Theta join with CrossFunction

[Arvid Heise]

Hi Flinker,

how can you implement a theta join with the revised cross API? In particular, how can you _not_ emit for a given input pair?

With the old interface, not calling the collector was quite easy and obvious. Returning null in the new interface seems to be a bad idea.

Or is there a different way to implement a theta join with the current Flink API? FlatMap with Broadcast?

Best,

Arvid

[Stephan Ewen]

Hi!

You can always use a cross without a function (that creates a Tuple2 of the input) and then use Filter on the result, or flatMap, or anything that implements your predicate.

Stephan