--- parser3/src/main/pa_string.C	2001/04/23 13:38:31	1.78
+++ parser3/src/main/pa_string.C	2001/10/19 12:43:30	1.112
@@ -2,15 +2,15 @@
 	Parser: string class. @see untasize_t.C.
 
 	Copyright (c) 2001 ArtLebedev Group (http://www.artlebedev.com)
-
 	Author: Alexander Petrosyan <paf@design.ru> (http://design.ru/paf)
 
-	$Id: pa_string.C,v 1.78 2001/04/23 13:38:31 paf Exp $
+	$Id: pa_string.C,v 1.112 2001/10/19 12:43:30 parser Exp $
 */
 
 #include "pa_config_includes.h"
 
 #include "pcre.h"
+#include "internal.h"
 
 #include "pa_pool.h"
 #include "pa_string.h"
@@ -20,14 +20,11 @@
 #include "pa_array.h"
 #include "pa_globals.h"
 #include "pa_table.h"
-#include "pa_threads.h"
-
-//#include "pa_sapi.h"
-
-// String
+#include "pa_dictionary.h"
 
 String::String(Pool& apool, const char *src, size_t src_size, bool tainted) :
-	Pooled(apool) {
+	Pooled(apool),
+	forigins_mode(false) {
 	last_chunk=&head;
 	head.count=CR_PREALLOCATED_COUNT;
 	append_here=head.rows;
@@ -42,23 +39,15 @@ String::String(Pool& apool, const char *
 			APPEND_CLEAN(src, src_size, 0, 0);
 }
 
-void String::expand() {
-	size_t new_chunk_count=last_chunk->count+last_chunk->count*CR_GROW_PERCENT/100;
-	last_chunk=static_cast<Chunk *>(
-		malloc(sizeof(size_t)+sizeof(Chunk::Row)*new_chunk_count+sizeof(Chunk *)));
-	last_chunk->count=new_chunk_count;
-	link_row->link=last_chunk;
-	append_here=last_chunk->rows;
-	link_row=&last_chunk->rows[last_chunk->count];
-	link_row->link=0;
-}
-
-String::String(const String& src) :	Pooled(src.pool()) {
+String::String(const String& src) :	
+	Pooled(src.pool()),
+	forigins_mode(false) {
 	head.count=CR_PREALLOCATED_COUNT;
 	
 	size_t src_used_rows=src.fused_rows;
 	if(src_used_rows<=head.count) {
 		// all new rows fit size_to preallocated area
+		last_chunk=&head;
 		size_t curr_chunk_rows=head.count;
 		memcpy(head.rows, src.head.rows, sizeof(Chunk::Row)*src_used_rows);
 		append_here=&head.rows[src_used_rows];
@@ -77,15 +66,15 @@ String::String(const String& src) :	Pool
 		memcpy(head.rows, src.head.rows, sizeof(Chunk::Row)*head.count);
 		// remaining rows size_to new_chunk
 		size_t curr_chunk_rows=src_used_rows-head.count;
-		Chunk *new_chunk=static_cast<Chunk *>(
+		last_chunk=static_cast<Chunk *>(
 			malloc(sizeof(size_t)+sizeof(Chunk::Row)*curr_chunk_rows+sizeof(Chunk *)));
-		new_chunk->count=curr_chunk_rows;
-		head.preallocated_link=new_chunk;
-		append_here=link_row=&new_chunk->rows[new_chunk->count];
+		last_chunk->count=curr_chunk_rows;
+		head.preallocated_link=last_chunk;
+		append_here=link_row=&last_chunk->rows[last_chunk->count];
 
 		Chunk *old_chunk=src.head.preallocated_link; 
-		Chunk::Row *new_rows=new_chunk->rows;
-		size_t rows_left_to_copy=new_chunk->count;
+		Chunk::Row *new_rows=last_chunk->rows;
+		size_t rows_left_to_copy=last_chunk->count;
 		while(true) {
 			size_t old_count=old_chunk->count;
 			Chunk *next_chunk=old_chunk->rows[old_count].link;
@@ -110,6 +99,17 @@ String::String(const String& src) :	Pool
 	fsize=src.fsize;
 }
 
+void String::expand() {
+	size_t new_chunk_count=last_chunk->count+CR_GROW_COUNT;
+	last_chunk=static_cast<Chunk *>(
+		malloc(sizeof(size_t)+sizeof(Chunk::Row)*new_chunk_count+sizeof(Chunk *)));
+	last_chunk->count=new_chunk_count;
+	link_row->link=last_chunk;
+	append_here=last_chunk->rows;
+	link_row=&last_chunk->rows[last_chunk->count];
+	link_row->link=0;
+}
+
 String& String::append(const String& src, Untaint_lang lang, bool forced) {
 	const Chunk *chunk=&src.head; 
 	do {
@@ -151,6 +151,15 @@ String& String::real_append(STRING_APPEN
 	return *this;
 }
 
+char String::first_char() const {
+	if(!fused_rows)
+		throw Exception(0, 0,
+			this,
+			"getting first char of empty string");
+
+	return *head.rows[0].item.ptr;
+}
+
 uint String::hash_code() const {
 	uint result=0;
 
@@ -186,16 +195,12 @@ int String::cmp(int& partial, const Stri
 	Chunk::Row *b_end=src.append_here;
 	size_t a_countdown=a_chunk->count;
 	size_t b_countdown=b_chunk->count;
-	bool a_break=false;
-	bool b_break=false;
 	size_t result;
 	size_t pos=0; 
-	while(true) {
-		a_break=a_row==a_end;
-		b_break=b_row==b_end;
-		if(a_break || b_break)
-			break;
 
+	bool a_break=size()==0;
+	bool b_break=src.size()==0;
+	if(!(a_break || b_break)) while(true) {
 		if(pos+a_row->item.size > this_offset) {
 			if(lang!=UL_UNSPECIFIED && a_row->item.lang!=lang) 
 				return -1; // wrong lang -- bail out
@@ -228,7 +233,10 @@ int String::cmp(int& partial, const Stri
 				pos+=a_row->item.size;
 				a_row++; a_countdown--; a_offset=0;
 			}
-			
+			if(b_break=b_row==b_end) {
+				a_break=a_row==a_end;
+				break;			
+			}
 			if(!b_countdown) {
 				b_chunk=b_row->link;
 				b_row=b_chunk->rows;
@@ -240,6 +248,10 @@ int String::cmp(int& partial, const Stri
 			a_row++; a_countdown--; 
 		}
 
+		if(a_break=a_row==a_end) {
+			b_break=b_row==b_end;
+			break;
+		}
 		if(!a_countdown) {
 			a_chunk=a_row->link;
 			a_row=a_chunk->rows;
@@ -268,14 +280,11 @@ int String::cmp(int& partial, const char
 	size_t b_offset=0;
 	Chunk::Row *a_end=append_here;
 	size_t a_countdown=a_chunk->count;
-	bool a_break=false;
-	bool b_break=false;
 	size_t pos=0;
-	while(true) {
-		a_break=a_row==a_end;
-		if(a_break || b_break)
-			break;
 
+	bool a_break=size()==0;
+	bool b_break=b_size==0;
+	if(!(a_break || b_break)) while(true) {
 		if(pos+a_row->item.size > this_offset) {
 			if(lang!=UL_UNSPECIFIED && a_row->item.lang!=lang) 
 				return -1; // wrong lang -- bail out
@@ -311,6 +320,9 @@ int String::cmp(int& partial, const char
 			a_row++; a_countdown--; 
 		}
 
+		a_break=a_row==a_end;
+		if(a_break || b_break)
+			break;
 		if(!a_countdown) {
 			a_chunk=a_row->link;
 			a_row=a_chunk->rows;
@@ -328,25 +340,29 @@ int String::cmp(int& partial, const char
 
 #ifndef NO_STRING_ORIGIN
 const Origin& String::origin() const { 
-	if(!fused_rows)
-		THROW(0, 0, 
-			0,
-			"String::origin() of empty string called");
+	if(!fused_rows) {
+		static const Origin empty_origin={"empty string"};
+		return empty_origin;
+	}
 	
 	// determining origin by last appended piece
 	// because first one frequently constant. 
 	// ex: ^load[/file] "document_root" + "/file"
-	return append_here[-1].item.origin; 
+	// when last peice is constant, 
+	// ex: parser_root_auto_path{dynamic} / auto.p{const}
+	// using first piece
+	Origin& last_origin=append_here[-1].item.origin;
+	return last_origin.file ? last_origin : head.rows[0].item.origin;
 }
 #endif
 
 String& String::mid(size_t start, size_t finish) const {
+	String& result=*NEW String(pool());
+
 	start=max(0, start);
 	finish=min(size(), finish);
 	if(start==finish)
-		return *empty_string;
-
-	String& result=*NEW String(pool());
+		return result;
 
 	size_t pos=0;
 	const Chunk *chunk=&head; 
@@ -463,7 +479,7 @@ static void regex_options(char *options,
 		int *result;
     } regex_option[]={
 		{'i', 0, PCRE_CASELESS, result}, // a=A
-		{'s', 0, PCRE_DOTALL, result}, // \n\n$
+		{'s', 0, PCRE_DOTALL, result}, // \n\n$ [default]
 		{'x', 0, PCRE_EXTENDED, result}, // whitespace in regex ignored
 		{'m', PCRE_DOTALL, PCRE_MULTILINE, result}, // ^aaa\n$^bbb\n$
 		{'g', 0, true, result+1}, // many rows
@@ -482,41 +498,38 @@ static void regex_options(char *options,
 			}
 }
 
-/**
-	returns true if fills table.
-	table format is defined and fixed[can be used by others]: 
-	@verbatim
-		pre-match/match/post-match/1/2/3/...
-	@endverbatim
-*/
+/// @todo maybe need speedup: some option to remove pre/match/post string generation
 bool String::match(const unsigned char *pcre_tables,
 				   const String *aorigin,
 				   const String& regexp, 
 				   const String *options,
 				   Table **table,
-				   Row_action row_action, void *info) const { 
+				   Row_action row_action, void *info,
+				   bool *was_global) const { 
 
 	if(!regexp.size())
-		THROW(0, 0,
+		throw Exception(0, 0,
 			aorigin,
 			"regexp is empty");
 	const char *pattern=regexp.cstr(UL_AS_IS);
 	const char *errptr;
 	int erroffset;
     int option_bits[2];  regex_options(options?options->cstr():0, option_bits);
+	if(was_global)
+		*was_global=option_bits[1]!=0;
 	pcre *code=pcre_compile(pattern, option_bits[0], 
 		&errptr, &erroffset,
 		pcre_tables);
 
 	if(!code)
-		THROW(0, 0,
+		throw Exception(0, 0,
 			&regexp.mid(erroffset, regexp.size()),
 			"regular expression syntax error - %s", errptr);
 	
 	int info_substrings=pcre_info(code, 0, 0);
 	if(info_substrings<0) {
-		(*pcre_free)(code);
-		THROW(0, 0,
+		pcre_free(code);
+		throw Exception(0, 0,
 			aorigin,
 			"pcre_info error (%d)", 
 				info_substrings);
@@ -549,36 +562,37 @@ bool String::match(const unsigned char *
 			exec_option_bits, ovector, ovecsize);
 		
 		if(exec_substrings==PCRE_ERROR_NOMATCH) {
-			(*pcre_free)(code);
-			(*row_action)(**table, 0/*last time, no row*/, 0, 0, info);
+			pcre_free(code);
+			row_action(**table, 0/*last time, no row*/, 0, 0, info);
 			return option_bits[1]!=0; // global=true+table, not global=false
 		}
 
 		if(exec_substrings<0) {
-			(*pcre_free)(code);
-			THROW(0, 0,
+			pcre_free(code);
+			throw Exception(0, 0,
 				aorigin,
 				"regular expression execute error (%d)", 
 					exec_substrings);
 		}
 
 		Array& row=*NEW Array(pool());
-		row+=&mid(0, ovector[0]); // .pre-match column value
+		row+=&mid(0, ovector[0]); // .prematch column value
 		row+=&mid(ovector[0], ovector[1]); // .match
-		row+=&mid(ovector[1], size()); // .post-match
+		row+=&mid(ovector[1], size()); // .postmatch
 		
 		for(int i=1; i<exec_substrings; i++) {
 			// -1:-1 case handled peacefully by mid() itself
 			row+=&mid(ovector[i*2+0], ovector[i*2+1]); // .i column value
 		}
 		
-		(*row_action)(**table, &row, startoffset, ovector[0], info);
+		row_action(**table, &row, startoffset, ovector[0], info);
 
-		if(!option_bits[1] || !(startoffset=ovector[1])) { // not global | going to hang
-			(*pcre_free)(code);
-			(*row_action)(**table, 0/*last time, no row*/, 0, 0, info);
+		if(!option_bits[1] || startoffset==ovector[1]) { // not global | going to hang
+			pcre_free(code);
+			row_action(**table, 0/*last time, no row*/, 0, 0, info);
 			return true;
 		}
+		startoffset=ovector[1];
 
 /*
 		if(option_bits[0] & PCRE_MULTILINE)
@@ -586,3 +600,243 @@ bool String::match(const unsigned char *
 */
 	}
 }
+
+String& String::change_case(Pool& pool, const unsigned char *tables, 
+							Change_case_kind kind) const {
+	String& result=*new(pool) String(pool);
+
+	const unsigned char *a;
+	const unsigned char *b;
+	switch(kind) {
+	case CC_UPPER:
+		a=tables+lcc_offset;
+		b=tables+fcc_offset;
+		break;
+	case CC_LOWER:
+		a=tables+lcc_offset;
+		b=0;
+		break;
+	default:
+		throw Exception(0, 0, 
+			this, 
+			"unknown change case kind #%d", 
+				static_cast<int>(kind)); // never
+		a=b=0; // calm, compiler
+		break; // never
+	}	
+
+	const Chunk *chunk=&head; 
+	do {
+		const Chunk::Row *row=chunk->rows;
+		for(size_t i=0; i<chunk->count; i++, row++) {
+			if(row==append_here)
+				goto break2;
+
+			char *new_cstr=(char *)pool.malloc(row->item.size);
+			char *dest=new_cstr;
+			const char *src=row->item.ptr; 
+			for(int size=row->item.size; size--; src++) {
+				unsigned char c=a[(unsigned char)*src];
+				if(b)
+					c=b[c];
+
+				*dest++=(char)c;
+			}
+			
+			result.APPEND(new_cstr, row->item.size, 
+				row->item.lang,
+				row->item.origin.file, row->item.origin.line);
+		}
+		chunk=row->link;
+	} while(chunk);
+break2:
+
+	return result;
+}
+
+void String::join_chain(Pool& pool, 
+					   size_t& ai, const Chunk*& achunk, const Chunk::Row*& arow,
+					   Untaint_lang& joined_lang, const char *& joined_ptr, size_t& joined_size) const {
+	joined_lang=arow->item.lang;
+	
+	// calc size
+	joined_size=0;
+	{
+		size_t start_i=ai;
+		const Chunk::Row *start_row=arow;
+		const Chunk *chunk=achunk;
+		do {
+			const Chunk::Row *row=start_row;
+			for(size_t i=start_i; i<chunk->count; i++, row++) {
+				if(row==append_here)
+					goto break21;
+				
+				if(row->item.lang==joined_lang)
+					joined_size+=row->item.size;
+				else
+					break;
+			}
+			if(chunk=row->link) {
+				start_i=0;
+				start_row=chunk->rows;
+			} else
+				break;
+		} while(true);
+break21:;
+	}
+
+	// if one row, return simply itself
+	if(joined_size==arow->item.size) {
+		joined_ptr=arow->item.ptr;
+		ai++; arow++;
+		if(ai==achunk->count)
+			achunk=arow->link;		
+	} else {
+		// join adjacent rows
+		char *ptr=(char *)pool.malloc(joined_size);
+		joined_ptr=ptr;
+		size_t start_i=ai;
+		const Chunk::Row *start_row=arow;
+		const Chunk *chunk=achunk;
+		size_t i;
+		const Chunk::Row *row;
+		do {
+			row=start_row;
+			for(i=start_i; i<chunk->count; i++, row++) {
+				if(row==append_here)
+					goto break22;
+				
+				if(row->item.lang==joined_lang) {
+					memcpy(ptr, row->item.ptr, row->item.size);
+					ptr+=row->item.size;
+				} else
+					break;
+			}
+			if(chunk=row->link) {
+				start_i=0;
+				start_row=chunk->rows;
+			} else
+				break;
+		} while(true);
+break22:;
+		
+		// return joined rows
+		ai=i;
+		arow=row;
+		achunk=chunk;
+	}
+}
+
+String& String::reconstruct(Pool& pool) const {
+	//_asm int 3;
+	String& result=*new(pool) String(pool);
+	const Chunk *chunk=&head; 
+	do {
+		const Chunk::Row *row=chunk->rows;
+		for(size_t i=0; i<chunk->count; ) {
+			if(row==append_here)
+				goto break2;
+
+			Untaint_lang joined_lang;
+			const char *joined_ptr;
+			size_t joined_size;
+			join_chain(pool, i, chunk, row,
+				joined_lang, joined_ptr, joined_size);
+
+			result.APPEND(joined_ptr, joined_size, 
+				joined_lang,
+				row->item.origin.file, row->item.origin.line);
+			if(!chunk)
+				goto break2;
+		}
+	} while(true);
+break2:
+
+	return result;
+};
+
+String& String::replace_in_reconstructed(Pool& pool, Dictionary& dict) const {
+	//_asm int 3;
+	String& result=*new(pool) String(pool);
+	const Chunk *chunk=&head; 
+	do {
+		const Chunk::Row *row=chunk->rows;
+		for(size_t i=0; i<chunk->count; i++, row++) {
+			if(row==append_here)
+				goto break2;
+
+			const char *src=row->item.ptr; 
+			size_t src_size=row->item.size;
+			char *new_cstr=(char *)pool.malloc((size_t)ceil(src_size*dict.max_ratio()));
+			char *dest=new_cstr;
+			while(src_size) {
+				// there is a row where first column starts 'src'
+				if(Table::Item *item=dict.first_that_starts(src, src_size)) {
+					// get a=>b values
+					const String& a=*static_cast<Array *>(item)->get_string(0);
+					const String& b=*static_cast<Array *>(item)->get_string(1);
+					// skip 'a' in 'src' && reduce work size
+					src+=a.size();  src_size-=a.size();
+					// write 'b' to 'dest' && skip 'b' in 'dest'
+					b.store_to(dest);  dest+=b.size();
+				} else {
+					// write a char to b && reduce work size
+					*dest++=*src++;  src_size--;
+				}
+			}
+
+			result.APPEND(new_cstr, dest-new_cstr, 
+				row->item.lang,
+				row->item.origin.file, row->item.origin.line);
+		}
+		chunk=row->link;
+	} while(chunk);
+
+break2:
+	return result;
+}
+
+String& String::replace(Pool& pool, Dictionary& dict) const {
+	return reconstruct(pool).replace_in_reconstructed(pool, dict);
+}
+
+double String::as_double() const { 
+	double result;
+	const char *cstr=this->cstr();
+	char *error_pos;
+	// 0xABC
+	if(cstr[0]=='0')
+		if(cstr[1]=='x' || cstr[1]=='X')
+			result=(double)(unsigned long)strtol(cstr, &error_pos, 0);
+		else
+			result=(double)strtod(cstr+1/*skip leading 0*/, &error_pos);
+	else
+		result=(double)strtod(cstr, &error_pos);
+
+	if(*error_pos/*not EOS*/)
+		throw Exception(0, 0,
+			this,
+			"invalid number (double)");
+
+	return result;
+}
+int String::as_int() const { 
+	int result;
+	const char *cstr=this->cstr();
+	char *error_pos;
+	// 0xABC
+	if(cstr[0]=='0')
+		if(cstr[1]=='x' || cstr[1]=='X')
+			result=(int)(unsigned long)strtol(cstr, &error_pos, 0);
+		else
+			result=(int)strtol(cstr+1/*skip leading 0*/, &error_pos, 0);
+	else
+		result=(int)strtol(cstr, &error_pos, 0);
+
+	if(*error_pos/*not EOS*/)
+		throw Exception(0, 0,
+			this,
+			"invalid number (int)");
+
+	return result;
+}