Updated the census notebook as much as seems feasible right now. No longer needs geoviews, since holoviews has long included tile support and the data has already been projected to web mercator anyway. Also now uses rasterize, Bokeh-based colorbars, and hovering in the final example, along with more recent tools in the environment. There are several remaining issues that don't block merging but are important to fix:
[ ] The environment pins to dask=2022.11.1 because later versions of dask hit #1206
[ ] pandas is pinned <2 because the earlier versions of dask don't support pandas 2
[ ] There's a (hopefully temporary) workround df = df.categorize('race') in there to avoid hitting #1202
[ ] The parquet reader is set to fastparquet due to issues with the default pyarrow engine, though #1202's fix may also fix that issue
[ ] The conda channel is set to conda-forge rather than defaults, because defaults has only an old and incompatible version of fastparquet (0.5.0, IIRC).
[ ] The environment file doesn't enable arm-osx64 for locking, because the environment fails to solve on that platform
Once these other issues are addressed it would be nice to update the census project again, removing the pins and workarounds.
Updated the census notebook as much as seems feasible right now. No longer needs geoviews, since holoviews has long included tile support and the data has already been projected to web mercator anyway. Also now uses
rasterize
, Bokeh-based colorbars, and hovering in the final example, along with more recent tools in the environment. There are several remaining issues that don't block merging but are important to fix:df = df.categorize('race')
in there to avoid hitting #1202fastparquet
due to issues with the defaultpyarrow
engine, though #1202's fix may also fix that issuefastparquet
(0.5.0, IIRC).arm-osx64
for locking, because the environment fails to solve on that platformOnce these other issues are addressed it would be nice to update the census project again, removing the pins and workarounds.