Name permanence via shallow copy (#172)

* added descendants property

* tests for descendants, lineage, ancestors, subtree

* added descendants to API docs

* whatsnew

* rerun tests

* rewrote copy method

* remove outdated mypy ignore error

* changed tests to reflect new expected behaviour

* shallow copy on insertion

* update test for checking isomorphism from root

* whatsnew

Author: TomNicholas

datatree/datatree.py

@@ -754,9 +754,8 @@ def _set(self, key: str, val: DataTree | CoercibleValue) -> None:
         Counterpart to the public .get method, and also only works on the immediate node, not other nodes in the tree.
         """
         if isinstance(val, DataTree):
-            # TODO shallow copy here so as not to alter name of node in original tree?
-            # new_node = copy.copy(val, deep=False)
-            new_node = val
+            # create and assign a shallow copy here so as not to alter original name of node in grafted tree
+            new_node = val.copy(deep=False)
             new_node.name = key
             new_node.parent = self
         else:

datatree/tests/test_datatree.py

@@ -68,12 +68,6 @@ def test_child_gets_named_on_attach(self):
         mary = DataTree(children={"Sue": sue})  # noqa
         assert sue.name == "Sue"
 
-    @pytest.mark.xfail(reason="requires refactoring to retain name")
-    def test_grafted_subtree_retains_name(self):
-        subtree = DataTree("original")
-        root = DataTree(children={"new_name": subtree})  # noqa
-        assert subtree.name == "original"
-
 
 class TestPaths:
     def test_path_property(self):
@@ -294,8 +288,11 @@ class TestSetItem:
     def test_setitem_new_child_node(self):
         john = DataTree(name="john")
         mary = DataTree(name="mary")
-        john["Mary"] = mary
-        assert john["Mary"] is mary
+        john["mary"] = mary
+
+        grafted_mary = john["mary"]
+        assert grafted_mary.parent is john
+        assert grafted_mary.name == "mary"
 
     def test_setitem_unnamed_child_node_becomes_named(self):
         john2 = DataTree(name="john2")
@@ -304,10 +301,19 @@ def test_setitem_unnamed_child_node_becomes_named(self):
 
     def test_setitem_new_grandchild_node(self):
         john = DataTree(name="john")
-        DataTree(name="mary", parent=john)
+        mary = DataTree(name="mary", parent=john)
         rose = DataTree(name="rose")
-        john["Mary/Rose"] = rose
-        assert john["Mary/Rose"] is rose
+        john["mary/rose"] = rose
+
+        grafted_rose = john["mary/rose"]
+        assert grafted_rose.parent is mary
+        assert grafted_rose.name == "rose"
+
+    def test_grafted_subtree_retains_name(self):
+        subtree = DataTree(name="original_subtree_name")
+        root = DataTree(name="root")
+        root["new_subtree_name"] = subtree  # noqa
+        assert subtree.name == "original_subtree_name"
 
     def test_setitem_new_empty_node(self):
         john = DataTree(name="john")

datatree/tests/test_mapping.py

@@ -68,8 +68,9 @@ def test_not_isomorphic_complex_tree(self, create_test_datatree):
     def test_checking_from_root(self, create_test_datatree):
         dt1 = create_test_datatree()
         dt2 = create_test_datatree()
-        real_root = DataTree()
-        real_root["fake_root"] = dt2
+        real_root = DataTree(name="real root")
+        dt2.name = "not_real_root"
+        dt2.parent = real_root
         with pytest.raises(TreeIsomorphismError):
             check_isomorphic(dt1, dt2, check_from_root=True)
 

docs/source/whats-new.rst

@@ -37,6 +37,8 @@ Breaking changes
 
 - :py:meth:`DataTree.copy` copy method now only copies the subtree, not the parent nodes (:pull:`171`).
   By `Tom Nicholas <https://github.com/TomNicholas>`_.
+- Grafting a subtree onto another tree now leaves name of original subtree object unchanged (:issue:`116`, :pull:`172`).
+  By `Tom Nicholas <https://github.com/TomNicholas>`_.
 
 Deprecations
 ~~~~~~~~~~~~