We have released GeoPandas 1.0! Yes, I am excited and a bit relieved as it took a bit longer than expected. Anyway, it’s out and we’re waiting to hear what we broke 🙃.
It is a major milestone for GeoPandas, not only in a semantic sense, but it literally closes a long development cycle. If you have been following the ecosystem for a while, you might know the story, but it is worth refreshing your memories. So, very briefly:
Originally (by that I mean 10 years ago when Kelsey Jordahl and a few others put
together the first versions), a typical geometric operation on a GeoPandas object was an
internal for loop over an array of scalar
Shapely geometries. This was slow and there was no
way to speed it up given the design of Shapely. So
PyGEOS was born - a new, fully vectorised connection
to GEOS. And boy, was PyGEOS fast. Of course, GeoPandas included support for PyGEOS, but
since it was considered a bit experimental, it was done alongside Shapely. Later,
the community got together and PyGEOS was merged with Shapely to become Shapely 2.0.
This was an absolutely amazing decision from which everyone benefited, but GeoPandas,
which tends to provide long-term support for older versions of dependencies, had to
implement third geometry engine support, exposing the same
GEOS library. You can imagine the mess we got into. Every new
geometry method had to be implemented three times, once for Shapely 1.x, once for PyGEOS
and once for Shapely 2. The latter two were super similar, but there were minor
differences.
This is all gone now. Hurray! With the 1.0 release, we have removed all this complexity and only support Shapely 2. While most users don’t care and never noticed any of this, simply because things just worked, it is an important step for us.
Performance was also an issue with I/O. Scalar-based Fiona was replaced by vectorised PyOGRIO, which now comes with GeoPandas by default.
There is a long list of improvements in the release notes. A notable one is API parity with Shapely (all relevant functions are exposed as GeoPandas methods), but there’s much more. Have a look for yourself.
What comes next? We’ll be complicating our lives again with another geometry engine. This time not GEOS, but S2, which supports spherical geometry and operations on spheres. We’re finally going to break the main limitation of GEOS, which is restricted to operations on the Euclidean plane. When will this happen? Some preliminary support may be ready by the end of the year, but knowing how things are going with our promises, probably sometime next year.
We’re also planning some other things, but you can check them out on the roadmap.
I also wrote a post earlier about the major API changes in GeoPandas. If you are wondering what might break your code, this is the post to check.
Let’s get updating!
pip install geopandas==1.0.0
conda install geopandas==1.0.0 -c conda-forge
PS: There will likely be bugs. Please report them! We can’t fix what we’re not aware of.