jonkeane commented on PR #13789: URL: https://github.com/apache/arrow/pull/13789#issuecomment-1216348088
I'm not wholly against requiring registration if it is significantly simpler or is the only way. But I'm not sure that requiring registration will necessarily help make errors easier to understand (we'll need to have some good, clear error messages when anything goes wrong here regardless of the path we take). And I might even argue that having only one kind of registration (for the user-defined scalar function) sets that apart in a positive way (crafting one of those will take care + do things that are surprising compared to all of our other bindings, after all). IMO: people shouldn't (need to) think too much about how the bindings work, just use the code you're used to, and if it works, great, if it gives you a (clear) error that some function isn't native to arrow yet, you can respond to that as is. If we have two different types of registration we'll need to find a way to explain how they are different (and even with great docs explaining that difference, there will be a contingent of people who will wonder what's the difference between the hypothetical `register_binding(...)` and `register_scalar_function(...)`. But, again, if autoregistering is a sticking point (though looking through the code that's here it doesn't look like it is...), then sure we can go the registration route first. (or if all the autoregistration stuff is overly burdensome from a maintenance perspective too, of course). -- 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
