Hi,

I did the update for the #71 as promised. This merge request introduces backward incompatibility with old cassettes, but I think it's worth it.

Improvements which I think should be done:

1. url matcher could be removed on on favor of uri . (I left just for backward compatibility)
2. ~~More integration and unit test should be added (I will do this after discussion of this pull request)~~ done.

Looking forward for review. Max

Fixes #338.

Also removes redundant code that was required to maintain support for Python 2.6.

Here's the pip installs for vcrpy-unittest from PyPI for the last month (via pypinfo --percent --pip vcrpy-unittest pyversion) showing nothing for Python 2.6 or 3.3.

| python_version | percent | download_count | | -------------- | ------: | -------------: | | 2.7 | 47.5% | 699 | | 3.6 | 43.7% | 643 | | 3.5 | 4.8% | 70 | | 3.4 | 4.0% | 59 |

Also includes @samuelfekete's CI fixes from 36b5162 to (mostly) get the CI green! Thanks! (I would have cherry-picked that commit but git couldn't find it.)

I'm trying to use vcrpy to accelerate the execution of my django application test suite. I'm using django 1.7 on Mac, with Python 2.7.

I added the following couple of lines to one of my tests:

     import vcr
     with vcr.use_cassette('recording.yaml'):

The result is an import error:

    import vcr
  File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/vcr/__init__.py", line 2, in <module>
    from .config import VCR
  File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/vcr/config.py", line 6, in <module>
    from .cassette import Cassette
  File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/vcr/cassette.py", line 12, in <module>
    from .patch import CassettePatcherBuilder
  File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/vcr/patch.py", line 8, in <module>
    from .stubs import VCRHTTPConnection, VCRHTTPSConnection
  File "/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/site-packages/vcr/stubs/__init__.py", line 9, in <module>
    from six.moves.http_client import (
ImportError: No module named http_client

The problematic code in stubs/init.py is :

import six
from six.moves.http_client import (
    HTTPConnection,
    HTTPSConnection,
    HTTPMessage,
    HTTPResponse,
)

This code seems to run fine when I'm just running it from a plain python console, but it results in the above ImportError under django.

boto3 bundles the requests library, so this basically is a copy and paste of the requests stubbing.

I'm a little lost in the testing of this, with some guidance and/or help I could complete this.

This PR adds support for httpx.

The stub and tests are heavily based in the aiohttp implementation. You won't see any synchronous implementation because it was not necessary.

Not sure if I edited the tox file correctly. (:

I receive this error when i upgraded to 1.6.0:

Traceback (most recent call last):
  File "run_tests.py", line 14, in <module>
    from tests.app_login_tests import MainSiteTest  # noqa
  File "/my_app/tests/app_login_tests.py", line 7, in <module>
    from vcr_setup import vcrlib
  File "/my_app/tests/vcr_setup.py", line 1, in <module>
    import vcr
  File "/usr/local/lib/python2.7/dist-packages/vcr/__init__.py", line 2, in <module>
    from .config import VCR
  File "/usr/local/lib/python2.7/dist-packages/vcr/config.py", line 8, in <module>
    from .compat import collections
  File "/usr/local/lib/python2.7/dist-packages/vcr/compat.py", line 4, in <module>
    import mock
ImportError: No module named mock

Looking at the commits i think this new conditionals imports are not working https://github.com/kevin1024/vcrpy/commit/41949f7dc64a2ddb8a9c2cb845eac7bb8b94380b#diff-2eeaed663bd0d25b7e608891384b7298R36 i'm installing the package using pip im on python 2.7.

PEP479 renders the _handle_generator function of CassetteContextDecorator object erroneous in python3.7. Yet, we can't rely properly on yield from as python2.7 compat is still mandatory. Try to find a good balance between these two facts

@graingert is it better for you?

This would close #396

usage requires edited conftest and execute with py.test --template FOO

conftest.py

import pytest
@pytest.fixture
def template(request):
    return request.config.getoption("--template")

def pytest_addoption(parser):
    parser.addoption("--template", action="store",type='string',
                     help="set the name of the template to use in tests")

test.py

import vcr
import pytest
@vcr.use_cassette('test_vapp_find.yaml')
def test_vapp_find(self, template):
        assert template == FOO

usage:

py.test -s test.py --template FOO

suggested fix

config.py

# pass kwargs we dont know about merged_config = dict(kwargs.items() + merged_config.items())
return Cassette.load(path, **merged_config)

#cassette.py::Cassette.init
**kwargs
):
super(Cassette,self).init(**kwargs)

This is a tricky issue I encountered when using the body matcher on xmlrpc requests.

Symptoms: Sometimes the request won't match, sometimes it will, and this only affects certain requests. This occurs on Python 3.4 and I believe other python 3 versions but not on 2.7

Cause: An XMLRPC request has a body with XML inside which is generated from the parameters passed to the function call. Some parameters can be of dict type. xmlrpclib (or xmlrpc.client) will loop over items of the dict and generate the appropriate XML which will be the body of our POST request. Now items order is not guaranteed in dict and the behavior changed in python 3 such that the order of the same dict can change anytime and is not more or less constant on the same computer as in python 2. So the generated XML won't be necessarily the same as the one you recorded.

Fix suggestion: A custom xmlrpc body matcher that takes that into account and will compare the struct XML elements in the correct order.

The gzip compression didn't help me debuging this as I couldn't even read directly the cassettes...

Previously request.headers was a normal dict (albeit with the request.add_header interface) which meant that some code paths would do case-sensitive matching, for example remove_post_data_parameters which tests for 'Content-Type'. This change allows all code paths to get the same case-insensitive treatment.

Additionally request.headers becomes a property to enforce upgrading it to a CaseInsensitiveDict even if assigned.

Fixes #434.

Python 2.7 reaches EOL on 2020-01-01. Many projects are pledging to drop support before or during 2020: https://python3statement.org/

It's also relatively little used. Here's the pip installs for vcrpy from PyPI for May 2019:

| category | percent | downloads | |----------|--------:|----------:| | 3.6 | 37.19% | 65,506 | | 3.7 | 26.36% | 46,432 | | 2.7 | 21.83% | 38,445 | | 3.5 | 11.53% | 20,303 | | 3.4 | 2.48% | 4,361 | | null | 0.31% | 553 | | 3.8 | 0.29% | 503 | | 3.3 | 0.01% | 14 | | Total | | 176,117 |

Source: pypistats python_minor vcrpy --last-month # pip install pypistats

There's one bit of six remaining in this PR, what's the best way to replace this?

    def test_case(self, predicate=None):
        predicate = predicate or self.is_test_method
        return six.with_metaclass(auto_decorate(self.use_cassette, predicate))

I have various hosts that I need to hit for a unit test. I'd like to be able to route the responses from each host to its own specific cassette file. The documentation does not appear to show an example of how to do this. Is there a recommended way of doing this?

Bumps actions/checkout from 3.1.0 to 3.2.0.

Dependabot will resolve any conflicts with this PR as long as you don't alter it yourself. You can also trigger a rebase manually by commenting @dependabot rebase.

Context

Per discussion here: Originally posted by @salomvary in https://github.com/kevin1024/vcrpy/issues/533#issuecomment-1269519939

CannotOverwriteExistingCassetteException can be raised along with 'no matchers failed' as the message details.

In many cases, setting vcr.allow_playback_repeats for the affected test will clear the issue.

Proposal

Update the error handler to recognize this corner case and emit a message encouraging developers to consider whether they need to set vcr.allow_playback_repeats

Does anyone know if there are any libraries that would allow to record and play queries to sql database? I don't per-see any obstacles in implementation. The question is does it even makes sense in terms of test speed improvement?

This sequence https://github.com/kevin1024/vcrpy/blob/f3f66086a149722812345aaf1bb304ff39d105ef/vcr/filters.py#L110-L116 assumes that anything that isn't a BytesIO or dict or with content-type json that is splittable by b"=" can be UTF-8 decoded. But any random unlucky binary data is likely to have a b"=" somewhere in it, and, if passed in as a raw byte array instead of a BytesIO, it then chokes that decode with 'utf-8' codec can't decode...blah blah blah if the user has defined filter_post_data_parameters for their recorder.

The BytesIO check in https://github.com/kevin1024/vcrpy/blob/f3f66086a149722812345aaf1bb304ff39d105ef/vcr/filters.py#L86 is insufficient for preventing exceptions from trying to decode binary data.