Fix issues with unwrapping longitudes in RangeXY stream

[philippjfr]

This PR introduced two separate issues: #722

Fixes #753

[codecov]

Codecov Report

Attention: Patch coverage is 87.50000% with 2 lines in your changes missing coverage. Please review.

Project coverage is 67.28%. Comparing base (5f02c71) to head (eea3ee9).

Files with missing lines	Patch %	Lines
geoviews/plotting/bokeh/callbacks.py	0.00%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #756      +/-   ##
==========================================
+ Coverage   67.21%   67.28%   +0.07%     
==========================================
  Files          44       44              
  Lines        4642     4649       +7     
==========================================
+ Hits         3120     3128       +8     
+ Misses       1522     1521       -1     

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

[ahuang11]

Something is still up in this PR

!wget -nc https://downloads.psl.noaa.gov/Datasets/ncep.reanalysis/Monthlies/surface/air.day.ltm.nc
import xarray as xr
import holoviews as hv
from holoviews.operation.datashader import rasterize
import geoviews as gv
gv.extension("bokeh")

ds = xr.open_dataset("air.day.ltm.nc", engine="h5netcdf", decode_times=False).isel(time=0)
rasterize((gv.Image(ds["air"], ["lon", "lat"], ["air"]))) * gv.feature.coastline()

[ahuang11]

I'm starting to think #722 wasn't the right solution because if the data extents are global from 0 to 360, the ranges wrap to -180 to 180, meaning only 0 to 180 remains while 180-360 is cropped and discarded, resulting in half image:

I think we need to either warn and project the lons to -180 to 180 or raise an error mentioning that data must be from -180 to 180 for operations to work properly, i.e.

ds["lon"] = (ds["lon"] + 180) % 360 - 180
ds = ds.sortby("lon")

Separately, there seems to be some padding with the x0/x1:

            self._unwrap_lons = 0 <= x0 <= 360 and 180 <= x1 <= 360
            print(x0, x1, "Xs", self._unwrap_lons)

# outputs
# -1.25 358.75 Xs False
# -1.25 358.75 Xs False

[ahuang11]

Actually, if we extend the range from -180 to 540, it seems to work as expected:

Screen.Recording.2024-10-16.at.10.38.58.AM.mov

Screen.Recording.2024-10-16.at.10.45.50.AM.mov

[hoxbro]

Can you add test cases for all the cases you have tested locally so we don't get any regressions?

If needed, I can set up UI tests.

[ahuang11]

Yes can you help me set up ui tests or show me a reference?

[ahuang11]

Yes can you help me set up ui tests or show me a reference?

[hoxbro]

I have opened a PR with the infrastructure here: #760.

I will also try to fix the CI in it and then merge it.