Re: LALR parser question

[Shark8 <onewingedshark@gmail.com>]

On Tuesday, April 30, 2013 3:18:56 PM UTC-6, Dmitry A. Kazakov wrote:
> On Tue, 30 Apr 2013 09:06:53 -0700 (PDT), Shark8 wrote:
> 
> 
> The way humans parse texts and consequently grammars of programming are
> uncorrelated to the classes of formal grammars.

That's part of why they're interesting.

> This is why formal grammars proved to be quite useless for compiler construction.

They're not the only thing that could be defined to be parsed w/ a grammar though -- a configuration formulation, or a DB-dump, or as part of a distributed system's serialize/deserialize specification could all benefit from being defined in a formal grammar (as opposed to an ad-hoc/"grown" method).

> Practically there is no language for which recursive descent parser would not work, provided that literals, identifiers and expressions are parsed by other means.

Perhaps; but this doesn't mean that recursive-decent is the best in every situation. After all, just because there is no collection of sortable items that can't be sorted via Bubblesort doesn't mean that that's the suitable method for the problem.

> Specifically, parsing Ada imposes no problems at all.

*nod* -- I didn't think there was. I was talking more on the general subject.