Closed brendan-ward closed 8 months ago
For #306, enabling the sql tests lead to a bug being found, so it would be good if it can make it into version 0.7 still.
I did fix the bug in the PR already, but it is still marked as draft because I still needed to add an optimization to use the fast code path for newer gdal versions as a gdal bug was solved to make this possible. However, it is a small change to make, so it shouldn 't take long.
What's the bug exactly? Because right now the skip_features for Arrow is only enabled for GDAL >= 3.8, I think? (and in GDAL 3.8 this bug will be fixed?)
I would say that both PRs are not super critical to get in for 0.7 (but of course if they are ready, let's merge them), and would prefer to do a release shortly.
What's the bug exactly? Because right now the skip_features for Arrow is only enabled for GDAL >= 3.8, I think? (and in GDAL 3.8 this bug will be fixed?)
On second thought you are right:
I would say that both PRs are not super critical to get in for 0.7 (but of course if they are ready, let's merge them), and would prefer to do a release shortly.
I agree.
I moved both #304 (we're not yet decided on the API/approach) and #306 (hard to test without recent version GDAL including fixes) to a later milestone, so we're ready to proceed with this version.
I'll try to push the release momentarily.
Release is out: https://github.com/geopandas/pyogrio/releases/tag/v0.7.0 wheels on PyPI: https://pypi.org/project/pyogrio/#files
One problem that turned up on https://github.com/conda-forge/pyogrio-feedstock/pull/38: the PR https://github.com/geopandas/pyogrio/pull/285 added a usage of from packaging.version import Version
in non-test code, and so essentially packaging
has become a hard dependency of pyogrio, while it is not listed in our dependencies.
So we should either add it as dependency, or remove that one usage (I think we could also easily remove it, it's just checking that shapely > 2, so that can be done manually as well)
See https://github.com/geopandas/pyogrio/pull/318 for avoiding usage of it. Of course, simply adding it to
https://github.com/geopandas/pyogrio/blob/f0c82b6cc012b36c98fcd6ac27fee7d7c6e117bb/setup.py#L230
is also a trivial change, and should be no problem (for example, also geopandas depends on it, so most users will already have it anyway)
Outstanding PRs:
304 : need to decide if we are going to proceed with
fids
support for Arrow for this version (either via internalwhere
or filtering by batch) or bump to next version306 : we haven't tagged to a milestone and this is still draft, so I assume it is OK to defer this to next version