--- parser3/src/include/pa_string.h 2004/05/24 12:21:18 1.162 +++ parser3/src/include/pa_string.h 2009/05/15 06:57:43 1.185 @@ -1,17 +1,16 @@ /** @file Parser: string class decl. - Copyright (c) 2001-2004 ArtLebedev Group (http://www.artlebedev.com) + Copyright (c) 2001-2009 ArtLebedev Group (http://www.artlebedev.com) Author: Alexandr Petrosian (http://paf.design.ru) */ #ifndef PA_STRING_H #define PA_STRING_H -static const char * const IDENT_STRING_H="$Date: 2004/05/24 12:21:18 $"; +static const char * const IDENT_STRING_H="$Date: 2009/05/15 06:57:43 $"; // includes - #include "pa_types.h" #include "pa_array.h" @@ -20,6 +19,11 @@ extern "C" { // cord's author forgot to #include "cord.h" }; +// defines + +// cache hash code in String::Body for faster hash access +#define HASH_CODE_CACHING + // cord extension /* Returns true if x does contain */ /* char not_c at positions i..i+n. Value i,i+n must be < CORD_len(x). */ @@ -35,6 +39,7 @@ class Dictionary; class Request_charsets; class String; typedef Array ArrayString; +class VRegex; // generally useful @@ -56,6 +61,8 @@ inline size_t get_length(T current) { - whether they are tainted or not, and the lang which should be used to detaint them */ + + class String: public PA_Object { public: @@ -76,25 +83,27 @@ public: enum Language { L_UNSPECIFIED=0, ///< no real string has parts of this lange: it's just convinient to check when string's empty // these two must go before others, there are checks for >L_AS_IS - L_CLEAN='0', ///< clean WARNING: read above warning before changing - L_AS_IS='A', ///< leave all characters intact WARNING: read above warning before changing + L_CLEAN='0', ///< clean WARNING: read above warning before changing + L_AS_IS='A', ///< leave all characters intact WARNING: read above warning before changing L_PASS_APPENDED='P', /**< leave lang built into string being appended. just a flag, that value not stored */ - L_TAINTED='T', ///< tainted, untaint lang as assigned later + L_TAINTED='T', ///< tainted, untaint lang as assigned later // untaint langs. assigned by ^untaint[lang]{...} - L_FILE_SPEC='F', ///< file specification - L_HTTP_HEADER='h', ///< text in HTTP response header - L_MAIL_HEADER='m', ///< text in mail header - L_URI='U', ///< text in uri - L_TABLE='L', ///< ^table:set body - L_SQL='Q', ///< ^table:sql body - L_JS='J', ///< JavaScript code + L_FILE_SPEC='F', ///< file specification + L_HTTP_HEADER='h', ///< text in HTTP response header + L_MAIL_HEADER='m', ///< text in mail header + L_URI='U', ///< text in uri + L_SQL='Q', ///< ^table:sql body + L_JS='J', ///< JavaScript code L_XML='X', ///< ^dom:set xml - L_HTML='H', ///< HTML code (for editing) + L_HTML='H', ///< HTML code + L_REGEX='R', ///< RegEx expression + L_HTTP_COOKIE='C', ///< cookies encoded as %uXXXX for compartibility with js functions encode/decode + L_FILE_POST='f', ///temporary escaping zero-char // READ WARNING ABOVE BEFORE ADDING ANYTHING L_OPTIMIZE_BIT = 0x80 ///< flag, requiring cstr whitespace optimization }; @@ -105,14 +114,15 @@ public: TRIM_END }; + union Languages { struct { #ifdef PA_LITTLE_ENDIAN Language lang:8; - int is_not_just_lang:sizeof(CORD)*8-8; + size_t is_not_just_lang:sizeof(CORD)*8-8; #elif defined(PA_BIG_ENDIAN) - int is_not_just_lang:sizeof(CORD)*8-8; + size_t is_not_just_lang:sizeof(CORD)*8-8; Language lang:8; #else # error word endianness not determined for some obscure reason @@ -153,6 +163,7 @@ public: public: const char* v() const; + void dump() const; Languages(): langs(0) {} Languages(Language alang) { @@ -168,7 +179,7 @@ public: if(!opt.is_not_just_lang) if(opt.lang) { - if(opt.lang==alang) // same length? ignoring + if(opt.lang==alang) // same language? ignoring return; } else { opt.lang=alang; // to uninitialized @@ -178,20 +189,34 @@ public: append(current, CORD_chars((char)alang, asize)); } - /// MUST be called exactly prior to modification of current [uses it's length] template - void append(C current, size_t appending_length, - const Languages src) { - assert(appending_length); + void appendHelper(C current, Language alang, C asize_helper) { + assert(alang); + + if(!opt.is_not_just_lang) + if(opt.lang) { + if(opt.lang==alang) // same language? ignoring + return; + } else { + opt.lang=alang; // to uninitialized + return; + } + + append(current, CORD_chars((char)alang, asize_helper.length())); + } + + template + void appendHelper(C current, C length_helper, const Languages src) { if(!langs) langs=src.langs; // to uninitialized else if(!src.opt.is_not_just_lang) - append(current, src.opt.lang, appending_length); // simplifying when simple source + appendHelper(current, src.opt.lang, length_helper); // simplifying when simple source else - append(current, src.make_langs(appending_length)); + append(current, src.make_langs(length_helper)); } - + + /// MUST be called exactly prior to modification of current [uses it's length] template void append(C current, @@ -218,7 +243,7 @@ public: if(opt.is_not_just_lang) return CORD_range_contains_chr_greater_then(langs, aoffset, alength, (unsigned)alang)==0; else - return opt.lang<=alang; + return (unsigned)opt.lang<=(unsigned)alang; } /// @returns count of blocks @@ -241,7 +266,7 @@ public: callback(opt.lang, get_length(current), info); } - bool invariant(size_t current_length) { + bool invariant(size_t current_length) const { if(!langs) return current_length==0; if(opt.is_not_just_lang) @@ -254,12 +279,25 @@ public: CORD body; +#ifdef HASH_CODE_CACHING + // cached hash code is not reseted on write operations as test shows + // that string body does not change after it is stored as a hash key + mutable uint hash_code; +#endif + public: const char* v() const; + void dump() const; +#ifdef HASH_CODE_CACHING + Body(): body(CORD_EMPTY), hash_code(0) {} + Body(CORD abody, uint ahash_code): body(abody), hash_code(ahash_code) {} + Body(CORD abody): body(abody), hash_code(0) { +#else Body(): body(CORD_EMPTY) {} Body(CORD abody): body(abody) { +#endif assert(!body // no body || *body // ordinary string || body[1]==1 // CONCAT_HDR @@ -267,17 +305,15 @@ public: || body[1]==6 // SUBSTR_HDR ); } - /// WARNING: length is only HELPER length, str in ANY case should be zero-terminated - Body(const char* str, size_t helper_length): body(CORD_EMPTY) { - append_know_length(str, helper_length?helper_length:strlen(str)); - } + static Body Format(int value); void clear() { body=CORD_EMPTY; } bool operator! () const { return is_empty(); } - uint hash_code() const; + CORD get_cord() const { return body; } + uint get_hash_code() const; const char* cstr() const { return CORD_to_const_char_star(body); } char* cstrm() const { return CORD_to_char_star(body); } @@ -314,7 +350,7 @@ public: Body mid(size_t index, size_t length) const { return CORD_substr(body, index, length); } size_t pos(const char* substr, size_t offset=0) const { return CORD_str(body, offset, substr); } size_t pos(const Body substr, size_t offset=0) const { - if(!substr.length()) + if(substr.is_empty()) return STRING_NOT_FOUND; // in this case CORD_str returns 0 [parser users got used to -1] // CORD_str checks for bad offset [CORD_chr does not] @@ -374,16 +410,28 @@ private: Languages langs; ///< string characters lang const char* v() const; - -#define ASSERT_STRING_INVARIANT(string) \ - assert((string).langs.invariant((string).body.length())) + void dump() const; + #define ASSERT_STRING_INVARIANT(string) \ + assert((string).langs.invariant((string).body.length())) public: static const String Empty; - explicit String(const char* cstr=0, size_t helper_length=0, bool tainted=false); - explicit String(const C cstr, bool tainted=false); + explicit String(){}; + explicit String(const char* cstr, Language alang=L_CLEAN){ + if(cstr && *cstr){ + body=cstr; + langs=alang; + } + } + explicit String(const String::C cstr, Language alang=L_CLEAN){ + if(cstr.length){ + body=cstr.str; + langs=alang; + } + } + String(int value, char *format); String(Body abody, Language alang): body(abody), langs(alang) { ASSERT_STRING_INVARIANT(*this); } @@ -392,10 +440,15 @@ public: } /// for convinient hash lookup +#ifdef HASH_CODE_CACHING + operator const Body&() const { return body; } +#else operator const Body() const { return body; } +#endif bool is_empty() const { return body.is_empty(); } size_t length() const { return body.length(); } + size_t length(Charset& charset) const; /// convert to CORD. if 'lang' known, forcing 'lang' to it Body cstr_to_string_body(Language lang=L_AS_IS, @@ -458,6 +511,7 @@ public: /// extracts [start, finish) piece of string String& mid(size_t substr_begin, size_t substr_end) const; + String& mid(Charset& charset, size_t from, size_t to, size_t helper_length=0) const; /** ignore lang if it's L_UNSPECIFIED @@ -473,6 +527,9 @@ public: size_t this_offset=0) const { return body.pos(c, this_offset); } + size_t pos(Charset& charset, + const String& substr, + size_t this_offset=0, Language lang=L_UNSPECIFIED) const; void split(ArrayString& result, size_t& pos_after, @@ -494,11 +551,9 @@ public: prematch/match/postmatch/1/2/3/... @endverbatim */ - Table* match(Charset& source_charset, - const String& regexp, - const String* options, + Table* match(VRegex* vregex, Row_action row_action, void *info, - bool& just_matched) const; + int& matches_count) const; enum Change_case_kind { CC_UPPER, CC_LOWER @@ -509,6 +564,8 @@ public: const String& trim(Trim_kind kind=TRIM_BOTH, const char* chars=0) const; double as_double() const { return pa_atod(cstr(), this); } int as_int() const { return pa_atoi(cstr(), this); } + bool as_bool() const { return as_int()!=0; } + const String& escape(Charset& source_charset) const; private: //disabled @@ -521,10 +578,12 @@ inline size_t get_length(S return body.length(); } +#ifndef HASH_CODE_CACHING /// simple hash code of string. used by Hash inline uint hash_code(const String::Body self) { - return self.hash_code(); + return self.get_hash_code(); } +#endif /// now that we've declared specialization we can use it