Thanks for the pointer. so a runner might do substitution and implement a primitive in terms of other primitives, but leave the original in the graph...
My take is that the graph should be correct (by some stricter definition) when it's submitted, but a runner can do things that make it "less" correct and we should perhaps not have as strict of checks after the fact. Treating transforms with primitive URNs but sub-transforms as composites makes sense here. FWIW, it's worth getting this in to unblock things. [ Full content available at: https://github.com/apache/beam/pull/6461 ] This message was relayed via gitbox.apache.org for [email protected]
