cannot create implicit token for string literal in non-combined grammar: '('

Question

I've broken up my combined grammar in two Lexer and Parser grammars:

The combined grammar is so easy like:

parser grammar Fluent;

fluent
 : MANDATORY
 | PARAMETRIZED '(' PARAMETER ')'
 | OPTIONS OPTION?
 | CHOOCING ( CHOOSE1 | CHOOSE2 )
 | PLUS+
 ;

MANDATORY : 'mandatory';
PARAMETRIZED : 'parameterized';
PARAMETER : 'parameter';
OPTIONS : 'options';
OPTION : 'option';
CHOOCING : 'choosing';
CHOOSE1: 'choise1';
CHOOSE2: 'choise2';
PLUS: 'plus';

Then I've created a FluentLexer.g4 with lexer rules and a FluentParser.g4 with parser rules.

Then I've imported FluentLexer grammar on FluentParser grammar.

I've tried two options:

With an import FluentLexer in Fluent Grammar: import FluentLexer
With an options structure:

options{ language = Java; tokenVocab = FluentLexer; }

Using the first one option ANTLR dumps me:

cannot create implicit token for string literal in non-combined grammar: '('

error(126): FluentParser.g4:7:30: cannot create implicit token for string literal in non-combined grammar: ')'

error(126): FluentParser.g4:3:12: cannot create implicit token for string literal in non-combined grammar: 'mandatory'

error(126): FluentParser.g4:4:15: cannot create implicit token for string literal in non-combined grammar: 'parameterized'

error(126): FluentParser.g4:5:12: cannot create implicit token for string literal in non-combined grammar: 'parameter'

...

It seems Lexer rules are imported.

with the second one option, ANTLR tells me only:

error(126): FluentParser.g4:10:16: cannot create implicit token for string literal in non-combined grammar: '('

error(126): FluentParser.g4:10:30: cannot create implicit token for string literal in non-combined grammar: ')'

Why do I need to define ')' and '(' in separated grammars, and why don't I need to define them on a combined grammar?

How could I solve that? I want to use '(' or ')' insted of a LEXER rule...

Bart Kiers Bart Kiers · Accepted Answer · 2015-12-28T09:34:43

It's a feature only available for combined grammars. Probably to help speed up development/prototyping. However, it's "better" to user lexer rules instead. See: this Q&A: Is "Implicit token definition in parser rule" something to worry about?

How could I solve that? I want to use '(' or ')' insted of a LEXER rule...

You can't in your case of separate grammars.

cannot create implicit token for string literal in non-combined grammar: '('

1 Answers