A call to re.sub() returns different results in Python 3.7 compared to versions 3.6 / 3.5 and 2.7

Example behavior in 2.7 / 3.5 and 3.6:
>>> re.sub(r'(([^/]*)(/.*)?)', r'\2.zip/\1/', 'example')
'example.zip/example/'

Example in 3.7.0 and 3.7.1rc2:
>>> re.sub(r'(([^/]*)(/.*)?)', r'\2.zip/\1/', 'example')
'example.zip/example/.zip//'

As you can see the returned string is different for the same regex. re.subn() confirms that 2 replacements are made instead of 1.

Is it intended to have different behaviour in 3.7+ or is this a bug?
Thanks

Thanks for the report. git bisect tells me this change was introduced with fbb490fd2f38bd817d99c20c05121ad0168a38ee (issue32308)

# ../backups/bpo34982.py
import re

print(re.sub(r'(([^/]*)(/.*)?)', r'\2.zip/\1/', 'example'))

# Running script at fbb490fd2f38bd817d99c20c05121ad0168a38ee

➜  cpython git:(fbb490fd2f) ./python.exe ../backups/bpo34982.py
example.zip/example/.zip//

# Script at fbb490fd2f38bd817d99c20c05121ad0168a38ee~1

➜  cpython git:(fbb490fd2f) git checkout -q fbb490fd2f38bd817d99c20c05121ad0168a38ee~1
➜  cpython git:(0cc99c8cd7) make > /dev/null
➜  cpython git:(0cc99c8cd7) ./python.exe ../backups/bpo34982.py
example.zip/example/

I think is an intended change as noted in the message that might break third party code (msg308229) . Adding Serhiy for thoughts.

Yes, this is an intended change. Your pattern matches an empty string at the end of the input string. It was a bug in earlier Python versions that re.sub() didn't replace empty matches adjacent to a previous non-empty match.

It is not clear what is the purpose of your code, but adding anchors or replacing * with + usually helps.

Great, thank you for explaining. My specific use case can be fixed by replacing * with + as per your suggestion.

So, just to make sure, that also means that

  re.sub('a*$', 'b', 'a')

returning 'bb' instead of 'b' is intended behavior?

Yes, it is.