Re: [Caml-list] Re: Syntax extensions without Camlp4

[Gerd Stolpmann <info@gerd-stolpmann.de>]

Am Donnerstag, den 31.05.2012, 19:03 +0100 schrieb Wojciech Meyer:
> Gerd Stolpmann <info@gerd-stolpmann.de> writes:
> 
> > Am Donnerstag, den 31.05.2012, 12:01 -0400 schrieb bob zhang:
> > I guess you are missing Alain's point here. He suggests to use normal
> > pattern matching directly on parsetrees, as in
> >
> > match expr with
> >  | Pexp_ident id -> Pexp_ident (tranform id)
> >  | Pexp_contant c -> Pexp_constant c
> >  | ...
> >
> > instead of using camlp4 syntax quotations (no <<...>> stuff for syntax
> > manipulation anymore).
> 
> I admit quotations are *way* better to construct the code, they are
> syntax aware but also reassemble string manipulation at the same
> time. If somebody has generated some more serious program using
> quotations then she can't stand using normal way by constructing via the
> data constructors. I can compare, because I have used both ways for the
> same project. Of course normal ASTs have it's wide applications, what is
> shown above is a transform, and not strictly a code generation - usually
> I have intermediate AST on the left hand side pattern match and on the
> right hand side quotations. So I can't say quotations are not useful,
> they are quite much. If the quotations could be as lightweight as
> brackets-escape notations of MetaOCaml it would be brilliant. (I'm
> really talking about code generation - and not extending the syntax).
> 
> Please see [1].
> 
> I'd support the idea of having quotations for the parse trees.

I'm wondering whether you could run the normal ocaml parser in a special
mode for constructing code. In this mode, all non-keyword identifiers
(and possibly other placeholders) can also be given in antiquotation
form $name$, and the values for such injected names are taken from a
table you have to supply when calling the parser in this mode. This
would then look like

let antiquot_map =
  StrMap.add "var" "x" StrMap.empty

let ast =
  parse_struct_item antiquot_map "let $var$ = x+1"

And, given quotations are available in a renovated parser, one could
even write <:struct_item< let $var$ = x+1 >>, which already looks very
camlp4-ish (but would not use camlp4 at all), and would just be a thin
layer on top of the standard parser.

I mean the whole point is that we can do a lot without camlp4 by just
pimping the normal parser a bit. Actually, we need camlp4 only for
extensible grammars and the revised syntax, and Alain is probably right
that this is not worth the effort.

Gerd
-- 
------------------------------------------------------------
Gerd Stolpmann, Darmstadt, Germany    gerd@gerd-stolpmann.de
Creator of GODI and camlcity.org.
Contact details:        http://www.camlcity.org/contact.html
Company homepage:       http://www.gerd-stolpmann.de
*** Searching for new projects! Need consulting for system
*** programming in Ocaml? Gerd Stolpmann can help you.
------------------------------------------------------------