felipecrv commented on PR #37202:
URL: https://github.com/apache/arrow/pull/37202#issuecomment-1680928731
> Right, it would add a bit of complexity, hopefully not too much. Most
types would probably be ok with 0xff for all buffer data. Perhaps
(list/binary/dense union) offsets and run-ends would better be zero, though?
If we were to got that route we would have to define the "default value" for
each type similar to how protobuf 3 is allowed to omit data and default values
are assumed.
- Boolean - `false`
- Number - `0.0`
- List - empty list (offsets would have to be 0 and not `0xffffffff`)
- Unions - the default value of the first type
> > This sounds reasonable, but we should clarify this more generally? It
sounds like we aren't precise enough with that field nullability means for the
physical data.
>
> Hmm, it seems like this could be discussed on the mailing-list. Perhaps we
don't want to impose undue constraints here.
I vote in favor of not imposing constraints. This is why I don't think the
way to go here is "fixing" `MakeArrayOfNull` but relaxing the constraint
checking that's failing on per-spec valid data.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
