reject unicode in zlib

[vstinner]

BPO	4757
Nosy	@malemburg, @pitrou, @vstinner, @florentx
Files	zlib_bytes.patch issue4757_zlib_bytes_v2.diff: Patch, apply to py3k

^{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/pitrou'
closed_at = <Date 2009-12-14.18:24:46.510>
created_at = <Date 2008-12-27.12:58:17.987>
labels = ['extension-modules', 'type-bug']
title = 'reject unicode in zlib'
updated_at = <Date 2010-01-10.02:31:01.289>
user = 'https://github.com/vstinner'

bugs.python.org fields:

activity = <Date 2010-01-10.02:31:01.289>
actor = 'vstinner'
assignee = 'pitrou'
closed = True
closed_date = <Date 2009-12-14.18:24:46.510>
closer = 'pitrou'
components = ['Extension Modules']
creation = <Date 2008-12-27.12:58:17.987>
creator = 'vstinner'
dependencies = []
files = ['12472', '15556']
hgrepos = []
issue_num = 4757
keywords = ['patch']
message_count = 15.0
messages = ['78356', '78357', '78360', '78365', '79090', '79110', '79124', '79225', '96376', '96378', '96382', '96383', '96384', '96389', '97491']
nosy_count = 5.0
nosy_names = ['lemburg', 'pitrou', 'vstinner', 'ebfe', 'flox']
pr_nums = []
priority = 'normal'
resolution = 'fixed'
stage = None
status = 'closed'
superseder = None
type = 'behavior'
url = 'https://bugs.python.org/issue4757'
versions = ['Python 3.1', 'Python 3.2']

[vstinner]

Python 2.x allows to encode any byte string (str) and ASCII unicode
string (unicode):

$ python
Python 2.5.1 (r251:54863, Jul 31 2008, 23:17:40)
>>> import zlib
>>> zlib.compress('abc')
"x\x9cKLJ\x06\x00\x02M\x01'"
>>> zlib.compress(u'abc')
"x\x9cKLJ\x06\x00\x02M\x01'"
>>> zlib.compress(u'abc\xe9')
...
UnicodeEncodeError: 'ascii' codec can't encode character u'\xe9' ...

I'm not sure that this behaviour was really wanted become the
decompress operation is not symetric (the result type is always byte
string):

$ python
Python 2.5.1 (r251:54863, Jul 31 2008, 23:17:40)
>>> import zlib
>>> zlib.decompress("x\x9cKLJ\x06\x00\x02M\x01'")
'abc'

---

Python 3.0 accepts any string: bytes or characters. But decompress
always produce bytes string:

$ ./python
Python 3.1a0 (py3k:67926M, Dec 26 2008, 23:59:07)
>>> import zlib
>>> zlib.compress(b'abc')
b"x\x9cKLJ\x06\x00\x02M\x01'"
>>> zlib.compress('abc')
b"x\x9cKLJ\x06\x00\x02M\x01'"
>>> zlib.compress('abc\xe9')
b'x\x9cKLJ>\xbc\x12\x00\x06\xca\x02\x93'
>>> zlib.compress('abc\xe9'.encode('utf-8'))
b'x\x9cKLJ>\xbc\x12\x00\x06\xca\x02\x93'
>>> zlib.decompress(b'x\x9cKLJ>\xbc\x12\x00\x06\xca\x02\x93')
b'abc\xc3\xa9'

The most strange operation is the decompression of an unicode string:

$ ./python
>>> zlib.decompress('x\x9cKLJ>\xbc\x12\x00\x06\xca\x02\x93')
...
zlib.error: Error -3 while decompressing data: incorrect header check

---

I propose to change zlib API to reject unicode string and use explicit
conversion to/from bytes. Functions/methods:

• compress(bytes, ...)
• decompress(bytes, ...)
• <compress object>.compress(bytes, ...)
• <decompress object>.decompress(bytes, ...)
• crc32(bytes, value=0)
• adler(bytes, value=1)

Note: binascii.crc32() already rejects unicode string.

The behaviour may kept in Python 3.0.x and only changed in Python 3.1.

[vstinner]

See also issue bpo-4738 (better threads support in zlib).

[malemburg]

On 2008-12-27 13:58, STINNER Victor wrote:
> Python 2.x allows to encode any byte string (str) and ASCII unicode 
> string (unicode):
> 
> $ python
> Python 2.5.1 (r251:54863, Jul 31 2008, 23:17:40)
>>>> import zlib
>>>> zlib.compress('abc')
> "x\x9cKLJ\x06\x00\x02M\x01'"
>>>> zlib.compress(u'abc')
> "x\x9cKLJ\x06\x00\x02M\x01'"
>>>> zlib.compress(u'abc\xe9')
> ...
> UnicodeEncodeError: 'ascii' codec can't encode character u'\xe9' ...
> 
> I'm not sure that this behaviour was really wanted become the 
> decompress operation is not symetric (the result type is always byte 
> string):
> 
> $ python
> Python 2.5.1 (r251:54863, Jul 31 2008, 23:17:40)
>>>> import zlib
>>>> zlib.decompress("x\x9cKLJ\x06\x00\x02M\x01'")
> 'abc'
> 

I don't see a problem with this. The fact that Python 2.x also
accepts Unicode ASCII strings where strings are normally expected
is intended to help with the migration to Unicode, so the above
is expected.

zlib itself doesn't care about whether the data to be encoded
is text or bytes.

In Python 3.x, it's probably better to use bytes throughout the
API.

[ebfe]

I don't think Python 2.x should be changed - but 3.0 or 3.1 should be:

• Characters don't mean a thing in zlib-land, all operations are based
  on bytes and their (implicit) default encoding. This behaviour is hidden
  and somewhat violates the rule of least surprise.
• type(zlib.decompress(zlib.compress('abc'))) == bytes anyway
• Changing from s* to y* forces the programmer to use .encode() on his
  strings (e.g. zlib.compress('abc'.encode()) which very clearly shows
  what's happening. If you want to compress and decompress Python3
  strings, you *must* share the same character encoding; think of
  zlib.compress('hôńè') and str(zlib.decompress(x)) with different locales.
• Other modules (hashlib comes to my mind...) already reject Unicode
  objects for the same argument.

[vstinner]

The fact that Python 2.x also accepts Unicode ASCII strings
where strings are normally expected is intended to help with
the migration to Unicode

I hate this behaviour. It doesn't help migration, it's the opposite! Sometimes
it works (ASCII), and somtimes it fails (just one non-ASCII character). And
then we will read "Unicode sucks!" because people doesn't understand the
error.

In Python 3.x, it's probably better to use bytes throughout the
API.

I propose to reject unicode in Python 3.x and display a warning for Python
2.x. A warning to prepare the migration... not to Unicode, but to Python3 ;-)

[ebfe]

The current behaviour may help the majority by ignorance and cause weird
errors for others. We tell people that Python distincts between Text and
Data but actually treat it all the same by implicit encoding.

Modules that only operate on Bytes should reject Unicode-objects in
Python3; it's a matter of 3 lines to display a warning in Python 2.
Those modules that usually operate on Text but have single functions
that operate on Bytes should display a warning but not enforce explicit
encoding.

Also see bpo-4821 and bpo-4818 where unicode already got rejected by the
openssl-driven classes but silently accepted by the build-in ones.

[malemburg]

On 2009-01-04 23:51, STINNER Victor wrote:

STINNER Victor <victor.stinner@haypocalc.com> added the comment:

> The fact that Python 2.x also accepts Unicode ASCII strings
> where strings are normally expected is intended to help with
> the migration to Unicode

I hate this behaviour. It doesn't help migration, it's the opposite! Sometimes
it works (ASCII), and somtimes it fails (just one non-ASCII character). And
then we will read "Unicode sucks!" because people doesn't understand the
error.

Well, that's your opinion.

The feature was added to get people
work with Unicode at all, since otherwise we would have had to do
all the Unicode porting we're doing now for Python 3 at the time
Unicode was introduced - which was in version Python 1.6, eight years
ago.

At the time the Python community was a lot smaller and there wasn't
all that much interest in Unicode anyway - the Unicode support I wrote
for Python 1.6 was partially financed by HP which needed it for an
application they had written in Python.

See the introduction in PEP-100 for the motivation behind the design
decisions:

http://www.python.org/dev/peps/pep-0100/

> In Python 3.x, it's probably better to use bytes throughout the
> API.

I propose to reject unicode in Python 3.x and display a warning for Python
2.x. A warning to prepare the migration... not to Unicode, but to Python3 ;-)

Fair enough.

[vstinner]

> I propose to reject unicode in Python 3.x and display a warning for
> Python 2.x. A warning to prepare the migration... not to Unicode, but to
> Python3 ;-)

Fair enough.

The patch for Python 3.x is already attached to this issue. We might only
apply this one and leave Python 2.x unchanged. Can someone review the patch?

[florentx]

Definitely, zlib.compress should raise a TypeError (like bz2 does).

>>> import bz2, zlib
>>> bz2.compress('abc')
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: argument 1 must be bytes or buffer, not str
>>> zlib.compress('abc')
b"x\x9cKLJ\x06\x00\x02M\x01'"

Someone can review the patch and merge it?

[pitrou]

The patch lacks a test that TypeError is raised on unicode input,
otherwise it's fine.

[florentx]

Patch from haypo updated for r76830 .

Additional tests for TypeError, and to check that bytearray objects are
accepted.

[pitrou]

The patch produces a number of errors in test_tarfile, test_distutils,
test_gzip and test_xmlrpc.

[florentx]

Fixed.

And some "bytearray" tests improved in test_zlib.

[pitrou]

The patch was committed to py3k and 3.1. Thank you!

[vstinner]

The patch was committed to py3k and 3.1. Thank you!

r76836 and r76838