alamb commented on issue #194: URL: https://github.com/apache/arrow-datafusion/issues/194#issuecomment-839650054
I have been thinking about this and did some digging in the code. There are probably two major strategies: 1. Implement arrow compute kernels (as @jorgecarleitao points to in https://github.com/jorgecarleitao/arrow2/blob/main/src/compute/arithmetics/time.rs#L208) 2. Implement the operations in datafusion itself (by casting the various time arrays to their primitive i64 versions, performing whatever math is needed, and then casting them back to the resulting type) Long term I think approach 1 is likely both the "cleanest" and fastest performance approach, but it will take non trivial time. I may try to bound the time required by just implementing subtraction (end to end in arrow + datafusion), and filing tickets to fill out the rest of the operations (the most useful ones are described in the [postgres docs](https://www.postgresql.org/docs/current/functions-datetime.html)) -- 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]
