Stacking results fails with odd number of frames #58
Description
I've got a trajectory (7367 frames) that was split across four CPUs, resulting in partial trajectories with 3x 1842 and 1x 1841 frames. Running np.hstack(self._results) does not work, it should either be np.vstack or np.concatenate.
Lines 102 to 103 in 723d021
Expected behaviour
Results are concatenated over the frame axis.
Actual behaviour
ValueError is raised.
Code to reproduce the behaviour
import numpy as np
data = np.zeros([7367, 10])
print(data.shape)
# (7367, 10)
data_frames = np.array_split(data, 4)
print([frames.shape[0] for frames in data_frames])
# [1842, 1842, 1842, 1841]
np.hstack(data_frames)
# [...]
# ValueError: all the input array dimensions except for the concatenation axis must match exactly
np.vstack(data_frames)
# array([[0., 0., 0., ..., 0., 0., 0.],
# [0., 0., 0., ..., 0., 0., 0.],
# [0., 0., 0., ..., 0., 0., 0.],
# ...,
# [0., 0., 0., ..., 0., 0., 0.],
# [0., 0., 0., ..., 0., 0., 0.],
# [0., 0., 0., ..., 0., 0., 0.]])Currently version of MDAnalysis:
(run python -c "import MDAnalysis as mda; print(mda.__version__)")
0.18.0
(run python -c "import pmda; print(pmda.__version__)")
0.1.1
(run python -c "import dask; print(dask.__version__)")
0.19.1