Re: [Caml-list] [Camlp4] Quotation expander with OCaml syntax

[Raphael Proust <raphlalou@gmail.com>]

On Tue, Jul 27, 2010 at 3:22 PM,  <Thomas.Gazagnaire@sophia.inria.fr> wrote:
> Are you trying to the same thing as HOP [1] ?

I haven't looked at HOP yet, but I heard a lot and will probably try to read
about it in the future.

>
> What kind of restrictions do you impose on the expressions you can
> transfer from the server to the client (only non-functional values) ? How
> do you ensure them in your program ?

We try to minimize restriction by providing users a way to define their own
[un]wrappers for exotic values. Such a value is (1) wrapped, (2) marshalled, (3)
sent to the client, (4) unmarshalled and (5) unwrapped. It is not possible to
send a value that has no associated wrapper (but their a basic wrapper for
generic values). The "weirder" type we currently support is probably ['a
React.E.t][1].

Types are currently checked (and inferred) via the wrap and unwrap function.

>
> Are you sure splitting the code into two parts is sufficient ? How will
> you handle branching than you can decide only at runtime, as example ? HOP
> is using a javascrip compiler embedded in a library to compile efficiently
> the right client code at runtime.

The client code is obtained with js_of_ocaml[2], an ocaml byte-code to
Javascript compiler. Code is compiled statically. Branching is handled by the
client at runtime.

>
> Btw, would be glad to have more details on what you have done and plan to
> do as I am working on similar things (not very actively currently) : I've
> been trying  to make camloo, a caml-light to bigloo compiler, working
> again (current version[2] is quite working - still few things to finish).
> At one point, the goal would be to extand the source language with similar
> constructs than yours  and to compile to HOP ...



[1] http://erratique.ch/software/react/doc/React
[2] http://ocsigen.org/js_of_ocaml
-- 
_______
Raphael