[wg-camlp4] My uses of syntax extension

[Alain Frisch]

On 01/29/2013 11:19 AM, Fabrice Le Fessant wrote:
> I would strongly argue for a convention that has to be obeyed by all
> preprocessors. If you leave it to the preprocessor, you will have to
> call the preprocessor during the parsing, not during the typing, i.e.
> you won't have access to the typing environment while expanding macros.

If we extend the grammar with quotations, they can be rewritten by -ppx 
filters.  This is useful, and not incompatible with OCaml Templates 
(which can interpret remaining quotations, not rewritten by -ppx).  Many 
current uses of quotations really don't need access to typing 
information, and there are advantages in doing the expansion during 
parsing, not type-checking:

  - ocamldep can run the -ppx filters in order to get a precise idea of 
dependencies, after expansion has occurred.

  - OCaml Templates requires to load expanders in the same process as 
the type-checker, but this would force to use Dynlink in the compiler 
(or re-link statically you compilers with your expanders).  Dynlink in 
native code is not available on all platform.

  - An expansion done by a -ppx rewriter can benefit from other 
rewriters applied afterwards.  It is not clear if we should apply the 
-ppx rewriters to each fragment produced by OCaml Templates.


I'm not arguing against OCaml Templates, which has very interesting 
usage scenarios, but it should be thought as an addition to expanders 
implemented with -ppx, for those cases where type information is 
actually required and the downsides mentioned above are not problematic. 
  Of course, we should share the same quotation syntax.


Alain