[
https://issues.apache.org/jira/browse/BEAM-11881?focusedWorklogId=563891&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-563891
]
ASF GitHub Bot logged work on BEAM-11881:
-----------------------------------------
Author: ASF GitHub Bot
Created on: 10/Mar/21 17:07
Start Date: 10/Mar/21 17:07
Worklog Time Spent: 10m
Work Description: TheNeuralBit commented on a change in pull request
#14135:
URL: https://github.com/apache/beam/pull/14135#discussion_r591707315
##########
File path: sdks/python/apache_beam/dataframe/expressions.py
##########
@@ -245,7 +285,7 @@ def requires_partition_by(self):
return partitionings.Nothing()
def preserves_partition_by(self):
- return partitionings.Nothing()
+ return partitionings.Index()
Review comment:
Yeah my motivation here was just that `preserves=Nothing()` is
potentially confusing. In retrospect maybe it's not much of a concern since
`PartitioningSession` should catch incorrect specifications. Regardless - I
like your idea (in the next comment) to solve this with naming.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
Issue Time Tracking
-------------------
Worklog Id: (was: 563891)
Time Spent: 3h (was: 2h 50m)
> DataFrame subpartitioning order is incorrect
> --------------------------------------------
>
> Key: BEAM-11881
> URL: https://issues.apache.org/jira/browse/BEAM-11881
> Project: Beam
> Issue Type: Bug
> Components: sdk-py-core
> Reporter: Brian Hulette
> Priority: P2
> Labels: dataframe-api
> Time Spent: 3h
> Remaining Estimate: 0h
>
> Currently we've defined
> Nothing() < Index([i]) < Index([i,j]) < .. < Index() < Singleton()
> s.t. Singleton is a subpartitoning of Index, is a subpartitioning of
> Index([i,j]), but this is incorrect. The order should be
> Singleton() < Index([i]) < Index([i,j]) < .. < Index() < Nothing()
> s.t. every other partitioning is a subpartitioning of Singleton. This is
> logical, since Singleton will collect the largest amount of data on a single
> node, partitioning by a single index will be alittle more distributed, and
> partitioning by the full Index() will be the most distribtued.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
