ESQL: ValueSourceReader sometimes tries to append empty blocks #128959
Description
Got this stack trace from Serverless from a simple query:
java.lang.IllegalArgumentException: Block [BytesRefVectorBlock[vector=BytesRefArrayVector[positions=0]]] does not have same position count: 0 != 10
at org.elasticsearch.compute.data.Page.<init>(Page.java:88)
at org.elasticsearch.compute.data.Page.appendBlocks(Page.java:174)
at org.elasticsearch.compute.lucene.ValuesSourceReaderOperator.process(ValuesSourceReaderOperator.java:174)
at org.elasticsearch.compute.operator.AbstractPageMappingOperator.getOutput(AbstractPageMappingOperator.java:85)
at org.elasticsearch.compute.operator.Driver.runSingleLoopIteration(Driver.java:265)
at org.elasticsearch.compute.operator.Driver.run(Driver.java:185)
at org.elasticsearch.compute.operator.Driver$1.doRun(Driver.java:406)
at org.elasticsearch.common.util.concurrent.AbstractRunnable.run(AbstractRunnable.java:27)
at org.elasticsearch.compute.operator.DriverScheduler$1.doRun(DriverScheduler.java:57)
at org.elasticsearch.common.util.concurrent.AbstractRunnable.run(AbstractRunnable.java:27)
at org.elasticsearch.common.util.concurrent.TimedRunnable.doRun(TimedRunnable.java:35)
at org.elasticsearch.common.util.concurrent.ThreadContext$ContextPreservingAbstractRunnable.doRun(ThreadContext.java:1044)
at org.elasticsearch.common.util.concurrent.AbstractRunnable.run(AbstractRunnable.java:27)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1095)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:619)
at java.lang.Thread.run(Thread.java:1447)
This indicates that the ValueSourceReaderOperator tries to append blocks with 0 positions, even though we should fetch field values for 10 positions.
The query that triggered this was simple and in essence
FROM idx | WHERE field == "value"
together with a specific time range picked via the filter request field. idx is actually a data stream, which may be relevant. Happened for some time ranges, but not for others - maybe due to rollover at some point in time and thus picking up different shards.
In the specific case, field was a constant keyword field; but we also saw problems with just
FROM idx
with a specific time range in the filter request field.