Return to pa_string.C CVS log

Up to [parser3project] / parser3 / src / main

uchar changed to String_UL in all places [was not everywhere]
untaint test: 032

/** @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: 2003/03/03 13:44:26 $";

#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 {
	Pool pool;
public:
	String_match_table_template_columns() {
		*this+=StringPtr(new String("prematch"));
		*this+=StringPtr(new String("match"));
		*this+=StringPtr(new String("postmatch"));
		for(int i=0; i<MAX_MATCH_GROUPS; i++) {
			char *cname=new(pool) char[3/*strlen("100")*/+1/*terminating 0*/];
			StringPtr sname(new String(cname, sprintf(cname, "%d", 1+i)));
			*this+=sname;
		}
	}
};

Table string_match_table_template(
								  Exception::undefined_source, 
								  Table::columns_type(new String_match_table_template_columns));

// methods

String::String(const char* src, size_t src_size, bool tainted): Array<String_fragment>(1), fsize(0) {
	if(src)
		if(tainted)
			APPEND_TAINTED(src, src_size, 0, 0);
		else
			APPEND_CLEAN(src, src_size, 0, 0);
}

String::String(const String& src): Array<String_fragment>(src.count()) {
	append(src, UL_PASS_APPENDED);
}

String& String::real_append(STRING_APPEND_PARAMS) {
	if(!src)
		return *this;
	if(!size)
		size=strlen(src);
	if(!size)
		return *this;

	if(is_full())
		expand(fdelta);

	String_fragment *fragment=&felements[fused++];
	fragment->ptr=src;
	fragment->size=size;
	fragment->lang=lang;
#ifndef NO_STRING_ORIGIN
	fragment->origin.file=file;
	fragment->origin.line=line;
#endif

	fsize+=size;

	return *this;
}

uint String::hash_code() const {
	uint result=0;
	STRING_FOREACH_FRAGMENT(
		result=generic_hash_code(result, fragment->ptr, fragment->size);
	);
	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;
	size_t a_size=size();
	this_offset=min(this_offset, a_size-1);

	const String_fragment *a_current=felements;
	const String_fragment *b_current=src.felements;
	size_t a_offset=this_offset;
	size_t b_offset=0;
	String_fragment *a_end=felements+fused;
	String_fragment *b_end=src.felements+src.fused;
	int result;
	size_t pos=0; 

	bool a_break;
	bool b_break;
	while(true) {
		a_break=(a_current==a_end);
		b_break=(b_current==b_end);
		if(a_break || b_break)
			break;

		if(pos+a_current->size > this_offset) {
			if(lang!=UL_UNSPECIFIED && a_current->lang>(String_UL)lang) 
				return -1; // wrong lang -- bail out

			int size_diff=
				(a_current->size-a_offset)-
				(b_current->size-b_offset);
			
			if(size_diff==0) { // a has same size as b
				result=memcmp(a_current->ptr+a_offset, b_current->ptr+b_offset, 
					a_current->size-a_offset);
				if(result)
					return result;
				pos+=a_current->size;
				a_current++; a_offset=0;
				b_current++; b_offset=0;
			} else if(size_diff>0) { // a longer
				result=memcmp(a_current->ptr+a_offset, b_current->ptr+b_offset, 
					b_current->size-b_offset);
				if(result)
					return result;
				a_offset+=b_current->size-b_offset;
				b_current++; b_offset=0;
			} else { // b longer
				result=memcmp(a_current->ptr+a_offset, b_current->ptr+b_offset, 
					a_current->size-a_offset);
				if(result)
					return result;
				b_offset+=a_current->size-a_offset;
				pos+=a_current->size;
				a_current++; a_offset=0;
			}
		} else {
			a_offset-=a_current->size;
			pos+=a_current->size;
			a_current++;
		}
	}
	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;
	}
}

/// @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 a_size=size();
	size_t b_size=src_size?src_size:b_ptr?strlen(b_ptr):0;
	this_offset=min(this_offset, a_size-1);

	const String_fragment *a_current=felements;
	size_t a_offset=this_offset;
	size_t b_offset=0;
	String_fragment *a_end=felements+fused;
	size_t pos=0;

	bool a_break;
	bool b_break=b_size==0;
	while(true) {
		a_break=(a_current==a_end);
		if(a_break || b_break)
			break;
		if(pos+a_current->size > this_offset) {
			if(lang!=UL_UNSPECIFIED && a_current->lang>(String_UL)lang) 
				return -1; // wrong lang -- bail out

			int size_diff=
				(a_current->size-a_offset)-
				(b_size-b_offset);
			
			if(size_diff==0) { // a has same size as b
				if(int result=memcmp(a_current->ptr+a_offset, b_ptr+b_offset, 
					a_current->size-a_offset)!=0)
					return result;
				pos+=a_current->size;
				a_current++; a_offset=0;
				b_break=true;
			} else if (size_diff>0) { // a longer
				if(int result=memcmp(a_current->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(int result=memcmp(a_current->ptr+a_offset, b_ptr+b_offset, 
					a_current->size-a_offset)!=0)
					return result;
				b_offset+=a_current->size-a_offset;
				pos+=a_current->size;
				a_current++; a_offset=0;
			}
		} else {
			a_offset-=a_current->size; 
			pos+=a_current->size;
			a_current++; 
		}
	}
	if(a_break==b_break) { // ended simultaneously
		partial=0; return 0;
	} else if(a_break) { // first bytes equal, but a ended before b
		partial=1; return -1;
	} else {
		partial=2; return +1;
	}
}

#ifndef NO_STRING_ORIGIN
const String_fragment::Origin& String::origin() const { 
	if(is_empty()) {
		static const String_fragment::Origin empty_origin={"empty string"};
		return empty_origin;
	}
	
	// determining origin by first piece or last appended piece
	// because any of them can be constant=without origin: 
	// ex: ^load[/file] "document_root" + "/file"
	// when last peice is constant, 
	// ex: parser_root_auto_path{dynamic} / auto.p{const}
	// using first piece
	String_fragment::Origin& first_origin=felements[0].origin;
	return first_origin.file ? first_origin : felements[fused-1].origin;
}
#endif

StringPtr String::mid(size_t start, size_t finish) const {
	StringPtr result(new String());

	start=min(start, size());
	finish=max(start, finish);
	if(start==finish)
		return result;

	size_t pos=0;
	STRING_FOREACH_FRAGMENT(
		size_t item_finish=pos+fragment->size;
		if(item_finish > start) { // started now or already?
			bool started=result->is_empty(); // started now?
			bool finished=finish <= item_finish; // finished now?
			size_t offset=started?start-pos:0;
			size_t size=finished?finish-pos:fragment->size;
			result->APPEND(
				fragment->ptr+offset, size-offset, 
				fragment->lang,
				fragment->origin.file, fragment->origin.line);
			if(finished)
				goto break2;
		}
		pos+=fragment->size;
	);
break2:
//	SAPI::log(pool(), "piece of '%s' from %d to %d is '%s'",
		//cstr(), start, finish, result.cstr());
	return result;
}

int String::pos(const String& substr, 
				int result, Untaint_lang lang) const {
	size_t self_size=size();
	for(; size_t(result)<self_size; result++) {
		int partial; cmp(partial, substr, result, lang);
		if(
			partial==0 || // full match
			partial==2) // 'substr' starts 'this'+'result'
			return result;
	}
	
	return -1;
}

int String::pos(const char* substr, size_t substr_size, 
				int result, Untaint_lang lang) const {
	size_t self_size=size();
	for(; size_t(result)<self_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;
}

void String::split(ArrayString& result, 
				   size_t* pos_after_ref, 
				   const char* delim, size_t delim_size, 
				   Untaint_lang lang, int limit) {
	size_t self_size=size();
	if(delim_size) {
		size_t pos_after=pos_after_ref?*pos_after_ref:0;
		int pos_before;
		// while we have 'delim'...
		for(; (pos_before=pos(delim, delim_size, 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;
		}
		if(pos_after_ref)
			*pos_after_ref=pos_after;
	} else { // empty delim
		result+=StringPtr(this);
		if(pos_after_ref)
			*pos_after_ref+=self_size;
	}
}

void String::split(ArrayString& result, 
				   size_t* pos_after_ref, 
				   const String& delim, Untaint_lang lang, 
				   int limit) {
	if(!delim.is_empty()) {
		size_t pos_after=pos_after_ref?*pos_after_ref:0;
		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();
		}
		if(pos_after_ref)
			*pos_after_ref=pos_after;
	} else { // empty delim
		result+=StringPtr(this);
		if(pos_after_ref)
			*pos_after_ref+=size();
	}
}

static void regex_options(StringPtr 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;
				}
			}
}

TablePtr String::match(Charset& source_charset,
					   StringPtr aorigin,		
					   const String& regexp, 
					   StringPtr options,
					   Row_action row_action, void *info,
					   bool *was_global) const { 

	if(regexp.is_empty())
		throw Exception(0,
			aorigin,
			"regexp is empty");

	CharPtr 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);
	if(was_global)
		*was_global=option_bits[1]!=0;
	pcre *code=pcre_compile(pattern.get(), option_bits[0], 
		&errptr, &erroffset,
		source_charset.pcre_tables);

	if(!code)
		throw Exception(0,
			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 Exception(0,
			aorigin,
			"pcre_info error (%d)", 
				info_substrings);
	}

	CharPtr subject=cstr();
	int length=size();
	const int ovecsize=(1/*match*/+MAX_MATCH_GROUPS)*3;
	int ovector[ovecsize];

	// create table
	TablePtr result(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.get(), length, prestart,
			exec_option_bits, ovector, ovecsize);
		
		if(exec_substrings==PCRE_ERROR_NOMATCH) {
			pcre_free(code);
			row_action(result, ArrayStringPtr(0)/*last time, no raw*/, 0, 0, poststart, postfinish, info);
			if(option_bits[1])
				return result; // global=true+result
			else
				return TablePtr(0);// not global=no result
		}

		if(exec_substrings<0) {
			pcre_free(code);
			throw Exception(0,
				aorigin,
				"regular expression execute error (%d)", 
					exec_substrings);
		}

		int prefinish=ovector[0];
		poststart=ovector[1];
		object_ptr<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+=StringPtr(0); // .prematch column value
			*row+=StringPtr(0); // .match
			*row+=StringPtr(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(result, row, prestart, prefinish, poststart, postfinish, info);

		if(!option_bits[1] || prestart==poststart) { // not global | going to hang
			pcre_free(code);
			row_action(result, ArrayStringPtr(0)/*last time, no row*/, 0, 0, poststart, postfinish, info);
			return result;
		}
		prestart=poststart;

/*
		if(option_bits[0] & PCRE_MULTILINE)
			exec_option_bits|=PCRE_NOTBOL; // start of subject+startoffset not BOL
*/
	}
}

StringPtr String::change_case(Pool& pool, Charset& source_charset, Change_case_kind kind) {
	StringPtr 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(0, 
			StringPtr(this), 
			"unknown change case kind #%d", 
				static_cast<int>(kind)); // never
		a=b=0; // calm, compiler
		break; // never
	}	

	STRING_FOREACH_FRAGMENT(
		char *new_cstr=new(pool) char[fragment->size];
		char *dest=new_cstr;
		const char* end=fragment->ptr+fragment->size;
		for(const char* current=fragment->ptr; current<end; current++) {
			unsigned char c=a[(unsigned char)*current];
			if(b)
				c=b[c];

			*dest++=(char)c;
		}
		
		result->APPEND(new_cstr, fragment->size, 
			fragment->lang,
			fragment->origin.file, fragment->origin.line);
	);

	return result;
}

/// @test if in some piece were found no dict words, append it, not it's duplicate
StringPtr String::replace(Pool& pool, Dictionary& dict) const {
	StringPtr result(new String());
	CharPtr lcstr(cstr());
	const char* current=lcstr.get();

	STRING_FOREACH_FRAGMENT(
IFNDEF_NO_STRING_ORIGIN(
		const char* joined_origin_file=fragment->origin.file;
		const size_t joined_origin_line=fragment->origin.line;
);
		String_UL joined_lang=fragment->lang;
		const char* joined_ptr=current;
		// calc size
		size_t joined_size=0;
		STRING_PREPARED_FOREACH_FRAGMENT(
			if(fragment->lang==joined_lang)
				joined_size+=fragment->size;
			else
				break; // before non-ours
		);
		current+=joined_size;

		// pointers are after joined piece
		// & one step back, see STRING_PREPARED_FOREACH_FRAGMENT
		--fragment;
		
		char *new_cstr=new(pool) char[(size_t)ceil(joined_size*dict.max_ratio())];
		char *dest=new_cstr;
		while(joined_size) {
			// there is a fragment where first column starts 'joined_ptr'
			if(Table::element_type row=dict.first_that_starts(joined_ptr, joined_size)) {
				// get a=>b values
				StringPtr a=row->get(0);
				StringPtr b=row->get(1);
				// skip 'a' in 'joined_ptr' && reduce work size
				joined_ptr+=a->size();  joined_size-=a->size();
				// write 'b' to 'dest' && skip 'b' in 'dest'
				b->store_to(dest, String::UL_AS_IS);  dest+=b->size();
			} else {
				// write a char to b && reduce work size
				*dest++=*joined_ptr++;  joined_size--;
			}
		}

		result->APPEND(new_cstr, dest-new_cstr, joined_lang,
			joined_origin_file, joined_origin_line);
	);

	return result;
}

StringPtr String::join_chains(Pool& pool) const {
	StringPtr result(new String());
	
	char *pooled_buf=new(pool) char[cstr_bufsize()];
	store_to(pooled_buf, String::UL_AS_IS);
	const char* current=pooled_buf;

	STRING_FOREACH_FRAGMENT(
IFNDEF_NO_STRING_ORIGIN(
		const char* joined_origin_file=fragment->origin.file;
		const size_t joined_origin_line=fragment->origin.line;
);
		String_UL joined_lang=fragment->lang;
		const char* joined_ptr=current;
		// calc size
		size_t joined_size=0;
		STRING_PREPARED_FOREACH_FRAGMENT(
			if(fragment->lang==joined_lang)
				joined_size+=fragment->size;
			else
				break; // before non-ours
		);
		current+=joined_size;

		// pointers are after joined piece
		// & one step back, see STRING_PREPARED_FOREACH_FRAGMENT
		--fragment;
		
		result->APPEND(joined_ptr, joined_size, joined_lang,
			joined_origin_file, joined_origin_line);
	);

	return result;
}

double String::as_double() { 
	double result;
	char buf[MAX_STRING];
	if(size()>MAX_STRING-1)
		throw Exception("number.format",
			StringPtr(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",
				StringPtr(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",
			StringPtr(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",
				StringPtr(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::serialize(Pool& pool, size_t prolog_size, char *& buf, size_t& buf_size) const {
	buf_size=
		prolog_size
		+fused*(sizeof(String_UL)+sizeof(size_t))
		+size();
	buf=new(pool) 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;
}