Generalize beam shift deflector and stage rotating speed interface

[Baharis]

Context

Below I would like to suggest some small changes to the interface that may warrant an adequately short discussion:

Firstly, the Deflector class, parent for BeamShift, BeamTilt, DiffShift etc. is type-hinted to accept and return only instances of int. However, FEI microscopes do not store these as integers, but rather floats between -1 and 1 exclusive. This carries major implications because some functions, e.g. CalibBeamShift.pixelcoord_to_beamshift explicitly map its result to int using beamshift.astype(int). rendering use on FEI machine impossible. So here I suggest modifying microscope/component/deflectors:Deflector base class hints to suggest that these values can be of either numeric type.

Secondly, although following the above CalibBeamShift should theoretically return shifts as int, it does not. The return type of pixelcoord_to_beamshift and the setting methods use as a container np.ndarray full of np.int. Consequently, if your server does not have numpy installed, np.int is an unknown type. This can be easily fixed by mapping all set values to float: ctrl.beamshift.set(*(float(b) for b in beamshift)). This completely fixes the issue for FEI, and should be benign for Jeol since its interface converts it back to int anyway, and the precision of converting int to float and back is likely good enough:

>>> a = 123456789123456789123456789
>>> abs(a - int(float(a))) / a
1.793587896398062e-17

Thirdly, the Stage class currently has two (unused) methods to work with stage speed: the set_rotation_speed setter and the rotating_speed context. I suggest adding method get_rotation_speed - otherwise, one can get it only via a camel case parameter of a private attribute. I also suggest renaming rotating_speed to rotation_speed so that it matches the other two. I could not find this context used anywhere outside my own code, so I doubt anyone will mind.

Finally, I have introduced here some QOL improvements: made imports absolute (otherwise calibration could not be run as script, plus relative imports are good only in dynamically developed packages, link), added or improved type hinting, made it so that you don't need to manually make a folder to save a tiff file during calibration.

Minor change

• CalibBeamShift: allow saving tiff in a new directory so that calibration can be run without manually creating the output directory for collected tiff files each time.

Bugfix

• CalibBeamShift: some methods return or set float instead of np.int type, so that you can use this calibration of FEI;
• calibrate_beamshift_live: can now be run as a script thanks to making imports absolute;
• calibrate_beamshift_live: fix typo so that calibration files are not all saved under the same name calib_beamshift_{i.tiff.

Code maintanance

• Deflector child classes are now properly type hinted to handle either int or float type values.

[Baharis]

@viljarjf Tagging you because you are the author of the MicroscopeBase that propagates the information that getBeamShift and setBeamShift operate on ints; these use float on FEI machines. I modified only hints for these two there, maybe I should actually generalize all of these. Also AFAIK you are an active Jeol machine user.

[stefsmeets]

Looks good to me, I left a minor comment.

[stefsmeets]

Suggested change

	def get_rotation_speed(self) -> Union[float, int]:
	def get_rotation_speed(self) -> Number:

[Baharis]

Can do though it will require adding declaration there as well. In the other place it was more needed since this patter repeated multiple times. 👍

[Baharis]

Alright, done. If that is the case and there are no further comments, I will merge this tomorrow.