I think this might just be overly cautious because it's not incorrect to check these. Also, in fact I think you could have "forward" with differing parallelism in some earlier versions which would lead to some unexpected results. Not sure if that is still possible now.
> On 24. Jan 2018, at 09:50, Marvin777 <xymaqingxiang...@gmail.com> wrote: > > Hi all: > > We all know that the conditions for chaining are 9, like this: > <image.png> > However, I have a question about 2, 3, 8 in the figure above. > As far as I know, the operator cannot be null, and If the partitioner of the > edge is forward, then their parallelism must be the same. Maybe I'm > completely missing something... > > Waiting for your reply. > > Best regard.
