♻️ PickledData: derive from Data, not SinglefileData (aiidateam#7600)

`ShellJob` writes every `SinglefileData` input into the working
directory, so passing a `PickledData` through the `nodes` port dropped
raw pickle bytes there as an input file. Pickled bytes are not a file
the caller supplied, and inheritance said they were.

Store them in the node repository directly, under `obj.pickle`. The
entry point still names the node type, so existing nodes keep loading;
only the inherited `filename` attribute goes.

Author: GeigerJ2

src/aiida/orm/nodes/data/pickled.py

@@ -12,22 +12,24 @@
 
 import importlib
 import importlib.metadata
-import io
 import sys
 import types
 import typing as t
 
 from aiida.common.log import AIIDA_LOGGER
-from aiida.orm.nodes.data.singlefile import SinglefileData
+from aiida.orm.nodes.data.data import Data
 
 __all__ = ('PickledData',)
 
 LOGGER = AIIDA_LOGGER.getChild('pickled_data')
 
 DEFAULT_PICKLER: str = 'dill'
 
+FILENAME: str = 'obj.pickle'
+"""Name the pickled bytes are stored under in the node repository."""
 
-class PickledData(SinglefileData):
+
+class PickledData(Data):
     """Data plugin to store (almost) any Python object by pickling it.
 
     The pickler is chosen per node and recorded on it, so that a node knows how to unpickle itself::
@@ -37,6 +39,10 @@ class PickledData(SinglefileData):
 
     Any module providing ``dumps`` and ``loads`` can be named. A pickler that does not follow that convention is
     used through a small adapter module rather than by subclassing this class.
+
+    This deliberately does not extend :class:`~aiida.orm.nodes.data.singlefile.SinglefileData`. Pickled bytes are
+    not a file the caller supplied, and ``ShellJob`` writes every ``SinglefileData`` input into the working
+    directory, which would dump the raw pickle there.
     """
 
     KEY_ATTRIBUTES_PICKLER: str = 'pickler'
@@ -68,8 +74,9 @@ def __init__(self, obj: t.Any, *, pickler: str = DEFAULT_PICKLER, **kwargs: t.An
         """
         module = self._load_pickler(pickler)
 
-        super().__init__(file=io.BytesIO(module.dumps(obj, **kwargs)))
+        super().__init__()
 
+        self.base.repository.put_object_from_bytes(module.dumps(obj, **kwargs), FILENAME)
         self.base.attributes.set(self.KEY_ATTRIBUTES_PICKLER, pickler)
         self.base.attributes.set(self.KEY_ATTRIBUTES_PICKLER_VERSION, self._get_distribution_version(pickler))
         self.base.attributes.set(self.KEY_ATTRIBUTES_PICKLER_KWARGS, kwargs)
@@ -189,7 +196,7 @@ def load(self) -> t.Any:
         :returns: The unpickled Python object.
         :raises ValueError: If the stored pickled object could not be unpickled.
         """
-        with self.open(mode='rb') as handle:
+        with self.base.repository.open(FILENAME, mode='rb') as handle:
             pickled = handle.read()
 
         try:

tests/orm/data/test_pickled.py

@@ -36,7 +36,6 @@ def test_records_pickling_environment():
     """Test that the constructor records everything needed to decide whether this node can be unpickled here."""
     node = PickledData(None)
     assert node.base.attributes.all == {
-        'filename': node.filename,
         'pickler': 'dill',
         'pickler_version': dill.__version__,
         'pickler_kwargs': {},

tests/orm/test_fields/fields_aiida.data.core.pickled.PickledData.yml

@@ -5,8 +5,6 @@ ctime: QbNumericField('ctime', dtype=<class 'datetime.datetime'>, doc='The creat
   time of the node')
 description: QbStrField('description', dtype=<class 'str'>, doc='The node description')
 extras: QbDictField('extras', dtype=dict[str, typing.Any], doc='The node extras')
-filename: QbStrField('attributes.filename', dtype=<class 'str'>, doc='The name of
-  the stored file')
 label: QbStrField('label', dtype=<class 'str'>, doc='The node label')
 mtime: QbNumericField('mtime', dtype=<class 'datetime.datetime'>, doc='The modification
   time of the node')