Question

I’m using antlr to analyse and re-write sql query.

I have:

select : SELECT ;

fragment S : 's' | 'S' ;
....
fragment LETTER : 'a'..'z' | 'A'..'Z' ;

SELECT : S E L E C T ;

IDENTIFIER : LETTER+ ;

to define reserved key words and let them to be case-insensitive.

My question is how can I define non-reserved key words?

Answer 1

Your problem seems similar to the problem we had when building the parser for the Drools (www.jboss.org/drools) language (DRL). In DRL, for instance, “rule” is a keyword, but could also be used by a java programmer as a property name in his POJO. So we can’t have that as a reserved keyword.

rule /*keyword*/ "my rule"
when
    SomeClass( rule /*property name*/ == "foo" )
...

We called these keywords “soft keywords”.

To do that in ANTLR, we defined only “true”/”false”/”null” as hard keywords in the LEXER:

https://github.com/droolsjbpm/drools/blob/master/drools-compiler/src/main/resources/org/drools/lang/DRLLexer.g#L132

Everything else is an ID. Then in the PARSER, we used semantic predicates for each soft keyword:

https://github.com/droolsjbpm/drools/blob/master/drools-compiler/src/main/resources/org/drools/lang/DRLExpressions.g#L597

This makes it possible to seamlessly integrate with java created POJOs without clashing property names and other things with Drools defined keywords.

Hope it helps.