Re: bison for nlp

[r0ller]

Hi Akim,

Thanks for everything again. One thing came into my mind concerning your 
question if there's any feature of bison I miss (though this may not belong to 
the scope of bison but another tool): a kind of reverse mode would be great:) I 
mean, generating the sequence of tokens from a grammar. I'm currently using 
nltk to test what my bison generated parser would accept: 
http://www.nltk.org/howto/generate.html

Or are you aware of any such c/c++ tool?

Best regards,
r0ller
 
-------- Eredeti levél --------
Feladó: Akim Demaille < address@hidden (Link -> mailto:address@hidden) >
Dátum: 2018 november 23 19:13:10
Tárgy: Re: bison for nlp
Címzett: r0ller < address@hidden (Link -> mailto:address@hidden) >
 
Hi r0ller,
> Le 21 nov. 2018 à 11:07, r0ller <address@hidden> a écrit :
>
> Hi Akim,
>
> Thanks again! After your answer I felt pretty dumb as I had to realize that 
> the problems I mentioned came most probably due to my carelessness.
:) No worries.
> This is actually one thing that could be pointed out in the Simple C++ 
> Example.
Well, I should probably not write « pay attention to the details » :)
> The other thing that could be added to the C++ part is a Simplest C++ 
> (migration) Example for those (like me) who'd migrate from a C parser using 
> ints just to show that there's almost nothing to do to get a c++ parser. The 
> Simplest C++ Example is fine for those who have a C parser with unions as far 
> as I can see.
Which is the typical case! It is meant to be an introductory example.
> To answer your question concerning using ints: it has historical reasons. 
> When I started the project (with yacc) I was not familiar with parser 
> generators (and am still not as knowledgeable as I’d like to be) so not 
> knowing what is possible and what not, I created a simple parser design and 
> coded the rest in c (later c++) to get the functionality I needed.
Ok.
> But the variants will be a great help once I manage to fulfill all the 
> prerequisites as the migration path for me has just begun:) I don’t want to 
> turn everything upside down in one step.
That’s reasonable :) But some steps cannot be done gradually (like changing the 
semantic value type).
> So I first need to get rid of numbering the tokens and then I'll see what 
> comes afterwards. Anyway, now I'm already happy that I have a c++ parser and 
> 5 warnings less when compiling the project. I'm pretty sure that I'll come 
> back with questions during the migration. By the way, I’ve recently added 
> English support to the Android app so if you check out the project page and 
> install it (for free) via the link pointing to the play store, you'll be able 
> to search for contacts and make phone calls even offline via voice control 
> having a bison parser running under the hood:)
That’s great :) But I don’t run Android.