Is it possible to have something like 'saved searches'

[apc]

At the moment, calling M-x bibtex-actions-open-pdf gives a list of all entries in bibtex-completion-bibliography. But it seems as if it should be possible to only see those entries that have an associated PDF file (the same applies to bibtex-actions-open-notes). I do not yet understand how completing-read-multiple works to see if this would be easy to do (to list only those entries that match a particular search, e.g.), but in case anyone knows of a way to accomplish this, I'd love to hear about it!

[bdarcus]

On your specific examples, I added some of the details currently in bibtex-actions to make this easier.

So, for example, in my orderless-based setup, I can type this to give me all entries with notes, but no link:

The substrings that are actually matched against are things like has:pdf.

Were you aware you can do that? Do you still think this would be valuable?

You're basically asking for pre-filtering, to obviate that.

I think all that would be required is to add an optional INITIAL_INPUT parameter to the main bibtex-actions-read function, and to feed those strings to that on the appropriate "open" commands.

If you're asking to also allow arbitrary saved search strings, I'm not sure. I forget: does ivy-bibtex and/or helm-bibtex have that? If yes, how does it work there?

PS - the consult--read function, which is a nicer wrapper around completing-read, is interesting to look at, and easier to understand. I'm also curious about "history" though have never used it.

https://github.com/minad/consult/blob/ded314c4a3ee48e812b206016501cb49b29259fc/consult.el#L1654-1677

[apc]

From what I can understand (which isn't much), using completion-table-in-turn might be the way to go. You'd first need to generate a completion table with all candidates with pdfs, and then pass that as well as bibtex-actions---candidates-cache as arguments to completion-table-in-turn. I don't know how to do the first thing (if you point me in the right direction I can give it a try), but this could be an interesting alternative.

EDIT: OK, I'm way out of my depth here. How do I check whether the 'text' of an element of bibtex-actions---candidates-cache contains a certain string? string-match doesn’t work because of the text properties I presume?

[bdarcus]

I don't think that's right.

I think the easy approach is just to use the initial value, because it's just a string.

And then if that doesn't work, use predicate with a filter function.

That function would use something like the built-in seq-filter or this dash function:

https://github.com/magnars/dash.el#-keep-fn-list

Both of those should work on the cached data.

I will open a branch/draft PR that points you in what I think is the right direction.

Edit: see #118.

[bdarcus]

string-match doesn’t work because of the text properties I presume?

On this, I think it does:

ELISP> (setq test-foo (propertize "test" 'face 'bar))
#("test" 0 4
  (face bar))

ELISP> (string-match "st" test-foo)
2 (#o2, #x2, ?\C-b)
ELISP> (string-match "sat" test-foo)
nil

[apc]

Hm. I guess then I do not understand how the predicate filtering works. What am I doing wrong here?

(defun my/has-pdf (string)
  (if (not (string-match "has:pdf" string))
      nil
      t))

(all-completions "Fodor" bibtex-actions--candidates-cache #'my/has-pdf)

I get (wrong-type-argument stringp (#("Fodor ... as an error.

The documentation for completion-table-with-predicate says that the predicate arguments should be "a function of one argument which returns non-nil if and only if the argument is an element of TABLE which should be considered for completion."

If I call my/has-pdf with the car of (all-completions "Fodor" bibtex-actions--candidates-cache) as argument, I get a non-nil result (since that first entry has an associated pdf). So I don't know what I'm doing wrong.

[bdarcus]

The documentation for completion-table-with-predicate says that the predicate arguments should be "a function of one argument" ...

I think this tripped me up earlier. For functions like affixation and such, I had to add an explicit argument to take the list of candidates, even though they weren't explicitly called from completing-read-multiple.

Something similar might be at issue here.

I pushed a commit just now for that filter function, but not sure this will work.

[bdarcus]

Not sure who's following this forum at this point, but just a head's up I merged support for this and for "preset" search strings.

See current README for details. Thanks to @apc for the idea, and the help.