--- parser3/src/main/pa_string.C	2001/04/23 13:38:31	1.78
+++ parser3/src/main/pa_string.C	2004/10/07 09:22:01	1.201
@@ -1,584 +1,497 @@
 /** @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 1.78 2001/04/23 13:38:31 paf Exp $
+	Copyright (c) 2001-2004 ArtLebedev Group (http://www.artlebedev.com)
+	Author: Alexandr Petrosian <paf@design.ru> (http://paf.design.ru)
 */
 
-#include "pa_config_includes.h"
+static const char * const IDENT_STRING_C="$Date: 2004/10/07 09:22:01 $";
 
 #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 "pa_sapi.h"
+const String String::Empty;
 
-// String
+int pa_atoi(const char* str, const String* problem_source) {
+	if(!str)
+		return 0;
+
+	while(*str && isspace((unsigned char)*str))
+		str++;
+	if(!*str)
+		return 0;
+
+	int result;
+	char *error_pos;
+	bool negative=false;
+	if(str[0]=='-') {
+		negative=true;
+		str++;
+	} else if(str[0]=='+') {
+		str++;
+	}
+	// 0xABC
+	if(str[0]=='0')
+		if(str[1]=='x' || str[1]=='X')
+			result=(int)(unsigned long)strtol(str, &error_pos, 0);
+		else {
+			 // skip leading 0000, to disable octal interpretation
+			do str++; while(*str=='0');				
+			result=(int)strtol(str, &error_pos, 0);
+		}
+	else
+		result=(int)strtol(str, &error_pos, 0);
+	if(negative)
+		result=-result;
+
+	while(char c=*error_pos++)
+		if(!isspace((unsigned char)c))
+			throw Exception("number.format",
+				problem_source,
+				problem_source?"invalid number (int)": "'%s' is invalid number (int)", str);
 
-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;
-
-	if(src)
-		if(tainted)
-			APPEND_TAINTED(src, src_size, 0, 0);
-		else
-			APPEND_CLEAN(src, src_size, 0, 0);
-}
-
-void String::expand() {
-	size_t new_chunk_count=last_chunk->count+last_chunk->count*CR_GROW_PERCENT/100;
-	last_chunk=static_cast<Chunk *>(
-		malloc(sizeof(size_t)+sizeof(Chunk::Row)*new_chunk_count+sizeof(Chunk *)));
-	last_chunk->count=new_chunk_count;
-	link_row->link=last_chunk;
-	append_here=last_chunk->rows;
-	link_row=&last_chunk->rows[last_chunk->count];
-	link_row->link=0;
+	return result;
 }
 
-String::String(const String& src) :	Pooled(src.pool()) {
-	head.count=CR_PREALLOCATED_COUNT;
-	
-	size_t src_used_rows=src.fused_rows;
-	if(src_used_rows<=head.count) {
-		// all new rows fit size_to preallocated area
-		size_t curr_chunk_rows=head.count;
-		memcpy(head.rows, src.head.rows, sizeof(Chunk::Row)*src_used_rows);
-		append_here=&head.rows[src_used_rows];
-		link_row=&head.rows[curr_chunk_rows];
-	} else {
-		// warning: 
-		//   heavily relies on the fact 
-		//   "preallocated area is the same for all strings"
-		//
-		// info:
-		//   allocating only enough mem to fit src string rows
-		//   next append would allocate a new chunk
-		//
-		// new rows don't fit size_to preallocated area: splitting size_to two chunks
-		// preallocated chunk src to constructing head
-		memcpy(head.rows, src.head.rows, sizeof(Chunk::Row)*head.count);
-		// remaining rows size_to new_chunk
-		size_t curr_chunk_rows=src_used_rows-head.count;
-		Chunk *new_chunk=static_cast<Chunk *>(
-			malloc(sizeof(size_t)+sizeof(Chunk::Row)*curr_chunk_rows+sizeof(Chunk *)));
-		new_chunk->count=curr_chunk_rows;
-		head.preallocated_link=new_chunk;
-		append_here=link_row=&new_chunk->rows[new_chunk->count];
-
-		Chunk *old_chunk=src.head.preallocated_link; 
-		Chunk::Row *new_rows=new_chunk->rows;
-		size_t rows_left_to_copy=new_chunk->count;
-		while(true) {
-			size_t old_count=old_chunk->count;
-			Chunk *next_chunk=old_chunk->rows[old_count].link;
-			if(next_chunk) {
-				// not last source chunk
-				// taking it all
-				memcpy(new_rows, old_chunk->rows, sizeof(Chunk::Row)*old_count);
-				new_rows+=old_count;
-				rows_left_to_copy-=old_count;
-
-				old_chunk=next_chunk;
-			} else {
-				// the last source chunk
-				// taking only those rows of chunk that _left_to_copy
-				memcpy(new_rows, old_chunk->rows, sizeof(Chunk::Row)*rows_left_to_copy);
-				break;
-			}
-		}
-	}
-	link_row->link=0;
-	fused_rows=src_used_rows;
-	fsize=src.fsize;
-}
-
-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;
-}
+double pa_atod(const char* str, const String* problem_source) {
+	if(!str)
+		return 0;
+
+	while(*str && isspace((unsigned char)*str))
+		str++;
+	if(!*str)
+		return 0;
+
+	double result;
+	char *error_pos;
+	bool negative=false;
+	if(str[0]=='-') {
+		negative=true;
+		str++;
+	} else if(str[0]=='+') {
+		str++;
+	}
+	// 0xABC
+	if(str[0]=='0')
+		if(str[1]=='x' || str[1]=='X')
+			result=(double)(unsigned long)strtol(str, &error_pos, 0);
+		else {
+			 // skip leading 0000, to disable octal interpretation
+			do str++; while(*str=='0');				
+			result=(double)strtod(str, &error_pos);
+		}
+	else
+		result=(double)strtod(str, &error_pos);
+	if(negative)
+		result=-result;
+
+	while(char c=*error_pos++)
+		if(!isspace((unsigned char)c))
+			throw Exception("number.format",
+				problem_source,
+				problem_source?"invalid number (double)": "'%s' is invalid number (double)", str);
 
-String& String::real_append(STRING_APPEND_PARAMS) {
-	if(!src)
-		return *this;
-	if(!size)
-		size=strlen(src);
-	if(!size)
-		return *this;
+	return result;
+}
 
-	if(chunk_is_full())
-		expand();
+// cord lib extension
 
-	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;
+#ifndef DOXYGEN
+typedef struct {
+    ssize_t countdown;
+    char target;	/* Character we're looking for	*/
+} chr_data;
 #endif
-	append_here++; fused_rows++;
-
-	return *this;
+static int CORD_range_contains_chr_greater_then_proc(char c, size_t size, void* client_data)
+{
+    register chr_data * d = (chr_data *)client_data;
+    
+    if (d -> countdown<=0) return(2);
+    d -> countdown -= size;
+    if (c > d -> target) return(1);
+    return(0);
 }
+int CORD_range_contains_chr_greater_then(CORD x, size_t i, size_t n, int c)
+{
+    chr_data d;
 
-uint String::hash_code() const {
-	uint result=0;
+    d.countdown = n;
+    d.target = c;
+    return(CORD_block_iter(x, i, CORD_range_contains_chr_greater_then_proc, &d) == 1/*alternatives: 0 normally ended, 2=struck 'n'*/);
+}
 
-	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:
-	return result;
+static int CORD_block_count_proc(char /*c*/, size_t /*size*/, void* client_data)
+{
+    int* result=(int*)client_data;
+    (*result)++;
+    return(0); // 0=continue
+}
+size_t CORD_block_count(CORD x)
+{
+	size_t result=0;
+	CORD_block_iter(x, 0, CORD_block_count_proc, &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;
+// helpers
 
-		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--; 
-		}
+/// String::match uses this as replace & global search table columns
+
+const int MAX_MATCH_GROUPS=100;
 
-		if(!a_countdown) {
-			a_chunk=a_row->link;
-			a_row=a_chunk->rows;
-			a_countdown=a_chunk->count;
+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++) {
+			*this+=new String(String::Body::Format(1+i), String::L_CLEAN);
 		}
 	}
-	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;
-	}
+};
+
+Table string_match_table_template(new String_match_table_template_columns);
+
+// String::Body methods
+
+String::Body String::Body::Format(int value) {
+	char local[MAX_NUMBER];
+	size_t length=snprintf(local, MAX_NUMBER, "%d", value);
+	return String::Body(pa_strdup(local, length), length);
 }
 
-/// @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;
+String::Body String::Body::trim(String::Trim_kind kind, const char* chars, 
+								size_t* out_start, size_t* out_length) const {
+	size_t our_length=length();
+	if(!our_length)
+		return *this;
+	if(!chars)
+		chars=" \t\n"; // white space
 
-		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--; 
+	size_t start=0;
+	size_t end=our_length;
+	// from left...
+	if(kind!=TRIM_END) {
+		CORD_pos pos; set_pos(pos, 0);
+		while(true) {
+			char c=CORD_pos_fetch(pos);
+			if(strchr(chars, c)) {
+				if(++start==our_length)
+					return 0; // all chars are empty, just return empty string
+			} else
+				break;			
+
+			CORD_next(pos);
 		}
+	}
+	// from right..
+	if(kind!=TRIM_START) {
+		CORD_pos pos; set_pos(pos, end-1);
+		while(true) {
+			char c=CORD_pos_fetch(pos);
+			if(strchr(chars, c)) {
+				if(--end==0) // optimization: NO need to check for 'end>=start', that's(<) impossible
+					return 0; // all chars are empty, just return empty string
+			} else
+				break;			
 
-		if(!a_countdown) {
-			a_chunk=a_row->link;
-			a_row=a_chunk->rows;
-			a_countdown=a_chunk->count;
+			CORD_prev(pos);
 		}
 	}
-	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;
-	}
+
+	if(start==0 && end==our_length) // nobody moved a thing
+		return *this;
+
+	if(out_start)
+		*out_start=start;
+	size_t new_length=end-start;
+	if(out_length)
+		*out_length=new_length;
+
+	return mid(start, new_length);
+}
+
+static int CORD_batched_iter_fn_generic_hash_code(char c, void * client_data) {
+	uint& result=*static_cast<uint*>(client_data);
+	generic_hash_code(result, c);
+	return 0;
+}
+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::Body::hash_code() const {
+	uint result=0;
+	CORD_iter5(body, 0,
+		CORD_batched_iter_fn_generic_hash_code, 
+		CORD_batched_iter_fn_generic_hash_code, &result);
+	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; 
+// String methods
+
+String::String(const char* cstr, size_t helper_length, bool tainted): body(CORD_EMPTY) {
+	append_help_length(cstr, helper_length, tainted?L_TAINTED:L_CLEAN);
+}
+String::String(const String::C cstr, bool tainted): body(CORD_EMPTY) {
+	append_know_length(cstr.str, cstr.length, tainted?L_TAINTED:L_CLEAN);
 }
-#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 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());
+String& String::append_know_length(const char* str, size_t known_length, Language lang) {
+	if(!known_length)
+		return *this;
+
+	// first: langs
+	langs.append(body, lang, known_length);
+	// next: letters themselves
+	body.append_know_length(str, known_length);
+
+	ASSERT_STRING_INVARIANT(*this);
+	return *this;
+}
+String& String::append_help_length(const char* str, size_t helper_length, Language lang) {
+	if(!str)
+		return *this;
+	size_t known_length=helper_length?helper_length:strlen(str);
+	if(!known_length)
+		return *this;
+
+	return append_know_length(str, known_length, lang);
+}
+String& String::append_strdup(const char* str, size_t helper_length, Language lang) {
+	size_t known_length=helper_length?helper_length:strlen(str);
+	if(!known_length)
+		return *this;
+
+	// first: langs
+	langs.append(body, lang, known_length);
+	// next: letters themselves
+	body.append_strdup_know_length(str, known_length);
+
+	ASSERT_STRING_INVARIANT(*this);
+	return *this;
+}
+
+/// @todo check in doc: whether it documents NOW bad situation "abc".mid(-1, 3) =were?="ab"
+String& String::mid(size_t substr_begin, size_t substr_end) const {
+	String& result=*new String;
+
+	size_t self_length=length();
+	substr_begin=min(substr_begin, self_length);
+	substr_end=min(max(substr_end, substr_begin), self_length);
+	size_t substr_length=substr_end-substr_begin;
+	if(!substr_length)
+		return result;
+
+	// first: their langs
+	result.langs.append(result.body, langs, substr_begin, substr_length);
+	// next: letters themselves
+	result.body=body.mid(substr_begin, substr_length);
+
+	ASSERT_STRING_INVARIANT(result);
 	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;
+size_t String::pos(const String::Body substr, size_t this_offset, Language lang) const {
+	size_t substr_length=substr.length();
+	while(true) {
+		size_t substr_begin=body.pos(substr, this_offset);
+		
+		if(substr_begin==CORD_NOT_FOUND)
+			return STRING_NOT_FOUND;
+
+		if(langs.check_lang(lang, substr_begin, substr_length))
+			return substr_begin;
+
+		this_offset=substr_begin+substr_length;
 	}
-	
-	return -1;
 }
 
-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, Language 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;
-		int pos_before;
+void String::split(ArrayString& result, 
+		   size_t& pos_after, 
+		   const char* delim, 
+		   Language lang, int limit) const {
+	size_t self_length=length();
+	if(size_t delim_length=strlen(delim)) {
+		size_t 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))!=STRING_NOT_FOUND && limit; limit--) {
 			result+=&mid(pos_after, pos_before);
-			pos_after=pos_before+delim_size;
+			pos_after=pos_before+delim_length;
 		}
 		// last piece
-		if(pos_after<size() && limit) {
-			result+=&mid(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, 
-				   const String& delim, Untaint_lang lang, 
-				   int limit) const {
-	if(delim.size()) {
-		size_t pos_after=pos_after_ref?*pos_after_ref:0;
-		int pos_before;
+void String::split(ArrayString& result, 
+		   size_t& pos_after, 
+		   const String& delim, Language lang, 
+		   int limit) const {
+	if(!delim.is_empty()) {
+		size_t pos_before;
 		// while we have 'delim'...
-		for(; (pos_before=pos(delim, pos_after, lang))>=0 && limit; limit--) {
+		for(; (pos_before=pos(delim, pos_after, lang))!=STRING_NOT_FOUND && limit; limit--) {
 			result+=&mid(pos_after, pos_before);
-			pos_after=pos_before+delim.size();
+			pos_after=pos_before+delim.length();
 		}
 		// last piece
-		if(pos_after<size() && limit) {
-			result+=&mid(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();
 	}
 }
 
-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, 0}, // a=A
+		{"s", "S", 0, PCRE_DOTALL, result, 0}, // \n\n$ [default]
+		{"x", "U", 0, PCRE_EXTENDED, result, 0}, // whitespace in regex ignored
+		{"m", "M", PCRE_DOTALL, PCRE_MULTILINE, result, 0}, // ^aaa\n$^bbb\n$
+		{"g", "G", 0, 1, result+1, 0}, // many rows
+		{"'", 0, 0, 0, 0, &need_pre_post_match},
+		{0, 0, 0, 0, 0, 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;
+    if(options && !options->is_empty()) 
+		for(Regex_option *o=regex_option; o->keyL; o++) 
+			if(options->pos(o->keyL)!=STRING_NOT_FOUND
+				|| (o->keyU && options->pos(o->keyU)!=STRING_NOT_FOUND)) {
+				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.is_empty())
+		throw Exception(0,
+			0,
 			"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]={0};  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,
-			&regexp.mid(erroffset, regexp.size()),
+		throw Exception(0,
+			&regexp.mid(erroffset, regexp.length()),
 			"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);
-	}
+	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::Action_options table_options;
+	Table& table=*new Table(string_match_table_template, table_options);
 
 	int exec_option_bits=0;
+	int prestart=0;
+	int poststart=0;
+	int postfinish=length();
 	while(true) {
 		int exec_substrings=pcre_exec(code, 0,
-			subject, length, startoffset,
-			exec_option_bits, ovector, ovecsize);
+			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 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(0,
+				&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+=&Empty; // .prematch column value
+			*row+=&Empty; // .match
+			*row+=&Empty; // .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 +499,245 @@ bool String::match(const unsigned char *
 */
 	}
 }
+
+String& String::change_case(Charset& source_charset, Change_case_kind kind) const {
+	String& result=*new String();
+	if(is_empty())
+		return result;
+
+	char* new_cstr=cstrm();
+	size_t new_cstr_len=length();
+	if(source_charset.isUTF8()) {
+		switch(kind) {
+		case CC_UPPER:
+			change_case_UTF8((const XMLByte*)new_cstr, new_cstr_len, (XMLByte*)new_cstr, new_cstr_len, UTF8CaseToUpper);
+			break;
+		case CC_LOWER:
+			change_case_UTF8((const XMLByte*)new_cstr, new_cstr_len, (XMLByte*)new_cstr, new_cstr_len, UTF8CaseToLower);
+			break;
+		default:
+			assert(!"unknown change case kind");
+			break; // never
+		}	
+		
+	} else {
+		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:
+			assert(!"unknown change case kind");
+			a=b=0; // calm, compiler
+			break; // never
+		}	
+
+		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;
+		}
+	}
+	result.langs=langs;
+	result.body=new_cstr;
+
+	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;
+
+	const char* current=old_cstr;
+	while(*current) {
+		if(Dictionary::Subst subst=dict.first_that_begins(current)) {
+			// prematch
+			if(size_t prematch_length=current-prematch_begin) {
+				result.langs.append(result.body, langs, prematch_begin-old_cstr, prematch_length);
+				result.body.append_strdup_know_length(prematch_begin, prematch_length);
+			}
+
+			// match
+			// skip 'a' in 'current'; move prematch_begin
+			current+=subst.from_length; prematch_begin=current;
+
+			if(const String* b=subst.to) // are there any b?
+				result<<*b;
+		} else // simply advance
+			current++; 
+	}
+
+	// postmatch
+	if(size_t postmatch_length=current-prematch_begin) {
+		result.langs.append(result.body, langs, prematch_begin-old_cstr, postmatch_length);
+		result.body.append_strdup_know_length(prematch_begin, postmatch_length);
+	}
+
+	ASSERT_STRING_INVARIANT(result);
+	return result;
+}
+
+static int serialize_body_char(char c, char** cur) {
+	*((*cur)++)=c;
+	return 0; // 0=continue
+};
+static int serialize_body_piece(const char* s, char** cur) {
+	size_t length=strlen(s);
+	memcpy(*cur, s, length);  *cur+=length;
+	return 0; // 0=continue
+};
+static int serialize_lang_piece(char alang, size_t asize, char** cur) {
+	// lang
+	**cur=alang; (*cur)++;
+	// length [WARNING: not cast, addresses must be %4=0 on sparc]
+	memcpy(*cur, &asize, sizeof(asize));  *cur+=sizeof(asize);
+
+	return 0; // 0=continue
+}
+String::Cm String::serialize(size_t prolog_length) const {
+	size_t fragments_count=langs.count();
+	size_t buf_length=
+		prolog_length //1
+		+sizeof(size_t) //2
+		+fragments_count*(sizeof(char)+sizeof(size_t)) //3
+		+body.length() //4
+		+1; // for zero terminator used in deserialize
+	String::Cm result(new(PointerFreeGC) char[buf_length], buf_length);
+
+	// 1: prolog
+	char *cur=result.str+prolog_length;
+	// 2: langs.count [WARNING: not cast, addresses must be %4=0 on sparc]
+	memcpy(cur, &fragments_count, sizeof(fragments_count));  cur+=sizeof(fragments_count);
+	// 3: lang info
+	langs.for_each(body, serialize_lang_piece, &cur);
+	// 4: letters
+	body.for_each(serialize_body_char, serialize_body_piece, &cur);
+	// 5: zero terminator
+	*cur=0;
+
+	return result;
+}
+bool String::deserialize(size_t prolog_length, void *buf, size_t buf_length) {
+	if(buf_length<=prolog_length)
+		return false;
+	buf_length-=prolog_length;
+	buf_length-=1; // 5: zero terminator
+
+	// 1: prolog
+	const char* cur=(const char* )buf+prolog_length;
+
+	// 2: langs.count
+	size_t fragments_count;
+	if(buf_length<sizeof(fragments_count)) // langs.count don't fit?
+		return false;
+	// [WARNING: not cast, addresses must be %4=0 on sparc]
+	memcpy(&fragments_count, cur, sizeof(fragments_count));  cur+=sizeof(fragments_count);
+	buf_length-=sizeof(fragments_count);
+	
+	if(fragments_count) {
+		// 3: lang info
+		size_t total_length=0;
+		for(size_t f=0; f<fragments_count; f++) {
+			char lang;
+			size_t fragment_length;
+			size_t piece_length=sizeof(lang)+sizeof(fragment_length);
+			if(buf_length<piece_length) // lang+length
+				return false;
+
+			// lang
+			lang=*cur++;
+			// length [WARNING: not cast, addresses must be %4=0 on sparc]
+			memcpy(&fragment_length, cur, sizeof(fragment_length));  cur+=sizeof(fragment_length);
+
+			// uchar needed to prevent propagating 0x80 bit to upper bytes
+			langs.append(total_length, (String::Language)(uchar)lang, fragment_length);
+			total_length+=fragment_length;
+
+			buf_length-=piece_length;
+		}
+
+		// 4: letters
+		if(buf_length!=total_length)
+			return false;
+
+		// serialize wrote extra zero byte there, we can rely on that
+		body=String::Body(cur, buf_length);
+	}
+
+	ASSERT_STRING_INVARIANT(*this);
+	return true;
+}
+
+const char* String::Body::v() const {
+	return CORD_to_const_char_star(body);
+}
+void String::Body::dump() const {
+	CORD_dump(body);
+}
+
+const char* String::Languages::v() const {
+	if(opt.is_not_just_lang)
+		return CORD_to_const_char_star(langs);
+	else
+		return (const char*)&langs;
+}
+void String::Languages::dump() const {
+	if(opt.is_not_just_lang)
+		CORD_dump(langs);
+	else
+		puts((const char*)&langs);
+}
+const char* String::v() const {
+	const uint LIMIT_VIEW=20;
+	char* buf=(char*)malloc(MAX_STRING);
+	const char*body_view=body.v();
+	const char*langs_view=langs.v();
+	snprintf(buf, MAX_STRING, 
+		"%d:%.*s%s}   "
+		"{%d:%s",
+		langs.count(), LIMIT_VIEW, langs_view, strlen(langs_view)>LIMIT_VIEW?"...":"",
+		strlen(body_view), body_view
+	);
+
+	return buf;
+}
+void String::dump() const {
+	body.dump();
+	langs.dump();
+}
+const String& String::trim(String::Trim_kind kind, const char* chars) const {
+	if(!length())
+		return *this;
+
+	size_t substr_begin, substr_length;
+	Body new_body=body.trim(kind, chars, &substr_begin, &substr_length);
+	if(new_body==body) // we received unchanged pointer, do likewise
+		return *this;
+	// new_body differs from body, adjust langs along
+
+	String& result=*new String;
+	if(!new_body) // body.trim produced empty result
+		return result;
+	// body.trim produced nonempty result
+
+	// first: their langs
+	result.langs.append(result.body, langs, substr_begin, substr_length);
+	// next: letters themselves
+	result.body=new_body;
+
+	ASSERT_STRING_INVARIANT(result);
+	return result;
+}