Re: project--completing-read-strict breaks ada-mode project completion table

[Dmitry Gutov]

On 19.01.2019 06:35, Stephen Leake wrote:

> > The latency seemed to be a bit bigger than with the default completion
> > table based on 'find', but not significantly so.
>
> Fixing the ignore would help that; it's processing all of .git.

It doesn't handle the contents of .gitignore, though. That would add 
some performance penalty by itself.

> > I don't think it can be *the* UI for everybody.
>
> Of course. And this is the main point; project.el must allow
> users/clients to choose the completion table and completion style.

Users, not necessarily backends. And let's say "completion style" in the 
general sense.

> So the recent change to project--completing-read-strict should be
> reverted, and implemented as a completion style/table instead.

Please see the rest of this discussion. I think it's now the consensus 
that the uniquify logic should move out of the completion table.

We can apply it inside project-find-file itself, depending on a user 
option. That will require changes in project--completing-read-strict 
too, of course.