[Twisted-Python] Refactoring of trial - call for feature requests and suggestions

[Andrew Bennetts]

On Thu, Jun 17, 2004 at 07:26:30AM -0400, Jp Calderone wrote:
> Andrew Bennetts wrote:
> >On Thu, Jun 17, 2004 at 02:38:28AM -0400, Jp Calderone wrote:
> >
> >> * Argument passing between unittest methods.  For example, if a test 
> >>method were defined as testFoo(self, x, y), the two-tuple returned from 
> >>setUp() would be passed in as values for the arguments.  Similarly for 
> >>tearDown().
> >
> >
> >I'm not sure why this is better than having setUp set instance attributes?
> >
> 
>   Shorter user code.

The difference seems minimal to me.

    def setUp(self):
        ...
        self.x = ...
        self.y = ...

    def testFoo(self):
        ...
        self.assertEqual((self.y, self.x), foo(self.x, self.y))

    def tearDown(self):
        self.x.close()

vs.

    def setUp(self):
        ...
        x = ...
        y = ...
        return x, y

    def testFoo(self, x, y):
        ...
        self.assertEqual((y, x), foo(x, y))

    def tearDown(self, x, y):
        x.close()
        
It doesn't seem like a big deal either way, so why change it?

If a test* or a tearDown method find prepending "self." to be too unwieldy,
they always have the option of doing "x = self.x".

Also, what would happen if setUpClass and setUp both returned some values?
Using instance variables, there's no ambiguity.

-Andrew.