bug#16577: Add regexp-based version of completion-ignored-extensions

[Stefan Monnier]

Lars Ingebrigtsen [2022-02-21 17:20:54] wrote:
> Stefan Monnier <monnier@iro.umontreal.ca> writes:
>>> I haven't looked at the various code snippets that use the resulting
>>> regexps, but I imagine that some are matching on the full file name and
>>> some aren't?
>> I think the important case only has the nondirectory part of the file
>> name at hand, which is why I propose for "/<RE>" to be defined as
>> matching only on that part (with a terminating / if it's a directory).
> Oh, so this wouldn't be a plain regexp, but a ... mini-language where
> "/" has a special meaning?

Yup.

> I think that sounds pretty confusing.

It's not great, indeed.  But, AFAICT something like that is the only way
to introduce regexps into `completion-ignored-extensions` without having
to touch too much of the code using that variable.

> Hm.  I think it'd a more straightforward interface to just keep
> regexps out of it all, but I see the attraction of having a regexp
> interface, too.

There are several file name patterns I've wanted to ignore but couldn't
because enumerating them all is somewhere between inconvenient
and impossible (whereas a regexp would cover them easily).


        Stefan