Issue SELF-MODIFYING-CODE Writeup

Forum:		Public Review

Issue:		SELF-MODIFYING-CODE

References:	Moon's public review comment #14 (X3J13/92-3214)

                Section 3.2.2.1.1 (purpose of compiler macros), p 3-17

                Section 3.1.2.1.2.2 (macro forms), p 3-7

                Issue DEFINE-COMPILER-MACRO

                Issue MACRO-CACHING

Category:	CLARIFICATION/CHANGE

Edit history:	21 Dec 1992, Version 1 by Loosemore

		08 Jan 1993, Version 2 by Loosemore (comment from Moon)

Status:         Proposal FORBID accepted 7-2-0, March 1993



Problem description:


  The last sentence of the second paragraph of section 3.2.2.1.1 on page

  3-17 says compiler-macro expanders aren't allowed to modify the source

  code.  Is there a similar restriction on normal macro expanders?  I

  expect that there would be, but I couldn't find it.



Proposal (SELF-MODIFYING-CODE:FORBID):


  Add to section 3.1.2.1.2.2 a statement that the consequences are 

  undefined if a macro function destructively modifies any part of its

  form argument.



Rationale:


  This would make the specification more complete.



Current practice:


  Unknown.



Cost to implementors:


  None.



Cost to users:


  Conceivably this proposal might cause some existing conforming programs

  to become non-conforming, but we don't know of any such programs.



Aesthetics:


  I think everybody agrees that writing self-modifying code is a

  bletcherous practice.



Editorial impact:


  Trivial.



Discussion:


  Loosemore says:

   The DEFINE-COMPILER-MACRO proposal included the explicit prohibition

   against modification of the form argument because of the way a

   compiler macro function must return the original form to indicate it

   declines to expand it.  Ordinary macros don't have to deal with this

   problem, but it's generally accepted that self-modifying macros are

   bad for other reasons.  The problem description section of issue

   MACRO-CACHING, for example, touches upon this problem: "Caching by

   displacement won't work because the same (EQ) macro call form may

   appear in distinct lexical contexts.  In addition, the macro call form

   may be a read-only constant."


  Moon says:

   I agree with SELF-MODIFYING-CODE:FORBID (Version 1).


   This is consistent with the chucking-out of displacing macros from

   Common Lisp.