Closed dazza-codes closed 4 years ago
Don't take this as authoritative commentary - it's learned from trial-n-error - part of the problem seems to be that conda-forge has renamed proj4<=6.1
to proj>6.1
and somewhere along the way the proj>6.1
project has split out the code project from the proj.db
database project, so if you only install proj>6.1
without the additional proj-datumgrid
stuff, then all the libs that look for CRS data can't find it. What makes a few things more complex is that often when working with CRS, it's used with GeoTIFF data files, so often the entire GIS/CRS ecosystem needs some combinations of geopandas
, pyproj
(with proj4<=6.1
and proj>6.1
+ proj-datumgrid*
), gdal
(with fiona
and rasterio
), with the underlying pandas
and numpy
libs. From this perspective, it's difficult to keep track of all the library versions that cooperate well together. It might help if the conda-forge project had some kind of packaging process that was similar to Debian packaging, where some top-level integrations try to identify and maintain some high-level packages that effectively perform integration tests on an entire GIS/CRS ecosystem.
I don't think that is the problem. Can you check what you have set as PROJ_LIB
?
echo $PROJ_LIB
I was unable to reproduce your issue with a fresh env and I get virtually the same packages.
I also cannot reproduce it reliably. IIRC, the PROJ_LIB
was not set explicitly (and should not need to be set???). If it was set and pointing to the wrong location, that could explain it. (I guess my work has moved on and hopefully the 👻 is gone.)
the PROJ_LIB was not set explicitly (and should not need to be set???).
It is automatically set by conda. Closing this b/c we don't have enough info to reproduce it.
Issue:
Various libraries cannot find the proj.db file in a conda-env; since geopandas often dictates what the "compatible" libraries are, this issue is posted here because proj is the core of all the CRS operations and finding a set of "compatible" versions will always need to involve geopandas.
Create conda env with latest versions
Environment (
conda list
):Details about
conda
and system (conda info
):