Github user mgaido91 commented on the issue: https://github.com/apache/spark/pull/20856 @cloud-fan I do like the idea of generating the code on driver side. The main issues I see are: - In the generated code we reference external objects. We have to send this array of references to all the executors; - We should implement a mechanism for identifying which code runs which part of the plan in order for the executors to be able to retrieve it... This is probably feasible but I think we have to be very careful at it. - nit: the generated code may be big, may sending it to all executors cause performance issues?
--- --------------------------------------------------------------------- To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org For additional commands, e-mail: reviews-h...@spark.apache.org