Q: read-file-name: Descend into subdirs while using a predicate?

Q: read-file-name: Descend into subdirs while using a predicate?

[T.V Raman]

This feels like a puzzle that should have an easy answer, but I've not
found it:-)

You want to:

1. Use read-file-name with completion.
2. You want to only return filenames with a given extension -- and
   that extension is a regex 
3. You also want to descend into  subdirs of the current directory if
   the user's input matches a subdir that exists 

So you provide 'must-match  and a predicate -- and in that predicate I
tested for the regexp --- but then the descend into subdirs fails, and
if you check for either the   pattern or file-directory-p, then the
read call just returns the directory.

one heavy-weight solution might be to call directory-files with a
pattern,  and pass that as a list to completing-read but that feels
heaviy-weight in that you could end up descending deeper into the
directory tree than you need to ...

-- 

Thanks,

--Raman(I Search, I Find, I Misplace, I Research)
♉ Id: kg:/m/0285kf1  🦮

-- 

Thanks,

--Raman(I Search, I Find, I Misplace, I Research)
♉ Id: kg:/m/0285kf1  🦮

Re: Q: read-file-name: Descend into subdirs while using a predicate?

[Stefan Monnier]

T.V Raman [2022-09-26 07:45:08] wrote:
> This feels like a puzzle that should have an easy answer, but I've not
> found it:-)

It's a long standing "todo" :-(

Basically, we need to change the notion of "predicate" during completion
so as to distinguish the "predicate to filter completions" from the
"predicate to check if a value is fit to be returned from
`completing-read`" (AKA the distinguish the predicate passed to
`test-completion` from the predicate passed to `all/try/completions`).

I've been toying with the idea of letting the predicate return a third
option (beside nil and t) to say "it's acceptable for completions but
not for `test-completion`".


        Stefan

Re: Q: read-file-name: Descend into subdirs while using a predicate?

[T.V Raman]

Thanks Stefan for the clear answer, and happy to know I wasn't
overlooking something simple.
Also, on  my laptop with an SSD, using 
directory-files-recursively appears to be blazingly fast for
relatively large directories --- so that "heavy hammer" is a good
intermediate solution for my usecase which was to filter media files,
epub books and the like.
Stefan Monnier writes:
 > T.V Raman [2022-09-26 07:45:08] wrote:
 > > This feels like a puzzle that should have an easy answer, but I've not
 > > found it:-)
 > 
 > It's a long standing "todo" :-(
 > 
 > Basically, we need to change the notion of "predicate" during completion
 > so as to distinguish the "predicate to filter completions" from the
 > "predicate to check if a value is fit to be returned from
 > `completing-read`" (AKA the distinguish the predicate passed to
 > `test-completion` from the predicate passed to `all/try/completions`).
 > 
 > I've been toying with the idea of letting the predicate return a third
 > option (beside nil and t) to say "it's acceptable for completions but
 > not for `test-completion`".
 > 
 > 
 >         Stefan

-- 

Thanks,

--Raman(I Search, I Find, I Misplace, I Research)
♉ Id: kg:/m/0285kf1  🦮

--

Thanks,

--Raman(I Search, I Find, I Misplace, I Research)
♉ Id: kg:/m/0285kf1  🦮

--

Thanks,

--Raman(I Search, I Find, I Misplace, I Research)
♉ Id: kg:/m/0285kf1  🦮