Orson Peters

> Btw DoS resistance is not advertised I disagree. The way it is worded it sounds like it has *some* DoS resistance when there is none at all. Similarly with...

> > ```python > > df.select(pl.all().get(pl.col.price.arg_max())) > > ``` > > Oh, that's even better. This method is efficient for the in-memory engine but will be less efficient for the...

> Sorry for the ignorance, but where can I see if an operation is suitable for the new streamong engine? Nowhere, currently. But you can infer it, if some operation...

@mcrumiller It was rather tricky, we also were doing things wrong with 'very negative' indices. For example `pl.Series(["abcd"]).str.slice(-100, 2)` should return `""`, not `"ab"`.

@mcrumiller This is the current behavior of `pl.Series(["abcd"]).str.slice(i, 2)`: ```python -7 ab -6 ab -5 ab 0 ab -4 ab 1 bc -3 bc 2 cd -2 cd 3 d...

Ugh but this is inconsistent with `.slice`... Will put this on draft for now until we've discussed it.

@ritchie46 This PR will be merged later, as part of this issue: https://github.com/pola-rs/polars/issues/15232. Unless you temporarily want to make `str.slice` inconsistent with `.slice` until we fix `.slice` as well. @mcrumiller...

Is this still a problem? I don't believe that panic can still occur.

@alexander-beedie I think it's fine for `describe`.

I think defining `prev_power_of_two(0) == 0` is significantly more useful, and faster to boot. I don't think the checked and wrapping versions are necessary. Similarly I think the name `highest_bit_set`...