Re: In-buffer completion of EBDB mail addresses with corfu

[Eric Abrahamsen <eric@ericabrahamsen.net>]

Arash Esbati <arash@gnu.org> writes:

> Eric Abrahamsen <eric@ericabrahamsen.net> writes:
>
>> I'm not completely clear what you're doing here -- you enter a full name
>> and email, then a comma, then TAB? EBDB will only look back as far as
>> the nearest comma for completion. In this case it will send in a blank
>> string, which should offer all emails in your EBDB database for
>> completion. 
>
> I think we can skip this (see below).
>
>> Have you created any records?
>
> Yes.
>
>> Are there strings in `ebdb-dwim-completion-cache'?
>
> Thanks, this is it.  It was nil.  If I do `M-x ebdb RET RET' to see my
> records and then hit 'q', `ebdb-dwim-completion-cache' seems to get
> populated and then hitting TAB in the header in Message mode offers
> in-buffer completion.  Is there a way to automate the setting of
> `ebdb-dwim-completion-cache'?

There are "is EBDB loaded" guards around every entrypoint I could think
of, except this! Luckily that's not hard to fix.

>> That's not a bad idea. Having them `require' libraries to activate
>> behavior isn't very Emacsy.
>
> Agreed.  And that was the reason I made that suggestion.
>
>> That would be a pretty annoying change for existing users, though...
>
> Are these two approaches mutually exclusive?  Users who want to
> `require' it can still do it, and others can load the package lazily if
> the autoload cookies are added.  Or am I missing something?

Yes, I suppose you're right. `add-hook' is idempotent, so no harm done
if that ends up getting called twice. And the "insinuate" functions are
a fine place to make sure that `ebdb-load' is called.

I don't think there should be anything complicated about that, I'll play
with it and release a new version soon.

Thanks!
Eric