You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Is your feature request related to a problem or challenge?
During testing of Version 21.0.0 RC1 we discovered that it inadvertently depends both on arrow (and its subcrates) version 34.0.0 as well as arrow-array version 35.0.0
This feature was fixed in #5724 (so datafusion only depends on arrow 34.0.0)
Thanks @alamb. There is a branch-21 already that we can merge the fix to, and release from there. Note that the release process for DataFusion differs slightly from the Arrow-rs release process (specifically with tags and branches), but hopefully this is documented well enough in dev/release/README, LMK if you have questions.
Ah, ok. I was seeing the mixed arrow 34/35 dependencies with datafusion 21.0.0 and assumed that it was intended to depend fully on arrow 35. Thanks for the clarification.
Ah, ok. I was seeing the mixed arrow 34/35 dependencies with datafusion 21.0.0 and assumed that it was intended to depend fully on arrow 35. Thanks for the clarification.
No problem -- we decided to skip arrow-35 due to several regressions in parsing timestamps (see #5441 (comment) for more details)
Edit: I also updated the PR description to make the ('only depend on arrow 34.0.0') more explicit
Activity
andygrove commentedon Mar 26, 2023
Thanks @alamb. There is a
branch-21already that we can merge the fix to, and release from there. Note that the release process for DataFusion differs slightly from the Arrow-rs release process (specifically with tags and branches), but hopefully this is documented well enough in dev/release/README, LMK if you have questions.alamb commentedon Mar 28, 2023
I have prepared PRs to backport the relevant parts to branch-21:
#5766
#5765
alamb commentedon Mar 29, 2023
I have created a release candidate here: https://lists.apache.org/thread/qw5067glfmxy4k42wty4vph7pjkgq24p
alamb commentedon Mar 30, 2023
I am now waiting for the 3 day voting period. Once that has elapsed, assuming no
-1votes, I will release this version to crates.ioalamb commentedon Apr 2, 2023
The release is approved and available: https://lists.apache.org/thread/tmfzkqnvkmccmy8w2ltpqh0sy4mgy6qq
However, unfortunately I am still working on publishing it to crates.io ((I do not have permissions to https://crates.io/crates/datafusion-execution/21.0.0)
Perhaps @andygrove can help here
alamb commentedon Apr 3, 2023
I have published the release to crates.io (thank you @andygrove )
Since this release only includes a dependency improvement (rather than any features) I am not planning to publish an update on homebrew.
alamb commentedon Apr 3, 2023
I have published the release to crates.io (thank you @andygrove )
Since this release only includes a dependency improvement (rather than any features) I am not planning to publish an update on homebrew.
jonmmease commentedon Apr 3, 2023
Thanks @alamb. I'm unforunately still seeing both Arrow 35 and arrow 34 pulled in. I may be missing something, but here's what I'm trying.
Fresh Cargo.toml
If I run
cargo tree, this is what I'm seeing:In particular:
Happy to open a new issue if that's easier than discussing here
alamb commentedon Apr 3, 2023
@jonmmease 🤔 I wonder if you can try runnning
cargo update-- does that fix things?alamb commentedon Apr 3, 2023
(note that datafusion-21 still depends on arrow-rs 34) -- only datafusion 22.0.0 upgrades to arrow-rs 36.
So in other words this is expected in my mind:
jonmmease commentedon Apr 3, 2023
Ah, ok. I was seeing the mixed arrow 34/35 dependencies with datafusion 21.0.0 and assumed that it was intended to depend fully on arrow 35. Thanks for the clarification.
alamb commentedon Apr 3, 2023
No problem -- we decided to skip arrow-35 due to several regressions in parsing timestamps (see #5441 (comment) for more details)
Edit: I also updated the PR description to make the ('only depend on arrow 34.0.0') more explicit