Mask convention: 0 or 1 for masked value

[t20100]

Not sure we can do anything in silx, but I still open it for discussion. Apparently bliss is using the opposite convention that is used in silx regarding masks. This forces users to invert the silx mask to be used.

[kif]

numpy.logical_not is their friend then. We can need multiple||overlapping masks

[vallsv]

That's about GUI and file. In fact people dont know anything about the way it is stored, nor read.

And Lima uses the reverse of silx.

Is that normalized in Nexus?

[vallsv]

BTW, thinking out loud, as it's about EDF mask, we could write the way the mask have to be read in the header.

So what Lima could read silx files properly. I can take care of it.

What do you think?

[vasole]

If an array denotes a mask, typically 0 is not masked and any other value allows to specify different levels of masks.

[vasole]

If 0 is masked, then you are not in a good situation.

[vasole]

Are you confusing mask with pixel_mask_applied?

https://manual.nexusformat.org/classes/base_classes/NXdetector.html#nxdetector

[vasole]

To me is crystal clear that 0 is a non-masked value. Dectris is following NeXus too. That should settle the question.

[vallsv]

You remember me this pixel_mask specification.

My proposal when we save mask as EDF is only to write something like source=silx or source=silx-mask in the header (or application= whatever). To me it is enough to know how to deal with the file data when it have to be read again. We don't have to change our workflow.

[vallsv]

I let you close it if you think #3310 is enough.

[vallsv]

Here is a MR on Lima to be able to properly read silx 0.14 masks without doing anything more: https://gitlab.esrf.fr/limagroup/Lima-tango-python/-/merge_requests/58

[t20100]

closed by PR #3310 on branch 0.14