Refactoring based on covr report

[bahadzie]

No description provided.

[bahadzie]

This draft is based on the covr report of epidemics. I have changed the default parameters for model_default_cpp and model_default_rand removed some code that is no longer necessary due to this change.

@pratikunterwegs, I would like your opinion on this and whether it's a good idea to proceed with the rest of the report. It is related to issue #105

[pratikunterwegs]

Thanks for opening this @bahadzie - I would say this is not the way to go, for a couple of reasons:

1. The convention in R functions we use is to indicate optional arguments as NULL. Due to the structure of internal C++ functions, we have to substitute NULL with no_*_intervention() within model_*_*(), but ideally this wouldn't be necessary. Here, the intervention set, vaccinations, and time dependence are all optional, and should be indicated by NULL.
2. As you have found in the vignette, the function now requires a dummy contact intervention to be specified even when the intention is to have only a rate intervention. This adds (a little) work for the user that the package can/should do for them, as it's an optional argument overall.

What you could do is to adopt the approach used in {finalsize}, where the default control settings list has its elements replaced if they are passed by the user. This would indeed cut down on the cross-checking required in argument checker functions .check_args_model_*().

Overall, what I would say is also that the {epidemics} API is changing substantially - see #160. It might be good to revisit #105 once the API has stabilised in a few weeks, as otherwise a number of the changes made now would soon be overwritten.

[pratikunterwegs]

Closing this PR as #176 introduced much better testing for all ODE models, and future PRs will improve coverage for the Ebola model as well.