 0x0a -#define U_NL 0x85 + range=gb18030Ranges[0]; + for(i=0; itoUBytes[0..length[ + * @return if(U_FAILURE) return the length (toULength, byteIndex) for the input + * else return 0 after output has been written to the target + */ +static int8_t +_extToU(UConverter *cnv, const UConverterSharedData *sharedData, + int8_t length, + const uint8_t **source, const uint8_t *sourceLimit, + UChar **target, const UChar *targetLimit, + int32_t **offsets, int32_t sourceIndex, + UBool flush, + UErrorCode *pErrorCode) { + const int32_t *cx; + + if( (cx=sharedData->mbcs.extIndexes)!=NULL && + ucnv_extInitialMatchToU( + cnv, cx, + length, (const char **)source, (const char *)sourceLimit, + target, targetLimit, + offsets, sourceIndex, + flush, + pErrorCode) + ) { + return 0; /* an extension mapping handled the input */ + } + + /* GB 18030 */ + if(length==4 && (cnv->options&_MBCS_OPTION_GB18030)!=0) { + const uint32_t *range; + uint32_t linear; + int32_t i; + + linear=LINEAR_18030(cnv->toUBytes[0], cnv->toUBytes[1], cnv->toUBytes[2], cnv->toUBytes[3]); + range=gb18030Ranges[0]; + for(i=0; iNL ------------------------------------------------------ */ + +/* + * This code modifies a standard EBCDIC<->Unicode mapping table for + * OS/390 (z/OS) Unix System Services (Open Edition). + * The difference is in the mapping of Line Feed and New Line control codes: + * Standard EBCDIC maps + * + * \x25 |0 + * \x15 |0 + * + * but OS/390 USS EBCDIC swaps the control codes for LF and NL, + * mapping + * + * \x15 |0 + * \x25 |0 + * + * This code modifies a loaded standard EBCDIC<->Unicode mapping table + * by copying it into allocated memory and swapping the LF and NL values. + * It allows to support the same EBCDIC charset in both versions without + * duplicating the entire installed table. + */ + +/* standard EBCDIC codes */ +#define EBCDIC_LF 0x25 +#define EBCDIC_NL 0x15 + +/* standard EBCDIC codes with roundtrip flag as stored in Unicode-to-single-byte tables */ +#define EBCDIC_RT_LF 0xf25 +#define EBCDIC_RT_NL 0xf15 + +/* Unicode code points */ +#define U_LF 0x0a +#define U_NL 0x85 + +static UBool +_EBCDICSwapLFNL(UConverterSharedData *sharedData, UErrorCode *pErrorCode) { + UConverterMBCSTable *mbcsTable; + + const uint16_t *table, *results; + const uint8_t *bytes; + + int32_t (*newStateTable)[256]; + uint16_t *newResults; + uint8_t *p; + char *name; + + uint32_t stage2Entry; uint32_t size, sizeofFromUBytes; - mbcsTable=&sharedData->table->mbcs; + mbcsTable=&sharedData->mbcs; table=mbcsTable->fromUnicodeTable; bytes=mbcsTable->fromUnicodeBytes; @@ -647,6 +1167,29 @@ _EBCDICSwapLFNL(UConverterSharedData *sharedData, UErrorCode *pErrorCode) { } } + if(mbcsTable->fromUBytesLength>0) { + /* + * We _know_ the number of bytes in the fromUnicodeBytes array + * starting with header.version 4.1. + */ + sizeofFromUBytes=mbcsTable->fromUBytesLength; + } else { + /* + * Otherwise: + * There used to be code to enumerate the fromUnicode + * trie and find the highest entry, but it was removed in ICU 3.2 + * because it was not tested and caused a low code coverage number. + * See Jitterbug 3674. + * This affects only some .cnv file formats with a header.version + * below 4.1, and only when swaplfnl is requested. + * + * ucnvmbcs.c revision 1.99 is the last one with the + * ucnv_MBCSSizeofFromUBytes() function. + */ + *pErrorCode=U_INVALID_FORMAT_ERROR; + return FALSE; + } + /* * The table has an appropriate format. * Allocate and build @@ -654,7 +1197,6 @@ _EBCDICSwapLFNL(UConverterSharedData *sharedData, UErrorCode *pErrorCode) { * - a modified from-Unicode output array * - a converter name string with the swap option appended */ - sizeofFromUBytes=_MBCSSizeofFromUBytes(mbcsTable); size= mbcsTable->countStates*1024+ sizeofFromUBytes+ @@ -711,471 +1253,634 @@ _EBCDICSwapLFNL(UConverterSharedData *sharedData, UErrorCode *pErrorCode) { return TRUE; } -/* MBCS setup functions ----------------------------------------------------- */ - -static void -_MBCSLoad(UConverterSharedData *sharedData, - const uint8_t *raw, - UErrorCode *pErrorCode) { - UDataInfo info; - UConverterMBCSTable *mbcsTable=&sharedData->table->mbcs; - _MBCSHeader *header=(_MBCSHeader *)raw; - - if(header->version[0]!=4) { - *pErrorCode=U_INVALID_TABLE_FORMAT; - return; - } +/* reconstitute omitted fromUnicode data ------------------------------------ */ - mbcsTable->countStates=(uint8_t)header->countStates; - mbcsTable->countToUFallbacks=header->countToUFallbacks; - mbcsTable->stateTable=(const int32_t (*)[256])(raw+sizeof(_MBCSHeader)); - mbcsTable->toUFallbacks=(const _MBCSToUFallback *)(mbcsTable->stateTable+header->countStates); - mbcsTable->unicodeCodeUnits=(const uint16_t *)(raw+header->offsetToUCodeUnits); +/* for details, compare with genmbcs.c MBCSAddFromUnicode() and transformEUC() */ +static UBool U_CALLCONV +writeStage3Roundtrip(const void *context, uint32_t value, UChar32 codePoints[32]) { + UConverterMBCSTable *mbcsTable=(UConverterMBCSTable *)context; + const uint16_t *table; + uint32_t *stage2; + uint8_t *bytes, *p; + UChar32 c; + int32_t i, st3; - mbcsTable->fromUnicodeTable=(const uint16_t *)(raw+header->offsetFromUTable); - mbcsTable->fromUnicodeBytes=(const uint8_t *)(raw+header->offsetFromUBytes); - mbcsTable->fromUBytesLength=header->fromUBytesLength; - mbcsTable->outputType=(uint8_t)header->flags; + table=mbcsTable->fromUnicodeTable; + bytes=(uint8_t *)mbcsTable->fromUnicodeBytes; - /* make sure that the output type is known */ + /* for EUC outputTypes, modify the value like genmbcs.c's transformEUC() */ switch(mbcsTable->outputType) { - case MBCS_OUTPUT_1: - case MBCS_OUTPUT_2: - case MBCS_OUTPUT_3: - case MBCS_OUTPUT_4: case MBCS_OUTPUT_3_EUC: + if(value<=0xffff) { + /* short sequences are stored directly */ + /* code set 0 or 1 */ + } else if(value<=0x8effff) { + /* code set 2 */ + value&=0x7fff; + } else /* first byte is 0x8f */ { + /* code set 3 */ + value&=0xff7f; + } + break; case MBCS_OUTPUT_4_EUC: - case MBCS_OUTPUT_2_SISO: - /* OK */ + if(value<=0xffffff) { + /* short sequences are stored directly */ + /* code set 0 or 1 */ + } else if(value<=0x8effffff) { + /* code set 2 */ + value&=0x7fffff; + } else /* first byte is 0x8f */ { + /* code set 3 */ + value&=0xff7fff; + } break; default: - *pErrorCode=U_INVALID_TABLE_FORMAT; - return; + break; } - /* - * converter versions 6.1 and up contain a unicodeMask that is - * used here to select the most efficient function implementations - */ - info.size=sizeof(UDataInfo); - udata_getInfo((UDataMemory *)sharedData->dataMemory, &info); - if(info.formatVersion[0]>6 || (info.formatVersion[0]==6 && info.formatVersion[1]>=1)) { - /* mask off possible future extensions to be safe */ - mbcsTable->unicodeMask=(uint8_t)(sharedData->staticData->unicodeMask&3); - } else { - /* for older versions, assume worst case: contains anything possible (prevent over-optimizations) */ - mbcsTable->unicodeMask=UCNV_HAS_SUPPLEMENTARY|UCNV_HAS_SURROGATES; - } -} + for(i=0; i<=0x1f; ++value, ++i) { + c=codePoints[i]; + if(c<0) { + continue; + } -static void -_MBCSUnload(UConverterSharedData *sharedData) { - UConverterMBCSTable *mbcsTable=&sharedData->table->mbcs; + /* locate the stage 2 & 3 data */ + stage2=((uint32_t *)table)+table[c>>10]+((c>>4)&0x3f); + p=bytes; + st3=(int32_t)(uint16_t)*stage2*16+(c&0xf); - if(mbcsTable->swapLFNLStateTable!=NULL) { - uprv_free(mbcsTable->swapLFNLStateTable); - } -} + /* write the codepage bytes into stage 3 */ + switch(mbcsTable->outputType) { + case MBCS_OUTPUT_3: + case MBCS_OUTPUT_4_EUC: + p+=st3*3; + p[0]=(uint8_t)(value>>16); + p[1]=(uint8_t)(value>>8); + p[2]=(uint8_t)value; + break; + case MBCS_OUTPUT_4: + ((uint32_t *)p)[st3]=value; + break; + default: + /* 2 bytes per character */ + ((uint16_t *)p)[st3]=(uint16_t)value; + break; + } -static void -_MBCSReset(UConverter *cnv, UConverterResetChoice choice) { - if(choice<=UCNV_RESET_TO_UNICODE) { - /* toUnicode */ - cnv->toUnicodeStatus=0; /* offset */ - cnv->mode=0; /* state */ - cnv->toULength=0; /* byteIndex */ - } - if(choice!=UCNV_RESET_TO_UNICODE) { - /* fromUnicode */ - cnv->fromUSurrogateLead=0; - cnv->fromUnicodeStatus=1; /* prevLength */ + /* set the roundtrip flag */ + *stage2|=(1UL<<(16+(c&0xf))); } -} + return TRUE; + } static void -_MBCSOpen(UConverter *cnv, - const char *name, - const char *locale, - uint32_t options, - UErrorCode *pErrorCode) { - if((options&UCNV_OPTION_SWAP_LFNL)!=0) { - /* do this because double-checked locking is broken */ - UBool isCached; - - umtx_lock(NULL); - isCached=cnv->sharedData->table->mbcs.swapLFNLStateTable!=NULL; - umtx_unlock(NULL); - - if(!isCached) { - if(!_EBCDICSwapLFNL(cnv->sharedData, pErrorCode)) { - /* the option does not apply, remove it */ - cnv->options&=~UCNV_OPTION_SWAP_LFNL; +reconstituteData(UConverterMBCSTable *mbcsTable, + uint32_t stage1Length, uint32_t stage2Length, + uint32_t fullStage2Length, /* lengths are numbers of units, not bytes */ + UErrorCode *pErrorCode) { + uint16_t *stage1; + uint32_t *stage2; + uint8_t *bytes; + uint32_t dataLength=stage1Length*2+fullStage2Length*4+mbcsTable->fromUBytesLength; + mbcsTable->reconstitutedData=(uint8_t *)uprv_malloc(dataLength); + if(mbcsTable->reconstitutedData==NULL) { + *pErrorCode=U_MEMORY_ALLOCATION_ERROR; + return; + } + uprv_memset(mbcsTable->reconstitutedData, 0, dataLength); + + /* copy existing data and reroute the pointers */ + stage1=(uint16_t *)mbcsTable->reconstitutedData; + uprv_memcpy(stage1, mbcsTable->fromUnicodeTable, stage1Length*2); + + stage2=(uint32_t *)(stage1+stage1Length); + uprv_memcpy(stage2+(fullStage2Length-stage2Length), + mbcsTable->fromUnicodeTable+stage1Length, + stage2Length*4); + + mbcsTable->fromUnicodeTable=stage1; + mbcsTable->fromUnicodeBytes=bytes=(uint8_t *)(stage2+fullStage2Length); + + /* indexes into stage 2 count from the bottom of the fromUnicodeTable */ + stage2=(uint32_t *)stage1; + + /* reconstitute the initial part of stage 2 from the mbcsIndex */ + { + int32_t stageUTF8Length=((int32_t)mbcsTable->maxFastUChar+1)>>6; + int32_t stageUTF8Index=0; + int32_t st1, st2, st3, i; + + for(st1=0; stageUTF8IndexmbcsIndex[stageUTF8Index++]; + if(st3!=0) { + /* an stage 2 entry's index is per stage 3 16-block, not per stage 3 entry */ + st3>>=4; + /* + * 4 stage 2 entries point to 4 consecutive stage 3 16-blocks which are + * allocated together as a single 64-block for access from the mbcsIndex + */ + stage2[st2++]=st3++; + stage2[st2++]=st3++; + stage2[st2++]=st3++; + stage2[st2++]=st3; + } else { + /* no stage 3 block, skip */ + st2+=4; + } + } + } else { + /* no stage 2 block, skip */ + stageUTF8Index+=16; } } } + /* reconstitute fromUnicodeBytes with roundtrips from toUnicode data */ + ucnv_MBCSEnumToUnicode(mbcsTable, writeStage3Roundtrip, mbcsTable, pErrorCode); +} - if(uprv_strstr(name, "18030")!=NULL) { - if(uprv_strstr(name, "gb18030")!=NULL || uprv_strstr(name, "GB18030")!=NULL) { - /* set a flag for GB 18030 mode, which changes the callback behavior */ - cnv->options|=_MBCS_OPTION_GB18030; - } +/* MBCS setup functions ----------------------------------------------------- */ + +static void +ucnv_MBCSLoad(UConverterSharedData *sharedData, + UConverterLoadArgs *pArgs, + const uint8_t *raw, + UErrorCode *pErrorCode) { + UDataInfo info; + UConverterMBCSTable *mbcsTable=&sharedData->mbcs; + _MBCSHeader *header=(_MBCSHeader *)raw; + uint32_t offset; + uint32_t headerLength; + UBool noFromU=FALSE; + + if(header->version[0]==4) { + headerLength=MBCS_HEADER_V4_LENGTH; + } else if(header->version[0]==5 && header->version[1]>=3 && + (header->options&MBCS_OPT_UNKNOWN_INCOMPATIBLE_MASK)==0) { + headerLength=header->options&MBCS_OPT_LENGTH_MASK; + noFromU=(UBool)((header->options&MBCS_OPT_NO_FROM_U)!=0); + } else { + *pErrorCode=U_INVALID_TABLE_FORMAT; + return; } - _MBCSReset(cnv, UCNV_RESET_BOTH); -} + mbcsTable->outputType=(uint8_t)header->flags; + if(noFromU && mbcsTable->outputType==MBCS_OUTPUT_1) { + *pErrorCode=U_INVALID_TABLE_FORMAT; + return; + } -static const char * -_MBCSGetName(const UConverter *cnv) { - if((cnv->options&UCNV_OPTION_SWAP_LFNL)!=0 && cnv->sharedData->table->mbcs.swapLFNLName!=NULL) { - return cnv->sharedData->table->mbcs.swapLFNLName; - } else { - return cnv->sharedData->staticData->name; + /* extension data, header version 4.2 and higher */ + offset=header->flags>>8; + if(offset!=0) { + mbcsTable->extIndexes=(const int32_t *)(raw+offset); } -} -/* MBCS-to-Unicode conversion functions ------------------------------------- */ + if(mbcsTable->outputType==MBCS_OUTPUT_EXT_ONLY) { + UConverterLoadArgs args={ 0 }; + UConverterSharedData *baseSharedData; + const int32_t *extIndexes; + const char *baseName; -static UChar32 -_MBCSGetFallback(UConverterMBCSTable *mbcsTable, uint32_t offset) { - const _MBCSToUFallback *toUFallbacks; - uint32_t i, start, limit; + /* extension-only file, load the base table and set values appropriately */ + if((extIndexes=mbcsTable->extIndexes)==NULL) { + /* extension-only file without extension */ + *pErrorCode=U_INVALID_TABLE_FORMAT; + return; + } - limit=mbcsTable->countToUFallbacks; - if(limit>0) { - /* do a binary search for the fallback mapping */ - toUFallbacks=mbcsTable->toUFallbacks; - start=0; - while(startnestedLoads!=1) { + /* an extension table must not be loaded as a base table */ + *pErrorCode=U_INVALID_TABLE_FILE; + return; } - /* did we really find it? */ - if(offset==toUFallbacks[start].offset) { - return toUFallbacks[start].codePoint; + /* load the base table */ + baseName=(const char *)header+headerLength*4; + if(0==uprv_strcmp(baseName, sharedData->staticData->name)) { + /* forbid loading this same extension-only file */ + *pErrorCode=U_INVALID_TABLE_FORMAT; + return; } - } - return 0xfffe; -} + /* TODO parse package name out of the prefix of the base name in the extension .cnv file? */ + args.size=sizeof(UConverterLoadArgs); + args.nestedLoads=2; + args.onlyTestIsLoadable=pArgs->onlyTestIsLoadable; + args.reserved=pArgs->reserved; + args.options=pArgs->options; + args.pkg=pArgs->pkg; + args.name=baseName; + baseSharedData=ucnv_load(&args, pErrorCode); + if(U_FAILURE(*pErrorCode)) { + return; + } + if( baseSharedData->staticData->conversionType!=UCNV_MBCS || + baseSharedData->mbcs.baseSharedData!=NULL + ) { + ucnv_unload(baseSharedData); + *pErrorCode=U_INVALID_TABLE_FORMAT; + return; + } + if(pArgs->onlyTestIsLoadable) { + /* + * Exit as soon as we know that we can load the converter + * and the format is valid and supported. + * The worst that can happen in the following code is a memory + * allocation error. + */ + ucnv_unload(baseSharedData); + return; + } -U_CFUNC void -_MBCSToUnicodeWithOffsets(UConverterToUnicodeArgs *pArgs, - UErrorCode *pErrorCode) { - UConverter *cnv; - const uint8_t *source, *sourceLimit; - UChar *target; - const UChar *targetLimit; - int32_t *offsets; + /* copy the base table data */ + uprv_memcpy(mbcsTable, &baseSharedData->mbcs, sizeof(UConverterMBCSTable)); - const int32_t (*stateTable)[256]; - const uint16_t *unicodeCodeUnits; + /* overwrite values with relevant ones for the extension converter */ + mbcsTable->baseSharedData=baseSharedData; + mbcsTable->extIndexes=extIndexes; - uint32_t offset; - uint8_t state; - int8_t byteIndex; - uint8_t *bytes; + /* + * It would be possible to share the swapLFNL data with a base converter, + * but the generated name would have to be different, and the memory + * would have to be free'd only once. + * It is easier to just create the data for the extension converter + * separately when it is requested. + */ + mbcsTable->swapLFNLStateTable=NULL; + mbcsTable->swapLFNLFromUnicodeBytes=NULL; + mbcsTable->swapLFNLName=NULL; - int32_t sourceIndex, nextSourceIndex; + /* + * The reconstitutedData must be deleted only when the base converter + * is unloaded. + */ + mbcsTable->reconstitutedData=NULL; - int32_t entry; - UChar c; - uint8_t action; - UConverterCallbackReason reason; + /* + * Set a special, runtime-only outputType if the extension converter + * is a DBCS version of a base converter that also maps single bytes. + */ + if( sharedData->staticData->conversionType==UCNV_DBCS || + (sharedData->staticData->conversionType==UCNV_MBCS && + sharedData->staticData->minBytesPerChar>=2) + ) { + if(baseSharedData->mbcs.outputType==MBCS_OUTPUT_2_SISO) { + /* the base converter is SI/SO-stateful */ + int32_t entry; + + /* get the dbcs state from the state table entry for SO=0x0e */ + entry=mbcsTable->stateTable[0][0xe]; + if( MBCS_ENTRY_IS_FINAL(entry) && + MBCS_ENTRY_FINAL_ACTION(entry)==MBCS_STATE_CHANGE_ONLY && + MBCS_ENTRY_FINAL_STATE(entry)!=0 + ) { + mbcsTable->dbcsOnlyState=(uint8_t)MBCS_ENTRY_FINAL_STATE(entry); + + mbcsTable->outputType=MBCS_OUTPUT_DBCS_ONLY; + } + } else if( + baseSharedData->staticData->conversionType==UCNV_MBCS && + baseSharedData->staticData->minBytesPerChar==1 && + baseSharedData->staticData->maxBytesPerChar==2 && + mbcsTable->countStates<=127 + ) { + /* non-stateful base converter, need to modify the state table */ + int32_t (*newStateTable)[256]; + int32_t *state; + int32_t i, count; + + /* allocate a new state table and copy the base state table contents */ + count=mbcsTable->countStates; + newStateTable=(int32_t (*)[256])uprv_malloc((count+1)*1024); + if(newStateTable==NULL) { + ucnv_unload(baseSharedData); + *pErrorCode=U_MEMORY_ALLOCATION_ERROR; + return; + } - /* use optimized function if possible */ - cnv=pArgs->converter; - if(cnv->sharedData->table->mbcs.countStates==1) { - if(!(cnv->sharedData->table->mbcs.unicodeMask&UCNV_HAS_SUPPLEMENTARY)) { - _MBCSSingleToBMPWithOffsets(pArgs, pErrorCode); - } else { - _MBCSSingleToUnicodeWithOffsets(pArgs, pErrorCode); - } - return; - } + uprv_memcpy(newStateTable, mbcsTable->stateTable, count*1024); - /* set up the local pointers */ - source=(const uint8_t *)pArgs->source; - sourceLimit=(const uint8_t *)pArgs->sourceLimit; - target=pArgs->target; - targetLimit=pArgs->targetLimit; - offsets=pArgs->offsets; + /* change all final single-byte entries to go to a new all-illegal state */ + state=newStateTable[0]; + for(i=0; i<256; ++i) { + if(MBCS_ENTRY_IS_FINAL(state[i])) { + state[i]=MBCS_ENTRY_TRANSITION(count, 0); + } + } - if((cnv->options&UCNV_OPTION_SWAP_LFNL)!=0) { - stateTable=(const int32_t (*)[256])cnv->sharedData->table->mbcs.swapLFNLStateTable; - } else { - stateTable=cnv->sharedData->table->mbcs.stateTable; - } - unicodeCodeUnits=cnv->sharedData->table->mbcs.unicodeCodeUnits; + /* build the new all-illegal state */ + state=newStateTable[count]; + for(i=0; i<256; ++i) { + state[i]=MBCS_ENTRY_FINAL(0, MBCS_STATE_ILLEGAL, 0); + } + mbcsTable->stateTable=(const int32_t (*)[256])newStateTable; + mbcsTable->countStates=(uint8_t)(count+1); + mbcsTable->stateTableOwned=TRUE; - /* get the converter state from UConverter */ - offset=cnv->toUnicodeStatus; - state=(uint8_t)(cnv->mode); - byteIndex=cnv->toULength; - bytes=cnv->toUBytes; + mbcsTable->outputType=MBCS_OUTPUT_DBCS_ONLY; + } + } - /* sourceIndex=-1 if the current character began in the previous buffer */ - sourceIndex=byteIndex==0 ? 0 : -1; - nextSourceIndex=0; + /* + * unlike below for files with base tables, do not get the unicodeMask + * from the sharedData; instead, use the base table's unicodeMask, + * which we copied in the memcpy above; + * this is necessary because the static data unicodeMask, especially + * the UCNV_HAS_SUPPLEMENTARY flag, is part of the base table data + */ + } else { + /* conversion file with a base table; an additional extension table is optional */ + /* make sure that the output type is known */ + switch(mbcsTable->outputType) { + case MBCS_OUTPUT_1: + case MBCS_OUTPUT_2: + case MBCS_OUTPUT_3: + case MBCS_OUTPUT_4: + case MBCS_OUTPUT_3_EUC: + case MBCS_OUTPUT_4_EUC: + case MBCS_OUTPUT_2_SISO: + /* OK */ + break; + default: + *pErrorCode=U_INVALID_TABLE_FORMAT; + return; + } + if(pArgs->onlyTestIsLoadable) { + /* + * Exit as soon as we know that we can load the converter + * and the format is valid and supported. + * The worst that can happen in the following code is a memory + * allocation error. + */ + return; + } + + mbcsTable->countStates=(uint8_t)header->countStates; + mbcsTable->countToUFallbacks=header->countToUFallbacks; + mbcsTable->stateTable=(const int32_t (*)[256])(raw+headerLength*4); + mbcsTable->toUFallbacks=(const _MBCSToUFallback *)(mbcsTable->stateTable+header->countStates); + mbcsTable->unicodeCodeUnits=(const uint16_t *)(raw+header->offsetToUCodeUnits); + + mbcsTable->fromUnicodeTable=(const uint16_t *)(raw+header->offsetFromUTable); + mbcsTable->fromUnicodeBytes=(const uint8_t *)(raw+header->offsetFromUBytes); + mbcsTable->fromUBytesLength=header->fromUBytesLength; - /* conversion loop */ - while(sourcedataMemory, &info); + if(info.formatVersion[0]>6 || (info.formatVersion[0]==6 && info.formatVersion[1]>=1)) { + /* mask off possible future extensions to be safe */ + mbcsTable->unicodeMask=(uint8_t)(sharedData->staticData->unicodeMask&3); + } else { + /* for older versions, assume worst case: contains anything possible (prevent over-optimizations) */ + mbcsTable->unicodeMask=UCNV_HAS_SUPPLEMENTARY|UCNV_HAS_SURROGATES; + } + /* + * _MBCSHeader.version 4.3 adds utf8Friendly data structures. + * Check for the header version, SBCS vs. MBCS, and for whether the + * data structures are optimized for code points as high as what the + * runtime code is designed for. + * The implementation does not handle mapping tables with entries for + * unpaired surrogates. + */ + if( header->version[1]>=3 && + (mbcsTable->unicodeMask&UCNV_HAS_SURROGATES)==0 && + (mbcsTable->countStates==1 ? + (header->version[2]>=(SBCS_FAST_MAX>>8)) : + (header->version[2]>=(MBCS_FAST_MAX>>8)) + ) + ) { + mbcsTable->utf8Friendly=TRUE; + + if(mbcsTable->countStates==1) { /* - * An if-else-if chain provides more reliable performance for - * the most common cases compared to a switch. + * SBCS: Stage 3 is allocated in 64-entry blocks for U+0000..SBCS_FAST_MAX or higher. + * Build a table with indexes to each block, to be used instead of + * the regular stage 1/2 table. */ - action=(uint8_t)(MBCS_ENTRY_FINAL_ACTION(entry)); - if(action==MBCS_STATE_VALID_16) { - offset+=MBCS_ENTRY_FINAL_VALUE_16(entry); - c=unicodeCodeUnits[offset]; - if(c<0xfffe) { - /* output BMP code point */ - *target++=c; - if(offsets!=NULL) { - *offsets++=sourceIndex; - } - } else if(c==0xfffe) { - if(UCNV_TO_U_USE_FALLBACK(cnv) && (entry=(int32_t)_MBCSGetFallback(&cnv->sharedData->table->mbcs, offset))!=0xfffe) { - /* output fallback BMP code point */ - *target++=(UChar)entry; - if(offsets!=NULL) { - *offsets++=sourceIndex; - } - } else { - /* callback(unassigned) */ - goto unassigned; - } - } else { - /* callback(illegal) */ - goto illegal; - } - } else if(action==MBCS_STATE_VALID_DIRECT_16) { - /* output BMP code point */ - *target++=(UChar)MBCS_ENTRY_FINAL_VALUE_16(entry); - if(offsets!=NULL) { - *offsets++=sourceIndex; - } - } else if(action==MBCS_STATE_VALID_16_PAIR) { - offset+=MBCS_ENTRY_FINAL_VALUE_16(entry); - c=unicodeCodeUnits[offset++]; - if(c<0xd800) { - /* output BMP code point below 0xd800 */ - *target++=c; - if(offsets!=NULL) { - *offsets++=sourceIndex; - } - } else if(UCNV_TO_U_USE_FALLBACK(cnv) ? c<=0xdfff : c<=0xdbff) { - /* output roundtrip or fallback surrogate pair */ - *target++=(UChar)(c&0xdbff); - if(offsets!=NULL) { - *offsets++=sourceIndex; - } - if(targetUCharErrorBuffer[0]=unicodeCodeUnits[offset]; - cnv->UCharErrorBufferLength=1; - *pErrorCode=U_BUFFER_OVERFLOW_ERROR; + int32_t i; + for(i=0; i<(SBCS_FAST_LIMIT>>6); ++i) { + mbcsTable->sbcsIndex[i]=mbcsTable->fromUnicodeTable[mbcsTable->fromUnicodeTable[i>>4]+((i<<2)&0x3c)]; + } + /* set SBCS_FAST_MAX to reflect the reach of sbcsIndex[] even if header->version[2]>(SBCS_FAST_MAX>>8) */ + mbcsTable->maxFastUChar=SBCS_FAST_MAX; + } else { + /* + * MBCS: Stage 3 is allocated in 64-entry blocks for U+0000..MBCS_FAST_MAX or higher. + * The .cnv file is prebuilt with an additional stage table with indexes + * to each block. + */ + mbcsTable->mbcsIndex=(const uint16_t *) + (mbcsTable->fromUnicodeBytes+ + (noFromU ? 0 : mbcsTable->fromUBytesLength)); + mbcsTable->maxFastUChar=(((UChar)header->version[2])<<8)|0xff; + } + } - offset=0; - byteIndex=0; - break; - } - } else if(UCNV_TO_U_USE_FALLBACK(cnv) ? (c&0xfffe)==0xe000 : c==0xe000) { - /* output roundtrip BMP code point above 0xd800 or fallback BMP code point */ - *target++=unicodeCodeUnits[offset]; - if(offsets!=NULL) { - *offsets++=sourceIndex; - } - } else if(c==0xffff) { - /* callback(illegal) */ - goto illegal; - } else { - /* callback(unassigned) */ - goto unassigned; - } - } else if(action==MBCS_STATE_VALID_DIRECT_20) { -valid20: - entry=MBCS_ENTRY_FINAL_VALUE(entry); - /* output surrogate pair */ - *target++=(UChar)(0xd800|(UChar)(entry>>10)); - if(offsets!=NULL) { - *offsets++=sourceIndex; - } - c=(UChar)(0xdc00|(UChar)(entry&0x3ff)); - if(targetUCharErrorBuffer[0]=c; - cnv->UCharErrorBufferLength=1; - *pErrorCode=U_BUFFER_OVERFLOW_ERROR; + /* calculate a bit set of 4 ASCII characters per bit that round-trip to ASCII bytes */ + { + uint32_t asciiRoundtrips=0xffffffff; + int32_t i; - offset=0; - byteIndex=0; - break; - } - } else if(action==MBCS_STATE_CHANGE_ONLY) { - /* - * This serves as a state change without any output. - * It is useful for reading simple stateful encodings, - * for example using just Shift-In/Shift-Out codes. - * The 21 unused bits may later be used for more sophisticated - * state transitions. - */ - } else if(action==MBCS_STATE_FALLBACK_DIRECT_16) { - if(!UCNV_TO_U_USE_FALLBACK(cnv)) { - /* callback(unassigned) */ - goto unassigned; - } - /* output BMP code point */ - *target++=(UChar)MBCS_ENTRY_FINAL_VALUE_16(entry); - if(offsets!=NULL) { - *offsets++=sourceIndex; - } - } else if(action==MBCS_STATE_FALLBACK_DIRECT_20) { - if(!UCNV_TO_U_USE_FALLBACK(cnv)) { - /* callback(unassigned) */ - goto unassigned; - } - goto valid20; - } else if(action==MBCS_STATE_UNASSIGNED) { - /* callback(unassigned) */ - goto unassigned; - } else if(action==MBCS_STATE_ILLEGAL) { - /* callback(illegal) */ - goto illegal; - } else { - /* reserved, must never occur */ + for(i=0; i<0x80; ++i) { + if(mbcsTable->stateTable[0][i]!=MBCS_ENTRY_FINAL(0, MBCS_STATE_VALID_DIRECT_16, i)) { + asciiRoundtrips&=~((uint32_t)1<<(i>>2)); } + } + mbcsTable->asciiRoundtrips=asciiRoundtrips; + } - /* normal end of action codes: prepare for a new character */ - offset=0; - byteIndex=0; - sourceIndex=nextSourceIndex; - continue; + if(noFromU) { + uint32_t stage1Length= + mbcsTable->unicodeMask&UCNV_HAS_SUPPLEMENTARY ? + 0x440 : 0x40; + uint32_t stage2Length= + (header->offsetFromUBytes-header->offsetFromUTable)/4- + stage1Length/2; + reconstituteData(mbcsTable, stage1Length, stage2Length, header->fullStage2Length, pErrorCode); + } + } -illegal: - reason=UCNV_ILLEGAL; - *pErrorCode=U_ILLEGAL_CHAR_FOUND; - goto callback; -unassigned: - reason=UCNV_UNASSIGNED; - *pErrorCode=U_INVALID_CHAR_FOUND; -callback: - /* call the callback function with all the preparations and post-processing */ - /* update the arguments structure */ - pArgs->source=(const char *)source; - pArgs->target=target; - pArgs->offsets=offsets; - - /* set the converter state in UConverter to deal with the next character */ - cnv->toUnicodeStatus=0; - cnv->mode=state; - cnv->toULength=0; + /* Set the impl pointer here so that it is set for both extension-only and base tables. */ + if(mbcsTable->utf8Friendly) { + if(mbcsTable->countStates==1) { + sharedData->impl=&_SBCSUTF8Impl; + } else { + if(mbcsTable->outputType==MBCS_OUTPUT_2) { + sharedData->impl=&_DBCSUTF8Impl; + } + } + } + + if(mbcsTable->outputType==MBCS_OUTPUT_DBCS_ONLY || mbcsTable->outputType==MBCS_OUTPUT_2_SISO) { + /* + * MBCS_OUTPUT_DBCS_ONLY: No SBCS mappings, therefore ASCII does not roundtrip. + * MBCS_OUTPUT_2_SISO: Bypass the ASCII fastpath to handle prevLength correctly. + */ + mbcsTable->asciiRoundtrips=0; + } +} + +static void +ucnv_MBCSUnload(UConverterSharedData *sharedData) { + UConverterMBCSTable *mbcsTable=&sharedData->mbcs; - /* call the callback function */ - toUCallback(cnv, cnv->toUContext, pArgs, (const char *)bytes, byteIndex, reason, pErrorCode); + if(mbcsTable->swapLFNLStateTable!=NULL) { + uprv_free(mbcsTable->swapLFNLStateTable); + } + if(mbcsTable->stateTableOwned) { + uprv_free((void *)mbcsTable->stateTable); + } + if(mbcsTable->baseSharedData!=NULL) { + ucnv_unload(mbcsTable->baseSharedData); + } + if(mbcsTable->reconstitutedData!=NULL) { + uprv_free(mbcsTable->reconstitutedData); + } +} - /* get the converter state from UConverter */ - offset=cnv->toUnicodeStatus; - state=(uint8_t)cnv->mode; - byteIndex=cnv->toULength; +static void +ucnv_MBCSOpen(UConverter *cnv, + UConverterLoadArgs *pArgs, + UErrorCode *pErrorCode) { + UConverterMBCSTable *mbcsTable; + const int32_t *extIndexes; + uint8_t outputType; + int8_t maxBytesPerUChar; - /* update target and deal with offsets if necessary */ - offsets=ucnv_updateCallbackOffsets(offsets, pArgs->target-target, sourceIndex); - target=pArgs->target; + if(pArgs->onlyTestIsLoadable) { + return; + } - /* update the source pointer and index */ - sourceIndex=nextSourceIndex+((const uint8_t *)pArgs->source-source); - source=(const uint8_t *)pArgs->source; + mbcsTable=&cnv->sharedData->mbcs; + outputType=mbcsTable->outputType; - /* - * If the callback overflowed the target, then we need to - * stop here with an overflow indication. - */ - if(*pErrorCode==U_BUFFER_OVERFLOW_ERROR) { - break; - } else if(U_FAILURE(*pErrorCode)) { - /* break on error */ - offset=0; - state=0; - byteIndex=0; - break; - } else if(cnv->UCharErrorBufferLength>0) { - /* target is full */ - *pErrorCode=U_BUFFER_OVERFLOW_ERROR; - break; + if(outputType==MBCS_OUTPUT_DBCS_ONLY) { + /* the swaplfnl option does not apply, remove it */ + cnv->options=pArgs->options&=~UCNV_OPTION_SWAP_LFNL; + } + + if((pArgs->options&UCNV_OPTION_SWAP_LFNL)!=0) { + /* do this because double-checked locking is broken */ + UBool isCached; + + umtx_lock(NULL); + isCached=mbcsTable->swapLFNLStateTable!=NULL; + umtx_unlock(NULL); + + if(!isCached) { + if(!_EBCDICSwapLFNL(cnv->sharedData, pErrorCode)) { + if(U_FAILURE(*pErrorCode)) { + return; /* something went wrong */ } - /* - * We do not need to repeat the statements from the normal - * end of the action codes because we already updated all the - * necessary variables. - */ + /* the option does not apply, remove it */ + cnv->options=pArgs->options&=~UCNV_OPTION_SWAP_LFNL; } - } else { - /* target is full */ - *pErrorCode=U_BUFFER_OVERFLOW_ERROR; - break; } } - if(pArgs->flush && source>=sourceLimit) { - /* reset the state for the next conversion */ - if(byteIndex>0 && U_SUCCESS(*pErrorCode)) { - /* a character byte sequence remains incomplete */ - *pErrorCode=U_TRUNCATED_CHAR_FOUND; + if(uprv_strstr(pArgs->name, "18030")!=NULL) { + if(uprv_strstr(pArgs->name, "gb18030")!=NULL || uprv_strstr(pArgs->name, "GB18030")!=NULL) { + /* set a flag for GB 18030 mode, which changes the callback behavior */ + cnv->options|=_MBCS_OPTION_GB18030; + } + } else if((uprv_strstr(pArgs->name, "KEIS")!=NULL) || (uprv_strstr(pArgs->name, "keis")!=NULL)) { + /* set a flag for KEIS converter, which changes the SI/SO character sequence */ + cnv->options|=_MBCS_OPTION_KEIS; + } else if((uprv_strstr(pArgs->name, "JEF")!=NULL) || (uprv_strstr(pArgs->name, "jef")!=NULL)) { + /* set a flag for JEF converter, which changes the SI/SO character sequence */ + cnv->options|=_MBCS_OPTION_JEF; + } else if((uprv_strstr(pArgs->name, "JIPS")!=NULL) || (uprv_strstr(pArgs->name, "jips")!=NULL)) { + /* set a flag for JIPS converter, which changes the SI/SO character sequence */ + cnv->options|=_MBCS_OPTION_JIPS; + } + + /* fix maxBytesPerUChar depending on outputType and options etc. */ + if(outputType==MBCS_OUTPUT_2_SISO) { + cnv->maxBytesPerUChar=3; /* SO+DBCS */ + } + + extIndexes=mbcsTable->extIndexes; + if(extIndexes!=NULL) { + maxBytesPerUChar=(int8_t)UCNV_GET_MAX_BYTES_PER_UCHAR(extIndexes); + if(outputType==MBCS_OUTPUT_2_SISO) { + ++maxBytesPerUChar; /* SO + multiple DBCS */ + } + + if(maxBytesPerUChar>cnv->maxBytesPerUChar) { + cnv->maxBytesPerUChar=maxBytesPerUChar; } - cnv->toUnicodeStatus=0; - cnv->mode=0; - cnv->toULength=0; + } + +#if 0 + /* + * documentation of UConverter fields used for status + * all of these fields are (re)set to 0 by ucnv_bld.c and ucnv_reset() + */ + + /* toUnicode */ + cnv->toUnicodeStatus=0; /* offset */ + cnv->mode=0; /* state */ + cnv->toULength=0; /* byteIndex */ + + /* fromUnicode */ + cnv->fromUChar32=0; + cnv->fromUnicodeStatus=1; /* prevLength */ +#endif +} + +static const char * +ucnv_MBCSGetName(const UConverter *cnv) { + if((cnv->options&UCNV_OPTION_SWAP_LFNL)!=0 && cnv->sharedData->mbcs.swapLFNLName!=NULL) { + return cnv->sharedData->mbcs.swapLFNLName; } else { - /* set the converter state back into UConverter */ - cnv->toUnicodeStatus=offset; - cnv->mode=state; - cnv->toULength=byteIndex; + return cnv->sharedData->staticData->name; } +} - /* write back the updated pointers */ - pArgs->source=(const char *)source; - pArgs->target=target; - pArgs->offsets=offsets; +/* MBCS-to-Unicode conversion functions ------------------------------------- */ + +static UChar32 +ucnv_MBCSGetFallback(UConverterMBCSTable *mbcsTable, uint32_t offset) { + const _MBCSToUFallback *toUFallbacks; + uint32_t i, start, limit; + + limit=mbcsTable->countToUFallbacks; + if(limit>0) { + /* do a binary search for the fallback mapping */ + toUFallbacks=mbcsTable->toUFallbacks; + start=0; + while(startconverter; @@ -1201,14 +1905,13 @@ _MBCSSingleToUnicodeWithOffsets(UConverterToUnicodeArgs *pArgs, offsets=pArgs->offsets; if((cnv->options&UCNV_OPTION_SWAP_LFNL)!=0) { - stateTable=(const int32_t (*)[256])cnv->sharedData->table->mbcs.swapLFNLStateTable; + stateTable=(const int32_t (*)[256])cnv->sharedData->mbcs.swapLFNLStateTable; } else { - stateTable=cnv->sharedData->table->mbcs.stateTable; + stateTable=cnv->sharedData->mbcs.stateTable; } /* sourceIndex=-1 if the current character began in the previous buffer */ sourceIndex=0; - nextSourceIndex=0; /* conversion loop */ while(source=targetLimit) { + /* target is full */ + *pErrorCode=U_BUFFER_OVERFLOW_ERROR; + break; + } - /* test the most common case first */ - if(MBCS_ENTRY_FINAL_IS_VALID_DIRECT_16(entry)) { - /* output BMP code point */ - *target++=(UChar)MBCS_ENTRY_FINAL_VALUE_16(entry); - if(offsets!=NULL) { - *offsets++=sourceIndex; - } + entry=stateTable[0][*source++]; + /* MBCS_ENTRY_IS_FINAL(entry) */ - /* normal end of action codes: prepare for a new character */ - sourceIndex=nextSourceIndex; - continue; + /* test the most common case first */ + if(MBCS_ENTRY_FINAL_IS_VALID_DIRECT_16(entry)) { + /* output BMP code point */ + *target++=(UChar)MBCS_ENTRY_FINAL_VALUE_16(entry); + if(offsets!=NULL) { + *offsets++=sourceIndex; } - /* - * An if-else-if chain provides more reliable performance for - * the most common cases compared to a switch. - */ - action=(uint8_t)(MBCS_ENTRY_FINAL_ACTION(entry)); - if(action==MBCS_STATE_VALID_DIRECT_20) { -valid20: - entry=MBCS_ENTRY_FINAL_VALUE(entry); - /* output surrogate pair */ - *target++=(UChar)(0xd800|(UChar)(entry>>10)); - if(offsets!=NULL) { - *offsets++=sourceIndex; - } - c=(UChar)(0xdc00|(UChar)(entry&0x3ff)); - if(targetUCharErrorBuffer[0]=c; - cnv->UCharErrorBufferLength=1; - *pErrorCode=U_BUFFER_OVERFLOW_ERROR; - break; - } - } else if(action==MBCS_STATE_FALLBACK_DIRECT_16) { - if(!UCNV_TO_U_USE_FALLBACK(cnv)) { - /* callback(unassigned) */ - goto unassigned; + /* normal end of action codes: prepare for a new character */ + ++sourceIndex; + continue; + } + + /* + * An if-else-if chain provides more reliable performance for + * the most common cases compared to a switch. + */ + action=(uint8_t)(MBCS_ENTRY_FINAL_ACTION(entry)); + if(action==MBCS_STATE_VALID_DIRECT_20 || + (action==MBCS_STATE_FALLBACK_DIRECT_20 && UCNV_TO_U_USE_FALLBACK(cnv)) + ) { + entry=MBCS_ENTRY_FINAL_VALUE(entry); + /* output surrogate pair */ + *target++=(UChar)(0xd800|(UChar)(entry>>10)); + if(offsets!=NULL) { + *offsets++=sourceIndex; + } + c=(UChar)(0xdc00|(UChar)(entry&0x3ff)); + if(targetUCharErrorBuffer[0]=c; + cnv->UCharErrorBufferLength=1; + *pErrorCode=U_BUFFER_OVERFLOW_ERROR; + break; + } + + ++sourceIndex; + continue; + } else if(action==MBCS_STATE_FALLBACK_DIRECT_16) { + if(UCNV_TO_U_USE_FALLBACK(cnv)) { /* output BMP code point */ *target++=(UChar)MBCS_ENTRY_FINAL_VALUE_16(entry); if(offsets!=NULL) { *offsets++=sourceIndex; } - } else if(action==MBCS_STATE_FALLBACK_DIRECT_20) { - if(!UCNV_TO_U_USE_FALLBACK(cnv)) { - /* callback(unassigned) */ - goto unassigned; - } - goto valid20; - } else if(action==MBCS_STATE_UNASSIGNED) { - /* callback(unassigned) */ - goto unassigned; - } else if(action==MBCS_STATE_ILLEGAL) { - /* callback(illegal) */ - reason=UCNV_ILLEGAL; - *pErrorCode=U_ILLEGAL_CHAR_FOUND; - goto callback; - } else { - /* reserved, must never occur */ - } - /* normal end of action codes: prepare for a new character */ - sourceIndex=nextSourceIndex; + ++sourceIndex; + continue; + } + } else if(action==MBCS_STATE_UNASSIGNED) { + /* just fall through */ + } else if(action==MBCS_STATE_ILLEGAL) { + /* callback(illegal) */ + *pErrorCode=U_ILLEGAL_CHAR_FOUND; + } else { + /* reserved, must never occur */ + ++sourceIndex; continue; + } -unassigned: - reason=UCNV_UNASSIGNED; - *pErrorCode=U_INVALID_CHAR_FOUND; -callback: - /* call the callback function with all the preparations and post-processing */ - /* update the arguments structure */ + if(U_FAILURE(*pErrorCode)) { + /* callback(illegal) */ + break; + } else /* unassigned sequences indicated with byteIndex>0 */ { + /* try an extension mapping */ pArgs->source=(const char *)source; - pArgs->target=target; - pArgs->offsets=offsets; - - /* call the callback function */ - toUCallback(cnv, cnv->toUContext, pArgs, (const char *)(source-1), 1, reason, pErrorCode); - - /* update target and deal with offsets if necessary */ - offsets=ucnv_updateCallbackOffsets(offsets, pArgs->target-target, sourceIndex); - target=pArgs->target; - - /* update the source pointer and index */ - sourceIndex=nextSourceIndex+((const uint8_t *)pArgs->source-source); - source=(const uint8_t *)pArgs->source; - - /* - * If the callback overflowed the target, then we need to - * stop here with an overflow indication. - */ - if(*pErrorCode==U_BUFFER_OVERFLOW_ERROR) { - break; - } else if(U_FAILURE(*pErrorCode)) { - /* break on error */ - break; - } else if(cnv->UCharErrorBufferLength>0) { - /* target is full */ - *pErrorCode=U_BUFFER_OVERFLOW_ERROR; + cnv->toUBytes[0]=*(source-1); + cnv->toULength=_extToU(cnv, cnv->sharedData, + 1, &source, sourceLimit, + &target, targetLimit, + &offsets, sourceIndex, + pArgs->flush, + pErrorCode); + sourceIndex+=1+(int32_t)(source-(const uint8_t *)pArgs->source); + + if(U_FAILURE(*pErrorCode)) { + /* not mappable or buffer overflow */ break; } - - /* - * We do not need to repeat the statements from the normal - * end of the action codes because we already updated all the - * necessary variables. - */ - } else { - /* target is full */ - *pErrorCode=U_BUFFER_OVERFLOW_ERROR; - break; } } @@ -1351,13 +2026,13 @@ callback: } /* - * This version of _MBCSSingleToUnicodeWithOffsets() is optimized for single-byte, single-state codepages + * This version of ucnv_MBCSSingleToUnicodeWithOffsets() is optimized for single-byte, single-state codepages * that only map to and from the BMP. * In addition to single-byte optimizations, the offset calculations * become much easier. */ static void -_MBCSSingleToBMPWithOffsets(UConverterToUnicodeArgs *pArgs, +ucnv_MBCSSingleToBMPWithOffsets(UConverterToUnicodeArgs *pArgs, UErrorCode *pErrorCode) { UConverter *cnv; const uint8_t *source, *sourceLimit, *lastSource; @@ -1371,20 +2046,19 @@ _MBCSSingleToBMPWithOffsets(UConverterToUnicodeArgs *pArgs, int32_t entry; uint8_t action; - UConverterCallbackReason reason; /* set up the local pointers */ cnv=pArgs->converter; source=(const uint8_t *)pArgs->source; sourceLimit=(const uint8_t *)pArgs->sourceLimit; target=pArgs->target; - targetCapacity=pArgs->targetLimit-pArgs->target; + targetCapacity=(int32_t)(pArgs->targetLimit-pArgs->target); offsets=pArgs->offsets; if((cnv->options&UCNV_OPTION_SWAP_LFNL)!=0) { - stateTable=(const int32_t (*)[256])cnv->sharedData->table->mbcs.swapLFNLStateTable; + stateTable=(const int32_t (*)[256])cnv->sharedData->mbcs.swapLFNLStateTable; } else { - stateTable=cnv->sharedData->table->mbcs.stateTable; + stateTable=cnv->sharedData->mbcs.stateTable; } /* sourceIndex=-1 if the current character began in the previous buffer */ @@ -1395,7 +2069,7 @@ _MBCSSingleToBMPWithOffsets(UConverterToUnicodeArgs *pArgs, * since the conversion here is 1:1 UChar:uint8_t, we need only one counter * for the minimum of the sourceLength and targetCapacity */ - length=sourceLimit-source; + length=(int32_t)(sourceLimit-source); if(length0) { + while(targetCapacity > 0 && source < sourceLimit) { entry=stateTable[0][*source++]; /* MBCS_ENTRY_IS_FINAL(entry) */ @@ -1497,30 +2171,23 @@ unrolled: */ action=(uint8_t)(MBCS_ENTRY_FINAL_ACTION(entry)); if(action==MBCS_STATE_FALLBACK_DIRECT_16) { - if(!UCNV_TO_U_USE_FALLBACK(cnv)) { - /* callback(unassigned) */ - reason=UCNV_UNASSIGNED; - *pErrorCode=U_INVALID_CHAR_FOUND; + if(UCNV_TO_U_USE_FALLBACK(cnv)) { + /* output BMP code point */ + *target++=(UChar)MBCS_ENTRY_FINAL_VALUE_16(entry); + --targetCapacity; + continue; } - /* output BMP code point */ - *target++=(UChar)MBCS_ENTRY_FINAL_VALUE_16(entry); - --targetCapacity; - continue; } else if(action==MBCS_STATE_UNASSIGNED) { - /* callback(unassigned) */ - reason=UCNV_UNASSIGNED; - *pErrorCode=U_INVALID_CHAR_FOUND; + /* just fall through */ } else if(action==MBCS_STATE_ILLEGAL) { /* callback(illegal) */ - reason=UCNV_ILLEGAL; *pErrorCode=U_ILLEGAL_CHAR_FOUND; } else { /* reserved, must never occur */ continue; } - /* call the callback function with all the preparations and post-processing */ - /* set offsets since the start or the last callback */ + /* set offsets since the start or the last extension */ if(offsets!=NULL) { int32_t count=(int32_t)(source-lastSource); @@ -1531,40 +2198,32 @@ unrolled: /* offset and sourceIndex are now set for the current character */ } - /* update the arguments structure */ - pArgs->source=(const char *)source; - pArgs->target=target; - pArgs->offsets=offsets; - - /* call the callback function */ - toUCallback(cnv, cnv->toUContext, pArgs, (const char *)(source-1), 1, reason, pErrorCode); - - /* update target and deal with offsets if necessary */ - offsets=ucnv_updateCallbackOffsets(offsets, pArgs->target-target, sourceIndex); - target=pArgs->target; - - /* update the source pointer and index */ - sourceIndex+=1+((const uint8_t *)pArgs->source-source); - source=lastSource=(const uint8_t *)pArgs->source; - targetCapacity=pArgs->targetLimit-target; - length=sourceLimit-source; - if(lengthUCharErrorBufferLength>0) { - /* target is full */ - *pErrorCode=U_BUFFER_OVERFLOW_ERROR; + if(U_FAILURE(*pErrorCode)) { + /* callback(illegal) */ break; + } else /* unassigned sequences indicated with byteIndex>0 */ { + /* try an extension mapping */ + lastSource=source; + cnv->toUBytes[0]=*(source-1); + cnv->toULength=_extToU(cnv, cnv->sharedData, + 1, &source, sourceLimit, + &target, pArgs->targetLimit, + &offsets, sourceIndex, + pArgs->flush, + pErrorCode); + sourceIndex+=1+(int32_t)(source-lastSource); + + if(U_FAILURE(*pErrorCode)) { + /* not mappable or buffer overflow */ + break; + } + + /* recalculate the targetCapacity after an extension mapping */ + targetCapacity=(int32_t)(pArgs->targetLimit-target); + length=(int32_t)(sourceLimit-source); + if(lengthoffsets=offsets; } -static UChar32 -_MBCSGetNextUChar(UConverterToUnicodeArgs *pArgs, - UErrorCode *pErrorCode) { - UChar buffer[UTF_MAX_CHAR_LENGTH]; +static UBool +hasValidTrailBytes(const int32_t (*stateTable)[256], uint8_t state) { + const int32_t *row=stateTable[state]; + int32_t b, entry; + /* First test for final entries in this state for some commonly valid byte values. */ + entry=row[0xa1]; + if( !MBCS_ENTRY_IS_TRANSITION(entry) && + MBCS_ENTRY_FINAL_ACTION(entry)!=MBCS_STATE_ILLEGAL + ) { + return TRUE; + } + entry=row[0x41]; + if( !MBCS_ENTRY_IS_TRANSITION(entry) && + MBCS_ENTRY_FINAL_ACTION(entry)!=MBCS_STATE_ILLEGAL + ) { + return TRUE; + } + /* Then test for final entries in this state. */ + for(b=0; b<=0xff; ++b) { + entry=row[b]; + if( !MBCS_ENTRY_IS_TRANSITION(entry) && + MBCS_ENTRY_FINAL_ACTION(entry)!=MBCS_STATE_ILLEGAL + ) { + return TRUE; + } + } + /* Then recurse for transition entries. */ + for(b=0; b<=0xff; ++b) { + entry=row[b]; + if( MBCS_ENTRY_IS_TRANSITION(entry) && + hasValidTrailBytes(stateTable, (uint8_t)MBCS_ENTRY_TRANSITION_STATE(entry)) + ) { + return TRUE; + } + } + return FALSE; +} + +/* + * Is byte b a single/lead byte in this state? + * Recurse for transition states, because here we don't want to say that + * b is a lead byte if all byte sequences that start with b are illegal. + */ +static UBool +isSingleOrLead(const int32_t (*stateTable)[256], uint8_t state, UBool isDBCSOnly, uint8_t b) { + const int32_t *row=stateTable[state]; + int32_t entry=row[b]; + if(MBCS_ENTRY_IS_TRANSITION(entry)) { /* lead byte */ + return hasValidTrailBytes(stateTable, (uint8_t)MBCS_ENTRY_TRANSITION_STATE(entry)); + } else { + uint8_t action=(uint8_t)(MBCS_ENTRY_FINAL_ACTION(entry)); + if(action==MBCS_STATE_CHANGE_ONLY && isDBCSOnly) { + return FALSE; /* SI/SO are illegal for DBCS-only conversion */ + } else { + return action!=MBCS_STATE_ILLEGAL; + } + } +} +U_CFUNC void +ucnv_MBCSToUnicodeWithOffsets(UConverterToUnicodeArgs *pArgs, + UErrorCode *pErrorCode) { UConverter *cnv; const uint8_t *source, *sourceLimit; + UChar *target; + const UChar *targetLimit; + int32_t *offsets; const int32_t (*stateTable)[256]; const uint16_t *unicodeCodeUnits; @@ -1609,245 +2328,424 @@ _MBCSGetNextUChar(UConverterToUnicodeArgs *pArgs, int8_t byteIndex; uint8_t *bytes; + int32_t sourceIndex, nextSourceIndex; + int32_t entry; - UChar32 c; + UChar c; uint8_t action; - UConverterCallbackReason reason; /* use optimized function if possible */ cnv=pArgs->converter; - if(cnv->sharedData->table->mbcs.unicodeMask&UCNV_HAS_SURROGATES) { + + if(cnv->preToULength>0) { /* - * Calling the inefficient, generic getNextUChar() lets us deal correctly - * with the rare case of a codepage that maps single surrogates - * without adding the complexity to this already complicated function here. + * pass sourceIndex=-1 because we continue from an earlier buffer + * in the future, this may change with continuous offsets */ - return ucnv_getNextUCharFromToUImpl(pArgs, _MBCSToUnicodeWithOffsets, TRUE, pErrorCode); - } else if(cnv->sharedData->table->mbcs.countStates==1) { - return _MBCSSingleGetNextUChar(pArgs, pErrorCode); + ucnv_extContinueMatchToU(cnv, pArgs, -1, pErrorCode); + + if(U_FAILURE(*pErrorCode) || cnv->preToULength<0) { + return; + } + } + + if(cnv->sharedData->mbcs.countStates==1) { + if(!(cnv->sharedData->mbcs.unicodeMask&UCNV_HAS_SUPPLEMENTARY)) { + ucnv_MBCSSingleToBMPWithOffsets(pArgs, pErrorCode); + } else { + ucnv_MBCSSingleToUnicodeWithOffsets(pArgs, pErrorCode); + } + return; } /* set up the local pointers */ source=(const uint8_t *)pArgs->source; sourceLimit=(const uint8_t *)pArgs->sourceLimit; + target=pArgs->target; + targetLimit=pArgs->targetLimit; + offsets=pArgs->offsets; if((cnv->options&UCNV_OPTION_SWAP_LFNL)!=0) { - stateTable=(const int32_t (*)[256])cnv->sharedData->table->mbcs.swapLFNLStateTable; + stateTable=(const int32_t (*)[256])cnv->sharedData->mbcs.swapLFNLStateTable; } else { - stateTable=cnv->sharedData->table->mbcs.stateTable; + stateTable=cnv->sharedData->mbcs.stateTable; } - unicodeCodeUnits=cnv->sharedData->table->mbcs.unicodeCodeUnits; + unicodeCodeUnits=cnv->sharedData->mbcs.unicodeCodeUnits; /* get the converter state from UConverter */ offset=cnv->toUnicodeStatus; - state=(uint8_t)(cnv->mode); byteIndex=cnv->toULength; bytes=cnv->toUBytes; + /* + * if we are in the SBCS state for a DBCS-only converter, + * then load the DBCS state from the MBCS data + * (dbcsOnlyState==0 if it is not a DBCS-only converter) + */ + if((state=(uint8_t)(cnv->mode))==0) { + state=cnv->sharedData->mbcs.dbcsOnlyState; + } + + /* sourceIndex=-1 if the current character began in the previous buffer */ + sourceIndex=byteIndex==0 ? 0 : -1; + nextSourceIndex=0; + /* conversion loop */ while(source=targetLimit) { + /* target is full */ + *pErrorCode=U_BUFFER_OVERFLOW_ERROR; + break; + } + + if(byteIndex==0) { + /* optimized loop for 1/2-byte input and BMP output */ + if(offsets==NULL) { + do { + entry=stateTable[state][*source]; + if(MBCS_ENTRY_IS_TRANSITION(entry)) { + state=(uint8_t)MBCS_ENTRY_TRANSITION_STATE(entry); + offset=MBCS_ENTRY_TRANSITION_OFFSET(entry); + + ++source; + if( source=sourceLimit) { + break; + } + if(target>=targetLimit) { + /* target is full */ + *pErrorCode=U_BUFFER_OVERFLOW_ERROR; + break; + } + + ++nextSourceIndex; + bytes[byteIndex++]=*source++; + } else /* byteIndex>0 */ { + ++nextSourceIndex; + entry=stateTable[state][bytes[byteIndex++]=*source++]; + } + if(MBCS_ENTRY_IS_TRANSITION(entry)) { state=(uint8_t)MBCS_ENTRY_TRANSITION_STATE(entry); offset+=MBCS_ENTRY_TRANSITION_OFFSET(entry); - } else { - /* set the next state early so that we can reuse the entry variable */ - state=(uint8_t)MBCS_ENTRY_FINAL_STATE(entry); /* typically 0 */ + continue; + } - /* - * An if-else-if chain provides more reliable performance for - * the most common cases compared to a switch. - */ - action=(uint8_t)(MBCS_ENTRY_FINAL_ACTION(entry)); - if(action==MBCS_STATE_VALID_16) { - offset+=MBCS_ENTRY_FINAL_VALUE_16(entry); - c=unicodeCodeUnits[offset]; - if(c<0xfffe) { - /* output BMP code point */ - goto finish; - } else if(c==0xfffe) { - if(UCNV_TO_U_USE_FALLBACK(cnv) && (c=_MBCSGetFallback(&cnv->sharedData->table->mbcs, offset))!=0xfffe) { - goto finish; + /* save the previous state for proper extension mapping with SI/SO-stateful converters */ + cnv->mode=state; + + /* set the next state early so that we can reuse the entry variable */ + state=(uint8_t)MBCS_ENTRY_FINAL_STATE(entry); /* typically 0 */ + + /* + * An if-else-if chain provides more reliable performance for + * the most common cases compared to a switch. + */ + action=(uint8_t)(MBCS_ENTRY_FINAL_ACTION(entry)); + if(action==MBCS_STATE_VALID_16) { + offset+=MBCS_ENTRY_FINAL_VALUE_16(entry); + c=unicodeCodeUnits[offset]; + if(c<0xfffe) { + /* output BMP code point */ + *target++=c; + if(offsets!=NULL) { + *offsets++=sourceIndex; + } + byteIndex=0; + } else if(c==0xfffe) { + if(UCNV_TO_U_USE_FALLBACK(cnv) && (entry=(int32_t)ucnv_MBCSGetFallback(&cnv->sharedData->mbcs, offset))!=0xfffe) { + /* output fallback BMP code point */ + *target++=(UChar)entry; + if(offsets!=NULL) { + *offsets++=sourceIndex; } - /* callback(unassigned) */ - goto unassigned; - } else { - /* callback(illegal) */ - goto illegal; + byteIndex=0; } - } else if(action==MBCS_STATE_VALID_DIRECT_16) { - /* output BMP code point */ - c=(UChar)MBCS_ENTRY_FINAL_VALUE_16(entry); - goto finish; - } else if(action==MBCS_STATE_VALID_16_PAIR) { - offset+=MBCS_ENTRY_FINAL_VALUE_16(entry); - c=unicodeCodeUnits[offset++]; - if(c<0xd800) { - /* output BMP code point below 0xd800 */ - goto finish; - } else if(UCNV_TO_U_USE_FALLBACK(cnv) ? c<=0xdfff : c<=0xdbff) { - /* output roundtrip or fallback supplementary code point */ - c=((c&0x3ff)<<10)+unicodeCodeUnits[offset]+(0x10000-0xdc00); - goto finish; - } else if(UCNV_TO_U_USE_FALLBACK(cnv) ? (c&0xfffe)==0xe000 : c==0xe000) { - /* output roundtrip BMP code point above 0xd800 or fallback BMP code point */ - c=unicodeCodeUnits[offset]; - goto finish; - } else if(c==0xffff) { - /* callback(illegal) */ - goto illegal; - } else { - /* callback(unassigned) */ - goto unassigned; + } else { + /* callback(illegal) */ + *pErrorCode=U_ILLEGAL_CHAR_FOUND; + } + } else if(action==MBCS_STATE_VALID_DIRECT_16) { + /* output BMP code point */ + *target++=(UChar)MBCS_ENTRY_FINAL_VALUE_16(entry); + if(offsets!=NULL) { + *offsets++=sourceIndex; + } + byteIndex=0; + } else if(action==MBCS_STATE_VALID_16_PAIR) { + offset+=MBCS_ENTRY_FINAL_VALUE_16(entry); + c=unicodeCodeUnits[offset++]; + if(c<0xd800) { + /* output BMP code point below 0xd800 */ + *target++=c; + if(offsets!=NULL) { + *offsets++=sourceIndex; } - } else if(action==MBCS_STATE_VALID_DIRECT_20) { - /* output supplementary code point */ - c=(UChar32)(MBCS_ENTRY_FINAL_VALUE(entry)+0x10000); - goto finish; - } else if(action==MBCS_STATE_CHANGE_ONLY) { - /* - * This serves as a state change without any output. - * It is useful for reading simple stateful encodings, - * for example using just Shift-In/Shift-Out codes. - * The 21 unused bits may later be used for more sophisticated - * state transitions. - */ - } else if(action==MBCS_STATE_FALLBACK_DIRECT_16) { - if(!UCNV_TO_U_USE_FALLBACK(cnv)) { - /* callback(unassigned) */ - goto unassigned; + byteIndex=0; + } else if(UCNV_TO_U_USE_FALLBACK(cnv) ? c<=0xdfff : c<=0xdbff) { + /* output roundtrip or fallback surrogate pair */ + *target++=(UChar)(c&0xdbff); + if(offsets!=NULL) { + *offsets++=sourceIndex; } - /* output BMP code point */ - c=(UChar)MBCS_ENTRY_FINAL_VALUE_16(entry); - goto finish; - } else if(action==MBCS_STATE_FALLBACK_DIRECT_20) { - if(!UCNV_TO_U_USE_FALLBACK(cnv)) { - /* callback(unassigned) */ - goto unassigned; + byteIndex=0; + if(targetUCharErrorBuffer[0]=unicodeCodeUnits[offset]; + cnv->UCharErrorBufferLength=1; + *pErrorCode=U_BUFFER_OVERFLOW_ERROR; + + offset=0; + break; } - /* output supplementary code point */ - c=(UChar32)(MBCS_ENTRY_FINAL_VALUE(entry)+0x10000); - goto finish; - } else if(action==MBCS_STATE_UNASSIGNED) { - /* callback(unassigned) */ - goto unassigned; - } else if(action==MBCS_STATE_ILLEGAL) { + } else if(UCNV_TO_U_USE_FALLBACK(cnv) ? (c&0xfffe)==0xe000 : c==0xe000) { + /* output roundtrip BMP code point above 0xd800 or fallback BMP code point */ + *target++=unicodeCodeUnits[offset]; + if(offsets!=NULL) { + *offsets++=sourceIndex; + } + byteIndex=0; + } else if(c==0xffff) { /* callback(illegal) */ - goto illegal; - } else { - /* reserved, must never occur */ + *pErrorCode=U_ILLEGAL_CHAR_FOUND; + } + } else if(action==MBCS_STATE_VALID_DIRECT_20 || + (action==MBCS_STATE_FALLBACK_DIRECT_20 && UCNV_TO_U_USE_FALLBACK(cnv)) + ) { + entry=MBCS_ENTRY_FINAL_VALUE(entry); + /* output surrogate pair */ + *target++=(UChar)(0xd800|(UChar)(entry>>10)); + if(offsets!=NULL) { + *offsets++=sourceIndex; } - - /* normal end of action codes: prepare for a new character */ - offset=0; byteIndex=0; - continue; - -illegal: - reason=UCNV_ILLEGAL; - *pErrorCode=U_ILLEGAL_CHAR_FOUND; - goto callback; -unassigned: - reason=UCNV_UNASSIGNED; - *pErrorCode=U_INVALID_CHAR_FOUND; -callback: - /* call the callback function with all the preparations and post-processing */ - /* update the arguments structure */ - pArgs->source=(const char *)source; - pArgs->target=buffer; - pArgs->targetLimit=buffer+UTF_MAX_CHAR_LENGTH; - - /* set the converter state in UConverter to deal with the next character */ - cnv->toUnicodeStatus=0; - cnv->mode=state; - cnv->toULength=0; - - /* call the callback function */ - toUCallback(cnv, cnv->toUContext, pArgs, (const char *)bytes, byteIndex, reason, pErrorCode); - - /* get the converter state from UConverter */ - offset=cnv->toUnicodeStatus; - state=(uint8_t)cnv->mode; - byteIndex=cnv->toULength; - - /* update the source pointer */ - source=(const uint8_t *)pArgs->source; - - /* - * return the first character if the callback wrote some - * we do not need to goto finish because the converter state is already set - */ - if(U_SUCCESS(*pErrorCode)) { - entry=pArgs->target-buffer; - if(entry>0) { - return ucnv_getUChar32KeepOverflow(cnv, buffer, entry); + c=(UChar)(0xdc00|(UChar)(entry&0x3ff)); + if(targetUCharErrorBuffer[0]=c; + cnv->UCharErrorBufferLength=1; + *pErrorCode=U_BUFFER_OVERFLOW_ERROR; + offset=0; + break; + } + } else if(action==MBCS_STATE_CHANGE_ONLY) { /* - * We do not need to repeat the statements from the normal - * end of the action codes because we already updated all the - * necessary variables. + * This serves as a state change without any output. + * It is useful for reading simple stateful encodings, + * for example using just Shift-In/Shift-Out codes. + * The 21 unused bits may later be used for more sophisticated + * state transitions. */ + if(cnv->sharedData->mbcs.dbcsOnlyState==0) { + byteIndex=0; + } else { + /* SI/SO are illegal for DBCS-only conversion */ + state=(uint8_t)(cnv->mode); /* restore the previous state */ + + /* callback(illegal) */ + *pErrorCode=U_ILLEGAL_CHAR_FOUND; + } + } else if(action==MBCS_STATE_FALLBACK_DIRECT_16) { + if(UCNV_TO_U_USE_FALLBACK(cnv)) { + /* output BMP code point */ + *target++=(UChar)MBCS_ENTRY_FINAL_VALUE_16(entry); + if(offsets!=NULL) { + *offsets++=sourceIndex; + } + byteIndex=0; + } + } else if(action==MBCS_STATE_UNASSIGNED) { + /* just fall through */ + } else if(action==MBCS_STATE_ILLEGAL) { + /* callback(illegal) */ + *pErrorCode=U_ILLEGAL_CHAR_FOUND; + } else { + /* reserved, must never occur */ + byteIndex=0; } - } - if(byteIndex>0) { - /* incomplete character byte sequence */ - *pErrorCode=U_TRUNCATED_CHAR_FOUND; - state=0; - } else { - /* no output because of empty input or only state changes and skipping callbacks */ - *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR; + /* end of action codes: prepare for a new character */ + offset=0; + + if(byteIndex==0) { + sourceIndex=nextSourceIndex; + } else if(U_FAILURE(*pErrorCode)) { + /* callback(illegal) */ + if(byteIndex>1) { + /* + * Ticket 5691: consistent illegal sequences: + * - We include at least the first byte in the illegal sequence. + * - If any of the non-initial bytes could be the start of a character, + * we stop the illegal sequence before the first one of those. + */ + UBool isDBCSOnly=(UBool)(cnv->sharedData->mbcs.dbcsOnlyState!=0); + int8_t i; + for(i=1; + isource); + byteIndex=i; /* length of reported illegal byte sequence */ + if(backOutDistance<=bytesFromThisBuffer) { + source-=backOutDistance; + } else { + /* Back out bytes from the previous buffer: Need to replay them. */ + cnv->preToULength=(int8_t)(bytesFromThisBuffer-backOutDistance); + /* preToULength is negative! */ + uprv_memcpy(cnv->preToU, bytes+i, -cnv->preToULength); + source=(const uint8_t *)pArgs->source; + } + } + } + break; + } else /* unassigned sequences indicated with byteIndex>0 */ { + /* try an extension mapping */ + pArgs->source=(const char *)source; + byteIndex=_extToU(cnv, cnv->sharedData, + byteIndex, &source, sourceLimit, + &target, targetLimit, + &offsets, sourceIndex, + pArgs->flush, + pErrorCode); + sourceIndex=nextSourceIndex+=(int32_t)(source-(const uint8_t *)pArgs->source); + + if(U_FAILURE(*pErrorCode)) { + /* not mappable or buffer overflow */ + break; + } + } } - c=0xffff; -finish: - /* set the converter state back into UConverter, ready for a new character */ - cnv->toUnicodeStatus=0; + /* set the converter state back into UConverter */ + cnv->toUnicodeStatus=offset; cnv->mode=state; - cnv->toULength=0; + cnv->toULength=byteIndex; - /* write back the updated pointer */ + /* write back the updated pointers */ pArgs->source=(const char *)source; - return c; + pArgs->target=target; + pArgs->offsets=offsets; } /* - * This version of _MBCSGetNextUChar() is optimized for single-byte, single-state codepages. - * We still need a conversion loop in case a skipping callback is called. + * This version of ucnv_MBCSGetNextUChar() is optimized for single-byte, single-state codepages. + * We still need a conversion loop in case we find reserved action codes, which are to be ignored. */ static UChar32 -_MBCSSingleGetNextUChar(UConverterToUnicodeArgs *pArgs, +ucnv_MBCSSingleGetNextUChar(UConverterToUnicodeArgs *pArgs, UErrorCode *pErrorCode) { - UChar buffer[UTF_MAX_CHAR_LENGTH]; - UConverter *cnv; const int32_t (*stateTable)[256]; const uint8_t *source, *sourceLimit; int32_t entry; uint8_t action; - UConverterCallbackReason reason; /* set up the local pointers */ cnv=pArgs->converter; source=(const uint8_t *)pArgs->source; sourceLimit=(const uint8_t *)pArgs->sourceLimit; if((cnv->options&UCNV_OPTION_SWAP_LFNL)!=0) { - stateTable=(const int32_t (*)[256])cnv->sharedData->table->mbcs.swapLFNLStateTable; + stateTable=(const int32_t (*)[256])cnv->sharedData->mbcs.swapLFNLStateTable; } else { - stateTable=cnv->sharedData->table->mbcs.stateTable; + stateTable=cnv->sharedData->mbcs.stateTable; } /* conversion loop */ @@ -1868,188 +2766,184 @@ _MBCSSingleGetNextUChar(UConverterToUnicodeArgs *pArgs, * the most common cases compared to a switch. */ action=(uint8_t)(MBCS_ENTRY_FINAL_ACTION(entry)); - if(action==MBCS_STATE_VALID_DIRECT_20) { + if( action==MBCS_STATE_VALID_DIRECT_20 || + (action==MBCS_STATE_FALLBACK_DIRECT_20 && UCNV_TO_U_USE_FALLBACK(cnv)) + ) { /* output supplementary code point */ return (UChar32)(MBCS_ENTRY_FINAL_VALUE(entry)+0x10000); } else if(action==MBCS_STATE_FALLBACK_DIRECT_16) { - if(!UCNV_TO_U_USE_FALLBACK(cnv)) { - /* callback(unassigned) */ - reason=UCNV_UNASSIGNED; - *pErrorCode=U_INVALID_CHAR_FOUND; - } else { + if(UCNV_TO_U_USE_FALLBACK(cnv)) { /* output BMP code point */ return (UChar)MBCS_ENTRY_FINAL_VALUE_16(entry); } - } else if(action==MBCS_STATE_FALLBACK_DIRECT_20) { - if(!UCNV_TO_U_USE_FALLBACK(cnv)) { - /* callback(unassigned) */ - reason=UCNV_UNASSIGNED; - *pErrorCode=U_INVALID_CHAR_FOUND; - } else { - /* output supplementary code point */ - return (UChar32)(MBCS_ENTRY_FINAL_VALUE(entry)+0x10000); - } } else if(action==MBCS_STATE_UNASSIGNED) { - /* callback(unassigned) */ - reason=UCNV_UNASSIGNED; - *pErrorCode=U_INVALID_CHAR_FOUND; + /* just fall through */ } else if(action==MBCS_STATE_ILLEGAL) { /* callback(illegal) */ - reason=UCNV_ILLEGAL; *pErrorCode=U_ILLEGAL_CHAR_FOUND; } else { /* reserved, must never occur */ - *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR; - return 0xffff; + continue; } - /* call the callback function with all the preparations and post-processing */ - /* update the arguments structure */ - pArgs->target=buffer; - pArgs->targetLimit=buffer+UTF_MAX_CHAR_LENGTH; - - /* call the callback function */ - toUCallback(cnv, cnv->toUContext, pArgs, (const char *)(source-1), 1, reason, pErrorCode); - - /* update the source pointer */ - source=(const uint8_t *)pArgs->source; - - /* - * return the first character if the callback wrote some - * we do not need to goto finish because the converter state is already set - */ - if(U_SUCCESS(*pErrorCode)) { - entry=pArgs->target-buffer; - if(entry>0) { - return ucnv_getUChar32KeepOverflow(cnv, buffer, entry); - } - /* else (callback did not write anything) continue */ - } else if(*pErrorCode==U_BUFFER_OVERFLOW_ERROR) { - *pErrorCode=U_ZERO_ERROR; - return ucnv_getUChar32KeepOverflow(cnv, buffer, UTF_MAX_CHAR_LENGTH); - } else { - /* break on error */ - /* ### what if a callback set an error but _also_ generated output?! */ - return 0xffff; + if(U_FAILURE(*pErrorCode)) { + /* callback(illegal) */ + break; + } else /* unassigned sequence */ { + /* defer to the generic implementation */ + pArgs->source=(const char *)source-1; + return UCNV_GET_NEXT_UCHAR_USE_TO_U; } } - /* no output because of empty input or only state changes and skipping callbacks */ + /* no output because of empty input or only state changes */ *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR; return 0xffff; } /* - * This is a simple version of getNextUChar() that is used - * by other converter implementations. - * It does not use state from the converter, nor error codes. - * It does not handle the EBCDIC swaplfnl option (set in UConverter). + * Version of _MBCSToUnicodeWithOffsets() optimized for single-character + * conversion without offset handling. * - * Return value: - * U+fffe unassigned - * U+ffff illegal - * otherwise the Unicode code point + * When a character does not have a mapping to Unicode, then we return to the + * generic ucnv_getNextUChar() code for extension/GB 18030 and error/callback + * handling. + * We also defer to the generic code in other complicated cases and have them + * ultimately handled by _MBCSToUnicodeWithOffsets() itself. + * + * All normal mappings and errors are handled here. */ -U_CFUNC UChar32 -_MBCSSimpleGetNextUChar(UConverterSharedData *sharedData, - const char **pSource, const char *sourceLimit, - UBool useFallback) { - const uint8_t *source; +static UChar32 +ucnv_MBCSGetNextUChar(UConverterToUnicodeArgs *pArgs, + UErrorCode *pErrorCode) { + UConverter *cnv; + const uint8_t *source, *sourceLimit, *lastSource; const int32_t (*stateTable)[256]; const uint16_t *unicodeCodeUnits; uint32_t offset; - uint8_t state, action; + uint8_t state; int32_t entry; + UChar32 c; + uint8_t action; - /* set up the local pointers */ - source=(const uint8_t *)*pSource; - if(source>=(const uint8_t *)sourceLimit) { - /* no input at all: "illegal" */ - return 0xffff; + /* use optimized function if possible */ + cnv=pArgs->converter; + + if(cnv->preToULength>0) { + /* use the generic code in ucnv_getNextUChar() to continue with a partial match */ + return UCNV_GET_NEXT_UCHAR_USE_TO_U; } -#if 0 -/* - * Code disabled 2002dec09 (ICU 2.4) because it is not currently used in ICU. markus - * TODO In future releases, verify that this function is never called for SBCS - * conversions, i.e., that sharedData->table->mbcs.countStates==1 is still true. - * Removal improves code coverage. - */ - /* use optimized function if possible */ - if(sharedData->table->mbcs.countStates==1) { - return _MBCSSingleSimpleGetNextUChar(sharedData, (uint8_t)(*(*pSource)++), useFallback); + if(cnv->sharedData->mbcs.unicodeMask&UCNV_HAS_SURROGATES) { + /* + * Using the generic ucnv_getNextUChar() code lets us deal correctly + * with the rare case of a codepage that maps single surrogates + * without adding the complexity to this already complicated function here. + */ + return UCNV_GET_NEXT_UCHAR_USE_TO_U; + } else if(cnv->sharedData->mbcs.countStates==1) { + return ucnv_MBCSSingleGetNextUChar(pArgs, pErrorCode); } -#endif - stateTable=sharedData->table->mbcs.stateTable; - unicodeCodeUnits=sharedData->table->mbcs.unicodeCodeUnits; + /* set up the local pointers */ + source=lastSource=(const uint8_t *)pArgs->source; + sourceLimit=(const uint8_t *)pArgs->sourceLimit; - /* converter state */ - offset=0; - state=0; + if((cnv->options&UCNV_OPTION_SWAP_LFNL)!=0) { + stateTable=(const int32_t (*)[256])cnv->sharedData->mbcs.swapLFNLStateTable; + } else { + stateTable=cnv->sharedData->mbcs.stateTable; + } + unicodeCodeUnits=cnv->sharedData->mbcs.unicodeCodeUnits; + + /* get the converter state from UConverter */ + offset=cnv->toUnicodeStatus; + + /* + * if we are in the SBCS state for a DBCS-only converter, + * then load the DBCS state from the MBCS data + * (dbcsOnlyState==0 if it is not a DBCS-only converter) + */ + if((state=(uint8_t)(cnv->mode))==0) { + state=cnv->sharedData->mbcs.dbcsOnlyState; + } /* conversion loop */ - do { + c=U_SENTINEL; + while(sourcemode=state; + + /* set the next state early so that we can reuse the entry variable */ + state=(uint8_t)MBCS_ENTRY_FINAL_STATE(entry); /* typically 0 */ /* * An if-else-if chain provides more reliable performance for * the most common cases compared to a switch. */ action=(uint8_t)(MBCS_ENTRY_FINAL_ACTION(entry)); - if(action==MBCS_STATE_VALID_16) { + if(action==MBCS_STATE_VALID_DIRECT_16) { + /* output BMP code point */ + c=(UChar)MBCS_ENTRY_FINAL_VALUE_16(entry); + break; + } else if(action==MBCS_STATE_VALID_16) { offset+=MBCS_ENTRY_FINAL_VALUE_16(entry); - entry=unicodeCodeUnits[offset]; - if(entry!=0xfffe) { - return (UChar32)entry; - } else if(UCNV_TO_U_USE_FALLBACK(cnv)) { - return _MBCSGetFallback(&sharedData->table->mbcs, offset); + c=unicodeCodeUnits[offset]; + if(c<0xfffe) { + /* output BMP code point */ + break; + } else if(c==0xfffe) { + if(UCNV_TO_U_USE_FALLBACK(cnv) && (c=ucnv_MBCSGetFallback(&cnv->sharedData->mbcs, offset))!=0xfffe) { + break; + } } else { - return 0xfffe; + /* callback(illegal) */ + *pErrorCode=U_ILLEGAL_CHAR_FOUND; } - } else if(action==MBCS_STATE_VALID_DIRECT_16) { - /* output BMP code point */ - return (UChar)MBCS_ENTRY_FINAL_VALUE_16(entry); } else if(action==MBCS_STATE_VALID_16_PAIR) { offset+=MBCS_ENTRY_FINAL_VALUE_16(entry); - entry=unicodeCodeUnits[offset++]; - if(entry<0xd800) { + c=unicodeCodeUnits[offset++]; + if(c<0xd800) { /* output BMP code point below 0xd800 */ - return (UChar32)entry; - } else if(UCNV_TO_U_USE_FALLBACK(cnv) ? entry<=0xdfff : entry<=0xdbff) { + break; + } else if(UCNV_TO_U_USE_FALLBACK(cnv) ? c<=0xdfff : c<=0xdbff) { /* output roundtrip or fallback supplementary code point */ - return (UChar32)(((entry&0x3ff)<<10)+unicodeCodeUnits[offset]+(0x10000-0xdc00)); - } else if(UCNV_TO_U_USE_FALLBACK(cnv) ? (entry&0xfffe)==0xe000 : entry==0xe000) { + c=((c&0x3ff)<<10)+unicodeCodeUnits[offset]+(0x10000-0xdc00); + break; + } else if(UCNV_TO_U_USE_FALLBACK(cnv) ? (c&0xfffe)==0xe000 : c==0xe000) { /* output roundtrip BMP code point above 0xd800 or fallback BMP code point */ - return unicodeCodeUnits[offset]; - } else if(entry==0xffff) { - return 0xffff; - } else { - return 0xfffe; - } - } else if(action==MBCS_STATE_VALID_DIRECT_20) { - /* output supplementary code point */ - return 0x10000+MBCS_ENTRY_FINAL_VALUE(entry); - } else if(action==MBCS_STATE_FALLBACK_DIRECT_16) { - if(!TO_U_USE_FALLBACK(useFallback)) { - return 0xfffe; - } - /* output BMP code point */ - return (UChar)MBCS_ENTRY_FINAL_VALUE_16(entry); - } else if(action==MBCS_STATE_FALLBACK_DIRECT_20) { - if(!TO_U_USE_FALLBACK(useFallback)) { - return 0xfffe; + c=unicodeCodeUnits[offset]; + break; + } else if(c==0xffff) { + /* callback(illegal) */ + *pErrorCode=U_ILLEGAL_CHAR_FOUND; } + } else if(action==MBCS_STATE_VALID_DIRECT_20 || + (action==MBCS_STATE_FALLBACK_DIRECT_20 && UCNV_TO_U_USE_FALLBACK(cnv)) + ) { /* output supplementary code point */ - return 0x10000+MBCS_ENTRY_FINAL_VALUE(entry); + c=(UChar32)(MBCS_ENTRY_FINAL_VALUE(entry)+0x10000); + break; } else if(action==MBCS_STATE_CHANGE_ONLY) { /* * This serves as a state change without any output. @@ -2058,26 +2952,94 @@ _MBCSSimpleGetNextUChar(UConverterSharedData *sharedData, * The 21 unused bits may later be used for more sophisticated * state transitions. */ - if(source==(const uint8_t *)sourceLimit) { - /* if there are only state changes, then return "unassigned" */ - return 0xfffe; + if(cnv->sharedData->mbcs.dbcsOnlyState!=0) { + /* SI/SO are illegal for DBCS-only conversion */ + state=(uint8_t)(cnv->mode); /* restore the previous state */ + + /* callback(illegal) */ + *pErrorCode=U_ILLEGAL_CHAR_FOUND; + } + } else if(action==MBCS_STATE_FALLBACK_DIRECT_16) { + if(UCNV_TO_U_USE_FALLBACK(cnv)) { + /* output BMP code point */ + c=(UChar)MBCS_ENTRY_FINAL_VALUE_16(entry); + break; } } else if(action==MBCS_STATE_UNASSIGNED) { - return 0xfffe; + /* just fall through */ } else if(action==MBCS_STATE_ILLEGAL) { - return 0xffff; + /* callback(illegal) */ + *pErrorCode=U_ILLEGAL_CHAR_FOUND; } else { - /* reserved, must never occur */ + /* reserved (must never occur), or only state change */ + offset=0; + lastSource=source; + continue; } - /* state change only - prepare for a new character */ - state=(uint8_t)MBCS_ENTRY_FINAL_STATE(entry); /* typically 0 */ + /* end of action codes: prepare for a new character */ offset=0; + + if(U_FAILURE(*pErrorCode)) { + /* callback(illegal) */ + break; + } else /* unassigned sequence */ { + /* defer to the generic implementation */ + cnv->toUnicodeStatus=0; + cnv->mode=state; + pArgs->source=(const char *)lastSource; + return UCNV_GET_NEXT_UCHAR_USE_TO_U; + } } - } while(source<(const uint8_t *)sourceLimit); + } - *pSource=(const char *)source; - return 0xffff; + if(c<0) { + if(U_SUCCESS(*pErrorCode) && source==sourceLimit && lastSourcetoUBytes; + cnv->toULength=(int8_t)(source-lastSource); + do { + *bytes++=*lastSource++; + } while(lastSourcesharedData->mbcs.dbcsOnlyState!=0); + uint8_t *bytes=cnv->toUBytes; + *bytes++=*lastSource++; /* first byte */ + if(lastSource==source) { + cnv->toULength=1; + } else /* lastSourcetoULength=i; + source=lastSource; + } + } else { + /* no output because of empty input or only state changes */ + *pErrorCode=U_INDEX_OUTOFBOUNDS_ERROR; + } + c=0xffff; + } + + /* set the converter state back into UConverter, ready for a new character */ + cnv->toUnicodeStatus=0; + cnv->mode=state; + + /* write back the updated pointer */ + pArgs->source=(const char *)source; + return c; } #if 0 @@ -2086,16 +3048,17 @@ _MBCSSimpleGetNextUChar(UConverterSharedData *sharedData, * Removal improves code coverage. */ /** - * This version of _MBCSSimpleGetNextUChar() is optimized for single-byte, single-state codepages. + * This version of ucnv_MBCSSimpleGetNextUChar() is optimized for single-byte, single-state codepages. * It does not handle the EBCDIC swaplfnl option (set in UConverter). + * It does not handle conversion extensions (_extToU()). */ U_CFUNC UChar32 -_MBCSSingleSimpleGetNextUChar(UConverterSharedData *sharedData, +ucnv_MBCSSingleSimpleGetNextUChar(UConverterSharedData *sharedData, uint8_t b, UBool useFallback) { int32_t entry; uint8_t action; - entry=sharedData->table->mbcs.stateTable[0][b]; + entry=sharedData->mbcs.stateTable[0][b]; /* MBCS_ENTRY_IS_FINAL(entry) */ if(MBCS_ENTRY_FINAL_IS_VALID_DIRECT_16(entry)) { @@ -2134,11 +3097,164 @@ _MBCSSingleSimpleGetNextUChar(UConverterSharedData *sharedData, } #endif +/* + * This is a simple version of _MBCSGetNextUChar() that is used + * by other converter implementations. + * It only returns an "assigned" result if it consumes the entire input. + * It does not use state from the converter, nor error codes. + * It does not handle the EBCDIC swaplfnl option (set in UConverter). + * It handles conversion extensions but not GB 18030. + * + * Return value: + * U+fffe unassigned + * U+ffff illegal + * otherwise the Unicode code point + */ +U_CFUNC UChar32 +ucnv_MBCSSimpleGetNextUChar(UConverterSharedData *sharedData, + const char *source, int32_t length, + UBool useFallback) { + const int32_t (*stateTable)[256]; + const uint16_t *unicodeCodeUnits; + + uint32_t offset; + uint8_t state, action; + + UChar32 c; + int32_t i, entry; + + if(length<=0) { + /* no input at all: "illegal" */ + return 0xffff; + } + +#if 0 +/* + * Code disabled 2002dec09 (ICU 2.4) because it is not currently used in ICU. markus + * TODO In future releases, verify that this function is never called for SBCS + * conversions, i.e., that sharedData->mbcs.countStates==1 is still true. + * Removal improves code coverage. + */ + /* use optimized function if possible */ + if(sharedData->mbcs.countStates==1) { + if(length==1) { + return ucnv_MBCSSingleSimpleGetNextUChar(sharedData, (uint8_t)*source, useFallback); + } else { + return 0xffff; /* illegal: more than a single byte for an SBCS converter */ + } + } +#endif + + /* set up the local pointers */ + stateTable=sharedData->mbcs.stateTable; + unicodeCodeUnits=sharedData->mbcs.unicodeCodeUnits; + + /* converter state */ + offset=0; + state=sharedData->mbcs.dbcsOnlyState; + + /* conversion loop */ + for(i=0;;) { + entry=stateTable[state][(uint8_t)source[i++]]; + if(MBCS_ENTRY_IS_TRANSITION(entry)) { + state=(uint8_t)MBCS_ENTRY_TRANSITION_STATE(entry); + offset+=MBCS_ENTRY_TRANSITION_OFFSET(entry); + + if(i==length) { + return 0xffff; /* truncated character */ + } + } else { + /* + * An if-else-if chain provides more reliable performance for + * the most common cases compared to a switch. + */ + action=(uint8_t)(MBCS_ENTRY_FINAL_ACTION(entry)); + if(action==MBCS_STATE_VALID_16) { + offset+=MBCS_ENTRY_FINAL_VALUE_16(entry); + c=unicodeCodeUnits[offset]; + if(c!=0xfffe) { + /* done */ + } else if(UCNV_TO_U_USE_FALLBACK(cnv)) { + c=ucnv_MBCSGetFallback(&sharedData->mbcs, offset); + /* else done with 0xfffe */ + } + break; + } else if(action==MBCS_STATE_VALID_DIRECT_16) { + /* output BMP code point */ + c=(UChar)MBCS_ENTRY_FINAL_VALUE_16(entry); + break; + } else if(action==MBCS_STATE_VALID_16_PAIR) { + offset+=MBCS_ENTRY_FINAL_VALUE_16(entry); + c=unicodeCodeUnits[offset++]; + if(c<0xd800) { + /* output BMP code point below 0xd800 */ + } else if(UCNV_TO_U_USE_FALLBACK(cnv) ? c<=0xdfff : c<=0xdbff) { + /* output roundtrip or fallback supplementary code point */ + c=(UChar32)(((c&0x3ff)<<10)+unicodeCodeUnits[offset]+(0x10000-0xdc00)); + } else if(UCNV_TO_U_USE_FALLBACK(cnv) ? (c&0xfffe)==0xe000 : c==0xe000) { + /* output roundtrip BMP code point above 0xd800 or fallback BMP code point */ + c=unicodeCodeUnits[offset]; + } else if(c==0xffff) { + return 0xffff; + } else { + c=0xfffe; + } + break; + } else if(action==MBCS_STATE_VALID_DIRECT_20) { + /* output supplementary code point */ + c=0x10000+MBCS_ENTRY_FINAL_VALUE(entry); + break; + } else if(action==MBCS_STATE_FALLBACK_DIRECT_16) { + if(!TO_U_USE_FALLBACK(useFallback)) { + c=0xfffe; + break; + } + /* output BMP code point */ + c=(UChar)MBCS_ENTRY_FINAL_VALUE_16(entry); + break; + } else if(action==MBCS_STATE_FALLBACK_DIRECT_20) { + if(!TO_U_USE_FALLBACK(useFallback)) { + c=0xfffe; + break; + } + /* output supplementary code point */ + c=0x10000+MBCS_ENTRY_FINAL_VALUE(entry); + break; + } else if(action==MBCS_STATE_UNASSIGNED) { + c=0xfffe; + break; + } + + /* + * forbid MBCS_STATE_CHANGE_ONLY for this function, + * and MBCS_STATE_ILLEGAL and reserved action codes + */ + return 0xffff; + } + } + + if(i!=length) { + /* illegal for this function: not all input consumed */ + return 0xffff; + } + + if(c==0xfffe) { + /* try an extension mapping */ + const int32_t *cx=sharedData->mbcs.extIndexes; + if(cx!=NULL) { + return ucnv_extSimpleMatchToU(cx, source, length, useFallback); + } + } + + return c; +} + /* MBCS-from-Unicode conversion functions ----------------------------------- */ -U_CFUNC void -_MBCSFromUnicodeWithOffsets(UConverterFromUnicodeArgs *pArgs, - UErrorCode *pErrorCode) { +/* This version of ucnv_MBCSFromUnicodeWithOffsets() is optimized for double-byte codepages. */ +static void +ucnv_MBCSDoubleFromUnicodeWithOffsets(UConverterFromUnicodeArgs *pArgs, + UErrorCode *pErrorCode) { UConverter *cnv; const UChar *source, *sourceLimit; uint8_t *target; @@ -2146,72 +3262,46 @@ _MBCSFromUnicodeWithOffsets(UConverterFromUnicodeArgs *pArgs, int32_t *offsets; const uint16_t *table; - const uint8_t *p, *bytes; - uint8_t outputType; + const uint16_t *mbcsIndex; + const uint8_t *bytes; UChar32 c; - int32_t prevSourceIndex, sourceIndex, nextSourceIndex; + int32_t sourceIndex, nextSourceIndex; - UConverterCallbackReason reason; uint32_t stage2Entry; + uint32_t asciiRoundtrips; uint32_t value; - int32_t length, prevLength; uint8_t unicodeMask; /* use optimized function if possible */ cnv=pArgs->converter; - outputType=cnv->sharedData->table->mbcs.outputType; - unicodeMask=cnv->sharedData->table->mbcs.unicodeMask; - if(outputType==MBCS_OUTPUT_1 && !(unicodeMask&UCNV_HAS_SURROGATES)) { - if(!(unicodeMask&UCNV_HAS_SUPPLEMENTARY)) { - _MBCSSingleFromBMPWithOffsets(pArgs, pErrorCode); - } else { - _MBCSSingleFromUnicodeWithOffsets(pArgs, pErrorCode); - } - return; - } else if(outputType==MBCS_OUTPUT_2) { - _MBCSDoubleFromUnicodeWithOffsets(pArgs, pErrorCode); - return; - } + unicodeMask=cnv->sharedData->mbcs.unicodeMask; /* set up the local pointers */ source=pArgs->source; sourceLimit=pArgs->sourceLimit; target=(uint8_t *)pArgs->target; - targetCapacity=pArgs->targetLimit-pArgs->target; + targetCapacity=(int32_t)(pArgs->targetLimit-pArgs->target); offsets=pArgs->offsets; - table=cnv->sharedData->table->mbcs.fromUnicodeTable; + table=cnv->sharedData->mbcs.fromUnicodeTable; + mbcsIndex=cnv->sharedData->mbcs.mbcsIndex; if((cnv->options&UCNV_OPTION_SWAP_LFNL)!=0) { - bytes=cnv->sharedData->table->mbcs.swapLFNLFromUnicodeBytes; + bytes=cnv->sharedData->mbcs.swapLFNLFromUnicodeBytes; } else { - bytes=cnv->sharedData->table->mbcs.fromUnicodeBytes; + bytes=cnv->sharedData->mbcs.fromUnicodeBytes; } + asciiRoundtrips=cnv->sharedData->mbcs.asciiRoundtrips; /* get the converter state from UConverter */ - c=cnv->fromUSurrogateLead; - prevLength=cnv->fromUnicodeStatus; + c=cnv->fromUChar32; /* sourceIndex=-1 if the current character began in the previous buffer */ - prevSourceIndex=-1; sourceIndex= c==0 ? 0 : -1; nextSourceIndex=0; /* conversion loop */ - /* - * This is another piece of ugly code: - * A goto into the loop if the converter state contains a first surrogate - * from the previous function call. - * It saves me to check in each loop iteration a check of if(c==0) - * and duplicating the trail-surrogate-handling code in the else - * branch of that check. - * I could not find any other way to get around this other than - * using a function call for the conversion and callback, which would - * be even more inefficient. - * - * Markus Scherer 2000-jul-19 - */ if(c!=0 && targetCapacity>0) { goto getTrail; } @@ -2233,377 +3323,150 @@ _MBCSFromUnicodeWithOffsets(UConverterFromUnicodeArgs *pArgs, */ c=*source++; ++nextSourceIndex; + if(c<=0x7f && IS_ASCII_ROUNDTRIP(c, asciiRoundtrips)) { + *target++=(uint8_t)c; + if(offsets!=NULL) { + *offsets++=sourceIndex; + sourceIndex=nextSourceIndex; + } + --targetCapacity; + c=0; + continue; + } /* - * This also tests if the codepage maps single surrogates. - * If it does, then surrogates are not paired but mapped separately. - * Note that in this case unmatched surrogates are not detected. + * utf8Friendly table: Test for <=0xd7ff rather than <=MBCS_FAST_MAX + * to avoid dealing with surrogates. + * MBCS_FAST_MAX must be >=0xd7ff. */ - if(UTF_IS_SURROGATE(c) && !(unicodeMask&UCNV_HAS_SURROGATES)) { - if(UTF_IS_SURROGATE_FIRST(c)) { + if(c<=0xd7ff) { + value=DBCS_RESULT_FROM_MOST_BMP(mbcsIndex, (const uint16_t *)bytes, c); + /* There are only roundtrips (!=0) and no-mapping (==0) entries. */ + if(value==0) { + goto unassigned; + } + /* output the value */ + } else { + /* + * This also tests if the codepage maps single surrogates. + * If it does, then surrogates are not paired but mapped separately. + * Note that in this case unmatched surrogates are not detected. + */ + if(UTF_IS_SURROGATE(c) && !(unicodeMask&UCNV_HAS_SURROGATES)) { + if(UTF_IS_SURROGATE_FIRST(c)) { getTrail: - if(sourcefromUnicodeStatus=prevLength; /* save the old state */ + /* get the bytes and the length for the output */ + /* MBCS_OUTPUT_2 */ value=MBCS_VALUE_2_FROM_STAGE_2(bytes, stage2Entry, c); - if(value<=0xff) { - if(prevLength==1) { - length=1; - } else { - /* change from double-byte mode to single-byte */ - value|=(uint32_t)UCNV_SI<<8; - length=2; - prevLength=1; - } - } else { - if(prevLength==2) { - length=2; + + /* is this code point assigned, or do we use fallbacks? */ + if(!(MBCS_FROM_U_IS_ROUNDTRIP(stage2Entry, c) || + (UCNV_FROM_U_USE_FALLBACK(cnv, c) && value!=0)) + ) { + /* + * We allow a 0 byte output if the "assigned" bit is set for this entry. + * There is no way with this data structure for fallback output + * to be a zero byte. + */ + +unassigned: + /* try an extension mapping */ + pArgs->source=source; + c=_extFromU(cnv, cnv->sharedData, + c, &source, sourceLimit, + &target, target+targetCapacity, + &offsets, sourceIndex, + pArgs->flush, + pErrorCode); + nextSourceIndex+=(int32_t)(source-pArgs->source); + + if(U_FAILURE(*pErrorCode)) { + /* not mappable or buffer overflow */ + break; } else { - /* change from single-byte mode to double-byte */ - value|=(uint32_t)UCNV_SO<<16; - length=3; - prevLength=2; + /* a mapping was written to the target, continue */ + + /* recalculate the targetCapacity after an extension mapping */ + targetCapacity=(int32_t)(pArgs->targetLimit-(char *)target); + + /* normal end of conversion: prepare for a new character */ + sourceIndex=nextSourceIndex; + continue; } } - break; - case MBCS_OUTPUT_3: - p=MBCS_POINTER_3_FROM_STAGE_2(bytes, stage2Entry, c); - value=((uint32_t)*p<<16)|((uint32_t)p[1]<<8)|p[2]; - if(value<=0xff) { - length=1; - } else if(value<=0xffff) { - length=2; - } else { - length=3; - } - break; - case MBCS_OUTPUT_4: - value=MBCS_VALUE_4_FROM_STAGE_2(bytes, stage2Entry, c); - if(value<=0xff) { - length=1; - } else if(value<=0xffff) { - length=2; - } else if(value<=0xffffff) { - length=3; - } else { - length=4; - } - break; - case MBCS_OUTPUT_3_EUC: - value=MBCS_VALUE_2_FROM_STAGE_2(bytes, stage2Entry, c); - /* EUC 16-bit fixed-length representation */ - if(value<=0xff) { - length=1; - } else if((value&0x8000)==0) { - value|=0x8e8000; - length=3; - } else if((value&0x80)==0) { - value|=0x8f0080; - length=3; - } else { - length=2; - } - break; - case MBCS_OUTPUT_4_EUC: - p=MBCS_POINTER_3_FROM_STAGE_2(bytes, stage2Entry, c); - value=((uint32_t)*p<<16)|((uint32_t)p[1]<<8)|p[2]; - /* EUC 16-bit fixed-length representation applied to the first two bytes */ - if(value<=0xff) { - length=1; - } else if(value<=0xffff) { - length=2; - } else if((value&0x800000)==0) { - value|=0x8e800000; - length=4; - } else if((value&0x8000)==0) { - value|=0x8f008000; - length=4; - } else { - length=3; - } - break; - default: - /* must not occur */ - /* - * To avoid compiler warnings that value & length may be - * used without having been initialized, we set them here. - * In reality, this is unreachable code. - * Not having a default branch also causes warnings with - * some compilers. - */ - value=0; - length=0; - break; - } - - /* is this code point assigned, or do we use fallbacks? */ - if(!(MBCS_FROM_U_IS_ROUNDTRIP(stage2Entry, c)!=0 || - (UCNV_FROM_U_USE_FALLBACK(cnv, c) && (value!=0 || c==0))) - ) { - /* - * We allow a 0 byte output if the Unicode code point is - * U+0000 and also if the "assigned" bit is set for this entry. - * There is no way with this data structure for fallback output - * for other than U+0000 to be a zero byte. - */ - /* callback(unassigned) */ - goto unassigned; } /* write the output character bytes from value and length */ /* from the first if in the loop we know that targetCapacity>0 */ - if(length<=targetCapacity) { - if(offsets==NULL) { - switch(length) { - /* each branch falls through to the next one */ - case 4: - *target++=(uint8_t)(value>>24); - case 3: - *target++=(uint8_t)(value>>16); - case 2: - *target++=(uint8_t)(value>>8); - case 1: - *target++=(uint8_t)value; - default: - /* will never occur */ - break; - } - } else { - switch(length) { - /* each branch falls through to the next one */ - case 4: - *target++=(uint8_t)(value>>24); - *offsets++=sourceIndex; - case 3: - *target++=(uint8_t)(value>>16); - *offsets++=sourceIndex; - case 2: - *target++=(uint8_t)(value>>8); - *offsets++=sourceIndex; - case 1: - *target++=(uint8_t)value; - *offsets++=sourceIndex; - default: - /* will never occur */ - break; - } - } - targetCapacity-=length; - } else { - uint8_t *charErrorBuffer; - - /* - * We actually do this backwards here: - * In order to save an intermediate variable, we output - * first to the overflow buffer what does not fit into the - * regular target. - */ - /* we know that 1<=targetCapacitycharErrorBuffer; - switch(length) { - /* each branch falls through to the next one */ - case 3: - *charErrorBuffer++=(uint8_t)(value>>16); - case 2: - *charErrorBuffer++=(uint8_t)(value>>8); - case 1: - *charErrorBuffer=(uint8_t)value; - default: - /* will never occur */ - break; + if(value<=0xff) { + /* this is easy because we know that there is enough space */ + *target++=(uint8_t)value; + if(offsets!=NULL) { + *offsets++=sourceIndex; } - cnv->charErrorBufferLength=(int8_t)length; - - /* now output what fits into the regular target */ - value>>=8*length; /* length was reduced by targetCapacity */ - switch(targetCapacity) { - /* each branch falls through to the next one */ - case 3: - *target++=(uint8_t)(value>>16); + --targetCapacity; + } else /* length==2 */ { + *target++=(uint8_t)(value>>8); + if(2<=targetCapacity) { + *target++=(uint8_t)value; if(offsets!=NULL) { *offsets++=sourceIndex; - } - case 2: - *target++=(uint8_t)(value>>8); - if(offsets!=NULL) { *offsets++=sourceIndex; } - case 1: - *target++=(uint8_t)value; + targetCapacity-=2; + } else { if(offsets!=NULL) { *offsets++=sourceIndex; } - default: - /* will never occur */ + cnv->charErrorBuffer[0]=(char)value; + cnv->charErrorBufferLength=1; + + /* target overflow */ + targetCapacity=0; + *pErrorCode=U_BUFFER_OVERFLOW_ERROR; + c=0; break; } - - /* target overflow */ - targetCapacity=0; - *pErrorCode=U_BUFFER_OVERFLOW_ERROR; - c=0; - break; } /* normal end of conversion: prepare for a new character */ c=0; - if(offsets!=NULL) { - prevSourceIndex=sourceIndex; - sourceIndex=nextSourceIndex; - } + sourceIndex=nextSourceIndex; continue; - - /* - * This is the same ugly trick as in ToUnicode(), for the - * same reasons... - */ -unassigned: - reason=UCNV_UNASSIGNED; - *pErrorCode=U_INVALID_CHAR_FOUND; -callback: - /* call the callback function with all the preparations and post-processing */ - /* update the arguments structure */ - pArgs->source=source; - pArgs->target=(char *)target; - pArgs->offsets=offsets; - - /* set the converter state in UConverter to deal with the next character */ - cnv->fromUSurrogateLead=0; - /* - * Do not save the prevLength SISO state because prevLength is set for - * the character that is now not output because it is unassigned or it is - * a fallback that is not taken. - * The above branch for MBCS_OUTPUT_2_SISO has saved the previous state already. - * See comments there. - */ - prevSourceIndex=sourceIndex; - - /* call the callback function */ - fromUCallback(cnv, cnv->fromUContext, pArgs, c, reason, pErrorCode); - - /* get the converter state from UConverter */ - c=cnv->fromUSurrogateLead; - prevLength=cnv->fromUnicodeStatus; - - /* update target and deal with offsets if necessary */ - offsets=ucnv_updateCallbackOffsets(offsets, ((uint8_t *)pArgs->target)-target, sourceIndex); - target=(uint8_t *)pArgs->target; - - /* update the source pointer and index */ - sourceIndex=nextSourceIndex+(pArgs->source-source); - source=pArgs->source; - targetCapacity=(uint8_t *)pArgs->targetLimit-target; - - /* - * If the callback overflowed the target, then we need to - * stop here with an overflow indication. - */ - if(*pErrorCode==U_BUFFER_OVERFLOW_ERROR) { - break; - } else if(U_FAILURE(*pErrorCode)) { - /* break on error */ - c=0; - break; - } else if(cnv->charErrorBufferLength>0) { - /* target is full */ - *pErrorCode=U_BUFFER_OVERFLOW_ERROR; - break; - } - - /* - * We do not need to repeat the statements from the normal - * end of the conversion because we already updated all the - * necessary variables. - */ } else { /* target is full */ *pErrorCode=U_BUFFER_OVERFLOW_ERROR; @@ -2611,41 +3474,8 @@ callback: } } - if(pArgs->flush && source>=sourceLimit && U_SUCCESS(*pErrorCode)) { - /* end of input stream */ - if(c!=0) { - /* a Unicode code point remains incomplete (only a first surrogate) */ - *pErrorCode=U_TRUNCATED_CHAR_FOUND; - /* the following may change with Jitterbug 2449: would prepare for callback instead of resetting */ - c=0; - prevLength=1; - } else if(outputType==MBCS_OUTPUT_2_SISO && prevLength==2) { - /* EBCDIC_STATEFUL ending with DBCS: emit an SI to return the output stream to SBCS */ - if(targetCapacity>0) { - *target++=(uint8_t)UCNV_SI; - if(offsets!=NULL) { - /* set the last source character's index (sourceIndex points at sourceLimit now) */ - *offsets++=prevSourceIndex; - } - } else { - /* target is full */ - cnv->charErrorBuffer[0]=(char)UCNV_SI; - cnv->charErrorBufferLength=1; - *pErrorCode=U_BUFFER_OVERFLOW_ERROR; - } - prevLength=1; /* we switched into SBCS */ - } - - /* reset the state for the next conversion */ - if(U_SUCCESS(*pErrorCode)) { - c=0; - prevLength=1; - } - } - /* set the converter state back into UConverter */ - cnv->fromUSurrogateLead=(UChar)c; - cnv->fromUnicodeStatus=prevLength; + cnv->fromUChar32=c; /* write back the updated pointers */ pArgs->source=source; @@ -2653,9 +3483,9 @@ callback: pArgs->offsets=offsets; } -/* This version of _MBCSFromUnicodeWithOffsets() is optimized for double-byte codepages. */ +/* This version of ucnv_MBCSFromUnicodeWithOffsets() is optimized for single-byte codepages. */ static void -_MBCSDoubleFromUnicodeWithOffsets(UConverterFromUnicodeArgs *pArgs, +ucnv_MBCSSingleFromUnicodeWithOffsets(UConverterFromUnicodeArgs *pArgs, UErrorCode *pErrorCode) { UConverter *cnv; const UChar *source, *sourceLimit; @@ -2664,39 +3494,41 @@ _MBCSDoubleFromUnicodeWithOffsets(UConverterFromUnicodeArgs *pArgs, int32_t *offsets; const uint16_t *table; - const uint8_t *bytes; + const uint16_t *results; UChar32 c; int32_t sourceIndex, nextSourceIndex; - UConverterCallbackReason reason; - uint32_t stage2Entry; - uint32_t value; - int32_t length, prevLength; - uint8_t unicodeMask; - - /* use optimized function if possible */ - cnv=pArgs->converter; - unicodeMask=cnv->sharedData->table->mbcs.unicodeMask; + uint16_t value, minValue; + UBool hasSupplementary; /* set up the local pointers */ + cnv=pArgs->converter; source=pArgs->source; sourceLimit=pArgs->sourceLimit; target=(uint8_t *)pArgs->target; - targetCapacity=pArgs->targetLimit-pArgs->target; + targetCapacity=(int32_t)(pArgs->targetLimit-pArgs->target); offsets=pArgs->offsets; - table=cnv->sharedData->table->mbcs.fromUnicodeTable; + table=cnv->sharedData->mbcs.fromUnicodeTable; if((cnv->options&UCNV_OPTION_SWAP_LFNL)!=0) { - bytes=cnv->sharedData->table->mbcs.swapLFNLFromUnicodeBytes; + results=(uint16_t *)cnv->sharedData->mbcs.swapLFNLFromUnicodeBytes; + } else { + results=(uint16_t *)cnv->sharedData->mbcs.fromUnicodeBytes; + } + + if(cnv->useFallback) { + /* use all roundtrip and fallback results */ + minValue=0x800; } else { - bytes=cnv->sharedData->table->mbcs.fromUnicodeBytes; + /* use only roundtrips and fallbacks from private-use characters */ + minValue=0xc00; } + hasSupplementary=(UBool)(cnv->sharedData->mbcs.unicodeMask&UCNV_HAS_SUPPLEMENTARY); /* get the converter state from UConverter */ - c=cnv->fromUSurrogateLead; - prevLength=cnv->fromUnicodeStatus; + c=cnv->fromUChar32; /* sourceIndex=-1 if the current character began in the previous buffer */ sourceIndex= c==0 ? 0 : -1; @@ -2724,12 +3556,7 @@ _MBCSDoubleFromUnicodeWithOffsets(UConverterFromUnicodeArgs *pArgs, */ c=*source++; ++nextSourceIndex; - /* - * This also tests if the codepage maps single surrogates. - * If it does, then surrogates are not paired but mapped separately. - * Note that in this case unmatched surrogates are not detected. - */ - if(UTF_IS_SURROGATE(c) && !(unicodeMask&UCNV_HAS_SURROGATES)) { + if(UTF_IS_SURROGATE(c)) { if(UTF_IS_SURROGATE_FIRST(c)) { getTrail: if(source0 */ - if(length==1) { - /* this is easy because we know that there is enough space */ - *target++=(uint8_t)value; - if(offsets!=NULL) { - *offsets++=sourceIndex; - } - --targetCapacity; - } else /* length==2 */ { - *target++=(uint8_t)(value>>8); - if(2<=targetCapacity) { - *target++=(uint8_t)value; - if(offsets!=NULL) { - *offsets++=sourceIndex; - *offsets++=sourceIndex; - } - targetCapacity-=2; - } else { - if(offsets!=NULL) { - *offsets++=sourceIndex; - } - cnv->charErrorBuffer[0]=(char)value; - cnv->charErrorBufferLength=1; - - /* target overflow */ - targetCapacity=0; - *pErrorCode=U_BUFFER_OVERFLOW_ERROR; - c=0; - break; - } - } - - /* normal end of conversion: prepare for a new character */ - c=0; - sourceIndex=nextSourceIndex; - continue; - - /* - * This is the same ugly trick as in ToUnicode(), for the - * same reasons... - */ -unassigned: - reason=UCNV_UNASSIGNED; - *pErrorCode=U_INVALID_CHAR_FOUND; -callback: - /* call the callback function with all the preparations and post-processing */ - /* update the arguments structure */ - pArgs->source=source; - pArgs->target=(char *)target; - pArgs->offsets=offsets; - - /* set the converter state in UConverter to deal with the next character */ - cnv->fromUSurrogateLead=0; - cnv->fromUnicodeStatus=prevLength; - - /* call the callback function */ - fromUCallback(cnv, cnv->fromUContext, pArgs, c, reason, pErrorCode); - - /* get the converter state from UConverter */ - c=cnv->fromUSurrogateLead; - prevLength=cnv->fromUnicodeStatus; - - /* update target and deal with offsets if necessary */ - offsets=ucnv_updateCallbackOffsets(offsets, ((uint8_t *)pArgs->target)-target, sourceIndex); - target=(uint8_t *)pArgs->target; - - /* update the source pointer and index */ - sourceIndex=nextSourceIndex+(pArgs->source-source); - source=pArgs->source; - targetCapacity=(uint8_t *)pArgs->targetLimit-target; - - /* - * If the callback overflowed the target, then we need to - * stop here with an overflow indication. - */ - if(*pErrorCode==U_BUFFER_OVERFLOW_ERROR) { - break; - } else if(U_FAILURE(*pErrorCode)) { - /* break on error */ - c=0; - break; - } else if(cnv->charErrorBufferLength>0) { - /* target is full */ - *pErrorCode=U_BUFFER_OVERFLOW_ERROR; - break; - } - - /* - * We do not need to repeat the statements from the normal - * end of the conversion because we already updated all the - * necessary variables. - */ - } else { - /* target is full */ - *pErrorCode=U_BUFFER_OVERFLOW_ERROR; - break; - } - } - - if(pArgs->flush && source>=sourceLimit) { - /* reset the state for the next conversion */ - if(c!=0 && U_SUCCESS(*pErrorCode)) { - /* a Unicode code point remains incomplete (only a first surrogate) */ - *pErrorCode=U_TRUNCATED_CHAR_FOUND; - } - cnv->fromUSurrogateLead=0; - cnv->fromUnicodeStatus=1; - } else { - /* set the converter state back into UConverter */ - cnv->fromUSurrogateLead=(UChar)c; - cnv->fromUnicodeStatus=prevLength; - } - - /* write back the updated pointers */ - pArgs->source=source; - pArgs->target=(char *)target; - pArgs->offsets=offsets; -} - -/* This version of _MBCSFromUnicodeWithOffsets() is optimized for single-byte codepages. */ -static void -_MBCSSingleFromUnicodeWithOffsets(UConverterFromUnicodeArgs *pArgs, - UErrorCode *pErrorCode) { - UConverter *cnv; - const UChar *source, *sourceLimit; - uint8_t *target; - int32_t targetCapacity; - int32_t *offsets; - - const uint16_t *table; - const uint16_t *results; - - UChar32 c; - - int32_t sourceIndex, nextSourceIndex; - - UConverterCallbackReason reason; - uint16_t value, minValue; - UBool hasSupplementary; - - /* set up the local pointers */ - cnv=pArgs->converter; - source=pArgs->source; - sourceLimit=pArgs->sourceLimit; - target=(uint8_t *)pArgs->target; - targetCapacity=pArgs->targetLimit-pArgs->target; - offsets=pArgs->offsets; - - table=cnv->sharedData->table->mbcs.fromUnicodeTable; - if((cnv->options&UCNV_OPTION_SWAP_LFNL)!=0) { - results=(uint16_t *)cnv->sharedData->table->mbcs.swapLFNLFromUnicodeBytes; - } else { - results=(uint16_t *)cnv->sharedData->table->mbcs.fromUnicodeBytes; - } - - if(cnv->useFallback) { - /* use all roundtrip and fallback results */ - minValue=0x800; - } else { - /* use only roundtrips and fallbacks from private-use characters */ - minValue=0xc00; - } - hasSupplementary=(UBool)(cnv->sharedData->table->mbcs.unicodeMask&UCNV_HAS_SUPPLEMENTARY); - - /* get the converter state from UConverter */ - c=cnv->fromUSurrogateLead; - - /* sourceIndex=-1 if the current character began in the previous buffer */ - sourceIndex= c==0 ? 0 : -1; - nextSourceIndex=0; - - /* conversion loop */ - if(c!=0 && targetCapacity>0) { - goto getTrail; - } - - while(source0) { - /* - * Get a correct Unicode code point: - * a single UChar for a BMP code point or - * a matched surrogate pair for a "supplementary code point". - */ - c=*source++; - ++nextSourceIndex; - if(UTF_IS_SURROGATE(c)) { - if(UTF_IS_SURROGATE_FIRST(c)) { -getTrail: - if(source=minValue) { - /* assigned, write the output character bytes from value and length */ - /* length==1 */ + if(value>=minValue) { + /* assigned, write the output character bytes from value and length */ + /* length==1 */ /* this is easy because we know that there is enough space */ *target++=(uint8_t)value; if(offsets!=NULL) { @@ -3040,65 +3608,31 @@ getTrail: /* normal end of conversion: prepare for a new character */ c=0; sourceIndex=nextSourceIndex; - continue; } else { /* unassigned */ - /* - * We allow a 0 byte output if the Unicode code point is - * U+0000 and also if the "assigned" bit is set for this entry. - * There is no way with this data structure for fallback output - * for other than U+0000 to be a zero byte. - */ - /* callback(unassigned) */ - } unassigned: - reason=UCNV_UNASSIGNED; - *pErrorCode=U_INVALID_CHAR_FOUND; -callback: - /* call the callback function with all the preparations and post-processing */ - /* update the arguments structure */ - pArgs->source=source; - pArgs->target=(char *)target; - pArgs->offsets=offsets; - - /* set the converter state in UConverter to deal with the next character */ - cnv->fromUSurrogateLead=0; - - /* call the callback function */ - fromUCallback(cnv, cnv->fromUContext, pArgs, c, reason, pErrorCode); - - /* get the converter state from UConverter */ - c=cnv->fromUSurrogateLead; - - /* update target and deal with offsets if necessary */ - offsets=ucnv_updateCallbackOffsets(offsets, ((uint8_t *)pArgs->target)-target, sourceIndex); - target=(uint8_t *)pArgs->target; + /* try an extension mapping */ + pArgs->source=source; + c=_extFromU(cnv, cnv->sharedData, + c, &source, sourceLimit, + &target, target+targetCapacity, + &offsets, sourceIndex, + pArgs->flush, + pErrorCode); + nextSourceIndex+=(int32_t)(source-pArgs->source); + + if(U_FAILURE(*pErrorCode)) { + /* not mappable or buffer overflow */ + break; + } else { + /* a mapping was written to the target, continue */ - /* update the source pointer and index */ - sourceIndex=nextSourceIndex+(pArgs->source-source); - source=pArgs->source; - targetCapacity=(uint8_t *)pArgs->targetLimit-target; + /* recalculate the targetCapacity after an extension mapping */ + targetCapacity=(int32_t)(pArgs->targetLimit-(char *)target); - /* - * If the callback overflowed the target, then we need to - * stop here with an overflow indication. - */ - if(*pErrorCode==U_BUFFER_OVERFLOW_ERROR) { - break; - } else if(U_FAILURE(*pErrorCode)) { - /* break on error */ - c=0; - break; - } else if(cnv->charErrorBufferLength>0) { - /* target is full */ - *pErrorCode=U_BUFFER_OVERFLOW_ERROR; - break; + /* normal end of conversion: prepare for a new character */ + sourceIndex=nextSourceIndex; + } } - - /* - * We do not need to repeat the statements from the normal - * end of the conversion because we already updated all the - * necessary variables. - */ } else { /* target is full */ *pErrorCode=U_BUFFER_OVERFLOW_ERROR; @@ -3106,17 +3640,8 @@ callback: } } - if(pArgs->flush && source>=sourceLimit) { - /* reset the state for the next conversion */ - if(c!=0 && U_SUCCESS(*pErrorCode)) { - /* a Unicode code point remains incomplete (only a first surrogate) */ - *pErrorCode=U_TRUNCATED_CHAR_FOUND; - } - cnv->fromUSurrogateLead=0; - } else { - /* set the converter state back into UConverter */ - cnv->fromUSurrogateLead=(UChar)c; - } + /* set the converter state back into UConverter */ + cnv->fromUChar32=c; /* write back the updated pointers */ pArgs->source=source; @@ -3125,13 +3650,18 @@ callback: } /* - * This version of _MBCSFromUnicode() is optimized for single-byte codepages + * This version of ucnv_MBCSFromUnicode() is optimized for single-byte codepages * that map only to and from the BMP. * In addition to single-byte/state optimizations, the offset calculations * become much easier. + * It would be possible to use the sbcsIndex for UTF-8-friendly tables, + * but measurements have shown that this diminishes performance + * in more cases than it improves it. + * See SVN revision 21013 (2007-feb-06) for the last version with #if switches + * for various MBCS and SBCS optimizations. */ static void -_MBCSSingleFromBMPWithOffsets(UConverterFromUnicodeArgs *pArgs, +ucnv_MBCSSingleFromBMPWithOffsets(UConverterFromUnicodeArgs *pArgs, UErrorCode *pErrorCode) { UConverter *cnv; const UChar *source, *sourceLimit, *lastSource; @@ -3146,7 +3676,7 @@ _MBCSSingleFromBMPWithOffsets(UConverterFromUnicodeArgs *pArgs, int32_t sourceIndex; - UConverterCallbackReason reason; + uint32_t asciiRoundtrips; uint16_t value, minValue; /* set up the local pointers */ @@ -3154,15 +3684,16 @@ _MBCSSingleFromBMPWithOffsets(UConverterFromUnicodeArgs *pArgs, source=pArgs->source; sourceLimit=pArgs->sourceLimit; target=(uint8_t *)pArgs->target; - targetCapacity=pArgs->targetLimit-pArgs->target; + targetCapacity=(int32_t)(pArgs->targetLimit-pArgs->target); offsets=pArgs->offsets; - table=cnv->sharedData->table->mbcs.fromUnicodeTable; + table=cnv->sharedData->mbcs.fromUnicodeTable; if((cnv->options&UCNV_OPTION_SWAP_LFNL)!=0) { - results=(uint16_t *)cnv->sharedData->table->mbcs.swapLFNLFromUnicodeBytes; + results=(uint16_t *)cnv->sharedData->mbcs.swapLFNLFromUnicodeBytes; } else { - results=(uint16_t *)cnv->sharedData->table->mbcs.fromUnicodeBytes; + results=(uint16_t *)cnv->sharedData->mbcs.fromUnicodeBytes; } + asciiRoundtrips=cnv->sharedData->mbcs.asciiRoundtrips; if(cnv->useFallback) { /* use all roundtrip and fallback results */ @@ -3173,7 +3704,7 @@ _MBCSSingleFromBMPWithOffsets(UConverterFromUnicodeArgs *pArgs, } /* get the converter state from UConverter */ - c=cnv->fromUSurrogateLead; + c=cnv->fromUChar32; /* sourceIndex=-1 if the current character began in the previous buffer */ sourceIndex= c==0 ? 0 : -1; @@ -3183,7 +3714,7 @@ _MBCSSingleFromBMPWithOffsets(UConverterFromUnicodeArgs *pArgs, * since the conversion here is 1:1 UChar:uint8_t, we need only one counter * for the minimum of the sourceLength and targetCapacity */ - length=sourceLimit-source; + length=(int32_t)(sourceLimit-source); if(length=minValue) { /* assigned, write the output character bytes from value and length */ @@ -3270,15 +3806,6 @@ unrolled: continue; } else if(!UTF_IS_SURROGATE(c)) { /* normal, unassigned BMP character */ - /* - * We allow a 0 byte output if the Unicode code point is - * U+0000 and also if the "assigned" bit is set for this entry. - * There is no way with this data structure for fallback output - * for other than U+0000 to be a zero byte. - */ - /* callback(unassigned) */ - reason=UCNV_UNASSIGNED; - *pErrorCode=U_INVALID_CHAR_FOUND; } else if(UTF_IS_SURROGATE_FIRST(c)) { getTrail: if(sourceflush) { + *pErrorCode=U_TRUNCATED_CHAR_FOUND; + } break; } } else { /* this is an unmatched trail code unit (2nd surrogate) */ /* callback(illegal) */ - reason=UCNV_ILLEGAL; *pErrorCode=U_ILLEGAL_CHAR_FOUND; + break; } - /* call the callback function with all the preparations and post-processing */ - /* get the number of code units for c to correctly advance sourceIndex after the callback call */ - length=UTF_CHAR_LENGTH(c); + /* c does not have a mapping */ + + /* get the number of code units for c to correctly advance sourceIndex */ + length=U16_LENGTH(c); - /* set offsets since the start or the last callback */ + /* set offsets since the start or the last extension */ if(offsets!=NULL) { int32_t count=(int32_t)(source-lastSource); - /* do not set the offset for the callback-causing character */ + /* do not set the offset for this character */ count-=length; while(count>0) { *offsets++=sourceIndex++; --count; } - /* offset and sourceIndex are now set for the current character */ - } - - /* update the arguments structure */ - pArgs->source=source; - pArgs->target=(char *)target; - pArgs->offsets=offsets; - - /* set the converter state in UConverter to deal with the next character */ - cnv->fromUSurrogateLead=0; - - /* call the callback function */ - fromUCallback(cnv, cnv->fromUContext, pArgs, c, reason, pErrorCode); - - /* get the converter state from UConverter */ - c=cnv->fromUSurrogateLead; - - /* update target and deal with offsets if necessary */ - offsets=ucnv_updateCallbackOffsets(offsets, ((uint8_t *)pArgs->target)-target, sourceIndex); - target=(uint8_t *)pArgs->target; - - /* update the source pointer and index */ - sourceIndex+=length+(pArgs->source-source); - source=lastSource=pArgs->source; - targetCapacity=(uint8_t *)pArgs->targetLimit-target; - length=sourceLimit-source; - if(lengthcharErrorBufferLength>0) { - /* target is full */ - *pErrorCode=U_BUFFER_OVERFLOW_ERROR; + /* try an extension mapping */ + lastSource=source; + c=_extFromU(cnv, cnv->sharedData, + c, &source, sourceLimit, + &target, (const uint8_t *)(pArgs->targetLimit), + &offsets, sourceIndex, + pArgs->flush, + pErrorCode); + sourceIndex+=length+(int32_t)(source-lastSource); + lastSource=source; + + if(U_FAILURE(*pErrorCode)) { + /* not mappable or buffer overflow */ break; + } else { + /* a mapping was written to the target, continue */ + + /* recalculate the targetCapacity after an extension mapping */ + targetCapacity=(int32_t)(pArgs->targetLimit-(char *)target); + length=(int32_t)(sourceLimit-source); + if(length 0 && *pErrorCode == U_TRUNCATED_CHAR_FOUND) { + /* + Caller gave us a partial supplementary character, + which this function couldn't convert in any case. + The callback will handle the offset. + */ + count--; + } while(count>0) { *offsets++=sourceIndex++; --count; } } - if(pArgs->flush && source>=sourceLimit) { - /* reset the state for the next conversion */ - if(c!=0 && U_SUCCESS(*pErrorCode)) { - /* a Unicode code point remains incomplete (only a first surrogate) */ - *pErrorCode=U_TRUNCATED_CHAR_FOUND; - } - cnv->fromUSurrogateLead=0; - } else { - /* set the converter state back into UConverter */ - cnv->fromUSurrogateLead=(UChar)c; - } + /* set the converter state back into UConverter */ + cnv->fromUChar32=c; /* write back the updated pointers */ pArgs->source=source; @@ -3407,192 +3917,1571 @@ getTrail: pArgs->offsets=offsets; } -/* - * This is another simple conversion function for internal use by other - * conversion implementations. - * It does not use the converter state nor call callbacks. - * It does not handle the EBCDIC swaplfnl option (set in UConverter). - * - * It converts one single Unicode code point into codepage bytes, encoded - * as one 32-bit value. The function returns the number of bytes in *pValue: - * 1..4 the number of bytes in *pValue - * 0 unassigned (*pValue undefined) - * -1 illegal (currently not used, *pValue undefined) - * - * *pValue will contain the resulting bytes with the last byte in bits 7..0, - * the second to last byte in bits 15..8, etc. - * Currently, the function assumes but does not check that 0<=c<=0x10ffff. - */ -U_CFUNC int32_t -_MBCSFromUChar32(UConverterSharedData *sharedData, - UChar32 c, uint32_t *pValue, - UBool useFallback) { - const uint16_t *table=sharedData->table->mbcs.fromUnicodeTable; - const uint8_t *p; +U_CFUNC void +ucnv_MBCSFromUnicodeWithOffsets(UConverterFromUnicodeArgs *pArgs, + UErrorCode *pErrorCode) { + UConverter *cnv; + const UChar *source, *sourceLimit; + uint8_t *target; + int32_t targetCapacity; + int32_t *offsets; + + const uint16_t *table; + const uint16_t *mbcsIndex; + const uint8_t *p, *bytes; + uint8_t outputType; + + UChar32 c; + + int32_t prevSourceIndex, sourceIndex, nextSourceIndex; + + uint32_t stage2Entry; + uint32_t asciiRoundtrips; + uint32_t value; + uint8_t si_value[2] = {0, 0}; + uint8_t so_value[2] = {0, 0}; + uint8_t si_value_length, so_value_length; + int32_t length = 0, prevLength; + uint8_t unicodeMask; + + cnv=pArgs->converter; + + if(cnv->preFromUFirstCP>=0) { + /* + * pass sourceIndex=-1 because we continue from an earlier buffer + * in the future, this may change with continuous offsets + */ + ucnv_extContinueMatchFromU(cnv, pArgs, -1, pErrorCode); + + if(U_FAILURE(*pErrorCode) || cnv->preFromULength<0) { + return; + } + } + + /* use optimized function if possible */ + outputType=cnv->sharedData->mbcs.outputType; + unicodeMask=cnv->sharedData->mbcs.unicodeMask; + if(outputType==MBCS_OUTPUT_1 && !(unicodeMask&UCNV_HAS_SURROGATES)) { + if(!(unicodeMask&UCNV_HAS_SUPPLEMENTARY)) { + ucnv_MBCSSingleFromBMPWithOffsets(pArgs, pErrorCode); + } else { + ucnv_MBCSSingleFromUnicodeWithOffsets(pArgs, pErrorCode); + } + return; + } else if(outputType==MBCS_OUTPUT_2 && cnv->sharedData->mbcs.utf8Friendly) { + ucnv_MBCSDoubleFromUnicodeWithOffsets(pArgs, pErrorCode); + return; + } + + /* set up the local pointers */ + source=pArgs->source; + sourceLimit=pArgs->sourceLimit; + target=(uint8_t *)pArgs->target; + targetCapacity=(int32_t)(pArgs->targetLimit-pArgs->target); + offsets=pArgs->offsets; + + table=cnv->sharedData->mbcs.fromUnicodeTable; + if(cnv->sharedData->mbcs.utf8Friendly) { + mbcsIndex=cnv->sharedData->mbcs.mbcsIndex; + } else { + mbcsIndex=NULL; + } + if((cnv->options&UCNV_OPTION_SWAP_LFNL)!=0) { + bytes=cnv->sharedData->mbcs.swapLFNLFromUnicodeBytes; + } else { + bytes=cnv->sharedData->mbcs.fromUnicodeBytes; + } + asciiRoundtrips=cnv->sharedData->mbcs.asciiRoundtrips; + + /* get the converter state from UConverter */ + c=cnv->fromUChar32; + + if(outputType==MBCS_OUTPUT_2_SISO) { + prevLength=cnv->fromUnicodeStatus; + if(prevLength==0) { + /* set the real value */ + prevLength=1; + } + } else { + /* prevent fromUnicodeStatus from being set to something non-0 */ + prevLength=0; + } + + /* sourceIndex=-1 if the current character began in the previous buffer */ + prevSourceIndex=-1; + sourceIndex= c==0 ? 0 : -1; + nextSourceIndex=0; + + /* Get the SI/SO character for the converter */ + si_value_length = getSISOBytes(SI, cnv->options, si_value); + so_value_length = getSISOBytes(SO, cnv->options, so_value); + + /* conversion loop */ + /* + * This is another piece of ugly code: + * A goto into the loop if the converter state contains a first surrogate + * from the previous function call. + * It saves me to check in each loop iteration a check of if(c==0) + * and duplicating the trail-surrogate-handling code in the else + * branch of that check. + * I could not find any other way to get around this other than + * using a function call for the conversion and callback, which would + * be even more inefficient. + * + * Markus Scherer 2000-jul-19 + */ + if(c!=0 && targetCapacity>0) { + goto getTrail; + } + + while(source0) { + /* + * Get a correct Unicode code point: + * a single UChar for a BMP code point or + * a matched surrogate pair for a "supplementary code point". + */ + c=*source++; + ++nextSourceIndex; + if(c<=0x7f && IS_ASCII_ROUNDTRIP(c, asciiRoundtrips)) { + *target++=(uint8_t)c; + if(offsets!=NULL) { + *offsets++=sourceIndex; + prevSourceIndex=sourceIndex; + sourceIndex=nextSourceIndex; + } + --targetCapacity; + c=0; + continue; + } + /* + * utf8Friendly table: Test for <=0xd7ff rather than <=MBCS_FAST_MAX + * to avoid dealing with surrogates. + * MBCS_FAST_MAX must be >=0xd7ff. + */ + if(c<=0xd7ff && mbcsIndex!=NULL) { + value=mbcsIndex[c>>6]; + + /* get the bytes and the length for the output (copied from below and adapted for utf8Friendly data) */ + /* There are only roundtrips (!=0) and no-mapping (==0) entries. */ + switch(outputType) { + case MBCS_OUTPUT_2: + value=((const uint16_t *)bytes)[value +(c&0x3f)]; + if(value<=0xff) { + if(value==0) { + goto unassigned; + } else { + length=1; + } + } else { + length=2; + } + break; + case MBCS_OUTPUT_2_SISO: + /* 1/2-byte stateful with Shift-In/Shift-Out */ + /* + * Save the old state in the converter object + * right here, then change the local prevLength state variable if necessary. + * Then, if this character turns out to be unassigned or a fallback that + * is not taken, the callback code must not save the new state in the converter + * because the new state is for a character that is not output. + * However, the callback must still restore the state from the converter + * in case the callback function changed it for its output. + */ + cnv->fromUnicodeStatus=prevLength; /* save the old state */ + value=((const uint16_t *)bytes)[value +(c&0x3f)]; + if(value<=0xff) { + if(value==0) { + goto unassigned; + } else if(prevLength<=1) { + length=1; + } else { + /* change from double-byte mode to single-byte */ + if (si_value_length == 1) { + value|=(uint32_t)si_value[0]<<8; + length = 2; + } else if (si_value_length == 2) { + value|=(uint32_t)si_value[1]<<8; + value|=(uint32_t)si_value[0]<<16; + length = 3; + } + prevLength=1; + } + } else { + if(prevLength==2) { + length=2; + } else { + /* change from single-byte mode to double-byte */ + if (so_value_length == 1) { + value|=(uint32_t)so_value[0]<<16; + length = 3; + } else if (so_value_length == 2) { + value|=(uint32_t)so_value[1]<<16; + value|=(uint32_t)so_value[0]<<24; + length = 4; + } + prevLength=2; + } + } + break; + case MBCS_OUTPUT_DBCS_ONLY: + /* table with single-byte results, but only DBCS mappings used */ + value=((const uint16_t *)bytes)[value +(c&0x3f)]; + if(value<=0xff) { + /* no mapping or SBCS result, not taken for DBCS-only */ + goto unassigned; + } else { + length=2; + } + break; + case MBCS_OUTPUT_3: + p=bytes+(value+(c&0x3f))*3; + value=((uint32_t)*p<<16)|((uint32_t)p[1]<<8)|p[2]; + if(value<=0xff) { + if(value==0) { + goto unassigned; + } else { + length=1; + } + } else if(value<=0xffff) { + length=2; + } else { + length=3; + } + break; + case MBCS_OUTPUT_4: + value=((const uint32_t *)bytes)[value +(c&0x3f)]; + if(value<=0xff) { + if(value==0) { + goto unassigned; + } else { + length=1; + } + } else if(value<=0xffff) { + length=2; + } else if(value<=0xffffff) { + length=3; + } else { + length=4; + } + break; + case MBCS_OUTPUT_3_EUC: + value=((const uint16_t *)bytes)[value +(c&0x3f)]; + /* EUC 16-bit fixed-length representation */ + if(value<=0xff) { + if(value==0) { + goto unassigned; + } else { + length=1; + } + } else if((value&0x8000)==0) { + value|=0x8e8000; + length=3; + } else if((value&0x80)==0) { + value|=0x8f0080; + length=3; + } else { + length=2; + } + break; + case MBCS_OUTPUT_4_EUC: + p=bytes+(value+(c&0x3f))*3; + value=((uint32_t)*p<<16)|((uint32_t)p[1]<<8)|p[2]; + /* EUC 16-bit fixed-length representation applied to the first two bytes */ + if(value<=0xff) { + if(value==0) { + goto unassigned; + } else { + length=1; + } + } else if(value<=0xffff) { + length=2; + } else if((value&0x800000)==0) { + value|=0x8e800000; + length=4; + } else if((value&0x8000)==0) { + value|=0x8f008000; + length=4; + } else { + length=3; + } + break; + default: + /* must not occur */ + /* + * To avoid compiler warnings that value & length may be + * used without having been initialized, we set them here. + * In reality, this is unreachable code. + * Not having a default branch also causes warnings with + * some compilers. + */ + value=0; + length=0; + break; + } + /* output the value */ + } else { + /* + * This also tests if the codepage maps single surrogates. + * If it does, then surrogates are not paired but mapped separately. + * Note that in this case unmatched surrogates are not detected. + */ + if(UTF_IS_SURROGATE(c) && !(unicodeMask&UCNV_HAS_SURROGATES)) { + if(UTF_IS_SURROGATE_FIRST(c)) { +getTrail: + if(sourcefromUnicodeStatus=prevLength; /* save the old state */ + /* callback(unassigned) */ + goto unassigned; + } + /* convert this supplementary code point */ + /* exit this condition tree */ + } else { + /* this is an unmatched lead code unit (1st surrogate) */ + /* callback(illegal) */ + *pErrorCode=U_ILLEGAL_CHAR_FOUND; + break; + } + } else { + /* no more input */ + break; + } + } else { + /* this is an unmatched trail code unit (2nd surrogate) */ + /* callback(illegal) */ + *pErrorCode=U_ILLEGAL_CHAR_FOUND; + break; + } + } + + /* convert the Unicode code point in c into codepage bytes */ + + /* + * The basic lookup is a triple-stage compact array (trie) lookup. + * For details see the beginning of this file. + * + * Single-byte codepages are handled with a different data structure + * by _MBCSSingle... functions. + * + * The result consists of a 32-bit value from stage 2 and + * a pointer to as many bytes as are stored per character. + * The pointer points to the character's bytes in stage 3. + * Bits 15..0 of the stage 2 entry contain the stage 3 index + * for that pointer, while bits 31..16 are flags for which of + * the 16 characters in the block are roundtrip-assigned. + * + * For 2-byte and 4-byte codepages, the bytes are stored as uint16_t + * respectively as uint32_t, in the platform encoding. + * For 3-byte codepages, the bytes are always stored in big-endian order. + * + * For EUC encodings that use only either 0x8e or 0x8f as the first + * byte of their longest byte sequences, the first two bytes in + * this third stage indicate with their 7th bits whether these bytes + * are to be written directly or actually need to be preceeded by + * one of the two Single-Shift codes. With this, the third stage + * stores one byte fewer per character than the actual maximum length of + * EUC byte sequences. + * + * Other than that, leading zero bytes are removed and the other + * bytes output. A single zero byte may be output if the "assigned" + * bit in stage 2 was on. + * The data structure does not support zero byte output as a fallback, + * and also does not allow output of leading zeros. + */ + stage2Entry=MBCS_STAGE_2_FROM_U(table, c); + + /* get the bytes and the length for the output */ + switch(outputType) { + case MBCS_OUTPUT_2: + value=MBCS_VALUE_2_FROM_STAGE_2(bytes, stage2Entry, c); + if(value<=0xff) { + length=1; + } else { + length=2; + } + break; + case MBCS_OUTPUT_2_SISO: + /* 1/2-byte stateful with Shift-In/Shift-Out */ + /* + * Save the old state in the converter object + * right here, then change the local prevLength state variable if necessary. + * Then, if this character turns out to be unassigned or a fallback that + * is not taken, the callback code must not save the new state in the converter + * because the new state is for a character that is not output. + * However, the callback must still restore the state from the converter + * in case the callback function changed it for its output. + */ + cnv->fromUnicodeStatus=prevLength; /* save the old state */ + value=MBCS_VALUE_2_FROM_STAGE_2(bytes, stage2Entry, c); + if(value<=0xff) { + if(value==0 && MBCS_FROM_U_IS_ROUNDTRIP(stage2Entry, c)==0) { + /* no mapping, leave value==0 */ + length=0; + } else if(prevLength<=1) { + length=1; + } else { + /* change from double-byte mode to single-byte */ + if (si_value_length == 1) { + value|=(uint32_t)si_value[0]<<8; + length = 2; + } else if (si_value_length == 2) { + value|=(uint32_t)si_value[1]<<8; + value|=(uint32_t)si_value[0]<<16; + length = 3; + } + prevLength=1; + } + } else { + if(prevLength==2) { + length=2; + } else { + /* change from single-byte mode to double-byte */ + if (so_value_length == 1) { + value|=(uint32_t)so_value[0]<<16; + length = 3; + } else if (so_value_length == 2) { + value|=(uint32_t)so_value[1]<<16; + value|=(uint32_t)so_value[0]<<24; + length = 4; + } + prevLength=2; + } + } + break; + case MBCS_OUTPUT_DBCS_ONLY: + /* table with single-byte results, but only DBCS mappings used */ + value=MBCS_VALUE_2_FROM_STAGE_2(bytes, stage2Entry, c); + if(value<=0xff) { + /* no mapping or SBCS result, not taken for DBCS-only */ + value=stage2Entry=0; /* stage2Entry=0 to reset roundtrip flags */ + length=0; + } else { + length=2; + } + break; + case MBCS_OUTPUT_3: + p=MBCS_POINTER_3_FROM_STAGE_2(bytes, stage2Entry, c); + value=((uint32_t)*p<<16)|((uint32_t)p[1]<<8)|p[2]; + if(value<=0xff) { + length=1; + } else if(value<=0xffff) { + length=2; + } else { + length=3; + } + break; + case MBCS_OUTPUT_4: + value=MBCS_VALUE_4_FROM_STAGE_2(bytes, stage2Entry, c); + if(value<=0xff) { + length=1; + } else if(value<=0xffff) { + length=2; + } else if(value<=0xffffff) { + length=3; + } else { + length=4; + } + break; + case MBCS_OUTPUT_3_EUC: + value=MBCS_VALUE_2_FROM_STAGE_2(bytes, stage2Entry, c); + /* EUC 16-bit fixed-length representation */ + if(value<=0xff) { + length=1; + } else if((value&0x8000)==0) { + value|=0x8e8000; + length=3; + } else if((value&0x80)==0) { + value|=0x8f0080; + length=3; + } else { + length=2; + } + break; + case MBCS_OUTPUT_4_EUC: + p=MBCS_POINTER_3_FROM_STAGE_2(bytes, stage2Entry, c); + value=((uint32_t)*p<<16)|((uint32_t)p[1]<<8)|p[2]; + /* EUC 16-bit fixed-length representation applied to the first two bytes */ + if(value<=0xff) { + length=1; + } else if(value<=0xffff) { + length=2; + } else if((value&0x800000)==0) { + value|=0x8e800000; + length=4; + } else if((value&0x8000)==0) { + value|=0x8f008000; + length=4; + } else { + length=3; + } + break; + default: + /* must not occur */ + /* + * To avoid compiler warnings that value & length may be + * used without having been initialized, we set them here. + * In reality, this is unreachable code. + * Not having a default branch also causes warnings with + * some compilers. + */ + value=stage2Entry=0; /* stage2Entry=0 to reset roundtrip flags */ + length=0; + break; + } + + /* is this code point assigned, or do we use fallbacks? */ + if(!(MBCS_FROM_U_IS_ROUNDTRIP(stage2Entry, c)!=0 || + (UCNV_FROM_U_USE_FALLBACK(cnv, c) && value!=0)) + ) { + /* + * We allow a 0 byte output if the "assigned" bit is set for this entry. + * There is no way with this data structure for fallback output + * to be a zero byte. + */ + +unassigned: + /* try an extension mapping */ + pArgs->source=source; + c=_extFromU(cnv, cnv->sharedData, + c, &source, sourceLimit, + &target, target+targetCapacity, + &offsets, sourceIndex, + pArgs->flush, + pErrorCode); + nextSourceIndex+=(int32_t)(source-pArgs->source); + prevLength=cnv->fromUnicodeStatus; /* restore SISO state */ + + if(U_FAILURE(*pErrorCode)) { + /* not mappable or buffer overflow */ + break; + } else { + /* a mapping was written to the target, continue */ + + /* recalculate the targetCapacity after an extension mapping */ + targetCapacity=(int32_t)(pArgs->targetLimit-(char *)target); + + /* normal end of conversion: prepare for a new character */ + if(offsets!=NULL) { + prevSourceIndex=sourceIndex; + sourceIndex=nextSourceIndex; + } + continue; + } + } + } + + /* write the output character bytes from value and length */ + /* from the first if in the loop we know that targetCapacity>0 */ + if(length<=targetCapacity) { + if(offsets==NULL) { + switch(length) { + /* each branch falls through to the next one */ + case 4: + *target++=(uint8_t)(value>>24); + case 3: + *target++=(uint8_t)(value>>16); + case 2: + *target++=(uint8_t)(value>>8); + case 1: + *target++=(uint8_t)value; + default: + /* will never occur */ + break; + } + } else { + switch(length) { + /* each branch falls through to the next one */ + case 4: + *target++=(uint8_t)(value>>24); + *offsets++=sourceIndex; + case 3: + *target++=(uint8_t)(value>>16); + *offsets++=sourceIndex; + case 2: + *target++=(uint8_t)(value>>8); + *offsets++=sourceIndex; + case 1: + *target++=(uint8_t)value; + *offsets++=sourceIndex; + default: + /* will never occur */ + break; + } + } + targetCapacity-=length; + } else { + uint8_t *charErrorBuffer; + + /* + * We actually do this backwards here: + * In order to save an intermediate variable, we output + * first to the overflow buffer what does not fit into the + * regular target. + */ + /* we know that 1<=targetCapacitycharErrorBuffer; + switch(length) { + /* each branch falls through to the next one */ + case 3: + *charErrorBuffer++=(uint8_t)(value>>16); + case 2: + *charErrorBuffer++=(uint8_t)(value>>8); + case 1: + *charErrorBuffer=(uint8_t)value; + default: + /* will never occur */ + break; + } + cnv->charErrorBufferLength=(int8_t)length; + + /* now output what fits into the regular target */ + value>>=8*length; /* length was reduced by targetCapacity */ + switch(targetCapacity) { + /* each branch falls through to the next one */ + case 3: + *target++=(uint8_t)(value>>16); + if(offsets!=NULL) { + *offsets++=sourceIndex; + } + case 2: + *target++=(uint8_t)(value>>8); + if(offsets!=NULL) { + *offsets++=sourceIndex; + } + case 1: + *target++=(uint8_t)value; + if(offsets!=NULL) { + *offsets++=sourceIndex; + } + default: + /* will never occur */ + break; + } + + /* target overflow */ + targetCapacity=0; + *pErrorCode=U_BUFFER_OVERFLOW_ERROR; + c=0; + break; + } + + /* normal end of conversion: prepare for a new character */ + c=0; + if(offsets!=NULL) { + prevSourceIndex=sourceIndex; + sourceIndex=nextSourceIndex; + } + continue; + } else { + /* target is full */ + *pErrorCode=U_BUFFER_OVERFLOW_ERROR; + break; + } + } + + /* + * the end of the input stream and detection of truncated input + * are handled by the framework, but for EBCDIC_STATEFUL conversion + * we need to emit an SI at the very end + * + * conditions: + * successful + * EBCDIC_STATEFUL in DBCS mode + * end of input and no truncated input + */ + if( U_SUCCESS(*pErrorCode) && + outputType==MBCS_OUTPUT_2_SISO && prevLength==2 && + pArgs->flush && source>=sourceLimit && c==0 + ) { + /* EBCDIC_STATEFUL ending with DBCS: emit an SI to return the output stream to SBCS */ + if(targetCapacity>0) { + *target++=(uint8_t)si_value[0]; + if (si_value_length == 2) { + if (targetCapacity<2) { + cnv->charErrorBuffer[0]=(uint8_t)si_value[1]; + cnv->charErrorBufferLength=1; + *pErrorCode=U_BUFFER_OVERFLOW_ERROR; + } else { + *target++=(uint8_t)si_value[1]; + } + } + if(offsets!=NULL) { + /* set the last source character's index (sourceIndex points at sourceLimit now) */ + *offsets++=prevSourceIndex; + } + } else { + /* target is full */ + cnv->charErrorBuffer[0]=(uint8_t)si_value[0]; + if (si_value_length == 2) { + cnv->charErrorBuffer[1]=(uint8_t)si_value[1]; + } + cnv->charErrorBufferLength=si_value_length; + *pErrorCode=U_BUFFER_OVERFLOW_ERROR; + } + prevLength=1; /* we switched into SBCS */ + } + + /* set the converter state back into UConverter */ + cnv->fromUChar32=c; + cnv->fromUnicodeStatus=prevLength; + + /* write back the updated pointers */ + pArgs->source=source; + pArgs->target=(char *)target; + pArgs->offsets=offsets; +} + +/* + * This is another simple conversion function for internal use by other + * conversion implementations. + * It does not use the converter state nor call callbacks. + * It does not handle the EBCDIC swaplfnl option (set in UConverter). + * It handles conversion extensions but not GB 18030. + * + * It converts one single Unicode code point into codepage bytes, encoded + * as one 32-bit value. The function returns the number of bytes in *pValue: + * 1..4 the number of bytes in *pValue + * 0 unassigned (*pValue undefined) + * -1 illegal (currently not used, *pValue undefined) + * + * *pValue will contain the resulting bytes with the last byte in bits 7..0, + * the second to last byte in bits 15..8, etc. + * Currently, the function assumes but does not check that 0<=c<=0x10ffff. + */ +U_CFUNC int32_t +ucnv_MBCSFromUChar32(UConverterSharedData *sharedData, + UChar32 c, uint32_t *pValue, + UBool useFallback) { + const int32_t *cx; + const uint16_t *table; +#if 0 +/* #if 0 because this is not currently used in ICU - reduce code, increase code coverage */ + const uint8_t *p; +#endif + uint32_t stage2Entry; + uint32_t value; + int32_t length; + + /* BMP-only codepages are stored without stage 1 entries for supplementary code points */ + if(c<=0xffff || (sharedData->mbcs.unicodeMask&UCNV_HAS_SUPPLEMENTARY)) { + table=sharedData->mbcs.fromUnicodeTable; + + /* convert the Unicode code point in c into codepage bytes (same as in _MBCSFromUnicodeWithOffsets) */ + if(sharedData->mbcs.outputType==MBCS_OUTPUT_1) { + value=MBCS_SINGLE_RESULT_FROM_U(table, (uint16_t *)sharedData->mbcs.fromUnicodeBytes, c); + /* is this code point assigned, or do we use fallbacks? */ + if(useFallback ? value>=0x800 : value>=0xc00) { + *pValue=value&0xff; + return 1; + } + } else /* outputType!=MBCS_OUTPUT_1 */ { + stage2Entry=MBCS_STAGE_2_FROM_U(table, c); + + /* get the bytes and the length for the output */ + switch(sharedData->mbcs.outputType) { + case MBCS_OUTPUT_2: + value=MBCS_VALUE_2_FROM_STAGE_2(sharedData->mbcs.fromUnicodeBytes, stage2Entry, c); + if(value<=0xff) { + length=1; + } else { + length=2; + } + break; +#if 0 +/* #if 0 because this is not currently used in ICU - reduce code, increase code coverage */ + case MBCS_OUTPUT_DBCS_ONLY: + /* table with single-byte results, but only DBCS mappings used */ + value=MBCS_VALUE_2_FROM_STAGE_2(sharedData->mbcs.fromUnicodeBytes, stage2Entry, c); + if(value<=0xff) { + /* no mapping or SBCS result, not taken for DBCS-only */ + value=stage2Entry=0; /* stage2Entry=0 to reset roundtrip flags */ + length=0; + } else { + length=2; + } + break; + case MBCS_OUTPUT_3: + p=MBCS_POINTER_3_FROM_STAGE_2(sharedData->mbcs.fromUnicodeBytes, stage2Entry, c); + value=((uint32_t)*p<<16)|((uint32_t)p[1]<<8)|p[2]; + if(value<=0xff) { + length=1; + } else if(value<=0xffff) { + length=2; + } else { + length=3; + } + break; + case MBCS_OUTPUT_4: + value=MBCS_VALUE_4_FROM_STAGE_2(sharedData->mbcs.fromUnicodeBytes, stage2Entry, c); + if(value<=0xff) { + length=1; + } else if(value<=0xffff) { + length=2; + } else if(value<=0xffffff) { + length=3; + } else { + length=4; + } + break; + case MBCS_OUTPUT_3_EUC: + value=MBCS_VALUE_2_FROM_STAGE_2(sharedData->mbcs.fromUnicodeBytes, stage2Entry, c); + /* EUC 16-bit fixed-length representation */ + if(value<=0xff) { + length=1; + } else if((value&0x8000)==0) { + value|=0x8e8000; + length=3; + } else if((value&0x80)==0) { + value|=0x8f0080; + length=3; + } else { + length=2; + } + break; + case MBCS_OUTPUT_4_EUC: + p=MBCS_POINTER_3_FROM_STAGE_2(sharedData->mbcs.fromUnicodeBytes, stage2Entry, c); + value=((uint32_t)*p<<16)|((uint32_t)p[1]<<8)|p[2]; + /* EUC 16-bit fixed-length representation applied to the first two bytes */ + if(value<=0xff) { + length=1; + } else if(value<=0xffff) { + length=2; + } else if((value&0x800000)==0) { + value|=0x8e800000; + length=4; + } else if((value&0x8000)==0) { + value|=0x8f008000; + length=4; + } else { + length=3; + } + break; +#endif + default: + /* must not occur */ + return -1; + } + + /* is this code point assigned, or do we use fallbacks? */ + if( MBCS_FROM_U_IS_ROUNDTRIP(stage2Entry, c) || + (FROM_U_USE_FALLBACK(useFallback, c) && value!=0) + ) { + /* + * We allow a 0 byte output if the "assigned" bit is set for this entry. + * There is no way with this data structure for fallback output + * to be a zero byte. + */ + /* assigned */ + *pValue=value; + return length; + } + } + } + + cx=sharedData->mbcs.extIndexes; + if(cx!=NULL) { + length=ucnv_extSimpleMatchFromU(cx, c, pValue, useFallback); + return length>=0 ? length : -length; /* return abs(length); */ + } + + /* unassigned */ + return 0; +} + + +#if 0 +/* + * This function has been moved to ucnv2022.c for inlining. + * This implementation is here only for documentation purposes + */ + +/** + * This version of ucnv_MBCSFromUChar32() is optimized for single-byte codepages. + * It does not handle the EBCDIC swaplfnl option (set in UConverter). + * It does not handle conversion extensions (_extFromU()). + * + * It returns the codepage byte for the code point, or -1 if it is unassigned. + */ +U_CFUNC int32_t +ucnv_MBCSSingleFromUChar32(UConverterSharedData *sharedData, + UChar32 c, + UBool useFallback) { + const uint16_t *table; + int32_t value; + + /* BMP-only codepages are stored without stage 1 entries for supplementary code points */ + if(c>=0x10000 && !(sharedData->mbcs.unicodeMask&UCNV_HAS_SUPPLEMENTARY)) { + return -1; + } + + /* convert the Unicode code point in c into codepage bytes (same as in _MBCSFromUnicodeWithOffsets) */ + table=sharedData->mbcs.fromUnicodeTable; + + /* get the byte for the output */ + value=MBCS_SINGLE_RESULT_FROM_U(table, (uint16_t *)sharedData->mbcs.fromUnicodeBytes, c); + /* is this code point assigned, or do we use fallbacks? */ + if(useFallback ? value>=0x800 : value>=0xc00) { + return value&0xff; + } else { + return -1; + } +} +#endif + +/* MBCS-from-UTF-8 conversion functions ------------------------------------- */ + +/* minimum code point values for n-byte UTF-8 sequences, n=0..4 */ +static const UChar32 +utf8_minLegal[5]={ 0, 0, 0x80, 0x800, 0x10000 }; + +/* offsets for n-byte UTF-8 sequences that were calculated with ((lead<<6)+trail)<<6+trail... */ +static const UChar32 +utf8_offsets[7]={ 0, 0, 0x3080, 0xE2080, 0x3C82080 }; + +static void +ucnv_SBCSFromUTF8(UConverterFromUnicodeArgs *pFromUArgs, + UConverterToUnicodeArgs *pToUArgs, + UErrorCode *pErrorCode) { + UConverter *utf8, *cnv; + const uint8_t *source, *sourceLimit; + uint8_t *target; + int32_t targetCapacity; + + const uint16_t *table, *sbcsIndex; + const uint16_t *results; + + int8_t oldToULength, toULength, toULimit; + + UChar32 c; + uint8_t b, t1, t2; + + uint32_t asciiRoundtrips; + uint16_t value, minValue; + UBool hasSupplementary; + + /* set up the local pointers */ + utf8=pToUArgs->converter; + cnv=pFromUArgs->converter; + source=(uint8_t *)pToUArgs->source; + sourceLimit=(uint8_t *)pToUArgs->sourceLimit; + target=(uint8_t *)pFromUArgs->target; + targetCapacity=(int32_t)(pFromUArgs->targetLimit-pFromUArgs->target); + + table=cnv->sharedData->mbcs.fromUnicodeTable; + sbcsIndex=cnv->sharedData->mbcs.sbcsIndex; + if((cnv->options&UCNV_OPTION_SWAP_LFNL)!=0) { + results=(uint16_t *)cnv->sharedData->mbcs.swapLFNLFromUnicodeBytes; + } else { + results=(uint16_t *)cnv->sharedData->mbcs.fromUnicodeBytes; + } + asciiRoundtrips=cnv->sharedData->mbcs.asciiRoundtrips; + + if(cnv->useFallback) { + /* use all roundtrip and fallback results */ + minValue=0x800; + } else { + /* use only roundtrips and fallbacks from private-use characters */ + minValue=0xc00; + } + hasSupplementary=(UBool)(cnv->sharedData->mbcs.unicodeMask&UCNV_HAS_SUPPLEMENTARY); + + /* get the converter state from the UTF-8 UConverter */ + c=(UChar32)utf8->toUnicodeStatus; + if(c!=0) { + toULength=oldToULength=utf8->toULength; + toULimit=(int8_t)utf8->mode; + } else { + toULength=oldToULength=toULimit=0; + } + + /* + * Make sure that the last byte sequence before sourceLimit is complete + * or runs into a lead byte. + * Do not go back into the bytes that will be read for finishing a partial + * sequence from the previous buffer. + * In the conversion loop compare source with sourceLimit only once + * per multi-byte character. + */ + { + int32_t i, length; + + length=(int32_t)(sourceLimit-source) - (toULimit-oldToULength); + for(i=0; i<3 && i0) { + utf8->toUnicodeStatus=0; + utf8->toULength=0; + goto moreBytes; + /* + * Note: We could avoid the goto by duplicating some of the moreBytes + * code, but only up to the point of collecting a complete UTF-8 + * sequence; then recurse for the toUBytes[toULength] + * and then continue with normal conversion. + * + * If so, move this code to just after initializing the minimum + * set of local variables for reading the UTF-8 input + * (utf8, source, target, limits but not cnv, table, minValue, etc.). + * + * Potential advantages: + * - avoid the goto + * - oldToULength could become a local variable in just those code blocks + * that deal with buffer boundaries + * - possibly faster if the goto prevents some compiler optimizations + * (this would need measuring to confirm) + * Disadvantage: + * - code duplication + */ + } + + /* conversion loop */ + while(source0) { + b=*source++; + if((int8_t)b>=0) { + /* convert ASCII */ + if(IS_ASCII_ROUNDTRIP(b, asciiRoundtrips)) { + *target++=(uint8_t)b; + --targetCapacity; + continue; + } else { + c=b; + value=SBCS_RESULT_FROM_UTF8(sbcsIndex, results, 0, c); + } + } else { + if(b<0xe0) { + if( /* handle U+0080..U+07FF inline */ + b>=0xc2 && + (t1=(uint8_t)(*source-0x80)) <= 0x3f + ) { + c=b&0x1f; + ++source; + value=SBCS_RESULT_FROM_UTF8(sbcsIndex, results, c, t1); + if(value>=minValue) { + *target++=(uint8_t)value; + --targetCapacity; + continue; + } else { + c=(c<<6)|t1; + } + } else { + c=-1; + } + } else if(b==0xe0) { + if( /* handle U+0800..U+0FFF inline */ + (t1=(uint8_t)(source[0]-0x80)) <= 0x3f && t1 >= 0x20 && + (t2=(uint8_t)(source[1]-0x80)) <= 0x3f + ) { + c=t1; + source+=2; + value=SBCS_RESULT_FROM_UTF8(sbcsIndex, results, c, t2); + if(value>=minValue) { + *target++=(uint8_t)value; + --targetCapacity; + continue; + } else { + c=(c<<6)|t2; + } + } else { + c=-1; + } + } else { + c=-1; + } + + if(c<0) { + /* handle "complicated" and error cases, and continuing partial characters */ + oldToULength=0; + toULength=1; + toULimit=utf8_countTrailBytes[b]+1; + c=b; +moreBytes: + while(toULengthsourceLimit) { + b=*source; + if(U8_IS_TRAIL(b)) { + ++source; + ++toULength; + c=(c<<6)+b; + } else { + break; /* sequence too short, stop with toULengthtoUBytes[oldToULength++]=*source++; + } + utf8->toUnicodeStatus=c; + utf8->toULength=toULength; + utf8->mode=toULimit; + pToUArgs->source=(char *)source; + pFromUArgs->target=(char *)target; + return; + } + } + + if( toULength==toULimit && /* consumed all trail bytes */ + (toULength==3 || toULength==2) && /* BMP */ + (c-=utf8_offsets[toULength])>=utf8_minLegal[toULength] && + (c<=0xd7ff || 0xe000<=c) /* not a surrogate */ + ) { + value=MBCS_SINGLE_RESULT_FROM_U(table, results, c); + } else if( + toULength==toULimit && toULength==4 && + (0x10000<=(c-=utf8_offsets[4]) && c<=0x10ffff) + ) { + /* supplementary code point */ + if(!hasSupplementary) { + /* BMP-only codepages are stored without stage 1 entries for supplementary code points */ + value=0; + } else { + value=MBCS_SINGLE_RESULT_FROM_U(table, results, c); + } + } else { + /* error handling: illegal UTF-8 byte sequence */ + source-=(toULength-oldToULength); + while(oldToULengthtoUBytes[oldToULength++]=*source++; + } + utf8->toULength=toULength; + pToUArgs->source=(char *)source; + pFromUArgs->target=(char *)target; + *pErrorCode=U_ILLEGAL_CHAR_FOUND; + return; + } + } + } + + if(value>=minValue) { + /* output the mapping for c */ + *target++=(uint8_t)value; + --targetCapacity; + } else { + /* valueUTF-16->charset conversion. + */ + static const UChar nul=0; + const UChar *noSource=&nul; + c=_extFromU(cnv, cnv->sharedData, + c, &noSource, noSource, + &target, target+targetCapacity, + NULL, -1, + pFromUArgs->flush, + pErrorCode); + + if(U_FAILURE(*pErrorCode)) { + /* not mappable or buffer overflow */ + cnv->fromUChar32=c; + break; + } else if(cnv->preFromUFirstCP>=0) { + /* + * Partial match, return and revert to pivoting. + * In normal from-UTF-16 conversion, we would just continue + * but then exit the loop because the extension match would + * have consumed the source. + */ + break; + } else { + /* a mapping was written to the target, continue */ + + /* recalculate the targetCapacity after an extension mapping */ + targetCapacity=(int32_t)(pFromUArgs->targetLimit-(char *)target); + } + } + } else { + /* target is full */ + *pErrorCode=U_BUFFER_OVERFLOW_ERROR; + break; + } + } + + /* + * The sourceLimit may have been adjusted before the conversion loop + * to stop before a truncated sequence. + * If so, then collect the truncated sequence now. + */ + if(U_SUCCESS(*pErrorCode) && source<(sourceLimit=(uint8_t *)pToUArgs->sourceLimit)) { + c=utf8->toUBytes[0]=b=*source++; + toULength=1; + toULimit=utf8_countTrailBytes[b]+1; + while(sourcetoUBytes[toULength++]=b=*source++; + c=(c<<6)+b; + } + utf8->toUnicodeStatus=c; + utf8->toULength=toULength; + utf8->mode=toULimit; + } + + /* write back the updated pointers */ + pToUArgs->source=(char *)source; + pFromUArgs->target=(char *)target; +} + +static void +ucnv_DBCSFromUTF8(UConverterFromUnicodeArgs *pFromUArgs, + UConverterToUnicodeArgs *pToUArgs, + UErrorCode *pErrorCode) { + UConverter *utf8, *cnv; + const uint8_t *source, *sourceLimit; + uint8_t *target; + int32_t targetCapacity; + + const uint16_t *table, *mbcsIndex; + const uint16_t *results; + + int8_t oldToULength, toULength, toULimit; + + UChar32 c; + uint8_t b, t1, t2; + uint32_t stage2Entry; - uint32_t value; - int32_t length; + uint32_t asciiRoundtrips; + uint16_t value, minValue; + UBool hasSupplementary; - /* BMP-only codepages are stored without stage 1 entries for supplementary code points */ - if(c>=0x10000 && !(sharedData->table->mbcs.unicodeMask&UCNV_HAS_SUPPLEMENTARY)) { - return 0; + /* set up the local pointers */ + utf8=pToUArgs->converter; + cnv=pFromUArgs->converter; + source=(uint8_t *)pToUArgs->source; + sourceLimit=(uint8_t *)pToUArgs->sourceLimit; + target=(uint8_t *)pFromUArgs->target; + targetCapacity=(int32_t)(pFromUArgs->targetLimit-pFromUArgs->target); + + table=cnv->sharedData->mbcs.fromUnicodeTable; + mbcsIndex=cnv->sharedData->mbcs.mbcsIndex; + if((cnv->options&UCNV_OPTION_SWAP_LFNL)!=0) { + results=(uint16_t *)cnv->sharedData->mbcs.swapLFNLFromUnicodeBytes; + } else { + results=(uint16_t *)cnv->sharedData->mbcs.fromUnicodeBytes; } + asciiRoundtrips=cnv->sharedData->mbcs.asciiRoundtrips; - /* convert the Unicode code point in c into codepage bytes (same as in _MBCSFromUnicodeWithOffsets) */ - if(sharedData->table->mbcs.outputType==MBCS_OUTPUT_1) { - value=MBCS_SINGLE_RESULT_FROM_U(table, (uint16_t *)sharedData->table->mbcs.fromUnicodeBytes, c); - /* is this code point assigned, or do we use fallbacks? */ - if(useFallback ? value>=0x800 : value>=0xc00) { - *pValue=value&0xff; - return 1; - } else { - return 0; - } + if(cnv->useFallback) { + /* use all roundtrip and fallback results */ + minValue=0x800; + } else { + /* use only roundtrips and fallbacks from private-use characters */ + minValue=0xc00; } + hasSupplementary=(UBool)(cnv->sharedData->mbcs.unicodeMask&UCNV_HAS_SUPPLEMENTARY); - stage2Entry=MBCS_STAGE_2_FROM_U(table, c); + /* get the converter state from the UTF-8 UConverter */ + c=(UChar32)utf8->toUnicodeStatus; + if(c!=0) { + toULength=oldToULength=utf8->toULength; + toULimit=(int8_t)utf8->mode; + } else { + toULength=oldToULength=toULimit=0; + } - /* get the bytes and the length for the output */ - switch(sharedData->table->mbcs.outputType) { - case MBCS_OUTPUT_2: - value=MBCS_VALUE_2_FROM_STAGE_2(sharedData->table->mbcs.fromUnicodeBytes, stage2Entry, c); - if(value<=0xff) { - length=1; - } else { - length=2; - } - break; - case MBCS_OUTPUT_3: - p=MBCS_POINTER_3_FROM_STAGE_2(sharedData->table->mbcs.fromUnicodeBytes, stage2Entry, c); - value=((uint32_t)*p<<16)|((uint32_t)p[1]<<8)|p[2]; - if(value<=0xff) { - length=1; - } else if(value<=0xffff) { - length=2; - } else { - length=3; - } - break; - case MBCS_OUTPUT_4: - value=MBCS_VALUE_4_FROM_STAGE_2(sharedData->table->mbcs.fromUnicodeBytes, stage2Entry, c); - if(value<=0xff) { - length=1; - } else if(value<=0xffff) { - length=2; - } else if(value<=0xffffff) { - length=3; - } else { - length=4; - } - break; - case MBCS_OUTPUT_3_EUC: - value=MBCS_VALUE_2_FROM_STAGE_2(sharedData->table->mbcs.fromUnicodeBytes, stage2Entry, c); - /* EUC 16-bit fixed-length representation */ - if(value<=0xff) { - length=1; - } else if((value&0x8000)==0) { - value|=0x8e8000; - length=3; - } else if((value&0x80)==0) { - value|=0x8f0080; - length=3; - } else { - length=2; - } - break; - case MBCS_OUTPUT_4_EUC: - p=MBCS_POINTER_3_FROM_STAGE_2(sharedData->table->mbcs.fromUnicodeBytes, stage2Entry, c); - value=((uint32_t)*p<<16)|((uint32_t)p[1]<<8)|p[2]; - /* EUC 16-bit fixed-length representation applied to the first two bytes */ - if(value<=0xff) { - length=1; - } else if(value<=0xffff) { - length=2; - } else if((value&0x800000)==0) { - value|=0x8e800000; - length=4; - } else if((value&0x8000)==0) { - value|=0x8f008000; - length=4; - } else { - length=3; + /* + * Make sure that the last byte sequence before sourceLimit is complete + * or runs into a lead byte. + * Do not go back into the bytes that will be read for finishing a partial + * sequence from the previous buffer. + * In the conversion loop compare source with sourceLimit only once + * per multi-byte character. + */ + { + int32_t i, length; + + length=(int32_t)(sourceLimit-source) - (toULimit-oldToULength); + for(i=0; i<3 && i0) { + utf8->toUnicodeStatus=0; + utf8->toULength=0; + goto moreBytes; + /* See note in ucnv_SBCSFromUTF8() about this goto. */ } -} + /* conversion loop */ + while(source0) { + b=*source++; + if((int8_t)b>=0) { + /* convert ASCII */ + if(IS_ASCII_ROUNDTRIP(b, asciiRoundtrips)) { + *target++=b; + --targetCapacity; + continue; + } else { + value=DBCS_RESULT_FROM_UTF8(mbcsIndex, results, 0, b); + if(value==0) { + c=b; + goto unassigned; + } + } + } else { + if(b>0xe0) { + if( /* handle U+1000..U+D7FF inline */ + (((t1=(uint8_t)(source[0]-0x80), b<0xed) && (t1 <= 0x3f)) || + (b==0xed && (t1 <= 0x1f))) && + (t2=(uint8_t)(source[1]-0x80)) <= 0x3f + ) { + c=((b&0xf)<<6)|t1; + source+=2; + value=DBCS_RESULT_FROM_UTF8(mbcsIndex, results, c, t2); + if(value==0) { + c=(c<<6)|t2; + goto unassigned; + } + } else { + c=-1; + } + } else if(b<0xe0) { + if( /* handle U+0080..U+07FF inline */ + b>=0xc2 && + (t1=(uint8_t)(*source-0x80)) <= 0x3f + ) { + c=b&0x1f; + ++source; + value=DBCS_RESULT_FROM_UTF8(mbcsIndex, results, c, t1); + if(value==0) { + c=(c<<6)|t1; + goto unassigned; + } + } else { + c=-1; + } + } else { + c=-1; + } -#if 0 -/** - * ################################################################ - * # - * # This function has been moved to ucnv2022.c for inlining. - * # This implementation is here only for documentation purposes - * # - * ################################################################ - */ + if(c<0) { + /* handle "complicated" and error cases, and continuing partial characters */ + oldToULength=0; + toULength=1; + toULimit=utf8_countTrailBytes[b]+1; + c=b; +moreBytes: + while(toULengthsourceLimit) { + b=*source; + if(U8_IS_TRAIL(b)) { + ++source; + ++toULength; + c=(c<<6)+b; + } else { + break; /* sequence too short, stop with toULengthtoUBytes[oldToULength++]=*source++; + } + utf8->toUnicodeStatus=c; + utf8->toULength=toULength; + utf8->mode=toULimit; + pToUArgs->source=(char *)source; + pFromUArgs->target=(char *)target; + return; + } + } -/** - * This version of _MBCSFromUChar32() is optimized for single-byte codepages. - * It does not handle the EBCDIC swaplfnl option (set in UConverter). - * - * It returns the codepage byte for the code point, or -1 if it is unassigned. - */ -U_CFUNC int32_t -_MBCSSingleFromUChar32(UConverterSharedData *sharedData, - UChar32 c, - UBool useFallback) { - const uint16_t *table; - int32_t value; + if( toULength==toULimit && /* consumed all trail bytes */ + (toULength==3 || toULength==2) && /* BMP */ + (c-=utf8_offsets[toULength])>=utf8_minLegal[toULength] && + (c<=0xd7ff || 0xe000<=c) /* not a surrogate */ + ) { + stage2Entry=MBCS_STAGE_2_FROM_U(table, c); + } else if( + toULength==toULimit && toULength==4 && + (0x10000<=(c-=utf8_offsets[4]) && c<=0x10ffff) + ) { + /* supplementary code point */ + if(!hasSupplementary) { + /* BMP-only codepages are stored without stage 1 entries for supplementary code points */ + stage2Entry=0; + } else { + stage2Entry=MBCS_STAGE_2_FROM_U(table, c); + } + } else { + /* error handling: illegal UTF-8 byte sequence */ + source-=(toULength-oldToULength); + while(oldToULengthtoUBytes[oldToULength++]=*source++; + } + utf8->toULength=toULength; + pToUArgs->source=(char *)source; + pFromUArgs->target=(char *)target; + *pErrorCode=U_ILLEGAL_CHAR_FOUND; + return; + } - /* BMP-only codepages are stored without stage 1 entries for supplementary code points */ - if(c>=0x10000 && !(sharedData->table->mbcs.unicodeMask&UCNV_HAS_SUPPLEMENTARY)) { - return -1; - } + /* get the bytes and the length for the output */ + /* MBCS_OUTPUT_2 */ + value=MBCS_VALUE_2_FROM_STAGE_2(results, stage2Entry, c); - /* convert the Unicode code point in c into codepage bytes (same as in _MBCSFromUnicodeWithOffsets) */ - table=sharedData->table->mbcs.fromUnicodeTable; + /* is this code point assigned, or do we use fallbacks? */ + if(!(MBCS_FROM_U_IS_ROUNDTRIP(stage2Entry, c) || + (UCNV_FROM_U_USE_FALLBACK(cnv, c) && value!=0)) + ) { + goto unassigned; + } + } + } - /* get the byte for the output */ - value=MBCS_SINGLE_RESULT_FROM_U(table, (uint16_t *)sharedData->table->mbcs.fromUnicodeBytes, c); - /* is this code point assigned, or do we use fallbacks? */ - if(useFallback ? value>=0x800 : value>=0xc00) { - return value&0xff; - } else { - return -1; + /* write the output character bytes from value and length */ + /* from the first if in the loop we know that targetCapacity>0 */ + if(value<=0xff) { + /* this is easy because we know that there is enough space */ + *target++=(uint8_t)value; + --targetCapacity; + } else /* length==2 */ { + *target++=(uint8_t)(value>>8); + if(2<=targetCapacity) { + *target++=(uint8_t)value; + targetCapacity-=2; + } else { + cnv->charErrorBuffer[0]=(char)value; + cnv->charErrorBufferLength=1; + + /* target overflow */ + *pErrorCode=U_BUFFER_OVERFLOW_ERROR; + break; + } + } + continue; + +unassigned: + { + /* + * Try an extension mapping. + * Pass in no source because we don't have UTF-16 input. + * If we have a partial match on c, we will return and revert + * to UTF-8->UTF-16->charset conversion. + */ + static const UChar nul=0; + const UChar *noSource=&nul; + c=_extFromU(cnv, cnv->sharedData, + c, &noSource, noSource, + &target, target+targetCapacity, + NULL, -1, + pFromUArgs->flush, + pErrorCode); + + if(U_FAILURE(*pErrorCode)) { + /* not mappable or buffer overflow */ + cnv->fromUChar32=c; + break; + } else if(cnv->preFromUFirstCP>=0) { + /* + * Partial match, return and revert to pivoting. + * In normal from-UTF-16 conversion, we would just continue + * but then exit the loop because the extension match would + * have consumed the source. + */ + break; + } else { + /* a mapping was written to the target, continue */ + + /* recalculate the targetCapacity after an extension mapping */ + targetCapacity=(int32_t)(pFromUArgs->targetLimit-(char *)target); + continue; + } + } + } else { + /* target is full */ + *pErrorCode=U_BUFFER_OVERFLOW_ERROR; + break; + } + } + + /* + * The sourceLimit may have been adjusted before the conversion loop + * to stop before a truncated sequence. + * If so, then collect the truncated sequence now. + */ + if(U_SUCCESS(*pErrorCode) && source<(sourceLimit=(uint8_t *)pToUArgs->sourceLimit)) { + c=utf8->toUBytes[0]=b=*source++; + toULength=1; + toULimit=utf8_countTrailBytes[b]+1; + while(sourcetoUBytes[toULength++]=b=*source++; + c=(c<<6)+b; + } + utf8->toUnicodeStatus=c; + utf8->toULength=toULength; + utf8->mode=toULimit; } + + /* write back the updated pointers */ + pToUArgs->source=(char *)source; + pFromUArgs->target=(char *)target; } -#endif /* miscellaneous ------------------------------------------------------------ */ static void -_MBCSGetStarters(const UConverter* cnv, +ucnv_MBCSGetStarters(const UConverter* cnv, UBool starters[256], UErrorCode *pErrorCode) { - const int32_t *state0=cnv->sharedData->table->mbcs.stateTable[0]; + const int32_t *state0; int i; + state0=cnv->sharedData->mbcs.stateTable[cnv->sharedData->mbcs.dbcsOnlyState]; for(i=0; i<256; ++i) { /* all bytes that cause a state transition from state 0 are lead bytes */ starters[i]= (UBool)MBCS_ENTRY_IS_TRANSITION(state0[i]); @@ -3604,12 +5493,12 @@ _MBCSGetStarters(const UConverter* cnv, * to check whether a byte is a lead byte. */ U_CFUNC UBool -_MBCSIsLeadByte(UConverterSharedData *sharedData, char byte) { - return (UBool)MBCS_ENTRY_IS_TRANSITION(sharedData->table->mbcs.stateTable[0][(uint8_t)byte]); +ucnv_MBCSIsLeadByte(UConverterSharedData *sharedData, char byte) { + return (UBool)MBCS_ENTRY_IS_TRANSITION(sharedData->mbcs.stateTable[0][(uint8_t)byte]); } static void -_MBCSWriteSub(UConverterFromUnicodeArgs *pArgs, +ucnv_MBCSWriteSub(UConverterFromUnicodeArgs *pArgs, int32_t offsetIndex, UErrorCode *pErrorCode) { UConverter *cnv=pArgs->converter; @@ -3618,18 +5507,24 @@ _MBCSWriteSub(UConverterFromUnicodeArgs *pArgs, int32_t length; /* first, select between subChar and subChar1 */ - if(cnv->subChar1!=0 && cnv->invalidUCharBuffer[0]<=0xff) { + if( cnv->subChar1!=0 && + (cnv->sharedData->mbcs.extIndexes!=NULL ? + cnv->useSubChar1 : + (cnv->invalidUCharBuffer[0]<=0xff)) + ) { /* select subChar1 if it is set (not 0) and the unmappable Unicode code point is up to U+00ff (IBM MBCS behavior) */ subchar=(char *)&cnv->subChar1; length=1; } else { /* select subChar in all other cases */ - subchar=(char *)cnv->subChar; + subchar=(char *)cnv->subChars; length=cnv->subCharLen; } - switch(cnv->sharedData->table->mbcs.outputType) { - case MBCS_OUTPUT_2_SISO: + /* reset the selector for the next code point */ + cnv->useSubChar1=FALSE; + + if (cnv->sharedData->mbcs.outputType == MBCS_OUTPUT_2_SISO) { p=buffer; /* fromUnicodeStatus contains prevLength */ @@ -3643,7 +5538,7 @@ _MBCSWriteSub(UConverterFromUnicodeArgs *pArgs, *p++=subchar[0]; break; case 2: - if(cnv->fromUnicodeStatus==1) { + if(cnv->fromUnicodeStatus<=1) { /* SBCS mode and DBCS sub char: change to DBCS */ cnv->fromUnicodeStatus=2; *p++=UCNV_SO; @@ -3655,24 +5550,19 @@ _MBCSWriteSub(UConverterFromUnicodeArgs *pArgs, *pErrorCode=U_ILLEGAL_ARGUMENT_ERROR; return; } - ucnv_cbFromUWriteBytes(pArgs, - buffer, (int32_t)(p-buffer), - offsetIndex, pErrorCode); - break; - default: - ucnv_cbFromUWriteBytes(pArgs, - subchar, length, - offsetIndex, pErrorCode); - break; + subchar=buffer; + length=(int32_t)(p-buffer); } + + ucnv_cbFromUWriteBytes(pArgs, subchar, length, offsetIndex, pErrorCode); } U_CFUNC UConverterType -_MBCSGetType(const UConverter* converter) { +ucnv_MBCSGetType(const UConverter* converter) { /* SBCS, DBCS, and EBCDIC_STATEFUL are replaced by MBCS, but here we cheat a little */ - if(converter->sharedData->table->mbcs.countStates==1) { + if(converter->sharedData->mbcs.countStates==1) { return (UConverterType)UCNV_SBCS; - } else if((converter->sharedData->table->mbcs.outputType&0xff)==MBCS_OUTPUT_2_SISO) { + } else if((converter->sharedData->mbcs.outputType&0xff)==MBCS_OUTPUT_2_SISO) { return (UConverterType)UCNV_EBCDIC_STATEFUL; } else if(converter->sharedData->staticData->minBytesPerChar==2 && converter->sharedData->staticData->maxBytesPerChar==2) { return (UConverterType)UCNV_DBCS; @@ -3680,132 +5570,90 @@ _MBCSGetType(const UConverter* converter) { return (UConverterType)UCNV_MBCS; } -static const UConverterImpl _MBCSImpl={ +static const UConverterImpl _SBCSUTF8Impl={ UCNV_MBCS, - _MBCSLoad, - _MBCSUnload, + ucnv_MBCSLoad, + ucnv_MBCSUnload, - _MBCSOpen, + ucnv_MBCSOpen, NULL, - _MBCSReset, - - _MBCSToUnicodeWithOffsets, - _MBCSToUnicodeWithOffsets, - _MBCSFromUnicodeWithOffsets, - _MBCSFromUnicodeWithOffsets, - _MBCSGetNextUChar, - - _MBCSGetStarters, - _MBCSGetName, - _MBCSWriteSub, NULL, - _MBCSGetUnicodeSet -}; + ucnv_MBCSToUnicodeWithOffsets, + ucnv_MBCSToUnicodeWithOffsets, + ucnv_MBCSFromUnicodeWithOffsets, + ucnv_MBCSFromUnicodeWithOffsets, + ucnv_MBCSGetNextUChar, -/* Static data is in tools/makeconv/ucnvstat.c for data-based - * converters. Be sure to update it as well. - */ + ucnv_MBCSGetStarters, + ucnv_MBCSGetName, + ucnv_MBCSWriteSub, + NULL, + ucnv_MBCSGetUnicodeSet, -const UConverterSharedData _MBCSData={ - sizeof(UConverterSharedData), 1, - NULL, NULL, NULL, FALSE, &_MBCSImpl, - 0 + NULL, + ucnv_SBCSFromUTF8 }; -/* GB 18030 special handling ------------------------------------------------ */ - -/* definition of LINEAR macros and gb18030Ranges see near the beginning of the file */ - -/* the callback functions handle GB 18030 specially */ -static void -fromUCallback(UConverter *cnv, - const void *context, UConverterFromUnicodeArgs *pArgs, - UChar32 codePoint, - UConverterCallbackReason reason, UErrorCode *pErrorCode) { - int32_t i; - - if((cnv->options&_MBCS_OPTION_GB18030)!=0 && reason==UCNV_UNASSIGNED) { - const uint32_t *range; - - range=gb18030Ranges[0]; - for(i=0; iinvalidUCharBuffer, i, codePoint); - cnv->invalidUCharLength=(int8_t)i; + ucnv_MBCSGetStarters, + ucnv_MBCSGetName, + ucnv_MBCSWriteSub, + NULL, + ucnv_MBCSGetUnicodeSet, - /* call the normal callback function */ - cnv->fromUCharErrorBehaviour(context, pArgs, cnv->invalidUCharBuffer, i, codePoint, reason, pErrorCode); -} + NULL, + ucnv_DBCSFromUTF8 +}; -static void -toUCallback(UConverter *cnv, - const void *context, UConverterToUnicodeArgs *pArgs, - const char *codeUnits, int32_t length, - UConverterCallbackReason reason, UErrorCode *pErrorCode) { - int32_t i; +static const UConverterImpl _MBCSImpl={ + UCNV_MBCS, - if((cnv->options&_MBCS_OPTION_GB18030)!=0 && reason==UCNV_UNASSIGNED && length==4) { - const uint32_t *range; - uint32_t linear; + ucnv_MBCSLoad, + ucnv_MBCSUnload, - linear=LINEAR_18030((uint8_t)codeUnits[0], (uint8_t)codeUnits[1], (uint8_t)codeUnits[2], (uint8_t)codeUnits[3]); - range=gb18030Ranges[0]; - for(i=0; iinvalidCharBuffer[i]=codeUnits[i]; - } - cnv->invalidCharLength=(int8_t)length; +/* Static data is in tools/makeconv/ucnvstat.c for data-based + * converters. Be sure to update it as well. + */ - /* call the normal callback function */ - cnv->fromCharErrorBehaviour(context, pArgs, codeUnits, length, reason, pErrorCode); -} +const UConverterSharedData _MBCSData={ + sizeof(UConverterSharedData), 1, + NULL, NULL, NULL, FALSE, &_MBCSImpl, + 0 +}; #endif /* #if !UCONFIG_NO_LEGACY_CONVERSION */