RE: [Sml-implementers] Could raise be non-expansive?

[Robert H. <Rob...@cs...>]

Yes it can, and it should.  The name "non-expansive" was already obsolete
and misleading by the SML 97 revision, but was retained for continuity's
sake.

This brings up a related question of whether other expressions that may
raise an exception are to be non-expansive.  A typical example is the
binding

	val x as _::_ = nil

whose right-hand side is deemed "non-expansive", yet the binding for x is
uncons(nil) (in the obvious sense), which may raise an exception.  The
binding for x should be considered non-expansive, and hence non-polymorphic.

More generally, for val bindings whose pattern is refutable (ie, there
exists a value not matching it), the variables involved should not be
generalized.  For irrefutable patterns, they should be generalized iff the
rhs is syntactically non-expansive in the current sense (which is, of
course, conservative).

Bob

-----Original Message-----
From: Dave Berry [mailto:da...@ta...]
Sent: Wednesday, September 05, 2001 5:13 PM
To: sml...@li...
Subject: [Sml-implementers] Could raise be non-expansive?


Am I right in thinking that the definition of non-expansive expressions
could be extended with "raise nexp", with no ill-effects?  I'm not
suggesting that such a minor change is worthwhile; I'm just curious.

Dave.

_______________________________________________
Sml-implementers mailing list
Sml...@li...
https://lists.sourceforge.net/lists/listinfo/sml-implementers