On Python 3, response.msg.keys() contains the same value multiple times if
there are multiple headers with the same value. Work around this by
converting to a set before iterating over it.
Let's have the serializer just worry about serializing the dict
that we hand it, and move the unicode stuff up to a serialize module.
This should hopefully let us move toward using a version string in
cassettes.
Add the ability to filter out sensitive data, using one of three
methods: from headers, from a query string, and by using a custom
callback to modify the request.
Closes#67
Before this change, vcrpy would not work with modules of Boto (e.g., boto.iam)
that use Boto's CertValidatingHTTPSConnection to connect to AWS (unless you
went through the extra effort of disabling certificate validation during the
tests). This change adds support for those modules.
This test will fail with the following error:
TypeError: unbound method __init__() must be called with VCRHTTPConnection
instance as first argument (got CertValidatingHTTPSConnection instance instead)
The TypeError is raised because the __init__ method of Boto's
CertValidatingHTTPSConnection (which extends httplib.HTTPConnection) calls
httplib.HTTPConnection.__init__, and during the test httplib.HTTPConnection
actually refers to the patched verison (i.e., VCRHTTPConnection). When
VCRHTTPConnection.__init__ is called, it expects to receive a
VCRHTTPConnection object as its first argument, but instead it receives a
CertValidatingHTTPSConnection object. Because the only ancestor class of
CertValidatingHTTPSConnection is the original, un-patched
httplib.HTTPConnection, the first argument is not considered to be a
VCRHTTPConnection object, so a TypeError is raised.
This is a backwards-incompatible change that will store headers
as a list rather than a dictionary. The reason being that you can
have multiple values for a single header, and concatenating them
together with commas can create an unparseable string (sometimes
the header values can have commas in them)
I wasn't emulating the stateful file-object in my response stub,
so urllib3 wasn't decompressing gzipped bodies properly. This
should fix that problem.
Thanks @bryanhelmig for the motivation to dig into this.
