Adapt pysteps to allow for postprocessing plugins #405
Conversation
pysteps/postprocessing/__init__.py
Outdated
| @@ -2,3 +2,4 @@ | |||
| """Methods for post-processing of forecasts.""" | |||
|
|
|||
| from . import ensemblestats | |||
| from postprocessors import * | |||
There was a problem hiding this comment.
| from postprocessors import * | |
| from .postprocessors import * |
There was a problem hiding this comment.
just wondering if "postprocessors" is a good name? sounds a bit too vague, particularly as a submodule of "postprocessing" ... can we try to be a bit more specific? perhaps use "diagnostic" instead? what do you think? @ladc ?
There was a problem hiding this comment.
Yes good point - "diagnostic" makes sense if we also want to add other postprocessors in the future which are not purely diagnostic (such as bias correction methods).
Codecov ReportAttention: Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## master #405 +/- ##
==========================================
+ Coverage 83.67% 84.01% +0.34%
==========================================
Files 159 162 +3
Lines 12649 12994 +345
==========================================
+ Hits 10584 10917 +333
- Misses 2065 2077 +12
Flags with carried forward coverage won't be shown. Click here to find out more. ☔ View full report in Codecov by Sentry. |
Also reformatted with black. Tests were added for the get_method and diagnostics_info functions in the postprocessing interface. Tests for the discover_diagnostics function will be written once these changes have been merged as then the cookiecutter plugins can then be properly tested.
|
Hi @dnerini, could you please review this pull request when you have the time? The extended version of the cookiecutter plugin template which can be used to create importer, diagnostic postprocessing, and nowcast plugins is nearly finished but I require the postprocessing interface to be merged to ensure that everything is behaving correctly when tested. Cheers |
|
Thanks for this contribution! We discussed how to make it more generic and @joeycasey87 will try to refactor it tomorrow to allow for various kinds of post-processors including new ensemblestats and bias correction methods, for example. To be continued. |
Changed the test file to match the test updated pysteps test file
These were necessary for the IO plugins, but less so for the diagnostics.
Avoid duplicate code, refactor into functions. Also fix a small typo causing a bug: postprocessor_s_
Added an infrastructure.py file to the postprocessing folder which should operate in the same way as the interface file does for the importers in the io folder. The postprocessor file is currently effectively empty as no postprocessor plugins have been added yet. A line has been added to the init file so that it should act similar to the init file in the io folder.