Unify regions stations

[cosunae]

When calling the showcase with a number of regions (for the plots) and locations for meteograms, currently we launch 1 task per region and location. While this maximizes parallelism, it is not efficient given these tasks are mostly IO bound that they will load the same data (entire horizontal plane) multiples times in order to select later on the region/location.

  params:
    - T_2M
    - SP_10M
    - TOT_PREC
  meteograms: false
  animations: true
  regions:
    - europe
    - switzerland
    - name: alpine_arc
      extent: [-16.0, 25.0, 30.0, 65.0]
      projection: orthographic

  stations: [JUN, COV, GOR, WFJ, SAE, SAM, DAV, ZER, ANT, VSBAS, BRT, LTB, GOS, CEV, BIA]

In this PR we modify the following:

• snakemake rule so that the plot_meteogram and plot_forecast_frame generate a list of plots (1 per region) and meteograms (1 per location)
• move the marimo files into a pure python script. This I would propose in order to make the code more readable in a pure VS code (not using marimo editor) and more homogeneous codebase (not mixing marimo and python). I checked with @dnerini, but let me know if you do not think this is a good change.
• adapt the python scripts to process multiple regions and select multiple locations.

[frazane]

Very nice. Fully support moving away from marimo notebooks here, I don't think they fit the use case. We might bring them back in the future for other more interactive things.

[frazane]

The try-except logic is duplicated

[jonasbhend]

Hi @cosunae Thanks for pushing this. Could you maybe rearrange the commits so that git(hub) understands that the files have been renamed? Otherwise it's impossible to identify changes made to the scripts. I could give you a hand if need be, just let me know (don't want to interfere with your work, as the fix will rewrite history).

[cosunae]

Hi @cosunae Thanks for pushing this. Could you maybe rearrange the commits so that git(hub) understands that the files have been renamed? Otherwise it's impossible to identify changes made to the scripts. I could give you a hand if need be, just let me know (don't want to interfere with your work, as the fix will rewrite history).

I will try, yes, but I think it wont help with the review because the files are very different (Inherently because the structure of a marimo and python file are).
I think In principle git should figure out these files are the same, but in this case it does not work because the diff is too large

[cosunae]

done, after rewriting the git history with a git mv and a force push, still similarity is very small, so PR will refuse to show it as a diff

[jonasbhend]

done, after rewriting the git history with a git mv and a force push, still similarity is very small, so PR will refuse to show it as a diff

Thanks and sorry for putting you on this. I didn't think about the python script / marimo issue

[cosunae]

anything missing here @frazane @jonasbhend ? could you take another look?

[jonasbhend]

Hi @cosunae, thanks for tackling this. It is looking good from what I can tell. However, when trying to run the interpolator config with meteograms enabled, I get an error. This is not coming from your PR, as the error is already in main (see below). Is this something we want to fix as part of this PR?

evalml showcase config/interpolators-ich1.yaml -n
host: balfrin-ln002
Building DAG of jobs...
InputFunctionException in rule plot_meteogram in file "/scratch/mch/bhendj/evalml-opr/workflow/rules/plot.smk", line 28:
Error:
  ValueError: No baseline zarr found for init time 202503010000
Wildcards:
  showcase=20260527_interpolators-ich1_7de8
  run_id=interpolator-tmp-d5aa-on-forecaster-c304-1e7e/7d5c
  init_time=202503010000
  param=T_2M
Traceback:
  File "/scratch/mch/bhendj/evalml-opr/workflow/rules/plot.smk", line 53, in <lambda>
  File "/scratch/mch/bhendj/evalml-opr/workflow/rules/plot.smk", line 24, in _get_available_baselines (rule plot_meteogram, line 30, /scratch/mch/bhendj/evalml-opr/workflow/rules/plot.smk)

[jonasbhend]

Hi @cosunae, thanks for tackling this. It is looking good from what I can tell. However, when trying to run the interpolator config with meteograms enabled, I get an error. This is not coming from your PR, as the error is already in main (see below). Is this something we want to fix as part of this PR?

evalml showcase config/interpolators-ich1.yaml -n
host: balfrin-ln002
Building DAG of jobs...
InputFunctionException in rule plot_meteogram in file "/scratch/mch/bhendj/evalml-opr/workflow/rules/plot.smk", line 28:
Error:
  ValueError: No baseline zarr found for init time 202503010000
Wildcards:
  showcase=20260527_interpolators-ich1_7de8
  run_id=interpolator-tmp-d5aa-on-forecaster-c304-1e7e/7d5c
  init_time=202503010000
  param=T_2M
Traceback:
  File "/scratch/mch/bhendj/evalml-opr/workflow/rules/plot.smk", line 53, in <lambda>
  File "/scratch/mch/bhendj/evalml-opr/workflow/rules/plot.smk", line 24, in _get_available_baselines (rule plot_meteogram, line 30, /scratch/mch/bhendj/evalml-opr/workflow/rules/plot.smk)

I'll try to fix this