[
https://issues.apache.org/jira/browse/PIG-5228?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16000233#comment-16000233
]
Rohini Palaniswamy commented on PIG-5228:
-----------------------------------------
bq. I think it may be dependent on HashMap implementation and thus JDK version
as well.
If running on different JDK version, it could be an issue as HashMap
implementation changed between 1.6 and 1.7. For the same jdk version behavior
of insertion is usually consistent.
bq. My wild guess is that either MR or Spark makes an extra filling of a map
somewhere under the hood and that's where the difference comes from.
Ordering of entries in map is not something we guarantee. But can you still
try to find why it is happening? I am surprised you are running into this with
just simple load and store statements for same jdk version. Could be something
to do with serialization as well.
bq. so I've created a new a test case where we project by each key from the map.
If figuring out the cause of change is taking more time, I am fine with the
current patch which has separate test for Spark.
> Orc_2 is failing with spark exec type
> -------------------------------------
>
> Key: PIG-5228
> URL: https://issues.apache.org/jira/browse/PIG-5228
> Project: Pig
> Issue Type: Sub-task
> Components: spark
> Reporter: Adam Szita
> Assignee: Adam Szita
> Fix For: spark-branch
>
> Attachments: PIG-5228.0.patch
>
>
> This test is failing due to mismatch in the actual and expected result. The
> difference is only related to the order of entries in Pig maps such as:
> Actual:
> {code}
> [name#alice, age#18]...
> {code}
> Expected:
> {code}
> [age#18, name#alice]...
> {code}
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
