Refactoring steps-nowcast

[sidekock]

I started with the steps nowcast because that seemed more straightforward to tackle first. Since both methods should look alike, they must both be changed. I can't figure out how to run the test locally, so I will have to wait and see what the pull request test gives me as feedback.

This is a first suggestion how to split up the code but look forward to suggestions and improvments

[codecov]

Codecov Report

Attention: Patch coverage is 85.23316% with 57 lines in your changes missing coverage. Please review.

Project coverage is 83.99%. Comparing base (0f859c1) to head (b6c47af).

Files with missing lines	Patch %	Lines
pysteps/nowcasts/steps.py	85.23%	57 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #436      +/-   ##
==========================================
+ Coverage   83.89%   83.99%   +0.10%     
==========================================
  Files         160      160              
  Lines       12902    13005     +103     
==========================================
+ Hits        10824    10924     +100     
- Misses       2078     2081       +3     

Flag	Coverage Δ
unit_tests	83.99% <85.23%> (+0.10%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[sidekock]

I have a question about the Codacy Static Code Analysis. It sais the method _check_inputs is to complex while this has barely been touched since the last version, I moved a few extra lines here because that just makes more sense. Should I split up that part of the code to remove the error or not?

[dnerini]

hi @sidekock , first of all, thanks for taking up this heroic task :-)

can't figure out how to run the test locally

which part? to run the test locally you need to setup the test data correctly, see https://pysteps.readthedocs.io/en/stable/user_guide/example_data.html#example-data and https://pysteps.readthedocs.io/en/stable/developer_guide/test_pysteps.html

I have a question about the Codacy Static Code Analysis. It sais the method _check_inputs is to complex while this has barely been touched since the last version, I moved a few extra lines here because that just makes more sense. Should I split up that part of the code to remove the error or not?

I agree with you, _check_inputs seems well in line with the complexity of the other methods, and feels like breaking it down into smaller pieces would somewhat overcrow your class. I silenced this issue on codacy.

[sidekock]

@mats-knmi @RubenImhoff Could you take a look at this and give feedback? That way I can go on to the blending version. I don't want to start working on that one if this one is not in a finalized state.

[RubenImhoff]

@sidekock, I will try to do so asap.

[RubenImhoff]

Hi @sidekock, super nice work! Thanks a lot for picking up this tremendous task. I like that we use a class now, it is much cleaner this way.

I think all looks good, but I've made some suggestions to change some variable names to a longer name.

In addition, perhaps utils.nowcast_main_loop also can use a refactor based on the refactor here. In this nowcast_main_loop, the subtimesteps are also used. I remember that we were discussing about this during the hackathon. Perhaps this is the time to clean that up a bit, too? I might be the most difficult/unreadable part of the code.

[RubenImhoff]

Should we add a test where measure_time is True (and False)? That would probably tick the box for the test coverage.

[sidekock]

We could to this but I don't know if it is worth it, its an internal method with almost no functionality...

[sidekock]

Hi @sidekock, super nice work! Thanks a lot for picking up this tremendous task. I like that we use a class now, it is much cleaner this way.

I think all looks good, but I've made some suggestions to change some variable names to a longer name.

In addition, perhaps utils.nowcast_main_loop also can use a refactor based on the refactor here. In this nowcast_main_loop, the subtimesteps are also used. I remember that we were discussing about this during the hackathon. Perhaps this is the time to clean that up a bit, too? I might be the most difficult/unreadable part of the code.

I agree this might be worth looking at but the thing I am a bit worried about is the that this loop is used for all nowcasters so minor changes could affect the code in a lot of different places. I was thinking to tackle the loop in the blending first (since that only affect one place) and then transfer what I learned back. Its a bit of double work but it seems easier to debug that way what works and does not work. Any feedback on the way I split up the code?