Have -b warn when directly comparing ints and bytes

[brettcannon]

BPO	23681
Nosy	@brettcannon, @pfmoore, @vstinner, @larryhastings, @PCManticore, @berkerpeksag, @vadmium, @serhiy-storchaka
Files	bytes_to_int_compare.patch bytes_to_int_compare_2.patch

^{Note: these values reflect the state of the issue at the time it was migrated and might not reflect the current state.}

Show more details

GitHub fields:

assignee = 'https://github.com/serhiy-storchaka'
closed_at = <Date 2015-03-20.14:59:45.984>
created_at = <Date 2015-03-16.17:05:30.088>
labels = ['interpreter-core', 'type-feature', 'release-blocker']
title = 'Have -b warn when directly comparing ints and bytes'
updated_at = <Date 2015-03-20.14:59:45.983>
user = 'https://github.com/brettcannon'

bugs.python.org fields:

activity = <Date 2015-03-20.14:59:45.983>
actor = 'serhiy.storchaka'
assignee = 'serhiy.storchaka'
closed = True
closed_date = <Date 2015-03-20.14:59:45.984>
closer = 'serhiy.storchaka'
components = ['Interpreter Core']
creation = <Date 2015-03-16.17:05:30.088>
creator = 'brett.cannon'
dependencies = []
files = ['38511', '38591']
hgrepos = []
issue_num = 23681
keywords = ['patch']
message_count = 10.0
messages = ['238228', '238236', '238237', '238261', '238501', '238626', '238636', '238637', '238665', '238687']
nosy_count = 9.0
nosy_names = ['brett.cannon', 'paul.moore', 'vstinner', 'larry', 'Claudiu.Popa', 'python-dev', 'berker.peksag', 'martin.panter', 'serhiy.storchaka']
pr_nums = []
priority = 'release blocker'
resolution = 'fixed'
stage = 'resolved'
status = 'closed'
superseder = None
type = 'enhancement'
url = 'https://bugs.python.org/issue23681'
versions = ['Python 3.4', 'Python 3.5']

[brettcannon]

To help writing Python 2/3 code the -b flag should switch on a warning when comparing an int to a bytes object in Python 2. This will help when someone writes something like b'abcd'[2] == b'c' and it always returns False thanks to the indexing returning 99 in Python 3.

[serhiy-storchaka]

6 tests failed:
test_buffer test_poplib test_quopri test_smtpd test_sunau
test_tokenize

And all of them look as bugs.

[serhiy-storchaka]

Here is a patch that adds required feature and fixes existing bugs in the stdlib and tests.

No one of failed tests was false positive.

[berkerpeksag]

[...] fixes existing bugs in the stdlib and tests.

These changes should probably be backported to 3.4.

[pfmoore]

LGTM, just one minor comment in review.

[vstinner]

I didn't understand the issue.

$ python2
>>> b'A'[0] == 65
False

$ python3
Python 3.4.1 (default, Nov  3 2014, 14:38:10) 
>>> b'A'[0] == 65
True

Oh ok, now I get it: Python 2 and Python 3 behaves differently when comparing a string of a single byte and an integer.

Ok to raise a warning when -b command line option is used.

[serhiy-storchaka]

@brett: Yes, the patch handles 42 == b'*' as well.

@victor: The problem in such code:

x = b'Array'
x[0] == b'A'

Added explicit tests for bytes on the right side of the comparison and replaced b'.'[0] to ord(b'.').

[vstinner]

This will help when someone writes something like b'abcd'[2] == b'c'

What if someone writes line[-1] == 0 and line is a Unicode string? Should we emit a warning?

I patched locally PyUnicode_RichCompare() to emit a warning. Hum, there are *many* warnings in argparse, http.client, and many other modules. I don't think that it's a good idea. It looks common to use a string for a sentinel (ex: state = "UNKNOWN") and then store an int (ex: state = 0). So comparison need to check the type (ex: isinstance(state, str) and state == "UNKNOWN") which is verbose and annoying.

So no, we should not emit a warning :-)

[brettcannon]

I did one final pass on the patch and only had wording comments, so tweak those and it LGTM. Assigned to Serhiy since it's obviously his patch. =)

Only other thing is to either open up a separate bug or at least apply the fixes to the stdlib in Python 3.4 as well as 3.5.