--- parser3/src/main/pa_string.C	2001/04/03 09:58:10	1.62
+++ parser3/src/main/pa_string.C	2003/03/20 16:57:43	1.172.2.21.2.13
@@ -1,527 +1,586 @@
 /** @file
-	Parser: string class. @see untasize_t.C.
+	Parser: string class. @see untalength_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 1.62 2001/04/03 09:58:10 paf Exp $
+	Copyright (c) 2001-2003 ArtLebedev Group (http://www.artlebedev.com)
+	Author: Alexandr Petrosian <paf@design.ru> (http://paf.design.ru)
 */
 
-#include "pa_config_includes.h"
+static const char* IDENT_STRING_C="$Date: 2003/03/20 16:57:43 $";
 
-#include <locale.h>
+#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_threads.h"
+#include "pa_dictionary.h"
+#include "pa_charset.h"
 
-#include "pcre.h"
+// helpers
 
-//#include "pa_sapi.h"
+/// String::match uses this as replace & global search table columns
 
-// consts
+const int MAX_MATCH_GROUPS=100;
 
-const int MAX_MATCH_COLUMNS=20;
+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(PointerFreeGC) char[3/*strlen("100")*/+1/*terminating 0*/];
+			*this+=new String(cname, sprintf(cname, "%d", 1+i));
+		}
+	}
+};
 
-// String
+Table string_match_table_template(new String_match_table_template_columns);
 
-String::String(Pool& apool, const char *src, bool tasize_ted) :
-	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;
-
-	if(src)
-		if(tasize_ted)
-			APPEND_TAINTED(src, 0, 0, 0);
-		else
-			APPEND_CONST(src);
-}
+// String::ArrayFragment methods
 
-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;
-}
+void String::ArrayFragment::append_positions(const ArrayFragment& src, 
+					     size_t substr_begin, size_t substr_end) {
+	if(substr_begin==substr_end)
+		return;
 
-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;
+	size_t fragment_begin=0;
+	size_t fragment_end;
+	for(Array_iterator<element_type> i(src); ; fragment_begin=fragment_end) {
+		const element_type& fragment=i.next();
+		fragment_end=fragment_begin+fragment.length;		
+
+		if(fragment_begin<substr_begin) // 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_begin);
+		
+		while(substr_end>fragment_end) { // are there more fragments including pieces of 'substr'?
+			fragment_begin=fragment_end;
+
+			const element_type& fragment=i.next();
+			fragment_end=fragment_begin+fragment.length;
+
+			if(substr_end>fragment_end) // are there still more?
+				append(Fragment(fragment.lang, fragment.length)); // appending whole fragment
+			else { // no, it was last
+				append(Fragment(fragment.lang, substr_end-fragment_begin));
+				return;
 			}
 		}
-	}
-	link_row->link=0;
-	fused_rows=src_used_rows;
-	fsize=src.fsize;
-}
-
-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;
+
+		break;
+	}	
+}
+/*
+void String::ArrayFragment::mid(ArrayFragment& result, size_t substr_begin, size_t substr_end) {
+}
+*/
+
+// String methods
+
+String::String(const char* cstr, size_t helper_length, bool tainted): body(CORD_EMPTY) {
+	append(cstr, helper_length, tainted?UL_TAINTED:UL_CLEAN);
 }
 
-String& String::real_append(STRING_APPEND_PARAMS) {
-	if(!src)
+String::String(const String& src): body(src.body), fragments(src.fragments) {}
+
+String& String::append(const char* str, size_t helper_length, Untaint_lang lang) {
+	if(!str)
 		return *this;
-	if(!size)
-		size=strlen(src);
-	if(!size)
+	size_t use_length=helper_length?helper_length:strlen(str);
+	if(!use_length)
 		return *this;
 
-	if(chunk_is_full())
-		expand();
-
-	append_here->item.ptr=src;
-	fsize+=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++;
+	body=CORD_cat_char_star(body, str, use_length);
+	fragments+=Fragment(lang, use_length);
 
 	return *this;
 }
 
+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;
-
-	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;
-
-			result=Hash::generic_code(result, row->item.ptr, row->item.size);
-			row++;
-		}
-		chunk=row->link;
-	} while(chunk);
-break2:
+	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;
-	this_offset=min(this_offset, size()-1);
+/// @todo check in doc: whether it documents NOW bad situation "abc".mid(-1, 3) =were?="ab"
+const String& String::mid(size_t substr_begin, size_t substr_end) const {
+	String& result=*new String;
+
+	size_t self_length=length();
+	substr_begin=max(min(substr_begin, self_length), (size_t)0);
+	substr_end=min(max(substr_end, substr_begin), self_length);
+	if(substr_begin==substr_end)
+		return result;
 
-	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;
-	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;
-
-		if(pos+a_row->item.size > this_offset) {
-			if(lang!=UL_UNKNOWN && 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_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--; 
-		}
+	// first: letters themselves
+	result.body=CORD_substr(body, substr_begin, substr_end-substr_begin);
 
-		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;
-	}
-}
+	// next: their langs
+	result.fragments.append_positions(fragments, substr_begin, substr_end);
 
-/// @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 b_size=src_size?src_size:b_ptr?strlen(b_ptr):0;
-	this_offset=min(this_offset, 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;
-	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;
-
-		if(pos+a_row->item.size > this_offset) {
-			if(lang!=UL_UNKNOWN && 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(size_t 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, 
-					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, 
-					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--; 
-		}
-
-		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;
-	}
-}
-
-#ifndef NO_STRING_ORIGIN
-const Origin& String::origin() const { 
-	if(!fused_rows)
-		THROW(0, 0, 
-			0,
-			"String::origin() of empty string called");
-	
-	// determining origin by last appended piece
-	// because first one frequently constant. 
-	// ex: ^load[/file] "document_root" + "/file"
-	return append_here[-1].item.origin; 
-}
-#endif
-
-String& String::piece(size_t start, size_t finish) const {
-	start=max(0, start);
-	finish=min(size(), finish);
-	if(start==finish)
-		return *empty_string;
-
-	String& result=*NEW String(pool());
-
-	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)
-				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);
-break2:
 //	SAPI::log(pool(), "piece of '%s' from %d to %d is '%s'",
-		//cstr(), start, finish, result.cstr());
+		//cstr(), substr_begin, substr_end, result.cstr());
 	return result;
 }
 
-int String::pos(const String& substr, 
-				size_t result, Untaint_lang lang) const {
-	for(; result<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_begin=CORD_str(body, this_offset, substr);
+	if(substr_begin==CORD_NOT_FOUND)
+		return STRING_NOT_FOUND;
+
+	// next: check the lang when specified
+
+	if(lang==UL_UNSPECIFIED) // ignore lang?
+		return substr_begin;
+
+	// substr must be in one fragment, and fragments' lang must = lang
+	size_t substr_end=substr_begin+CORD_len(substr);
+	size_t fragment_begin=0;
+	size_t fragment_end;
+	for(Array_iterator<ArrayFragment::element_type> i(fragments); i.has_next(); fragment_begin=fragment_end) {
+		const Fragment& fragment=i.next();
+		fragment_end=fragment_begin+fragment.length;
+
+		if(substr_begin<fragment_begin) // not reached fragments which may include 'result'?
+			continue;
+		if(substr_begin>=fragment_end) // begin of substr OUT of current fragment?
+			continue;
+		
+		if(substr_end>fragment_end) // end of substr OUT of current fragment?
+			throw Exception(0, // (*) see below
+				this,
+				"searching for '%s' starting from %ud problem: found begin in one fragment, but end in another",
+					CORD_to_const_char_star(substr), this_offset);
+
+		if(fragment.lang==lang)
+			return substr_begin;
+		else { // bad lang...
+			/// WARNING: this possibly skips assert (*), but it's fast
+			substr_begin=CORD_str(body, fragment_end/*...search AFTER for more*/, substr);
+			if(substr_begin==CORD_NOT_FOUND)
+				return STRING_NOT_FOUND;
+
+			size_t substr_end=substr_begin+CORD_len(substr);
+			// and continuing with next fragment
+		}
+	}	
+
+	return STRING_NOT_FOUND;
 }
 
-int String::pos(const char *substr, size_t substr_size, 
-				size_t result, Untaint_lang lang) const {
-	for(; result<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 {
-	if(delim_size) {
-		size_t pos_after=pos_after_ref?*pos_after_ref:0;
+void String::split(ArrayString& result, 
+				   size_t& pos_after, 
+				   const char* delim, 
+				   Untaint_lang lang, int limit) {
+	size_t self_length=length();
+	if(size_t delim_length=strlen(delim)) {
 		int pos_before;
 		// while we have 'delim'...
-		for(; (pos_before=pos(delim, delim_size, pos_after, lang))>=0 && limit; limit--) {
-			result+=&piece(pos_after, pos_before);
-			pos_after=pos_before+delim_size;
+		for(; (pos_before=pos(delim, pos_after, lang))>=0 && limit; limit--) {
+			result+=&mid(pos_after, pos_before);
+			pos_after=pos_before+delim_length;
 		}
 		// last piece
-		if(pos_after<size() && limit) {
-			result+=&piece(pos_after, size());
-			pos_after=size();
+		if(pos_after<self_length && limit) {
+			result+=&mid(pos_after, self_length);
+			pos_after=self_length;
 		}
-		if(pos_after_ref)
-			*pos_after_ref=pos_after;
 	} else { // empty delim
 		result+=this;
-		if(pos_after_ref)
-			*pos_after_ref+=size();
+		pos_after+=self_length;
 	}
 }
 
-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.size()) {
-		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--) {
-			result+=&piece(pos_after, pos_before);
-			pos_after=pos_before+delim.size();
+			result+=&mid(pos_after, pos_before);
+			pos_after=pos_before+delim.length();
 		}
 		// last piece
-		if(pos_after<size() && limit) {
-			result+=&piece(pos_after, size());
-			pos_after=size();
+		if(pos_after<length() && limit) {
+			result+=&mid(pos_after, length());
+			pos_after=length();
 		}
-		if(pos_after_ref)
-			*pos_after_ref=pos_after;
 	} else { // empty delim
 		result+=this;
-		if(pos_after_ref)
-			*pos_after_ref+=size();
+		pos_after+=length();
 	}
 }
 
-/// @test setlocale param to auto.p  @test pcre_malloc & pcre_free substs
-bool String::match(const String *aorigin,
-				   const String& regexp, 
-				   const String& options,
-				   Table **table) const { 
-	SYNCHRONIZED(true);
-	static const unsigned char *tables=0;
-	if(!tables) {
-		setlocale(LC_CTYPE, "ru");
-		tables = pcre_maketables();
-	}
-	const char *pattern=regexp.cstr();
-	const char *errptr;
+static void regex_options(const String& options, int *result, bool& need_pre_post_match){
+    struct Regex_option {
+		const char* keyL;
+		const char* keyU;
+		int clear, set;
+		int *result;
+		bool *flag;
+    } regex_option[]={
+		{"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 | PCRE_DOLLAR_ENDONLY;
+	result[1]=0;
+
+    if(options) 
+		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;
+				}
+			}
+}
+
+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(0,
+			0,
+			"regexp is empty");
+
+	const char* pattern=regexp.cstr();
+	const char* errptr;
 	int erroffset;
-	pcre *code=pcre_compile(pattern, 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,
-		tables);
+		source_charset.pcre_tables);
 
 	if(!code)
-		THROW(0, 0,
-			&regexp.piece(erroffset, regexp.size()),
-			errptr);
+		throw Exception(0,
+			&regexp.mid(erroffset, regexp.length()),
+			"regular expression syntax error - %s", errptr);
 	
-	int ovecsize;
-	int *ovector=(int *)malloc(sizeof(int)*(ovecsize=(1/*.match*/+MAX_MATCH_COLUMNS)*3));
-	const char *subject=cstr();
-	int length=strlen(subject);
-	int exec_result=pcre_exec(code, 0,
-          subject, length, 0/*startoffset*/,
-          0/*options*/, ovector, ovecsize);
+	int subpatterns=pcre_info(code, 0, 0);
+	if(subpatterns<0) {
+		pcre_free(code);
+		throw Exception(0,
+			&regexp,
+			"pcre_info error (%d)", 
+				subpatterns);
+	}
+
+	const char* subject=cstr();
+	size_t subject_length=strlen(subject);
+	const int oveclength=(1/*match*/+MAX_MATCH_GROUPS)*3;
+	int ovector[oveclength];
+
+	// create table
+	Table* table=new Table(string_match_table_template);
+
+	int exec_option_bits=0;
+	int prestart=0;
+	int poststart=0;
+	int postfinish=length();
+	while(true) {
+		int exec_substrings=pcre_exec(code, 0,
+			subject, subject_length, prestart,
+			exec_option_bits, ovector, oveclength);
+		
+		if(exec_substrings==PCRE_ERROR_NOMATCH) {
+			pcre_free(code);
+			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_result==PCRE_ERROR_NOMATCH) {
-		*table=0;
-		return false;
+		if(exec_substrings<0) {
+			pcre_free(code);
+			throw Exception(0,
+				&regexp,
+				"regular expression execute error (%d)", 
+					exec_substrings);
+		}
+
+		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_action(table, row, prestart, prefinish, poststart, postfinish, info);
+
+		if(!global || prestart==poststart) { // not global | going to hang
+			pcre_free(code);
+			row_action(table, 0/*last time, no row*/, 0, 0, poststart, postfinish, info);
+			return table;
+		}
+		prestart=poststart;
+
+/*
+		if(option_bits[0] & PCRE_MULTILINE)
+			exec_option_bits|=PCRE_NOTBOL; // start of subject+startoffset not BOL
+*/
 	}
+}
 
-	if(exec_result<0)
-		THROW(0, 0,
-			0,
-			"pcre_exec failed");
+String& String::change_case(Charset& source_charset, Change_case_kind kind) {
+	String& result=*new String();
 
-	if(exec_result==0)
-		THROW(0, 0,
-			aorigin,
-			"produced more substrings than maximum handled by Parser, which is %d",
-				MAX_MATCH_COLUMNS);
-
-	Array& columns=*NEW Array(pool());
-	columns+=string_match_name; // .match column name
-	Array& row=*NEW Array(pool());
-	row+=&piece(ovector[0], ovector[1]); // match column value
-	
-	for(int i=1; i<exec_result; i++) {
-		char *column=(char *)malloc(MAX_NUMBER);
-		snprintf(column, MAX_NUMBER, "%d", i);
-		columns+=NEW String(pool(), column); // .i column name
-		row+=&piece(ovector[i*2+0], ovector[i*2+1]); // .i column value
+	const unsigned char *tables=source_charset.pcre_tables;
+
+	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
+	}	
+
+	char* new_cstr=cstrm();
+	char *dest=new_cstr;
+	unsigned char index;
+	for(const char* current=new_cstr; index=(unsigned char)*current; current++) {
+		unsigned char c=a[index];
+		if(b)
+			c=b[c];
+
+		*dest++=(char)c;
 	}
-	
-	*table=NEW Table(pool(), aorigin, &columns);
-	(**table)+=&row;
+	result.body=new_cstr;
+	result.fragments.append(fragments);
+
+	return result;
+}
+
+const String& String::replace(const Dictionary& dict) const {
+	String& result=*new String();
+	const char* old_cstr=cstr();
+	const char* prematch_begin=old_cstr;
+
+	for(const char* current=old_cstr; *current; ) {
+		if(Table::element_type row=dict.first_that_begins(current)) {
+			// prematch
+			if(size_t prematch_length=current-prematch_begin) {
+				result.body=CORD_cat_char_star(result.body, prematch_begin, prematch_length);
+				result.fragments.append_positions(fragments, prematch_begin-old_cstr, prematch_length);
+			}
+
+			// match
+
+			const String* a=row->get(0);
+			// skip 'a' in 'current'; move prematch_begin
+			current+=a->length(); prematch_begin=current;
+
+			if(row->count()>1) { // are there any b?
+				const String* b=row->get(1);
+				if(size_t b_length=b->length()) {
+					result.body=CORD_cat(result.body, b->cstr_to_cord());
+					result.fragments.append(b->fragments);
+				}
+			}
+		} else // simply advance
+			current++; 
+	}
+
+	// postmatch
+	if(size_t postmatch_length=current-prematch_begin) {
+		result.body=CORD_cat_char_star(result.body, prematch_begin, postmatch_length);
+		result.fragments.append_positions(fragments, prematch_begin-old_cstr, postmatch_length);
+	}
+
+	return result;
+}
+
+double String::as_double() const { 
+	double result;
+	const char *str=cstr();
+
+	while(*str && isspace(*str))
+		str++;
+	if(!*str)
+		return 0;
+
+	char *error_pos;
+	// 0xABC
+	if(str[0]=='0')
+		if(str[1]=='x' || str[1]=='X')
+			result=(double)(unsigned long)strtol(str, &error_pos, 0);
+		else
+			result=(double)strtod(str+1/*skip leading 0*/, &error_pos);
+	else
+		result=(double)strtod(str, &error_pos);
+
+	while(char c=*error_pos++)
+		if(!isspace(c))
+			throw Exception("number.format",
+				this,
+				"invalid number (double)");
+
+	return result;
+}
+int String::as_int() const { 
+	int result;
+	const char *str=cstr();
+
+	while(*str && isspace(*str))
+		str++;
+	if(!*str)
+		return 0;
+
+	char *error_pos;
+	// 0xABC
+	if(str[0]=='0')
+		if(str[1]=='x' || str[1]=='X')
+			result=(int)(unsigned long)strtol(str, &error_pos, 0);
+		else
+			result=(int)strtol(str+1/*skip leading 0*/, &error_pos, 0);
+	else
+		result=(int)strtol(str, &error_pos, 0);
+
+	while(char c=*error_pos++)
+		if(!isspace(c))
+			throw Exception("number.format",
+				this,
+				"invalid number (int)");
+
+	return result;
+}
+
+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 length=strlen(s);
+	memcpy(cur, s, length);  cur+=length;
+	return 0;
+};
+String::C String::serialize(size_t prolog_length) const {
+	size_t buf_length=
+		prolog_length
+		+fragments.count()*(sizeof(Untaint_lang)+sizeof(size_t))
+		+length();
+	C result(new(PointerFreeGC) char[buf_length], buf_length);
+
+	// 1: prolog
+	char *cur=result.str+prolog_length;
+
+
+	// 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, &fragment.lang, sizeof(fragment.lang));  cur+=sizeof(fragment.lang);
+		// length
+		memcpy(cur, &fragment.length, sizeof(fragment.length));  cur+=sizeof(fragment.length);
+	}
+
+	// 4: letters
+	CORD_iter5(body, 0, 0, CORD_batched_iter_fn_append, &cur);
+
+	return result;
+}
+bool String::deserialize(size_t prolog_length, void *buf, size_t buf_length, const char* file) {
+	if(buf_length<=prolog_length)
+		return false;
+	buf_length-=prolog_length;
+
+	// 1: prolog
+	const char* cur=(const char* )buf+prolog_length;
+
+	// 2: fragments.count
+	if(buf_length<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_length-=sizeof(size_t);
+
+	// 3: lang info
+	size_t total_length=0;
+	for(size_t f=0; f<fragments_count; f++) {
+		size_t piece_length=sizeof(Untaint_lang)+sizeof(size_t);
+		if(buf_length<piece_length) // lang+length
+			return false;
+
+		Untaint_lang lang=*reinterpret_cast<const Untaint_lang *>(cur);  cur+=sizeof(Untaint_lang);
+		size_t fragment_length=*reinterpret_cast<const size_t*>(cur);  cur+=sizeof(size_t);
+		fragments+=Fragment(lang, fragment_length);
+		total_length+=fragment_length;
+
+		buf_length-=piece_length;
+	}
+
+	// 4: letters
+	if(buf_length!=total_length)
+		return false;
+
+	body=CORD_cat_char_star(CORD_EMPTY, cur, buf_length);
+
 	return true;
 }