search

geopandas / pyogrio

Vectorized vector I/O using OGR
https://pyogrio.readthedocs.io
MIT License
257 stars 22 forks source link

Release version 0.8.0 #410

Closed jorisvandenbossche closed 1 month ago

jorisvandenbossche commented 1 month ago

Let's release?

I think most items we wanted to finish for 0.8.0 have been merged now. The wheel builds are also all passing on main (including CircleCI) and have been updated for the latest GDAL bugfix release.

brendan-ward commented 1 month ago

Would be nice to get #411 , but that is oriented at tests and can happen whenever - it just needed to happen at a time when we didn't have a lot of outstanding PRs where it would cause git conflicts.

Otherwise agreed, finally ready for 0.8!

martinfleis commented 1 month ago

I suppose we should then pin geopandas to 0.8, right? Given the recent PRs.

brendan-ward commented 1 month ago

Released! https://github.com/geopandas/pyogrio/releases/tag/v0.8.0

Wheels on PyPI.

jorisvandenbossche commented 1 month ago

Thank @brendan-ward!

I suppose we should then pin geopandas to 0.8, right? Given the recent PRs.

Not necessarily? Yes you need the latest version if you want to write to BytesIO, but I don't know if that is enough reason to force everyone to have the latest version? (eg for conda-forge, pyogrio is only built with the latest GDAL, so if you have a bigger environment that for some reason still needs a previous build of gdal, then that might be a bit annoying?)

martinfleis commented 1 month ago

@brendan-ward I got a notification from PyPI that an API token was used to upload files to the project, even though the project has a Trusted Publisher configured. They say

This may have been in error. We recommend removing the API token and only using the Trusted Publisher to publish.

Was the setup of Trusted Publisher finished on GHA side?

brendan-ward commented 1 month ago

Sorry, I should have logged a note about that since this was the first time we've gotten notifications. We have a Trusted Publisher setup for GitHub, but that is not available for CircleCI (where we build linux Arm64 wheels), so our current deploy workflow involved downloading those wheels from CircleCI and uploading manually (using personal accounts) to PyPI.