Milan Klöwer

Thanks Aaron, that looks awesome. Good to see that all these information patches in the exponent bits disappear! For `dissicos` could you convert your arrays to [signed exponent bits](https://milankl.github.io/BitInformation.jl/dev/transformations/#Signed-exponent)? It...

Perfect, that looks really good to me. Now I'd keep these mantissa bits and then redo any analysis that you'd otherwise do with the full precision data set. Feel free...

Use [NetCDF.jl](https://github.com/JuliaGeo/NetCDF.jl), but switch lossless compression on, like [here](https://juliageo.org/NetCDF.jl/dev/intermediate/#Creating-files). If you then pass on arrays with rounded mantissas via `round(::Array,keepbits::Int)` from BitInformation.jl you'll get it accordingly compressed. Alternatively, use [nco](https://github.com/nco/nco)...

Branch #main now contains `bitinformation(::AbstractArray{T};masked_value::T) where T` so that you should be able to do directly ```julia julia> using BitInformation julia> A = rand(Float32,3,3); julia> round!(A,1) 3×3 Matrix{Float32}: 0.5 0.1875...

v0.5 got merged into the Julia registry now. Will leave this issue open to eventually pick some of the posts here for the documentation.

> https://gist.github.com/aaronspring/383dbbfe31baa4618c5b0dbef4f6d574 Many thanks for sharing this. Makes me very happy to see that (once we've sorted out the masked array issue, which I wanted to do for a long...

> But I think the potential for bitrounding is enormous. I agree 😄 Especially once netcdf supports more lossless codecs, I believe zstd is soon-ish going to be supported (discussion...

```julia julia> Posit8(1e8) Posit8(1.6777216e7) julia> Posit8(1e9) NaR julia> Posit8(1e10) Posit8(1.6777216e7) ``` So there's a "bad spot" where the conversion of x>maxpos is to NaR instead of to maxpos. Beyond that...

Turns out only a small change was needed to correct for this. The only test that's currently not passing is ```julia julia> Posit32(Inf32) Posit32(0.0) julia> Posit32(-Inf32) Posit32(0.0) julia> Posit32(NaN32) Posit32(0.0)...

Then there's apparently also this issue remaining (similar for other formats) ```julia julia> Posit8(1f-8) Posit8(9.536743e-7) julia> floatmin(Posit8) Posit8(5.9604645e-8) ``` Meaning that a float-value that should be converted to floatmin/minpos of...