Orson Peters
Orson Peters
> I'll just point out that if speed is of the essence then the current proposal already has 3 open source implementations I don't believe either proposal requires a lot...
There is actually a problem with the singular NaN count for data systems which use IEEE 754 total ordering (such as datafusion), they would need two counts for efficient page...
> I don't see adopting total ordering as a one way door, we can always add a nan count mechanism later. I do see it as a one-way door because...
The issue in the title ("Allow pl.Expr.get handle null index values") is fixed by https://github.com/pola-rs/polars/pull/15239, I've moved the more general issue of allowing out-of-bounds indices in get/gather to this dedicated...
It seems that it goes wrong in `crates/polars-lazy/src/physical_plan/expressions/take.rs`, where it considers nulls as out-of-bounds. @ritchie46 is that intended?
I'm sorry, I really don't see the added value of `lookup` over the mentioned 'usual approaches'.
We've discussed this and decided that we want to do a breaking change for this. It makes our explodes more expensive because we have to insert nulls where there previously...
@Object905 You can always get the current behavior of `x.explode()` with `pl.when(x.list.len() == 0).then(pl.lit([None])).otherwise(x).explode()`. In fact, you can start transitioning now because that also has the correct behavior today. If...
The problem here is the `sign` function, which currently always returns 64-bit signed integers. Then once the 64-bit signed integer is combined with the `Float32`, the `Float32` is upcast to...
Despite the implementation being written as an `if-else` chain, the resulting code is branchless: .