--- parser3/src/main/pa_string.C	2001/01/26 15:28:39	1.2
+++ parser3/src/main/pa_string.C	2003/07/24 11:31:24	1.174
@@ -1,77 +1,631 @@
-#include <string.h>
+/** @file
+	Parser: string class. @see untalength_t.C.
 
-#include "pa_pool.h"
+	Copyright (c) 2001-2003 ArtLebedev Group (http://www.artlebedev.com)
+	Author: Alexandr Petrosian <paf@design.ru> (http://paf.design.ru)
+*/
 
-void *String::operator new(size_t size, Pool *apool) {
-	return apool->alloc(size);
+static const char* IDENT_STRING_C="$Date: 2003/07/24 11:31:24 $";
+
+#include "pcre.h"
+
+#include "pa_string.h"
+#include "pa_exception.h"
+#include "pa_table.h"
+#include "pa_dictionary.h"
+#include "pa_charset.h"
+
+// helpers
+
+/// String::match uses this as replace & global search table columns
+
+const int MAX_MATCH_GROUPS=100;
+
+class String_match_table_template_columns: public ArrayString {
+public:
+	String_match_table_template_columns() {
+		*this+=new String("prematch");
+		*this+=new String("match");
+		*this+=new String("postmatch");
+		for(int i=0; i<MAX_MATCH_GROUPS; i++) {
+			*this+=new String(StringBody::Format(1+i), String::L_CLEAN);
+		}
+	}
+};
+
+Table string_match_table_template(new String_match_table_template_columns);
+
+// String::ArrayFragment methods
+
+void String::ArrayFragment::append_positions(const ArrayFragment& src, 
+					     size_t substr_begin, size_t substr_end) {
+	if(substr_begin==substr_end)
+		return;
+
+//	FILE *err=fopen("append.log", "wt");
+
+	size_t fragment_begin=0;
+	size_t fragment_end;
+	for(Array_iterator<element_type> i(src); ; fragment_begin=fragment_end) {
+		const Fragment fragment=i.next();
+		fragment_end=fragment_begin+fragment.length;		
+		//fprintf(err, "1end=%u\n", fragment_end);fflush(err);
+
+		// not reached fragments which may include 'substr'?
+		if(!(substr_begin>=fragment_begin && substr_begin<fragment_end))
+			continue;
+
+		// found first fragment including piece of 'substr'
+		if(substr_end<=fragment_end) // fits into first fragment?
+			*this+=Fragment(fragment.lang, substr_end-substr_begin);
+		else { // spans more then one fragment
+			*this+=Fragment(fragment.lang, fragment_end-substr_begin);
+			while(true) {
+				const Fragment fragment=i.next();
+				fragment_end=(fragment_begin=fragment_end)+fragment.length;
+				//fprintf(err, "2end=%u\n", fragment_end);fflush(err);
+
+				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));
+					//fclose(err);
+					return;
+				}
+			}
+		}
+
+		break;
+	}
+	//fclose(err);
+}
+
+// StringBody methods
+
+StringBody StringBody::Format(int value) {
+	char local[MAX_NUMBER];
+	size_t length=snprintf(local, MAX_NUMBER, "%d", value);
+	return StringBody(pa_strdup(local, length), 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 StringBody::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;
+}
+
+// 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);
 }
 
-void String::construct(Pool *apool) {
-	pool=apool;
-	head.count=curr_chunk_rows=CR_PREALLOCATED_COUNT;
-	append_here=head.first;
-	head.preallocated_link=0;
-	link_row=&head.first[curr_chunk_rows];
+String::String(const String& src): body(src.body) {
+	fragments.append(src.fragments);
+	ASSERT_STRING_INVARIANT(*this);
 }
 
-void String::expand() {
-	curr_chunk_rows=curr_chunk_rows*100/CR_GROW_PERCENT;
-	Chunk *chunk=static_cast<Chunk *>(
-		pool->calloc(sizeof(Chunk::Row)*curr_chunk_rows+sizeof(Chunk *)));
-	chunk->count=curr_chunk_rows;
-	link_row->link=chunk;
-	append_here=chunk->first;
-	link_row=&chunk->first[curr_chunk_rows];
+String& String::append_know_length(const char* str, size_t known_length, Language lang) {
+	if(!known_length)
+		return *this;
+
+	body.append_know_length(str, known_length);
+	fragments+=Fragment(lang, 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;
 
-String& String::operator += (char *src) {
-	if(chunk_is_full())
-		expand();
+	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;
 
-	append_here->item.ptr=src;
-	append_here->item.size=strlen(src);
-	append_here++;
+	body.append_strdup_know_length(str, known_length);
+	fragments+=Fragment(lang, known_length);
 
+	ASSERT_STRING_INVARIANT(*this);
 	return *this;
 }
 
-size_t String::size() {
-	int result=0;
-	Chunk *chunk=&head; 
-	do {
-		Chunk::Row *row=chunk->first;
-		for(int i=0; i<chunk->count; i++) {
-			if(row==append_here)
-				goto break2;
-
-			result+=row->item.size;
-			row++;
-		}
-		chunk=row->link;
-	} while(chunk);
-break2:
+/// @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);
+	if(substr_begin==substr_end)
+		return result;
+
+	// first: letters themselves
+	result.body=body.mid(substr_begin, substr_end-substr_begin);
+
+	// next: their langs
+	result.fragments.append_positions(fragments, substr_begin, substr_end);
+
+//	SAPI::log("piece of '%s' from %d to %d is '%s'",
+		//cstr(), substr_begin, substr_end, result.cstr());
+	ASSERT_STRING_INVARIANT(result);
+	return result;
+}
+
+size_t String::pos(const StringBody substr, 
+		   size_t this_offset, Language lang) const {
+	// first: letters themselves
+	size_t substr_begin=body.pos(substr, this_offset);
+	if(substr_begin==CORD_NOT_FOUND)
+		return STRING_NOT_FOUND;
+
+	// next: check the lang when specified
+
+	if(lang==L_UNSPECIFIED) // ignore lang?
+		return substr_begin;
+
+	// substr must be in one fragment, and fragments' lang must = lang
+	size_t substr_end=substr_begin+substr.length();
+	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",
+					substr.cstr(), this_offset);
+
+		if(fragment.lang<=lang)
+			return substr_begin;
+		else { // bad lang...
+			/// WARNING: this possibly skips assert (*), but it's fast
+			substr_begin=body.pos(substr, fragment_end/*...search AFTER for more*/);
+			if(substr_begin==CORD_NOT_FOUND)
+				return STRING_NOT_FOUND;
+
+			size_t substr_end=substr_begin+substr.length();
+			// and continuing with next fragment
+		}
+	}	
+
+	return STRING_NOT_FOUND;
+}
+
+size_t String::pos(const String& substr, 
+				size_t this_offset, Language lang) const {
+	return pos(substr.body, this_offset, lang);
+}
+
+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)) {
+		int pos_before;
+		// while we have 'delim'...
+		for(; (pos_before=pos(delim, pos_after, lang))!=STRING_NOT_FOUND && limit; limit--) {
+			result+=&mid(pos_after, pos_before);
+			pos_after=pos_before+delim_length;
+		}
+		// last piece
+		if(pos_after<self_length && limit) {
+			result+=&mid(pos_after, self_length);
+			pos_after=self_length;
+		}
+	} else { // empty delim
+		result+=this;
+		pos_after+=self_length;
+	}
+}
+
+void String::split(ArrayString& result, 
+		   size_t& pos_after, 
+		   const String& delim, Language lang, 
+		   int limit) const {
+	if(!delim.is_empty()) {
+		int pos_before;
+		// while we have 'delim'...
+		for(; (pos_before=pos(delim, pos_after, lang))!=STRING_NOT_FOUND && limit; limit--) {
+			result+=&mid(pos_after, pos_before);
+			pos_after=pos_before+delim.length();
+		}
+		// last piece
+		if(pos_after<length() && limit) {
+			result+=&mid(pos_after, length());
+			pos_after=length();
+		}
+	} else { // empty delim
+		result+=this;
+		pos_after+=length();
+	}
+}
+
+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 && !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;
+				}
+			}
+}
+
+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();
+	const char* errptr;
+	int erroffset;
+	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,
+		source_charset.pcre_tables);
+
+	if(!code)
+		throw Exception(0,
+			&regexp.mid(erroffset, regexp.length()),
+			"regular expression syntax error - %s", errptr);
+	
+	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::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, 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_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
+*/
+	}
+}
+
+String& String::change_case(Charset& source_charset, Change_case_kind kind) const {
+	String& result=*new String();
+	if(is_empty())
+		return result;
+
+	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;
+	}
+	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;
+
+	const char* current=old_cstr;
+	while(*current) {
+		if(Table::element_type row=dict.first_that_begins(current)) {
+			// prematch
+			if(size_t prematch_length=current-prematch_begin) {
+				result.body.append_strdup_know_length(prematch_begin, prematch_length);
+				result.fragments.append_positions(fragments, prematch_begin-old_cstr, current-old_cstr);
+			}
+
+			// 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);
+				result<<*b;
+			}
+		} else // simply advance
+			current++; 
+	}
+
+	// postmatch
+	if(size_t postmatch_length=current-prematch_begin) {
+		result.body.append_strdup_know_length(prematch_begin, postmatch_length);
+		result.fragments.append_positions(fragments, prematch_begin-old_cstr, current-old_cstr);
+	}
+
+	ASSERT_STRING_INVARIANT(result);
+	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;
 }
 
-char *String::c_str() {
-	char *result=static_cast<char *>(pool->alloc(size()+1));
-
-	char *copy_here=result;
-	Chunk *chunk=&head; 
-	do {
-		Chunk::Row *row=chunk->first;
-		for(int i=0; i<chunk->count; i++) {
-			if(row==append_here)
-				goto break2;
-
-			memcpy(copy_here, row->item.ptr, row->item.size);
-			copy_here+=row->item.size;
-			row++;
-		}
-		chunk=row->link;
-	} while(chunk);
-break2:
-	*copy_here=0;
+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 serialize_body_piece(const char* s, char** cur) {
+	size_t length=strlen(s);
+	memcpy(*cur, s, length);  *cur+=length;
+	return 0;
+};
+String::Cm String::serialize(size_t prolog_length) const {
+	size_t buf_length=
+		prolog_length
+		+fragments.count()*(sizeof(Language)+sizeof(size_t))
+		+length();
+	String::Cm 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
+	body.for_each(serialize_body_piece, &cur);
+
 	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;
+
+	// 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);
+	
+	if(fragments_count) {
+		// 3: lang info
+		size_t total_length=0;
+		for(size_t f=0; f<fragments_count; f++) {
+			size_t piece_length=sizeof(Language)+sizeof(size_t);
+			if(buf_length<piece_length) // lang+length
+				return false;
+
+			Language lang=*reinterpret_cast<const Language *>(cur);  cur+=sizeof(Language);
+			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=StringBody(cur, buf_length);
+	}
+
+	ASSERT_STRING_INVARIANT(*this);
+	return true;
+}