[
https://issues.apache.org/jira/browse/BEAM-6473?focusedWorklogId=190951&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-190951
]
ASF GitHub Bot logged work on BEAM-6473:
----------------------------------------
Author: ASF GitHub Bot
Created on: 28/Jan/19 12:55
Start Date: 28/Jan/19 12:55
Worklog Time Spent: 10m
Work Description: mxm commented on pull request #7640: [BEAM-6473]
Materialize all inputs of transforms with side input collections
URL: https://github.com/apache/beam/pull/7640#discussion_r251402120
##########
File path: sdks/python/apache_beam/runners/portability/fn_api_runner_test.py
##########
@@ -188,22 +188,21 @@ def test_pardo_windowed_side_inputs(self):
(9, list(range(7, 10)))]),
label='windowed')
- def test_flattened_side_input(self, extended=True):
+ def test_flattened_side_input(self):
with self.create_pipeline() as p:
main = p | 'main' >> beam.Create([None])
side1 = p | 'side1' >> beam.Create([('a', 1)])
side2 = p | 'side2' >> beam.Create([('b', 2)])
- side3 = p | 'side3' >> beam.Create(['another type'])
Review comment:
Figured it was intentional, was planning on letting you review this. As a
general rule, flattening only works for matching types and Flink does not
support flattening of different types. `side1`/`side2` use a KV list coder,
whereas `side3` used a string list coder which lead to different result types.
Transcoding can be implemented, but it requires rules for translating
different coders to a single result type. I think it would be worth adding a
dedicated test for testing flattening rules for non-matching coder types.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
Issue Time Tracking
-------------------
Worklog Id: (was: 190951)
Time Spent: 2h 20m (was: 2h 10m)
> Python Flink ValidatesRunner test_flattened_side_input fails
> ------------------------------------------------------------
>
> Key: BEAM-6473
> URL: https://issues.apache.org/jira/browse/BEAM-6473
> Project: Beam
> Issue Type: Test
> Components: runner-flink, sdk-py-core
> Reporter: Maximilian Michels
> Assignee: Maximilian Michels
> Priority: Major
> Time Spent: 2h 20m
> Remaining Estimate: 0h
>
> The {{test_flattened_side_input}} test fails after merging
> [https://github.com/apache/beam/pull/7456]
> {noformat}
> ERROR: test_flattened_side_input (_main_.FlinkRunnerTest)
> ----------------------------------------------------------------------
> Traceback (most recent call last)
> File "apache_beam/runners/portability/fn_api_runner_test.py", line 205, in
> test_flattened_side_input
> label='CheckFlattenOfSideInput')
> File "apache_beam/pipeline.py", line 425, in _exit_
> self.run().wait_until_finish()
> File "apache_beam/runners/portability/portable_runner.py", line 349, in
> wait_until_finish
> self._job_id, self._state, self._last_error_message()))
> RuntimeError: Pipeline
> test_flattened_side_input_1547859357.36_07dcde9b-acfc-4e8d-b930-582f7637a07e
> failed in state FAILED: java.lang.IllegalArgumentException: PCollectionNodes
> [PCollectionNode
> {id=ref_PCollection_PCollection_12, PCollection=unique_name:
> "17side3/Map(decode).None" coder_id: "ref_Coder_BytesCoder_1" is_bounded:
> BOUNDED windowing_strategy_id: "ref_Windowing_Windowing_1" }
> ] were consumed but never produced
> {noformat}
> [https://builds.apache.org/job/beam_PostCommit_Python_VR_Flink_PR/134/console]
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
