Re: Thoughts on Refactoring In-Buffer Completion In message.el

[Stefan Monnier]

> Why restrict it to a table? Perhaps we should allow both, functions and
> tables? Practically, that could mean checking whether the value
> satisfies `functionp`, and `funcall` it when that's the case; else use
> it as a ready-made table.
> It seems I am missing something?

A "completion table" can take several shapes, one of which is
a function.  But that function isn't used like a CAPF function.

E.g. `message-expand-name` is a CAPF function whereas
(message--name-table STR) returns a completion table (implemented as a 
function).

[ I'll get back to you later for the rest of your message.  ]


        Stefan