bug#50965: Allow nil third argument for mapconcat

[Stefan Kangas]

Eli Zaretskii <eliz@gnu.org> writes:

>> +Optional argument SEPARATOR must be a string, a vector, or a list of
>> +characters; nil is the empty string."
>
> nil is not a string, so saying that is not the best wording.  I
> suggest "nil stands for an empty string." instead.

Yes, that's better.  Fixed.

>> +(ert-deftest fns-tests-mapconcat ()
>> +  (should (string= (mapconcat #'identity '() "_") ""))
>> +  (should (string= (mapconcat #'identity '("A") "_") "A"))
>> +  (should (string= (mapconcat #'identity '("A" "B") "_") "A_B"))
>> +  (should (string= (mapconcat #'identity '("A" "B" "C") "_") "A_B_C"))
>> +  (should (string= (mapconcat #'identity '("a" "b")) "ab")))
>
> Please be sure to test with vectors, bool-vectors, list of characters,
> and non-ASCII strings as well.

I've added such tests.  Please check especially the bool-vector case,
as I was unsure about what exactly to test for.  I put this:

    (should (string= (mapconcat #'identity [nil nil] "") ""))
    (should-error (mapconcat #'identity [nil nil t])
                  :type 'wrong-type-argument)

>> I'm assuming this should go to master at this point.
>
> Yes, thanks.

New patch attached.

0001-Make-mapconcat-argument-separator-optional.patch
Description: Text Data