Fix for missing numpy data types on some platforms

[brownbaerchen]

I had some trouble with round-off errors today and noticed that my numpy doesn't actually have quadruple precision. So np.float128 will raise an error on my machine. np.longdouble, on the other hand, is just regular double on my machine. This is extremely weird behaviour that I neither like nor understand. I switched out the quadruple precision things for the word version so I can run the IO stuff with double precision at least.
Notice that the numpy documentation says that np.longdouble is an alias for np.float128 on Linux x86_64 machines. So this fix applies to Mac and Windows users.

[tlunet]

Interesting ... I'm always a bit confused on how numpy handle datatypes. In particular there is this thing :

import numpy as np
u = np.zeros(10, dtype=np.float64)
u.dtype == np.float64  # => True
u.dtype is np.float64  # => False

The way data types are handled can probably still be improved, however I think replacing np.float128 by np.longdouble will make the FieldsIO very platform dependent, and may lead to inconsistent file sizes (in particular if on some platform, np.longdouble are simply standart double precision numbers ...). The main advantage of keeping this size-based notation is that at least, it's explicit on how much bytes are used to store a field value.

I would rather see some import dependent definition, like this :

DTYPES = {
    0: np.float64,  # double precision
    1: np.complex128,
}
try:
    DTYPES.update({
        2: np.float128,  # quadruple precision
        3: np.complex256,
    })
except AttributeError:
    # Some warning on double precision not available on this platform
# And the same for single precision ...

[brownbaerchen]

You're right, that is a better solution! I went with the warning only on debug level and added a bit more elaborate message when you try to access a datatype that is not there.

[tlunet]

Can logging not be a module import ? Local import with black formating makes this very ugly ... 🤡

[tlunet]

Looks all good, thanks !

[pancetta]

Excellent solution, thanks!