Have citar list all file containing citekey in name as candidates

[oatmealm]

Is your feature request related to a problem? Please describe.
I'd like citar to list all note files containing the citekey

Describe the solution you'd like

I had an example code which I can no longer find here, which does this:

  (defun my-notes-items (citekeys)
    "Return a hash table mapping each CITEKEY to its note files."
    (let* ((all-files
            (apply #'append
                   (mapcar (lambda (dir)
                             (directory-files-recursively dir "\\.org\\'"))
                           citar-notes-paths)))
           (result (make-hash-table :test #'equal)))
      (dolist (key citekeys)
        (puthash key
                 (seq-filter (lambda (f) (string-match-p (regexp-quote key) f))
                             all-files)
                 result))
      result))
  (defun my-notes-has-p ()
    "Return a predicate that tests whether a key has note files."
    (lambda (citekey)
      (hash-table-p (gethash citekey (my-notes-items (list citekey))))))

  (citar-register-notes-source
   'my-notes
   '(:name "Notes"
     :category file
     :items my-notes-items
     :hasitems my-notes-has-p
     :open find-file
     :create citar-file--create-note))

  (setq citar-notes-source 'my-notes)

Describe alternatives you've considered
Would be nice to have citar enable something like this by default: any file with citekey not just base name, is a a note candidate.

Additional context
This works well though only been using it for a while.

[roshanshariff]

In Citar we've tried to not have any of the "-items" functions exhibit O(nm) running time, where in this case n is the number of citation keys and m is the number of files in the notes directory. In your code, there is a loop over all citekeys (the dolist) and a nested loop over all files in the directory (the seq-filter). This would quickly become unmanageable, if say, you had a few thousand citation keys and say a hundred files in the notes directory. The built-in functions have O(n+m) running time, because they do one loop over cite keys followed by a separate loop over files.

I would suggest putting this code in the wiki for people who're okay with the performance implications, but I would hesitate to make it the default behaviour.

[bdarcus]

I would suggest putting this code in the wiki for people who're okay with the performance implications, but I would hesitate to make it the default behaviour.

I'm going to close this for the reason that Roshan cites.