[
https://issues.apache.org/jira/browse/FLINK-39380?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18119739#comment-18119739
]
sepuri sai krishna commented on FLINK-39380:
--------------------------------------------
[~autophagian] is this one still on your list, or open for someone else to pick
up? Happy
either way, I just don't want to duplicate work while you are going through the
other
subtasks.
If it is free I would like to take it. From reading FLIP-567 and the current
harness, the
pieces a TestSnapshot needs look like they already exist:
{noformat}
per-partition state TestHarnessStateManager.stateByKey
pending / fired timers TestHarnessTimerManager.pendingTimersByPartition,
firedTimers
watermarks per table TestHarnessTimerManager.watermarkByTable
output rows ProcessTableFunctionTestHarness.output,
functionOutput
harness metadata arguments, defaultTableArgument, outputKind
{noformat}
Two things I would want your opinion on before writing anything:
1. FLIP-567 shows both {{Builder.restoreFromSnapshot(TestSnapshot)}} and a
static
{{ProcessTableFunctionTestHarness.restoreFromSnapshot(snapshot)}} in
different places.
Which did you intend, or both?
2. Whether the snapshot should deep-copy state values at snapshot() time. A
shallow copy
would let a mutable POJO state keep changing after the snapshot was taken,
which would
be surprising for the reusable-scenario use case the FLIP describes.
> Add support for test snapshotting/restoration to
> ProcessTableFunctionTestHarness
> --------------------------------------------------------------------------------
>
> Key: FLINK-39380
> URL: https://issues.apache.org/jira/browse/FLINK-39380
> Project: Flink
> Issue Type: Sub-task
> Reporter: Mika Naylor
> Priority: Major
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
