Fix propagation of global attributes to child datasets and variables

CHANGES.md

@@ -22,6 +22,9 @@
 - Load plugins from entry points allowing plugins to be discovered from installed libraries.
   - Automatically generate rule documentation removing the manual need to run `mkruleref.py`.
 
+- Fixed passing of global datatree attributes to children: attributes defined
+  on parent datatrees are now inherited by all descendants. (#63)
+
 
 ## Version 0.5.1 (from 2025-02-21)
 

tests/test_linter.py

@@ -138,6 +138,12 @@ def validate_datatree(self, ctx: RuleContext, node: DataTreeNode):
                 if len(node.datatree.data_vars) == 0:
                     ctx.report("DataTree does not have data variables")
 
+        @plugin.define_rule("datatree-children-must-have-title")
+        class DataTreeAttrsVer(RuleOp):
+            def validate_datatree(self, ctx: RuleContext, node: DataTreeNode):
+                if "title" not in node.datatree.attrs:
+                    ctx.report("DataTree must have a least a global title")
+
         @plugin.define_processor("multi-level-dataset")
         class MultiLevelDataset(ProcessorOp):
             def preprocess(
@@ -167,6 +173,7 @@ def test_rules_are_ok(self):
                 "data-var-dim-must-have-coord",
                 "dataset-without-data-vars",
                 "datatree-without-data-vars",
+                "datatree-children-must-have-title",
             ],
             list(self.linter.config.objects[0].plugins["test"].rules.keys()),
         )
@@ -308,6 +315,109 @@ def test_linter_recognized_datatree_rule(self):
         self.assertEqual(5, result.error_count)
         self.assertEqual(0, result.fatal_error_count)
 
+    def test_linter_missing_global_datatree_attrs(self):
+        result = self.linter.validate(
+            xr.DataTree(
+                children={
+                    "measurement": xr.DataTree(
+                        children={
+                            "r10m": xr.DataTree(
+                                dataset=xr.Dataset(
+                                    attrs={
+                                        "title": "10m resolution datatree",
+                                    }
+                                )
+                            ),
+                            "r20m": xr.DataTree(),
+                            "r60m": xr.DataTree(),
+                        }
+                    )
+                },
+            ),
+            rules={"test/datatree-children-must-have-title": 2},
+        )
+
+        self.assertEqual(
+            [
+                Message(
+                    message="DataTree must have a least a global title",
+                    node_path="dt",
+                    rule_id="test/datatree-children-must-have-title",
+                    severity=2,
+                    fatal=None,
+                    fix=None,
+                    suggestions=None,
+                ),
+                Message(
+                    message="DataTree must have a least a global title",
+                    node_path="dt/measurement",
+                    rule_id="test/datatree-children-must-have-title",
+                    severity=2,
+                    fatal=None,
+                    fix=None,
+                    suggestions=None,
+                ),
+                Message(
+                    message="DataTree must have a least a global title",
+                    node_path="dt/measurement/r20m",
+                    rule_id="test/datatree-children-must-have-title",
+                    severity=2,
+                    fatal=None,
+                    fix=None,
+                    suggestions=None,
+                ),
+                Message(
+                    message="DataTree must have a least a global title",
+                    node_path="dt/measurement/r60m",
+                    rule_id="test/datatree-children-must-have-title",
+                    severity=2,
+                    fatal=None,
+                    fix=None,
+                    suggestions=None,
+                ),
+            ],
+            result.messages,
+        )
+        self.assertEqual(0, result.warning_count)
+        self.assertEqual(4, result.error_count)
+        self.assertEqual(0, result.fatal_error_count)
+
+    def test_linter_global_datatree_attrs(self):
+        result = self.linter.validate(
+            xr.DataTree(
+                dataset=xr.Dataset(
+                    attrs={
+                        "title": "Global datatree title",
+                    }
+                ),
+                children={
+                    "measurement": xr.DataTree(
+                        children={
+                            "r10m": xr.DataTree(
+                                dataset=xr.Dataset(
+                                    attrs={
+                                        "title": "10m resolution datatree",
+                                    }
+                                )
+                            ),
+                            "r20m": xr.DataTree(),
+                            "r60m": xr.DataTree(),
+                        }
+                    )
+                },
+            ),
+            rules={"test/datatree-children-must-have-title": 2},
+        )
+
+        print(result.messages)
+        self.assertEqual(
+            [],
+            result.messages,
+        )
+        self.assertEqual(0, result.warning_count)
+        self.assertEqual(0, result.error_count)
+        self.assertEqual(0, result.fatal_error_count)
+
     def test_linter_real_life_scenario(self):
         dataset = xr.Dataset(
             attrs={

xrlint/_linter/apply.py

@@ -62,21 +62,33 @@ def apply_rule(
 
 
 def _visit_datatree_node(rule_op: RuleOp, context: RuleContextImpl, node: DataTreeNode):
+    # Get a copy of the current node's attrs.
+    # These will be merged into each child's attrs so that attributes
+    # defined on parent nodes are inherited by all descendants.
+    attrs = node.datatree.attrs.copy()
+
     with context.use_state(node=node):
         rule_op.validate_datatree(context, node)
+
         if node.datatree.is_leaf:
+            # Inherit attrs from the parent datatree into the child dataset
+            dataset = node.datatree.dataset.copy()
+            dataset.attrs = {**attrs, **dataset.attrs}
             _visit_dataset_node(
                 rule_op,
                 context,
                 DatasetNode(
                     parent=node,
                     path=f"{node.path}/{node.datatree.name}",
                     name=node.datatree.name,
-                    dataset=node.datatree.dataset,
+                    dataset=dataset,
                 ),
             )
         else:
             for name, datatree in node.datatree.children.items():
+                # Inherit attrs from the parent datatree into the child datatree
+                datatree = datatree.copy()
+                datatree.attrs = {**attrs, **datatree.attrs}
                 _visit_datatree_node(
                     rule_op,
                     context,