> > First of all I would like to ask why we use both type_codes and child_ids > for Union types. It seems that we can already cover the logical types a > union has using type_codes alone. What’s the point of using child_ids?
The two are inverses of each other: https://github.com/apache/arrow/blob/590541d251075b8c0bfabc6e574b25928f386bb5/cpp/src/arrow/type.cc#L586 Arrow supports unions containing more than one of each logic type. Secondly I would like to ask about the maximum amount of time permitted > when running unit tests. I will definitely profile and speed up my tests > prior to the pull request so I would like to know about the expectation > first. I can't answer this specifically. But if you are planning on submitting a PR, I would encourage you to think about possibly trying to split it up into multiple smaller PRs. We have nightly tests that we can use for larger stress test cases, but I would assume any normal unit tests should run pretty quickly (i.e. if timing is a concern, it seems like there might be some issues with the tests being run). On Sun, Nov 8, 2020 at 6:56 PM Ying Zhou <yzhou7...@gmail.com> wrote: > The work of converting Arrow Arrays, ChunkedArrays, RecordBatches and > Tables to ORC files is about 50% done. Now I have two questions. > > First of all I would like to ask why we use both type_codes and child_ids > for Union types. It seems that we can already cover the logical types a > union has using type_codes alone. What’s the point of using child_ids? > > Secondly I would like to ask about the maximum amount of time permitted > when running unit tests. I will definitely profile and speed up my tests > prior to the pull request so I would like to know about the expectation > first. > > Thanks, > Ying Zhou
