--- parser3/src/main/pa_string.C	2001/04/23 13:38:31	1.78
+++ parser3/src/main/pa_string.C	2003/03/19 12:22:36	1.172.2.21.2.3
@@ -1,132 +1,95 @@
 /** @file
 	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 $
+	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: 2003/03/19 12:22:36 $";
 
 #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_sapi.h"
+#include "pa_dictionary.h"
+#include "pa_charset.h"
 
-// String
+// helpers
 
-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;
+/// String::match uses this as replace & global search table columns
 
-	if(src)
-		if(tainted)
-			APPEND_TAINTED(src, src_size, 0, 0);
-		else
-			APPEND_CLEAN(src, src_size, 0, 0);
-}
+const int MAX_MATCH_GROUPS=100;
 
-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()) {
-	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;
-			}
+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));
 		}
 	}
-	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)
+Table string_match_table_template(new String_match_table_template_columns);
+
+// String::ArrayFragment methods
+
+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;
-			
-			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);
+
+		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);
+		else
+			APPEND_CLEAN(src, src_size, 0, 0);
+}
+
+String::String(const String& src): body(src.body), fragments(src.fragments) {}
 
 String& String::real_append(STRING_APPEND_PARAMS) {
 	if(!src)
@@ -136,306 +99,106 @@ String& String::real_append(STRING_APPEN
 	if(!size)
 		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, src, size);
+	fragments+=Fragment(lang, size);
 
 	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);
+};
 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);
-
-	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_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_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--; 
-		}
+const String& String::mid(size_t substr_start, size_t substr_end) const {
+	String& result=*new String;
 
-		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;
-	}
-}
+	substr_start=min(substr_start, size());
+	substr_end=max(substr_start, substr_end);
+	if(substr_start==substr_end)
+		return result;
 
-/// @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;
+	// first: letters themselves
+	result.body=CORD_substr(body, substr_start, substr_end-substr_start);
 
-		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(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--; 
-		}
+	// next: their langs
+	result.fragments.append(fragments, substr_start, substr_end);
 
-		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;
-	}
+//	SAPI::log(pool(), "piece of '%s' from %d to %d is '%s'",
+		//cstr(), substr_start, substr_end, result.cstr());
+	return result;
 }
 
-#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::mid(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 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;
 
-	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;
+		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;
 
-			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());
-	return result;
-}
+		return fragment.lang==lang?substr_end:NOT_FOUND;
+	}	
 
-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;
+	return 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_size=size();
+	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;
 		}
 		// 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+=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.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--) {
@@ -447,138 +210,138 @@ 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$
-		{'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;
+				}
 			}
 }
 
-/**
-	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 unsigned char *pcre_tables,
-				   const String *aorigin,
-				   const String& regexp, 
-				   const String *options,
-				   Table **table,
-				   Row_action row_action, void *info) const { 
-
-	if(!regexp.size())
-		THROW(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(UL_AS_IS);
-	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);
+	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,
-		pcre_tables);
+		source_charset.pcre_tables);
 
 	if(!code)
-		THROW(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) {
-		(*pcre_free)(code);
-		THROW(0, 0,
-			aorigin,
+	int subpatterns=pcre_info(code, 0, 0);
+	if(subpatterns<0) {
+		pcre_free(code);
+		throw Exception(0,
+			&regexp,
 			"pcre_info error (%d)", 
-				info_substrings);
+				subpatterns);
 	}
 
-	int startoffset=0;
-	const char *subject=cstr(UL_AS_IS);
-	int length=strlen(subject);
-	int ovecsize;
-	int *ovector=(int *)malloc(sizeof(int)*
-		(ovecsize=(1/*match*/+info_substrings)*3));
-
-	{ // 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
+			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_substrings<0) {
-			(*pcre_free)(code);
-			THROW(0, 0,
-				aorigin,
+			pcre_free(code);
+			throw Exception(Exception::undefined_type,
+				&regexp,
 				"regular expression execute error (%d)", 
 					exec_substrings);
 		}
 
-		Array& row=*NEW Array(pool());
-		row+=&mid(0, ovector[0]); // .pre-match column value
-		row+=&mid(ovector[0], ovector[1]); // .match
-		row+=&mid(ovector[1], size()); // .post-match
+		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
-			(*pcre_free)(code);
-			(*row_action)(**table, 0/*last time, no row*/, 0, 0, info);
-			return true;
+		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)
@@ -586,3 +349,223 @@ bool String::match(const unsigned char *
 */
 	}
 }
+
+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;
+	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(Exception::undefined_type, 
+			this, 
+			"unknown change case kind #%d", 
+				static_cast<int>(kind)); // never
+		a=b=0; // calm, compiler
+		break; // never
+	}	
+
+	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.body=new_cstr;
+	result.fragments.append(fragments, 0, fragments.count());
+
+	return result;
+}
+
+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;
+				}
+			}
+
+			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);
+
+	{ // 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);
+	}
+
+	result.body=new_cstr;
+	return result;
+}
+
+double String::as_double() { 
+	double result;
+	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')
+		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);
+
+	while(char c=*error_pos++)
+		if(!isspace(c))
+			throw Exception("number.format",
+				this,
+				"invalid number (double)");
+
+	return result;
+}
+int String::as_int() { 
+	int result;
+	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')
+		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);
+
+	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];
+}
+
+void String::serializesize_t prolog_size, char *& buf, size_t& buf_size) const {
+	buf_size=
+		prolog_size
+		+fused*(sizeof(String_UL)+sizeof(size_t))
+		+size();
+	buf=new char[buf_size];
+	char *cur=buf+prolog_size;
+
+	STRING_FOREACH_FRAGMENT(
+		// lang
+		memcpy(cur, &fragment->lang, sizeof(fragment->lang));
+		cur+=sizeof(fragment->lang);
+		// size
+		// bug on some sparc platform [you can't work with integers on odd pointers]
+		// forces us to use byte array instead
+		uchar bytes[4];
+		uint2uchars(fragment->size, bytes);
+		memcpy(cur, &bytes, sizeof(bytes)); cur+=sizeof(bytes);
+		// bytes
+		memcpy(cur, fragment->ptr, fragment->size);
+		cur+=fragment->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(String_UL)+sizeof(size_t)>buf_size) // lang+size
+			return false;
+
+		String_UL lang=*(String_UL *)(cur);		
+		size_t size=uchars2uint((uchar *)cur);
+
+		size_t piece_size=sizeof(String_UL)+sizeof(size_t)+size;
+		if(piece_size>buf_size) // buffer overrun, can be on incomplete cache files
+			return false;
+
+		const char* ptr=(const char*)(cur+sizeof(String_UL)+sizeof(size_t)); 
+		APPEND(ptr, size, lang, file, 0);
+
+		cur+=piece_size;
+		buf_size-=piece_size;
+	}
+	return true;
+}