I am attempting to re-organize our test libraries for automation and nose seems really promising. My question is, what is the best strategy for passing Python objects into nose tests?

Our tests are organized in a testlib with a bunch of modules that exercise different types of request operations. Something like this:

testlib
\-testmoda
\-testmodb
\-testmodc

In some cases the test modules (i.e. testmoda) is nothing but test_something(), test_something2() functions while in some cases we have a TestModB class in testmob with the test_anotherthing1(), test_anotherthing2() functions. The cool thing is that nose easily finds both.

Most of those test functions are request factory stuff that can easily share a single connection to our server farm. Thus we do a lot of test_something1(cnn), TestModB.test_anotherthing2(cnn), etc.

Currently we don't use nose, instead we have a hodge-podge of homegrown driver scripts with hard-coded lists of tests to execute. Each of those driver scripts creates its own connection object. Maintaining those scripts and the connection minutia is painful.

I'd like to take free advantage of nose's beautiful discovery functionality, passing in a connection object of my choosing.

Thanks in advance!

Rob

P.S. The connection objects are not pickle-able. :(