Adding more float constants for when
https://github.com/rust-lang/rust/issues/103883 is accepted and merged.
And fixing a small conflation in the description of the Euler number.
Please take a look and let me know if I've missed or screwed up
anything.
# Description
I copied the `math ln` command and replaced the relevant parts to
implement `math exp`.
# User-Facing Changes
The `math exp` command was added. Now one can do `[1, 2, 3] | math exp`
to get e to the power of these numbers.
# Tests + Formatting
I only wrote example tests, same as for `math ln`, which also does not
have special tests. I have ran into an issue with the tests but it seems
completely unrelated (see #8687)
# After Submitting
This PR was done in order to make the documentation complete, so I'm not
adding any documentation except `math ln`.
Reasoning:
Most missing math commands are implemented with #7258.
The `meval` crate itself declares that it doesn't strive to stringent
standards (https://docs.rs/meval/latest/meval/#related-projects).
For example no particular special casing or transformations are
performed to ensure numerical stability. It uses the same rust `std`
library functions we use or have access to (and `f64`).
While the command call syntax in nushell may be a bit more verbose,
having a single source of truth and common commands is beneficial.
Furthermore the `math` commands can themselves implement broadcasting
over lists (or table columns).
Closes#7073
Removed dependencies:
- `meval`
- `nom 1.2.4` (duplicate)
User-Facing Changes:
Scripts using `math eval` will break.
We remove a further `eval` like behavior to get results through runtime evaluation (albeit limited in scope)
Tests:
- Updated tests that internally used `math eval`.
- Removed one test that primarily used `math eval` to obtain a result from `str join`
Adds new command `math log` that takes as a required positional argument
a base.
Specialized for `math log 2` and `math log 10` for better performance
and precision that matches the expectations there. This leads to
discontinuities in numerical error but should make a better trade-off
for common usecases.
Example testing of the happy path
