bug#35564: [PATCH v2] Tweak dired warning about "wildcard" characters

[Kévin Le Gouguec]

Noam Postavsky <npostavs@gmail.com> writes:

> I think this might be better expressed in regexp:

I guess it is; I just tend to get superstitious about messing with match
data.

> Although this gives different results for things like:
>
>     (dired--isolated-char-p "?`foo`" 0)
>
> Do we care about that?  (if yes, then that's a missing test case)

I think we do care; if I look at what the existing code says,

    (dired--star-or-qmark-p "?`foo`" "?")
    ;; => nil

Off the top of my head, this is the best I can come up with to satisfy
this edge case:

    (defun dired--isolated-char-p (command pos)
      "Assert whether the character at POS is isolated within COMMAND.
    A character is isolated if:
    - it is surrounded by whitespace, the start of the command, or
      the end of the command,
    - it is surrounded by `\\=`' characters."
      (let ((start (max 0 (1- pos)))
            (char (string (aref command pos))))
        (and (or (string-match
                  (rx-to-string '(seq (or bos blank)
                                      (group char)
                                      (or eos blank)))
                  command start)
                 (string-match
                  (rx-to-string '(seq ?`
                                      (group char)
                                      ?`))
                  command start))
             (= pos (match-beginning 1)))))

> And perhaps the regexp above (if it's correct) can be integrated into
> this search?  (maybe not though, since negation isn't straightforward in
> regexps)

I will look into that.


Stefan Monnier <monnier@iro.umontreal.ca> writes:

> Can't we use `(... ,.. ..) instead of `eval`?
>
>
>         Stefan

That works too from what I tested, although it's not necessary anymore
with the new version above.


Thank you both for the review!  I will come back with an updated patch
(with the new test case) Some Time Later™.