[BEAM-10226] BeamSQL might fail on test_distinct_bytes after migrated to Sets transform

• Type: Task
• Status: Open
• Priority: P3
• Resolution: Unresolved
• Component/s: sdk-java-core
• Labels: Clarified
• Assignee: Unassigned
• Reporter: Rui Wang
• Created: 09/Jun/20 20:34
• Updated: 04/Jun/22 16:37

Description

This query breaks Java code import. Need to diagnose the root cause and have a fix properly.

 public void test_distinct_bytes() {
    String sql = "SELECT DISTINCT val.BYTES\n"
        + "from (select b\"1\" BYTES union all\n"
        + "      select cast(NULL as bytes) union all\n"
        + "      select b\"-1\" union all\n"
        + "      select b\"1\" union all\n"
        + "      select cast(NULL as bytes)) val";

    ZetaSQLQueryPlanner zetaSQLQueryPlanner = new ZetaSQLQueryPlanner(config);
    BeamRelNode beamRelNode = zetaSQLQueryPlanner.convertToBeamRel(sql);
    PCollection<Row> stream = BeamSqlRelUtils.toPCollection(pipeline, beamRelNode);

    Schema singleField = Schema.builder().addByteArrayField("field1").build();
    PAssert.that(stream)
        .containsInAnyOrder(
            Row.withSchema(singleField).addValues("123".getBytes(UTF_8)).build());
    pipeline.run().waitUntilFinish(Duration.standardMinutes(PIPELINE_EXECUTION_WAITTIME_MINUTES));
  }

This query will throws

Forbidden IOException when reading from InputStream
java.lang.IllegalArgumentException: Forbidden IOException when reading from InputStream
	at org.apache.beam.sdk.util.CoderUtils.decodeFromSafeStream(CoderUtils.java:118)
	at org.apache.beam.sdk.util.CoderUtils.decodeFromByteArray(CoderUtils.java:98)
	at org.apache.beam.sdk.util.CoderUtils.decodeFromByteArray(CoderUtils.java:92)
	at org.apache.beam.sdk.util.CoderUtils.clone(CoderUtils.java:141)