Output attribute assignment is done solely via name. This adds overhead of calculating the hash from the attribute name and hash lookup for each output attribute for each submitted tuple. Determining the attribute index once in initialize() and assign attributes via index would be more performant.
ghost
commented
4 years ago
Output attribute assignment is done solely via name. This adds overhead of calculating the hash from the attribute name and hash lookup for each output attribute for each submitted tuple. Determining the attribute index once in
initialize()and assign attributes via index would be more performant.https://github.com/IBMStreams/streamsx.kafka/blob/d92513cdd74a86a186074dc8d2d18148735e04e8/com.ibm.streamsx.kafka/impl/java/src/com/ibm/streamsx/kafka/operators/AbstractKafkaConsumerOperator.java#L1029