Great work!
> On Jan 19, 2018, at 05:43, Arina Ielchiieva <[email protected]> wrote:
>
> I am happy to announce that Apache Drill has rebased on Calcite 1.15. All
> information can be found in DRILL-3993 and in Google Doc [1].
> Currently changes are in master and are intended to be part of Drill 1.13.0
> release unless something critical pops up. Feel free to test them and give
> your feedback.
>
> Thanks Roman Kulyk and Volodymyr Vysotskyi for making this happen.
> Aman Sinha and Chunhui Shi for the code review.
> And of course Julian Hyde for helping to address / review Calcite specific
> issues.
>
> [1]
> https://docs.google.com/document/d/1Lqk9NoKQviz0YimBmov4z1pui7QjJGjDVwMa1p0emPk/edit#heading=h.i3rowg20vxv4
>
>
> Kind regards
> Arina
>
> On Fri, Sep 1, 2017 at 8:01 PM, Roman Kulyk <[email protected]> wrote:
>
>> Team, thanks for your answers!
>>
>> Julian, сontinuing the topic of Calcite tests I just want to note that I
>> have ignored 7 test in 5ba68240 commit in Calcite. Here is a list of
>> ignored tests:
>>
>> TraitConversionTest.testTraitConversion
>> JdbcAdapterTest.testCrossJoinWithJoinKeyPlan
>> JdbcAdapterTest.testCrossJoinWithJoinKeyAndFilterPlan
>> JdbcTest.testJoinManyWay (it was also ignored by 8a87ad1e5)
>> LatticeTest.testTileAlgorithm (it was also ignored by 2e076e1a)
>> LatticeTest.testTileAlgorithm2 (it was also ignored by 8a87ad1e5)
>> LatticeTest.testTwoLattices
>>
>> You can see detailed information in my comments. Also, I want to notice
>> that there are 2 required commits in my Calcite branch: 5ba68240 and
>> 1f5145b1. Without them, we lose a big part of Drill functionality.
>>
>> Best regards,
>> Roman Kulyk
>>
>> 2017-08-31 18:43 GMT+00:00 Julian Hyde <[email protected]>:
>>
>>> +dev@calcite.
>>>
>>> Excellent work, Roman.
>>>
>>> Your changes to Calcite don’t seem to be very major, and I ran the test
>>> suite on your https://github.com/KulykRoman/incubator-calcite/commits/
>>> DrillCalcite1.13.0_rc1 branch and all the tests pass. So when you are
>>> ready, I think we can expedite getting these changes back into Calcite.
>>>
>>> Calcite will release 1.14 probably in the next two weeks. So, when you’re
>>> on to 1.13 let’s continue the momentum and get up to the latest release.
>> I
>>> would love to get to the situation that we can ask Drill to verify a
>>> Calcite release before we make it. There are possibly breaking changes in
>>> the area of schema and grouping-sets aggregate but it should be mostly
>>> straightforward.
>>>
>>> Julian
>>>
>>>
>>>> On Aug 31, 2017, at 8:12 AM, Jinfeng Ni <[email protected]> wrote:
>>>>
>>>> It's great to hear that someone is working on moving Drill on new
>> Calcite
>>>> release! As someone who did similar job in last two rebase work, I
>> could
>>>> see that there would be many regressions to fix. Please let me know if
>> I
>>>> can provide any help.
>>>>
>>>> I thought we have pushed almost Drill specific commits to Calcite. I'll
>>>> take a look at your Calcite branch. Ideally, we want to get rid of all
>>>> Drill specific commits, after the rebase work this time.
>>>>
>>>> Thanks,
>>>>
>>>> Jinfeng
>>>>
>>>>
>>>>
>>>>
>>>> On Thu, Aug 31, 2017 at 8:04 AM, Aman Sinha <[email protected]>
>>> wrote:
>>>>
>>>>> Yes, this is long overdue ! Thanks for working on it Roman. If
>>> needed,
>>>>> we can do a separate hangout with a few selected folks who have worked
>>> on
>>>>> Drill+Calcite to provide feedback on how to resolve the test failures.
>>>>> Although, at this stage you probably know a lot more about the
>>> integration
>>>>> than many of us.
>>>>>
>>>>> -Aman
>>>>>
>>>>> On Thu, Aug 31, 2017 at 7:33 AM, Roman Kulyk <[email protected]>
>>> wrote:
>>>>>
>>>>>> Hi, team!
>>>>>>
>>>>>> I want to inform that I am working on updating Drill Calcite version
>> to
>>>>>> 1.13. I forked from Calcite "branch-1.13" and added some commits:
>>>>>> - DRILL-1455: Add return type-inference strategy for arithmetic
>>> operators
>>>>>> when one of the arguments is ANY type. (was not in Calcite)
>>>>>> - Add new method to ViewExpander interface to allow passing
>> SchemaRoot.
>>>>>> (was not in Calcite)
>>>>>> - Allow a MAP literal type. (was not in Calcite)
>>>>>> - DRILL-4047: Modify table functions to allow querying a table with
>>>>> options
>>>>>> in Drill (was not in Calcite)
>>>>>> - Drill-specific change: Add back AbstractConverter in RelSet.java
>> ...
>>>>>> (small changes after a3bc0d8 commit)
>>>>>> - CALCITE-628 related but not fix the problem: Ensure target traits
>> are
>>>>>> simple when use Frameworks or RelOptRule.convert() method. (small
>>> changes
>>>>>> after a3bc0d8 commit)
>>>>>> - Support select * from schema-less table in execution engine like
>>> Drill
>>>>>> (small changes after 9bd7d75 commit)
>>>>>>
>>>>>> Only after adding these commits to Calcite, a part of Drill
>>>>> functionality
>>>>>> started to work as before. Currently, I am at Drill integration
>> stage.
>>> I
>>>>> am
>>>>>> fixing Drill unit tests. Progress can be tracked in my branches.
>>>>>>
>>>>>> Drill: https://github.com/KulykRoman/drill/commits/
>>> CalciteForkRebase_rc1
>>>>>> Calcite:
>>>>>> https://github.com/KulykRoman/incubator-calcite/commits/
>>>>>> DrillCalcite1.13.0_rc1
>>>>>>
>>>>>> Currently, I face ~160 errors in java-exec module unit tests. I have
>>>>>> divided them into seven groups:
>>>>>> 1) SYSTEM ERROR: ClassCastException
>>>>>> 2) Unexpected column errors
>>>>>> 3) Runtime Exceptions
>>>>>> 4) validation errors
>>>>>> 5) RpcExceptions (generalized group: some of them should be fixed by
>>> the
>>>>>> previous groups)
>>>>>> 6) IllegalState errors (generalized group: some of them should be
>> fixed
>>>>> by
>>>>>> the previous group)
>>>>>> 7) Other errors
>>>>>>
>>>>>> Best regards,
>>>>>> Roman Kulyk
>>>>>>
>>>>>
>>>
>>>
>>
