-
Notifications
You must be signed in to change notification settings - Fork 3k
Revert "Spark 4.0 integration" #13006
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Conversation
|
@pan3793 I think this over, actually, the key to preserve history is what happens in step 1, right? Ideally, I should have 3 commits, but as long as I have 1 and 2, it's OK to have more than 3 commits?
|
|
commit and revert may mess the commit history worse, do you have permission to force push(with dropping the previous commit) the main branch directly? (this might affect a few contributors) otherwise I would suggest just keeping it as-is. |
|
I'd prefer to just keep things as is so we don't complicate history. It's a miss on my part to not catch the breakdown of the move/copy/initial support and miss preserving history there but I think it's ultimately OK to have the single commit that we have now. |
|
Thanks @amogh-jahagirdar @pan3793 . I will close this PR for now |
I know this is abandoned, but for future reference, why do you want to have more than 3 commits? If you have 1 and 2 for sure, then to have more than 3 commits means you want to spread the initial support for the new version over several commits? What benefit would that bring? For a PR itself, the initial support may well occur over many commits due to iterations, but for merging into the repo, we'd want a single commit for 3. |
|
@wypoon right, we should always have 3 commits. I forgot to squash all the remaining commits into one and was trying to figure out whether the commit history was still preserved. |
|
I had an offline discussion with @amogh-jahagirdar. We have decided to revert the integration PR and redo it, so we can keep all the commit history. |
Thanks for the heads up! |
|
Thanks @wypoon for the suggestion! I will try to resolve the conflicts on my side and see how it goes |
This reverts commit ad7f5c4.
21d1459 to
11e1873
Compare
|
@amogh-jahagirdar @pan3793 @wypoon @szehon-ho |
amogh-jahagirdar
left a comment
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks, and it looks like the reason this is not exactly the inverse of the original is because the revert also includes the 4.0 changes that were made in 809a232#diff-ba3f1e91e786949eb31ca8e396d7101ce78adc569350c0a1a0deb76ab9aa3074 right @huaxingao ? I'll leave it up for a bit before merging so that others can also take a look
|
I see that in #12494, there are 584 changed files, while in this PR, there are 580. Are the 4 files that are not reverted the build.gradle and CometColumnReader.java in spark/v3.4 and spark/v3.5? and that is because you just want to keep the Comet version at 0.8.1? (I checked gradle/libs.versions.toml in this PR and I see that the Comet version is not reverted.) |
|
@huaxingao Please update the PR description to summarize all the changes. |
Updated. Thanks! |
|
Thank you all very much! @amogh-jahagirdar @wypoon @pan3793 @manuzhang |

In order to keep all the git history on Spark4.0, we need to follow the proposal in this discussion.
I will reverts #12494 and redo the 4.0 integration PR. I will have three commits:
This PR reverted everything of #12494 except the Comet version bump. The 4.0 changes that were made in 809a232#diff-ba3f1e91e786949eb31ca8e396d7101ce78adc569350c0a1a0deb76ab9aa3074 were manually remove.