🚀 go-pugleaf

In an application I'm developing, it would be very useful to allow my
methods to have specializers different than the standard class or eql
specializers, but it's not clear whether the AMOP supports it.  In
particular, I would like to have a class of methods that can use
"AND" specializers, as in:

(defmethod foo ((self (and class1-mixin class2-mixin))) ...)

I need this because class1-mixin and class2-mixin can be used
independently of each other, but when used together require a
completely different algorithm for foo.  I could, of course, create
a class1-and-class2-mixin class that inherits both and does what is
necessary.  However, I'm trying to design my class hierarchy so that
users can build their classes out of whatever mixins they need, without
having to worry about their potential interdependencies (e.g. knowing
they have to include a class1-and-class2-mixin).

From what I can tell, the AMOP basically mentions nothing at all
about using these kind of alternative specializers.  It seems like
it *does* support it, or is close to supporting it, in that one would
only have to specialize COMPUTE-APPLICABLE-METHODS and
COMPUTE-APPLICABLE-METHODS-USING-CLASSES for the new method class
to work with the new specializer type.

My worry is that the specification of CLOS and the AMOP only
mentions specializers as being class or eql specializers.  In fact,
the text for DEFMETHOD on p. 839 of Steele Vol. 2 explicitly says:
"A parameter specializer name is a symbol that names a class or
(eql eql-specializer-form)."

Does this rule out what I want to do?  I can easily make the change
to support alternative specializers for the next version of July 92 PCL,
but is it supported by the AMOP (and other implementations)?

- Trent Lange
  lange@cs.ucla.edu