homepage

A common usage pattern is to prevent clean up in tearDown() if the test failed. However, it requires a hack: http://stackoverflow.com/questions/4414234/getting-pythons-unittest-results-in-a-teardown-method. Would be nice to have an officially documented feature.

Another solution would be to add a new method named addSuccessCleanup.

Or call tearDown() with some parameter that it's able to retrieve.

Changing the signature of tearDown would be backwards incompatible. addSuccessCleanup is an interesting idea - cleanup functions that are only executed if the test passes. (What to do if a cleanup function raises an exception though? And when do these get executed - before or after standard cleanups.) 
In general cleanup functions are an improvement on tearDown.

addSuccessCleanup is not the best name as it assumes that some cleanup is being added, which is confusing.
Additional consideration that the need to leave the test results is a user run time preference, which may only be actual when debugging or working on the particular test.

Well, addSuccessCleanup *would* be an api for adding a cleanup - one that is only called on success. So the cleanup is skipped on failure or error, which was the original use case.
"Additional consideration that the need to leave the test results is a user run time preference, which may only be actual when debugging or working on the particular test."
I can't parse that sentence, can you rephrase please. (Did you mean another word other than "leave", and I don't understand "may only be actual".)

Why do you need it? You can add a test result depended code at the end of the test method. For example set a special flag
 def setUp(self):
 self.success = False
 def tearDown(self):
 if not self.success:
 ...
 def test_xxx(self):
 ...
 self.success = True
I don't think any special support in stdlib needed.

On Tue, Dec 4, 2012 at 3:43 PM, Michael Foord <report@bugs.python.org>wrote:
>
> Well, addSuccessCleanup *would* be an api for adding a cleanup - one that
> is only called on success. So the cleanup is skipped on failure or error,
> which was the original use case.
>
I thought it will be a callback method.
> "Additional consideration that the need to leave the test results is a
> user run time preference, which may only be actual when debugging or
> working on the particular test."
>
> I can't parse that sentence, can you rephrase please. (Did you mean
> another word other than "leave", and I don't understand "may only be
> actual".)
>
Most of the time users need to clean up mess. It mostly during debug
session you need to leave it. So the workaround should be easy and obvious
to enable and disable. Serhiy's method is good. I was too busy with other
stuff to think about it. Thanks.

Anatoly, if this is good for you, please close the issue.

I agree with Serhiy here. This use case seems too specialized, and there are easy ways to achieve the same thing in code.

The solution suggested in msg176909 seems accepted all around so can we please close this.