[Python-Dev] Proposal for Python 3.3: dependence injection

[Xavier Morel]

On 2011-03-25, at 10:22 , Simon Cross wrote:
> On Fri, Mar 25, 2011 at 1:25 AM, Nick Coghlan <ncoghlan at gmail.com> wrote:
>> As an example of the last point, perhaps rather than modifying all the
>> *clients* of the socket module, it may make more sense to have tools
>> in the socket module itself to temporarily customise the socket
>> creation process in the current thread. The advantage of such an
>> approach is that it would then work for 3rd party libraries that
>> create sockets, without requiring any further modification.
> 
> -2.  That wouldn't allow one to use normal sockets in one 3rd party
> library and special sockets in another 3rd party library.
> 
> Schiavo
> Simon

Or even in "first-party" code (in the stdlib) to set different timeouts on different APIs (say, an xmlrpclib.ServerProxy and an IMAP4 client).

For instance, currently as far as I can tell setting a socket timeout on an xmlrpclib.ServerProxy without setting a global timeout involves:

* subclassing all of xmlrpclib.Serverproxy, xmlrpclib.Transport, httplib.HTTP and httplib.HTTPConnection
* overloading __init__ on the ServerProxy subclass (and on Transport if the timeout is to be a parameter)
* overloading make_connection on the Transport subclass (in order to use the HTTP subclass and propagate the timeout)
* overloading _connection_class on the HTTP subclass
* overloading connect on the HTTPConnection class

This *could* be solved by wrapping a socket-related thread-local context manager around each call resulting in the creation of a socket, but these call sites may not even be under control.