bug#12022: [PATCH] Unable to use anonymous functions for Key Translations

bug#12022: [PATCH] Unable to use anonymous functions for Key Translations

[Barry OReilly]

[-- Attachment #1.1: Type: text/plain, Size: 1248 bytes --]

> Using an anonymous function to translate keys as described in the Elisp
manual at
http://www.gnu.org/software/emacs/manual/html_node/elisp/Translation-Keymaps.htmldoes
not work as expected.
>
> Using a named function to map C-e to C-c, the Key Translation is
successful:
>   (progn (fset 'foo (lambda (prompt) (kbd "C-c")))
>          (define-key key-translation-map (kbd "C-e") 'foo))
>
> Using an anonymous function for another keymap binds C-e to the
lambda-defined command.
>   (define-key global-map
>               (kbd "C-e")
>               (lambda () (interactive) (message "Inside C-e's lambda")))
>
> But using an anonymous function for key-translation-map does not change
behavior of inputting C-e:
>   (define-key key-translation-map
>               (kbd "C-e")
>               (lambda (prompt) (kbd "C-c")))
>
> I find no reason why the last Elisp shouldn't define a Key Translation.
The Elisp manual section on Anonymous Functions states "Anonymous functions
are valid wherever function names are."  (
http://www.gnu.org/software/emacs/manual/html_node/elisp/Anonymous-Functions.html)

Hi,
I found where the definition of a key translation is called, and made a
patch.  The quoted test case works correctly now.  Is the change ok?

[-- Attachment #1.2: Type: text/html, Size: 1729 bytes --]

[-- Attachment #2: lambdaKeyTranslation.txt --]
[-- Type: text/plain, Size: 2198 bytes --]

# Bazaar merge directive format 2 (Bazaar 0.90)
# revision_id: gundaetiapo@gmail.com-20120812022358-9hf38s3yqrngmwj0
# target_branch: bzr://bzr.savannah.gnu.org/emacs/trunk/
# testament_sha1: 97ac7d396cfcbf692e8e88df42589ecba0fc8c70
# timestamp: 2012-08-11 22:24:03 -0400
# base_revision_id: jasonr@gnu.org-20120811145028-ax1qarn6z0mj5gtr
# 
# Begin patch
=== modified file 'src/keyboard.c'
--- src/keyboard.c	2012-08-08 15:53:52 +0000
+++ src/keyboard.c	2012-08-12 02:23:58 +0000
@@ -8837,7 +8837,8 @@
   /* If the keymap gives a function, not an
      array, then call the function with one arg and use
      its value instead.  */
-  if (SYMBOLP (next) && !NILP (Ffboundp (next)) && do_funcall)
+  if ((SYMBOLP (next) && !NILP (Ffboundp (next)) || FUNCTIONP(next))
+      && do_funcall)
     {
       Lisp_Object tem;
       tem = next;

# Begin bundle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Re: bug#12022: [PATCH] Unable to use anonymous functions for Key Translations

[Stefan Monnier]

> I found where the definition of a key translation is called, and made a
> patch.  The quoted test case works correctly now.  Is the change ok?

Yes, good spotting, thank you.  I installed your patch slightly
simplified (FUNCTIONP already does the Ffboundp and SYMBOLP checks).
Thanks,


        Stefan

Re: bug#12022: [PATCH] Unable to use anonymous functions for Key Translations

[Barry OReilly]

[-- Attachment #1: Type: text/plain, Size: 464 bytes --]

> I installed your patch slightly simplified (FUNCTIONP already does the
Ffboundp and SYMBOLP checks).

What I'm really doing involves the need for a closure, so I proceeded to
enable lexical binding.  This causes the test case to fail again.  I looked
into it and found that DEFUN("functionp", ...) checks EQ (car, Qclosure)
but FUNCTIONP does not.  We could add the same check to FUNCTIONP, but I'm
wondering if one of these functions should leverage the other?

[-- Attachment #2: Type: text/html, Size: 502 bytes --]

Re: bug#12022: [PATCH] Unable to use anonymous functions for Key Translations

[Stefan Monnier]

> What I'm really doing involves the need for a closure, so I proceeded to
> enable lexical binding.  This causes the test case to fail again.  I looked
> into it and found that DEFUN("functionp", ...) checks EQ (car, Qclosure)
> but FUNCTIONP does not.

Oops, sounds like a bug.

> We could add the same check to FUNCTIONP, but I'm
> wondering if one of these functions should leverage the other?

Yes.  The bulk of Ffunctionp should probably be moved to lisp.h into
a "functionp" inlinable function, then make FUNCTIONP an alias
for functionp and make Ffunctionp use functionp.


        Stefan