jorisvandenbossche commented on PR #34645:
URL: https://github.com/apache/arrow/pull/34645#issuecomment-1481293048

   > For the short term I don't have a strong opinion. As far as I know 
@icexelloss is the only one consuming cast via Substrait in this way currently 
so I'd be ok with matching spark over postgres.
   
   That's certainly a good reason to go with how it is now, no objections here.
   
   ---
   
   If we would have a discussion what we think longer term is the better 
"default" behaviour for substrait, I personally think the more sensible 
behaviour for users is to not allow overflow. In a DataFrame context, if this 
happens, I think it is almost always accidental and not intentional, and so I 
think it's better to not do that silently (but error, and let the user specific 
explicitly they want overflow behaviour). 
   But also, it might not matter that much what the default is for substrait? 
Because it can't have the "good" default for various use cases at the same 
time. So what matters is that there is a way to specify the behaviour, so that 
producers of substrait can specify the behaviour so they can choose whathever 
default they want for the users of their system, and ensure this gets reflected 
correctly in the substrait plan?
   
   


-- 
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]

Reply via email to