[FLINK-15801] Timestamp extractor created from properties does not work for some physical fields

• Type: Sub-task
• Status: Closed
• Priority: Major
• Resolution: Won't Fix
• Affects Version/s: 1.9.0, 1.10.0
• Component/s: Table SQL / API
• Labels: pull-request-available
• Assignee: Unassigned
• Reporter: Dawid Wysakowicz
• Created: 29/Jan/20 09:47
• Updated: 29/Apr/21 16:44
• Resolved: 29/Apr/21 16:44

Description

If a timestamp extractor is created from properties it can not use a physical field if the name of that field is equal to the logical field of the rowtime field.

The code below fails:

		StreamExecutionEnvironment streamExecutionEnvironment = StreamExecutionEnvironment.getExecutionEnvironment();
		streamExecutionEnvironment.setStreamTimeCharacteristic(TimeCharacteristic.EventTime);
		StreamTableEnvironment fsTableEnv = StreamTableEnvironment.create(streamExecutionEnvironment);
		String allEventsTable = "allEventsTable";
		fsTableEnv.connect(new Kafka()
			.version("universal")
			.topic("events")
			.property("zookeeper.connect", "")
			.property("bootstrap.servers", "localhost:9092")
			.property("group.id", "dummyquery").startFromLatest())
			.withSchema(new Schema()
				.field("rule_id", Types.INT)
				.field("sourceAddress", Types.STRING)
				.field("deviceProduct", Types.STRING)
				.field("destHost", Types.STRING)
				.field("extra", Types.STRING)
				.field("rowtime", Types.SQL_TIMESTAMP)
				.rowtime(new Rowtime().timestampsFromField("rowtime").watermarksPeriodicBounded(2000))

			)
			.withFormat(new Json().failOnMissingField(false).deriveSchema())
			.inAppendMode()
			.registerTableSource(allEventsTable);

		Table result = fsTableEnv.sqlQuery("select * from allEventsTable where sourceAddress='12345431'");

		DataStream alert = fsTableEnv.toAppendStream(result, Row.class);
		alert.print();

with exception:

Exception in thread "main" org.apache.flink.table.api.ValidationException: Field 'rowtime' could not be resolved by the field mapping.
    at org.apache.flink.table.sources.TableSourceValidation.resolveField(TableSourceValidation.java:245)
    at org.apache.flink.table.sources.TableSourceValidation.lambda$validateTimestampExtractorArguments$6(TableSourceValidation.java:202)
    at java.util.stream.ReferencePipeline$3$1.accept(ReferencePipeline.java:193)
    at java.util.Spliterators$ArraySpliterator.forEachRemaining(Spliterators.java:948)
    at java.util.stream.AbstractPipeline.copyInto(AbstractPipeline.java:481)
    at java.util.stream.AbstractPipeline.wrapAndCopyInto(AbstractPipeline.java:471)
    at java.util.stream.AbstractPipeline.evaluate(AbstractPipeline.java:545)
    at java.util.stream.AbstractPipeline.evaluateToArrayNode(AbstractPipeline.java:260)
    at java.util.stream.ReferencePipeline.toArray(ReferencePipeline.java:438)

The problem is that the field is removed from the FieldMapping in org.apache.flink.table.descriptors.SchemaValidator#deriveFieldMapping.

One possible solution could be to add:

				if (isRowtime) {
					Optional<String> timestampSource = properties.getOptionalString(SCHEMA + "." + i + "." + ROWTIME_TIMESTAMPS_FROM);
					timestampSource.ifPresent(s -> mapping.put(s, s));
				}

We should also consider the case what happens if we compute generated columns on fields that were pruned in a similar way.

Reported by a user: https://stackoverflow.com/questions/59857057/how-to-define-an-apache-flink-table-with-row-time-attribute

Issue Links

• links to: GitHub Pull Request #11782