On 05/04/2009, at 12:26 AM, Emmanuel Onzon wrote:
>
>
> 2009/4/4 Emmanuel Onzon <e.on...@gmail.com>
> 2009/4/4 john skaller <skal...@users.sourceforge.net>
>
>
> On 05/04/2009, at 12:09 AM, Emmanuel Onzon wrote:
>
>
> Dyp.Undefined_nt("comma")
>
> PARSE ERROR
> Unknown exception Parsing File
>
>
> U mean like this? but I still don't understand how "comma" can be
> undefined..
>
> Yes, I mean this exception.
> comma is well defined, but dypgen believes it is not.
>
> Why? It is even listed:
>
> Basically, it is because the way dypgen decides to raise
> Undefined_nt is ill-designed and must be changed.
Exceptions suck .. a good idea to eliminate almost all of them
and return a sum type instead, IMHO.
>
> I'm going to work on that for the next version.
> dypgen decides to raise Undefined_nt if
> it finds in the extension a non terminal that is only in a
> right hand side and not in a left hand side, disregarding
> whether it is in a left hand side in the initial grammar.
> You can call it a bug :)
Indeed :)
BTW: I get this processing the initial grammar:
Warning: non terminal `strue' is never in a right-hand side.
I would like to *selectively* suppress these, say by writing
@strue = { ... }
I.e. not by a global flag to turn the warning off, but a way
to say "yes, this nonterminal is intended to be used by extensions"/
With that, Dypgen could throw out unused productions. This is useful
if you have a large file of definitions to be used as a "library" which
you include then define something using only a few of the symbols.
Actually that suggests:
%keep strue
as a better syntax, since it would allow defining not only
something based on a subset of a library, but also providing
extensions with a set of non-terminals, but still eliding others
that aren't significant.
With this kind of "garbage collection" together with some namespace
management (some kind of structured names for non-terminals
resembling say X.Y .. perhaps X'Y since that's a legal Ocaml identifier)
this would make grammar libraries more viable, particularly if
they could be saved "pre-compiled".. both these things are possible
because Felix already does both of them :)
>
> The build system is now doing this:
> [...]
> checking build/test/regress/rt/rt-1.01.30-0: ok
> * build/bin/flx.py --test=build : build/test/regress/rt/
> rt-1.01.31-0.flx -> build/test/regress/rt/rt-1.01.31-0.so
> checking build/test/regress/rt/rt-1.01.31-0.so: ok
> * build/bin/flx.py --test=build : build/test/regress/rt/
> rt-1.01.31-0.flx -> build/test/regress/rt/rt-1.01.31-0
> checking build/test/regress/rt/rt-1.01.31-0: ok
> * build/bin/flx.py --test=build : build/test/regress/rt/
> rt-1.01.32-0.flx -> build/test/regress/rt/rt-1.01.32-0.so
> checking build/test/regress/rt/rt-1.01.32-0.so: ok
> * build/bin/flx.py --test=build : build/test/regress/rt/
> rt-1.01.32-0.flx -> build/test/regre
> [...]
Running regression tests :)
Hmmm .. that doesn't have my patch to "Bind_to_cons" all the RHS non-
terminals ..!! I haven't
committed that yet! It works anyhow .. hmm :)
--
john skaller
skal...@users.sourceforge.net
------------------------------------------------------------------------------
_______________________________________________
Felix-language mailing list
Felix-language@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/felix-language
