Spot-count histogram: use math.ceil instead of round

[rjgildea]

Python uses a rounding half to even strategy, which means that 1.5 and 2.5 both round to 2, which resulted inconsistency in the spot-count depending on whether there was an even or odd number of images.

[graeme-winter]

🧐 It was that easy?

I spent a while looking at this in the past, was evidently looking in ${WRONG-PLACE} 🤬

[graeme-winter]

Thanks for fixing this - it was one of the more annoying "features" and I am genuinely shocked that the fix was this simple, but it presumably was so that's great!

[graeme-winter]

Is it worth having this test on e.g. 1...9 images of centroid test data to make sure that it works right for all the cases?

[rjgildea]

The bug here was that the behaviour was inconsistent between even- and odd-length imagesets. We already had a test for output of even-length imagesets (1..10 above). I have added a test for an odd-length imageset below (1..3). We can change that to 1..9 if you think that would be better, but I don't think it would make any difference to the outcome.

[rjgildea]

You can see the old behaviour by doing:

$ dials.import $(dials.data get -q x4wide)/X4_wide_M1S4_2_*.cbf

$ dials.find_spots imported.expt scan_range=1,2
...
Histogram of per-image spot count for imageset 0:
738 spots found on 2 images (max 396 / bin)
* 
**
**
**
**
**
**
**
**
**
12

$ dials.find_spots imported.expt scan_range=1,3
...
Histogram of per-image spot count for imageset 0:
958 spots found on 2 images (max 514 / bin)
* 
**
**
**
**
**
**
**
**
**
12

$ dials.find_spots imported.expt scan_range=1,4
...
Histogram of per-image spot count for imageset 0:
1190 spots found on 4 images (max 433 / bin)
*  *
*  *
*  *
*  *
*  *
*  *
** *
****
****
****
1  4

$ dials.find_spots imported.expt scan_range=1,5
...
Histogram of per-image spot count for imageset 0:
1449 spots found on 4 images (max 515 / bin)
*  *
*  *
*  *
*  *
*  *
*  *
****
****
****
****
1  4

With the changes in this PR, the behaviour is now as expected.

[codecov]

Codecov Report

Merging #1660 (5c6750f) into main (e5213f4) will decrease coverage by 0.00%.
The diff coverage is 42.85%.

❗ Current head 5c6750f differs from pull request most recent head fdb5067. Consider uploading reports for the commit fdb5067 to get more accurate results

@@            Coverage Diff             @@
##             main    #1660      +/-   ##
==========================================
- Coverage   66.70%   66.69%   -0.01%     
==========================================
  Files         616      616              
  Lines       69057    69062       +5     
  Branches     9606     9607       +1     
==========================================
- Hits        46063    46062       -1     
- Misses      21060    21065       +5     
- Partials     1934     1935       +1