make_edc_from_energy fixture

[SimonBuechner]

fixture added in conftest.py, test_parameters.py adjusted to contain fixture-tests for clarity

Which issue(s) are closed by this pull request?

Closes #81

Changes proposed in this pull request:

• make_edc_from_energy fixture implemented in conftest.py
• test_parameters.py added and fixture-adjusted clarity-tests added.
• newly proposed testing structure reducing redundancy

[SimonBuechner]

As requested this is the proposed fixture for the parameter tests. A few conceptual problems though:

• it basically just builds a pyfar.TimeData object from provided energy and samplerate. It also normalizes the energy although, as we talked about this before, this is not neccessary.
• In that case though, this could be done inline by just calling the TimeData-Constructor – especially because we need the energy values in many cases anyways to calculate the solution.
• We should think about keeping it anyways for future cases.
• The whole thing will be restructured anyways because as soon as the _energy_balance (V1 parameter definition #80) base function is done we will need slightly different tests

Before continuing though I'd love some feedback on this.

[ahms5]

nice, thanks, I have some conseptual questions.
would it make sence to define different fixtures with more describing names for energy, "geometric" and "real"?
could you also add more details to what geometic and real means?

[SimonBuechner]

nice, thanks, I have some conseptual questions. would it make sence to define different fixtures with more describing names for energy, "geometric" and "real"? could you also add more details to what geometic and real means?

Yes, will do. I think I will also allow handover of e.g. the parameters for the geometrical decay because we usually need them in the tests anyway.

Right now it's a rough sketch to get the idea across. I was thinking about implementing different fixtures for those cases but I like the idea of having one universal fixture to get different EDCs for these kind of tests.

[mberz]

Thanks for getting this started.

In general I feel that the fixture is more complex than it needs to be and can be simplified, see the comments below.
I would propose to reduce the number of different parameterizations to for example:

• passing energy values which represent an edc
• passing a reverberation time/ intercept (simple exponential decay function)

In most test cases, I'd expect that an edc with a dynamic range between 0 and -65 (or a bit more headroom) is sufficient. I don't think there's need for -inf dB dynamic range for the average test case. This in my opinion could be a more representative default.

Currently, the fixture is not tested, please add some basic tests for the fixture.
Please also do not mix up adding this fixture with other changes such as changing the tests for other functions at this point.

[SimonBuechner]

I restructured the factory so that we follow a more generalized approach for an exponential decay as requested. For test convenience it now returns a simple empty edc for typechecks etc.

dynamic range can also be adjusted - not sure though if this is neccessary but it leaves flexibility.

More tests for the fixture will follow.

[mberz]

Thanks for the updates. It looks much cleaner.
I only have a couple of comments left.

[mberz]

Could you please move the tests for the fixture to a separate file (something like test_edc_fixture or similar)?

[mberz]

I'm not sure what is happening here? If the input does not raise an error there should be no reason to test this here, or am I misunderstanding this?
If that is correct, It would make more sense to remove None from the parameters, so there's no need for the skip functionality.

[mberz]

Suggested change

	snr = 90 # -90 dBFS
	edc = make_edc_from_energy(energy=np.zeros(10), dynamic_range=snr)
	assert np.allclose(edc.time, 10 ** (-snr / 10))
	dynamic_range = 90 # -90 dBFS
	edc = make_edc_from_energy(
	energy=np.zeros(10), dynamic_range=dynamic_range)
	assert np.allclose(edc.time, 10 ** (-dynamic_range / 10))

The term SNR has another meaning and could be confusing here

[mberz]

Suggested change

	def make_edc_from_energy():
	def make_edc():

I feel that the _from_energy extension is confusing here, since it is only one of the parameterizations which can be used.