[FLINK-40949][table-runtime] Update UnsliceAssigners NULL window exception message by ykisana · Pull Request #29424 · apache/flink

ykisana · 2026-10-07T20:13:36Z

What is the purpose of the change

UnsliceAssigners.WindowedUnsliceAssigner should not have a NULL Window edge elements at either (windowStartIndex and windowStartIndex).

The current exception messaging (RowTime field should not be null.) isn't clear. We can make this more clear.

As per: FLINK-40949

Brief change log

CreateWindow in WindowedUnsliceAssigner has an updated message.
Old: RowTime field should not be null.
New: Both window start and window end should not be null.

Verifying this change

This is just a simple renaming of a string. We don't need explicit tests.

Does this pull request potentially affect one of the following parts:

Dependencies (does it add or upgrade a dependency): no
The public API, i.e., is any changed class annotated with @Public(Evolving): no
The serializers: no
The runtime per-record code paths (performance sensitive): no
Anything that affects deployment or recovery: JobManager (and its components), Checkpointing, Kubernetes/Yarn, ZooKeeper: no
The S3 file system connector: no

Documentation

Does this pull request introduce a new feature? no
If yes, how is the feature documented? not applicable

Was generative AI tooling used to co-author this PR?

Yes (please specify the tool below)

While AI was used before for PR desc/comments, current state of everything has no AI.

flinkbot · 2026-10-07T20:21:45Z

CI report:

8086cad Azure: SUCCESS

Bot commands

The @flinkbot bot supports the following commands:

@flinkbot run azure re-run the last Azure build

MartijnVisser · 2026-10-08T07:26:45Z

@ykisana For the next time, please wait with opening a PR until you're assigned to the Jira ticket

snuyanzin · 2026-10-08T10:13:51Z

        public long assignSliceEnd(RowData element, ClockService clock) {
+            if (element.isNullAt(sliceEndIndex)) {
+                throw new IllegalStateException(
+                        "Slice end of a sliced input row is null. This is a bug.");


and what is call to action to user?

this is a bug where?
flink, jdk, data, something else?
if I see it as a user how can this message help me?

@snuyanzin I tried to follow bug messages in the rest of the repo 'This is a bug' is found in other places. It's an IllegalStateException. User can't do anything about it, bug on our side.
Updating to make this more clear though. Let me know if you have suggestions.

User can't do anything about it, bug on our side.

if somewhere is NPE it is also bug in Flink, however I hardly tend to think that we need to insert tons of checks for NPE and for every tell that bug in our side because something is null...

Can we better try to understand cases we come here and why?

@snuyanzin Yeah I agree. What's the process for that?
The ticket is for error messaging. I can do a follow up for looking at paths where we actually enter this state.

snuyanzin

Slice end of the input row is null, but it is always set by the upstream local window aggregate. This is a bug in Flink. Please file an issue.

this is false

after looking into that it is data/user issue rather than Flink bug

snuyanzin · 2026-10-08T19:34:58Z

you can repro this with a query like

SELECT a, window_start, window_end, COUNT(*) FROM TUMBLE(TABLE src, DESCRIPTOR(rt), INTERVAL '5' SECOND) GROUP BY a, window_start, window_end

and then pass null for one of the values of rt

ykisana · 2026-10-08T21:30:22Z

you can repro this with a query like
SELECT a, window_start, window_end, COUNT(*) FROM TUMBLE(TABLE src, DESCRIPTOR(rt), INTERVAL '5' SECOND) GROUP BY a, window_start, window_end
and then pass null for one of the values of rt

Thanks for looking into this and finding this! Yeah those code paths are still reachable. I was able to confirm with that query.

I believe WindowedUnsliceAssigner seeing NULL bounds is still unreachable, so I kept that in. Updated message following other examples in Flink.

snuyanzin · 2026-10-08T21:40:07Z

+                throw new IllegalStateException(
+                        "Window start or window end of the input row is null. "
+                                + "Both should be set by the upstream window operator. "
+                                + "This is a bug. Please file an issue.");


why do we need this?
i'm pretty sure same story

@snuyanzin At this stage we only expect valid windows, NULL rowtimes are rejected upstream. I believe this path is unreachable.

I believe this path is unreachable.

it is not a proof, need something more reliable as a proof

I believe this path is unreachable.

also even if it is true, why do we even touch unreachable message?

@snuyanzin I'm happy to leave message as is, or maybe just change to invalid state exception.
The purpose of ticket was clearer messaging: https://issues.apache.org/jira/browse/FLINK-40949

As for why it should be unreachable:

First Check:
http://github.com/apache/flink/blob/master/flink-table/flink-table-runtime/src/main/java/org/apache/flink/table/runtime/operators/wmassigners/WatermarkAssignerOperator.java#L136-L140

Second Check:
http://github.com/apache/flink/blob/master/flink-table/flink-table-runtime/src/main/java/org/apache/flink/table/runtime/operators/window/tvf/operator/UnalignedWindowTableFunctionOperator.java#L200-L205

If survived these checks, they get real bounds:
http://github.com/apache/flink/blob/master/flink-table/flink-table-runtime/src/main/java/org/apache/flink/table/runtime/operators/window/tvf/operator/WindowTableFunctionOperatorBase.java#L109-L121

Expected real bounds here:
http://github.com/apache/flink/blob/master/flink-table/flink-table-planner/src/main/scala/org/apache/flink/table/planner/plan/rules/physical/stream/StreamPhysicalWindowAggregateRule.scala#L111-L115

Thus real bounds here (which calls the function I changed):
http://github.com/apache/flink/blob/master/flink-table/flink-table-planner/src/main/java/org/apache/flink/table/planner/plan/nodes/exec/stream/StreamExecWindowAggregateBase.java#L77-L90

Note, your counter example only proved for SliceAssigners in a DIFFERENT path.
In this path, SliceAssigners also is expected real bounds.

UnsliceAssigners only lives here. It doesn't live in other paths without real bound pruning like here.

Let me know if I missed anything or how I should handle this.
Change exception? Or just leave as is? Ticket wanted something different here.

cc @MartijnVisser

change only message like

throw new RuntimeException("Both window start and window end should not be null.");

@snuyanzin updated, thanks for confirming. simple pr now

snuyanzin · 2026-10-09T07:23:07Z

the commit/pr message is still vague

Don't report NULL window bounds as a NULL rowtime

reporting it as NOT NULL rowtime also satisfying this message however it is not true

snuyanzin · 2026-10-09T07:24:03Z

also ci is red

…ption message

snuyanzin reviewed Oct 8, 2026

View reviewed changes

Comment thread ...ava/org/apache/flink/table/runtime/operators/window/tvf/slicing/SlicedSliceAssignerTest.java Outdated

ykisana force-pushed the FLINK-40949 branch 2 times, most recently from d725e52 to 271fe01 Compare October 8, 2026 11:31

snuyanzin requested changes Oct 8, 2026

View reviewed changes

ykisana force-pushed the FLINK-40949 branch from 271fe01 to b8a3fc1 Compare October 8, 2026 21:13

ykisana changed the title ~~[FLINK-40949][table-runtime] Fail with a clear error on NULL window bounds in windowed assigners~~ [FLINK-40949][table-runtime] Don't report NULL window bounds as a NULL rowtime in WindowedUnsliceAssigner Oct 8, 2026

ykisana force-pushed the FLINK-40949 branch from b8a3fc1 to 77e6ab6 Compare October 8, 2026 21:18

ykisana requested a review from snuyanzin October 8, 2026 21:30

snuyanzin reviewed Oct 8, 2026

View reviewed changes

Comment thread ...g/apache/flink/table/runtime/operators/window/tvf/unslicing/WindowedUnsliceAssignerTest.java Outdated

ykisana force-pushed the FLINK-40949 branch 2 times, most recently from c7f2459 to 874c00b Compare October 9, 2026 06:02

ykisana force-pushed the FLINK-40949 branch from 874c00b to 6c6a8ba Compare October 9, 2026 16:31

ykisana changed the title ~~[FLINK-40949][table-runtime] Don't report NULL window bounds as a NULL rowtime in WindowedUnsliceAssigner~~ [FLINK-40949][table-runtime] Update UnsliceAssigners NULL window exception message Oct 9, 2026

[FLINK-40949][table-runtime] Update UnsliceAssigners NULL window exce…

8086cad

…ption message

ykisana force-pushed the FLINK-40949 branch from 6c6a8ba to 8086cad Compare October 9, 2026 16:40

ykisana requested a review from snuyanzin October 10, 2026 02:30

Conversation

ykisana commented Oct 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What is the purpose of the change

Brief change log

Verifying this change

Does this pull request potentially affect one of the following parts:

Documentation

Was generative AI tooling used to co-author this PR?

Uh oh!

flinkbot commented Oct 7, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

CI report:

Uh oh!

MartijnVisser commented Oct 8, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

snuyanzin left a comment

Choose a reason for hiding this comment

Uh oh!

snuyanzin commented Oct 8, 2026

Uh oh!

ykisana commented Oct 8, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

snuyanzin commented Oct 9, 2026

Uh oh!

snuyanzin commented Oct 9, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

ykisana commented Oct 7, 2026 •

edited

Loading

flinkbot commented Oct 7, 2026 •

edited

Loading