Community for the Daft project and all things distributed data

Distributed Data Community

image.png

I have a generic question. Maybe ping <@U0664J23JMP> or <@U06N1KAUCUV> for visibility.

Why are we using the `entry` API here? The way this API reads is that "if min_id does not exist in this map, insert it; but *if* it does, then _use the already existing value_". Do we want that functionality? Shouldn't we expect that `min_id` does *not* exist at the time of insertion (and more strongly that we should panic in the case that it somehow does happen to exist)?

I was thinking the logic flow should be something more like `assert!(irst_stage_aggs.insert(min_id), "this value should not exist!");`

could you link the line on github you are referring to?

For example:
<https://github.com/Eventual-Inc/Daft/blob/main/src/daft-plan/src/physical_planner/translate.rs#L802>

Yeah, i'm not sure either. If you always insert it, does it break any tests? lol

e.g. if you do a groupby and then a sum + mean on the same values, you can reuse the same local sum for both aggs