alamb opened a new issue #558: URL: https://github.com/apache/arrow-rs/issues/558
**Is your feature request related to a problem or challenge? Please describe what you are trying to do.** I would like to do arithmetic on temporal arrys like `TimestampNanosecondArray` https://docs.rs/arrow/4.4.0/arrow/compute/kernels/arithmetic/fn.add.html The current compute kernels have the following signatures: ```rust pub fn add<T>( left: &PrimitiveArray<T>, right: &PrimitiveArray<T> ) -> Result<PrimitiveArray<T>> ... ``` So if I have two `ArrayRefs` (aka `Arc<dyn Array>`) I am forced to downcast them to specific known types like ```rust let left = left.as_any().downcast_ref<TimestampNanosecondArray>().unwrap(); let right = right.as_any().downcast_ref<DurationArray>().unwrap(); let result = compute::add(left, right).unwrap(); ``` However, the cast kernel is much easier to use as it does this type dispatch / downcasting internally for me: ```rust let cast_result = compute::cast(&left, DataType::Int32)` ``` **Describe the solution you'd like** I would like for the `add` (and other compute kernels) to take a `&ArrayRef` as argument and then do any type matching / downcasting necessary internally, so I, as a user, don't have to think about it **Describe alternatives you've considered** An alternate is to take an `&dyn Array` and avoid the need for wrapping the array in an `Arc` -- however, given the prevalence of `ArrayRef` / `Arc` in Arrow APIs already I would personally prefer a uniform interface with the kernels like `cast` which already take `&ArrayRef` **Additional context** There is additional discussion and context at https://github.com/apache/arrow-rs/issues/527#issuecomment-880649326 -- 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
