--- parser3/src/include/pa_string.h	2003/07/24 11:31:21	1.145
+++ parser3/src/include/pa_string.h	2004/05/24 12:21:18	1.162
@@ -1,14 +1,14 @@
 /** @file
 	Parser: string class decl.
 
-	Copyright (c) 2001-2003 ArtLebedev Group (http://www.artlebedev.com)
+	Copyright (c) 2001-2004 ArtLebedev Group (http://www.artlebedev.com)
 	Author: Alexandr Petrosian <paf@design.ru> (http://paf.design.ru)
 */
 
 #ifndef PA_STRING_H
 #define PA_STRING_H
 
-static const char* IDENT_STRING_H="$Date: 2003/07/24 11:31:21 $";
+static const char * const IDENT_STRING_H="$Date: 2004/05/24 12:21:18 $";
 
 // includes
 
@@ -20,6 +20,12 @@ extern "C" { // cord's author forgot to
 #include "cord.h"
 };
 
+// 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_range_contains_chr_greater_then(CORD x, size_t i, size_t n, int c);
+size_t CORD_block_count(CORD x);
+
 // forwards
 
 class Charset;
@@ -30,184 +36,327 @@ class Request_charsets;
 class String;
 typedef Array<const String*> ArrayString;
 
+// generally useful
+
+int pa_atoi(const char* str, const String* problem_source=0);
+double pa_atod(const char* str, const String* problem_source=0);
+
 /// this is result of pos functions which mean that substr were not found
 #define STRING_NOT_FOUND ((size_t)-1)
 
-class StringBody {
+template<typename T>
+inline size_t get_length(T current) {
+	return current;
+}
 
-	CORD body;
+/** 
+	String which knows the lang of all it's langs.
 
+	All pieces remember 
+	- whether they are tainted or not, 
+	  and the lang which should be used to detaint them
+*/
+class String: public PA_Object {
 public:
 
-	StringBody(): body(CORD_EMPTY) {}
-	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; }
+	/** piece is tainted or not. the lang to use when detaint
+		remember to change String_Untaint_lang_name @ untaint.C along
 
-	bool operator! () const { return is_empty(); }
+		WARNING WARNING WARNING WARNING WARNING WARNING 
+		
+		pos function compares(<=) languages, that is used in searching
+		for table column separator being L_CLEAN or L_AS_IS.
+		they search for AS_IS, meaning AS_IS|CLEAN [doing <=L_AS_IS check].
+		
+		letters assigned for debugging, but it's important for no language-letter
+		come before L_AS_IS other then L_CLEAN
 
-	uint hash_code() const;
+		WARNING WARNING WARNING WARNING WARNING WARNING 
+	*/
+	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
 
-	const char* cstr() const { return CORD_to_const_char_star(body); }
-	char* cstrm() const { return CORD_to_char_star(body); }
+		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 
+		// 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_XML='X',		///< ^dom:set xml
+		L_HTML='H',      ///< HTML code (for editing)
+		// READ WARNING ABOVE BEFORE ADDING ANYTHING
+		L_OPTIMIZE_BIT = 0x80  ///< flag, requiring cstr whitespace optimization
+	};
 
-	size_t length() const { return CORD_len(body); }
+	enum Trim_kind {
+		TRIM_BOTH,
+		TRIM_START,
+		TRIM_END
+	};
 
-	bool is_empty() const { return body==CORD_EMPTY; }
+	union Languages {
 
-	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); }
-	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; }
+		struct {
+#ifdef PA_LITTLE_ENDIAN
+			Language lang:8;
+			int is_not_just_lang:sizeof(CORD)*8-8;
+#elif defined(PA_BIG_ENDIAN)
+			int is_not_just_lang:sizeof(CORD)*8-8;
+			Language lang:8;
+#else
+#	error word endianness not determined for some obscure reason
+#endif
+		} opt;
+		CORD langs;
 
-	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);
-	}
+		template<typename C>
+		CORD make_langs(C current) const {
+			return opt.is_not_just_lang?
+				langs
+				:CORD_chars((char)opt.lang, get_length(current));
+		}
 
-	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 { 
-		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);
-	}
+		CORD make_langs(size_t aoffset, size_t alength)  const {
+			return opt.is_not_just_lang?
+				CORD_substr(langs, aoffset, alength)
+				:CORD_chars((char)opt.lang, alength);
+		}
 
-	template<typename I> void for_each(int (*callback)(const char* s, I), I info) const {
-		CORD_iter5(body, 0, 0, (CORD_batched_iter_fn)callback, info);
-	}
+		/// appending when 'langs' already contain something [simple cases handled elsewhere]
+		template<typename C>
+		void append(C current, 
+			const CORD to_nonempty_target_langs) {
+			assert(langs);
+
+			if(opt.is_not_just_lang)
+				langs=CORD_cat(langs, to_nonempty_target_langs);
+			else { // we were "just lang"
+				size_t current_size=get_length(current);
+				assert(current_size);
+				langs=CORD_cat(
+					CORD_chars((char)opt.lang, current_size),  // first piece [making from just 'lang']
+					to_nonempty_target_langs); // new piece
+			}
+		}
 
-	void set_pos(CORD_pos& pos, size_t index) const { CORD_set_pos(pos, body, index); }
+	public:
 
-	StringBody normalize() const {
-		return StringBody(CORD_balance(body));
-	}
+		const char* v() const;
 
-	void dump() const {
-		CORD_dump(body);
-	}
-};
+		Languages(): langs(0) {}
+		Languages(Language alang) {
+			opt.lang=alang;
+			opt.is_not_just_lang=0;
+		}
 
-/** 
-	String which knows the language of all it's fragments.
+		/// MUST be called exactly prior to modification of current [uses it's length]
+		template<typename C>
+		void append(C current, Language alang, size_t asize) {
+			assert(alang);
+			assert(asize);
+
+			if(!opt.is_not_just_lang)
+				if(opt.lang) {
+					if(opt.lang==alang) // same length? ignoring
+						return;
+				} else {
+					opt.lang=alang; // to uninitialized
+					return;
+				}
 
-	All pieces remember 
-	- whether they are tainted or not, 
-	  and the language which should be used to detaint them
-*/
-class String: public PA_Object {
+			append(current, CORD_chars((char)alang, asize));
+		}
 
-//	friend class StringBody;
+		/// MUST be called exactly prior to modification of current [uses it's length]
+		template<typename C>
+		void append(C current, size_t appending_length, 
+			const Languages src) {
+			assert(appending_length);
+
+			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
+			else
+				append(current, src.make_langs(appending_length));
+		}
+		
+		/// MUST be called exactly prior to modification of current [uses it's length]
+		template<typename C>
+		void append(C current,
+			const Languages src, size_t aoffset, size_t alength) {
+			assert(alength);
+
+			if(!langs) // to uninitialized?
+				if(src.opt.is_not_just_lang)
+					langs=CORD_substr(src.langs, aoffset, alength); // to uninitialized complex
+				else
+					opt.lang=src.opt.lang; // to uninitialized simple
+			else 
+				if(!opt.is_not_just_lang && !src.opt.is_not_just_lang && opt.lang==src.opt.lang) // both simple & of same language?
+					return; // ignoring
+				else
+					append(current, src.make_langs(aoffset, alength));
+		}
 
-public:
+		/// checks if we have lang<=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(opt.is_not_just_lang)
+				return CORD_range_contains_chr_greater_then(langs, aoffset, alength, (unsigned)alang)==0;
+			else
+				return opt.lang<=alang;
+		}
 
-	/** piece is tainted or not. the language 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
-		// these two must go before others, there are checks for >L_AS_IS
-		L_CLEAN, ///< clean
-		L_AS_IS,     ///< leave all characters intact
+		/// @returns count of blocks
+		/// @todo currently there can be adjucent blocks of same language. someday merge them
+		size_t count() const {
+			return opt.is_not_just_lang?
+				CORD_block_count(langs)
+				: opt.lang?
+					1
+					: 0;
+		};
+
+		template<typename C, typename I> 
+		void for_each(C current, 
+			int callback(char, size_t, I), I info) const {
+			
+			if(opt.is_not_just_lang)
+				CORD_block_iter(langs, 0, (CORD_block_iter_fn)callback, info);
+			else
+				callback(opt.lang, get_length(current), info);
+		}
 
-		L_PASS_APPENDED,
-			/**<
-				leave language 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_FILE_SPEC, ///< file specification
-		L_HTTP_HEADER,    ///< text in HTTP response header
-		L_MAIL_HEADER,    ///< text in mail header
-		L_URI,       ///< text in uri
-		L_TABLE,     ///< ^table:set body
-		L_SQL,       ///< ^table:sql body
-		L_JS,        ///< JavaScript code
-		L_XML,		///< ^dom:set xml
-		L_HTML,      ///< HTML code (for editing)
-		L_OPTIMIZE_BIT = 0x8000  ///< flag, requiring cstr whitespace optimization
+		bool invariant(size_t current_length) {
+			if(!langs)
+				return current_length==0;
+			if(opt.is_not_just_lang)
+				return CORD_len(langs)==current_length;
+			return true; // uncheckable, actually
+		}
 	};
 
-	struct Fragment { 
-		Language lang; ///< untaint flag, later untaint language
-		size_t length;  ///< length
-		Fragment(Language alang, size_t asize): lang(alang), length(asize) {
-			assert(alang!=L_UNSPECIFIED);
-			assert(asize!=0);
-			assert(asize!=(size_t)-1);
+	class Body {
+
+		CORD body;
+
+	public:
+
+		const char* v() const;
+
+		Body(): body(CORD_EMPTY) {}
+		Body(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
+		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;
+
+		const char* cstr() const { return CORD_to_const_char_star(body); }
+		char* cstrm() const { return CORD_to_char_star(body); }
 
-	class ArrayFragment: public Array<Fragment> {
-		void append(element_type src) {
-			*static_cast<Array<Fragment> *>(this)+=src;
-		}
-		/// hiding from accidental USE, use append_positions
-		void append(const ArrayFragment& src, int offset, int limit) { 
-			static_cast<Array<Fragment> *>(this)->append(src, offset, limit);
+		size_t length() const { return CORD_len(body); }
+
+		bool is_empty() const { return body==CORD_EMPTY; }
+
+		void append_know_length(const char *str, size_t known_length) {
+			if(known_length)
+				body=CORD_cat_char_star(body, str, known_length);
 		}
-	public:
-		ArrayFragment& operator += (element_type src) {
-			if(size_t lcount=count()) { // not empty?
-				// try to join with last
-				Fragment& last=get_ref(lcount-1);
-				if(last.lang==src.lang) {
-					last.length+=src.length;
-					return *this;
-				}
-			}
-			append(src);
-			return *this;
+		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(const ArrayFragment& src) { append(src, 0, ARRAY_OPTION_LIMIT_ALL); }
-		void append_positions(const ArrayFragment& src, size_t substr_begin, size_t substr_end);
+		void append(char c) { body=CORD_cat_char(body, c); }
+		Body& operator << (const Body src) { body=CORD_cat(body, src.body); return *this; }
+		Body& 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 Body src) const { return CORD_cmp(body, src.body)<0; }
+		bool operator > (const Body src) const { return CORD_cmp(body, src.body)>0; }
+		bool operator <= (const Body src) const { return CORD_cmp(body, src.body)<=0; }
+		bool operator >= (const Body src) const { return CORD_cmp(body, src.body)>=0; }
+		bool operator != (const Body src) const { return CORD_cmp(body, src.body)!=0; }
+		bool operator == (const Body src) const { return CORD_cmp(body, src.body)==0; }
 
-		size_t length() {
-			size_t result=0;
-			for(Array_iterator<element_type> i(*this); i.has_next(); ) {
-				const Fragment fragment=i.next();
-				result+=fragment.length;
-			}
-			return result;
+		int ncmp(size_t x_begin, const Body y, size_t y_begin, size_t size) const {
+			return CORD_ncmp(body, x_begin, y.body, y_begin, size);
 		}
+
+		char fetch(size_t index) const { return CORD_fetch(body, index); }
+		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())
+				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]
+			return CORD_str(body, offset, substr.body); 
+		}
+		size_t pos(char c, 
+			size_t offset=0) const {
+			if(offset>=length()) // CORD_chr does not check that [and ABORT's in that case]
+				return STRING_NOT_FOUND;
+
+			return CORD_chr(body, offset, c);
+		}
+
+		template<typename I> int for_each(
+			int (*f)(char c, I), 
+			I info) const {
+			return CORD_iter(body, (CORD_iter_fn)f, (void*)info);
+		}
+
+		template<typename I> int for_each(
+			int (*f1)(char c, I), 
+			int (*f2)(const char* s, I), 
+			I info) const {
+			return CORD_iter5(body, 0, (CORD_iter_fn)f1, (CORD_batched_iter_fn)f2, info);
+		}
+
+		void set_pos(CORD_pos& pos, size_t index) const { CORD_set_pos(pos, body, index); }
+
+		/*Body normalize() const {
+			return Body(CORD_balance(body));
+		}*/
+
+		/// @returns this or 0 or mid. if returns this or 0 out_* are not filled
+		Body trim(Trim_kind kind=TRIM_BOTH, const char* chars=0,
+			size_t* out_start=0, size_t* out_length=0) const;
 	};
 
 	struct C {
 		const char *str;
 		size_t length;
 		operator const char *() { return str; }
+		C(): str(0), length(0) {}
 		C(const char *astr, size_t asize): str(astr), length(asize) {}
 	};
 
@@ -215,34 +364,41 @@ public:
 		char *str;
 		size_t length;
 		//operator char *() { return str; }
+		Cm(): str(0), length(0) {}
 		Cm(char *astr, size_t asize): str(astr), length(asize) {}
 	};
 
 private:
 
-	StringBody body; ///< all characters of string
-	ArrayFragment fragments; ///< fragment language+length info
+	Body body; ///< all characters of string
+	Languages langs; ///< string characters lang
+
+	const char* v() 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);
-	String(const String& src);
-	String(StringBody abody, Language alang): body(abody) {
-		fragments+=Fragment(alang, abody.length());
+	String(Body abody, Language alang): body(abody), langs(alang) {
+		ASSERT_STRING_INVARIANT(*this);
+	}
+	String(const String& src): body(src.body), langs(src.langs) {
+		ASSERT_STRING_INVARIANT(*this);
 	}
-
-#define ASSERT_STRING_INVARIANT(string) \
-	assert((string).body.length()==(string).fragments.length())
 
 	/// for convinient hash lookup
-	operator const StringBody() const { return body; }
+	operator const Body() const { return body; }
 
 	bool is_empty() const { return body.is_empty(); }
 	size_t length() const { return body.length(); }
 
 	/// convert to CORD. if 'lang' known, forcing 'lang' to it
-	StringBody cstr_to_string_body(Language lang=L_AS_IS, 
+	Body cstr_to_string_body(Language lang=L_AS_IS, 
 		SQL_Connection* connection=0,
 		const Request_charsets *charsets=0) const;
 
@@ -262,22 +418,22 @@ public:
 	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);
-	/// @see StringBody::append_know_length
+	/// @see Body::append_know_length
 	String& append_know_length(const char* str, size_t known_length, Language lang);
-	/// @see StringBody::append_help_length
+	/// @see Body::append_help_length
 	String& append_help_length(const char* str, size_t helper_length, Language lang);
 	String& append_strdup(const char* str, size_t helper_length, Language lang);
 
-	bool operator == (const char* y) const { return body==StringBody(y); }
-	bool operator != (const char* y) const { return body!=StringBody(y); }
+	bool operator == (const char* y) const { return body==Body(y); }
+	bool operator != (const char* y) const { return body!=Body(y); }
 
 	/// this starts with y
 	bool starts_with(const char* y) const {
-		return body.ncmp(0/*x_begin*/, StringBody(y), 0/*y_begin*/, strlen(y))==0;
+		return body.ncmp(0/*x_begin*/, Body(y), 0/*y_begin*/, strlen(y))==0;
 	}
 	/// x starts with this
 	bool this_starts(const char* x) const {
-		return StringBody(x).ncmp(0/*x_begin*/, body, 0/*y_begin*/, length())==0;
+		return Body(x).ncmp(0/*x_begin*/, body, 0/*y_begin*/, length())==0;
 	}
 
 	String& append_to(String& dest, Language lang, bool forced) const;
@@ -286,11 +442,7 @@ public:
 	}
 	String& operator << (const String& src) { return append(src, L_PASS_APPENDED); }
 	String& operator << (const char* src) { return append_help_length(src, 0, L_AS_IS); }
-	String& operator << (const StringBody src) { 
-		body<<src;
-		fragments+=Fragment(L_AS_IS, src.length());
-		return *this;
-	}
+	String& operator << (const Body src);
 
 	/// extracts first char of a string, if any
 	char first_char() const {
@@ -312,7 +464,7 @@ public:
 		but when specified: look for substring that lies in ONE fragment in THAT lang
 		@return position of substr in string, -1 means "not found" [const char* version]
 	*/
-	size_t pos(const StringBody substr, 
+	size_t pos(const Body substr, 
 		size_t this_offset=0, Language lang=L_UNSPECIFIED) const;
 	/// String version of @see pos(const char*, int, Language)
 	size_t pos(const String& substr, 
@@ -354,8 +506,9 @@ public:
 	String& change_case(Charset& source_charset,
 		Change_case_kind kind) const;
 	const String& replace(const Dictionary& dict) const;
-	double as_double() const;
-	int as_int() const;
+	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); }
 
 private: //disabled
 
@@ -363,9 +516,23 @@ private: //disabled
 
 };
 
+template<>
+inline size_t get_length<String::Body>(String::Body body) {
+	return body.length();
+}
+
 /// simple hash code of string. used by Hash
-inline uint hash_code(const StringBody self) {
+inline uint hash_code(const String::Body self) {
 	return self.hash_code();
 }
 
+
+/// now that we've declared specialization we can use it
+inline String& String::operator << (const String::Body src) { 
+	langs.append(body, L_AS_IS, src.length());
+	body<<src;
+	return *this;
+}
+
+
 #endif