--- parser3/src/include/pa_string.h 2003/03/24 13:16:12 1.144.2.28.2.27 +++ parser3/src/include/pa_string.h 2003/09/23 12:14:00 1.145.4.3 @@ -8,7 +8,7 @@ #ifndef PA_STRING_H #define PA_STRING_H -static const char* IDENT_STRING_H="$Date: 2003/03/24 13:16:12 $"; +static const char* IDENT_STRING_H="$Date: 2003/09/23 12:14:00 $"; // includes @@ -16,9 +16,20 @@ static const char* IDENT_STRING_H="$Date #include "pa_array.h" extern "C" { // cord's author forgot to do that +#define CORD_NO_IO #include "cord.h" }; +/// must use simple version, without optimization for short pieces +/// it's much more convinient for iterating. we rely on this here: CORD_block_iter +#define CORD_chars(c, i) CORD_chars_block((c), (i)) +extern "C" CORD CORD_chars_block(char c, size_t i); + +// 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). */ +int CORD_nnchr(CORD x, size_t i, size_t n, int not_c); + // forwards class Charset; @@ -29,17 +40,6 @@ class Request_charsets; class String; typedef Array ArrayString; -// helpers -/* -/// appends clean piece to String @see String::append -#define APPEND_CLEAN(src, length) append(src, length, String::L_CLEAN) -/// appends piece to String as-is @see String::append -#define APPEND_AS_IS(src, length) append(src, length, String::L_AS_IS) -/// appends tainted piece to String @see String::append -#define APPEND_TAINTED(src, length) append(src, length, String::L_TAINTED) -/// handy: appends const char* piece to String @see String::append -#define APPEND_CONST(src) APPEND_AS_IS(src, 0) -*/ /// this is result of pos functions which mean that substr were not found #define STRING_NOT_FOUND ((size_t)-1) @@ -50,10 +50,19 @@ class StringBody { public: StringBody(): body(CORD_EMPTY) {} - StringBody(CORD abody): body(abody) {} - explicit StringBody(const char* str, size_t length): body(CORD_EMPTY) { - append(str, length); + StringBody(CORD abody): body(abody) { + assert(!body // no body + || *body // ordinary string + || body[1]==1 // CONCAT_HDR + || body[1]==4 // FN_HDR + || body[1]==6 // SUBSTR_HDR + ); + } + /// WARNING: length is only HELPER length, str in ANY case should be zero-terminated + StringBody(const char* str, size_t helper_length): body(CORD_EMPTY) { + append_know_length(str, helper_length?helper_length:strlen(str)); } + static StringBody Format(int value); void clear() { body=CORD_EMPTY; } @@ -68,19 +77,25 @@ public: bool is_empty() const { return body==CORD_EMPTY; } - void append(const char *str, size_t length) { - body=CORD_cat_char_star(body, str, length); + void append_know_length(const char *str, size_t known_length) { + if(known_length) + body=CORD_cat_char_star(body, str, known_length); + } + void append_strdup_know_length(const char* str, size_t known_length) { + if(known_length) + append_know_length(pa_strdup(str, known_length), known_length); } void append(char c) { body=CORD_cat_char(body, c); } - void append(StringBody src) { body=CORD_cat(body, src.body); } + StringBody& operator << (const StringBody src) { body=CORD_cat(body, src.body); return *this; } + StringBody& operator << (const char* str) { append_know_length(str, strlen(str)); return *this; } // could not figure out why this operator is needed [should do this chain: string->simple->==] - bool operator < (const StringBody& src) const { return CORD_cmp(body, src.body)<0; } - bool operator > (const StringBody& src) const { return CORD_cmp(body, src.body)>0; } - bool operator <= (const StringBody& src) const { return CORD_cmp(body, src.body)<=0; } - bool operator >= (const StringBody& src) const { return CORD_cmp(body, src.body)>=0; } - bool operator != (const StringBody& src) const { return CORD_cmp(body, src.body)!=0; } - bool operator == (const StringBody& src) const { return CORD_cmp(body, src.body)==0; } + bool operator < (const StringBody src) const { return CORD_cmp(body, src.body)<0; } + bool operator > (const StringBody src) const { return CORD_cmp(body, src.body)>0; } + bool operator <= (const StringBody src) const { return CORD_cmp(body, src.body)<=0; } + bool operator >= (const StringBody src) const { return CORD_cmp(body, src.body)>=0; } + bool operator != (const StringBody src) const { return CORD_cmp(body, src.body)!=0; } + bool operator == (const StringBody src) const { return CORD_cmp(body, src.body)==0; } int ncmp(size_t x_begin, const StringBody y, size_t y_begin, size_t size) const { return CORD_ncmp(body, x_begin, y.body, y_begin, size); @@ -89,45 +104,57 @@ public: char fetch(size_t index) const { return CORD_fetch(body, index); } StringBody 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 StringBody substr, size_t offset=0) const { return CORD_str(body, offset, substr.body); } + size_t pos(const StringBody substr, size_t offset=0) const { + if(!substr.length()) + return STRING_NOT_FOUND; // in this case CORD_str returns 0 [parser users got used to -1] + return CORD_str(body, offset, substr.body); + } + size_t pos(char c, + size_t offset=0) const { + return CORD_chr(body, offset, c); + } - template void for_each(int (*callback)(const char* s, I), I info) const { +/* template void for_each(int (*callback)(const char* s, I), I info) const { CORD_iter5(body, 0, 0, (CORD_batched_iter_fn)callback, info); - } + }*/ void set_pos(CORD_pos& pos, size_t index) const { CORD_set_pos(pos, body, index); } + /*StringBody normalize() const { + return StringBody(CORD_balance(body)); + }*/ + + void dump() const { + CORD_dump(body); + } }; /** - String which knows the language of all it's fragments. + String which knows the lang of all it's langs. All pieces remember - whether they are tainted or not, - and the language which should be used to detaint them + and the lang which should be used to detaint them */ class String: public PA_Object { - -// friend class StringBody; - public: - /** piece is tainted or not. the language to use when detaint + /** piece is tainted or not. the lang to use when detaint remember to change String_Untaint_lang_name @ untaint.C along */ enum Language { - L_UNSPECIFIED=0, ///< zero value handy for hash lookup @see untaint_lang_name2enum + 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, ///< clean L_AS_IS, ///< leave all characters intact L_PASS_APPENDED, /**< - leave language built into string being appended. + leave lang built into string being appended. just a flag, that value not stored */ - L_TAINTED, ///< tainted, untaint language as assigned later - // untaint languages. assigned by ^untaint[lang]{...} + L_TAINTED, ///< tainted, untaint lang as assigned later + // untaint langs. assigned by ^untaint[lang]{...} L_FILE_SPEC, ///< file specification L_HTTP_HEADER, ///< text in HTTP response header L_MAIL_HEADER, ///< text in mail header @@ -137,60 +164,145 @@ public: L_JS, ///< JavaScript code L_XML, ///< ^dom:set xml L_HTML, ///< HTML code (for editing) - L_OPTIMIZE_BIT = 0x8000 ///< flag, requiring cstr whitespace optimization + L_OPTIMIZE_BIT = 0x80 ///< flag, requiring cstr whitespace optimization }; - struct Fragment { - Language lang; ///< untaint flag, later untaint language - size_t length; ///< length - Fragment(Language alang, size_t asize): lang(alang), length(asize) {} - }; + class Languages { + + union { + struct { + Language lang:8; + int is_not_just_lang:16-8; + }; + CORD langs; + }; + + CORD make_langs(const StringBody body) const { + return is_not_just_lang? + langs + :CORD_chars((char)lang, body.length()); + } - class ArrayFragment: public Array { - void append(element_type src) { - *static_cast *>(this)+=src; + CORD make_langs(size_t aoffset, size_t alength) const { + return is_not_just_lang? + CORD_substr(langs, aoffset, alength) + :CORD_chars((char)lang, alength); } - /// hiding from accidental USE, use append_positions - void append(const ArrayFragment& src, int offset, int limit) { - static_cast *>(this)->append(src, offset, limit); + + /// appending when 'langs' already contain something [simple cases hanled elsewhere] + void append(const StringBody current_body, + const CORD to_nonempty_target_langs) { + assert(langs); + + if(is_not_just_lang) + langs=CORD_cat(langs, to_nonempty_target_langs); + else { // we were "just lang" + size_t current_size=current_body.length(); + assert(current_size); + langs=CORD_cat( + CORD_chars((char)lang, current_size), // first piece [making from just 'lang'] + to_nonempty_target_langs); // new piece + } } + public: - ArrayFragment& operator += (element_type src) { - if(count()) { // not empty? - // try to join with last - Fragment& last=get(count()-1); - if(last.lang==src.lang) { - last.length+=src.length; - return *this; + + Languages(): langs(0) {} + Languages(Language alang): lang(alang), is_not_just_lang(0) {} + + /// MUST be called prior to modification of current_body [uses it's original length] + void append(const StringBody current_body, Language alang, size_t asize) { + assert(alang); + assert(asize); + + if(!is_not_just_lang) + if(lang) { + if(lang==alang) // same length? ignoring + return; + } else { + lang=alang; // to uninitialized + return; } - } - append(src); - return *this; + + append(current_body, CORD_chars((char)alang, asize)); + } + + /// MUST be called prior to modification of current_body [uses it's original length] + void append(const StringBody current_body, const StringBody appending_body, + const Languages src) { + assert(appending_body.length()); + + if(!langs) + langs=src.langs; // to uninitialized + else if(!src.is_not_just_lang) + append(current_body, src.lang, appending_body.length()); // simplifying when simple source + else + append(current_body, src.make_langs(appending_body)); + } + + /// MUST be called prior to modification of current_body [uses it's original length] + void append(const StringBody current_body, + const Languages src, size_t aoffset, size_t alength) { + assert(alength); + + if(!langs) // to uninitialized? + if(src.is_not_just_lang) + langs=CORD_substr(src.langs, aoffset, alength); // to uninitialized complex + else + lang=src.lang; // to uninitialized simple + else + if(!is_not_just_lang && !src.is_not_just_lang && lang==src.lang) // both simple & of same language? + return; // ignoring + else + append(current_body, src.make_langs(aoffset, alength)); + } + + /// checks if we have alang all from aoffset to aoffset+alength + bool check_lang(Language alang, size_t aoffset, size_t alength) const { + if(alang==L_UNSPECIFIED) // ignore lang? + return true; + + if(is_not_just_lang) + return CORD_nnchr(langs, aoffset, alength, (unsigned)alang)!=0; + else + return lang==alang; + } + + template void for_each(const StringBody current_body, + int callback(char, size_t, I), I info) const { + + if(is_not_just_lang) + CORD_block_iter(langs, 0, (CORD_block_iter_fn)callback, info); + else + callback(lang, current_body.length(), info); } - void append(const ArrayFragment& src) { append(src, 0, 0); } - void append_positions(const ArrayFragment& src, size_t substr_begin, size_t substr_end); }; struct C { + const char *str; + size_t length; + operator const char *() { return str; } + C(const char *astr, size_t asize): str(astr), length(asize) {} + }; + + struct Cm { char *str; size_t length; - operator char *() { return str; } - C(char *astr, size_t asize): str(astr), length(asize) {} + //operator char *() { return str; } + Cm(char *astr, size_t asize): str(astr), length(asize) {} }; private: + Languages langs; ///< string characters lang StringBody body; ///< all characters of string - ArrayFragment fragments; ///< fragment language+length info public: - /// @todo check all places for length to be only HELPER length, and never be !=strlen(cstr) explicit String(const char* cstr=0, size_t helper_length=0, bool tainted=false); - String(const String& src); - String(StringBody abody, Language alang): body(abody) { - fragments+=Fragment(alang, abody.length()); - } + explicit String(const C cstr, bool tainted=false); + String(StringBody abody, Language alang): body(abody), langs(alang) {} + String(const String& src): body(src.body), langs(src.langs) {} /// for convinient hash lookup operator const StringBody() const { return body; } @@ -216,14 +328,14 @@ public: return cstr_to_string_body(lang, connection, charsets).cstrm(); } /// puts pieces to buf - C serialize(size_t prolog_size) const; + Cm serialize(size_t prolog_size) const; /// appends pieces from buf to self bool deserialize(size_t prolog_size, void *buf, size_t buf_size); - /** append fragment, - @c length can be zero, it will be autocalced - when length