completion-extra-properties

[daimrod]

completion-extra-properties doesn't seem to be supported by ido.

https://github.com/daimrod/manual-tagging/blob/master/manual-tagging.el#L85

I use it here to customize the completion buffer, it lets me display additional information.

[DarwinAwardWinner]

Comparing the completion code between stock completion and ido completion, it looks like both standard Emacs completion and ido call display-completion-list through minibuffer-completion-help and ido-completion-help respectively, but minibuffer-completion-help implements some preprocessing including adding annotation while ido-completion-help does not.

The easy solution would be to fall back to standard completing-read if completion-extra-propertiesis non-nil, and I think that's the first change I will make. The better solution would be to implement support for at least the two existing properties :annotation-function and :exit-function. I think it should be possible to implement both.

What do you think? Is the fallback strategy viable for now?

[daimrod]

Thanks for the quick reply.

Yes, I think that using completing-read when ido-completing-read can not
do the job is a correct approach.

ATM I don't have much time, but I plan to work on this issue ASAP. I've
just report the issue because I've started to use ido-ubiquitous and it

says that every incompatibilities should be reported :)

Daimrod/Greg

[DarwinAwardWinner]

Ok, I will push an update for now that disables ido-ubiquitous when completion-extra-properties is non-nil. Then we can work on a real solution. I'm thinking that one could write a wrapper display-completion-list-with-annotations that handles the :annotation-function property and then calls display-completion-list. Then define some advice on ido-completion-help that lexically replaces the binding of display-completion-list with the wrapper. (Can you lexically modify the function slot of a symbol?).

[DarwinAwardWinner]

Is completion-extra-properties a newly-added variable?

[DarwinAwardWinner]

Can you test out the latest master to see if it works as you expect? If so, I will make it a release. Thanks.

[daimrod]

Though it does detect the completion-extra-properties stuff, it still
uses ido and I don't understand why because it does execute the correct
ad-do-it.

Daimrod/Greg

[DarwinAwardWinner]

Ok, I'll have to investigate this when I have more time.

[daimrod]

I don't know if it was recently added, I've started to play with it the

last month. I think it has been here for a long time but is rarely used.

Daimrod/Greg

[DarwinAwardWinner]

Actually, it looks to have been added in Emacs 24. I'm still on Emacs 23 and I don't have it. Time for me to upgrade, I guess.

[daimrod]

Ok, I'll have more time to try to fix this issue the next week.

[DarwinAwardWinner]

Whoops, I forgot that ido-ubiquitous now has two code paths, one for Emacs 23 and below and another (cleaner) one for Emacs 24 and above. I put the fix in the code path for 23, where it had no effect. I've just pushed 0dff868 which moves the fix to the correct code path, and it works for me in Emacs 24.

[daimrod]

Nice, it works as expected.

Thank you.

Daimrod/Greg

[cichli]

@DarwinAwardWinner adding annotation support to ido-completion-help sounds like a great solution, and :exit-function seems certainly doable. Do you still plan to implement support for this? :-)

If not, I'm happy to take a stab at a PR.

[DarwinAwardWinner]

I'm not currently planning to implement this, but I would certainly be willing to merge a patch implementing it. Make sure you check all the properties in completion-extra-properties and fall back if any of them are ones your code can't handle.