Re: POSIX and reduce/reduce conflicts

[Hans Aberg]

At 17:50 -0700 2002/04/09, Paul Eggert wrote:
>So POSIX does not allow precedence or associativity to count for
>reduce/reduce conflicts.
>
>> But if some day Bison supports other paradigms, such as back
>> tracking, then precedence can be seen as a means to order the
>> search, which can be good.
>
>Yes, but I would wait for those other paradigms before worrying about
>this.

I think that one needs to create the underlying theory, perhaps via the
ref's I gave, before one can use the token precedence rules for something
wider.

If one should leave the idea of not only using LALR(1), the one wants a
language specification L(G, X), where G is a grammar and X other grammar
related specifications, but which does not include the algorithm itself.
That way, one can ensure that changing the algorithm does not alter the
parsed language.

Of course, algorithm dependent features, like %expect, can help the
implementation of a language, and if one has not intent to change the
algorithm, one needs not to worry about the picture above:

But I think that this is the picture: One needs to isolate a specific core
of only language dependent features, separating it from algorithm dependent
features, which are add-on's to that picture.

  Hans Aberg