You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@tinkerpop.apache.org by Marko Rodriguez <ok...@gmail.com> on 2016/02/26 16:35:31 UTC
[DICUSS] Making SparkGraphComputerProcessIntegrateTest not Integration
Hi,
Bob Briody is doing some work on profile()-step and said:
"SparkGraphComputerProcessIntegrateTest only takes about 30 seconds to execute. Why is that integration testing?"
I thought about it, and I agree. We already have various Spark tests that are not integration-based. Would people object if I made both:
SparkGraphComputerProcessIntegrateTest
SparkGraphComputerGroovyProcessIntegrateTest
---->
SparkGraphComputerProcessTest
SparkGraphComputerGroovyProcessTest
This will add about 1 minute to the "mvn clean install" but will be a lot safer for people as lots of Serialization/VertexProgram Configuration/etc. issues get exposed here that you don't see with TinkerGraphComputerProcessTest. Definitely beneficial to people trying to submit PRs.
Thoughts?,
Marko.
http://markorodriguez.com
Re: [DICUSS] Making SparkGraphComputerProcessIntegrateTest not Integration
Posted by Stephen Mallette <sp...@gmail.com>.
oops - hit send too soon - also - the spark integration tests dont' require
external dependencies (e.g. hdfs) in the background either so that would
further disqualify them for being integration tests.
On Fri, Feb 26, 2016 at 10:38 AM, Stephen Mallette <sp...@gmail.com>
wrote:
> +1 - integration tests typically mean "long" in our definition and one
> extra minute isn't really a hurt for local builds or CI
>
> On Fri, Feb 26, 2016 at 10:35 AM, Marko Rodriguez <ok...@gmail.com>
> wrote:
>
>> Hi,
>>
>> Bob Briody is doing some work on profile()-step and said:
>>
>> "SparkGraphComputerProcessIntegrateTest only takes about 30
>> seconds to execute. Why is that integration testing?"
>>
>> I thought about it, and I agree. We already have various Spark tests that
>> are not integration-based. Would people object if I made both:
>>
>> SparkGraphComputerProcessIntegrateTest
>> SparkGraphComputerGroovyProcessIntegrateTest
>>
>> ---->
>>
>> SparkGraphComputerProcessTest
>> SparkGraphComputerGroovyProcessTest
>>
>> This will add about 1 minute to the "mvn clean install" but will be a lot
>> safer for people as lots of Serialization/VertexProgram Configuration/etc.
>> issues get exposed here that you don't see with
>> TinkerGraphComputerProcessTest. Definitely beneficial to people trying to
>> submit PRs.
>>
>> Thoughts?,
>> Marko.
>>
>> http://markorodriguez.com
>>
>>
>
Re: [DICUSS] Making SparkGraphComputerProcessIntegrateTest not Integration
Posted by Stephen Mallette <sp...@gmail.com>.
+1 - integration tests typically mean "long" in our definition and one
extra minute isn't really a hurt for local builds or CI
On Fri, Feb 26, 2016 at 10:35 AM, Marko Rodriguez <ok...@gmail.com>
wrote:
> Hi,
>
> Bob Briody is doing some work on profile()-step and said:
>
> "SparkGraphComputerProcessIntegrateTest only takes about 30
> seconds to execute. Why is that integration testing?"
>
> I thought about it, and I agree. We already have various Spark tests that
> are not integration-based. Would people object if I made both:
>
> SparkGraphComputerProcessIntegrateTest
> SparkGraphComputerGroovyProcessIntegrateTest
>
> ---->
>
> SparkGraphComputerProcessTest
> SparkGraphComputerGroovyProcessTest
>
> This will add about 1 minute to the "mvn clean install" but will be a lot
> safer for people as lots of Serialization/VertexProgram Configuration/etc.
> issues get exposed here that you don't see with
> TinkerGraphComputerProcessTest. Definitely beneficial to people trying to
> submit PRs.
>
> Thoughts?,
> Marko.
>
> http://markorodriguez.com
>
>
Re: [DICUSS] Making SparkGraphComputerProcessIntegrateTest not Integration
Posted by Marko Rodriguez <ok...@gmail.com>.
Hi,
I just did this. Its in master/ and will come out with 3.2.0.
[INFO] ------------------------------------------------------------------------
[INFO] Reactor Summary:
[INFO]
[INFO] Apache TinkerPop .................................. SUCCESS [4.296s]
[INFO] Apache TinkerPop :: Gremlin Shaded ................ SUCCESS [2.217s]
[INFO] Apache TinkerPop :: Gremlin Core .................. SUCCESS [37.355s]
[INFO] Apache TinkerPop :: Gremlin Test .................. SUCCESS [12.544s]
[INFO] Apache TinkerPop :: Gremlin Groovy ................ SUCCESS [39.642s]
[INFO] Apache TinkerPop :: Gremlin Groovy Test ........... SUCCESS [6.172s]
[INFO] Apache TinkerPop :: TinkerGraph Gremlin ........... SUCCESS [1:50.495s]
[INFO] Apache TinkerPop :: Hadoop Gremlin ................ SUCCESS [5:06.528s]
[INFO] Apache TinkerPop :: Spark Gremlin ................. SUCCESS [3:58.827s]
[INFO] Apache TinkerPop :: Giraph Gremlin ................ SUCCESS [5.803s]
[INFO] Apache TinkerPop :: Neo4j Gremlin ................. SUCCESS [3.014s]
[INFO] Apache TinkerPop :: Gremlin Driver ................ SUCCESS [8.365s]
[INFO] Apache TinkerPop :: Gremlin Server ................ SUCCESS [6.168s]
[INFO] Apache TinkerPop :: Gremlin Console ............... SUCCESS [17.548s]
[INFO] Apache TinkerPop :: Gremlin Archetype ............. SUCCESS [0.070s]
[INFO] Apache TinkerPop :: Archetype - TinkerGraph ....... SUCCESS [4.974s]
[INFO] Apache TinkerPop :: Archetype - Server ............ SUCCESS [11.271s]
[INFO] ------------------------------------------------------------------------
[INFO] BUILD SUCCESS
[INFO] ------------------------------------------------------------------------
[INFO] Total time: 13:35.762s
[INFO] Finished at: Tue Mar 01 10:30:01 MST 2016
[INFO] Final Memory: 140M/804M
[INFO] ------------------------------------------------------------------------
Marko.
http://markorodriguez.com
On Feb 26, 2016, at 9:14 AM, Daniel Kuppitz <me...@gremlin.guru> wrote:
> +1
>
> On Fri, Feb 26, 2016 at 4:35 PM, Marko Rodriguez <ok...@gmail.com>
> wrote:
>
>> Hi,
>>
>> Bob Briody is doing some work on profile()-step and said:
>>
>> "SparkGraphComputerProcessIntegrateTest only takes about 30
>> seconds to execute. Why is that integration testing?"
>>
>> I thought about it, and I agree. We already have various Spark tests that
>> are not integration-based. Would people object if I made both:
>>
>> SparkGraphComputerProcessIntegrateTest
>> SparkGraphComputerGroovyProcessIntegrateTest
>>
>> ---->
>>
>> SparkGraphComputerProcessTest
>> SparkGraphComputerGroovyProcessTest
>>
>> This will add about 1 minute to the "mvn clean install" but will be a lot
>> safer for people as lots of Serialization/VertexProgram Configuration/etc.
>> issues get exposed here that you don't see with
>> TinkerGraphComputerProcessTest. Definitely beneficial to people trying to
>> submit PRs.
>>
>> Thoughts?,
>> Marko.
>>
>> http://markorodriguez.com
>>
>>
Re: [DICUSS] Making SparkGraphComputerProcessIntegrateTest not Integration
Posted by Daniel Kuppitz <me...@gremlin.guru>.
+1
On Fri, Feb 26, 2016 at 4:35 PM, Marko Rodriguez <ok...@gmail.com>
wrote:
> Hi,
>
> Bob Briody is doing some work on profile()-step and said:
>
> "SparkGraphComputerProcessIntegrateTest only takes about 30
> seconds to execute. Why is that integration testing?"
>
> I thought about it, and I agree. We already have various Spark tests that
> are not integration-based. Would people object if I made both:
>
> SparkGraphComputerProcessIntegrateTest
> SparkGraphComputerGroovyProcessIntegrateTest
>
> ---->
>
> SparkGraphComputerProcessTest
> SparkGraphComputerGroovyProcessTest
>
> This will add about 1 minute to the "mvn clean install" but will be a lot
> safer for people as lots of Serialization/VertexProgram Configuration/etc.
> issues get exposed here that you don't see with
> TinkerGraphComputerProcessTest. Definitely beneficial to people trying to
> submit PRs.
>
> Thoughts?,
> Marko.
>
> http://markorodriguez.com
>
>
Re: [DICUSS] Making SparkGraphComputerProcessIntegrateTest not
Integration
Posted by pieter-gmail <pi...@gmail.com>.
+1, will make people like me that are somewhat afraid of the distributed
config pain to look at the code/features.
On 26/02/2016 17:35, Marko Rodriguez wrote:
> Hi,
>
> Bob Briody is doing some work on profile()-step and said:
>
> "SparkGraphComputerProcessIntegrateTest only takes about 30 seconds to execute. Why is that integration testing?"
>
> I thought about it, and I agree. We already have various Spark tests that are not integration-based. Would people object if I made both:
>
> SparkGraphComputerProcessIntegrateTest
> SparkGraphComputerGroovyProcessIntegrateTest
>
> ---->
>
> SparkGraphComputerProcessTest
> SparkGraphComputerGroovyProcessTest
>
> This will add about 1 minute to the "mvn clean install" but will be a lot safer for people as lots of Serialization/VertexProgram Configuration/etc. issues get exposed here that you don't see with TinkerGraphComputerProcessTest. Definitely beneficial to people trying to submit PRs.
>
> Thoughts?,
> Marko.
>
> http://markorodriguez.com
>
>