sort out tests

[drbenvincent]

At the moment I've got a strange ad hoc mixture of python scripts which I use to debug (using MS Visual Studio Code) and some unit tests (using unittest). The debugging and stepping through code is very handy, but I've not figured out how to do that in MS VS Code yet.

• [x] 1) move all debugging scripts into a /dev folder
• [x] 2) See if it's possible to to the debugging process as invoked through proper test code. If so, then I can get rid of my scripts and turn them into proper tests. Answer: yes, in VS Code we can set breakpoints and do Debug Test rather than Run Test and it works well.

And I'm not convinced that unittest is the best way forward. It might be that pytest is simpler.

• [x] Quickly look at pytest vs unittest and decide. Decision: go for pytest.

Tests

Models

• [x] model instantiation
• [x] test models by calling their calc_decision_variable method. Will need to set up example data
• [x] test update_beliefs()
• [x] getting a simulated response (for a made up design) from models. This will require having true parameters. We could require example made up true parameters to be required for each model, but this is annoying and confusing. So instead, the test should generate it's own true parameters made from sampling from the model's priors. Taking a random sample might result in some stochasticity in the results of the tests (although it should work for any valid parameters), so briefly look into whether we can extract the central tendency from all prior distributions. If not, it's a pain, so just sample.

Designs

• [x] creating a design instance DARC_Designs using a variety of approaches (ie customisations of the design space

[drbenvincent]

Decision: move forward with pytest

[drbenvincent]

Have made a pretty good start. Picked up pytest relatively quickly, got some nice parameterised tests going on. Got some good unit-level tests of models and designs going on.

As of now, we have 48 passing, 3 failing tests. We'll deal with this in another issue. For now, we have achieved the objective of laying down a good initial spread of unit tests.