Re: Problem with multiple parsers

[Laurence Finston]

Hello,

I tried defining YY_DECL as you suggested, but I still couldn't get it to work. 
 The program is already fairly complex so it's not easy to test just the 
problem of getting multiple parsers to work together.  If I find the time, I'll 
try to set up a minimal example.

One problem is a `friend' declaration in a class declaration:

friend int yylex(YYSTYPE*, YYLTYPE*, yyscan_t);

(I'm generating C scanners and parsers and using them in a C++ program.)

Using `%name-prefix', this works for the other scanners (xxlex and zzlex), too. 
 That is, this single `friend' declaration works for all three scanner 
functions.

When I try to use `%define api.prefix', I get various errors pertaining to 
YYSTYPE, YYLTYPE, YY_DO_BEFORE_ACTION, YY_NEW_FILE, etc.  I've tried using 
`typedef' and `#define' to get rid of the errors, but nothing I've tried so far 
works.

Would it be possible to "undeprecate" `%name-prefix'?  I've been using it for 
several years now and have been happy with the results.  `%define api.prefix' 
doesn't work the same way and if `%name-prefix' were to disappear, upgrading to 
a new version of Bison would break my code.  I dont know how many other people 
would be affected by this.

Thanks,

Laurence




> Gesendet: Donnerstag, 29. Januar 2015 um 21:15 Uhr
> Von: "Akim Demaille" <address@hidden>
> An: "Laurence Finston" <address@hidden>
> Cc: "Bison Bugs" <address@hidden>
> Betreff: Re: Problem with multiple parsers
>
> 
> > Le 29 janv. 2015 à 10:32, Laurence Finston <address@hidden> a écrit :
> > 
> > Hello,
> > 
> >> What version of Bison are you using?
> > 
> > 2.6.5 and 3.0.2.  I plan on installing 3.0.4 before too long.
> 
> 3.0.2 is close enough :)
> 
> >> prefix I guess.
> > 
> > No, I actually meant a suffix, e.g., MESSAGE_YY, MESSAGE_ZZ, MESSAGE_XX for 
> > tokens in three different parsers.  I was assuming that the tokens were 
> > defined as preprocessor macros.  I think this was once the case, but I'm 
> > not completely sure.
> 
> Yes it was, but it also depends on the skeleton you
> use.
> 
> >  At any rate, I've now checked and found that they are not now.
> 
> Well, indeed, it's probably not needed, and if you really
> want something like that, I would recommend checking
> api.token.prefix.
> 
> http://www.gnu.org/software/bison/manual/html_node/_0025define-Summary.html
> 
> 
>