Re: push parser

[Joel E. Denny]

On Mon, 25 Jun 2007, Bob Rossi wrote:

> Section 3.7.10 describes the pure parser. I think it would be
> appropriate to describe the push parser in section 3.7.11.

Seems fine.

> In section 4.2.4 the manual describes the "Calling Conventions for 
> Pure Parsers". Off the top of my head, I don't think the push parser
> effects this particular mode so I don't think we would need to change
> this section. Do you?

I believe all the information in this section is correct even if you use 
%push-parser or %push-pull-parser.  Of course, we'll want to document the 
effect of %pure-parser in those cases, but that documentation probably 
belongs in your new section 3.7.11.

> In the section "Appendix A Bison Symbols", some variables are described.
> It states when certain variables will be global, so that the user could
> potentially get access to them.

For a pure push parser, yynerrs is a member of yypstate, but yychar, 
yylloc, and yylval are local variables in yypush_parse.  So, that 
documentation needs to be updated.  There may be other places besides 
Appendix A.

For an impure parser, regardless of whether it's push, all four of those 
variables are global.

> In a few of these, it describes the
> difference when the push parser option is used. I can't really remember
> how you eventually got the push parser working. What are the valid
> combinations of options with pure/push/normal parsing? Should we
> describe them all in a manual?

The user can declare either %push-parser or %push-pull-parser.  If he 
declares both, the last one declared has precedence, but I'm not sure 
that's permanent behavior.  The user should just declare one or the other.

You can use %pure-parser in combination with either %push-parser or 
%push-pull-parser.

Actually, we would get automatic checking for conflicting push 
declarations if we converted %push-* to a single %define.  Maybe:

  %define push_pull "pull" // default
  %define push_pull "push"
  %define push_pull "both"

I even wonder if %pure-parser should be deprecated and we should add:

  %define pure_parser

Bison would then complain automatically when any of these declarations 
have no effect for the chosen skeleton.

> One other question, is it important to have an example in the manual
> showing how the push parser works?

I think that would be great.

Thanks for working on this.