utterless 0.4.0

Extend Python's "unittest" to better handle "logging" messages

Utterless

Extend Python's "unittest" to better handle "logging" messages

There's a problem when you try to run a unittest test suite in a codebase which also uses the logging framework: You see your nice output of dots (...F....SS....F..) clogged up with lots of log messages.

For occasional cases you can supress log messages in a block of code with self.assertLogs(...). But that doesn't scale when almost every test in your suite generates messages.

Alternatively you can fiddle your log levels so that all log messages get supressed. But when you have a failing test the log messages can be very helpful in debugging it.

Wouldn't it be great to have an alternative log runner that captures all log messages. If the test fails the messages get displayed with the error traceback. But if the test succeeds then the log messages get silently discarded.

Look no further! This is possibly the package you want.

Usage

We assume you already have a test suite that you run something like this:

$ python -m unittest discover

Having installed utterless you just need to replace that with:

$ python -m utterless discover

Optional Usage

Utterless has mechanisms to integrate with some frameworks that don't run tests through unittest directly.

For example, with Django you can specify the Utterless test runner on the command line:

$ ./manage.py test --testrunner utterless.contrib.django.DiscoverRunner

or by editing settings.py to contain:

TEST_RUNNER = "utterless.contrib.django.DiscoverRunner"

Licence

This software copyright P. S. Clarke and is licensed in accordance with the BSD Three-Clause Licence.

Limitations and Roadmap

• Currently it completely ignores the warnings module, so output from that (e.g. a DeprecationWarning or a ResourceWarning) will still splurts across the screen. I could possibly collect that up during tests and include it at the end. The complication is that warnings should probably still be emited regardless of whether the test passed or failed. Treating warnings as errors might be the better option.