RE: Any objection to adding completing-read-function?

[Drew Adams]

1. Leo's original patch said this near the beginning of the doc string for
`completing-read':

"This function calls `completing-read-function' to do the work"

That's very clear.  What finally got implemented removed that explanation and
just added this to the very end of the (very long) doc string:

"See also `completing-read-function'."

That's insufficiently clear, IMO.  There is a big difference between the two in
terms of communication to users.  The former makes clear what
`completing-read-function' is about - its role for `completing-read'.  The
latter relegates `completing-read-function' to a "see also".

I have the same comment wrt `read-file-name'.  IMO it would be preferable to say
that it calls `read-file-name-function' to do the work, instead of just adding a
"see also" at the very end.


2. I would also prefer it if the code for `completing-read-function' and its
treatment were in Lisp, not C.  No, I won't be submitting a patch. ;-)  Dunno
how easy it would be to move this to Lisp, but IMO it would be preferable if it,
just like `read-file-name-function', were in minibuffer.el.  These two variables
are parallel.

At this stage in the game it doesn't seem like we should be adding completion
code to the C sources.  Ideally we should be moving the last remaining pieces of
the completion code (e.g., `completing-read') completely to Lisp instead.  How
about putting this on the TODO list?