Re: [COMMIT] Autoload within #'keymapp, as documented.

Aidan Kehoe writes:

 > An alternative would be what #'functionp does, which is return t for
 > autoload objects without loading them.

Sure, but functionp objects only have one other generically useful
API: call them, and that will load them.  Anything else is an
extraordinary case, and I think it's reasonable to ask those cases to
deal (since they're going to have to special-case like crazy if
they're going to introspect the objects that are returned).

keymapp objects have several useful APIs, and they are less likely to
*require* a ton of other stuff in such uses (ie, it would usually make
sense to put the keymap(s) in a separate foo-maps.el library so only
the keymaps would get loaded by #'keymapp).  So I think it's
reasonable to load the keymap library in #'keymapp rather changing all
the other code.

 > But, it’s currently documented as autoloading and that’s what the
 > other code assumes.

Hm.  I guess there's a possibility that code could assume something is
a keymap and invoke an API that only works on loaded keymaps without
checking for keymapp first.  Maybe there is something to discuss (ie,
auditing all the low-level calls for whether they should autoload).

Not your problem, your patch is an improvement for the current code
and nobody's complained that I can recall, so commit it.  If you care
to come back and audit it at some point, it's issue871, currently
unassigned.



_______________________________________________
XEmacs-Patches mailing list
[email protected]
http://lists.xemacs.org/mailman/listinfo/xemacs-patches