alamb commented on pull request #9376: URL: https://github.com/apache/arrow/pull/9376#issuecomment-774460448
> @alamb I am not sure I understand why folding should be done on the physical plan level or really depending on this PR , it should be possible without changes in this PR (just by having some Expr -> Expr rules)? They could share the same evaluation code though? @Dandandan -- yes, I think constant folding looks like `Expr` --> `Expr`. But as you hint at, if the physical evaluation of functions is entirely separate from the plan time evaluation, we will end up with two parallel implementations for evaluation that need to be kept in sync (one for `Expr` and one for the physical runtime) -- using the same implementation for both I think avoids a lot of potential inconsistency (and source of bugs) We'll see if that turns out to be possible, but I think it should be ---------------------------------------------------------------- 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. For queries about this service, please contact Infrastructure at: [email protected]
