Clustering ensemble of in memory universes

[fluver]

Expected behavior

Given a list of in-memory universe objects c = MDAnalysis.analysis.encore.clustering.cluster.cluster(us) should return a ClustersCollection.

Actual behavior

Call fails with
TypeError: Cannot determine file format for file 'None'. You can set the format explicitly with 'Universe(..., format=FORMAT)'.

Code to reproduce the behavior

def create_subverse(pdb, xtc, selection):
    big_u = load_trj(pdb, xtc)
    goi = big_u.select_atoms(selection)
    subu = mda.Merge(goi).load_new(
                AnalysisFromFunction(lambda ag: ag.positions.copy(),
                                    goi).run().results,
                format=MemoryReader)
    return(subu)

us = [create_subverse(pdb='../small.pdb', xtc=f'../windows/{r}/trj.cat.xtc', selection='segid A and backbone') for r in range(3)]
c = cluster(us)

Possible fix
merge_universes in MDAnalysis/analysis/encore/utils.py does not properly check if the filename property of the first universe in the list is set, the following code does.

if universes[0].filename:
        return mda.Universe(
            universes[0].filename,
            np.concatenate(tuple([e.trajectory.timeseries(order='fac') for e in universes]),
                        axis=0),
            format=MemoryReader)

elif universes[0].trajectory.format == 'MEMORY' and universes[0].filename == None:
        trj = np.concatenate([e.trajectory.timeseries(order='fac') for e in universes])
        o = mda.Merge(universes[0].atoms)
        o.load_new(trj, format=MemoryReader)
        return o

Alternative:

return mda.Merge(
    universes[0].atoms
    ).load_new(
        np.concatenate(
            [e.trajectory.timeseries(order='fac') for e in universes]), 
            format=MemoryReader
            )

Currently version of MDAnalysis

• Which version are you using? (run python -c "import MDAnalysis as mda; print(mda.__version__)") -> 0.20.1
• Which version of Python (python -V)? -> Python 3.6.7
• Which operating system? -> Debian 8