On 12/01/22 16:48 PM, Alexander Adolf wrote:
> Hello Eric,
>
> Eric Abrahamsen <eric@ericabrahamsen.net> writes:
>
>> [...]
>> In fact this whole message-mode thing is an impossible tangle, burritos
>> within burritos, and it would be great to get rid of it altogether.
>> [...]
>> So I need to clobber `message-expand-name' altogether.
>>
>> And EUDC having a function on `completion-at-point-functions' is
>> wrapping yet another burrito outside the existing burritos, because now
>> EUDC has a completion function both inside and outside message-mode's
>> own completion machinery.
>>
>> In fact the docstring of `eudc-capf-message-expand-name' makes it sound
>> like it thinks it's being called as part of `message-expand-name', but
>> now it isn't -- it's being called as part of the capf machinery. Or
>> rather, it could potentially be called in both places.
>>
>> I think a half-stick of dynamite is the only real solution.
>> [...]
>
> Perhaps we can be slightly more CONstructive that this. ;-)))
>
> I am preparing a patch to message.el which refactors
> `message-completion-alist` along the lines of this:
>
> ---------------------------- Begin Quote -----------------------------
> (defcustom message-completion-alist
>   `((,message-newgroups-header-regexp . (:category newsgroup
>                                          :fieldsep-re "\\([:,]\\|^\\)[ \t]*"
>                                          :completions ,#'message-expand-group))
>     (,message-email-recipient-header-regexp . (:category email
>                                                :fieldsep-re "\\([:,]\\|^\\)[ \t]*"
>                                                :completions ,#'eudc-capf-message-expand-name)))
>   "Alist of (RE . RECIPE), defining completion contexts.
> This variable controls how `message-completion-function' performs
> in-buffer completion.  RECIPE is either a function (deprecated),
> or a plist.
>
> When `message-completion-function' is invoked, and point is on a
> line matching one of the REs in the alist, the settings in the
> corresponding RECIPE are applied.
>
> When RECIPE is a function, it is called for completion.  RECIPE
> should be a function that obeys the same rules as those of
> `completion-at-point-functions'.
>
> When RECIPE is a plist, the properties control how in-buffer
> completion is performed.  The following properties are currently
> defined:
>
> :category
>
>     The symbol defining the category in
>     `completion-category-defaults' to use for completion.  Also
>     see `completion-category-overrides', and `completion-styles'.
>
> :fieldsep-re
>
>     The regular expression to use when scanning backwards in the
>     buffer.  All text between point, and any preceding text
>     matching this regular expression, will be used as the prefix
>     for finding completion candidates.
>
> :completions
>
>     The function that provides completions, and that obeys the
>     same rules as those of `completion-at-point-functions'.
>     In-buffer completion will be performed as if
>     `completion-at-point-functions' had been set to this value."
>   :version "29.1"
>   :group 'message
>   :type '(alist :key-type regexp
>                 :value-type (choice (plist)
>                                     (function
>                                      :tag "Completion function (deprecated)"))))
> ----------------------------- End Quote ------------------------------
>
> As you can see, `eudc-capf-message-expand-name` effectively replaces
> `message-expand-name` altogether:
>
> ---------------------------- Begin Quote -----------------------------
> (make-obsolete 'message-expand-name 'eudc-capf-message-expand-name
>                "29.1")
> ----------------------------- End Quote ------------------------------
>
> The patch goes on to remove everything relating to ecomplete,
> mailabbrev, bbdb, and the likes from message.el, too. Get rid of all the
> burritos, except one. The one and only source for email addresses in
> message.el IMHO should be EUDC, and the one and only completion UI
> should be whatever CAPF uses. Any and all sources for email addresses
> should implement back-ends for EUDC, so they can be queried via EUDC,
> which does the aggregation of results from the different sources, and
> delivers it back to message.el as one lump.
>
>
> EUDC backends:   ldap   ecomplete   mailabbrev   bbdb    you-name-it
>                \________________________ ____________________________/
>                                         V
>                                         |
>                                         V
> eudc-capf.el:            eudc-capf-message-expand-name
>                                         |
>                                         V
> message.el:               message-completion-function
>                                         |
>                                         V
> minibuffer.el                completion-at-point
>                                         |
>                                         V
>                    [ optional completion UI (for example corfu) ]
>
>
> As you may imagine, this is a bigger patch, and I am discussing it on
> emacs-devel with Stefan Monnier. So it'll still be a little while until
> it might hopefully get merged, and the burritos unwrapped.
>
>
> Many thanks and looking forward to your thoughts,

My only thought is that we already have a mechanism for combining
results from multiple contact-management packages, and it's called
`message--name-table'. I don't see why we should be obliged to add EUDC
as an additional and obligatory point of collation. I'm attaching a
patch I floated earlier, showing how I think it could works -- it's very
simple.

Stefan was the one who added `message--name-table', and if you're
talking to him I will obviously defer to whatever you (plural) decide.
But that's my two cents.

Eric