Closed knaaptime closed 5 years ago
We now have ncld rasters for different years as well as 2000 and 2010 block-level census data in the quilt bucket (i've got housing and population counts for 2010 but no attributes yet for 2000) so there are lots of ways to improve this quickly
hm.. interesting, but if this is the case, I think we need to tweak the harmonize's tobler function since I believe it only presupposes one NLCD raster.
ps.: I think that "numpy
not found" issue with the travis build is happening again.
hm.. interesting, but if this is the case, I think we need to tweak the harmonize's tobler function since I believe it only presupposes one NLCD raster.
I just meant that we have them available internally so that we can use convenience methods for providing the paths and the users dont need to download the data manually (then they can choose which raster is more appropriate for their problem)
ps.: I think that "numpy not found" issue with the travis build is happening again.
yeah, the tests are broken until the quilt conda-forge recipe gets merged
the second commit here uses the if __name__ == "__main__"
convention to mask the dash app and the utilities that need internet connection to (1) reduce memory overhead and (2) make sure that geosnap still works without an internet connection
tests are passing locally
one thing to note is that we're still going to run into a CI build error even when quilt is added to conda forge because we now import tobler, which has unspecified dependencies (shap, xgboost) at the moment
one thing to note is that we're still going to run into a CI build error even when quilt is added to conda forge because we now import tobler, which has unspecified dependencies (shap, xgboost) at the moment
Good catch! Will fix that.
this is a quick and dirty start to the harmonization module. Currently, all it does is shade the harmonize function from tobler and make it available as a method on the Community class.
We now have ncld rasters for different years as well as 2000 and 2010 block-level census data in the quilt bucket (i've got housing and population counts for 2010 but no attributes yet for 2000) so there are lots of ways to improve this quickly