[GitHub] flink issue #3021: [FLINK-3617] Prevent NPE in CaseClassSerializer and infor...

2017-01-23 Thread StephanEwen
Github user StephanEwen commented on the issue:

https://github.com/apache/flink/pull/3021
  
This has been merged a while back in 
fdce1f319c512fc845b64cbb7cbfb10f9d899021
I forgot to add the keywork that the bot closes the pull request.

Can you manually close this pull request?


---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---


[GitHub] flink issue #3021: [FLINK-3617] Prevent NPE in CaseClassSerializer and infor...

2017-01-13 Thread StephanEwen
Github user StephanEwen commented on the issue:

https://github.com/apache/flink/pull/3021
  
Looks good, thanks, will merge this...



---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---


[GitHub] flink issue #3021: [FLINK-3617] Prevent NPE in CaseClassSerializer and infor...

2017-01-10 Thread chermenin
Github user chermenin commented on the issue:

https://github.com/apache/flink/pull/3021
  
@StephanEwen What do you think about merging this PR?


---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---


[GitHub] flink issue #3021: [FLINK-3617] Prevent NPE in CaseClassSerializer and infor...

2016-12-21 Thread chermenin
Github user chermenin commented on the issue:

https://github.com/apache/flink/pull/3021
  
I've got results and here are:

Benchmark  | Mode  | Cnt  |Score  |  Error  | Units
---|---|--|---|-|
Just operation  | avgt  | 200  |   74.672  |  0.944  | ns/op
if-else (_not_ null value)  | avgt  | 200  |  915.718  |  5.534  | ns/op
try-catch (_not_ null value)| avgt  | 200  |   74.183  |  0.405  | ns/op
if-else (null value) | avgt  | 200  |  901.775  |  5.368  | ns/op
try-catch (null value)   | avgt  | 200  | 8649.317  | 40.462  | ns/op

In this way `if-else-check` always adds more than 800 nanoseconds per 
operation to run-time.
While `try-catch-block` doesn't slow down normal performance. We really 
need this option, because null value isn't normal value for us now. I will 
change the code in the closest future.


---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---


[GitHub] flink issue #3021: [FLINK-3617] Prevent NPE in CaseClassSerializer and infor...

2016-12-20 Thread chermenin
Github user chermenin commented on the issue:

https://github.com/apache/flink/pull/3021
  
Yep, i'll do a benchmark to test approaches and compare them each other at 
this week. I will write the results here. Big thanks for your idea)


---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---


[GitHub] flink issue #3021: [FLINK-3617] Prevent NPE in CaseClassSerializer and infor...

2016-12-20 Thread StephanEwen
Github user StephanEwen commented on the issue:

https://github.com/apache/flink/pull/3021
  
Looks like a good solution.
Since these serializers are very performance critical, I am curious what is 
more efficient:
  - The if-null-check you implemented here
  - a try/catch block that catches the nullpointer exception and throws the 
NullFieldException

Are you interested in microbenchmarking that?


---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---