Add log/2 and exponential/1 to Series #549
Conversation
This only accepts numbers as bases.
| @doc type: :element_wise | ||
| @spec log(argument :: Series.t(), base :: number()) :: Series.t() | ||
| def log(argument, base) when is_number(base) do | ||
| if base <= 0, do: raise(ArgumentError, "base must be a positive number") |
There was a problem hiding this comment.
I think that it will be better, in general, to return negative infinity for 0 in the logarithm. However, it looks like there is no support for infinity in Explorer Series/DataFrame.
There was a problem hiding this comment.
@msluszniak we do support it, it seems the error is elsewhere. Likely Kino.Explorer.
There was a problem hiding this comment.
Yes, probably the aggregate functions like min, max, and mean don't work properly. I'll send an issue to Kino.Explorer.
There was a problem hiding this comment.
The bug is in Explorer as well. Reported here: #550
|
|
||
| """ | ||
| @doc type: :element_wise | ||
| @spec log(argument :: Series.t()) :: Series.t() |
There was a problem hiding this comment.
Instead of having a default value, I think we should really call log/1 in the backend. This will help us avoid precision issues if we support multiple float types in the future and I believe the natural log may be more optimized. For example, it seems Rust implements all other logs are implemented on top of ln: https://doc.rust-lang.org/src/std/f64.rs.html#428-430
May close #548
This version only accepts
log(series, number), since Polars implements the equivalent for series.Do you think we need
log(series, series)andlog(number, series)? If so, I can implement in upcoming PRs.Also, I don't know much about
exponential, so please help me documenting that.cc/ @msluszniak @kimjoaoun