-
Notifications
You must be signed in to change notification settings - Fork 152
APEXMALHAR-2314 Improper functioning in partitioning for sequentialFileRead for FSRecord #468
Conversation
@yogidevendra: Could you please review ? |
Let us reuse SequentialFileBlockMetadataCodec from FSInputModule instead of defining separate one. |
@yogidevendra : Thanks for suggestion. Makes sense. Only thing that worried me about creating unnecessary dependency of FSInputModule's StreamCodec on FSRecordReaderModule. So if someone changes FSInput StreamCodec he must consider same in FSRecoredReaderModule. |
This dependency should be OK. FSRecordReader is closely aligned with FSInputModule. |
fe81d79
to
75c4f75
Compare
Incorporated Yogi's suggestion. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Could you please shorten the commit message?
Keep details to 1-2 lines per change.
Additional details can be captured in the JIRA description. Git log showing long messages would be hard to read.
Also please fix the spelling mistake in the JIRA title and commit message. |
75c4f75
to
259cc5b
Compare
Done. Thanks ! On Wed, Oct 26, 2016 at 11:15 AM, yogidevendra notifications@github.com
Thanks & Regards Deepak Narkhede |
259cc5b
to
36d18cc
Compare
36d18cc
to
87473a8
Compare
[ERROR] src/main/java/org/apache/apex/malhar/lib/fs/FSRecordReaderModule.java:32,8 UnusedImports: Unused import - com.datatorrent.common.partitioner.StatelessPartitioner. please fix above checkstyle error. |
Sorry my bad. After rebase it occurred. Will resolve the conflicts. On Nov 7, 2016 3:33 PM, "Tushar R. Gosavi" notifications@github.com wrote:
|
…eRead property of FSRecordReaderModule. Modified the StreamCodec to work with hashcode of filepath rather than blockId. Conflicts: library/src/main/java/org/apache/apex/malhar/lib/fs/FSRecordReaderModule.java
87473a8
to
3f97304
Compare
Fix the StreamCodec for FSRecordReader, initially it was hashcode of blockId's mostly always unique.
Hence unable to satisfy the sequentialFileRead property. Now the StreamCodec is modified to work
with hashcode of filePath. So all blocks related to a file would be partitioned on same operator.
Tested with recordReader and verified for sequentialFileRead that all blocks related to a file are partitioned to single operator.