Adding unit tests to statistics

[jamescrowley]

Replaces #35 to solve half of #32

I made some changes from @Phuc-Hong-Tran's approach, hope this is ok @chienleng

1. I added tests to try and capture intended behaviour more explicitly, rather than relying on large JSON files and a large expected output where (at least as someone new to the project), it wasn't clear.
2. Added tests for Statistic class itself, total minus loads.
3. Tidied up some of the type definitions to remove errors
4. Fixed a condition that allowed for an infinite loop for interpolate (last commit)
5. Added a test coverage npm command
6. I placed the tests alongside their corresponding file (personal preference as for me makes it much clearer which files have corresponding tests, and they are right there when editing) - happy to switch if you disagree.

I should also say

• I'm not entirely clear on the intended behaviour for the interpolate-data steps (and correspondingly in the Statistic class). I've added tests but it would be good to review that the inputs/outputs actually make sense, or if there can be some more meaningful examples.
• I have not yet added any tests for timeseries
• If you have any preference re unit tests on the individual functions, vs the 'public' interface boundary of Statistics let me know - I ended up doing both but personally would prioritise the latter if doing time series, in hindsight.

[jamescrowley]

See here re intended behaviour. Why 4 values? Why does last get set to 09:38?

[chienleng]

not entirely sure, but I am guessing this is trying to test a 1min interval interpolation.

[jamescrowley]

@chienleng yes - I wrote the test, in case that wasn't clear :D This was just related to not fully understanding the intent behind the behaviour around start/last and the number of interpolated values - and therefore harder to write a meaningful test for it. So long as you think the test actually makes sense, then 👍

[chienleng]

ah, got it. :) Understood about the confusion of those values. The Stats class will need a refactor at some point, I am mostly porting this from a previous code base from OpenNEM.

I will take a closer look at reviewing this some time this week.

thanks for your help though.

[jamescrowley]

And here - I wasn't clear why there needed to be another entry in the data array. Would be great to make the test explicit to explain further.

[chienleng]

it's because the function only performs the interpolation if it detects there is a difference in the intervals returned. Will add that comment to the interpolate-data function.

[jamescrowley]

@chienleng thanks :) Given the desired minInterval is passed in, the context I think I'm missing was around the relationship of the start/end date of those StatsData objects that match the desired minInterval and the corresponding filtering - ie around https://github.com/opennem/openelectricity/pull/43/files/bc0fbba86875f99da82bcbcf2881b4f07c6e1946#diff-fc1bb786b10f9d5da7716d320830a20042c0a7e5b60670ff93f3f9798ad74beeR56-R66 - as that results in an empty data field being returned otherwise. This only stood out while trying to build test cases - I haven't yet dug into the broader usage when I'm sure this would be clear, so no need to address as such :)