> > 1. Is there any reason to expect these will need to be batched into one new > version of the Arrow format? Or would we have no problem adding (as an > theoretical example) RLE arrays in format version 2, and then later string > views in version 3?
These shouldn't require a major version bump if they are new enum types. 2. Is the general process that two implementations will implement the > vector type with integration tests prior to final voting on adding the > array type to the format? Yes, at least the last time we made this decision it was decided that C++ and Java were the required ones. On Thu, Apr 14, 2022 at 1:04 PM Will Jones <will.jones...@gmail.com> wrote: > Hi all, > > I have a few questions to understand expectations for how work on these > could proceed: > > 1. Is there any reason to expect these will need to be batched into one new > version of the Arrow format? Or would we have no problem adding (as an > theoretical example) RLE arrays in format version 2, and then later string > views in version 3? > 2. Is the general process that two implementations will implement the > vector type with integration tests prior to final voting on adding the > array type to the format? > > Thanks, > > Will Jones >