--- parser3/src/main/pa_string.C	2001/04/10 11:24:00	1.76
+++ parser3/src/main/pa_string.C	2002/04/10 08:53:55	1.151
@@ -1,15 +1,12 @@
 /** @file
 	Parser: string class. @see untasize_t.C.
 
-	Copyright (c) 2001 ArtLebedev Group (http://www.artlebedev.com)
+	Copyright (c) 2001, 2002 ArtLebedev Group (http://www.artlebedev.com)
+	Author: Alexandr Petrosian <paf@design.ru> (http://paf.design.ru)
 
-	Author: Alexander Petrosyan <paf@design.ru> (http://design.ru/paf)
-
-	$Id: pa_string.C,v 1.76 2001/04/10 11:24:00 paf Exp $
+	$Id: pa_string.C,v 1.151 2002/04/10 08:53:55 paf Exp $
 */
 
-#include "pa_config_includes.h"
-
 #include "pcre.h"
 
 #include "pa_pool.h"
@@ -20,20 +17,21 @@
 #include "pa_array.h"
 #include "pa_globals.h"
 #include "pa_table.h"
-#include "pa_threads.h"
+#include "pa_dictionary.h"
+#include "pa_charset.h"
 
-//#include "pa_sapi.h"
+#define DEBUG_STRING_APPENDS_VS_EXPANDS
 
-// String
+
+#ifdef DEBUG_STRING_APPENDS_VS_EXPANDS
+ulong string_piece_appends=0;
+#endif
 
 String::String(Pool& apool, const char *src, size_t src_size, bool tainted) :
 	Pooled(apool) {
-	last_chunk=&head;
-	head.count=CR_PREALLOCATED_COUNT;
-	append_here=head.rows;
-	head.preallocated_link=0;
-	link_row=&head.rows[head.count];
-	fused_rows=fsize=0;
+	last_chunk=&head.chunk;
+	head.chunk.count=CR_PREALLOCATED_COUNT;
+	append_here=head.chunk.rows;
 
 	if(src)
 		if(tainted)
@@ -42,93 +40,56 @@ 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()) {
+	last_chunk=&head.chunk;
+	head.chunk.count=CR_PREALLOCATED_COUNT;
+	append_here=head.chunk.rows;
+
+	append(src, UL_UNSPECIFIED);
 }
 
-String::String(const String& src) :	Pooled(src.pool()) {
-	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
-		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];
-		link_row=&head.rows[curr_chunk_rows];
-	} else {
-		// warning: 
-		//   heavily relies on the fact 
-		//   "preallocated area is the same for all strings"
-		//
-		// info:
-		//   allocating only enough mem to fit src string rows
-		//   next append would allocate a new chunk
-		//
-		// new rows don't fit size_to preallocated area: splitting size_to two chunks
-		// preallocated chunk src to constructing head
-		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 *>(
-			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];
-
-		Chunk *old_chunk=src.head.preallocated_link; 
-		Chunk::Row *new_rows=new_chunk->rows;
-		size_t rows_left_to_copy=new_chunk->count;
-		while(true) {
-			size_t old_count=old_chunk->count;
-			Chunk *next_chunk=old_chunk->rows[old_count].link;
-			if(next_chunk) {
-				// not last source chunk
-				// taking it all
-				memcpy(new_rows, old_chunk->rows, sizeof(Chunk::Row)*old_count);
-				new_rows+=old_count;
-				rows_left_to_copy-=old_count;
-
-				old_chunk=next_chunk;
-			} else {
-				// the last source chunk
-				// taking only those rows of chunk that _left_to_copy
-				memcpy(new_rows, old_chunk->rows, sizeof(Chunk::Row)*rows_left_to_copy);
-				break;
-			}
-		}
-	}
-	link_row->link=0;
-	fused_rows=src_used_rows;
-	fsize=src.fsize;
+size_t  String::size() const {
+	size_t result=0;
+	STRING_FOREACH_ROW(
+			result+=row->item.size;
+	);
+	return result;
 }
 
-String& String::append(const String& src, Untaint_lang lang, bool forced) {
-	const Chunk *chunk=&src.head; 
-	do {
-		const Chunk::Row *row=chunk->rows;
-		for(size_t i=0; i<chunk->count; i++, row++) {
-			if(row==src.append_here)
-				goto break2;
-			
-			APPEND(row->item.ptr, row->item.size, 
-				(lang!=UL_PASS_APPENDED && (row->item.lang==UL_TAINTED || forced))?lang:row->item.lang,
-				row->item.origin.file, row->item.origin.line);
-		}
-		chunk=row->link;
-	} while(chunk);
-break2:
-	return *this;
+/// @todo not very optimal
+uint String::used_rows() const {
+	uint result=0;
+	STRING_FOREACH_ROW(
+		result++;
+	);
+	return result;
+}
+void String::expand() {
+	uint new_chunk_count=last_chunk->count+CR_GROW_COUNT;
+	if(new_chunk_count>max_integral(Chunk::count_type))
+		new_chunk_count=max_integral(Chunk::count_type);
+
+	Chunk *new_chunk=static_cast<Chunk *>(
+		malloc(
+			sizeof(Chunk)// count+interpadding(?)+rows[CR_PREALLOCATED_COUNT]+tailpadding(??)
+			-sizeof(Chunk::rows_type) // PREALLOCATED rows
+			+sizeof(Chunk::Row)*new_chunk_count // neaded rows
+			+sizeof(Chunk *) // link size
+		, 10));
+	new_chunk->rows[new_chunk->count=new_chunk_count].link=0;
+	last_chunk->rows[last_chunk->count].link=new_chunk;
+	
+	last_chunk=new_chunk;
+	append_here=last_chunk->rows;
 }
 
 String& String::real_append(STRING_APPEND_PARAMS) {
+	if(!last_chunk) // growth stopped [we're appended as string to somebody]
+		throw Exception(0,
+			this,
+			"string growth stopped (append cstr)");
+
 	if(!src)
 		return *this;
 	if(!size)
@@ -136,37 +97,57 @@ String& String::real_append(STRING_APPEN
 	if(!size)
 		return *this;
 
+#ifdef DEBUG_STRING_APPENDS_VS_EXPANDS
+	string_piece_appends++;
+#endif
+
+	// manually unrolled to avoid extra check
+	while(size>max_integral(Chunk::Row::item_size_type)) {
+		if(chunk_is_full())
+			expand();
+
+		append_here->item.ptr=src;
+		append_here->item.size=max_integral(Chunk::Row::item_size_type);
+		append_here->item.lang=lang;
+#ifndef NO_STRING_ORIGIN
+		append_here->item.origin.file=file;
+		append_here->item.origin.line=line;
+#endif
+		append_here++;
+
+		src+=max_integral(Chunk::Row::item_size_type);
+		size-=max_integral(Chunk::Row::item_size_type);
+	}
+
 	if(chunk_is_full())
 		expand();
 
 	append_here->item.ptr=src;
-	fsize+=append_here->item.size=size;
+	append_here->item.size=size;
 	append_here->item.lang=lang;
 #ifndef NO_STRING_ORIGIN
 	append_here->item.origin.file=file;
 	append_here->item.origin.line=line;
 #endif
-	append_here++; fused_rows++;
+	append_here++;
 
 	return *this;
 }
 
-uint String::hash_code() const {
-	uint result=0;
+char String::first_char() const {
+	if(is_empty())
+		throw Exception(0,
+			this,
+			"getting first char of empty string");
 
-	const Chunk *chunk=&head; 
-	do {
-		const Chunk::Row *row=chunk->rows;
-		for(size_t i=0; i<chunk->count; i++) {
-			if(row==append_here)
-				goto break2;
+	return *head.chunk.rows[0].item.ptr;
+}
 
+uint String::hash_code() const {
+	uint result=0;
+	STRING_FOREACH_ROW(
 			result=Hash::generic_code(result, row->item.ptr, row->item.size);
-			row++;
-		}
-		chunk=row->link;
-	} while(chunk);
-break2:
+	);
 	return result;
 }
 
@@ -174,30 +155,27 @@ break2:
 int String::cmp(int& partial, const String& src, 
 				size_t this_offset, Untaint_lang lang) const {
 	partial=-1;
-	this_offset=min(this_offset, size()-1);
+	size_t a_size=size();
+	this_offset=min(this_offset, a_size-1);
 
-	const Chunk *a_chunk=&head;
-	const Chunk *b_chunk=&src.head;
+	const Chunk *a_chunk=&head.chunk;
+	const Chunk *b_chunk=&src.head.chunk;
 	const Chunk::Row *a_row=a_chunk->rows;
 	const Chunk::Row *b_row=b_chunk->rows;
 	size_t a_offset=this_offset;
 	size_t b_offset=0;
 	Chunk::Row *a_end=append_here;
 	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;
+	uint a_countdown=a_chunk->count;
+	uint b_countdown=b_chunk->count;
+	int 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=a_size==0;
+	bool b_break=src.is_empty();
+	if(!(a_break || b_break)) while(true) {
 		if(pos+a_row->item.size > this_offset) {
-			if(lang!=UL_UNSPECIFIED && a_row->item.lang!=lang) 
+			if(lang!=UL_UNSPECIFIED && a_row->item.lang>lang) 
 				return -1; // wrong lang -- bail out
 
 			int size_diff=
@@ -228,7 +206,10 @@ 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 +221,10 @@ 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;
@@ -259,25 +244,23 @@ int String::cmp(int& partial, const Stri
 int String::cmp(int& partial, const char* b_ptr, size_t src_size, 
 				size_t this_offset, Untaint_lang lang) const {
 	partial=-1;
+	size_t a_size=size();
 	size_t b_size=src_size?src_size:b_ptr?strlen(b_ptr):0;
-	this_offset=min(this_offset, size()-1);
+	this_offset=min(this_offset, a_size-1);
 
-	const Chunk *a_chunk=&head;
+	const Chunk *a_chunk=&head.chunk;
 	const Chunk::Row *a_row=a_chunk->rows;
 	size_t a_offset=this_offset;
 	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;
+	uint a_countdown=a_chunk->count;
 	size_t pos=0;
-	while(true) {
-		a_break=a_row==a_end;
-		if(a_break || b_break)
-			break;
 
+	bool a_break=a_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) 
+			if(lang!=UL_UNSPECIFIED && a_row->item.lang>lang) 
 				return -1; // wrong lang -- bail out
 
 			int size_diff=
@@ -285,20 +268,20 @@ int String::cmp(int& partial, const char
 				(b_size-b_offset);
 			
 			if(size_diff==0) { // a has same size as b
-				if(size_t result=memcmp(a_row->item.ptr+a_offset, b_ptr+b_offset, 
+				if(int result=memcmp(a_row->item.ptr+a_offset, b_ptr+b_offset, 
 					a_row->item.size-a_offset)!=0)
 					return result;
 				pos+=a_row->item.size;
 				a_row++; a_countdown--; a_offset=0;
 				b_break=true;
 			} else if (size_diff>0) { // a longer
-				if(size_t result=memcmp(a_row->item.ptr+a_offset, b_ptr+b_offset, 
+				if(int result=memcmp(a_row->item.ptr+a_offset, b_ptr+b_offset, 
 					b_size-b_offset)!=0)
 					return result;
 				a_offset+=b_size-b_offset;
 				b_break=true;
 			} else { // b longer
-				if(size_t result=memcmp(a_row->item.ptr+a_offset, b_ptr+b_offset, 
+				if(int result=memcmp(a_row->item.ptr+a_offset, b_ptr+b_offset, 
 					a_row->item.size-a_offset)!=0)
 					return result;
 				b_offset+=a_row->item.size-a_offset;
@@ -311,6 +294,9 @@ 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;
@@ -328,50 +314,47 @@ 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(is_empty()) {
+		static const Origin empty_origin={"empty string"};
+		return empty_origin;
+	}
 	
-	// determining origin by last appended piece
-	// because first one frequently constant. 
+	// determining origin by first piece or last appended piece
+	// because any of them can be constant=without origin: 
 	// 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& first_origin=head.chunk.rows[0].item.origin;
+	return first_origin.file ? first_origin : append_here[-1].item.origin;
 }
 #endif
 
 String& String::mid(size_t start, size_t finish) const {
-	start=max(0, start);
+	String& result=*NEW String(pool());
+
+	start=max(size_t(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; 
-	do {
-		const Chunk::Row *row=chunk->rows;
-		for(size_t i=0; i<chunk->count; pos+=row->item.size, i++, row++) {
-			if(row==append_here)
+	STRING_FOREACH_ROW(
+		size_t item_finish=pos+row->item.size;
+		if(item_finish > start) { // started now or already?
+			bool started=result.is_empty(); // started now?
+			bool finished=finish <= item_finish; // finished now?
+			size_t offset=started?start-pos:0;
+			size_t size=finished?finish-pos:row->item.size;
+			result.APPEND(
+				row->item.ptr+offset, size-offset, 
+				row->item.lang,
+				row->item.origin.file, row->item.origin.line);
+			if(finished)
 				goto break2;
-
-			size_t item_finish=pos+row->item.size;
-			if(item_finish > start) { // started now or already?
-				bool started=result.size()==0; // started now?
-				bool finished=finish <= item_finish; // finished now?
-				size_t offset=started?start-pos:0;
-				size_t size=finished?finish-pos:row->item.size;
-				result.APPEND(
-					row->item.ptr+offset, size-offset, 
-					row->item.lang,
-					row->item.origin.file, row->item.origin.line);
-				if(finished)
-					goto break2;
-			}
 		}
-		chunk=row->link;
-	} while(chunk);
+		pos+=row->item.size;
+	);
 break2:
 //	SAPI::log(pool(), "piece of '%s' from %d to %d is '%s'",
 		//cstr(), start, finish, result.cstr());
@@ -379,8 +362,9 @@ break2:
 }
 
 int String::pos(const String& substr, 
-				size_t result, Untaint_lang lang) const {
-	for(; result<size(); result++) {
+				int result, Untaint_lang lang) const {
+	size_t self_size=size();
+	for(; size_t(result)<self_size; result++) {
 		int partial; cmp(partial, substr, result, lang);
 		if(
 			partial==0 || // full match
@@ -392,8 +376,9 @@ int String::pos(const String& substr,
 }
 
 int String::pos(const char *substr, size_t substr_size, 
-				size_t result, Untaint_lang lang) const {
-	for(; result<size(); result++) {
+				int result, Untaint_lang lang) const {
+	size_t self_size=size();
+	for(; size_t(result)<self_size; result++) {
 		int partial; cmp(partial, substr, substr_size, result, lang);
 		if(
 			partial==0 || // full match
@@ -408,6 +393,7 @@ void String::split(Array& result,
 				   size_t* pos_after_ref, 
 				   const char *delim, size_t delim_size, 
 				   Untaint_lang lang, int limit) const {
+	size_t self_size=size();
 	if(delim_size) {
 		size_t pos_after=pos_after_ref?*pos_after_ref:0;
 		int pos_before;
@@ -417,16 +403,16 @@ void String::split(Array& result,
 			pos_after=pos_before+delim_size;
 		}
 		// last piece
-		if(pos_after<size() && limit) {
-			result+=&mid(pos_after, size());
-			pos_after=size();
+		if(pos_after<self_size && limit) {
+			result+=&mid(pos_after, self_size);
+			pos_after=self_size;
 		}
 		if(pos_after_ref)
 			*pos_after_ref=pos_after;
 	} else { // empty delim
 		result+=this;
 		if(pos_after_ref)
-			*pos_after_ref+=size();
+			*pos_after_ref+=self_size;
 	}
 }
 
@@ -434,7 +420,7 @@ void String::split(Array& result,
 				   size_t* pos_after_ref, 
 				   const String& delim, Untaint_lang lang, 
 				   int limit) const {
-	if(delim.size()) {
+	if(!delim.is_empty()) {
 		size_t pos_after=pos_after_ref?*pos_after_ref:0;
 		int pos_before;
 		// while we have 'delim'...
@@ -456,7 +442,6 @@ void String::split(Array& result,
 	}
 }
 
-/// @test really @b test: s x m [tested: i & g ]
 static void regex_options(char *options, int *result){
     struct Regex_option {
 		char key;
@@ -464,7 +449,7 @@ 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
@@ -483,51 +468,49 @@ 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
-*/
-bool String::match(const String *aorigin,
+/// @todo maybe need speedup: some option to remove pre/match/post string generation
+bool String::match(
+				   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,
+	if(regexp.is_empty())
+		throw Exception(0,
 			aorigin,
 			"regexp is empty");
-	const char *pattern=regexp.cstr(UL_AS_IS);
+	const char *pattern=regexp.cstr();
 	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);
+		pool().get_source_charset().pcre_tables);
 
 	if(!code)
-		THROW(0, 0,
+		throw Exception(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,
 			aorigin,
 			"pcre_info error (%d)", 
 				info_substrings);
 	}
 
 	int startoffset=0;
-	const char *subject=cstr(UL_AS_IS);
+	const char *subject=cstr();
 	int length=strlen(subject);
 	int ovecsize;
 	int *ovector=(int *)malloc(sizeof(int)*
-		(ovecsize=(1/*match*/+info_substrings)*3));
+		(ovecsize=(1/*match*/+info_substrings)*3), 11);
 
 	{ // create table
 		Array& columns=*NEW Array(pool());
@@ -549,36 +532,37 @@ bool String::match(const String *aorigin
 			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,
 				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 +570,255 @@ bool String::match(const String *aorigin
 */
 	}
 }
+
+String& String::change_case(Pool& pool, 
+							Change_case_kind kind) const {
+	const unsigned char *tables=pool.get_source_charset().pcre_tables;
+	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, 
+			this, 
+			"unknown change case kind #%d", 
+				static_cast<int>(kind)); // never
+		a=b=0; // calm, compiler
+		break; // never
+	}	
+
+	STRING_FOREACH_ROW(
+		char *new_cstr=(char *)pool.malloc(row->item.size, 12);
+		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);
+	);
+
+	return result;
+}
+
+void String::join_chain(Pool& pool, 
+					   const Chunk*& achunk, const Chunk::Row*& arow, uint& acountdown, 
+					   uchar& joined_lang, const char *& joined_ptr, size_t& joined_size) const {
+	joined_lang=arow->item.lang;
+	
+	// calc size
+	joined_size=0;
+	{
+		const Chunk* chunk=achunk;
+		const Chunk::Row* row=arow;
+		uint countdown=acountdown;
+		STRING_PREPARED_FOREACH_ROW(*this, 
+			if(row->item.lang==joined_lang)
+				joined_size+=row->item.size;
+			else
+				break;
+		);
+	}
+
+	// if one row, return simply itself
+	if(joined_size==arow->item.size) {
+		joined_ptr=arow->item.ptr;
+	} else {
+		// join adjacent rows
+		char *ptr=(char *)pool.malloc(joined_size,13);  
+		joined_ptr=ptr;
+
+		const Chunk* chunk=achunk;
+		const Chunk::Row* row=arow;
+		uint countdown=acountdown;
+		STRING_PREPARED_FOREACH_ROW(*this, 
+			if(row->item.lang==joined_lang) {
+				memcpy(ptr, row->item.ptr, row->item.size);  ptr+=row->item.size;
+			} else
+				break; // before non-ours
+		);
+		
+		// set pointers after joined piece
+		achunk=chunk;  arow=row;  acountdown=countdown;
+		// & one step back, see String::reconstruct
+		--arow;  ++acountdown;
+	}
+}
+
+/// @test if in some piece were found no dict words, append it, not it's duplicate
+String& String::replace(Pool& pool, Dictionary& dict) const {
+//	return reconstruct(pool).replace_in_reconstructed(pool, dict);
+	String& result=*new(pool) String(pool);
+
+	STRING_FOREACH_ROW(
+		uchar joined_lang;
+		const char *joined_ptr;
+		size_t joined_size;
+IFNDEF_NO_STRING_ORIGIN(
+		const char *joined_origin_file=row->item.origin.file;
+		const size_t joined_origin_line=row->item.origin.line;
+);
+		join_chain(pool, chunk, row, countdown,
+			joined_lang, joined_ptr, joined_size);
+		
+		const char *src=joined_ptr; 
+		size_t src_size=joined_size;
+		char *new_cstr=(char *)pool.malloc((size_t)ceil(src_size*dict.max_ratio()), 14);
+		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, joined_lang,
+			joined_origin_file, joined_origin_line);
+	);
+	return result;
+}
+
+double String::as_double() const { 
+	double result;
+	const char *cstr;
+	char buf[MAX_NUMBER];
+	if(head.chunk.rows+1==append_here) {
+		int size=min(head.chunk.rows[0].item.size, MAX_NUMBER-1);
+		memcpy(buf, head.chunk.rows[0].item.ptr, size);
+		buf[size]=0;
+		cstr=buf;
+	} else
+		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("number.format",
+			this,
+			"invalid number (double)");
+
+	return result;
+}
+int String::as_int() const { 
+	int result;
+	const char *cstr;
+	char buf[MAX_NUMBER];
+	if(head.chunk.rows+1==append_here) {
+		int size=min(head.chunk.rows[0].item.size, MAX_NUMBER-1);
+		memcpy(buf, head.chunk.rows[0].item.ptr, size);
+		buf[size]=0;
+		cstr=buf;
+	} else
+		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("number.format",
+			this,
+			"invalid number (int)");
+
+	return result;
+}
+
+inline void ushort2uchars(ushort word, uchar& byte1, uchar& byte2) {
+	byte1=word&0xFF;
+	byte2=word>>8;
+}
+inline ushort uchars2ushort(uchar byte1, uchar byte2) {
+	return (byte2<<8) | byte1;
+}
+/* @todo maybe network order worth spending some effort?
+	don't bothering myself with network byte order,
+	am not planning to be able to move resulting file across platforms
+	for now
+*/
+void String::serialize(size_t prolog_size, void *& buf, size_t& buf_size) const {
+	buf_size=
+		prolog_size
+		+used_rows()*(sizeof(uchar)+sizeof(ushort))
+		+size();
+	buf=malloc(buf_size,15);
+	char *cur=(char *)buf+prolog_size;
+
+	STRING_FOREACH_ROW(
+		// lang
+		memcpy(cur, &row->item.lang, sizeof(uchar));
+		cur+=sizeof(uchar);
+		// size
+		uchar byte1; uchar byte2;
+		ushort2uchars(row->item.size, byte1, byte2);
+		memcpy(cur, &byte1, sizeof(uchar)); cur+=sizeof(uchar);
+		memcpy(cur, &byte2, sizeof(uchar)); cur+=sizeof(uchar);
+		// bytes
+		memcpy(cur, row->item.ptr, row->item.size);
+		cur+=row->item.size;
+	);
+}
+bool String::deserialize(size_t prolog_size, void *buf, size_t buf_size, const char *file) {
+	if(buf_size<=prolog_size)
+		return false;
+
+	char *cur=(char *)buf+prolog_size;
+	buf_size-=prolog_size;
+
+	while(buf_size) {
+		if(sizeof(uchar)+sizeof(ushort)>buf_size) // lang+size
+			return false;
+
+		uchar lang=*(uchar *)(cur);		
+		ushort size=uchars2ushort(
+			*(uchar*)(cur+sizeof(uchar)*1),
+			*(uchar*)(cur+sizeof(uchar)*2)
+		);
+
+		size_t piece_size=sizeof(uchar)+sizeof(ushort)+size;
+		if(piece_size>buf_size) // buffer overrun, can be on incomplete cache files
+			return false;
+
+		const char *ptr=(const char*)(cur+sizeof(uchar)*3); 
+		APPEND(ptr, size, lang, file, 0);
+
+		cur+=piece_size;
+		buf_size-=piece_size;
+	}
+	return true;
+}