Excerpts from Hendrik Tews's message of Fri Apr 04 23:30:58 +0200 2008:
> "Nicolas Pouillard" <nicolas.pouillard@gmail.com> writes:
> 
>    I think that's doable (not tested), basically you need three
>    primitive functors:
> 
> I don't see how. Let me clarify again what I mean. And let me
> clarify that I only pursue this discussion out of curiosity, to
> see, what can be done with the functorial Camlp4 interface. 
> 
> Assume we have a table like this:
> 
>     s1.cmo s2.cmo ... : f1.ml f2.ml ...
>     s3.cmo s4.cmo ... : f3.ml f4.ml ...
> 
> The sx.cmo files are Camlp4 syntax extentions and the fx.ml are
> files to be parsed using those syntax extentions. There are a
> number of possible solutions:
> - starting camlp4 for each file or each row
> - clearing PreCast.Syntax before each row and then calling the
>   callbacks that the syntax extensions registered
> 
> But I would like to know if the functorial Camlp4 design is
> strong enough to solve this exercise within one program (without
> fork/exec) and without reusing/clearing one syntax module.
> 
> It is clear that the registration in Register.SyntaxExtension is
> not good enough, because it always applies the syntax extension
> to PreCast.Syntax.
> 
> However, even with changing Register.SyntaxExtension or how
> syntax extensions register themselves, I can't see a solution.

That's  some  corner  case where first-class modules would help a lot, however
one  can  use some black magic under the hood to get the same effect. One need
to  store  extensions once and for all, in a map from names to "modules". Then
one  can  have  some function (functor) that takes a list modules names and an
initial module, and will apply these extensions to it.

-- 
Nicolas Pouillard aka Ertai