H Bond count_by_time gives a dtype error when analysis was run with a FrameIterator

[capritchard]

Hi MD Analysts,
I have conducted an H bond analysis on a part of my trajectory using a FrameIterator to select the frames I wanted:

frames_to_use = []
for ts in u.trajectory[0:1819:1]:
	frames_to_use.append(u.trajectory.frame)
for ts in u.trajectory[7000:7029:1]:
	frames_to_use.append(u.trajectory.frame)
hbonds_1 = HydrogenBondAnalysis(universe=u, donors_sel=None,  hydrogens_sel='name *H* and around 1.1 name N* O*', acceptors_sel='protein and name N* O*')
hbonds_1.run(frames=frames_to_use)
plt.plot(hbonds_1.times, hbonds_1.count_by_time())

This produces the error:

    plt.plot(hbonds_1.times, hbonds_1.count_by_time())
                             ^^^^^^^^^^^^^^^^^^^^^^^^
  File "/opt/conda/envs/mdanalysis-tools/lib/python3.11/site-packages/MDAnalysis/analysis/hydrogenbonds/hbond_analysis.py", line 901, in count_by_time
    indices -= self.start
numpy.core._exceptions._UFuncOutputCastingError: Cannot cast ufunc 'subtract' output from dtype('O') to dtype('float64') with casting rule 'same_kind'

I do not get this error if I run the analysis on the entire trajectory [ie hbonds_1.run()], so I guess the issue comes from using a FrameIterator. Does anyone have any ideas on how to fix this? Thanks!

[orbeckst]

What you're proposing should have work (at least give our docs for run()). This looks like a bug. Can you please raise an issue in the issue tracker?

Ideally include an example that makes use of the test files, e.g., using the test files

from MDAnalysisTests.datafiles import waterPSF, waterDCD
u = mda.Universe(waterPSF, waterDCD)