Good point! I just wanted to compare equivalent substitutions.

Adding testC:

  'testC' => sub { my $f = $test; $f =~ s/\s+,/,/g; $f =~ s/,\s+/,/g; 
+},

100_000:
        Rate test1 test2 test3 testA testC testB
test1  6.5/s    --  -28%  -66%  -76%  -80%  -81%
test2  9.0/s   39%    --  -53%  -67%  -73%  -73%
test3 19.1/s  196%  113%    --  -29%  -42%  -43%
testA 27.0/s  318%  201%   41%    --  -18%  -19%
testC 33.1/s  412%  269%   73%   23%    --   -1%
testB 33.3/s  416%  272%   74%   23%    1%    --
[download]

C is 23% faster than A, even running twice over some of the commas in the test string. B doesn't touch a comma twice, but is only 1% faster than C.