Github user tzulitai commented on a diff in the pull request:
https://github.com/apache/flink/pull/5393#discussion_r165083343
--- Diff:
flink-connectors/flink-connector-kinesis/src/main/java/org/apache/flink/streaming/connectors/kinesis/internals/KinesisDataFetcher.java
---
@@ -584,17 +594,34 @@ private static ShardMetricsReporter
registerShardMetrics(MetricGroup metricGroup
// Miscellaneous utility functions
//
------------------------------------------------------------------------
+ /**
+ * Function to map a Kinesis shard to a Flink subtask index.
+ */
+ public interface ShardToSubtaskIndexFn {
+ /**
+ * Function to map a Kinesis shard to a Flink subtask index.
+ *
+ * @param shard the shard to determine
+ * @param totalNumberOfSubtasks total number of subtasks
+ * @return index or hash code
+ */
+ // TODO: extra parameter can be eliminated by creating hash
function after runtime context is present
+ int getSubTaskIndex(StreamShardHandle shard, int
totalNumberOfSubtasks);
--- End diff --
I actually prefer passing the `totalNumberOfSubtasks` value independently,
instead of passing in `runtimeContext`. IMO, it provides more context on the
nature of the assignment.
Moreover, IMO, having a factory as the API is much more complicating for
the user.
---
