implement group/dataset id as a string

examples/example1.py

@@ -22,3 +22,9 @@
 with pynwb.NWBHDF5IO(file=client, mode="r") as io:
     nwbfile = io.read()
     print(nwbfile)
+
+    print('Electrode group at shank0:')
+    print(nwbfile.electrode_groups["shank0"])  # type: ignore
+
+    print('Electrode group at index 0:')
+    print(nwbfile.electrodes.group[0])  # type: ignore

lindi/LindiH5pyFile/LindiH5pyDataset.py

@@ -14,11 +14,6 @@
     from .LindiH5pyFile import LindiH5pyFile  # pragma: no cover
 
 
-class LindiH5pyDatasetId:
-    def __init__(self, _h5py_dataset_id):
-        self._h5py_dataset_id = _h5py_dataset_id
-
-
 # This is a global list of external hdf5 clients, which are used by
 # possibly multiple LindiH5pyFile objects. The key is the URL of the
 # external hdf5 file, and the value is the h5py.File object.
@@ -32,6 +27,9 @@ def __init__(self, _dataset_object: Union[h5py.Dataset, zarr.Array], _file: "Lin
         self._file = _file
         self._readonly = _file.mode not in ['r+']
 
+        # see comment in LindiH5pyGroup
+        self._id = f'{id(self._file)}/{self._dataset_object.name}'
+
         # See if we have the _COMPOUND_DTYPE attribute, which signifies that
         # this is a compound dtype
         if isinstance(_dataset_object, zarr.Array):
@@ -74,10 +72,8 @@ def __init__(self, _dataset_object: Union[h5py.Dataset, zarr.Array], _file: "Lin
 
     @property
     def id(self):
-        if isinstance(self._dataset_object, h5py.Dataset):
-            return LindiH5pyDatasetId(self._dataset_object.id)
-        else:
-            return LindiH5pyDatasetId(None)
+        # see comment in LindiH5pyGroup
+        return self._id
 
     @property
     def shape(self):  # type: ignore

lindi/LindiH5pyFile/LindiH5pyFile.py

@@ -25,6 +25,9 @@ def __init__(self, _file_object: Union[h5py.File, zarr.Group], *, _zarr_store: U
         self._mode: Literal['r', 'r+'] = _mode
         self._the_group = LindiH5pyGroup(_file_object, self)
 
+        # see comment in LindiH5pyGroup
+        self._id = f'{id(self._file_object)}/'
+
     @staticmethod
     def from_reference_file_system(rfs: Union[dict, str], mode: Literal["r", "r+"] = "r"):
         """
@@ -298,7 +301,8 @@ def __contains__(self, name):
 
     @property
     def id(self):
-        return self._the_group.id
+        # see comment in LindiH5pyGroup
+        return self._id
 
     @property
     def file(self):

lindi/LindiH5pyFile/LindiH5pyGroup.py

@@ -11,17 +11,23 @@
     from .LindiH5pyFile import LindiH5pyFile  # pragma: no cover
 
 
-class LindiH5pyGroupId:
-    def __init__(self, _h5py_group_id):
-        self._h5py_group_id = _h5py_group_id
-
-
 class LindiH5pyGroup(h5py.Group):
     def __init__(self, _group_object: Union[h5py.Group, zarr.Group], _file: "LindiH5pyFile"):
         self._group_object = _group_object
         self._file = _file
         self._readonly = _file.mode not in ['r+']
 
+        # In h5py, the id property is an object that exposes low-level
+        # operations specific to the HDF5 library. LINDI aims to override the
+        # high-level methods such that the low-level operations on id are not
+        # needed. However, sometimes packages (e.g., pynwb) use the id as a
+        # unique identifier for purposes of caching. Therefore, we make the id
+        # to be a string that is unique for each object. If any of the low-level
+        # operations are attempted on this id string, then an exception will be
+        # raised, which will usually indicate that one of the high-level methods
+        # should be overridden.
+        self._id = f'{id(self._file)}/{self._group_object.name}'
+
         # The self._write object handles all the writing operations
         from .writers.LindiH5pyGroupWriter import LindiH5pyGroupWriter  # avoid circular import
         if self._readonly:
@@ -131,16 +137,8 @@ def __repr__(self):
 
     @property
     def id(self):
-        if isinstance(self._group_object, h5py.Group):
-            return LindiH5pyGroupId(self._group_object.id)
-        elif isinstance(self._group_object, zarr.Group):
-            # This is commented out for now because pynwb gets the id of a group
-            # in at least one place. But that could be avoided in the future, at
-            # which time, we could uncomment this.
-            # print('WARNING: Accessing low-level id of LindiH5pyGroup. This should be avoided.')
-            return LindiH5pyGroupId('')
-        else:
-            raise Exception(f'Unexpected group object type: {type(self._group_object)}')
+        # see comment above
+        return self._id
 
     @property
     def file(self):