Re: Non-greedy wildcard possible? (Long)

[Magnus Lie Hetland]

Magnus Lie Hetland <address@hidden>:
>
[snip]
> Hm. I suppose so. I was worried that some rules might give me trouble;
> for example, if a rule ends with a wildcard, finding out which
> terminating token is in effect would require knowledge about the parse
> state which cannot be inferred statically (at least as far as I can
> see), because the terminating token belongs to another production.
> 
> But as long as I stay away from rules that end with wildcards, I guess
> the analysis is quite straightforward. (Are there other pitfalls?) And
> this was, in fact, the approach I considered earlier.

I remember some of my other conserns, now. Most importantly, the
possibility of empty productions is a problem. If a wildcard is
followed by a possibly empty production (as in "zero or more" of
something) I have to include the set of legal first tokens of the
following production into the set of terminating tokens of the
wildcard. Same thing happens if *that*, again, is empty, and so on.

I'm sure the analysis can be done automatically/mechanically, but it
doesn't seem exactly straightforward (unless the two problems I've
pointed out are the only ones -- somehow I doubt it ;)

-- 
Magnus Lie Hetland              "Wake up!"  - Rage Against The Machine
http://hetland.org              "Shut up!"  - Linkin Park