Re: [PATCH] master: doc: token.prefix

[Joel E. Denny]

Hi Akim.

On Thu, 14 May 2009, Joel E. Denny wrote:

> On Thu, 14 May 2009, Akim Demaille wrote:
> 
> > Something slightly unrelated: I think "%define variables" is not pleasant to
> > read.
> 
> I believe I wrote it that way for symmetry with "%code qualifiers", but 
> the user probably doesn't care about that connection.
>
> > Unless I'm mistaken there are no other variable kinds in the Bison
> > world, so just "variable" seems enough to me.  I would not be surprised that
> > playing with -D might trigger messages about "%define variables", although
> > %define was not explicitly used.
> 
> Agreed.  I'll work on that eventually unless you get there first.

On second thought, I'm not sure "variables" is clear.  There are also 
variables in the literal code and semantic actions in the grammar file, 
and there are more variables in the generated code.  I fear the manual and 
NEWS are going to become confusing if we don't distinguish between these.

"%define variables" is the clearest name I can think of especially while 
these variables are still a new concept in Bison.  Users will hopefully 
equate -D with %define, so the error messages for -D might be clear enough 
as they are.

> > Maybe "Bison variables" if you think there are risks of confusion.
> 
> Off the top of my head, I don't see a need, but I'll keep it in mind.

I'm not sure "Bison variables" is any clearer.  For example, it's easy to 
think of yylval as a Bison variable as opposed to a user variable.