Do not set the collections transform

[andrewdnolan]

Not setting the collection transform, for GeoAxes instances, results in ~50% (or more) speed up in rendering. I reached out to the cartopy dev's about whether this was really needed, but have not heard. If we were making an obvious mistake here, I think they would have said something.

Native mesh contouring (#52) also produces matplotlib collections on map projected figures. We do not set the transform there and have not had any issues. So, I'm feeling comfortable moving forward with this (for now at least).

If we get a response or issues arise, we can always revert these changes.

Closes #33.

[andrewdnolan]

Testing: IcoswISC30E3r5

Without polycollection transform set:

$ hyperfine ./IcoswISC30E3r5.py 
Benchmark 1: ./IcoswISC30E3r5.py
  Time (mean ± σ):      4.947 s ±  0.120 s    [User: 4.267 s, System: 0.379 s]
  Range (min … max):    4.831 s …  5.132 s    10 runs

With polycollection transform set:

$ hyperfine ./IcoswISC30E3r5.py 
Benchmark 1: ./IcoswISC30E3r5.py
  Time (mean ± σ):     11.314 s ±  0.172 s    [User: 10.683 s, System: 0.387 s]
  Range (min … max):   11.039 s … 11.588 s    10 runsuns

./IcoswISC30E3r5.py

#!/usr/bin/env python

import cartopy.crs as ccrs
import matplotlib.pyplot as plt
import tempfile

from pathlib import Path

import mosaic


projection = ccrs.PlateCarree()
transform = ccrs.Geodetic()

ds = mosaic.datasets.open_dataset("mpaso.IcoswISC30E3r5")

descriptor = mosaic.Descriptor(ds, projection=projection, transform=transform)

fig, ax = plt.subplots(subplot_kw={"projection": projection})

mosaic.polypcolor(ax, descriptor, ds.bottomDepth)

with tempfile.TemporaryDirectory() as temp_dir:
    fig.savefig(
        Path(temp_dir) / "IcoswISC30E3r5.png", dpi=300, bbox_inches="tight"
    )

[andrewdnolan]

Testing: RRSwISC6to18E3r5

Without polycollection transform set:

$ hyperfine ./RRSwISC6to18E3r5.py 
Benchmark 1: ./RRSwISC6to18E3r5.py
  Time (mean ± σ):     28.381 s ±  0.463 s    [User: 26.426 s, System: 1.353 s]
  Range (min … max):   27.811 s … 29.530 s    10 runs

With polycollection transform set:

$ hyperfine ./RRSwISC6to18E3r5.py      
Benchmark 1: ./RRSwISC6to18E3r5.py
  Time (mean ± σ):     83.419 s ±  1.204 s    [User: 81.553 s, System: 1.345 s]
  Range (min … max):   81.944 s … 85.999 s    10 runs

./RRSwISC6to18E3r5.py

#!/usr/bin/env python

import cartopy.crs as ccrs
import matplotlib.pyplot as plt
import tempfile
import xarray as xr

from pathlib import Path

import mosaic

projection = ccrs.PlateCarree()
transform = ccrs.Geodetic()

ds = xr.open_dataset(
    "/Users/anolan/Downloads/mpaso.RRSwISC6to18E3r5.20250205.nc"
)

descriptor = mosaic.Descriptor(ds, projection=projection, transform=transform)

fig, ax = plt.subplots(subplot_kw={"projection": projection})

mosaic.polypcolor(ax, descriptor, ds.bottomDepth)

# fig.savefig("RRSwISC6to18E3r5.png", dpi=300, bbox_inches="tight")
with tempfile.TemporaryDirectory() as temp_dir:
    fig.savefig(
        Path(temp_dir) / "RRSwISC6to18E3r5.png", dpi=300, bbox_inches="tight"
    )

[andrewdnolan]

Testing

Using cmp on png files written with the collections transform set and unset produce bit-for-bit matches.