Skip observables contained in particle groups

package/AUTHORS

@@ -240,6 +240,7 @@ Chronological list of authors
   - Sampurna Mukherjee
   - Leon Wehrhan
   - Valerij Talagayev
+  - Fabian Zills
   - Kurt McKee
 
 

package/CHANGELOG

@@ -17,11 +17,13 @@ The rules for this file:
 ??/??/?? IAlibay, HeetVekariya, marinegor, lilyminium, RMeli,
          ljwoods2, aditya292002, pstaerk, PicoCentauri, BFedder,
          tyler.je.reddy, SampurnaM, leonwehrhan, kainszs, orionarcher,
-         yuxuanzhuang
+         yuxuanzhuang, PythonFZ
 
  * 2.8.0
 
 Fixes
+ * Do not raise an Error reading H5MD files with datasets like
+   `observables/<particle>/<property>` (part of Issue #4598, PR #4615)
  * Fix PSFParser error when encoutering string-like resids
  * (Issue #2053, Issue #4189 PR #4582)
  * Fix `MDAnalysis.analysis.align.AlignTraj` not accepting writer kwargs

package/MDAnalysis/coordinates/H5MD.py

@@ -207,6 +207,7 @@
    :members:
 
 """
+import warnings
 
 import numpy as np
 import MDAnalysis as mda
@@ -673,10 +674,25 @@
     def _copy_to_data(self):
         """assigns values to keys in data dictionary"""
 
-        if 'observables' in self._file:
-            for key in self._file['observables'].keys():
-                self.ts.data[key] = self._file['observables'][key][
-                    'value'][self._frame]
+        if "observables" in self._file:
+            for key in self._file["observables"].keys():
+                try:
+                    # if has value as subkey read directly into data
+                    if "value" in self._file["observables"][key]:
+                        self.ts.data[key] = self._file["observables"][key][
+                            "value"
+                        ][self._frame]
+                    # if value is not a subkey, read dict of subkeys
+                    else:
+                        for subkey in self._file["observables"][key].keys():
+                            self.ts.data[key + "/" + subkey] = self._file[
+                                "observables"
+                            ][key][subkey]["value"][self._frame]
+                except KeyError:
+                    warnings.warn(
+                        f"Could not read {key} from observables group. "
+                        "Possibly not a legal H5MD observable specification."
+                    )
 
         # pulls 'time' and 'step' out of first available parent group
         for name, value in self._has.items():

testsuite/MDAnalysisTests/coordinates/test_h5md.py

@@ -5,16 +5,24 @@
 import os
 import MDAnalysis as mda
 from MDAnalysis.coordinates.H5MD import HAS_H5PY
+
 if HAS_H5PY:
     import h5py
+    from MDAnalysis.coordinates.H5MD import H5MDReader
 from MDAnalysis.exceptions import NoDataError
-from MDAnalysisTests import make_Universe
-from MDAnalysisTests.datafiles import (H5MD_xvf, TPR_xvf, TRR_xvf,
-                                       COORDINATES_TOPOLOGY,
-                                       COORDINATES_H5MD)
-from MDAnalysisTests.coordinates.base import (MultiframeReaderTest,
-                                              BaseReference, BaseWriterTest,
-                                              assert_timestep_almost_equal)
+from MDAnalysisTests.datafiles import (
+    H5MD_xvf,
+    TPR_xvf,
+    TRR_xvf,
+    COORDINATES_TOPOLOGY,
+    COORDINATES_H5MD,
+    H5MD_energy,
+)
+from MDAnalysisTests.coordinates.base import (
+    MultiframeReaderTest,
+    BaseReference,
+    BaseWriterTest,
+)
 
 
 @pytest.mark.skipif(not HAS_H5PY, reason="h5py not installed")
@@ -894,3 +902,23 @@ def test_writer_no_h5py(self, Writer, outfile):
                         u.atoms.n_atoms) as W:
                 for ts in u.trajectory:
                     W.write(universe)
+
+
+@pytest.mark.skipif(not HAS_H5PY, reason="h5py not installed")
+class TestH5MDReaderWithObservables(object):
+    """Read H5MD file with 'observables/atoms/energy'."""
+
+    prec = 3
+    ext = 'h5md'
+
+    def test_read_h5md_issue4598(self):
+        """Read a H5MD file with observables.
+
+        The reader will ignore the 'observables/atoms/energy'.
+        """
+
+        u = mda.Universe.empty(n_atoms=108, trajectory=True)
+        reader = H5MDReader(H5MD_energy)
+        u.trajectory = reader
+        for ts in u.trajectory:
+            assert "atoms/energy" in ts.data

testsuite/MDAnalysisTests/datafiles.py

@@ -77,6 +77,7 @@
     "GRO_sameresid_diffresname",  # Case where two residues share the same resid
     "PDB_xvf", "TPR_xvf", "TRR_xvf",  # Gromacs coords/veloc/forces (cobrotoxin, OPLS-AA, Gromacs 4.5.5 tpr)
     "H5MD_xvf",  # TPR_xvf + TRR_xvf converted to h5md format
+    "H5MD_energy",  # H5MD trajectory with observables/atoms/energy
     "XVG_BZ2",  # Compressed xvg file about cobrotoxin
     "PDB_xlserial",
     "TPR400", "TPR402", "TPR403", "TPR404", "TPR405", "TPR406", "TPR407",
@@ -373,6 +374,7 @@
 TPR_xvf = (_data_ref / 'cobrotoxin.tpr').as_posix()
 TRR_xvf = (_data_ref / 'cobrotoxin.trr').as_posix()
 H5MD_xvf = (_data_ref / 'cobrotoxin.h5md').as_posix()
+H5MD_energy = (_data_ref / 'cu.h5md').as_posix()
 XVG_BZ2 = (_data_ref / 'cobrotoxin_protein_forces.xvg.bz2').as_posix()
 
 XPDB_small = (_data_ref / '5digitResid.pdb').as_posix()