You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@polaris.apache.org by Robert Stupp <sn...@snazy.de> on 2026/01/07 11:24:56 UTC

Re: [DISCUSS] Removal of (Before/After)CommitTableEvent

Hi all,

I am not objecting to the change it, but I think the same refactoring
should happen to corresponding events for views as well to retain
consistency.

Robert

On Wed, Dec 3, 2025 at 10:24 PM Adnan Hemani via dev
<de...@polaris.apache.org> wrote:
>
> Hi folks,
>
> I've cut #3195 <https://github.com/apache/polaris/pull/3195> on the basis
> of this discussion. PTAL!
>
> Best,
> Adnan Hemani
>
> On Wed, Nov 26, 2025 at 12:33 PM Adnan Hemani <ad...@snowflake.com>
> wrote:
>
> > Hi all,
> >
> > Yes Oleg, that's a good point out - at that time I hadn't questioned the
> > role that the *CommitTableEvents play while looking into the transaction
> > bugs that had been reported. This recommendation is 100% influenced by my
> > findings on that side of things :)
> >
> > > Are there any that are not covered by existing events?
> > The CommitTransaction API is a bit odd in this manner - today it only
> > generates the (Before/After)CommitTransaction events. So while a user may
> > have updated one/multiple tables(s) as part of the transaction, there is
> > not a specific *UpdateTableEvent that is generated. To maintain parity and
> > specificity while removing the *CommitTableEvents, I will work on
> > generating the necessary *UpdateTableEvent for changes made during a
> > transaction.
> >
> > Best,
> > Adnan Hemani
> >
> > On Tue, Nov 25, 2025 at 3:58 PM Yufei Gu <fl...@gmail.com> wrote:
> >
> >> +1 on the removal. Here are the reasons: they are not used anywhere AFAIK,
> >> and there are replacements(*UpdateTableEvent, *RegisterTableEvent) with
> >> more consistent names.
> >>
> >> Yufei
> >>
> >>
> >> On Tue, Nov 25, 2025 at 6:11 AM Oleg Soloviov <ol...@gmail.com> wrote:
> >>
> >> > Hi Adnan,
> >> >
> >> > I was planning to use them since you mentioned you were not going to
> >> delete
> >> > it on Polaris slack channel :)
> >> > But I do not mind removing them, if it helps to keep Events API more
> >> clear
> >> > and concise, especially in the context of multiple table commit
> >> scenarios.
> >> >
> >> > > Most of the APIs that the commit events were signaling now have their
> >> own
> >> > instrumented events (such as `UpdateTable`)
> >> > Are there any that are not covered by existing events?
> >> >
> >> > Thanks,
> >> > Oleg
> >> >
> >> > On Mon, Nov 24, 2025 at 9:57 PM Adnan Hemani via dev <
> >> > dev@polaris.apache.org>
> >> > wrote:
> >> >
> >> > > Hi all,
> >> > >
> >> > > I wanted to gauge the community's thoughts on removing
> >> > > (Before/After)CommitTableEvent altogether. Most of the APIs that the
> >> > commit
> >> > > events were signaling now have their own instrumented events (such as
> >> > > `UpdateTable`). Does anyone still have a use of the commit events on
> >> > their
> >> > > own then?
> >> > >
> >> > > Best,
> >> > > Adnan Hemani
> >> > >
> >> >
> >>
> >

Re: [DISCUSS] Removal of (Before/After)CommitTableEvent

Posted by Adnan Hemani via dev <de...@polaris.apache.org>.
Hi Robert,

I did not see your message on the PR. It is a fair comment. Can you please
cut an issue so someone can pick this up? If I get some time, I may do so
but cannot commit to it right now.

Best,
Adnan Hemani

On Wed, Jan 7, 2026 at 3:25 AM Robert Stupp <sn...@snazy.de> wrote:

> Hi all,
>
> I am not objecting to the change it, but I think the same refactoring
> should happen to corresponding events for views as well to retain
> consistency.
>
> Robert
>
> On Wed, Dec 3, 2025 at 10:24 PM Adnan Hemani via dev
> <de...@polaris.apache.org> wrote:
> >
> > Hi folks,
> >
> > I've cut #3195 <https://github.com/apache/polaris/pull/3195> on the
> basis
> > of this discussion. PTAL!
> >
> > Best,
> > Adnan Hemani
> >
> > On Wed, Nov 26, 2025 at 12:33 PM Adnan Hemani <
> adnan.hemani@snowflake.com>
> > wrote:
> >
> > > Hi all,
> > >
> > > Yes Oleg, that's a good point out - at that time I hadn't questioned
> the
> > > role that the *CommitTableEvents play while looking into the
> transaction
> > > bugs that had been reported. This recommendation is 100% influenced by
> my
> > > findings on that side of things :)
> > >
> > > > Are there any that are not covered by existing events?
> > > The CommitTransaction API is a bit odd in this manner - today it only
> > > generates the (Before/After)CommitTransaction events. So while a user
> may
> > > have updated one/multiple tables(s) as part of the transaction, there
> is
> > > not a specific *UpdateTableEvent that is generated. To maintain parity
> and
> > > specificity while removing the *CommitTableEvents, I will work on
> > > generating the necessary *UpdateTableEvent for changes made during a
> > > transaction.
> > >
> > > Best,
> > > Adnan Hemani
> > >
> > > On Tue, Nov 25, 2025 at 3:58 PM Yufei Gu <fl...@gmail.com> wrote:
> > >
> > >> +1 on the removal. Here are the reasons: they are not used anywhere
> AFAIK,
> > >> and there are replacements(*UpdateTableEvent, *RegisterTableEvent)
> with
> > >> more consistent names.
> > >>
> > >> Yufei
> > >>
> > >>
> > >> On Tue, Nov 25, 2025 at 6:11 AM Oleg Soloviov <ol...@gmail.com>
> wrote:
> > >>
> > >> > Hi Adnan,
> > >> >
> > >> > I was planning to use them since you mentioned you were not going to
> > >> delete
> > >> > it on Polaris slack channel :)
> > >> > But I do not mind removing them, if it helps to keep Events API more
> > >> clear
> > >> > and concise, especially in the context of multiple table commit
> > >> scenarios.
> > >> >
> > >> > > Most of the APIs that the commit events were signaling now have
> their
> > >> own
> > >> > instrumented events (such as `UpdateTable`)
> > >> > Are there any that are not covered by existing events?
> > >> >
> > >> > Thanks,
> > >> > Oleg
> > >> >
> > >> > On Mon, Nov 24, 2025 at 9:57 PM Adnan Hemani via dev <
> > >> > dev@polaris.apache.org>
> > >> > wrote:
> > >> >
> > >> > > Hi all,
> > >> > >
> > >> > > I wanted to gauge the community's thoughts on removing
> > >> > > (Before/After)CommitTableEvent altogether. Most of the APIs that
> the
> > >> > commit
> > >> > > events were signaling now have their own instrumented events
> (such as
> > >> > > `UpdateTable`). Does anyone still have a use of the commit events
> on
> > >> > their
> > >> > > own then?
> > >> > >
> > >> > > Best,
> > >> > > Adnan Hemani
> > >> > >
> > >> >
> > >>
> > >
>