Re: [C++] Sparse Unions and CICD tests

[Wes McKinney]

Regarding our CI: these builds should be consistently green (which
they are, only the Python 3.5 CI entry is failing for known reasons,
see e.g. 
https://github.com/apache/arrow/commit/64f9b3fbe9ef4c718449a735435b53ab992ca852).
We have a couple of flaky tests, but if you are seeing other failures
that aren't occuring in CI builds on master, then the issue is likely
your patch and so that does need to be fixed before merging.

On Sun, Nov 29, 2020 at 8:59 PM Ying Zhou  wrote:
>
> Hi,
>
> Really thanks for the help you guys gave me in the past! Tonight I would like 
> to ask two questions.
>
> First of all it seems that in the C++ implementation of sparse unions it is 
> possible to construct a union array of length 8 from two child arrays of 
> length 4 with dense union-like behavior. Is it intended? Can I throw an error 
> in my Arrow2ORC API when a sparse union array has some child array with a 
> different length?
>
> Secondly I would like to know whether all CICD tests need to pass before a 
> pull request can be merged. The C++ part of Arrow2ORC development 
> (https://issues.apache.org/jira/browse/ARROW-3014 
> ) is almost over. I’m doing 
> very thorough testing (BTW I have fixed the long unit test problem I asked 
> some time ago) and will seek the merge fairly soon. Right now it seems that 
> all the C++ and lint tests pass. However some python and glib tests have 
> failed. All the python and cglib code I currently have are from master. Can I 
> still ask for the PR merge if CICD tests unrelated to my C++ code fail?
>
> Thanks,
> Ying

[C++] Sparse Unions and CICD tests

[Ying Zhou]

Hi,

Really thanks for the help you guys gave me in the past! Tonight I would like 
to ask two questions.

First of all it seems that in the C++ implementation of sparse unions it is 
possible to construct a union array of length 8 from two child arrays of length 
4 with dense union-like behavior. Is it intended? Can I throw an error in my 
Arrow2ORC API when a sparse union array has some child array with a different 
length?

Secondly I would like to know whether all CICD tests need to pass before a pull 
request can be merged. The C++ part of Arrow2ORC development 
(https://issues.apache.org/jira/browse/ARROW-3014 
) is almost over. I’m doing 
very thorough testing (BTW I have fixed the long unit test problem I asked some 
time ago) and will seek the merge fairly soon. Right now it seems that all the 
C++ and lint tests pass. However some python and glib tests have failed. All 
the python and cglib code I currently have are from master. Can I still ask for 
the PR merge if CICD tests unrelated to my C++ code fail?

Thanks,
Ying