Using pathlib Path object

[mimischi]

Expected behavior

Providing a PosixPath from pathlib to mda.Universe should create the Universe. Instead, we currently have to wrap the PosixPath object in a str() call.

Actual behavior

Code throws an AttributeError:

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-154-4c04e03d6c7d> in <module>
      2 gro = base_path / "ref.gro"
      3 xtc = base_path / "trjcat_p1p10.mc.xtc"
----> 4 u = mda.Universe(gro, xtc)

~/.conda/envs/mda020/lib/python3.7/site-packages/MDAnalysis/core/universe.py in __init__(self, *args, **kwargs)
    335                 coordinatefile = None
    336 
--> 337             self.load_new(coordinatefile, **kwargs)
    338             # parse transformations
    339             trans_arg = kwargs.pop('transformations', None)

~/.conda/envs/mda020/lib/python3.7/site-packages/MDAnalysis/core/universe.py in load_new(self, filename, format, in_memory, **kwargs)
    602         kwargs['n_atoms'] = self.atoms.n_atoms
    603 
--> 604         self.trajectory = reader(filename, **kwargs)
    605         if self.trajectory.n_atoms != len(self.atoms):
    606             raise ValueError("The topology and {form} trajectory files don't"

~/.conda/envs/mda020/lib/python3.7/site-packages/MDAnalysis/coordinates/XDR.py in __init__(self, filename, convert_units, sub, refresh_offsets, **kwargs)
    134                                             convert_units=convert_units,
    135                                             **kwargs)
--> 136         self._xdr = self._file(self.filename)
    137 
    138         self._sub = sub

MDAnalysis/lib/formats/libmdaxdr.pyx in MDAnalysis.lib.formats.libmdaxdr._XDRFile.__cinit__()

AttributeError: 'PosixPath' object has no attribute 'encode'

Code to reproduce the behavior

from pathlib import Path 
import MDAnalysis as mda

base_path = Path(".")
gro = base_path / "example.gro"

u = mda.Universe(gro)

Currently version of MDAnalysis

• Which version are you using? 0.20.1
• Which version of Python (python -V)? 3.7.3
• Which operating system? Scientific Linux

[orbeckst]

Is pathlib in the standard library or do we install it as part of our other dependencies?

[mattwthompson]

It's in the standard library 3.4+

In principle it means os.path should never happen, although that's easier said than done

[orbeckst]

And in Python 2?

This version of MDA still supports 2.

[mattwthompson]

It's not in the standard library there, but it can still be installed. I just tested that it can be installed via pip and conda in a fresh 2.7 environment and it installed and imported fine. It can also be installed as a separate packing with Python 3 (although that seems a little silly). I don't know how you all deal with these 2/3 conflicts, but I would guess an "if Python < 3: conda/pip install pathlib" approach could work

[richardjgowers]

In python 3.4+ then python os functions can take Pathlib and strings interchangeably. The quoted module however (and probably a few others) is a C extension and so isn't opening things via python. So we need to go through our bindings to C extensions (DCD, XDR, MMTF to name a few) and check that they seamlessly handle Pathlib objects (which seems to be the new "Pythonic" way to deal with paths instead of strings).

…

On Mon, 3 Feb 2020 at 19:47, Oliver Beckstein ***@***.***> wrote: And in Python 2? This version of MDA still supports 2. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#2497?email_source=notifications&email_token=ACGSGB5WFXKBNFVWX5VT4HTRBBYEFA5CNFSM4KPCL6T2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEKVEXYI#issuecomment-581585889>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACGSGBY24ULZ6UCKSIRXDPLRBBYEFANCNFSM4KPCL6TQ> .