Updated test suite

[duncanmmacleod]

This PR updates the test suite to include a second test case module! and lots of other things:

• renamed test/test_gwin.py -> test/test_sampler.py
• added test/test_io.py (incomplete)
• added test/test_likelihood.py (incomplete)
• added test/test_option_utils.py (incomplete)
• added test/test_results.py (incomplete)
• added test/utils module with lots of fixtures
• added conftest.py with --approximants command-line options

The final point means that the suite can now be run using the default approximants with

python setup.py test  # OR
py.test

but you can specify which approximants to use:

py.test --approximants IMRPhenomA IMRPhenomB IMRPhenomC IMRPhenomD

Closes #11.
Fixes #16.

[coveralls]

Pull Request Test Coverage Report for Build 70

• 1 of 1 (100.0%) changed or added relevant line in 1 file are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage increased (+19.06%) to 41.088%

---

Totals
Change from base Build 67:	19.06%
Covered Lines:	899
Relevant Lines:	2188

---

💛 - Coveralls

[duncanmmacleod]

ok, @cmbiwer, I should probably stop here and add more unit tests in future PRs, you were automatically assigned because I modified gwpy/io/hdf.py to fix #16.

That commit (790ccfd) is the only to touch module code, all other modify build files, or test files.

[vivienr]

Could you add SEOBNRv4 to the default list? How long do the test take as is?

[duncanmmacleod]

@vivienr, on my macbook I get the following error:

ValueError: Approximant SEOBNRv4 not available

Is there something special that needs to be done to set up this approximant?

[duncanmmacleod]

This is probably a consequence of the test only being set up for frequency-domain waveforms, at the moment. Hopefully I can work with @cdcapano in a future PR to add support for TD waveform approximants.

[vivienr]

It's only available in the time domain, and the tests only do calls in the frequency domain. If that's too much troubles, let's look at implementing time domain versions in another patch.

[cmbiwer]

@duncanmmacleod I feel like some of the codeclimate thresholds are a bit low. Like this is considered a failure:

Function test_add_sampler_option_group has 26 lines of code (exceeds 25 allowed). Consider refactoring.

Maybe at some future point these could be relaxed a bit.

Otherwise, this PR looks fine.

[vivienr]

@cmbiwer: I agree. In particular, I'd up the lines of code allowed to 50, the cognitive complexity to 10, allow up to 8 arguments to functions, and 750 lines of codes per file.