Shoeboxes displayed wrong for rotated detector models

[phyy-nx]

Generally we have panels close to 90 degrees so It's not too surprising we haven't noticed this. Here are two examples:

Here I used rotate_detector.py (see #1182) to rotate dials_regression/image_examples/ALS_831/q315r_lyso_001.img 90 degrees, then ran dials.find_spots rotated.expt, dials.index rotated.expt strong.refl, and dials.image_viewer indexed.*. Interestingly, because the detector is not constrained in dials.index by design, the detector Z drifted ~20 degrees-ish. You can see the pixels are marked correctly, but the centroid cross is a bit funny and the red shoebox is wrong.

Another example:

Here I used rotate_detector.py on dials_regression/image_examples/LCLS_cspad_nexus/idx-20130301060858401.cbf, but changed the angle to 45 degrees, then used dials.find_spots rotated.expt gain=25 filter.min_spot_size=2 and dials.image_viewer rotated.expt strong.refl. Now you can see the centroid cross and shoebox are both flattened.

Note, this bug is in the master now, but can only be seen on multi-panel images until #1182 is merged.

[dagewa]

Interestingly, because the detector is not constrained in dials.index by design, the detector Z drifted ~20 degrees-ish

Could this be because the detector was rotated, but the rotation axis was not?

[phyy-nx]

Interestingly, because the detector is not constrained in dials.index by design, the detector Z drifted ~20 degrees-ish

Could this be because the detector was rotated, but the rotation axis was not?

You mean the goniometer axis? Likely! Not relevant to the second case of course which is a still, but it's a good point for the single image sweep.

[dagewa]

This issue seems to be because the shoebox display only takes the coordinates of diametric corners and draws a rectangle between them.

dials/util/image_viewer/spotfinder_frame.py

Lines 1387 to 1388 in 2d515a3

	x0_, y0_ = map_coords(x0, y0, panel)
	x1_, y1_ = map_coords(x1, y1, panel)

[dagewa]

With the fix in 62dab08, the original example:

dials.import $(libtbx.find_in_repositories dials_regression)/image_examples/ALS_831/q315r_lyso_001.img
dials.python $(libtbx.find_in_repositories dials)/util/image_viewer/slip_viewer/rotate_detector.py imported.expt
dials.find_spots rotated.expt
dials.index rotated.expt strong.refl
dials.image_viewer indexed.*

now gives this:

[phyy-nx]

oooooooo