Allow dials.predict to work without images

[dagewa]

If we just want to predict for some geometry, without having image data, we need check_format=False

[codecov]

Codecov Report

Merging #2553 (70a5482) into main (c6bf4cd) will increase coverage by 0.10%.
Report is 1 commits behind head on main.
The diff coverage is n/a.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2553      +/-   ##
==========================================
+ Coverage   78.49%   78.59%   +0.10%     
==========================================
  Files         609      609              
  Lines       74853    75385     +532     
  Branches    10674    10816     +142     
==========================================
+ Hits        58753    59250     +497     
- Misses      13937    13956      +19     
- Partials     2163     2179      +16     

[dagewa]

Actually setting check_format=False breaks the (untested) ignore_shadows=False option, added for #349. In 863044f it was noted that we really need a way to set check_format=auto (and the relevant issue is tracked here cctbx/dxtbx#187).

There is currently no good way to do this other than having a fallback parser

[rjgildea]

There’s a pattern/workaround used in dials.show to reinstantiate the experiment object with check_format=True when required for options that need access to the image - similar could work here?

dials/src/dials/command_line/show.py

Line 322 in 189fe5a

experiments.as_json(), check_format=True

[dagewa]

Oh my, there's a feeling of déjà vu... Yes, I'll try this. Thanks!

[dagewa]

@graeme-winter added you for review as this touches on the shadow filtering stuff that you added. I think this needs a test case. Can you suggest a good one? (otherwise I guess I can use almost anything just to exercise the mechanics of it)

[dagewa]

@graeme-winter I tried testing this with any old data set:

dials.predict $(dials.data get -q centroid_test_data)/indexed.expt ignore_shadows=False
The following parameters have been modified:

ignore_shadows = False
input {
  experiments = /home/fcx32934/sw/cctbx/build/dials_data/centroid_test_data/indexed.expt
}

Traceback (most recent call last):
  File "/home/fcx32934/sw/cctbx/build/../modules/dials/src/dials/command_line/predict.py", line 141, in <module>
    run()
  File "/home/fcx32934/sw/cctbx/conda_base/lib/python3.10/contextlib.py", line 79, in inner
    return func(*args, **kwds)
  File "/home/fcx32934/sw/cctbx/build/../modules/dials/src/dials/command_line/predict.py", line 137, in run
    script.run(args)
  File "/home/fcx32934/sw/cctbx/build/../modules/dials/src/dials/command_line/predict.py", line 118, in run
    shadowed = filter_shadowed_reflections(
  File "/home/fcx32934/sw/cctbx/modules/dials/src/dials/algorithms/shadowing/filter.py", line 19, in filter_shadowed_reflections
    shadow = masker.project_extrema(
AttributeError: Please report this error to dials-support@lists.sourceforge.net: 'NoneType' object has no attribute 'project_extrema'

Seems to me that the shadowing stuff you added is broken? This error is the same on main.

[graeme-winter]

Looks like a sensible change set, thank you, highlighted one area for API improvement for ExperimentList but that can wait for another day

[graeme-winter]

This makes me think we need a better way in the API of making this change

Not suggesting that we change it here just that the ExperimentList needs to have an add_format() or something

[dagewa]

Thanks @graeme-winter. I don't need this merged urgently, so I wonder if we should look at the failures with ignore_shadows=False (#2553 (comment)) in this PR, or spin it out as a separate issue?