CODEWITHSUNDEEP
grammarrascal
ThomasH
ThomasH

Reputation: 23506

Rascal: Grammar to Parse BNF

I want to write a concrete grammar to parse BNF-like syntax definitions.

Looking at the EXP Concrete Syntax recipe I created this very simple first version:

module BNFParser

lexical Identifier = [a-z]+ ;

syntax GrammarRule = left RuleHead ":" RuleCase* ";" ;
syntax RuleHead = Identifier ;
syntax RuleCase =  Identifier ;

and invoked it in the Repl like this:

import BNFParser;
import ParseTree;
parse(#GrammarRule, "foo : bar baz ;");

But this results in a rather arcane error message:

|std:///ParseTree.rsc|(13035,1963,<393,0>,<439,114>): ParseError(|unknown:///|(3,1,<1,3>,<1,4>))
        at *** somewhere ***(|std:///ParseTree.rsc|(13035,1963,<393,0>,<439,114>))
        at parse(|std:///ParseTree.rsc|(14991,5,<439,107>,<439,112>))
ok

I also tried using the start keyword ahead of GrammarRule, but that didn't help. What am I doing wrong?

Upvotes: 1

Views: 140

Answers (1)

Jurgen Vinju
Jurgen Vinju

Reputation: 6696

lexical Identifier = [a-z]+ !>> [a-z];

That helps for ambiguous lists of identifiers. The additional !>> constraint declares that identifiers are only acceptable if no further characters can be consumed.

Also this is required for fixing the parse error:

layout Whitespace = [\ \n\r]*;

For all syntax rules in scope it will intermix this nonterminal between all symbols. It leaves the lexical rules alone.

Upvotes: 1

Related Questions