Fix ErrorTree polluting contents on index access

ErrorTree used a defaultdict for _contents, which meant accessing
a non-existent index via __getitem__ would silently create an empty
entry. This caused __iter__ and __contains__ to return incorrect
results after accessing an index with no errors.

Replaced defaultdict with a regular dict and changed __init__ to
explicitly create subtrees. __getitem__ now returns a new empty
ErrorTree for valid indices without storing it in _contents.

Author: bysiber

commit_msg.txt

@@ -0,0 +1,10 @@
+Fix ErrorTree polluting contents on index access
+
+ErrorTree used a defaultdict for _contents, which meant accessing
+a non-existent index via __getitem__ would silently create an empty
+entry. This caused __iter__ and __contains__ to return incorrect
+results after accessing an index with no errors.
+
+Replaced defaultdict with a regular dict and changed __init__ to
+explicitly create subtrees. __getitem__ now returns a new empty
+ErrorTree for valid indices without storing it in _contents.

jsonschema/exceptions.py

@@ -3,7 +3,7 @@
 """
 from __future__ import annotations
 
-from collections import defaultdict, deque
+from collections import deque
 from pprint import pformat
 from textwrap import dedent, indent
 from typing import TYPE_CHECKING, Any, ClassVar
@@ -321,12 +321,14 @@ class ErrorTree:
 
     def __init__(self, errors: Iterable[ValidationError] = ()):
         self.errors: MutableMapping[str, ValidationError] = {}
-        self._contents: Mapping[str, ErrorTree] = defaultdict(self.__class__)
+        self._contents: dict[str | int, ErrorTree] = {}
 
         for error in errors:
             container = self
             for element in error.path:
-                container = container[element]
+                if element not in container._contents:
+                    container._contents[element] = self.__class__()
+                container = container._contents[element]
             container.errors[error.validator] = error
 
             container._instance = error.instance
@@ -348,7 +350,9 @@ def __getitem__(self, index):
         """
         if self._instance is not _unset and index not in self:
             self._instance[index]
-        return self._contents[index]
+        if index in self._contents:
+            return self._contents[index]
+        return self.__class__()
 
     def __setitem__(self, index: str | int, value: ErrorTree):
         """

jsonschema/tests/test_exceptions.py

@@ -506,6 +506,23 @@ def test_repr_empty(self):
         tree = exceptions.ErrorTree([])
         self.assertEqual(repr(tree), "<ErrorTree (0 total errors)>")
 
+    def test_accessing_index_without_error_does_not_pollute_contents(self):
+        error = exceptions.ValidationError(
+            "not a number",
+            validator="type",
+            path=[0],
+            instance=["spam", 2],
+        )
+        tree = exceptions.ErrorTree([error])
+        self.assertEqual(list(tree), [0])
+        self.assertNotIn(1, tree)
+
+        # accessing an index with no error should not add it to the tree
+        tree[1]
+
+        self.assertEqual(list(tree), [0])
+        self.assertNotIn(1, tree)
+
 
 class TestErrorInitReprStr(TestCase):
     def make_error(self, **kwargs):