Return to pa_string.C CVS log

Up to [parser3project] / parser3 / src / main

string reimplementation with cord+array<fragment>: 70%

/** @file
	Parser: string class. @see untasize_t.C.

	Copyright (c) 2001-2003 ArtLebedev Group (http://www.artlebedev.com)
	Author: Alexandr Petrosian <paf@design.ru> (http://paf.design.ru)
*/

static const char* IDENT_STRING_C="$Date: 2003/03/19 12:22:36 $";

#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++) {
			char *cname=new char[3/*strlen("100")*/+1/*terminating 0*/];
			*this+=new String(cname, sprintf(cname, "%d", 1+i));
		}
	}
};

Table string_match_table_template(new String_match_table_template_columns);

// String::ArrayFragment methods

String::ArrayFragment& String::ArrayFragment::append(const ArrayFragment& src, 
						     size_t substr_start, size_t substr_end) {
	if(substr_start==substr_end)
		return *this;

	size_t fragment_start=0;
	size_t fragment_end;
	for(Array_iterator<element_type> i(src); ; fragment_start=fragment_end) {
		const element_type& fragment=i.next();
		fragment_end=fragment_start+fragment.size;		

		if(substr_start<fragment_start) // not reached fragments which may include 'substr'?
			continue;

		// found first fragment including piece of 'substr'
		size_t piece_end=min(substr_end, fragment_end);
		*this+=Fragment(fragment.lang, piece_end-substr_start);
		
		while(substr_end>fragment_end) { // are there more fragments including pieces of 'substr'?
			const element_type& fragment=i.next();
			fragment_end=fragment_start+fragment.size;

			if(substr_end>fragment_end) // are there still more?
				simple_append(Fragment(fragment.lang, fragment.size)); // appending whole fragment
			else { // no, it was last
				simple_append(Fragment(fragment.lang, substr_end-fragment_start));
				goto break2;
			}
		}

		break;
	}	
break2:
	
	return *this;
}
/*
void String::ArrayFragment::mid(ArrayFragment& result, size_t substr_start, size_t substr_end) {
}
*/

// String methods

String::String(const char* src, size_t src_size, bool tainted): body(CORD_EMPTY) {
	if(src)
		if(tainted)
			APPEND_TAINTED(src, src_size, 0, 0);
		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)
		return *this;
	if(!size)
		size=strlen(src);
	if(!size)
		return *this;

	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;
	CORD_iter5(body, 0, 0, CORD_batched_iter_fn_generic_hash_code, &result);
	return result;
}

const String& String::mid(size_t substr_start, size_t substr_end) const {
	String& result=*new String;

	substr_start=min(substr_start, size());
	substr_end=max(substr_start, substr_end);
	if(substr_start==substr_end)
		return result;

	// first: letters themselves
	result.body=CORD_substr(body, substr_start, substr_end-substr_start);

	// next: their langs
	result.fragments.append(fragments, substr_start, substr_end);

//	SAPI::log(pool(), "piece of '%s' from %d to %d is '%s'",
		//cstr(), substr_start, substr_end, result.cstr());
	return result;
}

size_t String::pos(CORD substr, 
		   size_t this_offset, Untaint_lang lang) const {
	// first: letters themselves
	size_t substr_start=CORD_str(body, this_offset, substr);
	if(substr_start==CORD_NOT_FOUND)
		return NOT_FOUND;

	// next: check the lang when specified

	if(lang==UL_UNSPECIFIED) // ignore lang?
		return substr_start;

	// substr must be in one fragment, and fragments' lang must = lang
	size_t substr_end=substr_start+CORD_len(substr);
	size_t fragment_start=0;
	size_t fragment_end;
	for(Array_iterator<ArrayFragment::element_type> i(fragments); i.has_next(); fragment_start=fragment_end) {
		const Fragment& fragment=i.next();
		fragment_end=fragment_start+fragment.size;

		if(substr_start<fragment_start) // not reached fragments which may include 'result'?
			continue;
		
		if(substr_end>fragment_end) // end of substr OUT of current fragment?
			break;

		return fragment.lang==lang?substr_end:NOT_FOUND;
	}	

	return NOT_FOUND;
}

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(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, pos_after, lang))>=0 && limit; limit--) {
			result+=&mid(pos_after, pos_before);
			pos_after=pos_before+delim_size;
		}
		// last piece
		if(pos_after<self_size && limit) {
			result+=&mid(pos_after, self_size);
			pos_after=self_size;
		}
	} else { // empty delim
		result+=this;
		pos_after+=self_size;
	}
}

void String::split(ArrayString& result, 
				   size_t& pos_after, 
				   const String& delim, Untaint_lang lang, 
				   int limit) const {
	if(delim) {
		int pos_before;
		// while we have 'delim'...
		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();
		}
	} else { // empty delim
		result+=this;
		pos_after+=size();
	}
}

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(Exception::undefined_type,
			Exception::undefined_source,
			"regexp is empty");

	const char* pattern=regexp.cstr();
	const char* errptr;
	int erroffset;
	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,
		source_charset.pcre_tables);

	if(!code)
		throw Exception(Exception::undefined_type,
			&regexp.mid(erroffset, regexp.size()),
			"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);
	}

	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, subject.size, prestart,
			exec_option_bits, ovector, ovecsize);
		
		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(Exception::undefined_type,
				&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) {
	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;
}