Coordinates don't exist in output error - document hdf5 heirarchy

[samiransen23]

Coordinates don't exist in output error for utils/hdf_to_gro.py

python3 utils/hdf_to_gro.py sim.h5 --out traj.gro
>
Traceback (most recent call last):
  File "utils/hdf_to_gro.py", line 18, in <module>
    shape = f_hd5["coordinates"].shape[0]
...
KeyError: "Unable to open object (object 'coordinates' doesn't exist)"

sim.h5 is a short trajectory output of the sample dppc.gro in HyMD-2021/

[samiransen23]

> f_hd5 = h5py.File(args.hdf5_file, "r")
> f_hdf5.keys()
<KeysViewHDF5 ['connectivity', 'h5md', 'observables', 'parameters', 'particles']>
> particles = f_hd5.get('particles')
> particles
<HDF5 group "/particles" (1 members)>
> np.array(particles)
['all']

A 'particles'-like attribute called 'coordinates' should exist.

[samiransen23]

'coordinates' is an attribute for the input .h5 file, not for the output sim.h5
'time' exists for the output sim.h5 and not for the input .h5 of course, so I don't know why utils/hdf_to_gro.py uses both time and coordinates as attributes

[Lun4m]

Isn't there a "position" attribute inside "particles"?

[samiransen23]

Something like this: f_hd5['particles/all/position’] Working on it to make the conversion work. Now it doesn’t. Samiran Sen --------------------------------------- Hylleraas Centre for Quantum Molecular Sciences, University of Oslo, Norway

…

On 30 Mar 2021, at 11:45, Manuel Carrer ***@***.***> wrote: Isn't there a "position" attribute inside "particles"? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#54 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AOHBZUVOSFG3HIZ6J4EUFQ3TGGMSPANCNFSM42BOFY7A>.

[mortele]

Isn't there a "position" attribute inside "particles"?

Actually, the H5MD format specification states that the structure should be

particles
 \-- <group1>
      \-- box
      \-- (position)
      |    \-- step: Integer[variable]
      |    \-- time: Float[variable]
      |    \-- value: <type>[variable][N][D]

with the particle coordinates being called position and not coordinates. Not sure why this is the case in our output, probably because the old output Sigbjørn wrote had them named as such. We should change this at some point. Glancing over the VMD plugin code, it seems to also expect the coordinates in an HDF5 property called /particles/all/position.

[mortele]

Something like this: f_hd5['particles/all/position’] Working on it to make the conversion work. Now it doesn’t. Samiran Sen --------------------------------------- Hylleraas Centre for Quantum Molecular Sciences, University of Oslo, Norway
…
On 30 Mar 2021, at 11:45, Manuel Carrer @.***> wrote: Isn't there a "position" attribute inside "particles"? — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#54 (comment)>, or unsubscribe https://github.com/notifications/unsubscribe-auth/AOHBZUVOSFG3HIZ6J4EUFQ3TGGMSPANCNFSM42BOFY7A.

Not entirely sure if you fixed this yet or what you want to use it for, but #5 and #31 might be useful for functionality to convert HyMD output to HyMD input if that is your intended use.

[samiransen23]

Isn't there a "position" attribute inside "particles"?

Actually, the H5MD format specification states that the structure should be

particles
 \-- <group1>
      \-- box
      \-- (position)
      |    \-- step: Integer[variable]
      |    \-- time: Float[variable]
      |    \-- value: <type>[variable][N][D]

with the particle coordinates being called position and not coordinates. Not sure why this is the case in our output, probably because the old output Sigbjørn wrote had them named as such. We should change this at some point. Glancing over the VMD plugin code, it seems to also expect the coordinates in an HDF5 property called /particles/all/position.

Exactly...there is a nomenclature mismatch between input and output files, though both are .h5 files.

[mortele]

@samiransen23 Yes, I think we talked about matching them up in one of our previous meetings, just no one got around to actually doing it yet. It is an artifact of Sigbjørn using a different standard of trajectory output in the earlier code. H5MD is just one of several different trajectory format specifications that people have defined under the HDF5 file type. I believe the one Sigbjørn was (somewhat loosely) following was MDTraj HDF5.

[mortele]

Closing this, moving discussion about format and changing input/output to #79. Discussion about hdf_to_gro.py not working moved to #80.