Add tests for #9306

[kaushalmodi]

#9306 got fixed in 8ab6fa1 but tests need to be added.

[Araq]

What are the odds this pops up again? Close to zero, hence I didn't add tests.

[skilchen]

Unfortunately 8ab6fa1 fixes only part of the issue #9306:

import re
echo replace("bar", re"^", "foo")
echo replace("foo", re"", "-")

both replacements still end in an infinite loop instead of producing:

foobar
-f-o-o-

and:

import re
echo "".replacef(re"([^\d]*)(\d*)([^\w]*)", "$1 - $2 - $3")

produces no output, but using nre and in other languages (Perl, Gawk, Python, JavaScript, ruby, elixir, erlang) equivalent code produces:

 - - 

i.e. 3 zero-length matches resulting in empty-string captures.

[skilchen]

Thanks for your work on this issue!
The good thing: there are no longer any infinite loops. Thats excellent, but:

import re
echo replace("foo", re"", "-")

produces:
-foo
While in the common understanding of regular expressions there is an empty string in front of any string, one between every substring of length 1 and one at the end of the string.
nre.nim produces the usually expected -f-o-o-
With nre as well as in the other languages I know the following is also possible:

import nre
echo replace("foo", re"^|$", "-")

which produces
-foo-
as expected, but using re you get: -foo which doesn't seem correct.

import re
echo "".replacef(re"([^\d]*)(\d*)([^\w]*)", "$1 - $2 - $3")

still produces an empty line, which doesn't seem correct, I would expect - -. However on a string that usually should be split using this pattern, it works:

import re
echo "abc123###".replacef(re"([^\d]*)(\d*)([^\w]*)", "$1 - $2 - $3")

correctly produces:
abc - 123 - ###

[Araq]

Now that you know by my commits how to fix/touch re.nim, feel free to submit PRs.