foxygit / GZDoom Log in
commits tags

/tools/re2c/src/parse/parser.ypp · 14.28 KB

raw
%{

#include "src/util/c99_stdint.h"
#include <assert.h>
#include <stddef.h>
#include <stdlib.h>
#include <limits>
#include <list>
#include <map>
#include <set>
#include <string>
#include <utility>
#include <vector>

#include "src/codegen/output.h"
#include "src/conf/opt.h"
#include "src/globals.h"
#include "src/ir/compile.h"
#include "src/ir/adfa/adfa.h"
#include "src/ir/regexp/encoding/enc.h"
#include "src/ir/regexp/encoding/range_suffix.h"
#include "src/ir/regexp/regexp.h"
#include "src/ir/regexp/regexp_cat.h"
#include "src/ir/regexp/regexp_close.h"
#include "src/ir/regexp/regexp_null.h"
#include "src/ir/regexp/regexp_rule.h"
#include "src/ir/rule_rank.h"
#include "src/ir/skeleton/skeleton.h"
#include "src/parse/code.h"
#include "src/parse/extop.h"
#include "src/parse/loc.h"
#include "src/parse/parser.h"
#include "src/parse/scanner.h"
#include "src/parse/spec.h"
#include "src/util/counter.h"
#include "src/util/free_list.h"
#include "src/util/range.h"
#include "src/util/smart_ptr.h"

#define YYMALLOC malloc
#define YYFREE free

using namespace re2c;

extern "C"
{
int yylex();
void yyerror(const char*);
}

static counter_t<rule_rank_t> rank_counter;
static std::vector<std::string> condnames;
static re2c::SpecMap  specMap;
static Spec spec;
static RuleOp *specNone = NULL;
static RuleOpList       specStar;
static RuleOp * star_default = NULL;
static Scanner          *in = NULL;
static Scanner::ParseMode  parseMode;
static SetupMap            ruleSetupMap;
static bool                foundRules;
static symbol_table_t symbol_table;

/* Bison version 1.875 emits a definition that is not working
 * with several g++ version. Hence we disable it here.
 */
#if defined(__GNUC__)
#define __attribute__(x)
#endif

void context_check(CondList *clist)
{
	if (!opts->cFlag)
	{
		delete clist;
		in->fatal("conditions are only allowed when using -c switch");
	}
}

void context_none(CondList *clist)
{
	delete clist;
	context_check(NULL);
	in->fatal("no expression specified");
}

void context_rule
	( CondList * clist
	, const Loc & loc
	, RegExp * expr
	, RegExp * look
	, const Code * code
	, const std::string * newcond
	)
{
	context_check(clist);
	for(CondList::const_iterator it = clist->begin(); it != clist->end(); ++it)
	{
		if (specMap.find(*it) == specMap.end())
		{
			condnames.push_back (*it);
		}

		RuleOp * rule = new RuleOp
			( loc
			, expr
			, look
			, rank_counter.next ()
			, code
			, newcond
			);
		specMap[*it].add (rule);
	}
	delete clist;
	delete newcond;
}

void setup_rule(CondList *clist, const Code * code)
{
	assert(clist);
	assert(code);
	context_check(clist);
	for(CondList::const_iterator it = clist->begin(); it != clist->end(); ++it)
	{
		if (ruleSetupMap.find(*it) != ruleSetupMap.end())
		{
			in->fatalf_at(code->loc.line, "code to setup rule '%s' is already defined", it->c_str());
		}
		ruleSetupMap[*it] = std::make_pair(code->loc.line, code->text);
	}
	delete clist;
}

void default_rule(CondList *clist, const Code * code)
{
	assert(clist);
	assert(code);
	context_check(clist);
	for(CondList::const_iterator it = clist->begin(); it != clist->end(); ++it)
	{
		RuleOp * def = new RuleOp
			( code->loc
			, in->mkDefault ()
			, new NullOp
			, rule_rank_t::def ()
			, code
			, NULL
			);
		if (!specMap[*it].add_def (def))
		{
			in->fatalf_at(code->loc.line, "code to default rule '%s' is already defined", it->c_str());
		}
	}
	delete clist;
}

%}

%start	spec

%union {
	re2c::RegExp * regexp;
	const re2c::Code * code;
	char op;
	re2c::ExtOp extop;
	std::string * str;
	re2c::CondList * clist;
};

%token TOKEN_CLOSE
%token TOKEN_CLOSESIZE
%token TOKEN_CODE
%token TOKEN_CONF
%token TOKEN_ID
%token TOKEN_FID
%token TOKEN_FID_END
%token TOKEN_NOCOND
%token TOKEN_REGEXP
%token TOKEN_SETUP
%token TOKEN_STAR

%type <op>      TOKEN_CLOSE TOKEN_STAR TOKEN_SETUP close
%type <extop>   TOKEN_CLOSESIZE
%type <code>    TOKEN_CODE
%type <regexp>  TOKEN_REGEXP rule look expr diff term factor primary
%type <str>     TOKEN_ID TOKEN_FID newcond
%type <clist>   cond clist

%%

spec:
		/* empty */
		{
		}
	|	spec rule
		{
			foundRules = true;
		}
	|	spec decl
;

decl:
		TOKEN_ID '=' expr ';'
		{
			if (!symbol_table.insert (std::make_pair (* $1, $3)).second)
			{
				in->fatal("sym already defined");
			}
			delete $1;
		}
	|	TOKEN_FID expr TOKEN_FID_END
		{
			if (!symbol_table.insert (std::make_pair (* $1, $2)).second)
			{
				in->fatal("sym already defined");
			}
			delete $1;
		}
	|	TOKEN_ID '=' expr '/'
		{
			in->fatal("trailing contexts are not allowed in named definitions");
		}
	|	TOKEN_FID expr '/'
		{
			in->fatal("trailing contexts are not allowed in named definitions");
		}
	|	TOKEN_CONF {}
;

rule:
		expr look TOKEN_CODE
		{
			if (opts->cFlag)
			{
				in->fatal("condition or '<*>' required when using -c switch");
			}
			RuleOp * rule = new RuleOp
				( $3->loc
				, $1
				, $2
				, rank_counter.next ()
				, $3
				, NULL
				);
			spec.add (rule);
		}
	|	TOKEN_STAR TOKEN_CODE /* default rule */
		{
			if (opts->cFlag)
				in->fatal("condition or '<*>' required when using -c switch");
			RuleOp * def = new RuleOp
				( $2->loc
				, in->mkDefault ()
				, new NullOp
				, rule_rank_t::def ()
				, $2
				, NULL
				);
			if (!spec.add_def (def))
			{
				in->fatal("code to default rule is already defined");
			}
		}
	|	'<' cond '>' expr look newcond TOKEN_CODE
		{
			context_rule ($2, $7->loc, $4, $5, $7, $6);
		}
	|	'<' cond '>' expr look ':' newcond
		{
			assert($7);
			Loc loc (in->get_fname (), in->get_cline ());
			context_rule ($2, loc, $4, $5, NULL, $7);
		}
	|	'<' cond '>' look newcond TOKEN_CODE
		{
			context_none($2);
			delete $5;
		}
	|	'<' cond '>' look ':' newcond
		{
			assert($6);
			context_none($2);
			delete $6;
		}
	|	'<' cond '>' TOKEN_STAR TOKEN_CODE /* default rule for conditions */
		{
			default_rule($2, $5);
		}
	|	'<' TOKEN_STAR '>' expr look newcond TOKEN_CODE
		{
			context_check(NULL);
			RuleOp * rule = new RuleOp
				( $7->loc
				, $4
				, $5
				, rank_counter.next ()
				, $7
				, $6
				);
			specStar.push_back (rule);
			delete $6;
		}
	|	'<' TOKEN_STAR '>' expr look ':' newcond
		{
			assert($7);
			context_check(NULL);
			Loc loc (in->get_fname (), in->get_cline ());
			RuleOp * rule = new RuleOp
				( loc
				, $4
				, $5
				, rank_counter.next ()
				, NULL
				, $7
				);
			specStar.push_back (rule);
			delete $7;
		}
	|	'<' TOKEN_STAR '>' look newcond TOKEN_CODE
		{
			context_none(NULL);
			delete $5;
		}
	|	'<' TOKEN_STAR '>' look ':' newcond
		{
			assert($6);
			context_none(NULL);
			delete $6;
		}
	|	'<' TOKEN_STAR '>' TOKEN_STAR TOKEN_CODE /* default rule for all conditions */
		{
			if (star_default)
			{
				in->fatal ("code to default rule '*' is already defined");
			}
			star_default = new RuleOp
				( $5->loc
				, in->mkDefault ()
				, new NullOp
				, rule_rank_t::def ()
				, $5
				, NULL
				);
		}
	|	TOKEN_NOCOND newcond TOKEN_CODE
		{
			context_check(NULL);
			if (specNone)
			{
				in->fatal("code to handle illegal condition already defined");
			}
			$$ = specNone = new RuleOp
				( $3->loc
				, new NullOp
				, new NullOp
				, rank_counter.next ()
				, $3
				, $2
				);
			delete $2;
		}
	|	TOKEN_NOCOND ':' newcond
		{
			assert($3);
			context_check(NULL);
			if (specNone)
			{
				in->fatal("code to handle illegal condition already defined");
			}
			Loc loc (in->get_fname (), in->get_cline ());
			$$ = specNone = new RuleOp
				( loc
				, new NullOp
				, new NullOp
				, rank_counter.next ()
				, NULL
				, $3
				);
			delete $3;
		}
	|	TOKEN_SETUP TOKEN_STAR '>' TOKEN_CODE
		{
			CondList *clist = new CondList();
			clist->insert("*");
			setup_rule(clist, $4);
		}
	|	TOKEN_SETUP cond '>' TOKEN_CODE
		{
			setup_rule($2, $4);
		}
;

cond:
		/* empty */
		{
			in->fatal("unnamed condition not supported");
		}
	|	clist
		{
			$$ = $1;
		}
	;

clist:
		TOKEN_ID
		{
			$$ = new CondList();
			$$->insert(* $1);
			delete $1;
		}
	|	clist ',' TOKEN_ID
		{
			$1->insert(* $3);
			delete $3;
			$$ = $1;
		}
;

newcond:
		/* empty */
		{
			$$ = NULL;
		}
	|	'=' '>' TOKEN_ID
		{
			$$ = $3;
		}
;

look:
		/* empty */
		{
			$$ = new NullOp;
		}
	|	'/' expr
		{
			$$ = $2;
		}
;

expr:
		diff
		{
			$$ = $1;
		}
	|	expr '|' diff
		{
			$$ = mkAlt($1, $3);
		}
;

diff:
		term
		{
			$$ = $1;
		}
	|	diff '\\' term
		{
			$$ = in->mkDiff($1, $3);
		}
;

term:
		factor
		{
			$$ = $1;
		}
	|	term factor
		{
			$$ = new CatOp($1, $2);
		}
;

factor:
		primary
		{
			$$ = $1;
		}
	|	primary close
		{
			switch($2)
			{
			case '*':
				$$ = new CloseOp($1);
				break;
			case '+':
				$$ = new CatOp (new CloseOp($1), $1);
				break;
			case '?':
				$$ = mkAlt($1, new NullOp());
				break;
			}
		}
	|	primary TOKEN_CLOSESIZE
		{
			if ($2.max == std::numeric_limits<uint32_t>::max())
			{
				$$ = repeat_from ($1, $2.min);
			}
			else if ($2.min == $2.max)
			{
				$$ = repeat ($1, $2.min);
			}
			else
			{
				$$ = repeat_from_to ($1, $2.min, $2.max);
			}
			$$ = $$ ? $$ : new NullOp;
		}
;

close:
		TOKEN_CLOSE
		{
			$$ = $1;
		}
	|	TOKEN_STAR
		{
			$$ = $1;
		}
	|	close TOKEN_CLOSE
		{
			$$ = ($1 == $2) ? $1 : '*';
		}
	|	close TOKEN_STAR
		{
			$$ = ($1 == $2) ? $1 : '*';
		}
;

primary:
		TOKEN_ID
		{
			symbol_table_t::iterator i = symbol_table.find (* $1);
			delete $1;
			if (i == symbol_table.end ())
			{
				in->fatal("can't find symbol");
			}
			$$ = i->second;
		}
	|	TOKEN_REGEXP
		{
			$$ = $1;
		}
	|	'(' expr ')'
		{
			$$ = $2;
		}
;

%%

extern "C" {
void yyerror(const char* s)
{
	in->fatal(s);
}

int yylex(){
	return in ? in->scan() : 0;
}
} // end extern "C"

namespace re2c
{

void parse(Scanner& i, Output & o)
{
	std::map<std::string, smart_ptr<DFA> >  dfa_map;
	ScannerState rules_state;

	in = &i;

	o.source.wversion_time ()
		.wline_info (in->get_cline (), in->get_fname ().c_str ());
	if (opts->target == opt_t::SKELETON)
	{
		Skeleton::emit_prolog (o.source);
	}

	Enc encodingOld = opts->encoding;
	
	while ((parseMode = i.echo()) != Scanner::Stop)
	{
		o.source.new_block ();
		bool bPrologBrace = false;
		ScannerState curr_state;

		i.save_state(curr_state);
		foundRules = false;

		if (opts->rFlag && parseMode == Scanner::Rules && dfa_map.size())
		{
			in->fatal("cannot have a second 'rules:re2c' block");
		}
		if (parseMode == Scanner::Reuse)
		{
			if (dfa_map.empty())
			{
				in->fatal("got 'use:re2c' without 'rules:re2c'");
			}
		}
		else if (parseMode == Scanner::Rules)
		{
			i.save_state(rules_state);
		}
		else
		{
			dfa_map.clear();
		}
		rank_counter.reset ();
		spec.clear ();
		in->set_in_parse(true);
		yyparse();
		in->set_in_parse(false);
		if (opts->rFlag && parseMode == Scanner::Reuse)
		{
			if (foundRules || opts->encoding != encodingOld)
			{
				// Re-parse rules
				parseMode = Scanner::Parse;
				i.restore_state(rules_state);
				i.reuse();
				dfa_map.clear();
				parse_cleanup();
				spec.clear ();
				rank_counter.reset ();
				in->set_in_parse(true);
				yyparse();
				in->set_in_parse(false);

				// Now append potential new rules
				i.restore_state(curr_state);
				parseMode = Scanner::Parse;
				in->set_in_parse(true);
				yyparse();
				in->set_in_parse(false);
			}
			encodingOld = opts->encoding;
		}
		o.source.set_block_line (in->get_cline ());
		uint32_t ind = opts->topIndent;
		if (opts->cFlag)
		{
			SpecMap::iterator it;
			SetupMap::const_iterator itRuleSetup;

			if (parseMode != Scanner::Reuse)
			{
				// <*> rules must have the lowest priority
				// now that all rules have been parsed, we can fix it
				for (RuleOpList::const_iterator itOp = specStar.begin(); itOp != specStar.end(); ++itOp)
				{
					(*itOp)->rank = rank_counter.next ();
				}
				// merge <*> rules to all conditions
				for (it = specMap.begin(); it != specMap.end(); ++it)
				{
					for (RuleOpList::const_iterator itOp = specStar.begin(); itOp != specStar.end(); ++itOp)
					{
						it->second.add (*itOp);
					}
					if (star_default)
					{
						it->second.add_def (star_default);
					}
				}

				if (specNone)
				{
					specMap["0"].add (specNone);
					// Note that "0" inserts first, which is important.
					condnames.insert (condnames.begin (), "0");
				}
				o.types = condnames;
			}

			size_t nCount = specMap.size();

			for (it = specMap.begin(); it != specMap.end(); ++it)
			{
				if (parseMode != Scanner::Reuse)
				{
					itRuleSetup = ruleSetupMap.find(it->first);				
					if (itRuleSetup != ruleSetupMap.end())
					{
						yySetupRule = itRuleSetup->second.second;
					}
					else
					{
						itRuleSetup = ruleSetupMap.find("*");
						if (itRuleSetup != ruleSetupMap.end())
						{
							yySetupRule = itRuleSetup->second.second;
						}
						else
						{
							yySetupRule = "";
						}
					}

					dfa_map[it->first] = compile(it->second, o, it->first, opts->encoding.nCodeUnits ());
				}
				if (parseMode != Scanner::Rules && dfa_map.find(it->first) != dfa_map.end())
				{
					dfa_map[it->first]->emit(o, ind, !--nCount, bPrologBrace);
				}
			}
		}
		else
		{
			if (spec.re || !dfa_map.empty())
			{
				if (parseMode != Scanner::Reuse)
				{
					dfa_map[""] = compile(spec, o, "", opts->encoding.nCodeUnits ());
				}
				if (parseMode != Scanner::Rules && dfa_map.find("") != dfa_map.end())
				{
					dfa_map[""]->emit(o, ind, 0, bPrologBrace);
				}
			}
		}
		o.source.wline_info (in->get_cline (), in->get_fname ().c_str ());
		/* restore original char handling mode*/
		opts.reset_encoding (encodingOld);
	}

	if (opts->cFlag)
	{
		SetupMap::const_iterator itRuleSetup;
		for (itRuleSetup = ruleSetupMap.begin(); itRuleSetup != ruleSetupMap.end(); ++itRuleSetup)
		{
			if (itRuleSetup->first != "*" && specMap.find(itRuleSetup->first) == specMap.end())
			{
				in->fatalf_at(itRuleSetup->second.first, "setup for non existing rule '%s' found", itRuleSetup->first.c_str());
			}
		}
		if (specMap.size() < ruleSetupMap.size())
		{
			uint32_t line = in->get_cline();
			itRuleSetup = ruleSetupMap.find("*");
			if (itRuleSetup != ruleSetupMap.end())
			{
				line = itRuleSetup->second.first;
			}
			in->fatalf_at(line, "setup for all rules with '*' not possible when all rules are setup explicitly");
		}
	}

	if (opts->target == opt_t::SKELETON)
	{
		Skeleton::emit_epilog (o.source, o.skeletons);
	}

	parse_cleanup();
	in = NULL;
}

void parse_cleanup()
{
	RegExp::vFreeList.clear();
	Range::vFreeList.clear();
	RangeSuffix::freeList.clear();
	Code::freelist.clear();
	symbol_table.clear ();
	condnames.clear ();
	specMap.clear();
	specStar.clear();
	star_default = NULL;
	specNone = NULL;
}

} // end namespace re2c