Avoid using ? with get_item to handle unhashable inputs properly (#…

…1089)

Co-authored-by: Samuel Colvin <s@muelcolvin.com>

src/validators/literal.rs

@@ -136,7 +136,10 @@ impl<T: Debug> LiteralLookup<T> {
         }
         // must be an enum or bytes
         if let Some(expected_py) = &self.expected_py {
-            if let Some(v) = expected_py.as_ref(py).get_item(input)? {
+            // We don't use ? to unpack the result of `get_item` in the next line because unhashable
+            // inputs will produce a TypeError, which in this case we just want to treat equivalently
+            // to a failed lookup
+            if let Ok(Some(v)) = expected_py.as_ref(py).get_item(input) {
                 let id: usize = v.extract().unwrap();
                 return Ok(Some((input, &self.values[id])));
             }

tests/validators/test_union.py

@@ -771,3 +771,35 @@ class BinaryEnum(IntEnum):
     assert validator.validate_python(1) is not BinaryEnum.ONE
     assert validator.validate_python(BinaryEnum.ZERO) is BinaryEnum.ZERO
     assert validator.validate_python(BinaryEnum.ONE) is BinaryEnum.ONE
+
+
+def test_model_and_literal_union() -> None:
+    # see https://github.com/pydantic/pydantic/issues/8183
+    class ModelA:
+        pass
+
+    validator = SchemaValidator(
+        {
+            'type': 'union',
+            'choices': [
+                {
+                    'type': 'model',
+                    'cls': ModelA,
+                    'schema': {
+                        'type': 'model-fields',
+                        'fields': {
+                            'a': {'type': 'model-field', 'schema': {'type': 'int'}},
+                        },
+                    },
+                },
+                {'type': 'literal', 'expected': [True]},
+            ],
+        }
+    )
+
+    # validation against Literal[True] fails bc of the unhashable dict
+    # A ValidationError is raised, not a ValueError, which allows the validation against the union to continue
+    m = validator.validate_python({'a': 42})
+    assert isinstance(m, ModelA)
+    assert m.a == 42
+    assert validator.validate_python(True) is True