--- parser3/src/main/pa_charset.C 2008/07/15 12:53:10 1.60 +++ parser3/src/main/pa_charset.C 2009/05/13 07:35:41 1.73 @@ -1,11 +1,11 @@ /** @file Parser: Charset connection implementation. - Copyright(c) 2001-2005 ArtLebedev Group (http://www.artlebedev.com) + Copyright(c) 2001-2009 ArtLebedev Group (http://www.artlebedev.com) Author: Alexander Petrosyan(http://paf.design.ru) */ -static const char * const IDENT_CHARSET_C="$Date: 2008/07/15 12:53:10 $"; +static const char * const IDENT_CHARSET_C="$Date: 2009/05/13 07:35:41 $"; #include "pa_charset.h" #include "pa_charsets.h" @@ -15,6 +15,8 @@ static const char * const IDENT_CHARSET_ #endif //#define PA_PATCHED_LIBXML_BACKWARD + +// reduce memory usage by pre-calculation utf-8 string length #define PRECALCULATE_DEST_LENGTH // globals @@ -86,7 +88,6 @@ static void element2case(unsigned char f // methods -extern "C" unsigned char pcre_default_tables[]; // pcre/chartables.c Charset::Charset(Request_charsets* charsets, const String::Body ANAME, const String* afile_spec): FNAME(ANAME), FNAME_CSTR(ANAME.cstrm()) { @@ -100,7 +101,7 @@ Charset::Charset(Request_charsets* chars } else { fisUTF8=true; // grab default onces [for UTF-8 so to be able to make a-z =>A-Z - memcpy(pcre_tables, pcre_default_tables, sizeof(pcre_tables)); + memcpy(pcre_tables, _pcre_default_tables, sizeof(pcre_tables)); } #ifdef XML @@ -269,9 +270,9 @@ static const XMLByte gFirstByteMark[7] = 0x00, 0x00, 0xC0, 0xE0, 0xF0, 0xF8, 0xFC }; -static int transcodeToUTF8(const XMLByte* srcData, size_t& srcLen, - XMLByte *toFill, size_t& toFillLen, - const Charset::Tables& tables) { +static int transcodeToUTF8(const XMLByte* srcData, int& srcLen, + XMLByte *toFill, int& toFillLen, + const Charset::Tables& tables) { const XMLByte* srcPtr=srcData; const XMLByte* srcEnd=srcData+srcLen; XMLByte* outPtr=toFill; @@ -352,9 +353,9 @@ of ocetes consumed. return 0; } /// @todo digital entites only when xml/html output [at output in html/xml mode, in html part of a letter] -static int transcodeFromUTF8(const XMLByte* srcData, size_t& srcLen, - XMLByte* toFill, size_t& toFillLen, - const Charset::Tables& tables) { +static int transcodeFromUTF8(const XMLByte* srcData, int& srcLen, + XMLByte* toFill, int& toFillLen, + const Charset::Tables& tables) { const XMLByte* srcPtr=srcData; const XMLByte* srcEnd=srcData+srcLen; XMLByte* outPtr=toFill; @@ -432,19 +433,21 @@ of ocetes consumed. return 0; } -static bool is_escaped(char c){ +static bool isEscaped(XMLByte c){ return - !(c<=127 - && ( - ((c>='0') && (c<='9')) - || ((c>='A') && (c<='Z')) - || ((c>='a') && (c<='z')) - || strchr("*@-_+./", c)!=0 - )); + !( + (c<=127) + && ( + ((c>='0') && (c<='9')) + || ((c>='A') && (c<='Z')) + || ((c>='a') && (c<='z')) + || strchr("*@-_+./", c)!=0 + ) + ); } -// read one utf8 character, return number of bytes needed for store it -static unsigned int readChar(const XMLByte*& srcPtr, const XMLByte*& srcEnd, XMLByte& firstByte, XMLCh& UTF8Char){ +// read one UTF8 char and return length of this char (in bytes) +static unsigned int readUTF8Char(const XMLByte*& srcPtr, const XMLByte* srcEnd, XMLByte& firstByte, XMLCh& UTF8Char){ if(!srcPtr || !*srcPtr || srcPtr>=srcEnd) return 0; @@ -478,8 +481,19 @@ static unsigned int readChar(const XMLBy return trailingBytes+1; } -// read char, return number of bytes needed for store it as UTF8 -static unsigned int readChar(const XMLByte*& srcPtr, const XMLByte*& srcEnd, XMLByte& firstByte, XMLCh& UTF8Char, const Charset::Tables& tables){ +// skip UTF8 char and return length of this char (in bytes) +static unsigned int skipUTF8Char(const XMLByte*& srcPtr, const XMLByte* srcEnd){ + if(!srcPtr || !*srcPtr || srcPtr>=srcEnd) + return 0; + + unsigned int trailingBytes=gUTFBytes[*srcPtr]+1; + srcPtr+=trailingBytes; + + return trailingBytes; +} + +// read non-UTF8 char, and return number of bytes needed for store this char in UTF8 +static unsigned int readChar(const XMLByte*& srcPtr, const XMLByte* srcEnd, XMLByte& firstByte, XMLCh& UTF8Char, const Charset::Tables& tables){ if(!srcPtr || !*srcPtr || srcPtr>=srcEnd) return 0; @@ -504,7 +518,7 @@ static unsigned int readChar(const XMLBy return 1; } -static int escape(const XMLByte* srcData, size_t& srcLen, +static int escapeUTF8(const XMLByte* srcData, size_t& srcLen, XMLByte* toFill, size_t& toFillLen) { const XMLByte* srcPtr=srcData; const XMLByte* srcEnd=srcData+srcLen; @@ -515,9 +529,9 @@ static int escape(const XMLByte* srcData uint charSize; // loop until we either run out of input data, or room to store - while((outPtr < outEnd) && (charSize=readChar(srcPtr, srcEnd, firstByte, UTF8Char))){ + while((outPtr < outEnd) && (charSize=readUTF8Char(srcPtr, srcEnd, firstByte, UTF8Char))){ if(charSize==1){ - if(is_escaped(firstByte)) // %XX + if(isEscaped(firstByte)) // %XX outPtr+=sprintf((char*)outPtr, "%%%02X", firstByte); else *outPtr++=firstByte; @@ -540,7 +554,6 @@ static int escape(const XMLByte* srcData const XMLByte* srcPtr=srcData; const XMLByte* srcEnd=srcData+srcLen; XMLByte* outPtr=toFill; - //XMLByte* outEnd=toFill+toFillLen; XMLByte firstByte; XMLCh UTF8Char; uint charSize; @@ -548,7 +561,7 @@ static int escape(const XMLByte* srcData while(charSize=readChar(srcPtr, srcEnd, firstByte, UTF8Char, tables)){ if(charSize==1){ if(firstByte){ - if(is_escaped(firstByte)) // %XX + if(isEscaped(firstByte)) // %XX outPtr+=sprintf((char*)outPtr, "%%%02X", firstByte); else *outPtr++=firstByte; @@ -581,16 +594,16 @@ String::C Charset::escape(const String:: XMLCh UTF8Char; if(source_charset.isUTF8()){ - while(uint charSize=readChar(srcPtr, srcEnd, firstByte, UTF8Char)){ + while(uint charSize=readUTF8Char(srcPtr, srcEnd, firstByte, UTF8Char)){ if(charSize==1) - dest_length+=!is_escaped(firstByte)?1:3/*%XX*/; + dest_length+=!isEscaped(firstByte)?1:3/*%XX*/; else dest_length+=6; // '%uXXXX' } } else { while(uint charSize=readChar(srcPtr, srcEnd, firstByte, UTF8Char, source_charset.tables)){ if(charSize==1) - dest_length+=(!firstByte/*replacement char '?'*/ || !is_escaped(firstByte))?1:3/*'%XX'*/; + dest_length+=(!firstByte/*replacement char '?'*/ || !isEscaped(firstByte))?1:3/*'%XX'*/; else dest_length+=6; // '%uXXXX' } @@ -608,7 +621,7 @@ String::C Charset::escape(const String:: int status; if(source_charset.isUTF8()){ - status=::escape((XMLByte *)src.str, src_length, dest_body, dest_length); + status=::escapeUTF8((XMLByte *)src.str, src_length, dest_body, dest_length); } else { status=::escape((XMLByte *)src.str, src_length, dest_body, dest_length, source_charset.tables); } @@ -616,33 +629,48 @@ String::C Charset::escape(const String:: if(status<0) throw Exception(0, 0, - "Charset::escapeString buffer overflow"); + "Charset::escape buffer overflow"); assert(dest_length<=saved_dest_length); dest_body[dest_length]=0; // terminator return String::C((char*)dest_body, dest_length); } +String::Body Charset::escape(const String::Body src, const Charset& source_charset) { + const char *src_ptr=src.cstr(); + size_t src_size=strlen(src_ptr); + + String::C dest=Charset::escape(String::C(src_ptr, src_size), source_charset); + + return String::Body(dest.str); +} + +String& Charset::escape(const String& src, const Charset& source_charset) { + if(src.is_empty()) + return *new String(); + + return *new String(escape((String::Body)src, source_charset), String::L_CLEAN); +} const String::C Charset::transcodeToUTF8(const String::C src) const { - size_t src_length=src.length; + int src_length=src.length; #ifdef PRECALCULATE_DEST_LENGTH - size_t dest_length=0; + int dest_length=0; const XMLByte* srcPtr=(XMLByte*)src.str; const XMLByte* srcEnd=srcPtr+src_length; - XMLByte firstByte; - XMLCh UTF8Char; + XMLByte firstByte; + XMLCh UTF8Char; while(uint charSize=readChar(srcPtr, srcEnd, firstByte, UTF8Char, tables)) dest_length+=charSize; #else - size_t dest_length=src_length*6; // so that surly enough (max utf8 seq len=6) but too memory-hyngry + int dest_length=src_length*6; // so that surly enough (max utf8 seq len=6) but too memory-hyngry #endif //throw Exception(0,0,"%u",dest_length); #ifndef NDEBUG - size_t saved_dest_length=dest_length; + int saved_dest_length=dest_length; #endif XMLByte *dest_body=new(PointerFreeGC) XMLByte[dest_length+1/*for terminator*/]; @@ -800,15 +828,15 @@ static size_t getDecNumLength(XMLCh UTF8 } const String::C Charset::transcodeFromUTF8(const String::C src) const { - size_t src_length=src.length; + int src_length=src.length; #ifdef PRECALCULATE_DEST_LENGTH - size_t dest_length=0; + int dest_length=0; const XMLByte* srcPtr=(XMLByte*)src.str; const XMLByte* srcEnd=srcPtr+src_length; XMLByte firstByte; XMLCh UTF8Char; - while(uint charSize=readChar(srcPtr, srcEnd, firstByte, UTF8Char)){ + while(uint charSize=readUTF8Char(srcPtr, srcEnd, firstByte, UTF8Char)){ if(charSize==1) dest_length++; else @@ -820,13 +848,11 @@ const String::C Charset::transcodeFromUT } #else // so that surly enough, "&#XXX;" has max ratio (huh? 8 bytes needed for '&#XXXXX;') - size_t dest_length=src_length*6; + int dest_length=src_length*6; #endif - //throw Exception(0,0,"%u",dest_length); - #ifndef NDEBUG - size_t saved_dest_length=dest_length; + int saved_dest_length=dest_length; #endif XMLByte *dest_body=new(PointerFreeGC) XMLByte[dest_length+1/*for terminator*/]; @@ -884,16 +910,16 @@ static const Charset::Tables* tables[MAX unsigned char *out, int *outlen, \ const unsigned char *in, int *inlen, void*) { \ return transcodeToUTF8( \ - in, *(size_t*)inlen, \ - out, *(size_t*)outlen, \ + in, *inlen, \ + out, *outlen, \ *tables[i]); \ } \ static int xml256CharEncodingOutputFunc##i( \ unsigned char *out, int *outlen, \ const unsigned char *in, int *inlen, void*) { \ return transcodeFromUTF8( \ - in, *(size_t*)inlen, \ - out, *(size_t*)outlen, \ + in, *inlen, \ + out, *outlen, \ *tables[i]); \ } @@ -904,16 +930,16 @@ static const Charset::Tables* tables[MAX unsigned char *out, int *outlen, \ const unsigned char *in, int *inlen) { \ return transcodeToUTF8( \ - in, *(size_t*)inlen, \ - out, *(size_t*)outlen, \ + in, *inlen, \ + out, *outlen, \ *tables[i]); \ } \ static int xml256CharEncodingOutputFunc##i( \ unsigned char *out, int *outlen, \ const unsigned char *in, int *inlen) { \ return transcodeFromUTF8( \ - in, *(size_t*)inlen, \ - out, *(size_t*)outlen, \ + in, *inlen, \ + out, *outlen, \ *tables[i]); \ } @@ -1009,7 +1035,7 @@ String::C Charset::transcode_cstr(const } const String& Charset::transcode(const xmlChar* s) { String::C cstr=transcode_cstr(s); - return *new String(cstr.str, cstr.length, true); + return *new String(cstr.str, true); } /// @test less memory using -maybe- xmlParserInputBufferCreateMem @@ -1074,14 +1100,14 @@ String::Body Charset::transcode(const St source_transcoder, dest_transcoder); - return String::Body(dest.str, dest.length); + return String::Body(dest.str); } String& Charset::transcode(const String& src, const Charset& source_transcoder, const Charset& dest_transcoder) { - if(!src.length()) - return *new String("", 0, false); + if(src.is_empty()) + return *new String(); return *new String(transcode((String::Body)src, source_transcoder, dest_transcoder), String::L_CLEAN); } @@ -1106,9 +1132,41 @@ static void transcode_pair(const String: *info->source_transcoder, *info->dest_transcoder); } + void Charset::transcode(HashStringString& src, const Charset& source_transcoder, const Charset& dest_transcoder) { Transcode_pair_info info={&source_transcoder, &dest_transcoder}; src.for_each_ref(transcode_pair, &info); } + +size_t getUTF8BytePos(const XMLByte* srcBegin, const XMLByte* srcEnd, size_t charPos){ + const XMLByte* ptr=srcBegin; + while(charPos-- && skipUTF8Char(ptr, srcEnd)); + + return ptr-srcBegin; +} + +size_t getUTF8CharPos(const XMLByte* srcBegin, const XMLByte* srcEnd, size_t bytePos){ + size_t charPos=0; + const XMLByte* ptr=srcBegin; + const XMLByte* ptrEnd=srcBegin+bytePos; + while(skipUTF8Char(ptr, srcEnd)){ + if(ptr>ptrEnd) + return charPos; + charPos++; + } + + // scan till end but position in bytes still too low + throw Exception(0, + 0, + "Error convertion byte pos to char pos"); +} + +size_t lengthUTF8(const XMLByte* srcBegin, const XMLByte* srcEnd){ + size_t size=0; + while(skipUTF8Char(srcBegin, srcEnd)) + size++; + + return size; +}