Merge pull request 'schemagen: reject required entries with no matching properties key' (#65) from weaselbot/weaseljson:weaselbot/issue-63 into main

Reviewed-on: weaselab/weaseljson#65

contrib/schemagen/README.md

@@ -62,7 +62,8 @@ into the result, so it is non-movable.
 `oneOf` / `anyOf` / `allOf` / `not` / `if`-`then`-`else`,
 `patternProperties`, `additionalProperties` absent or set to `true`,
 `additionalProperties` with a schema (typed map), `prefixItems` (tuples),
-`const`,
+`const`, `required` entries with no matching `properties` key (the schema
+would be unsatisfiable),
 `dependentSchemas`/`dependentRequired`, union `type` lists other than
 `["T", "null"]`, non-string enums, and remote (`$ref` to other documents).
 

contrib/schemagen/test_schemagen.py

@@ -287,6 +287,94 @@ class SchemagenAdditionalPropertiesTest(unittest.TestCase):
         self.assertEqual(rc, 0, msg=stderr)
 
 
+class SchemagenRequiredPropertyTest(unittest.TestCase):
+    """Regression tests for issue #63: `required` entries with no matching
+    `properties` key must be rejected at generation time. With the mandatory
+    `additionalProperties: false` the schema is unsatisfiable, and the
+    generated parser has no seen bit for such keys, so it would otherwise
+    silently accept documents missing the required key."""
+
+    def run_schemagen(self, schema, args=None):
+        """Run schemagen on a schema dict. Returns (returncode, stdout, stderr)."""
+        with tempfile.NamedTemporaryFile("w", suffix=".json", delete=False) as fp:
+            json.dump(schema, fp)
+            schema_path = fp.name
+        try:
+            cmd = [sys.executable, SCRIPT, schema_path]
+            if args:
+                cmd.extend(args)
+            result = subprocess.run(cmd, capture_output=True, text=True, check=False)
+            return result.returncode, result.stdout, result.stderr
+        finally:
+            os.unlink(schema_path)
+
+    def test_required_key_without_properties_entry_rejected(self):
+        schema = {
+            "type": "object",
+            "additionalProperties": False,
+            "properties": {"a": {"type": "integer"}},
+            "required": ["b"],
+        }
+        rc, stdout, stderr = self.run_schemagen(schema)
+        self.assertNotEqual(rc, 0)
+        self.assertIn("required key 'b' has no matching properties entry", stderr)
+
+    def test_multiple_dangling_required_keys_reported(self):
+        schema = {
+            "type": "object",
+            "additionalProperties": False,
+            "properties": {"a": {"type": "integer"}},
+            "required": ["a", "x", "y"],
+        }
+        rc, stdout, stderr = self.run_schemagen(schema)
+        self.assertNotEqual(rc, 0)
+        self.assertIn(
+            "required keys 'x', 'y' have no matching properties entries",
+            stderr,
+        )
+
+    def test_required_without_properties_rejected(self):
+        schema = {
+            "type": "object",
+            "additionalProperties": False,
+            "required": ["any"],
+        }
+        rc, stdout, stderr = self.run_schemagen(schema)
+        self.assertNotEqual(rc, 0)
+        self.assertIn("required key 'any' has no matching properties entry", stderr)
+
+    def test_nested_required_key_without_properties_entry_rejected(self):
+        schema = {
+            "type": "object",
+            "additionalProperties": False,
+            "properties": {"child": {"$ref": "#/$defs/Child"}},
+            "$defs": {
+                "Child": {
+                    "type": "object",
+                    "additionalProperties": False,
+                    "properties": {"x": {"type": "string"}},
+                    "required": ["misspelled"],
+                }
+            },
+        }
+        rc, stdout, stderr = self.run_schemagen(schema)
+        self.assertNotEqual(rc, 0)
+        self.assertIn(
+            "required key 'misspelled' has no matching properties entry",
+            stderr,
+        )
+
+    def test_required_key_with_properties_entry_accepted(self):
+        schema = {
+            "type": "object",
+            "additionalProperties": False,
+            "properties": {"a": {"type": "integer"}, "b": {"type": "integer"}},
+            "required": ["b"],
+        }
+        rc, stdout, stderr = self.run_schemagen(schema)
+        self.assertEqual(rc, 0, msg=stderr)
+
+
 class SchemagenCyclicArrayTest(unittest.TestCase):
     """Regression tests for issue #33: cyclic array $ref targets."""
 

contrib/schemagen/weaseljson_schemagen.py

@@ -432,6 +432,22 @@ class Builder:
             )
         required = set(node.get("required", []))
         props = node.get("properties", {})
+        # A `required` entry with no matching `properties` key makes the schema
+        # unsatisfiable: `additionalProperties: false` (the only supported
+        # object mode) rejects any key not listed in `properties`, so the
+        # required key can never be present. No seen bit is emitted for such a
+        # key, so the generated parser would silently accept every document.
+        # Follow the other unsupported constructs and reject at generation time.
+        dangling = sorted(required - set(props))
+        if dangling:
+            quoted = ", ".join(f"'{k}'" for k in dangling)
+            if len(dangling) == 1:
+                raise GenError(
+                    f"required key {quoted} has no matching properties entry"
+                )
+            raise GenError(
+                f"required keys {quoted} have no matching properties entries"
+            )
         seen_cpp = set()
         for key, sub in props.items():
             ty, nullable = self._unpack(self.build_type(sub, key))