BUG: Compatibility with numpy 1.16.x

[pekof]

comtypes not working with numpy 1.16.x due to PR numpy/numpy#12769

automation.py and safearray.py reference numpy.ctypeslib._typecodes which is removed in that PR

[mobiusklein]

This appears to be solvable in a non-ideal fashion by replacing numpy.ctypeslib._typecodes with numpy.ctypeslib._scalar_type_map.

The optimal solution would be to build the mapping from ctypes type to numpy.dtype using the public API. Something along the lines of:

import numpy as np
from numpy.ctypeslib import as_ctypes_type

ctypes_to_dtypes = {}

for tp in set(np.sctypeDict.values()):
    try:
        ctype_for = as_ctypes_type(tp)
        ctypes_to_dtypes[ctype_for] = tp
    except NotImplementedError:
        continue

This yields:

{ctypes.c_long: numpy.int32,
 ctypes.c_ushort: numpy.uint16,
 ctypes.c_ulonglong: numpy.uint64,
 ctypes.c_longlong: numpy.int64,
 ctypes.c_ubyte: numpy.uint8,
 ctypes.c_byte: numpy.int8,
 ctypes.c_double: numpy.float64,
 ctypes.c_float: numpy.float32,
 ctypes.c_short: numpy.int16,
 ctypes.c_bool: numpy.bool_,
 ctypes.c_ulong: numpy.uint32}

Were there other types supported previously that are missing from this?

[fockez]

May I ask, when will this fix be merged?

[mobiusklein]

If there are no other types to add support for, I can write a short patch to add this behavior. At the time I posted this solution, I thought that the repository had been abandoned.

[fockez]

I thought too, but I can still see the new mergers although no new release for a long time.
The other problem is that there is no replacement.