Refactor _resampled_scene() and _reduce_data() methods of the Scene class

[pnuu]

This PR refactors Scene._resampled_scene() that @gerritholl reported to be complicated in #3168 (comment) . As there was a change to _reduce_data(), I did some additional refactoring to it too.

• Closes #xxxx
• Tests added
• Fully documented
• Add your name to AUTHORS.md if not there already

[codecov]

Codecov Report

❌ Patch coverage is 92.85714% with 5 lines in your changes missing coverage. Please review.
✅ Project coverage is 96.33%. Comparing base (fbfbe21) to head (f70d80f).
⚠️ Report is 6 commits behind head on main.

Files with missing lines	Patch %	Lines
satpy/scene.py	92.85%	5 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #3178      +/-   ##
==========================================
+ Coverage   96.32%   96.33%   +0.01%     
==========================================
  Files         466      467       +1     
  Lines       59096    59143      +47     
==========================================
+ Hits        56922    56977      +55     
+ Misses       2174     2166       -8     

Flag	Coverage Δ
behaviourtests	3.61% <24.28%> (+0.02%)	⬆️
unittests	96.42% <92.85%> (+0.01%)	⬆️

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

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

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[pnuu]

I'll do some more refactoring to _reduce_data().

[coveralls]

Pull Request Test Coverage Report for Build 16566762203

Details

• 51 of 56 (91.07%) changed or added relevant lines in 1 file are covered.
• No unchanged relevant lines lost coverage.
• Overall coverage remained the same at 96.381%

---

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
satpy/scene.py	51	56	91.07%

Totals
Change from base Build 16528353281:	0.0%
Covered Lines:	56082
Relevant Lines:	58188

---

💛 - Coveralls

[gerritholl]

Thanks! I've merged this into #3168 but there is still an issue with missing test coverage.

[pnuu]

I made the Fine Choice(tm) of asking Copilot (Claude Sonnet 4.6) to refactor the method. After couple of iterations we ended up with this version that uses a data class to hold resampling context and a method that is called for each resampled dataset.

[djhoese]

Interesting...I've never personally used data classes beyond toy programs and watching programming videos. What it is doing here makes me wonder if this deserves a full blown helper class where instead of context in these sub-methods it is just self..

[djhoese]

And sorry, I don't fully remember all the previous context for this PR and past thoughts I had about it.

[pnuu]

Unrelated test failure:

ERROR satpy/tests/reader_tests/test_mviri_l1b_fiduceo_nc.py::TestFiduceoMviriFileHandlers - satpy/tests/reader_tests/test_mviri_l1b_fiduceo_nc.py::TestFiduceoMviriFileHandlers::test_init: duplicate parametrization of 'projection_longitude'

I can't replicate this locally with package versions I have.

[pnuu]

Seems to be somehow related to pytest-dev/pytest#13974 that have several links from the few last days.

[pnuu]

Possible fix in #3415

[gerritholl]

Needs more test coverage?

[gerritholl]

This PR increases the overhead/runtime of _reduce_data by a factor 5. Test script:

import os
from satpy import Scene
from glob import glob
from satpy.tests.utils import CustomScheduler
from satpy.writers.core.compute import compute_writer_results
import dask.config
import time
reduce = True
with dask.config.set(scheduler=CustomScheduler(max_computes=0)):
    seviri_files = glob("/media/nas/x21308/scratch/SEVIRI/202103300900/H-000*")
    sc = Scene(filenames={"seviri_l1b_hrit": seviri_files})
    nms = {"ash", "natural_color", "dwd_RGB_12_12_9i_N", "convection",
           "dwd_IR_108", "dwd_HRV", "airmass", "dust"}
    sc.load(nms, generate=False)
    tic = time.time()
    ls = sc.resample("nqeuro3km",
                     resampler="gradient_search", reduce_data=reduce)
    print("resample", time.time() - tic, "seconds")
    res_all = []
    for nm in nms:
        res_all.append(ls.save_dataset(
                nm,
                "{platform_name}-{sensor}-{name}-{area.area_id}-{start_time:%Y%m%d%H%M}-geotiff.tif",
                writer="ninjogeotiff",
                ChannelID="IR 10.8",
                DataType="GORN",
                PhysicValue="Temperature",
                PhysicUnit="K",
                SatelliteNameID="METEOSAT Europe-Atlantic",
                fill_value=0,
                #dynamic_fields={"mean_time"},
                compute=False))

Satpy main: 8.2 seconds

This PR: 44.0 seconds

[gerritholl]

Profiling reveals:

• the time per call for _reduce_data increases from 1.02 to 5.97 seconds.
• The number of calls to spherical.py:492(get_next_intersection) increases from 302 to 1812.
• The number of calls to geometry.py:2654(get_area_slices) (first entry point into pyresample) increases from 2 to 12.

Here both profiles:

profile-results.zip

[gerritholl]

Serious performance regression in _reduce_data. The reductions dictionary appears no longer functional, leading to _reduce_data being called for every dataset. For my test script, that increases the time for calling resample from ca. 8 seconds to ca. 40 seconds. Without computing anything.