Issue SETF-FIND-CLASS Writeup

Issue:         SETF-FIND-CLASS


References:    FIND-CLASS in 88-002R


Category:      CLARIFICATION


Edit history:  29-Apr-90, Version 1 by Moon

               30-Apr-90, Version 2 by Moon (update current practice)

               2-May-90, Version 3 by Moon (remove incorrect reference

                     to Loosemore issue #7 of 27 Feb 90)

               4-May-90, Version 4 by Moon (update discussion and current

                                            practice)


Problem description:


  Is (SETF (FIND-CLASS 'FOO) NIL) permitted as a way to break the link from

  a name to a class?  I can't find anything in the CLOS spec for or against

  this.


  This is Symbolics issue #27.


Proposal (SETF-FIND-CLASS:ALLOW-NIL):


  Define (SETF (FIND-CLASS 'FOO) NIL) to cause FOO no longer to name a class.

  If FOO already does not name a class, the operation has no effect.


  This does not affect the class formerly named by FOO, if any, except that

  if FOO was that class's proper name, the class no longer has a proper name.


Rationale:


  This is simpler than putting in a new function to break the association

  between a name and a class.  When CMAKUNBOUND was removed from an early

  draft of 88-002R, it was probably intended to add this SETF of FIND-CLASS

  feature to replace it, but the document was accidentally not updated.


Current practice:


  Symbolics Genera 8.0, TI Explorer release 6.0, and Lucid 4.0.0 Beta-1

  implement the proposal.  Other CLOS implementations were not surveyed.


Cost to Implementors:


  Easy.


Cost to Users:


  None.


Cost of non-adoption:


  There will be no portable way to break the association between a name and

  a class.


Performance impact:


  None.


Benefits:


  Adds a useful feature that was probably intended to be there all along.


Esthetics:


  More complete language.


Discussion:


  Gregor Kiczales says he supports SETF-FIND-CLASS:ALLOW-NIL.