Fix dials.cosym target fail bug

[jbeilstenedmands]

In dials.cosym, I found a case where a low resolution cutoff in the resolution filter caused at least one full dataset to be removed, leading to a stacktrace during the minimisation at:

>       assert (x.size // self.dim) == (len(self._lattices) * len(self.sym_ops))
E       AssertionError: Please report this error to dials-support@lists.sourceforge.net:

/dials/modules/dials/algorithms/symmetry/cosym/target.py:271: AssertionError

Of course a suitable d_min could just be set, but I still thought it would be good to avoid crashing out and handle this more gracefully.

I have added a test to reproduce this, and the fix is fairly straightforward, just making sure to use the correct number of datasets in the procedure (which could be reduced compared to upon initialisation). In the example test, only 7/8 datasets remain after the resolution filter, so only 7 reindexing operators can be determined. The only outstanding question is whether the filtered-out dataset should remain in the output refl, expt files. As no reindexing operator can be determined, I don't believe they should, so this proposed changeset removes them as they may not be consistently indexed.

p.s. I updated all the tests to use pathlib objects

[codecov]

Codecov Report

Merging #1993 (0824621) into main (92daf13) will decrease coverage by 0.00%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main    #1993      +/-   ##
==========================================
- Coverage   68.82%   68.82%   -0.01%     
==========================================
  Files         649      649              
  Lines       75294    75318      +24     
  Branches    10782    10787       +5     
==========================================
+ Hits        51824    51835      +11     
- Misses      21446    21458      +12     
- Partials     2024     2025       +1     

[rjgildea]

Thanks for the bug fix and moving tests to use pathlib.Path. A few very minor comments about the pathlib usage, but otherwise looks good to me 👍