Remove cache of Kabsch debugging

[graeme-winter]

Was broken anyway for stills as image.index is always 0 so did not update when moving to next frame.

Fixes #1734

Removed now redundant request dict as well

[phyy-nx]

Wouldn't this re-open #1213? (@dagewa)

[graeme-winter]

Wouldn't this re-open #1213? (@dagewa)

Could be, but working and slow seems better to me than broken and quick

We know the image viewer as a whole needs speeding up 🙄

[dagewa]

Yes, it would reopen #1213. What does change when moving through stills if not image.index? It seems this could be fixed less destructively if the right item is put in the self._dispersion_debug_memo dictionary.

[dagewa]

Is there an example dataset available somewhere I can look at to demonstrate #1734?

[codecov]

Codecov Report

Merging #1761 (50284ad) into main (e9b1e2f) will increase coverage by 0.00%.
The diff coverage is 0.00%.

❗ Current head 50284ad differs from pull request most recent head d731ca1. Consider uploading reports for the commit d731ca1 to get more accurate results

@@           Coverage Diff           @@
##             main    #1761   +/-   ##
=======================================
  Coverage   67.22%   67.23%           
=======================================
  Files         616      616           
  Lines       69021    69012    -9     
  Branches     9618     9618           
=======================================
  Hits        46399    46399           
+ Misses      20680    20671    -9     
  Partials     1942     1942           

[graeme-winter]

Yes, it would reopen #1213. What does change when moving through stills if not image.index? It seems this could be fixed less destructively if the right item is put in the self._dispersion_debug_memo dictionary.

Still process imported experiments each have length 1 imageset -> index always 0, you iterate through imagesets. I will come up with a less destructive fix

[graeme-winter]

Reproduce: pick any sequence of still images, run through dials.stills_process then

dials.image_viewer *imported.expt

stills processing need not have worked as the data will still have been imported

[graeme-winter]

@phyy-nx @dagewa I have restored the fix to #1213 but restructured the code rather than aiming for a minimal change. Motivation:

• before I spent a good chunk of time trying to work out where request was sent - it was sent nowhere
• computing a hash rather than a dictionary comparison seems more idiomatic for "what was intended"

FWIW simply including the imageset pointer in whatever the fix was would have fixed the initial bug but would mean a future explorer in these woods would have had to retrace steps.

[dagewa]

This looks like a better solution. I did wonder if we could use functools.lru_cache for this though, but I'm not sure if it would work with the type passed as image to _calculate_dispersion_debug (which might be something like a "chooser_wrapper")

[graeme-winter]

@ndevenish are the build errors related to PR or some external influence? I cannot tell 🤔

[Anthchirp]

cctbx/cctbx_project#627 to blame