On Wed, 24 Oct 2007, Waldek Hebisch wrote:

| 
| Gabriel Dos Reis wrote:
| > 
| > BTW, a general approach I have been working on for some time now is to
| > have a domain ParseForm, for parse forms i.e. parse trees after they
| > have been property annotated, at the Spad level, and define a
| > protocol to construct new entities out of ParseForms.  This ParseForm
| > domain is different from InputForm (which represents only expressions).
| > That way people can extend the interpreter in ways unimagined by
| > OpenAxiom developers, and move lot of code out of the interpreter itself.
| > The tricky part, of course, is to nail down the protocol so that it is
| > both useful and safe enough.
| > 
| 
| I wonder how do you want to handle typechecking:

That is part of the protocol.

| is ParseForm intended
| to be essentially untyped representation which is passed instead of
| strings to evaluator (which annotates it with types)?

It is a a partially typed abstract syntax tree (like InputForm), and
operations that take ParseForm are responsible to conduct the typing
and return something of a domain that be used for further typing.

|  Or is ParseForm
| (including its evaluation) intended to obey normal Spad type rules?

-- Gaby


_______________________________________________
Axiom-math mailing list
[email protected]
http://lists.nongnu.org/mailman/listinfo/axiom-math

Reply via email to