There are reasonable implementations that might not always be able to request a next interval (imagine state passed in from the args, we can make one request, but not the next). Now there is no way to do this without exceptions or null.
Since the ScaldingPlatform just returns the WaitingState from RunningState.succeed, that second one can be null or can throw on .prepare, and it seems things will work, just be a bit ugly.
https://github.com/twitter/summingbird/blob/develop/summingbird-batch-hadoop/src/main/scala/com/twitter/summingbird/batch/JobState.scala#L25
There are reasonable implementations that might not always be able to request a next interval (imagine state passed in from the args, we can make one request, but not the next). Now there is no way to do this without exceptions or null.
Since the ScaldingPlatform just returns the WaitingState from RunningState.succeed, that second one can be null or can throw on
.prepare, and it seems things will work, just be a bit ugly.