[antlr-interest] Gated semantic predicates behaving like validating predicates

[Jim Idle]

Actually you are probably better matching the token letters that overlap 
in one rule, then changing the type in an action afterwards. Otherwise 
you are specifying the same sequence over and over again and duplicate 
the predicate and its converse and so on.

Jim

> -----Original Message-----
> From: Steve Bennett [mailto:stevagewp at gmail.com]
> Sent: Sunday, February 03, 2008 6:39 AM
> To: antlr-interest
> Subject: Re: [antlr-interest] Gated semantic predicates behaving like
> validating predicates
> 
> Dammit, I found the error. I didn't realise (or forgot) that gated
> predicates apply at the subrule level, not just the rule level.
> 
> ESCAPED_PUNC     : {this.in_noparse}? => ('#' | '*' | ':' | ';' | '|'
> | '[' | ']' | '=' | '/');
> 
> is correct.
> 
> ESCAPED_PUNC     : {this.in_noparse}? => '#' | '*' | ':' | ';' | '|' |
> '[' | ']' | '=' | '/';
> 
> is incorrect.
> 
> Ouch. What a lot of time spent on what turned out to be missing
> parentheses.
> 
> I have also learnt that this kind of construct is incorrect:
> 
> FOO: {flag}? => 'X';
> UNFOO: 'X';
> 
> You need to specify the reverse semantic predicate.
> 
> Anyway, thanks for listening everyone, you've been very supportive :)
> 
> Steve