Debugging grammar

[Šimerda]

Hello folks,

I'm a relatively new user of modgrammar and a *potential* Fedora
maintainer of a new modgrammar package.

Thanks for this very nice work... the only thing I need in order to
use modgrammar, is a way to debug. I was testing some grammars and it
happens to me that particular grammars don't actually parse (it takes
longer than I would wait).

I'm looking for a way to debug modgrammar parsing to see what's going
on and why backtracking
takes bad tracks in some cases. Do you have any suggestions?

Pavel

[Alex Stewart]

Hey Pavel,

Sorry about the delay responding.. been a busy week..

Unfortunately, debugging modgrammar can be tricky and it's something I haven't really worked out a good set of tools for yet..  The good news is that, I believe, at this point the modgrammar parser shouldn't have a lot of cases where it's too hard to figure out what's going on with a little careful thought about the input grammars..

If your parsing is not coming back within a reasonable time, the odds are very likely that it's because you have some left-recursion somewhere in your grammar.  This will cause an infinite parsing loop, and modgrammar will just keep chugging away indefinitely (until eventually it exhausts the python interpreter's stack depth and crashes with a huge traceback, but that can take quite a while).

(Left-recursion is where you have a grammar whose first (leftmost) element can (directly or indirectly) refer to itself, so you get something like "grammar A could start with grammar A, which could start with grammar A, which could...", etc.  This is one of the only regular grammar constructs that modgrammar doesn't currently support.)

Anyway, that would be my advice for the first thing to look for, as it's the only thing I know of which would cause modgrammar to not return from parsing fairly quickly.  If that isn't the case, maybe you could post (or send me directly) your problematic grammar definition and I could help you try to figure out where the problem is..

I have been meaning to improve the debugging/troubleshooting aspects of modgrammar, but because of the complexity inherent in the internal parsing process it's very difficult to come up with tools that provide information that's really useful to anybody who isn't intimately familiar with the parser internals (and even then often not).  If you have any suggestions, though, for things which would make troubleshooting modgrammar easier, please let me know and I can try to figure out whether there's a way to include them..

I'm thinking I might also start some documentation on troubleshooting tips, if I can come up with some useful stuff to put in it..

--Alex

PS: Regarding the Fedora package, if you'd be willing to do that that would be great :)  I'd be happy to include any spec file or source updates you want to contribute to support that into the main source distribution as well, just let me know.