ESROGUE-684 - Frame.getNumpy now allows user defined data type

[bengineerd]

Description

Changes to Frame.getNumpy()

The Frame.getNumpy() method now takes an additional dtype argument, which allows a numpy type to be specified for the returned array.

array = frame.getNumpy(0, frame.getPayload(), np.uint32)

Additionally, the getNumpy() method now has default arguments for all parameters

offset = 0
count = 0, 
dtype = np.uint8

Allowing for calls such as

# Read the entire frame into a np.uint8 array
array = frame.getNumpy()

# Read the entire frame into a np.uint32 array
array = frame.getNumpy(dtype=np.uint32)

The count argument defaults to 0, which functionally will return the entire array starting at offset.

The offset and count parameters are still specified in bytes, not in the dtype size.

Other changes

The offset argument of Frame.readPy() and Frame.writePy() has also been given a default of 0 when called from python.

A new Frame.getBa() method has been added that will allocate the bytearray internally and return it. This makes things a bit more concise in python when reading from a Frame.

# New
ba = frame.getBa() # Create a bytearray, fill it, and return it

#Old
ba = bytearray(frame.getPayload())
frame.read(ba)

The getBa() method takes an offset and count which function the same as in getNumpy() with the same defaults.

Also, a new Frame.getMemoryview() method has been added that allocates a bytearray but returns it as a memoryview. This allows for efficient slicing of the frame data without any copying. In most cases where copyless slicing is needed, a getNumpy() would be preferred, but getMemoryview() is useful for efficiently parsing complex header structures.

memory_view = frame.getMemoryview()

# Parse the header (4 bytes), length (2 bytes), and checksum (2 bytes)
header, length, checksum = struct.unpack_from('IHH', memory_view, 0)