GeoPandas 1.0 release

[martinfleis]

An issue to track the remaining tasks we'd like to finish before GeoPandas 1.0.

The roadmap says:

• Fully vectorized geometry engine (done in Migration from PyGEOS to Shapely 2.0 #2691)
• Feature parity with shapely (tracked in ENH: expose all shapely functions as GeoSeries/GeometryArray methods #2010)
• Clarity of the API (probably primarily API: use union_all() and deprecate unary_union #3007 and some finished deprecations)
• Pruned dependencies (GDAL is optional, PROJ not yet - Make pyproj an optional dependency (still required for CRS) #2077)

Explicit to-do I think we should ensure is in:

• ENH: expose all shapely functions as GeoSeries/GeometryArray methods #2010
• pyogrio engine as a default
• pyogrio functionality exposed from geopandas (e.g. list_layers ENH: Function to show layers in a particular file #2837)
• soft-dependency on pyproj (Make pyproj an optional dependency (still required for CRS) #2077)
• BUG/API: also preserve right index name in spatial join #2144 (API spatial join: need for index retention? #498)

Discussion required:

• updated roadmap for post-1.0
• API: default behaviour of alignment in binary ops #1832

And finally, the milestone itself containing the remaining issues/PRs:

• https://github.com/geopandas/geopandas/milestones/1.0

The release is currently planned for March 31, 2024. If we go a bit beyond that is fine I guess but would try to ensure we don't push it too far.

[m-richards]

I think it would be good to try and solve #1038 for 1.0 - the discussion there is old but the proposed solution is potentially breaking, so better to try and include and resolve for 1.0.
I took a scan through the issues / open PRs and noticed this one, probably also worth revisting / deciding whether to action #3019 / #1975 sooner than later.
(not sure if I will be able to make the scheduled meeting tomorrow/ today)

[kylebarron]

I'd be motivated to get #3156 in to 1.0 if we can agree on a proposed API. I'd be happy to make a PR

[martinfleis]

@kylebarron I don't know much about that but if you can get @jorisvandenbossche on board I won't object.

[martinfleis]

I suggest the following release plan:

Start with 1.0-alpha1 by the end of March that contains all the API-related changes but not necessarily all the bugfixes and enhancements. For that, I believe we need

• Change the default IO engine to pyogrio #3223
• API: change the align default to None allowing silencing alignment warning in binary ops #3212
• ENH: harmonize read_file column selection keywords (support columns for fiona, support ignore_fields/include_fields for pyogrio for back compat) #3133
• Look into numpy / pandas dev failures
• CLN: remove plotting compatibility aliases #3190
• BUG/API: Set geometry api where col is named / with drop=True #3237
• REGR: fix order of left sjoin with within predicate #3240

Then, ideally by mid April, we could do 1.0-rc1 that shall contain the enhancements and deprecations and cut final 1.0 by end of April. That should give people enough time to test against 1.0 before it is released and provide feedback if needed. It should also give us enough time to finish what needs to be finished as end of March is unrealistic at this point. Thoughts?

[martinfleis]

The last thing that touches the existing API in some way is #3190. Once we get that in, I propose cutting 1.0-alpha1.

[m-richards]

I still think #1038 would be good to resolve before 1.0 since it will likely involve an API break. Will have a look and see if I can move that forward. Edit: opened #3237 as a way to take that merge request/ discussion forward.

[martinfleis]

I am planning to cut 1.0.0-alpha1 tonight as we're already a bit behind the anticipated schedule.

Next milestone is rc1 with all the new functionality. I suppose it will be pushed to the end of April, with 1.0 coming (optimistically) in early May. We do have a hard deadline 24/5 due to GeoPython starting a few days after that.

[martinfleis]

It is out!

• https://github.com/geopandas/geopandas/releases/tag/v1.0.0-alpha1
• https://pypi.org/project/geopandas/1.0.0a1/

I wrote a short blog post with some highlights - https://martinfleischmann.net/geopandas-1.0-is-coming.-what-will-change/

[m-richards]

@martinfleis thanks for cutting the alpha and writing a summary up!. Just fyi in your blog post, there's a typo in the fiona to pyogrio migration guide link, latert-> latest I think, might be good to fix if you are sharing the link around to promote the alpha to test.

[martinfleis]

@m-richards good catch! The typo was not the only issue, #3206 was not merged :D. I was writing that last week, assuming it will be in.

[martinfleis]

All good now - https://geopandas.org/en/latest/docs/user_guide/fiona_to_pyogrio.html.

[martinfleis]

As an aside, I wonder if it'd be good to consider reorganising the docs pages for these "all geometries together" apis so they sit collectively under one heading, to make it clearer at a glance we've made a distinction about how they api should map to a geoseries, and it's not just verbatim porting of shapely

@m-richards's comment from #3273 - we should look at that once all shapely parity PRs are in.

[jorisvandenbossche]

Shall we try to do an RC early coming week?

[martinfleis]

@jorisvandenbossche can we get remaining new features in? Would be nice, given we need 1.0 ideally by the end of the next week to have it our for GeoPython.

[jorisvandenbossche]

The main remaining features that I am focusing on are the GeoParquet / GeoArrow related ones.

Further, I think all shapely-parity related PRs are in?
Are there other issues/PRs from the milestone (https://github.com/geopandas/geopandas/milestone/4) that we should still give some priority?

Given that we still made a few API changes, would it be helpful to do another pre-release now to at least get some testing of it?
(the geoparquet/geoarrow are generally new and so should in theory not break existing code)

[martinfleis]

I would like to get #3231 as it is ready. Apart from all of the GeoArrow and GeoParquet stuff, I have no strong feelings about the rest and wouldn't mind moving it to 1.1.

[jorisvandenbossche]

I think the main issue that is left (that has impact on API, and is not just an enhancement of bug fix) is #3312

We still have a bunch of other issues labeled with the milestone https://github.com/geopandas/geopandas/issues?q=is%3Aopen+is%3Aissue+milestone%3A1.0, but we will have to bump most of them to 1.1, I think.

I think we should try to do an RC this week. We could do that with keeping the change discussed in #3312 to get more feedback on that, but on the other hand in case we would have consensus of reverting it for GeoDataFrame(..), it would be a bit stupid to have people that test the RC to update their code potentially wrongly. So ideally we would resolve that discussion this week, I think.

[jorisvandenbossche]

I am planning to cut 1.0.0rc0 later today

[jorisvandenbossche]

RC1 is released (https://github.com/geopandas/geopandas/releases/tag/v1.0.0rc1) and on PyPI (https://pypi.org/project/geopandas/1.0.0rc1/)

I am doing a PR to also add it to conda-forge (not super important since it is pure python, but then we can already check the changes to the dependencies) -> conda-forge/geopandas-feedstock#114

[jorisvandenbossche]

And the 1.0.0 release is finally out! 🎉

Release: https://github.com/geopandas/geopandas/releases/tag/v1.0.0
PyPI: https://pypi.org/project/geopandas/1.0.0/
Conda-forge: https://anaconda.org/conda-forge/geopandas/files

Let the issue reports start coming in ;)

[martinfleis]

Thank you @jorisvandenbossche!

[EwoutH]

Congratulations on the 1.0 release!

[martinfleis]

I suggest cutting 1.0.1 with #3363 and #3360 as the regression in buffer can be quite surprising.