Reason or Problem
datashader is currently listed under the examples extra in setup.cfg. Once the source code, user-guide notebooks, and example notebooks no longer import it, the dependency declaration should go away so a fresh pip install xarray-spatial[examples] no longer pulls datashader.
This is the closing step of the wider effort to remove datashader as a dependency. xarray-spatial's own rasterize() (xrspatial/rasterize.py, available as arr.xrs.rasterize(...)) covers the vector-rasterization use cases, and matplotlib covers the visualization use cases.
Proposal
- Remove the
datashader >= 0.15.0 line from setup.cfg (the examples extra).
- Make sure
matplotlib (and any colormap source the notebooks now rely on) is present in the appropriate extra.
- Remove the
datashader row from docs/source/getting_started/installation.rst.
Done when
- A clean environment installs xarray-spatial without datashader, and
python -c "import xrspatial" succeeds.
grep -rn datashader xrspatial/ setup.cfg docs/source/getting_started/ returns nothing.
The only datashader references left in the repo after this are the comparison benchmark and the historical-credit prose, both kept on purpose.
Blocked by
This must land last, after every issue below, otherwise [examples] users lose datashader while the notebooks still import it.
Source and benchmark:
User-guide notebooks:
Example notebooks and script:
Reason or Problem
datashader is currently listed under the
examplesextra insetup.cfg. Once the source code, user-guide notebooks, and example notebooks no longer import it, the dependency declaration should go away so a freshpip install xarray-spatial[examples]no longer pulls datashader.This is the closing step of the wider effort to remove datashader as a dependency. xarray-spatial's own
rasterize()(xrspatial/rasterize.py, available asarr.xrs.rasterize(...)) covers the vector-rasterization use cases, and matplotlib covers the visualization use cases.Proposal
datashader >= 0.15.0line fromsetup.cfg(theexamplesextra).matplotlib(and any colormap source the notebooks now rely on) is present in the appropriate extra.datashaderrow fromdocs/source/getting_started/installation.rst.Done when
python -c "import xrspatial"succeeds.grep -rn datashader xrspatial/ setup.cfg docs/source/getting_started/returns nothing.The only datashader references left in the repo after this are the comparison benchmark and the historical-credit prose, both kept on purpose.
Blocked by
This must land last, after every issue below, otherwise
[examples]users lose datashader while the notebooks still import it.Source and benchmark:
User-guide notebooks:
Example notebooks and script: