update infrastructure

[knaaptime]

@martinfleis should we go ahead and make the switch from query_bulk to query here as well?

[codecov[bot]]

Codecov Report

Merging #174 (f78bfbb) into master (97205be) will decrease coverage by 23.05%. The diff coverage is 91.66%.

@@             Coverage Diff             @@
##           master     #174       +/-   ##
===========================================
- Coverage   72.91%   49.87%   -23.05%     
===========================================
  Files          20       14        -6     
  Lines        1045      786      -259     
===========================================
- Hits          762      392      -370     
- Misses        283      394      +111     

Impacted Files	Coverage Δ
tobler/area_weighted/area_interpolate.py	86.89% <66.66%> (ø)
tobler/__init__.py	100.00% <100.00%> (ø)
tobler/area_weighted/area_join.py	100.00% <100.00%> (ø)
tobler/pycno/pycno.py	82.97% <100.00%> (+0.18%)	:arrow_up:

... and 8 files with indirect coverage changes

[martinfleis]

@martinfleis should we go ahead and make the switch from query_bulk to query here as well?

If you simply replace it, then you need to pin geopandas to 0.13 or newer. You can do it conditionally based on the version but not sure if we need to go that far. But yeah, it would be nice to do something about it to silence the warning.

[jGaboardi]

Do we think these differences can be simply attributes to precision? Seems a bit high for that.

FAILED tobler/tests/test_utils.py::test_h3_multipoly - AssertionError: 
Arrays are not almost equal to 0 decimals
 ACTUAL: 1106846455924.778
 DESIRED: 1106844905155.1118

[knaaptime]

not sure what it is. Agree too high to be precision, and too odd that it's only windows. I'm not terribly worried though. Those aren't really appropriate interpolators anyway. They were highly experimental when we put them in and havent had time to return to them. I'm confident nobody is actually using those

[jGaboardi]

Seems very likely the failures is directly related to h3py on windows. Maybe simply skipping that one test for now until finding the time to dig into it?

[knaaptime]

any chance there's like a #nowin flag for that or something? (like how conda recipes work)

[jGaboardi]

You could:

• decorate the test with an xfail --> see here in spopt
• do a skipif --> see geopandas

[knaaptime]

ok, skipif(platform.system()=='Windows') would be easy but i think its better to just xfail here. Good refs, thanks