--- parser3/src/main/pa_string.C	2002/02/22 12:20:27	1.145
+++ parser3/src/main/pa_string.C	2003/03/19 13:16:56	1.172.2.21.2.4
@@ -1,38 +1,87 @@
 /** @file
 	Parser: string class. @see untasize_t.C.
 
-	Copyright (c) 2001, 2002 ArtLebedev Group (http://www.artlebedev.com)
+	Copyright (c) 2001-2003 ArtLebedev Group (http://www.artlebedev.com)
 	Author: Alexandr Petrosian <paf@design.ru> (http://paf.design.ru)
-
-	$Id: pa_string.C,v 1.145 2002/02/22 12:20:27 paf Exp $
 */
 
+static const char* IDENT_STRING_C="$Date: 2003/03/19 13:16:56 $";
+
 #include "pcre.h"
 
-#include "pa_pool.h"
 #include "pa_string.h"
-#include "pa_hash.h"
 #include "pa_exception.h"
-#include "pa_common.h"
-#include "pa_array.h"
-#include "pa_globals.h"
 #include "pa_table.h"
 #include "pa_dictionary.h"
 #include "pa_charset.h"
 
-#define DEBUG_STRING_APPENDS_VS_EXPANDS
+// helpers
+
+/// String::match uses this as replace & global search table columns
+
+const int MAX_MATCH_GROUPS=100;
+
+class String_match_table_template_columns: public ArrayString {
+public:
+	String_match_table_template_columns() {
+		*this+=new String("prematch");
+		*this+=new String("match");
+		*this+=new String("postmatch");
+		for(int i=0; i<MAX_MATCH_GROUPS; i++) {
+			char *cname=new char[3/*strlen("100")*/+1/*terminating 0*/];
+			*this+=new String(cname, sprintf(cname, "%d", 1+i));
+		}
+	}
+};
+
+Table string_match_table_template(new String_match_table_template_columns);
 
+// String::ArrayFragment methods
 
-#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;
+String::ArrayFragment& String::ArrayFragment::append(const ArrayFragment& src, 
+						     size_t substr_start, size_t substr_end) {
+	if(substr_start==substr_end)
+		return *this;
 
+	size_t fragment_start=0;
+	size_t fragment_end;
+	for(Array_iterator<element_type> i(src); ; fragment_start=fragment_end) {
+		const element_type& fragment=i.next();
+		fragment_end=fragment_start+fragment.size;		
+
+		if(substr_start<fragment_start) // not reached fragments which may include 'substr'?
+			continue;
+
+		// found first fragment including piece of 'substr'
+		size_t piece_end=min(substr_end, fragment_end);
+		*this+=Fragment(fragment.lang, piece_end-substr_start);
+		
+		while(substr_end>fragment_end) { // are there more fragments including pieces of 'substr'?
+			const element_type& fragment=i.next();
+			fragment_end=fragment_start+fragment.size;
+
+			if(substr_end>fragment_end) // are there still more?
+				simple_append(Fragment(fragment.lang, fragment.size)); // appending whole fragment
+			else { // no, it was last
+				simple_append(Fragment(fragment.lang, substr_end-fragment_start));
+				goto break2;
+			}
+		}
+
+		break;
+	}	
+break2:
+	
+	return *this;
+}
+/*
+void String::ArrayFragment::mid(ArrayFragment& result, size_t substr_start, size_t substr_end) {
+}
+*/
+
+// String methods
+
+String::String(const char* src, size_t src_size, bool tainted): body(CORD_EMPTY) {
 	if(src)
 		if(tainted)
 			APPEND_TAINTED(src, src_size, 0, 0);
@@ -40,51 +89,9 @@ String::String(Pool& apool, const char *
 			APPEND_CLEAN(src, src_size, 0, 0);
 }
 
-String::String(const String& src) :	
-	Pooled(src.pool()) {
-	last_chunk=&head;
-	head.count=CR_PREALLOCATED_COUNT;
-	append_here=head.rows;
-
-	append(src, UL_UNSPECIFIED);
-}
-
-size_t  String::size() const {
-	size_t result=0;
-	STRING_FOREACH_ROW(
-			result+=row->item.size;
-	);
-	return result;
-}
-
-/// @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_type)+sizeof(Chunk::Row)*new_chunk_count+sizeof(Chunk *), 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(const String& src): body(src.body), fragments(src.fragments) {}
 
 String& String::real_append(STRING_APPEND_PARAMS) {
-	if(!last_chunk) // growth stopped [we're appended as string to somebody]
-		throw Exception(0, 0,
-			this,
-			"string growth stopped (append cstr)");
-
 	if(!src)
 		return *this;
 	if(!size)
@@ -92,308 +99,88 @@ 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;
-	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++;
+	body=CORD_cat_char_star(body, src, size);
+	fragments+=Fragment(lang, size);
 
 	return *this;
 }
 
-char String::first_char() const {
-	if(is_empty())
-		throw Exception(0, 0,
-			this,
-			"getting first char of empty string");
-
-	return *head.rows[0].item.ptr;
-}
-
+static int CORD_batched_iter_fn_generic_hash_code(const char * s, void * client_data) {
+	uint& result=*static_cast<uint*>(client_data);
+	generic_hash_code(result, s);
+	return 0;
+};
 uint String::hash_code() const {
 	uint result=0;
-	STRING_FOREACH_ROW(
-			result=Hash::generic_code(result, row->item.ptr, row->item.size);
-	);
+	CORD_iter5(body, 0, 0, CORD_batched_iter_fn_generic_hash_code, &result);
 	return result;
 }
 
-/// @todo move 'lang' skipping to pos
-int String::cmp(int& partial, const String& src, 
-				size_t this_offset, Untaint_lang lang) const {
-	partial=-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::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;
-	uint a_countdown=a_chunk->count;
-	uint b_countdown=b_chunk->count;
-	int result;
-	size_t pos=0; 
-
-	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) 
-				return -1; // wrong lang -- bail out
-
-			int size_diff=
-				(a_row->item.size-a_offset)-
-				(b_row->item.size-b_offset);
-			
-			if(size_diff==0) { // a has same size as b
-				result=memcmp(a_row->item.ptr+a_offset, b_row->item.ptr+b_offset, 
-					a_row->item.size-a_offset);
-				if(result)
-					return result;
-				pos+=a_row->item.size;
-				a_row++; a_countdown--; a_offset=0;
-				b_row++; b_countdown--; b_offset=0;
-			} else if (size_diff>0) { // a longer
-				result=memcmp(a_row->item.ptr+a_offset, b_row->item.ptr+b_offset, 
-					b_row->item.size-b_offset);
-				if(result)
-					return result;
-				a_offset+=b_row->item.size-b_offset;
-				b_row++; b_countdown--; b_offset=0;
-			} else { // b longer
-				result=memcmp(a_row->item.ptr+a_offset, b_row->item.ptr+b_offset, 
-					a_row->item.size-a_offset);
-				if(result)
-					return result;
-				b_offset+=a_row->item.size-a_offset;
-				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;
-				b_countdown=b_chunk->count;
-			}
-		} else {
-			a_offset-=a_row->item.size;
-			pos+=a_row->item.size;
-			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;
-			a_countdown=a_chunk->count;
-		}
-	}
-	if(a_break==b_break) { // ended simultaneously
-		partial=0; return 0;
-	} else if(a_break) { // first bytes equal, but a ended before b
-		partial=1; return -1;
-	} else {
-		partial=2; return +1;
-	}
-}
+const String& String::mid(size_t substr_start, size_t substr_end) const {
+	String& result=*new String;
 
-/// @todo move 'lang' skipping to pos
-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, a_size-1);
-
-	const Chunk *a_chunk=&head;
-	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;
-	uint a_countdown=a_chunk->count;
-	size_t pos=0;
-
-	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) 
-				return -1; // wrong lang -- bail out
-
-			int size_diff=
-				(a_row->item.size-a_offset)-
-				(b_size-b_offset);
-			
-			if(size_diff==0) { // a has same size as b
-				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(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(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;
-				pos+=a_row->item.size;
-				a_row++; a_countdown--; a_offset=0;
-			}
-		} else {
-			a_offset-=a_row->item.size; 
-			pos+=a_row->item.size;
-			a_row++; a_countdown--; 
-		}
+	substr_start=min(substr_start, size());
+	substr_end=max(substr_start, substr_end);
+	if(substr_start==substr_end)
+		return result;
 
-		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;
-			a_countdown=a_chunk->count;
-		}
-	}
-	if(a_break==b_break) { // ended simultaneously
-		partial=0; return 0;
-	} else if(a_break) { // first bytes equal, but a ended before b
-		partial=1; return -1;
-	} else {
-		partial=2; return +1;
-	}
-}
+	// first: letters themselves
+	result.body=CORD_substr(body, substr_start, substr_end-substr_start);
 
-#ifndef NO_STRING_ORIGIN
-const Origin& String::origin() const { 
-	if(is_empty()) {
-		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"
-	// 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(size_t(0), start);
-	finish=min(size(), finish);
-	if(start==finish)
-		return result;
+	// next: their langs
+	result.fragments.append(fragments, substr_start, substr_end);
 
-	size_t pos=0;
-	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;
-		}
-		pos+=row->item.size;
-	);
-break2:
 //	SAPI::log(pool(), "piece of '%s' from %d to %d is '%s'",
-		//cstr(), start, finish, result.cstr());
+		//cstr(), substr_start, substr_end, result.cstr());
 	return result;
 }
 
-int String::pos(const String& substr, 
-				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
-			partial==2) // 'substr' starts 'this'+'result'
-			return result;
-	}
-	
-	return -1;
+size_t String::pos(CORD substr, 
+		   size_t this_offset, Untaint_lang lang) const {
+	// first: letters themselves
+	size_t substr_start=CORD_str(body, this_offset, substr);
+	if(substr_start==CORD_NOT_FOUND)
+		return NOT_FOUND;
+
+	// next: check the lang when specified
+
+	if(lang==UL_UNSPECIFIED) // ignore lang?
+		return substr_start;
+
+	// substr must be in one fragment, and fragments' lang must = lang
+	size_t substr_end=substr_start+CORD_len(substr);
+	size_t fragment_start=0;
+	size_t fragment_end;
+	for(Array_iterator<ArrayFragment::element_type> i(fragments); i.has_next(); fragment_start=fragment_end) {
+		const Fragment& fragment=i.next();
+		fragment_end=fragment_start+fragment.size;
+
+		if(substr_start<fragment_start) // not reached fragments which may include 'result'?
+			continue;
+		
+		if(substr_end>fragment_end) // end of substr OUT of current fragment?
+			break;
+
+		return fragment.lang==lang?substr_end:NOT_FOUND;
+	}	
+
+	return NOT_FOUND;
 }
 
-int String::pos(const char *substr, size_t substr_size, 
-				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
-			partial==2) // 'substr' starts 'this'+'result'
-			return result;
-	}
-	
-	return -1;
+size_t String::pos(const String& substr, 
+				size_t this_offset, Untaint_lang lang) const {
+	return pos(substr.body, this_offset, lang);
 }
 
-void String::split(Array& result, 
-				   size_t* pos_after_ref, 
-				   const char *delim, size_t delim_size, 
-				   Untaint_lang lang, int limit) const {
+void String::split(ArrayString& result, 
+				   size_t& pos_after, 
+				   const char* delim, 
+				   Untaint_lang lang, int limit) {
 	size_t self_size=size();
-	if(delim_size) {
-		size_t pos_after=pos_after_ref?*pos_after_ref:0;
+	if(size_t delim_size=strlen(delim)) {
 		int pos_before;
 		// while we have 'delim'...
-		for(; (pos_before=pos(delim, delim_size, pos_after, lang))>=0 && limit; limit--) {
+		for(; (pos_before=pos(delim, pos_after, lang))>=0 && limit; limit--) {
 			result+=&mid(pos_after, pos_before);
 			pos_after=pos_before+delim_size;
 		}
@@ -402,21 +189,17 @@ void String::split(Array& result,
 			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+=self_size;
+		pos_after+=self_size;
 	}
 }
 
-void String::split(Array& result, 
-				   size_t* pos_after_ref, 
+void String::split(ArrayString& result, 
+				   size_t& pos_after, 
 				   const String& delim, Untaint_lang lang, 
 				   int limit) const {
-	if(!delim.is_empty()) {
-		size_t pos_after=pos_after_ref?*pos_after_ref:0;
+	if(delim) {
 		int pos_before;
 		// while we have 'delim'...
 		for(; (pos_before=pos(delim, pos_after, lang))>=0 && limit; limit--) {
@@ -428,136 +211,138 @@ void String::split(Array& result,
 			result+=&mid(pos_after, size());
 			pos_after=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+=size();
 	}
 }
 
-static void regex_options(char *options, int *result){
+static void regex_options(const String& options, int *result, bool& need_pre_post_match){
     struct Regex_option {
-		char key;
+		const char* keyL;
+		const char* keyU;
 		int clear, set;
 		int *result;
+		bool *flag;
     } regex_option[]={
-		{'i', 0, PCRE_CASELESS, result}, // a=A
-		{'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
-		{0},
+		{"i", "I", 0, PCRE_CASELESS, result}, // a=A
+		{"s", "S", 0, PCRE_DOTALL, result}, // \n\n$ [default]
+		{"x", "U", 0, PCRE_EXTENDED, result}, // whitespace in regex ignored
+		{"m", "M", PCRE_DOTALL, PCRE_MULTILINE, result}, // ^aaa\n$^bbb\n$
+		{"g", "G", 0, true, result+1}, // many rows
+		{"'", 0, 0, 0, 0, &need_pre_post_match},
+		{0}
     };
-	result[0]=PCRE_EXTRA | PCRE_DOTALL;
+	result[0]=PCRE_EXTRA | PCRE_DOTALL | PCRE_DOLLAR_ENDONLY;
 	result[1]=0;
 
     if(options) 
-		for(Regex_option *o=regex_option; o->key; o++) 
-			if(
-				strchr(options, o->key) || 
-				strchr(options, toupper(o->key))) {
-				*(o->result)&=~o->clear;
-				*(o->result)|=o->set;
+		for(Regex_option *o=regex_option; o->keyL; o++) 
+			if(options.pos(o->keyL)>=0
+				|| (o->keyU && options.pos(o->keyU)>=0)) {
+				if(o->flag)
+					*o->flag=true;
+				else { // result
+					*o->result &= ~o->clear;
+					*o->result |= o->set;
+				}
 			}
 }
 
-/// @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,
-				   bool *was_global) const { 
-
-	if(regexp.is_empty())
-		throw Exception(0, 0,
-			aorigin,
+Table* String::match(Charset& source_charset,
+					   const String& regexp, 
+					   const String& options,
+					   Row_action row_action, void *info,
+					   bool& just_matched) const { 
+	if(!regexp)
+		throw Exception(Exception::undefined_type,
+			Exception::undefined_source,
 			"regexp is empty");
-	const char *pattern=regexp.cstr();
-	const char *errptr;
+
+	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;
+	bool need_pre_post_match=false;
+	int option_bits[2];  regex_options(options, option_bits, need_pre_post_match);
+	bool global=option_bits[1]!=0;
 	pcre *code=pcre_compile(pattern, option_bits[0], 
 		&errptr, &erroffset,
-		pool().get_source_charset().pcre_tables);
+		source_charset.pcre_tables);
 
 	if(!code)
-		throw Exception(0, 0,
+		throw Exception(Exception::undefined_type,
 			&regexp.mid(erroffset, regexp.size()),
 			"regular expression syntax error - %s", errptr);
 	
-	int info_substrings=pcre_info(code, 0, 0);
-	if(info_substrings<0) {
+	int subpatterns=pcre_info(code, 0, 0);
+	if(subpatterns<0) {
 		pcre_free(code);
-		throw Exception(0, 0,
-			aorigin,
+		throw Exception(0,
+			&regexp,
 			"pcre_info error (%d)", 
-				info_substrings);
+				subpatterns);
 	}
 
-	int startoffset=0;
-	const char *subject=cstr();
-	int length=strlen(subject);
-	int ovecsize;
-	int *ovector=(int *)malloc(sizeof(int)*
-		(ovecsize=(1/*match*/+info_substrings)*3), 11);
-
-	{ // create table
-		Array& columns=*NEW Array(pool());
-		columns+=string_pre_match_name;
-		columns+=string_match_name;
-		columns+=string_post_match_name;
-		for(int i=1; i<=info_substrings; i++) {
-			char *column=(char *)malloc(MAX_NUMBER);
-			snprintf(column, MAX_NUMBER, "%d", i);
-			columns+=NEW String(pool(), column); // .i column name
-		}
-		*table=NEW Table(pool(), aorigin, &columns);
-	}
+	C subject=cstr();
+	const int ovecsize=(1/*match*/+MAX_MATCH_GROUPS)*3;
+	int ovector[ovecsize];
+
+	// create table
+	Table* table=new Table(string_match_table_template);
 
 	int exec_option_bits=0;
+	int prestart=0;
+	int poststart=0;
+	int postfinish=size();
 	while(true) {
 		int exec_substrings=pcre_exec(code, 0,
-			subject, length, startoffset,
+			subject, subject.size, prestart,
 			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);
-			return option_bits[1]!=0; // global=true+table, not global=false
+			row_action(table, 0/*last time, no raw*/, 0, 0, poststart, postfinish, info);
+			if(global || subpatterns)
+				return table; // global or with subpatterns=true+result
+			else {
+				just_matched=false; return 0; // not global=no result
+			}
 		}
 
 		if(exec_substrings<0) {
 			pcre_free(code);
-			throw Exception(0, 0,
-				aorigin,
+			throw Exception(Exception::undefined_type,
+				&regexp,
 				"regular expression execute error (%d)", 
 					exec_substrings);
 		}
 
-		Array& row=*NEW Array(pool());
-		row+=&mid(0, ovector[0]); // .prematch column value
-		row+=&mid(ovector[0], ovector[1]); // .match
-		row+=&mid(ovector[1], size()); // .postmatch
+		int prefinish=ovector[0];
+		poststart=ovector[1];
+		ArrayString* row=new ArrayString;
+		if(need_pre_post_match) {
+			*row+=&mid(0, prefinish); // .prematch column value
+			*row+=&mid(prefinish, poststart); // .match
+			*row+=&mid(poststart, postfinish); // .postmatch
+		} else {
+			*row+=0; // .prematch column value
+			*row+=0; // .match
+			*row+=0; // .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+=&mid(ovector[i*2+0], ovector[i*2+1]); // .i column value
 		}
 		
-		row_action(**table, &row, startoffset, ovector[0], info);
+		row_action(table, row, prestart, prefinish, poststart, postfinish, info);
 
-		if(!option_bits[1] || startoffset==ovector[1]) { // not global | going to hang
+		if(!global || prestart==poststart) { // not global | going to hang
 			pcre_free(code);
-			row_action(**table, 0/*last time, no row*/, 0, 0, info);
-			return true;
+			row_action(table, 0/*last time, no row*/, 0, 0, poststart, postfinish, info);
+			return table;
 		}
-		startoffset=ovector[1];
+		prestart=poststart;
 
 /*
 		if(option_bits[0] & PCRE_MULTILINE)
@@ -566,10 +351,10 @@ bool String::match(
 	}
 }
 
-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);
+String& String::change_case(Charset& source_charset, Change_case_kind kind) {
+	String& result=*new String();
+
+	const unsigned char *tables=source_charset.pcre_tables;
 
 	const unsigned char *a;
 	const unsigned char *b;
@@ -583,7 +368,7 @@ String& String::change_case(Pool& pool,
 		b=0;
 		break;
 	default:
-		throw Exception(0, 0, 
+		throw Exception(Exception::undefined_type, 
 			this, 
 			"unknown change case kind #%d", 
 				static_cast<int>(kind)); // never
@@ -591,139 +376,86 @@ String& String::change_case(Pool& pool,
 		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];
+	C new_cstr=cstr();
+	const char *end=new_cstr+new_cstr.size;
+	char *dest=new_cstr;
+	for(const char* current=new_cstr; current<end; current++) {
+		unsigned char c=a[(unsigned char)*current];
+		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);
-	);
+		*dest++=(char)c;
+	}
+	result.body=new_cstr;
+	result.fragments.append(fragments, 0, fragments.count());
 
 	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 {
-				--row;  ++countdown; // step back from not-ours
-				break;
+const String& String::replace(const Dictionary& dict) const {
+	String& result=*new String();
+	C old_cstr=cstr();
+	const char* current=old_cstr;
+	size_t old_pos=0;
+
+	char *new_cstr=new char[(size_t)ceil(old_cstr.size*dict.max_ratio())];
+	char *dest=new_cstr;
+	size_t remaining_size=old_cstr.size;
+	do {
+		if(Table::element_type row=dict.first_that_starts(current, remaining_size)) {
+			{ // prematch
+				size_t now_pos=current-old_cstr;
+				if(now_pos!=old_pos) {
+					result.fragments.append(fragments, old_pos, now_pos);
+					old_pos=now_pos;
+				}
 			}
-		);
-		
-		// set pointers after joined piece
-		achunk=chunk;  arow=row;  acountdown=countdown;
-		// & one step back, see String::reconstruct
-		--row;  ++countdown;
-	}
-}
-
-String& String::reconstruct(Pool& pool) const {
-	//_asm int 3;
-	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);
-
-		result.APPEND(joined_ptr, joined_size, joined_lang,
-			joined_origin_file, joined_origin_line);
-	);
-
-	return result;
-};
 
-String& String::replace_in_reconstructed(Pool& pool, Dictionary& dict) const {
-	//_asm int 3;
-	String& result=*new(pool) String(pool);
-	STRING_FOREACH_ROW(
-		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()), 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--;
+			const String* a=row->get(0); size_t a_size=a->size();
+			// skip 'a' in 'current', in fragment lang index && reduce work size
+			current+=a_size;  old_pos+=a_size;  remaining_size-=a_size;
+
+			if(row->count()>1) { // are there any b?
+				const String* b=row->get(1); size_t b_size=b->size();
+				if(b_size) {
+					// write 'b' to 'dest' && skip 'b' in 'dest'
+					b->store_to(dest, String::UL_AS_IS);  dest+=b_size;
+
+					// append match fragments
+					result.fragments.append(b->fragments, 0, b->fragments.count());
+				}
 			}
+		} else {
+			// write a char to b && reduce work size
+			*dest++=*current++;  remaining_size--;
 		}
+	} while(remaining_size);
 
-		result.APPEND(new_cstr, dest-new_cstr, 
-			row->item.lang,
-			row->item.origin.file, row->item.origin.line);
-	);
-	return result;
-}
+	{ // postmatch
+		size_t now_pos=current-old_cstr;
+		if(now_pos!=old_pos) // have something at end?
+			result.fragments.append(fragments, old_pos, now_pos-old_pos);
+	}
 
-String& String::replace(Pool& pool, Dictionary& dict) const {
-	return reconstruct(pool).replace_in_reconstructed(pool, dict);
+	result.body=new_cstr;
+	return result;
 }
 
-double String::as_double() const { 
+double String::as_double() { 
 	double result;
-	const char *cstr;
-	char buf[MAX_NUMBER];
-	if(head.rows+1==append_here) {
-		int size=min(head.rows[0].item.size, MAX_NUMBER-1);
-		memcpy(buf, head.rows[0].item.ptr, size);
-		buf[size]=0;
-		cstr=buf;
-	} else
-		cstr=this->cstr();
+	char buf[MAX_STRING];
+	if(size()>MAX_STRING-1)
+		throw Exception("number.format",
+			this,
+			"invalid number too long a string (%u>%u)", size(), MAX_STRING-1);
+	char *eol=store_to(buf, String::UL_AS_IS); *eol=0;
+	const char* cstr=buf;
+
+	while(*cstr && isspace(*cstr))
+		cstr++;
+	if(!*cstr)
+		return 0;
+
 	char *error_pos;
 	// 0xABC
 	if(cstr[0]=='0')
@@ -734,24 +466,29 @@ double String::as_double() const {
 	else
 		result=(double)strtod(cstr, &error_pos);
 
-	if(*error_pos/*not EOS*/)
-		throw Exception(0, 0,
-			this,
-			"invalid number (double)");
+	while(char c=*error_pos++)
+		if(!isspace(c))
+			throw Exception("number.format",
+				this,
+				"invalid number (double)");
 
 	return result;
 }
-int String::as_int() const { 
+int String::as_int() { 
 	int result;
-	const char *cstr;
-	char buf[MAX_NUMBER];
-	if(head.rows+1==append_here) {
-		int size=min(head.rows[0].item.size, MAX_NUMBER-1);
-		memcpy(buf, head.rows[0].item.ptr, size);
-		buf[size]=0;
-		cstr=buf;
-	} else
-		cstr=this->cstr();
+	char buf[MAX_STRING];
+	if(size()>MAX_STRING-1)
+		throw Exception("number.format",
+			this,
+			"invalid number too long a string (%u>%u)", size(), MAX_STRING-1);
+	char *eol=store_to(buf, String::UL_AS_IS); *eol=0;
+	const char* cstr=buf;
+
+	while(*cstr && isspace(*cstr))
+		cstr++;
+	if(!*cstr)
+		return 0;
+
 	char *error_pos;
 	// 0xABC
 	if(cstr[0]=='0')
@@ -762,68 +499,98 @@ int String::as_int() const {
 	else
 		result=(int)strtol(cstr, &error_pos, 0);
 
-	if(*error_pos/*not EOS*/)
-		throw Exception(0, 0,
-			this,
-			"invalid number (int)");
+	while(char c=*error_pos++)
+		if(!isspace(c))
+			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=
+inline void uint2uchars(uint word, uchar *bytes) {
+	bytes[0]=word&0xFF;
+	bytes[1]=(word>>8)&0xFF;
+	bytes[2]=(word>>16)&0xFF;
+	bytes[3]=(word>>24)&0xFF;
+}
+inline uint uchars2uint(uchar *bytes) {
+	return bytes[3]<<24
+		| bytes[2]<<16
+		| bytes[1]<<8
+		| bytes[0];
+}
+
+static int CORD_batched_iter_fn_append(const char* s, void* client_data) {
+	char*& cur=*static_cast<char**>(client_data);
+
+	size_t size=strlen(s);
+	memcpy(cur, s, size);  cur+=size;
+	return 0;
+};
+String::C String::serialize(size_t prolog_size) const {
+	size_t buf_size=
 		prolog_size
-		+used_rows()*(sizeof(uchar)+sizeof(ushort))
+		+fragments.count()*(sizeof(Untaint_lang)+sizeof(size_t))
 		+size();
-	buf=malloc(buf_size,15);
-	char *cur=(char *)buf+prolog_size;
+	C result(new char[buf_size], buf_size);
 
-	STRING_FOREACH_ROW(
+	// 1: prolog
+	char *cur=result.str+prolog_size;
+
+
+	// 2: fragments.count
+	size_t fragments_count=fragments.count();
+	memcpy(cur, &fragments_count, sizeof(fragments_count));  cur+=sizeof(fragments_count);
+
+	// 3: lang info
+	for(Array_iterator<ArrayFragment::element_type> i(fragments); i.has_next(); ) {
+		const Fragment& fragment=i.next();
 		// lang
-		memcpy(cur, &row->item.lang, sizeof(uchar));
-		cur+=sizeof(uchar);
+		memcpy(cur, &fragment.lang, sizeof(fragment.lang));  cur+=sizeof(fragment.lang);
 		// 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;
-	);
+		memcpy(cur, &fragment.size, sizeof(fragment.size));  cur+=sizeof(fragment.size);
+	}
+
+	// 4: letters
+	CORD_iter5(body, 0, 0, CORD_batched_iter_fn_append, &cur);
+
+	return result;
 }
-void String::deserialize(size_t prolog_size, void *buf, size_t buf_size, const char *file) {
+bool String::deserialize(size_t prolog_size, void *buf, size_t buf_size, const char* file) {
 	if(buf_size<=prolog_size)
-		return;
-
-	char *cur=(char *)buf+prolog_size;
+		return false;
 	buf_size-=prolog_size;
 
-	while(buf_size) {
-		uchar lang=*(uchar *)(cur);
-		
-		ushort size=uchars2ushort(
-			*(uchar*)(cur+sizeof(uchar)*1),
-			*(uchar*)(cur+sizeof(uchar)*2)
-		);
+	// 1: prolog
+	const char *cur=(const char *)buf+prolog_size;
 
-		const char *ptr=(const char*)(cur+sizeof(uchar)*3); 
-		APPEND(ptr, size, lang, file, 0);
+	// 2: fragments.count
+	if(buf_size<sizeof(size_t)) // fragments.count don't fit?
+		return false;
+	size_t fragments_count=*reinterpret_cast<const size_t*>(cur);  cur+=sizeof(size_t);
+	buf_size-=sizeof(size_t);
+
+	// 3: lang info
+	size_t total_size=0;
+	for(size_t f=0; f<fragments_count; f++) {
+		size_t piece_size=sizeof(Untaint_lang)+sizeof(size_t);
+		if(buf_size<piece_size) // lang+size
+			return false;
+
+		Untaint_lang lang=*reinterpret_cast<const Untaint_lang *>(cur);  cur+=sizeof(Untaint_lang);
+		size_t fragment_size=*reinterpret_cast<const size_t*>(cur);  cur+=sizeof(size_t);
+		fragments+=Fragment(lang, fragment_size);
+		total_size+=fragment_size;
 
-		size_t piece_size=sizeof(uchar)+sizeof(ushort)+size;
-		cur+=piece_size;
 		buf_size-=piece_size;
 	}
+
+	// 4: letters
+	if(buf_size!=total_size)
+		return false;
+
+	body=CORD_cat_char_star(CORD_EMPTY, cur, buf_size);
+
+	return true;
 }