[jira] [Commented] (FLINK-931) Remove redundant (old) operator tests

[Shang Yuanchun (Jira)]

    [ https://issues.apache.org/jira/browse/FLINK-931?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14029561#comment-14029561 ]

Fabian Hueske commented on FLINK-931:
-------------------------------------

The old tests are NOT subsumed by the new tests.
The tests under eu.stratosphere.test.operators test all physical execution strategies for each (basic) operator. The tests under eu.stratosphere.test.japiOperators basically test that the transformation before the optimization and the optimizer postpass are working correctly (assembly of program, serializers and comparators). They do not run all physical strategies.

I would leave the tests in place as they are.

> Remove redundant (old) operator tests
> -------------------------------------
>
>                 Key: FLINK-931
>                 URL: https://issues.apache.org/jira/browse/FLINK-931
>             Project: Flink
>          Issue Type: Bug
>            Reporter: Stephan Ewen
>            Assignee: Stephan Ewen
>            Priority: Minor
>
> I would remove all tests in the package "eu.stratosphere.test.operators".
> There are many tests in "eu.stratosphere.test.javaApiOperators" that cover the same operator code paths.
> The redundant tests take a lot of time without really adding test coverage.

--
This message was sent by Atlassian JIRA
(v6.2#6252)

[Stephan Ewen]

Okay.

But it seems that executing runtime code in the
eu.stratosphere.test.japiOperators
tests is actually not what the tests mean to do. Would they cover their
intention better if they just invoked translation / optimization and tested
the result of the optimized plan in a fine granular fashion?

[Robert Metzger]

Hey Stephan, comments from the mailing list are not mirrored back to JIRA.
(And I think your intention was to answer to Fabian's comment)


On Thu, Jun 12, 2014 at 11:59 PM, Stephan Ewen <[hidden email]> wrote:

> Okay.
>
> But it seems that executing runtime code in the
> eu.stratosphere.test.japiOperators
> tests is actually not what the tests mean to do. Would they cover their
> intention better if they just invoked translation / optimization and tested
> the result of the optimized plan in a fine granular fashion?
>

[Ufuk Celebi]

On 13 Jun 2014, at 09:50, Robert Metzger <[hidden email]> wrote:

> Hey Stephan, comments from the mailing list are not mirrored back to JIRA.
> (And I think your intention was to answer to Fabian's comment)

I think this will take some time to get used to as a lot of people did this often with GH. Does anyone know if JIRA supports this? Can we ask INFRA?