[Feature Addition]: NumPy (Python) datatype for posit

[anubane]

I have created a wrapper library to utilize the Posit datatype defined by this project as a custom datatype in NumPy (Python).

In this example file you may observe:

from posit8_2 import posit8_2 as p8    #importing the datatype library in Python

Then the datatype is used as follows:

a = np.array([0.33, 0.25, 0.025, 0.0025, 0.00025], dtype=np.float32)

b = a.astype(p8)    # implicit cast from NumPy float32 to Posit<8,2>

Current problems with the library:

1. I use an old build procedure
2. The current vesion is specifically for Posit<8,2>, will need help though, to make it generic

Notwithstanding the above, the library is ready to use, once built (using the present method).

I wanted to inquire the following:

1. If this is useful for the community
2. Whether stillwater would want to add this officially

[Ravenwater]

@anubane Thank you for your contribution. We have been looking for a Numpy integration for half a decade, so this work is much appreciated.

We have been looking at cppyy as a mechanism to integrate the full parameterization of the C++ types.

Do you have a reviewer you can suggest to evaluate your approach in the context of upstream integration into the Python ecosystem?

[anubane]

@Ravenwater, I just wish to place on record that the data type wrapper that I have prepared will work only within the NumPy library for Python. It will not provide native data type support in Python itself.

That being said, I am not sure whom may I recommend to review my work. I was hoping that someone from stillwater will be able to handle that!