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 2025/03/18 10:14:33 UTC
Re: [DISCUSS] Event listener interface
Thanks Andrew for the effort. I'm very much in favor of having the
ability to receive events.
From the doc and from the implementation it's hard to reason about the
actual use cases / problems that the approach is going to tackle.
Can you please enhance the document and describe the use cases and what
information those need?
Technically speaking (from the PR) it's nice to know that e.g. a table
has been changed, but I think that the information is incomplete,
because important attributes like realm, catalog, table identifier and
actor are missing.
I'm also curious how the approach is intended to be extensible and
backwards compatible when adding new features (generic tables, policies,
etc).
Since we're talking about event payload here, I think it's necessary to
give users a direction / proposal about which attributes must/may be
serialized in which way to be "future proof".
Robert
On 27.02.25 00:09, Andrew Guterman wrote:
> Just opened a PR https://github.com/apache/polaris/pull/922
>
> Best,
> Andrew
>
> On Thu, Jan 30, 2025 at 12:24 PM Andrew Guterman <an...@gmail.com>
> wrote:
>
>> Hi all,
>>
>> Apologies for the delay.
>>
>> Here's my doc
>> <https://docs.google.com/document/d/1sJiFKeMlPVlqRUj8Rv4YufMMrfCq_3ZFtDuNv_8eOQ0/edit?usp=sharing>
>> for this along with 2 prototypes: one using Jakarta Events and the other
>> using a custom interface:
>>
>> Please leave comments directly on the doc or reply to this thread.
>>
>> Best,
>> Andrew
>>
>> On Fri, Jan 24, 2025 at 9:59 AM Andrew Guterman <
>> andrew.guterman1@gmail.com> wrote:
>>
>>> I've got a draft in progress and am hoping to share it next week to
>>> collect feedback.
>>>
>>> Best,
>>> Andrew
>>>
>>> On Tue, Jan 21, 2025 at 6:01 PM Yufei Gu <fl...@gmail.com> wrote:
>>>
>>>> Quakus runtime(PR469) has been merged. Do we plan to proceed with
>>>> this proposal? Is there any design doc we can refer to?
>>>>
>>>> Yufei
>>>>
>>>>
>>>> On Tue, Jan 7, 2025 at 3:10 PM Andrew Guterman <
>>>> andrew.guterman1@gmail.com>
>>>> wrote:
>>>>
>>>>> Happy new year everyone.
>>>>>
>>>>> I'm still interested in this but waiting for the migration to Quarkus
>>>> to be
>>>>> fully done before proposing anything concrete, as I'd like to see how
>>>> much
>>>>> of this will already be possible.
>>>>>
>>>>> Best,
>>>>> Andrew
>>>>>
>>>>> On Tue, Dec 17, 2024 at 12:23 AM Robert Stupp <sn...@snazy.de> wrote:
>>>>>
>>>>>> On 14.12.24 11:11, Alex Dutra wrote:
>>>>>>> ServiceLoader
>>>>>>> works, but all the implementations you intend to use must be
>>>> present at
>>>>>>> build time.
>>>>>> Correct (my previous email was wrong wrt to class loading).
>>>>>>
>>>>>>
>>>>>> --
>>>>>> Robert Stupp
>>>>>> @snazy
>>>>>>
>>>>>>
--
Robert Stupp
@snazy
Re: [DISCUSS] Event listener interface
Posted by Andrew Guterman <an...@gmail.com>.
Thanks for the review. I've responded to the specific comments on the PR.
By definition events enable a large degree of user-specific customization
so not all use cases are known in advance. It's important to view them as
iterative and the initial PR doesn't need to try to cover everything.
Re: future proofing, I don't think event listeners are a special case. We
should always consider how our changes affect existing users. There's a
balance, as users pulling a new version of Polaris can't expect nothing to
have changed.
Best,
Andrew
On Tue, Mar 18, 2025 at 3:47 AM Robert Stupp <sn...@snazy.de> wrote:
> Thanks Andrew for the effort. I'm very much in favor of having the
> ability to receive events.
>
> From the doc and from the implementation it's hard to reason about the
> actual use cases / problems that the approach is going to tackle.
>
> Can you please enhance the document and describe the use cases and what
> information those need?
>
> Technically speaking (from the PR) it's nice to know that e.g. a table
> has been changed, but I think that the information is incomplete,
> because important attributes like realm, catalog, table identifier and
> actor are missing.
>
> I'm also curious how the approach is intended to be extensible and
> backwards compatible when adding new features (generic tables, policies,
> etc).
>
> Since we're talking about event payload here, I think it's necessary to
> give users a direction / proposal about which attributes must/may be
> serialized in which way to be "future proof".
>
> Robert
>
>
> On 27.02.25 00:09, Andrew Guterman wrote:
> > Just opened a PR https://github.com/apache/polaris/pull/922
> >
> > Best,
> > Andrew
> >
> > On Thu, Jan 30, 2025 at 12:24 PM Andrew Guterman <
> andrew.guterman1@gmail.com>
> > wrote:
> >
> >> Hi all,
> >>
> >> Apologies for the delay.
> >>
> >> Here's my doc
> >> <
> https://docs.google.com/document/d/1sJiFKeMlPVlqRUj8Rv4YufMMrfCq_3ZFtDuNv_8eOQ0/edit?usp=sharing
> >
> >> for this along with 2 prototypes: one using Jakarta Events and the other
> >> using a custom interface:
> >>
> >> Please leave comments directly on the doc or reply to this thread.
> >>
> >> Best,
> >> Andrew
> >>
> >> On Fri, Jan 24, 2025 at 9:59 AM Andrew Guterman <
> >> andrew.guterman1@gmail.com> wrote:
> >>
> >>> I've got a draft in progress and am hoping to share it next week to
> >>> collect feedback.
> >>>
> >>> Best,
> >>> Andrew
> >>>
> >>> On Tue, Jan 21, 2025 at 6:01 PM Yufei Gu <fl...@gmail.com> wrote:
> >>>
> >>>> Quakus runtime(PR469) has been merged. Do we plan to proceed with
> >>>> this proposal? Is there any design doc we can refer to?
> >>>>
> >>>> Yufei
> >>>>
> >>>>
> >>>> On Tue, Jan 7, 2025 at 3:10 PM Andrew Guterman <
> >>>> andrew.guterman1@gmail.com>
> >>>> wrote:
> >>>>
> >>>>> Happy new year everyone.
> >>>>>
> >>>>> I'm still interested in this but waiting for the migration to Quarkus
> >>>> to be
> >>>>> fully done before proposing anything concrete, as I'd like to see how
> >>>> much
> >>>>> of this will already be possible.
> >>>>>
> >>>>> Best,
> >>>>> Andrew
> >>>>>
> >>>>> On Tue, Dec 17, 2024 at 12:23 AM Robert Stupp <sn...@snazy.de>
> wrote:
> >>>>>
> >>>>>> On 14.12.24 11:11, Alex Dutra wrote:
> >>>>>>> ServiceLoader
> >>>>>>> works, but all the implementations you intend to use must be
> >>>> present at
> >>>>>>> build time.
> >>>>>> Correct (my previous email was wrong wrt to class loading).
> >>>>>>
> >>>>>>
> >>>>>> --
> >>>>>> Robert Stupp
> >>>>>> @snazy
> >>>>>>
> >>>>>>
> --
> Robert Stupp
> @snazy
>
>