Zachary Moon

This is that section of the code for reference: https://github.com/Unidata/netcdf4-python/blob/0ad67b05358d12e7223235686ec694d70fca681b/netCDF4/utils.py#L67:L87

@fraserwg what do you think about this?

I added a binary rounder (following @milankl) and a significant digit in base-10 rounder (see [here](https://github.com/zmoon92/elePyant/blob/round-bin/elePyant/core.py) in my branch). I did some testing of the impacts on size/fidelity ([notebook](https://nbviewer.jupyter.org/gist/zmoon92/6c1a6ebe81713781c47c22fd603e6d55)). Interesting...

> Why do you need to convert to an `np.array` here? For the `shavemask` lines I used `np.array` just to be able to get an unsigned integer (in order to...

Here is [one benchmark example](https://gist.github.com/zmoon92/ddfa99b710f7c6d5611993003a3f36db) running on one thread with an array that is big but fits fine in RAM.

Ok, thanks @milankl, I think I fixed this in my binary rounder by [changing the return](https://github.com/zmoon92/elePyant/commit/629dcdf7345dde50bfd82a3bc4f30273ec70842e). According to line-by-line memory profiling with `%mprun` the memory increment is now almost 0...

Yeah, although much of NumPy is written in C and compiled ahead of time, my function using NumPy's array and operations doesn't get the same treatment. I [tried Numba](https://gist.github.com/zmoon92/161b61b6a263e1b0de4bce32b97b6709) and...

Could it be shown without the square brackets if there is only one type choice?

> Choices are always in brackets regardless of number. I noticed that, was merely making a suggestion. In my personal opinion, I feel for type choices (not literal choices), it...

I noticed that currently there is . Following installing that with ``` pip install ipyflow --no-deps ``` I was able to get the "Python 3 (ipyflow)" kernel option.