Issue: GET-MACRO-CHARACTER-READTABLE (version 1)

[Jon L White <jonl@lucid.com>]

Issue: GET-MACRO-CHARACTER-READTABLE (version 1)

References:  CLtL p.361: COPY-READTABLE, SET-SYNTAX-FROM-CHAR, and
                         GET-MACRO-CHARACTER
             CLtL p.364: GET-DISPATCH-MACRO-CHARACTER, 
             CLtL p.378: Example in middle of page

Category:    CLARIFICATION/CHANGE

Edit history:  Version 1, 16-Nov-88, by JonL


Problem description:

The 'readtable' argument to GET-DISPATCH-MACRO-CHARACTER and to
GET-MACRO-CHARACTER must be of type READTABLE; this may have been simply
an oversight, since it makes more sense for it to refer to values from
the standard readtable.  Both COPY-READTABLE and SET-SYNTAX-FROM-CHAR
explicitly say that a NIL in the 'from-readtable' argument refers to the
standard readtable.  Also, an example in the middle of the page, CLtL
p.378, supplies a NIL to GET-MACRO-CHARACTER, and is clearly intending
to access the standard readtable values.


Proposal (GET-MACRO-CHARACTER-READTABLE:NIL-STANDARD)

Specify that the 'readtable' argument to GET-DISPATCH-MACRO-CHARACTER
and to GET-MACRO-CHARACTER mean the same thing it does for COPY-READTABLE,
and SET-SYNTAX-FROM-CHAR; namely a reference to the standard readtable.  
Thus (GET-MACRO-CHARACTER <char> NIL) would be equivalent to 
(GET-MACRO-CHARACTER <char> (COPY-READTABLE)), but without consing.

Test Cases:

(let ((standard-rt (copy-readtable))
      (chars '(#\* #\= #\| #\A #\  #\( #\# #\1)))
  ;; Test Case 1
  (dolist (char chars)
    (assert (eq (get-macro-character char nil)
                (get-macro-character char standard-rt))
            () "Lose on character ~C" char))
  ;; Test Case 2
  (dolist (char chars)
    (assert (eq (get-dispatch-macro-character #\# char nil)
                (get-dispatch-macro-character #\# char standard-rt))
            () "Lose on #\# dispatch character ~C" char))
  ;; Test Case 3
  (assert (eq (get-dispatch-macro-character #\# #\A nil)
              (get-dispatch-macro-character #\# #\a nil))
          () "Lose on #\# dispatch character ~C" char)
 )

Rationale:

Probably was the original intent; somebody wants it; also see "Esthetics".

Current practice:

Symbolics and Xerox have already implemented the proposal; Lucid, VAXLISP,
and KCL stuck to the more rigid interpretation.


Cost to Implementors:

Trivial.

Cost to Users:

None.

Cost of non-adoption:

Minor worry about porting between implementations that support the
generalization and those that don't; minor worry about consing when
calling (COPY-READTABLE) to get at standard readtable semantics.

Performance impact:

See "Cost of non-adoption".

Benefits:

See "Cost of non-adoption".

Esthetics:

Increases parallel design among similar readtable functions.

Discussion:

This question was raised in the Common Lisp mailing last summer:
    Date: 19 Jul 88 13:35
    Subject: Question about readtable null arguments
    From: quiroz%cs.rochester:EDU:Xerox
    To: common-lisp%sail.stanford:EDU:Xerox