[antlr-interest] Keywords as identifiers in ANTLR 3.0

[Gavin Lambert]

At 00:20 30/08/2007, Ayende Rahien wrote:
>Thanks for the help. I do have the need to handle select (and 
>other SQL elements).
>How can I handle that?

I answered that one in the last paragraph :)  The way I prefer is 
to do something like this:

1. Create a new parser rule 'identifier' that accepts IDENTIFIER 
as well as any of the (non-fragment) keyword tokens that can also 
potentially be identifiers when used out-of-context.
2. Change any existing parser rules to use 'identifier' instead of 
'IDENTIFIER'.

Incidentally, a different approach you might want to take for the 
original "order by" issue is to convert ORDER_BY into a parser 
rule and make both ORDER and BY non-fragment rules.  That way you 
don't need to worry about the whitespace any more (it will have 
already been removed by that stage), and (combined with the above) 
makes it a bit easier to treat every keyword as an identifier when 
used out of place.

Are you really sure you need to do this, though?  I didn't think 
regular SQL permitted this sort of thing, since it leads to evil 
typo behaviour.