ydb/contrib/tools/ragel6/rlparse.kl

/*
 *  Copyright 2001-2007 Adrian Thurston <thurston@complang.org>
 */

/*  This file is part of Ragel.
 *
 *  Ragel is free software; you can redistribute it and/or modify
 *  it under the terms of the GNU General Public License as published by
 *  the Free Software Foundation; either version 2 of the License, or
 *  (at your option) any later version.
 * 
 *  Ragel is distributed in the hope that it will be useful,
 *  but WITHOUT ANY WARRANTY; without even the implied warranty of
 *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 *  GNU General Public License for more details.
 * 
 *  You should have received a copy of the GNU General Public License
 *  along with Ragel; if not, write to the Free Software
 *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA 
 */

#include "rlparse.h"
#include "ragel.h"
#include <iostream>
#include <errno.h>
#include <stdlib.h>

using std::cout;
using std::cerr;
using std::endl;

%%{

parser Parser;

include "rlparse.kh";

start: section_list;

section_list: section_list statement_list TK_EndSection;
section_list: ;

statement_list: statement_list statement;
statement_list: ;

statement: assignment commit;
statement: instantiation commit;
statement: action_spec commit;
statement: alphtype_spec commit;
statement: range_spec commit;
statement: getkey_spec commit;
statement: access_spec commit;
statement: variable_spec commit;
statement: export_block commit;
statement: pre_push_spec commit;
statement: post_pop_spec commit;
statement: length_spec commit;

length_spec:
	KW_Length TK_Word ';'
	final {
		LengthDef *lengthDef = new LengthDef( $2->data );
		pd->lengthDefList.append( lengthDef );

		/* Generic creation of machine for instantiation and assignment. */
		MachineDef *machineDef = new MachineDef( lengthDef );
		tryMachineDef( $2->loc, $2->data, machineDef, false );
	};

pre_push_spec:
	KW_PrePush '{' inline_block '}' 
	final {
		if ( pd->prePushExpr != 0 ) {
			/* Recover by just ignoring the duplicate. */
			error($2->loc) << "pre_push code already defined" << endl;
		}

		pd->prePushExpr = $3->inlineList;
	};


post_pop_spec:
	KW_PostPop '{' inline_block '}' 
	final {
		if ( pd->postPopExpr != 0 ) {
			/* Recover by just ignoring the duplicate. */
			error($2->loc) << "post_pop code already defined" << endl;
		}

		pd->postPopExpr = $3->inlineList;
	};


export_open: KW_Export 
	final {
		exportContext.append( true );
	};

nonterm opt_export
{
	bool isSet;
};

opt_export: export_open final { $$->isSet = true; };
opt_export: final { $$->isSet = false; };

export_block: export_open '{' statement_list '}' 
	final {
		exportContext.remove( exportContext.length()-1 );
	};

assignment:
	opt_export machine_name '=' join ';' final {
		/* Main machine must be an instance. */
		bool isInstance = false;
		if ( strcmp($2->token.data, mainMachine) == 0 ) {
			warning($2->token.loc) << 
					"main machine will be implicitly instantiated" << endl;
			isInstance = true;
		}

		/* Generic creation of machine for instantiation and assignment. */
		MachineDef *machineDef = new MachineDef( $4->join );
		tryMachineDef( $2->token.loc, $2->token.data, machineDef, isInstance );

		if ( $1->isSet )
			exportContext.remove( exportContext.length()-1 );

		$4->join->loc = $3->loc;
	};

instantiation: 
	opt_export machine_name TK_ColonEquals join_or_lm ';' final {
		/* Generic creation of machine for instantiation and assignment. */
		tryMachineDef( $2->token.loc, $2->token.data, $4->machineDef, true );

		if ( $1->isSet )
			exportContext.remove( exportContext.length()-1 );

		/* Pass a location to join_or_lm */
		if ( $4->machineDef->join != 0 )
			$4->machineDef->join->loc = $3->loc;
	};

type token_type
{
	Token token;
};

nonterm machine_name uses token_type;

machine_name: 
	TK_Word final {
		/* Make/get the priority key. The name may have already been referenced
		 * and therefore exist. */
		PriorDictEl *priorDictEl;
		if ( pd->priorDict.insert( $1->data, pd->nextPriorKey, &priorDictEl ) )
			pd->nextPriorKey += 1;
		pd->curDefPriorKey = priorDictEl->value;

		/* Make/get the local error key. */
		LocalErrDictEl *localErrDictEl;
		if ( pd->localErrDict.insert( $1->data, pd->nextLocalErrKey, &localErrDictEl ) )
			pd->nextLocalErrKey += 1;
		pd->curDefLocalErrKey = localErrDictEl->value;

		$$->token = *$1;
	};

action_spec:
	KW_Action TK_Word '{' inline_block '}' final {
		if ( pd->actionDict.find( $2->data ) ) {
			/* Recover by just ignoring the duplicate. */
			error($2->loc) << "action \"" << $2->data << "\" already defined" << endl;
		}
		else {
			//cerr << "NEW ACTION " << $2->data << " " << $4->inlineList << endl;
			/* Add the action to the list of actions. */
			Action *newAction = new Action( $3->loc, $2->data, 
					$4->inlineList, pd->nextCondId++ );

			/* Insert to list and dict. */
			pd->actionList.append( newAction );
			pd->actionDict.insert( newAction );
		}
	};

# Specifies the data type of the input alphabet. One or two words followed by a
# semi-colon.
alphtype_spec:
	KW_AlphType TK_Word TK_Word ';' final {
		if ( ! pd->setAlphType( $1->loc, $2->data, $3->data ) ) {
			// Recover by ignoring the alphtype statement.
			error($2->loc) << "\"" << $2->data << 
					" " << $3->data << "\" is not a valid alphabet type" << endl;
		}
	};

alphtype_spec:
	KW_AlphType TK_Word ';' final {
		if ( ! pd->setAlphType( $1->loc, $2->data ) ) {
			// Recover by ignoring the alphtype statement.
			error($2->loc) << "\"" << $2->data << 
					"\" is not a valid alphabet type" << endl;
		}
	};

# Specifies a range to assume that the input characters will fall into.
range_spec:
	KW_Range alphabet_num alphabet_num ';' final {
		// Save the upper and lower ends of the range and emit the line number.
		pd->lowerNum = $2->token.data;
		pd->upperNum = $3->token.data;
		pd->rangeLowLoc = $2->token.loc;
		pd->rangeHighLoc = $3->token.loc;
	};

getkey_spec:
	KW_GetKey inline_expr ';' final {
		pd->getKeyExpr = $2->inlineList;
	};

access_spec:
	KW_Access inline_expr ';' final {
		pd->accessExpr = $2->inlineList;
	};

variable_spec:
	KW_Variable opt_whitespace TK_Word inline_expr ';' final {
		/* FIXME: Need to implement the rest of this. */
		bool wasSet = pd->setVariable( $3->data, $4->inlineList );
		if ( !wasSet )
			error($3->loc) << "bad variable name" << endl;
	};

opt_whitespace: opt_whitespace IL_WhiteSpace;
opt_whitespace: ;

#
# Expressions
#

nonterm join_or_lm
{
	MachineDef *machineDef;
};

join_or_lm: 
	join final {
		$$->machineDef = new MachineDef( $1->join );
	};
join_or_lm:
	TK_BarStar lm_part_list '*' '|' final {
		/* Create a new factor going to a longest match structure. Record
		 * in the parse data that we have a longest match. */
		LongestMatch *lm = new LongestMatch( $1->loc, $2->lmPartList );
		pd->lmList.append( lm );
		for ( LmPartList::Iter lmp = *($2->lmPartList); lmp.lte(); lmp++ )
			lmp->longestMatch = lm;
		$$->machineDef = new MachineDef( lm );
	};

nonterm lm_part_list
{
	LmPartList *lmPartList;
};

lm_part_list:
	lm_part_list longest_match_part
	final {
		if ( $2->lmPart != 0 ) 
			$1->lmPartList->append( $2->lmPart );
		$$->lmPartList = $1->lmPartList;
	};
lm_part_list:
	longest_match_part
	final {
		/* Create a new list with the part. */
		$$->lmPartList = new LmPartList;
		if ( $1->lmPart != 0 )
			$$->lmPartList->append( $1->lmPart );
	};

nonterm longest_match_part
{
	LongestMatchPart *lmPart;
};

longest_match_part: 
	action_spec final { $$->lmPart = 0; };
longest_match_part: 
	assignment final { $$->lmPart = 0; };
longest_match_part: 
	join opt_lm_part_action ';' final {
		$$->lmPart = 0;
		Action *action = $2->action;
		if ( action != 0 )
			action->isLmAction = true;
		$$->lmPart = new LongestMatchPart( $1->join, action, 
				$3->loc, pd->nextLongestMatchId++ );

		/* Provide a location to join. Unfortunately We don't
		 * have the start of the join as in other occurances. Use the end. */
		$1->join->loc = $3->loc;
	};

nonterm opt_lm_part_action
{
	Action *action;
};

opt_lm_part_action:
	TK_DoubleArrow action_embed final { 
		$$->action = $2->action;
	};
opt_lm_part_action:
	action_embed_block final {
		$$->action = $1->action;
	};
opt_lm_part_action:
	final {
		$$->action = 0;
	};


nonterm join
{
	Join *join;
};

join: 
	join ',' expression final {
		/* Append the expression to the list and return it. */
		$1->join->exprList.append( $3->expression );
		$$->join = $1->join;
	};
join: 
	expression final {
		$$->join = new Join( $1->expression );
	};

nonterm expression
{
	Expression *expression;
};

expression: 
	expression '|' term_short final {
		$$->expression = new Expression( $1->expression, 
				$3->term, Expression::OrType );
	};
expression: 
	expression '&' term_short final {
		$$->expression = new Expression( $1->expression, 
				$3->term, Expression::IntersectType );
	};
expression: 
	expression '-' term_short final {
		$$->expression = new Expression( $1->expression, 
				$3->term, Expression::SubtractType );
	};
expression: 
	expression TK_DashDash term_short final {
		$$->expression = new Expression( $1->expression, 
				$3->term, Expression::StrongSubtractType );
	};
expression: 
	term_short final {
		$$->expression = new Expression( $1->term );
	};

# This is where we resolve the ambiguity involving -. By default ragel tries to
# do a longest match, which gives precedence to a concatenation because it is
# innermost. What we need is to force term into a shortest match so that when -
# is seen it doesn't try to extend term with a concatenation, but ends term and
# goes for a subtraction.
#
# The shortest tag overrides the default longest match action ordering strategy
# and instead forces a shortest match stragegy. The wrap the term production in
# a new nonterminal 'term_short' to guarantee the shortest match behaviour.

shortest term_short;
nonterm term_short
{
	Term *term;
};

term_short:
	term final {
		$$->term = $1->term;
	};

nonterm term
{
	Term *term;
};

term:
	term factor_with_label final {
		$$->term = new Term( $1->term, $2->factorWithAug );
	};
term:
	term '.' factor_with_label final {
		$$->term = new Term( $1->term, $3->factorWithAug );
	};
term:
	term TK_ColonGt factor_with_label final {
		$$->term = new Term( $1->term, $3->factorWithAug, Term::RightStartType );
	};
term:
	term TK_ColonGtGt factor_with_label final {
		$$->term = new Term( $1->term, $3->factorWithAug, Term::RightFinishType );
	};
term:
	term TK_LtColon factor_with_label final {
		$$->term = new Term( $1->term, 
				$3->factorWithAug, Term::LeftType );
	};
term:
	factor_with_label final {
		$$->term = new Term( $1->factorWithAug );
	};

nonterm factor_with_label
{
	FactorWithAug *factorWithAug;
};

factor_with_label: 
	TK_Word ':' factor_with_label final { 
		/* Add the label to the list and pass the factor up. */
		$3->factorWithAug->labels.prepend( Label($1->loc, $1->data) );
		$$->factorWithAug = $3->factorWithAug; 
	};
factor_with_label: 
	factor_with_ep final {
		$$->factorWithAug = $1->factorWithAug;
	};

nonterm factor_with_ep
{
	FactorWithAug *factorWithAug;
};

factor_with_ep: 
	factor_with_ep TK_Arrow local_state_ref final { 
		/* Add the target to the list and return the factor object. */
		$1->factorWithAug->epsilonLinks.append( EpsilonLink( $2->loc, nameRef ) );
		$$->factorWithAug = $1->factorWithAug; 
	};
factor_with_ep: 
	factor_with_aug final {
		$$->factorWithAug = $1->factorWithAug;
	};

nonterm factor_with_aug
{
	FactorWithAug *factorWithAug;
};

factor_with_aug:
	factor_with_aug aug_type_base action_embed final {
		/* Append the action to the factorWithAug, record the refernce from 
		 * factorWithAug to the action and pass up the factorWithAug. */
		$1->factorWithAug->actions.append( 
				ParserAction( $2->loc, $2->augType, 0, $3->action ) );
		$$->factorWithAug = $1->factorWithAug;
	};
factor_with_aug:
	factor_with_aug aug_type_base priority_aug final {
		/* Append the named priority to the factorWithAug and pass it up. */
		$1->factorWithAug->priorityAugs.append( 
				PriorityAug( $2->augType, pd->curDefPriorKey, $3->priorityNum ) );
		$$->factorWithAug = $1->factorWithAug;
	};
factor_with_aug:
	factor_with_aug aug_type_base '(' priority_name ',' priority_aug ')' final {
		/* Append the priority using a default name. */
		$1->factorWithAug->priorityAugs.append( 
				PriorityAug( $2->augType, $4->priorityName, $6->priorityNum ) );
		$$->factorWithAug = $1->factorWithAug;
	};
factor_with_aug:
	factor_with_aug aug_type_cond action_embed final {
		$1->factorWithAug->conditions.append( ConditionTest( $2->loc, 
				$2->augType, $3->action, true ) );
		$$->factorWithAug = $1->factorWithAug;
	};
factor_with_aug:
	factor_with_aug aug_type_cond '!' action_embed final {
		$1->factorWithAug->conditions.append( ConditionTest( $2->loc, 
				$2->augType, $4->action, false ) );
		$$->factorWithAug = $1->factorWithAug;
	};
factor_with_aug:
	factor_with_aug aug_type_to_state action_embed final {
		/* Append the action, pass it up. */
		$1->factorWithAug->actions.append( ParserAction( $2->loc, 
				$2->augType, 0, $3->action ) );
		$$->factorWithAug = $1->factorWithAug;
	};
factor_with_aug:
	factor_with_aug aug_type_from_state action_embed final {
		/* Append the action, pass it up. */
		$1->factorWithAug->actions.append( ParserAction( $2->loc,
				$2->augType, 0, $3->action ) );
		$$->factorWithAug = $1->factorWithAug;
	};
factor_with_aug:
	factor_with_aug aug_type_eof action_embed final {
		/* Append the action, pass it up. */
		$1->factorWithAug->actions.append( ParserAction( $2->loc,
				$2->augType, 0, $3->action ) );
		$$->factorWithAug = $1->factorWithAug;
	};
factor_with_aug:
	factor_with_aug aug_type_gbl_error action_embed final {
		/* Append the action to the factorWithAug, record the refernce from 
		 * factorWithAug to the action and pass up the factorWithAug. */
		$1->factorWithAug->actions.append( ParserAction( $2->loc,
				$2->augType, pd->curDefLocalErrKey, $3->action ) );
		$$->factorWithAug = $1->factorWithAug;
	};
factor_with_aug:
	factor_with_aug aug_type_local_error action_embed final {
		/* Append the action to the factorWithAug, record the refernce from 
		 * factorWithAug to the action and pass up the factorWithAug. */
		$1->factorWithAug->actions.append( ParserAction( $2->loc, 
				$2->augType, pd->curDefLocalErrKey, $3->action ) );
		$$->factorWithAug = $1->factorWithAug;
	};
factor_with_aug:
	factor_with_aug aug_type_local_error '(' local_err_name ',' action_embed ')' final {
		/* Append the action to the factorWithAug, record the refernce from
		 * factorWithAug to the action and pass up the factorWithAug. */
		$1->factorWithAug->actions.append( ParserAction( $2->loc,
				$2->augType, $4->error_name, $6->action ) );
		$$->factorWithAug = $1->factorWithAug;
	};
factor_with_aug:
	factor_with_rep final {
		$$->factorWithAug = new FactorWithAug( $1->factorWithRep );
	};

type aug_type
{
	InputLoc loc;
	AugType augType;
};

#  Classes of transtions on which to embed actions or change priorities.
nonterm aug_type_base uses aug_type;

aug_type_base: '@' final { $$->loc = $1->loc; $$->augType = at_finish; };
aug_type_base: '%' final { $$->loc = $1->loc; $$->augType = at_leave; };
aug_type_base: '$' final { $$->loc = $1->loc; $$->augType = at_all; };
aug_type_base: '>' final { $$->loc = $1->loc; $$->augType = at_start; };

# Embedding conditions.
nonterm aug_type_cond uses aug_type;

aug_type_cond: TK_StartCond final { $$->loc = $1->loc; $$->augType = at_start; };
aug_type_cond: '>' KW_When final { $$->loc = $1->loc; $$->augType = at_start; };
aug_type_cond: TK_AllCond final { $$->loc = $1->loc; $$->augType = at_all; };
aug_type_cond: '$' KW_When final { $$->loc = $1->loc; $$->augType = at_all; };
aug_type_cond: TK_LeavingCond final { $$->loc = $1->loc; $$->augType = at_leave; };
aug_type_cond: '%' KW_When final { $$->loc = $1->loc; $$->augType = at_leave; };
aug_type_cond: KW_When final { $$->loc = $1->loc; $$->augType = at_all; };
aug_type_cond: KW_InWhen final { $$->loc = $1->loc; $$->augType = at_start; };
aug_type_cond: KW_OutWhen final { $$->loc = $1->loc; $$->augType = at_leave; };

#
# To state actions.
#

nonterm aug_type_to_state uses aug_type;

aug_type_to_state: TK_StartToState 
		final { $$->loc = $1->loc; $$->augType = at_start_to_state; };
aug_type_to_state: '>' KW_To 
		final { $$->loc = $1->loc; $$->augType = at_start_to_state; };

aug_type_to_state: TK_NotStartToState 
		final { $$->loc = $1->loc; $$->augType = at_not_start_to_state; };
aug_type_to_state: '<' KW_To 
		final { $$->loc = $1->loc; $$->augType = at_not_start_to_state; };

aug_type_to_state: TK_AllToState 
		final { $$->loc = $1->loc; $$->augType = at_all_to_state; };
aug_type_to_state: '$' KW_To 
		final { $$->loc = $1->loc; $$->augType = at_all_to_state; };

aug_type_to_state: TK_FinalToState
		final { $$->loc = $1->loc; $$->augType = at_final_to_state; };
aug_type_to_state: '%' KW_To
		final { $$->loc = $1->loc; $$->augType = at_final_to_state; };

aug_type_to_state: TK_NotFinalToState
		final { $$->loc = $1->loc; $$->augType = at_not_final_to_state; };
aug_type_to_state: '@' KW_To
		final { $$->loc = $1->loc; $$->augType = at_not_final_to_state; };

aug_type_to_state: TK_MiddleToState
		final { $$->loc = $1->loc; $$->augType = at_middle_to_state; };
aug_type_to_state: TK_Middle KW_To
		final { $$->loc = $1->loc; $$->augType = at_middle_to_state; };

#
# From state actions.
#

nonterm aug_type_from_state uses aug_type;

aug_type_from_state: TK_StartFromState 
		final { $$->loc = $1->loc; $$->augType = at_start_from_state; };
aug_type_from_state: '>' KW_From 
		final { $$->loc = $1->loc; $$->augType = at_start_from_state; };

aug_type_from_state: TK_NotStartFromState 
		final { $$->loc = $1->loc; $$->augType = at_not_start_from_state; };
aug_type_from_state: '<' KW_From 
		final { $$->loc = $1->loc; $$->augType = at_not_start_from_state; };

aug_type_from_state: TK_AllFromState 
		final { $$->loc = $1->loc; $$->augType = at_all_from_state; };
aug_type_from_state: '$' KW_From 
		final { $$->loc = $1->loc; $$->augType = at_all_from_state; };

aug_type_from_state: TK_FinalFromState 
		final { $$->loc = $1->loc; $$->augType = at_final_from_state; };
aug_type_from_state: '%' KW_From 
		final { $$->loc = $1->loc; $$->augType = at_final_from_state; };

aug_type_from_state: TK_NotFinalFromState 
		final { $$->loc = $1->loc; $$->augType = at_not_final_from_state; };
aug_type_from_state: '@' KW_From 
		final { $$->loc = $1->loc; $$->augType = at_not_final_from_state; };

aug_type_from_state: TK_MiddleFromState 
		final { $$->loc = $1->loc; $$->augType = at_middle_from_state; };
aug_type_from_state: TK_Middle KW_From 
		final { $$->loc = $1->loc; $$->augType = at_middle_from_state; };

#
# Eof state actions.
#

nonterm aug_type_eof uses aug_type;

aug_type_eof: TK_StartEOF 
		final { $$->loc = $1->loc; $$->augType = at_start_eof; };
aug_type_eof: '>' KW_Eof
		final { $$->loc = $1->loc; $$->augType = at_start_eof; };

aug_type_eof: TK_NotStartEOF
		final { $$->loc = $1->loc; $$->augType = at_not_start_eof; };
aug_type_eof: '<' KW_Eof
		final { $$->loc = $1->loc; $$->augType = at_not_start_eof; };

aug_type_eof: TK_AllEOF
		final { $$->loc = $1->loc; $$->augType = at_all_eof; };
aug_type_eof: '$' KW_Eof
		final { $$->loc = $1->loc; $$->augType = at_all_eof; };

aug_type_eof: TK_FinalEOF
		final { $$->loc = $1->loc; $$->augType = at_final_eof; };
aug_type_eof: '%' KW_Eof
		final { $$->loc = $1->loc; $$->augType = at_final_eof; };

aug_type_eof: TK_NotFinalEOF
		final { $$->loc = $1->loc; $$->augType = at_not_final_eof; };
aug_type_eof: '@' KW_Eof
		final { $$->loc = $1->loc; $$->augType = at_not_final_eof; };

aug_type_eof: TK_MiddleEOF
		final { $$->loc = $1->loc; $$->augType = at_middle_eof; };
aug_type_eof: TK_Middle KW_Eof
		final { $$->loc = $1->loc; $$->augType = at_middle_eof; };

#
# Global error actions.
#

nonterm aug_type_gbl_error uses aug_type;

aug_type_gbl_error: TK_StartGblError 
		final { $$->loc = $1->loc; $$->augType = at_start_gbl_error; };
aug_type_gbl_error: '>' KW_Err 
		final { $$->loc = $1->loc; $$->augType = at_start_gbl_error; };

aug_type_gbl_error: TK_NotStartGblError 
		final { $$->loc = $1->loc; $$->augType = at_not_start_gbl_error; };
aug_type_gbl_error: '<' KW_Err
		final { $$->loc = $1->loc; $$->augType = at_not_start_gbl_error; };

aug_type_gbl_error: TK_AllGblError
		final { $$->loc = $1->loc; $$->augType = at_all_gbl_error; };
aug_type_gbl_error: '$' KW_Err
		final { $$->loc = $1->loc; $$->augType = at_all_gbl_error; };

aug_type_gbl_error: TK_FinalGblError
		final { $$->loc = $1->loc; $$->augType = at_final_gbl_error; };
aug_type_gbl_error: '%' KW_Err
		final { $$->loc = $1->loc; $$->augType = at_final_gbl_error; };

aug_type_gbl_error: TK_NotFinalGblError
		final { $$->loc = $1->loc; $$->augType = at_not_final_gbl_error; };
aug_type_gbl_error: '@' KW_Err
		final { $$->loc = $1->loc; $$->augType = at_not_final_gbl_error; };

aug_type_gbl_error: TK_MiddleGblError
		final { $$->loc = $1->loc; $$->augType = at_middle_gbl_error; };
aug_type_gbl_error: TK_Middle KW_Err
		final { $$->loc = $1->loc; $$->augType = at_middle_gbl_error; };


#
# Local error actions.
#

nonterm aug_type_local_error uses aug_type;

aug_type_local_error: TK_StartLocalError
		final { $$->loc = $1->loc; $$->augType = at_start_local_error; };
aug_type_local_error: '>' KW_Lerr
		final { $$->loc = $1->loc; $$->augType = at_start_local_error; };

aug_type_local_error: TK_NotStartLocalError
		final { $$->loc = $1->loc; $$->augType = at_not_start_local_error; };
aug_type_local_error: '<' KW_Lerr
		final { $$->loc = $1->loc; $$->augType = at_not_start_local_error; };

aug_type_local_error: TK_AllLocalError
		final { $$->loc = $1->loc; $$->augType = at_all_local_error; };
aug_type_local_error: '$' KW_Lerr
		final { $$->loc = $1->loc; $$->augType = at_all_local_error; };

aug_type_local_error: TK_FinalLocalError
		final { $$->loc = $1->loc; $$->augType = at_final_local_error; };
aug_type_local_error: '%' KW_Lerr
		final { $$->loc = $1->loc; $$->augType = at_final_local_error; };

aug_type_local_error: TK_NotFinalLocalError
		final { $$->loc = $1->loc; $$->augType = at_not_final_local_error; };
aug_type_local_error: '@' KW_Lerr
		final { $$->loc = $1->loc; $$->augType = at_not_final_local_error; };

aug_type_local_error: TK_MiddleLocalError
		final { $$->loc = $1->loc; $$->augType = at_middle_local_error; };
aug_type_local_error: TK_Middle KW_Lerr
		final { $$->loc = $1->loc; $$->augType = at_middle_local_error; };


type action_ref
{
	Action *action;
};

# Different ways to embed actions. A TK_Word is reference to an action given by
# the user as a statement in the fsm specification. An action can also be
# specified immediately.
nonterm action_embed uses action_ref;

action_embed: action_embed_word final { $$->action = $1->action; };
action_embed: '(' action_embed_word ')' final { $$->action = $2->action; };
action_embed: action_embed_block final { $$->action = $1->action; };

nonterm action_embed_word uses action_ref;

action_embed_word:
	TK_Word final {
		/* Set the name in the actionDict. */
		Action *action = pd->actionDict.find( $1->data );
		if ( action != 0 ) {
			/* Pass up the action element */
			$$->action = action;
		}
		else {
			/* Will recover by returning null as the action. */
			error($1->loc) << "action lookup of \"" << $1->data << "\" failed" << endl;
			$$->action = 0;
		}
	};

nonterm action_embed_block uses action_ref;

action_embed_block:
	'{' inline_block '}' final {
		/* Create the action, add it to the list and pass up. */
		Action *newAction = new Action( $1->loc, 0, $2->inlineList, pd->nextCondId++ );
		pd->actionList.append( newAction );
		$$->action = newAction;
	};

nonterm priority_name
{
	int priorityName;
};

# A specified priority name. Looks up the name in the current priority
# dictionary.
priority_name:
	TK_Word final {
		// Lookup/create the priority key.
		PriorDictEl *priorDictEl;
		if ( pd->priorDict.insert( $1->data, pd->nextPriorKey, &priorDictEl ) )
			pd->nextPriorKey += 1;

		// Use the inserted/found priority key.
		$$->priorityName = priorDictEl->value;
	};

nonterm priority_aug
{
	int priorityNum;
};

# Priority change specs.
priority_aug: 
	priority_aug_num final {
		// Convert the priority number to a long. Check for overflow.
		errno = 0;
		//cerr << "PRIOR AUG: " << $1->token.data << endl;
		long aug = strtol( $1->token.data, 0, 10 );
		if ( errno == ERANGE && aug == LONG_MAX ) {
			/* Priority number too large. Recover by setting the priority to 0. */
			error($1->token.loc) << "priority number " << $1->token.data << 
					" overflows" << endl;
			$$->priorityNum = 0;
		}
		else if ( errno == ERANGE && aug == LONG_MIN ) {
			/* Priority number too large in the neg. Recover by using 0. */
			error($1->token.loc) << "priority number " << $1->token.data << 
					" underflows" << endl;
			$$->priorityNum = 0;
		}
		else {
			/* No overflow or underflow. */
			$$->priorityNum = aug;
		}
	};

nonterm priority_aug_num uses token_type;

priority_aug_num:
	TK_UInt final {
		$$->token = *$1;
	};
priority_aug_num:
	'+' TK_UInt final {
		$$->token.set( "+", 1 );
		$$->token.loc = $1->loc;
		$$->token.append( *$2 );
	};
priority_aug_num:
	'-' TK_UInt final {
		$$->token.set( "-", 1 );
		$$->token.loc = $1->loc;
		$$->token.append( *$2 );
	};

nonterm local_err_name
{
	int error_name;
};

local_err_name:
	TK_Word final {
		/* Lookup/create the priority key. */
		LocalErrDictEl *localErrDictEl;
		if ( pd->localErrDict.insert( $1->data, pd->nextLocalErrKey, &localErrDictEl ) )
			pd->nextLocalErrKey += 1;

		/* Use the inserted/found priority key. */
		$$->error_name = localErrDictEl->value;
	};



# The fourth level of precedence. These are the trailing unary operators that
# allow for repetition.

nonterm factor_with_rep
{
	FactorWithRep *factorWithRep;
};

factor_with_rep:
	factor_with_rep '*' final {
		$$->factorWithRep = new FactorWithRep( $2->loc, $1->factorWithRep, 
				0, 0, FactorWithRep::StarType );
	};
factor_with_rep:
	factor_with_rep TK_StarStar final {
		$$->factorWithRep = new FactorWithRep( $2->loc, $1->factorWithRep, 
				0, 0, FactorWithRep::StarStarType );
	};
factor_with_rep:
	factor_with_rep '?' final {
		$$->factorWithRep = new FactorWithRep( $2->loc, $1->factorWithRep, 
				0, 0, FactorWithRep::OptionalType );
	};
factor_with_rep:
	factor_with_rep '+' final {
		$$->factorWithRep = new FactorWithRep( $2->loc, $1->factorWithRep, 
				0, 0, FactorWithRep::PlusType );
	};
factor_with_rep:
	factor_with_rep '{' factor_rep_num '}' final {
		$$->factorWithRep = new FactorWithRep( $2->loc, $1->factorWithRep, 
				$3->rep, 0, FactorWithRep::ExactType );
	};
factor_with_rep:
	factor_with_rep '{' ',' factor_rep_num '}' final {
		$$->factorWithRep = new FactorWithRep( $2->loc, $1->factorWithRep, 
				0, $4->rep, FactorWithRep::MaxType );
	};
factor_with_rep:
	factor_with_rep '{' factor_rep_num ',' '}' final {
		$$->factorWithRep = new FactorWithRep( $2->loc, $1->factorWithRep,
				$3->rep, 0, FactorWithRep::MinType );
	};
factor_with_rep:
	factor_with_rep '{' factor_rep_num ',' factor_rep_num '}' final {
		$$->factorWithRep = new FactorWithRep( $2->loc, $1->factorWithRep, 
				$3->rep, $5->rep, FactorWithRep::RangeType );
	};
factor_with_rep:
	factor_with_neg final {
		$$->factorWithRep = new FactorWithRep( $1->factorWithNeg );
	};

nonterm factor_rep_num
{
	int rep;
};

factor_rep_num:
	TK_UInt final {
		// Convert the priority number to a long. Check for overflow.
		errno = 0;
		long rep = strtol( $1->data, 0, 10 );
		if ( errno == ERANGE && rep == LONG_MAX ) {
			// Repetition too large. Recover by returing repetition 1. */
			error($1->loc) << "repetition number " << $1->data << " overflows" << endl;
			$$->rep = 1;
		}
		else {
			// Cannot be negative, so no overflow.
			$$->rep = rep;
 		}
	};


#
# The fifth level up in precedence. Negation.
#

nonterm factor_with_neg
{
	FactorWithNeg *factorWithNeg;
};

factor_with_neg:
	'!' factor_with_neg final {
		$$->factorWithNeg = new FactorWithNeg( $1->loc,
				$2->factorWithNeg, FactorWithNeg::NegateType );
	};
factor_with_neg:
	'^' factor_with_neg final {
		$$->factorWithNeg = new FactorWithNeg( $1->loc,
				$2->factorWithNeg, FactorWithNeg::CharNegateType );
	};
factor_with_neg:
	factor final {
		$$->factorWithNeg = new FactorWithNeg( $1->factor );
	};

nonterm factor
{
	Factor *factor;
};

factor: 
	TK_Literal final {
		/* Create a new factor node going to a concat literal. */
		$$->factor = new Factor( new Literal( *$1, Literal::LitString ) );
	};
factor: 
	alphabet_num final {
		/* Create a new factor node going to a literal number. */
		$$->factor = new Factor( new Literal( $1->token, Literal::Number ) );
	};
factor:
	TK_Word final {
		/* Find the named graph. */
		GraphDictEl *gdNode = pd->graphDict.find( $1->data );
		if ( gdNode == 0 ) {
			/* Recover by returning null as the factor node. */
			error($1->loc) << "graph lookup of \"" << $1->data << "\" failed" << endl;
			$$->factor = 0;
		}
		else if ( gdNode->isInstance ) {
			/* Recover by retuning null as the factor node. */
			error($1->loc) << "references to graph instantiations not allowed "
					"in expressions" << endl;
			$$->factor = 0;
		}
		else {
			/* Create a factor node that is a lookup of an expression. */
			$$->factor = new Factor( $1->loc, gdNode->value );
		}
	};
factor:
	RE_SqOpen regular_expr_or_data RE_SqClose final {
		/* Create a new factor node going to an OR expression. */
		$$->factor = new Factor( new ReItem( $1->loc, $2->reOrBlock, ReItem::OrBlock ) );
	};
factor:
	RE_SqOpenNeg regular_expr_or_data RE_SqClose final {
		/* Create a new factor node going to a negated OR expression. */
		$$->factor = new Factor( new ReItem( $1->loc, $2->reOrBlock, ReItem::NegOrBlock ) );
	};
factor:
	RE_Slash regular_expr RE_Slash final {
		if ( $3->length > 1 ) {
			for ( char *p = $3->data; *p != 0; p++ ) {
				if ( *p == 'i' )
					$2->regExpr->caseInsensitive = true;
			}
		}

		/* Create a new factor node going to a regular exp. */
		$$->factor = new Factor( $2->regExpr );
	};
factor:
	range_lit TK_DotDot range_lit final {
		/* Create a new factor node going to a range. */
		$$->factor = new Factor( new Range( $1->literal, $3->literal ) );
	};
factor:
	'(' join ')' final {
		/* Create a new factor going to a parenthesized join. */
		$$->factor = new Factor( $2->join );
		$2->join->loc = $1->loc;
	};

nonterm range_lit
{
	Literal *literal;
};

# Literals which can be the end points of ranges.
range_lit:
	TK_Literal final {
		/* Range literas must have only one char. We restrict this in the parse tree. */
		$$->literal = new Literal( *$1, Literal::LitString );
	};
range_lit:
	alphabet_num final {
		/* Create a new literal number. */
		$$->literal = new Literal( $1->token, Literal::Number );
	};

nonterm alphabet_num uses token_type;

# Any form of a number that can be used as a basic machine. */
alphabet_num:
	TK_UInt final { 
		$$->token = *$1;
	};
alphabet_num: 
	'-' TK_UInt final { 
		$$->token.set( "-", 1 );
		$$->token.loc = $1->loc;
		$$->token.append( *$2 );
	};
alphabet_num: 
	TK_Hex final { 
		$$->token = *$1;
	};
#
# Regular Expressions.
#

nonterm regular_expr
{
	RegExpr *regExpr;
};

# Parser for regular expression fsms. Any number of expression items which
# generally gives a machine one character long or one character long stared.
regular_expr:
	regular_expr regular_expr_item final {
		/* An optimization to lessen the tree size. If a non-starred char is
		 * directly under the left side on the right and the right side is
		 * another non-starred char then paste them together and return the
		 * left side. Otherwise just put the two under a new reg exp node. */
		if ( $2->reItem->type == ReItem::Data && !$2->reItem->star &&
			$1->regExpr->type == RegExpr::RecurseItem &&
			$1->regExpr->item->type == ReItem::Data && !$1->regExpr->item->star )
		{
			/* Append the right side to the right side of the left and toss the
			 * right side. */
			$1->regExpr->item->token.append( $2->reItem->token );
			delete $2->reItem;
			$$->regExpr = $1->regExpr;
		}
		else {
			$$->regExpr = new RegExpr( $1->regExpr, $2->reItem );
		}
	};
regular_expr:
	final {
		/* Can't optimize the tree. */
		$$->regExpr = new RegExpr();
	};

nonterm regular_expr_item
{
	ReItem *reItem;
};

# RegularExprItems can be a character spec with an optional staring of the char.
regular_expr_item:
	regular_expr_char RE_Star final {
		$1->reItem->star = true;
		$$->reItem = $1->reItem;
	};
regular_expr_item:
	regular_expr_char final {
		$$->reItem = $1->reItem;
	};

nonterm regular_expr_char
{
	ReItem *reItem;
};

# A character spec can be a set of characters inside of square parenthesis, a
# dot specifying any character or some explicitly stated character.
regular_expr_char:
	RE_SqOpen regular_expr_or_data RE_SqClose final {
		$$->reItem = new ReItem( $1->loc, $2->reOrBlock, ReItem::OrBlock );
	};
regular_expr_char:
	RE_SqOpenNeg regular_expr_or_data RE_SqClose final {
		$$->reItem = new ReItem( $1->loc, $2->reOrBlock, ReItem::NegOrBlock );
	};
regular_expr_char:
	RE_Dot final {
		$$->reItem = new ReItem( $1->loc, ReItem::Dot );
	};
regular_expr_char:
	RE_Char final {
		$$->reItem = new ReItem( $1->loc, *$1 );
	};

# The data inside of a [] expression in a regular expression. Accepts any
# number of characters or ranges. */
nonterm regular_expr_or_data
{
	ReOrBlock *reOrBlock;
};

regular_expr_or_data:
	regular_expr_or_data regular_expr_or_char final {
		/* An optimization to lessen the tree size. If an or char is directly
		 * under the left side on the right and the right side is another or
		 * char then paste them together and return the left side. Otherwise
		 * just put the two under a new or data node. */
		if ( $2->reOrItem->type == ReOrItem::Data &&
				$1->reOrBlock->type == ReOrBlock::RecurseItem &&
				$1->reOrBlock->item->type == ReOrItem::Data )
		{
			/* Append the right side to right side of the left and toss the
			 * right side. */
			$1->reOrBlock->item->token.append( $2->reOrItem->token );
			delete $2->reOrItem;
			$$->reOrBlock = $1->reOrBlock;
		}
		else {
			/* Can't optimize, put the left and right under a new node. */
			$$->reOrBlock = new ReOrBlock( $1->reOrBlock, $2->reOrItem );
		}
	};
regular_expr_or_data:
	final {
		$$->reOrBlock = new ReOrBlock();
	};

# A single character inside of an or expression. Can either be a character or a
# set of characters.
nonterm regular_expr_or_char
{
	ReOrItem *reOrItem;
};

regular_expr_or_char:
	RE_Char final {
		$$->reOrItem = new ReOrItem( $1->loc, *$1 );
	};
regular_expr_or_char:
	RE_Char RE_Dash RE_Char final {
		$$->reOrItem = new ReOrItem( $2->loc, $1->data[0], $3->data[0] );
	};

#
# Inline Lists for inline host code.
#

type inline_list
{
	InlineList *inlineList;
};

nonterm inline_block uses inline_list;

inline_block:
	inline_block inline_block_item 
	final {
		/* Append the item to the list, return the list. */
		$$->inlineList = $1->inlineList;
		$$->inlineList->append( $2->inlineItem );
	};

inline_block:
	final {
		/* Start with empty list. */
		$$->inlineList = new InlineList;
	};

type inline_item
{
	InlineItem *inlineItem;
};

nonterm inline_block_item uses inline_item;
nonterm inline_block_interpret uses inline_item;

inline_block_item:
	inline_expr_any 
	final {
		$$->inlineItem = new InlineItem( $1->token.loc, $1->token.data, InlineItem::Text );
	};

inline_block_item:
	inline_block_symbol 
	final {
		$$->inlineItem = new InlineItem( $1->token.loc, $1->token.data, InlineItem::Text );
	};

inline_block_item:
	inline_block_interpret
	final {
		/* Pass the inline item up. */
		$$->inlineItem = $1->inlineItem;
	};

nonterm inline_block_symbol uses token_type;

inline_block_symbol: ',' final { $$->token = *$1; };
inline_block_symbol: ';' final { $$->token = *$1; };
inline_block_symbol: '(' final { $$->token = *$1; };
inline_block_symbol: ')' final { $$->token = *$1; };
inline_block_symbol: '*' final { $$->token = *$1; };
inline_block_symbol: TK_NameSep final { $$->token = *$1; };

# Interpreted statements in a struct block. */
inline_block_interpret:
	inline_expr_interpret final {
		/* Pass up interpreted items of inline expressions. */
		$$->inlineItem = $1->inlineItem;
	};
inline_block_interpret:
	KW_Hold ';' final {
		$$->inlineItem = new InlineItem( $1->loc, InlineItem::Hold );
	};
inline_block_interpret:
	KW_Exec inline_expr ';' final {
		$$->inlineItem = new InlineItem( $1->loc, InlineItem::Exec );
		$$->inlineItem->children = $2->inlineList;
	};
inline_block_interpret:
	KW_Goto state_ref ';' final { 
		$$->inlineItem = new InlineItem( $1->loc, 
				new NameRef(nameRef), InlineItem::Goto );
	};
inline_block_interpret:
	KW_Goto '*' inline_expr ';' final {
		$$->inlineItem = new InlineItem( $1->loc, InlineItem::GotoExpr );
		$$->inlineItem->children = $3->inlineList;
	};
inline_block_interpret:
	KW_Next state_ref ';' final { 
		$$->inlineItem = new InlineItem( $1->loc, new NameRef(nameRef), InlineItem::Next );
	};
inline_block_interpret:
	KW_Next '*' inline_expr ';' final {
		$$->inlineItem = new InlineItem( $1->loc, InlineItem::NextExpr );
		$$->inlineItem->children = $3->inlineList;
	};
inline_block_interpret:
	KW_Call state_ref ';' final {
		$$->inlineItem = new InlineItem( $1->loc, new NameRef(nameRef), InlineItem::Call );
	};
inline_block_interpret:
	KW_Call '*' inline_expr ';' final {
		$$->inlineItem = new InlineItem( $1->loc, InlineItem::CallExpr );
		$$->inlineItem->children = $3->inlineList;
	};
inline_block_interpret:
	KW_Ret ';' final {
		$$->inlineItem = new InlineItem( $1->loc, InlineItem::Ret );
	};
inline_block_interpret:
	KW_Break ';' final {
		$$->inlineItem = new InlineItem( $1->loc, InlineItem::Break );
	};

nonterm inline_expr uses inline_list;

inline_expr:
	inline_expr inline_expr_item 
	final {
		$$->inlineList = $1->inlineList;
		$$->inlineList->append( $2->inlineItem );
	};
inline_expr:
	final {
		/* Init the list used for this expr. */
		$$->inlineList = new InlineList;
	};

nonterm inline_expr_item uses inline_item;

inline_expr_item: 
	inline_expr_any 
	final {
		/* Return a text segment. */
		$$->inlineItem = new InlineItem( $1->token.loc, $1->token.data, InlineItem::Text );
	};
inline_expr_item:
	inline_expr_symbol
	final {
		/* Return a text segment, must heap alloc the text. */
		$$->inlineItem = new InlineItem( $1->token.loc, $1->token.data, InlineItem::Text );
	};
inline_expr_item:
	inline_expr_interpret
	final{
		/* Pass the inline item up. */
		$$->inlineItem = $1->inlineItem;
	};

nonterm inline_expr_any uses token_type;

inline_expr_any: IL_WhiteSpace try { $$->token = *$1; };
inline_expr_any: IL_Comment try { $$->token = *$1; };
inline_expr_any: IL_Literal try { $$->token = *$1; };
inline_expr_any: IL_Symbol try { $$->token = *$1; };
inline_expr_any: TK_UInt try { $$->token = *$1; };
inline_expr_any: TK_Hex try { $$->token = *$1; };
inline_expr_any: TK_Word try { $$->token = *$1; };

# Anything in a ExecValExpr that is not dynamically allocated. This includes
# all special symbols caught in inline code except the semi.

nonterm inline_expr_symbol uses token_type;

inline_expr_symbol: ',' try { $$->token = *$1; };
inline_expr_symbol: '(' try { $$->token = *$1; };
inline_expr_symbol: ')' try { $$->token = *$1; };
inline_expr_symbol: '*' try { $$->token = *$1; };
inline_expr_symbol: TK_NameSep try { $$->token = *$1; };

nonterm inline_expr_interpret uses inline_item;

inline_expr_interpret:
	KW_PChar 
	final {
		$$->inlineItem = new InlineItem( $1->loc, InlineItem::PChar );
	};
inline_expr_interpret:
	KW_Char 
	final {
		$$->inlineItem = new InlineItem( $1->loc, InlineItem::Char );
	};
inline_expr_interpret:
	KW_CurState 
	final {
		$$->inlineItem = new InlineItem( $1->loc, InlineItem::Curs );
	};
inline_expr_interpret:
	KW_TargState 
	final {
		$$->inlineItem = new InlineItem( $1->loc, InlineItem::Targs );
	};
inline_expr_interpret:
	KW_Entry '(' state_ref ')' 
	final {
		$$->inlineItem = new InlineItem( $1->loc, 
			new NameRef(nameRef), InlineItem::Entry );
	};

#  A local state reference. Cannot have :: prefix.
local_state_ref:
	no_name_sep state_ref_names;

# Clear the name ref structure.
no_name_sep:
	final {
		nameRef.empty();
	};

# A qualified state reference.
state_ref: opt_name_sep state_ref_names;

# Optional leading name separator.
opt_name_sep:
	TK_NameSep 
	final {
		/* Insert an initial null pointer val to indicate the existence of the
		 * initial name seperator. */
		nameRef.setAs( 0 );
	};
opt_name_sep:
	final {
		nameRef.empty();
	};

# List of names separated by ::
state_ref_names:
	state_ref_names TK_NameSep TK_Word
	final {
		nameRef.append( $3->data );
	};
state_ref_names:
	TK_Word 
	final {
		nameRef.append( $1->data );
	};

}%%

%%{
	write types;
	write data;
}%%

void Parser::init()
{
	%% write init;
}

int Parser::parseLangEl( int type, const Token *token )
{
	%% write exec;
	return errCount == 0 ? 0 : -1;
}

void Parser::tryMachineDef( InputLoc &loc, char *name, 
		MachineDef *machineDef, bool isInstance )
{
	GraphDictEl *newEl = pd->graphDict.insert( name );
	if ( newEl != 0 ) {
		/* New element in the dict, all good. */
		newEl->value = new VarDef( name, machineDef );
		newEl->isInstance = isInstance;
		newEl->loc = loc;
		newEl->value->isExport = exportContext[exportContext.length()-1];

		/* It it is an instance, put on the instance list. */
		if ( isInstance )
			pd->instanceList.append( newEl );
	}
	else {
		// Recover by ignoring the duplicate.
		error(loc) << "fsm \"" << name << "\" previously defined" << endl;
	}
}

ostream &Parser::parse_error( int tokId, Token &token )
{
	/* Maintain the error count. */
	gblErrorCount += 1;

	cerr << token.loc << ": ";
	cerr << "at token ";
	if ( tokId < 128 )
		cerr << "\"" << Parser_lelNames[tokId] << "\"";
	else 
		cerr << Parser_lelNames[tokId];
	if ( token.data != 0 )
		cerr << " with data \"" << token.data << "\"";
	cerr << ": ";
	
	return cerr;
}

int Parser::token( InputLoc &loc, int tokId, char *tokstart, int toklen )
{
	Token token;
	token.data = tokstart;
	token.length = toklen;
	token.loc = loc;
	int res = parseLangEl( tokId, &token );
	if ( res < 0 ) {
		parse_error(tokId, token) << "parse error" << endl;
		exit(1);
	}
	return res;
}