Status

Current state: "Under Discussion"

Discussion thread: here

JIRA: KAFKA-8953

Please keep the discussion on the mailing list rather than commenting on the wiki (wiki discussions get unwieldy fast).

Motivation

Kafka Streams ships couple of different timestamp extractors, one named `UsePreviousTimeOnInvalidTimestamp`.

Given the latest improvements with regard to time tracking, it seems appropriate to rename this class to `UsePartitionTimeOnInvalidTimestamp`, as we now have fixed definition of partition time, and also passed partition time in extract method of class ExtractRecordMetadataTimestamp.

Public Interfaces

UsePreviousTimeOnInvalidTimeStamp will be deprecated if class name is changed.

public class UsePreviousTimeOnInvalidTimestamp extends ExtractRecordMetadataTimestamp {
...
}

Current method signature for onInvalidTimestamp

public long onInvalidTimestamp(final ConsumerRecord<Object, Object> record, final long recordTimestamp, final long partitionTime)

throws StreamsException {
...
}

Proposed Changes

Change 'UsePartitionTimeOnInvalidTimeStamp'  to 'UsePartitionTimeOnInvalidTimestamp'

public class UsePreviousTimeOnInvalidTimestamp extends ExtractRecordMetadataTimestamp {
...
}

public class UsePartitionTimeOnInvalidTimestamp extends ExtractRecordMetadataTimestamp {
...
}

If we check the implementation of onInvalidTimestamp present in 'UsePreviousTimeOnInvalidTimestamp' which is abstract method of super class  ExtractRecordMetadataTimestamp argument long previousTimestamp has been changed to long partitionTime so it seems reasonable for us to change the class name as main functionality of this class is to return the valid time stamp of record's partition.

Also, there will be additional minor changes to be made if community agrees on the mentioned proposal which are listed below:

• UsePreviousTimeOnInvalidTimestampTest
• javadoc changes in ExtractRecordMetaDataTimeStamp.java, FailOnInvalidTimeStamp.java, LogAndSkipOnInvalidTimeStamp.java, WallClockTimeStampExtractor.java

Compatibility, Deprecation, and Migration Plan

UsePreviousTimeOnInvalidTimestamp won't be present instead 'UsePartitionOnInvalidTimeStamp' will be there in next major version without change in the functionality but it's clear from the name that this will be used for partition time. But we might need to deprecate UsePreviousTimeOnInvalidTimeStamp first and then introduce new one.

Rejected Alternatives

None.