Zhipeng Zhang created FLINK-31191:
-------------------------------------
Summary: VectorIndexer should check whether doublesByColumn is
null before snapshot
Key: FLINK-31191
URL: https://issues.apache.org/jira/browse/FLINK-31191
Project: Flink
Issue Type: Bug
Components: Library / Machine Learning
Affects Versions: ml-2.2.0
Reporter: Zhipeng Zhang
Currently VectorIndexer would lead to NPE when doing checkpoint. It should
check whether `doublesByColumn` is null before calling snapshot.
logview:
[https://github.com/apache/flink-ml/actions/runs/4249415318/jobs/7389547039]
details:
[735|https://github.com/apache/flink-ml/actions/runs/4249415318/jobs/7389547039#step:4:736]Caused
by: java.lang.NullPointerException
[736|https://github.com/apache/flink-ml/actions/runs/4249415318/jobs/7389547039#step:4:737]
at
org.apache.flink.ml.feature.vectorindexer.VectorIndexer$ComputeDistinctDoublesOperator.convertToListArray(VectorIndexer.java:232)
[737|https://github.com/apache/flink-ml/actions/runs/4249415318/jobs/7389547039#step:4:738]
at
org.apache.flink.ml.feature.vectorindexer.VectorIndexer$ComputeDistinctDoublesOperator.snapshotState(VectorIndexer.java:228)
[738|https://github.com/apache/flink-ml/actions/runs/4249415318/jobs/7389547039#step:4:739]
at
org.apache.flink.streaming.api.operators.StreamOperatorStateHandler.snapshotState(StreamOperatorStateHandler.java:222)
[739|https://github.com/apache/flink-ml/actions/runs/4249415318/jobs/7389547039#step:4:740]
... 33 more
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
