Milan Klöwer

And the issue as already descriped in the issue remains ```julia julia> Posit8(floatmin(Float32)) Posit8(0.0) julia> Posit8(10*floatmin(Float32)) Posit8(5.9604645e-8) ``` floats within a multiple of floatmin underflow instead of being converted to...

@Alexander-Barth sorry, only saw this now, yes, please zenodo for the time being and once the JOSS paper is out I'll let you know! If we're supposed to use another...

> If you want to measure it further you can time over > > https://github.com/JuliaDocs/Documenter.jl/blob/97f1a9f663f63e66d4d4d13d2cdf200fa30c5dfc/src/Expanders.jl#L37-L45 I was also facing insanely long times on `ExpandTemplates` (15min+) for something that used to...

For your information, [nco](http://nco.sourceforge.net/) has that option.

Typical compression factors I achieved with Zstandard are about 2-3x higher than with Zlib.

Interesting, yeah, I realised that netCDF is always behind the functionality that hdf5 offers. So I guess people want to standardisation of netCDF but the features of HDF5, which puts...

Technically, a proper round-to-nearest (tie-to-even) in binary only requires a few bitwise operations, you can find an example [here.](https://github.com/esowc/Elefridge.jl/blob/6f4404361fbd5e438b58d78aaa173b41f6fdca8c/src/bitrounding.jl#L15) In case you either want to implement it yourself, or two...

As a work-around you may do something like this ```python def binary_round(x,nsb): return round(x*2**nsb)/2**nsb In [57]: binary(binary_round(np.pi,7)) Out[57]: '01000000010010010000000000000000' In [58]: binary_round(np.pi,7) Out[58]: 3.140625 ``` But I doubt it works...

Also note the [pynco](https://github.com/nco/pynco) project. [NCO](https://github.com/nco/nco) has a similar lossy+lossless compression strategy, that is [explained here](http://nco.sourceforge.net/nco.html#Precision_002dPreserving-Compression)

@zmoon92 Great job, translating my Julia functions to python. Unfortunately, I don't know python well enough to understand how it handles bittypes. In my experience there is a lot of...