Re: [Grammatica-users] Questions about Parsers

[D. R. E. Moonfire]

Per Cederberg wrote:
> Well, this is actually expected... :-)  Thing is, the tokenizer is completely 
> context-insensitive. So it generates a flow of tokens from your input chars
> all by itself without caring about the current production. And as the 
> CONSTRAINT token is placed above the IDENTIFIER token in the list of
> token definitions, it is chosen.
>
> The fix here, if you really mean that a constraint can be called CONSTRAINT,
> is to change the production:
>
> bob = CONSTRAINT (IDENTIFIER | CONSTRAINT) ;
>
> Or even nicer:
>
> Bob = CONSTRAINT String ;
>
> String = IDENTIFIER
>        | CONSTRAINT ;
>
> Hope this helps!

Oh, that helps, I just was hoping to avoid it. :) But, that is the 
nature of tokenizers, so I'm not going to complain too much. Apparently 
Oracle allows CONSTRAINT and DEFAULT as column names (parsing a 80k line 
DDL right now) along with a lot of other keywords. But, I'll just break 
out identifier and that should fix it. Thank you.