AWSPreparedRequest.prepare_body broken for unicode parameters

[kacholia]

818d235 broke AWSPreparedRequest.prepare_body() when body is a dict with unicode values. This is common for calls to SES or other AWS services.

Previously the items in body would be utf-8 encoded in vendored/requests/models.py; with this change it directly calls urllib.urlencode before encoding unicode values.

Offending line: 818d235?diff=split#diff-cb00fc815f8f80ae15e7d0cecc74b395R438

[joguSD]

@kacholia This definitely seems like a possible regression.

1. What exact error are you getting?
2. Could you provide a minimal sample to reproduce the issue?

[kacholia]

This is a major regression for anyone sending emails via SES and passing in unicode strings as body/subject. Run prepare_body on a dict with unicode values vs a str that's utf8 encoded.

>> urllib.urlencode({'foo': u'Abbé'}, doseq=True)

'foo=Abb%3F'

>> urllib.urlencode({'foo': u'Abbé'.encode('utf-8')}, doseq=True)

'foo=Abb%C3%A9'

…

On Mon, Sep 17, 2018 at 6:45 PM, joguSD ***@***.***> wrote: @kacholia <https://github.com/kacholia> This definitely seems like a possible regression. 1. What exact error are you getting? 2. Could you provide a minimal sample to reproduce the issue? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1557 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AA13MNC29KVWgoLHZp7QDpm8bq50my6lks5ucFBEgaJpZM4Ws6c6> .

[joguSD]

I wasn't able to reproduce this on Python 3.
On Python 2 I get the following:

Python 2.7.13 (default, Jul 18 2017, 13:54:29) 
[GCC 4.2.1 Compatible Apple LLVM 8.1.0 (clang-802.0.42)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> from botocore.compat import urlencode
>>> urlencode({'foo': u'Abbé'}, doseq=True)
'foo=Abb%3F'
>>> urlencode({'foo': u'Abbé'.encode('utf8')}, doseq=True)
'foo=Abb%C3%A9'

Looking into a fix.

[kacholia]

Yes python3 behavior is different. Keep in mind requests/compat.py defines str as unicode for python2.

…

On Tue, Sep 18, 2018 at 10:08 AM, joguSD ***@***.***> wrote: I wasn't able to reproduce this on Python 3. On Python 2 I get the following: Python 2.7.13 (default, Jul 18 2017, 13:54:29) [GCC 4.2.1 Compatible Apple LLVM 8.1.0 (clang-802.0.42)] on darwin Type "help", "copyright", "credits" or "license" for more information.>>> from botocore.compat import urlencode>>> urlencode({'foo': u'Abbé'}, doseq=True)'foo=Abb%3F'>>> urlencode({'foo': u'Abbé'.encode('utf8')}, doseq=True)'foo=Abb%C3%A9'>>> type(u'Abbé')<type 'unicode'>>>> isinstance(u'Abbé', str)False Looking into a fix. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1557 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AA13MP1jG-HqplJUq6LnPIF5QUgYyRKDks5ucSiXgaJpZM4Ws6c6> .

[joguSD]

@kacholia This should be fixed and going out in the release today.