[Gcl-devel] cons types in the Baker algorithm

[Camm Maguire]

Greetings!  Your very helpful paper has provided the basis for GCL's
type calculations we you may recall discussing previously.  I
undertand it was written before cons types were invented.  These have
some limited use in the current GCL compiler to carry list length
information to allow apply to be converted to funcall for short final
lists.  Right now GCL implements cons types as effectively outer joins over
general types, and uses the same orthogonality concept in your paper
to do its computations.

One thing we never got working were types of the form '(member (1
. 2)), which of course is a subset but not equal to `(cons (member 1)
(member 2)).  Is there any hopefully elegant way of representing this
in the orthogonal system you conceived?

The goal is to defer constant list detection in the same fashion as
above by carrying the information in the type.

Take care,
-- 
Camm Maguire                                            address@hidden
==========================================================================
"The earth is but one country, and mankind its citizens."  --  Baha'u'llah