Issue on get_azimuthal_integral for MIB data format

[AZangiabadi]

Hi,
I noticed Pyxem versions 0.18.0 and 0.19.0 are not able to process u2 dtype during the azimuthal integration in example 7. Version 0.17.0 doesn't have this error.
Link to the raw mib data: Nano Diamond Diffraction Pattern
See below for more details:
calib = 0.00785
integration = dp.get_azimuthal_integral1d(npt=512)
integration2d = dp.get_azimuthal_integral2d(npt=512)

[CSSFrancis]

@AZangiabadi Thanks for the bug report! The azimuthal integration code changed from 0.17.0 to 0.18.0 so we are using some internal code. Unfortunately that introduced some errors appearently.

The issue is the same as hyperspy/hyperspy#2668 or numba/numba#2243.

For now you should be able to do s.data = s.data.astype(np.unit16) I think which should switch the endianess to little endian.

[CSSFrancis]

Just a further note that you should look at https://pyxem.readthedocs.io/en/stable/examples/processing/azimuthal_integration.html#sphx-glr-examples-processing-azimuthal-integration-py for more information about how to use the new azimuthal integration :)

[AZangiabadi]

For now you should be able to do s.data = s.data.astype(np.unit16) I think which should switch the endianess to little endian.

Thank You @CSSFrancis !
the uint16 twists the data in a strange way. I am guessing the header information maybe is getting lost.
This is how it looks like after azimuthal integration.
For now I'll use the v0.17.0.

Also thank you for referring me to the new user guide.

[CSSFrancis]

@AZangiabadi I think that the distortion is actually a result of using the .ai object which is deprecated (although maybe the warning isn't showing up?).

Following the user guide should help you calibrate the dataset in a way such that the azimuthal integral gives you what you would expect.