Sorting bounded data stream

2020-01-30 Thread Łukasz Jędrzejewski
Hi all, In Flink 1.9 couple of changes was introduced to deal with bounded streams e.g.  BoundedOneInput interface. I'm wondering would it be doable to do some kind of global sort after receiving end input event on finished data stream source, using only DataStream API? We have made some exp

Re: Sorting bounded data stream

2020-01-30 Thread Jingsong Li
Hi Łukasz, First, we are planning to design and implement the BoundedStream story, which will be discussed further in 1.11 or 1.12. SortedMapState was discussed in FLINK-6219 [1], But there are some problems that can not be solved well, so they have not been introduced. If it is a pure BoundedSt

Re: Sorting bounded data stream

2020-01-30 Thread Łukasz Jędrzejewski
Hi, Thank you very much for suggestions. I will check out the UnilateralSortMerge. However in our case we are using checkpoints. Kind regards, Łukasz W dniu 31.01.2020 o 07:54, Jingsong Li pisze: Hi Łukasz, First, we are planning to design and implement the BoundedStream story, which will

Re: Sorting bounded data stream

2020-01-30 Thread Jingsong Li
Hi, If you are using checkpoints, I think a simple way is using a ListState to store all coming records. And in endInput(), drain all records from ListState to a sorter to sort all records. Best, Jingsong Lee On Fri, Jan 31, 2020 at 3:10 PM Łukasz Jędrzejewski wrote: > Hi, > > Thank you very m