Issue 24284: Inconsistency in startswith/endswith

Messages (8) msg244027 - (view) Author: Serhiy Storchaka (serhiy.storchaka) * Date: 2015-05-25 11:41

The behavior of startswith in corner case is inconsistent between str and bytes in Python 3, str and unicode in Python 2, and between str in Python 2 and Python 3.

Python 3:
>>> ''.startswith('', 1, 0)
True
>>> b''.startswith(b'', 1, 0)
False

Python 2:
>>> ''.startswith('', 1, 0)
False
>>> u''.startswith(u'', 1, 0)
True

If define s1.startswith(s2, start, end) for non-negative indices and non-tuple s2 as an equivalent to the expression `start + len(s2) <= end and s2[start: start + len(s2)] == s2` or to `s1.find(s2, start, end) == start`, "".startswith("", 1, 0) should be False.

The same issue exists for endswith. See issue24243 for more detailed discussion.

Proposed patch fixes str.startswith and str.endswith.

msg244036 - (view) Author: R. David Murray (r.david.murray) * Date: 2015-05-25 14:02

I think this can only be applied in a feature release (and I think it should be, because of the backward-compatibility-with-python2 issue).  However, since this is potentially controversial, we need some more opinions.

msg244068 - (view) Author: Martin Panter (martin.panter) * Date: 2015-05-26 00:19

I can’t imagine much code would rely on either old or new behaviour. If you only put it into a feature release, would you have to document it as a change in behaviour?

msg244106 - (view) Author: R. David Murray (r.david.murray) * Date: 2015-05-26 14:43

Just in the what's new porting section, I think.  The fact that there "should" be very little to no code that relies on this is why I'd like to see it fixed.  The fact that the report was a theoretical one, and not one that broke code, is why I think we should fix it only in 3.5.

msg244124 - (view) Author: Serhiy Storchaka (serhiy.storchaka) * Date: 2015-05-26 18:56

Could you please help me with wording?

msg244139 - (view) Author: Martin Panter (martin.panter) * Date: 2015-05-26 22:14

How about this for What’s New:

* The :meth:`str.startswith` and :meth:`str.endswith` methods no longer return ``True`` when finding the empty string and the indexes are completely out of range. See :issue:`24284`.

Perhaps that is good enough, but if you wanted to be more precise, I think the actual condition is if the *start* index is beyond the end of the string, or *end* is before *start*.

msg244520 - (view) Author: Serhiy Storchaka (serhiy.storchaka) * Date: 2015-05-31 06:09

Thank you Martin.

msg244521 - (view) Author: Roundup Robot (python-dev) Date: 2015-05-31 06:17

New changeset a82498f424fe by Serhiy Storchaka in branch '3.5':
Issue #24284: The startswith and endswith methods of the str class no longer
https://hg.python.org/cpython/rev/a82498f424fe

New changeset 04162e06323f by Serhiy Storchaka in branch 'default':
Issue #24284: The startswith and endswith methods of the str class no longer
https://hg.python.org/cpython/rev/04162e06323f

History Date User Action Args 2015-05-31 06:17:31serhiy.storchakasetstatus: open -> closed
resolution: fixed
stage: patch review -> resolved 2015-05-31 06:17:05python-devsetnosy: + python-dev
messages: + msg244521
2015-05-31 06:09:05serhiy.storchakasetassignee: serhiy.storchaka
messages: + msg244520
versions: + Python 3.6 2015-05-26 22:14:53martin.pantersetmessages: + msg244139 2015-05-26 18:56:46serhiy.storchakasetmessages: + msg244124 2015-05-26 14:43:20r.david.murraysetmessages: + msg244106 2015-05-26 00:19:28martin.pantersetnosy: + martin.panter
messages: + msg244068
2015-05-25 22:01:43rhettingersetnosy: + rhettinger
2015-05-25 14:02:32r.david.murraysetnosy: + r.david.murray

messages: + msg244036
versions: - Python 2.7, Python 3.4, Python 3.6

2015-05-25 11:41:27serhiy.storchakacreate