Hi Antoine, Can you expand on why that restriction is necessary/makes things easier? It seems a little strange since each dictionary batch has the ID attached, I wouldn't think it would be hard for the reader to track their arrival in any order.
Thanks, Micah On Thu, May 21, 2020 at 10:25 AM Antoine Pitrou <anto...@python.org> wrote: > > Hello, > > I'm working on getting nested dictionaries to work in the C++ IPC > implementation, together with integration tests. > > My current implementation introduces a restriction. Let's say we have > the following schema field: > > - type: List > - dictionaryEncoding: > - id: 123 > - indexType: Int32 > - children[0]: > - type: String > - dictionaryEncoding: > - id: 456 > - indexType: Int32 > > then my C++ patch requires that the dictionary batch for the inner > dictionary (id 456) appears before the dictionary batch for the outer > dictionary (id 123). It seems like a reasonable restriction, but I'd > like to check if that's ok. Also, should we add it to the spec? > > Regards > > Antoine. >