density_from_Universe does not work with the MemoryReader

[jbarnoud]

Expected behaviour

density_from_Universe creates a Density object from an Universe that use the MemoryReader.

Actual behaviour

density_from_Universe calculate the density, but fails to create the resulting object as it is missing a filename to write as metadata.

Code to reproduce the behaviour

import MDAnalysis as mda
from MDAnalysisTests.datafiles import TPR, XTC
import MDAnalysis.analysis.density as mad

u = mda.Universe(TPR, XTC)
u.transfer_to_memory()
mad.density_from_Universe(u, atomselection='protein')

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-7-6400dc1944f6> in <module>()
      5 u = mda.Universe(TPR, XTC)
      6 u.transfer_to_memory()
----> 7 mad.density_from_Universe(u, atomselection='protein')

/home/jon/dev/mdanalysis/package/MDAnalysis/analysis/density.pyc in density_from_Universe(universe, delta, atomselection, start, stop, step, metadata, padding, cutoff, soluteselection, use_kdtree, update_selection, verbose, interval, quiet, parameters)
    636     metadata = metadata if metadata is not None else {}
    637     metadata['psf'] = u.filename
--> 638     metadata['dcd'] = u.trajectory.filename
    639     metadata['atomselection'] = atomselection
    640     metadata['n_frames'] = n_frames

AttributeError: 'MemoryReader' object has no attribute 'filename'

Currently version of MDAnalysis:

(run python -c "import MDAnalysis as mda; print(mda.__version__)")
'0.16.0-dev0'

[richardjgowers]

I think MemoryReader.filename came up before, because things like this might happen. Not sure what our Reader API states exactly, but maybe MemoryReader should have a default value of something like MEMORY

[kain88-de]

See #1027

[orbeckst]

The quick fix, however, is to make the metadata creation more fault tolerant (or decide to scrap its autogeneration and leave it to usercode to provide the metadata dict). The reason why it's there in the first place was because it can be rather difficult to track provenance of densities; metadata was a simple way to maintain context.

p.s.: bonus points to @jbarnoud for the acronym for importing MDAnalysis.analysis.density ;-)

[jbarnoud]

Fixed by #1252.