Out-of-order fails when `completing-read` has initial input

[oantolin]

Out-of-order fails when `completing-read` has initial input

[protesilaos]

Of notice is also the fact that Icomplete no longer highlights the matching characters (beside the matching candidate), but we can check that later.

[oantolin]

I thought the completion-boundaries function had my back for this initial input business. I guess not, or rather, not in all cases, because it certainly fixed issue #1.

[oantolin]

Of notice is also the fact that Icomplete no longer highlights the matching characters (beside the matching candidate), but we can check that later.

Yes, that's issue #4, kindly submitted by @jixiuf.

[oantolin]

When completing-read is given an initial input, such as one that represents a directory path, out-of-order matching will yield no results, unless the user's input starts with a space.

I don't think this is actually true. What I think is going on in your image is that /home/prot/the is being taken as a single regexp and it doesn't match any candidate. If, instead of the mo, you had typed G the mo, for example, it would use regexps /home/prot/G, the and mo, and some candidates do match all three. That would happen even though you started with G and not with a space.

So, I think it's working as advertised, so maybe this isn't technically a bug. It can be inconvenient, though, and we can try to fix it. Let me think about it a bit.

[oantolin]

Maybe a good solution is to have the regexp separator configurable. For completing paths you'd want to split the pattern on whitespace and slashes.

[oantolin]

i just realized the separators are already configurable. :) You can wrap your completing-read call in:

(let ((split-string-default-separators "[/ ]"))
   ...)

to have both slashes and spaces separate the regexps.

[protesilaos]

You are right, this is not a faulty behaviour per se. Just a matter of easing the user experience.

With regard to split-string-default-separators, it does indeed work. Could this be exposed as a user-facing option for when orderless is being used?

[oantolin]

I was first inclined to just document the usage split-string-default-separators, but I decided to add a variable orderless-regexp-separator, mostly because I think people might want to change its value for their entire Emacs session, and it seems like a terrible idea to fiddle with split-string-default-separators (as it affects every call to split-string which does not specify a separator).

Does adding that option adequately address this issue?

[protesilaos]

Yes, I can confirm this works!

[oantolin]

Excellent, I'll close this issue then.