liukun4515 commented on issue #3325: URL: https://github.com/apache/arrow-datafusion/issues/3325#issuecomment-1236470104
> I would summarize this issue as "sql engines like postgres and spark" avoid supporting signed and unsigned variants of types (presumably because for most usecases the difference in ranges are not important enough to justify the additional implementation and user comlexity). > > However, Arrow **does** support unsigned variants (e.g `UInt64` and `Int64`, `UInt32` and `Int32`, etc) and DataFusion has a non trivial amount of code to support those unsigned types (e.g. check out the uses of `UInt64`: https://github.com/apache/arrow-datafusion/search?q=UInt64) > > I think it would be a reasonable discussion to start about 'should DataFusion support unsigned types at all' but given that datafusion does support those types internally, I think it is inconsistent to disallow them in the SQL syntax and I can't think of any reason not to support them > > Thus after some additional thought I believe we should merge #3309 > > @liukun4515 if you disagree, do you have any additional concern you have (apart from unsigned types not being supported in spark or postgres)? > > cc @liukun4515 @kmitchener > > I would also be interested in any other opinions from the community (@houqp @andygrove ) I know in the core of datafusion we have supported some operation for the unsigned value. It's good to support unsigned value for datafusion, but I just concern it will cause some undefined behavior. If some users want this feature, I think it can be supported, and also need to update and add document to describe them. @alamb @kmitchener -- 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: github-unsubscr...@arrow.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
