--- parser3/src/include/pa_string.h 2001/10/29 16:29:07 1.112 +++ parser3/src/include/pa_string.h 2002/02/20 12:40:24 1.129 @@ -1,34 +1,24 @@ /** @file Parser: string class decl. - Copyright (c) 2001 ArtLebedev Group (http://www.artlebedev.com) - Author: Alexander Petrosyan (http://design.ru/paf) + Copyright (c) 2001, 2002 ArtLebedev Group (http://www.artlebedev.com) + Author: Alexandr Petrosian (http://paf.design.ru) - $Id: pa_string.h,v 1.112 2001/10/29 16:29:07 paf Exp $ + $Id: pa_string.h,v 1.129 2002/02/20 12:40:24 paf Exp $ */ #ifndef PA_STRING_H #define PA_STRING_H -#include "pa_config_includes.h" #include "pa_pool.h" #include "pa_types.h" class Table; -/** - $MAIN:html-typo table elements must enlarge string not more that that. - - that's a tradeoff - otherwise we'd have to scan string twice: - - first for buffer length - - second for replacements themselves -*/ -#define UNTAINT_TIMES_BIGGER 10 - #ifndef NO_STRING_ORIGIN # define STRING_APPEND_PARAMS \ const char *src, size_t size, \ - String::Untaint_lang lang, \ + uchar lang, \ const char *file, uint line /// appends piece to String @see String::real_append # define APPEND(src, size, lang, file, line) \ @@ -37,7 +27,7 @@ class Table; # define STRING_APPEND_PARAMS \ const char *src, \ size_t size, \ - String::Untaint_lang lang + uchar lang /// appends piece to String @see String::real_append # define APPEND(src, size, lang, file, line) \ real_append(src, size, lang) @@ -78,12 +68,13 @@ class Dictionary; - whether they are tainted or not, and the language which should be used to detaint them */ +#include "pa_pragma_pack_begin.h" class String : public Pooled { public: enum { CR_PREALLOCATED_COUNT=2, ///< default preallocated item count - CR_GROW_COUNT=2 ///< each time the String chunk_is_full() string expanded() + CR_GROW_COUNT=1 ///< each time the String chunk_is_full() string expanded() }; /** piece is tainted or not. the language to use when detaint @@ -91,15 +82,17 @@ public: */ enum Untaint_lang { UL_UNSPECIFIED=0, ///< zero value handy for hash lookup @see untaint_lang_name2enum + // these two must go before others, there are checks for >UL_AS_IS UL_CLEAN, ///< clean - UL_TAINTED, ///< tainted, untaint language as assigned later - // untaint languages. assigned by ^untaint[lang]{...} + UL_AS_IS, ///< leave all characters intact + UL_PASS_APPENDED, /**< leave language built into string being appended. just a flag, that value not stored */ - UL_AS_IS, ///< leave all characters intact + UL_TAINTED, ///< tainted, untaint language as assigned later + // untaint languages. assigned by ^untaint[lang]{...} UL_FILE_SPEC, ///< file specification UL_HTTP_HEADER, ///< text in HTTP response header UL_MAIL_HEADER, ///< text in mail header @@ -109,26 +102,26 @@ public: UL_JS, ///< JavaScript code UL_XML, ///< ^dom:set xml UL_HTML, ///< HTML code (for editing) - UL_USER_HTML ///< HTML code with USER chars + UL_OPTIMIZE_BIT = 0x80 ///< flag, requiring cstr whitespace optimization }; public: String(Pool& apool, const char *src=0, size_t src_size=0, bool tainted=false); String(const String& src); - size_t size() const { return fsize; } - /// makes store_to produce origin:text\n output - void set_origins_mode(bool aorigins_mode) const { forigins_mode=aorigins_mode; } + bool is_empty() const { return append_here==head.rows; } + size_t size() const; /// convert to C string. if 'lang' known, forcing 'lang' to it char *cstr(Untaint_lang lang=UL_AS_IS, SQL_Connection *connection=0, - const char *charset=0) const { + Charset *cstr_charset=0) const { - char *result=(char *)malloc(cstr_bufsize(lang)); - char *eol=store_to(result, lang, connection, charset); + char *result=(char *)malloc(cstr_bufsize(lang, connection, cstr_charset)); + char *eol=store_to(result, lang, connection, cstr_charset); *eol=0; return result; } + char *cstr_debug_origins() const; /// puts pieces to buf void serialize(size_t prolog_size, void *& buf, size_t& buf_size) const; /// appends pieces from buf to self @@ -173,14 +166,7 @@ public: return cmp(partial, src_ptr, 0)!=0; } - - /** - appends other String. - - marking all tainted pieces of it with @a lang. - or marking ALL pieces of it with a @a lang when @a forced to. - */ - String& append(const String& src, Untaint_lang lang, bool forced=false); + String& append(const String& src, uchar lang, bool forced=false); String& operator << (const String& src) { return append(src, UL_PASS_APPENDED); } String& operator << (const char *src) { return APPEND_CONST(src); } @@ -218,7 +204,7 @@ public: prematch/match/postmatch/1/2/3/... @endverbatim */ - bool match(const unsigned char *pcre_tables, + bool match( const String *aorigin, const String& regexp, const String *options, @@ -229,7 +215,7 @@ public: CC_UPPER, CC_LOWER }; - String& change_case(Pool& pool, const unsigned char *pcre_tables, + String& change_case(Pool& pool, Change_case_kind kind) const; String& replace(Pool& pool, Dictionary& dict) const; double as_double() const; @@ -242,62 +228,61 @@ public: private: - /// several String fragments - struct Chunk { - uint count; ///< the number of rows in chunk + /** several String fragments + + 'mutable' because can write after it's end, after it was appended to somebody + @see String::append + */ + mutable struct Chunk { + typedef uchar count_type; + count_type count; ///< the number of rows in chunk /// string fragment or a link to next chunk union union Row { + typedef uchar item_size_type; /// fragment struct { const char *ptr; ///< pointer to the start - size_t size; ///< length - Untaint_lang lang; ///< untaint flag, later untaint language + item_size_type size; ///< length + uchar/*Untaint_lang*/ lang; ///< untaint flag, later untaint language #ifndef NO_STRING_ORIGIN Origin origin; ///< origin #endif } item; Chunk *link; ///< link to the next chunk in chain } rows[CR_PREALLOCATED_COUNT]; - Chunk *preallocated_link; ///< next rows are here } head; ///< the head chunk of the chunk chain + Chunk *initial_head_link; ///< next rows initially are here[can shrink! @see String::append]. must go right after 'head' /// next append would write to this record Chunk::Row *append_here; - /** the address of place where lies address - of the link to the next chunk to allocate - */ - Chunk::Row *link_row; - private: /// last chunk - Chunk *last_chunk; - - /// string size - size_t fsize; - - /// origins_mode changes behaviour of store_to[cstr()] - mutable bool forigins_mode; + mutable Chunk *last_chunk; private: bool chunk_is_full() { - return append_here == link_row; + return append_here == last_chunk->rows+last_chunk->count; } uint used_rows() const; void expand(); + + Untaint_lang lang_of(size_t offset) const; - size_t cstr_bufsize(Untaint_lang lang) const; + size_t cstr_bufsize(Untaint_lang lang, + SQL_Connection *connection, + Charset *buf_charset) const; /// convert to C string, store to 'dest' which must be big enough for proper untaint char *store_to(char *dest, Untaint_lang lang=UL_UNSPECIFIED, SQL_Connection *connection=0, - const char *charset=0) const; + Charset *store_to_charset=0) const; String& reconstruct(Pool& pool) const; void join_chain(Pool& pool, uint& ai, const Chunk*& achunk, const Chunk::Row*& arow, - Untaint_lang& joined_lang, const char *& joined_ptr, size_t& joined_size) const; + uchar& joined_lang, const char *& joined_ptr, size_t& joined_size) const; String& replace_in_reconstructed(Pool& pool, Dictionary& dict) const; private: //disabled @@ -305,5 +290,23 @@ private: //disabled String& operator = (const String&) { return *this; } }; +#include "pa_pragma_pack_end.h" + +#define STRING_PREFIX_FOREACH_ROW(src, body) { \ + const Chunk *chunk=&(src).head; \ + do { \ + const Chunk::Row *row=chunk->rows; \ + for(uint i=0; icount; i++, row++) { \ + if(row==(src).append_here) \ + goto break2; \ + \ + body \ + } \ + chunk=row->link; \ + } while(chunk); \ +} + +#define STRING_FOREACH_ROW(body) STRING_PREFIX_FOREACH_ROW(*this, body) +#define STRING_SRC_FOREACH_ROW(body) STRING_PREFIX_FOREACH_ROW(src, body) #endif