Add __hash__ to DataSliceManagerView.

This makes DataSliceManagerView instances hashable, allowing them to be used in sets and as dictionary keys. The hash is based on the underlying data slice manager and the path from the root.

PiperOrigin-RevId: 864834042
Change-Id: I2ad93d76b50777914596bd976dfb21452a38be3a

Author: svsgoogle

py/koladata/ext/persisted_data/data_slice_manager_view.py

@@ -68,17 +68,31 @@ def __init__(
       path_from_root: The view path. It must be a valid path for the manager,
         i.e. manager.exists(path_from_root) must be True.
     """
+    # Conceptually, this class is a frozen dataclass with two attributes:
+    # _data_slice_manager and _path_from_root. However, since we override
+    # __getattr__() and __setattr__(), we cannot use a frozen dataclass, and we
+    # use __dict__ here to avoid triggering __setattr__().
     self.__dict__['_data_slice_manager'] = manager
     self.__dict__['_path_from_root'] = path_from_root
     self._check_path_from_root_is_valid()
 
   def __eq__(self, other: Any) -> bool:
+    # A frozen dataclass would give this behavior for free. As mentioned above,
+    # we cannot use a frozen dataclass, so we implement __eq__ manually here.
     return (
         type(self) is type(other)
         and self._data_slice_manager == other._data_slice_manager
         and self._path_from_root == other._path_from_root
     )
 
+  def __hash__(self) -> int:
+    # A frozen dataclass would give this behavior for free. As mentioned above,
+    # we cannot use a frozen dataclass, so we implement __hash__ manually here
+    # to be able to use sets and dicts with DataSliceManagerView instances.
+    # Implementing __eq__ removes the default hash behavior, so we need to
+    # implement __hash__ explicitly.
+    return hash((self._data_slice_manager, self._path_from_root))
+
   def __repr__(self) -> str:
     return (
         f'DataSliceManagerView({repr(self._data_slice_manager)},'

py/koladata/ext/persisted_data/data_slice_manager_view_test.py

@@ -1873,6 +1873,62 @@ def test_get_with_populate_arguments(self):
         ids_equality=True,
     )
 
+  def test_views_are_hashable(self):
+    persistence_dir = self.create_tempdir().full_path
+    manager = pidsm.PersistedIncrementalDataSliceManager.create_new(
+        persistence_dir
+    )
+    root = DataSliceManagerView(manager)
+    root.query = (
+        kd.list([
+            kd.named_schema('query').new(query_id=0, text='How tall is Obama'),
+            kd.named_schema('query').new(
+                query_id=1, text='How high is the Eiffel tower'
+            ),
+        ]),
+        'Added queries with query_id and text',
+    )
+    query = root.query[:]
+    query.doc = (
+        kd.list([
+            kd.named_schema('doc').new(doc_id=0, title='Barack Obama'),
+            kd.named_schema('doc').new(doc_id=1, title='Michelle Obama'),
+            kd.named_schema('doc').new(doc_id=2, title='George W. Bush'),
+        ]),
+        'Added docs with doc_id and title',
+    )
+    doc = query.doc[:]
+
+    self.assertEqual(hash(root), hash(root))
+    self.assertEqual(hash(root), hash(root.get_root()))
+    self.assertEqual(hash(root.query[:]), hash(query))
+    self.assertEqual(hash(root.query[:]), hash(root.query[:]))
+    self.assertEqual(hash(root.query[:].text), hash(query.text))
+    self.assertEqual(hash(root.query[:].text), hash(root.query[:].text))
+    self.assertEqual(hash(root.query[:].doc[:].title), hash(doc.title))
+
+    my_view_set = {root, query, doc, doc.title, root.query[:].doc[:]}
+    self.assertLen(my_view_set, 4)
+
+    another_manager = (
+        pidsm.PersistedIncrementalDataSliceManager.create_from_dir(
+            persistence_dir
+        )
+    )
+    another_root = DataSliceManagerView(another_manager)
+
+    self.assertNotEqual(hash(another_root), hash(root))
+    self.assertNotEqual(hash(another_root), hash(root.get_root()))
+    self.assertNotEqual(hash(another_root.query[:]), hash(query))
+    self.assertNotEqual(hash(another_root.query[:]), hash(root.query[:]))
+    self.assertNotEqual(hash(another_root.query[:].text), hash(query.text))
+    self.assertNotEqual(
+        hash(another_root.query[:].text), hash(root.query[:].text)
+    )
+    self.assertNotEqual(
+        hash(another_root.query[:].doc[:].title), hash(doc.title)
+    )
+
 
 if __name__ == '__main__':
   absltest.main()