Fix filter matching with bypass tree (#719)

* Make auto rule tester work with bypass rule tree
* Fix FilterExpression to prevent incorrect matching

CHANGELOG.md

@@ -29,7 +29,9 @@
 
 * fix `confluent_kafka.store_offsets` if `last_valid_record` is `None`, can happen if a rebalancing happens
   before the first message was pulled.
-- fix pseudonymizer cache metrics not updated
+* fix pseudonymizer cache metrics not updated
+* fix `_get_value` in `FilterExpression` so that keys don't match on values
+* fix `auto_rule_tester` to work with `LOGPREP_BYPASS_RULE_TREE` enabled
 
 ## 14.0.0
 ### Breaking

logprep/filter/expression/filter_expression.py

@@ -85,6 +85,8 @@ def _get_value(key: List[str], document: dict) -> Any:
 
         current = document
         for item in key:
+            if not isinstance(current, dict):
+                raise KeyDoesNotExistError
             if item not in current:
                 raise KeyDoesNotExistError
             current = current[item]

logprep/util/auto_rule_tester/auto_rule_tester.py

@@ -403,6 +403,8 @@ def _load_rules(self, processor: "Processor", rule_type: str):
             processor.load_rules(self._empty_rules_dirs, [])
         elif rule_type == "generic_rules":
             processor.load_rules([], self._empty_rules_dirs)
+        if processor._bypass_rule_tree:
+            processor._rules = processor.rules
         processor.setup()
 
     def _prepare_test_eval(

tests/unit/filter/test_filter_expression.py

@@ -36,6 +36,11 @@ def test_get_value_fails_empty_key(self):
         with pytest.raises(KeyDoesNotExistError):
             FilterExpression._get_value([], {"some": "value"})
 
+    def test_get_value_fails_when_key_matches_value_instead_of_key(self):
+        document = {"key": "not_a_key"}
+        with pytest.raises(KeyDoesNotExistError):
+            FilterExpression._get_value(["key", "not_a_key"], document)
+
     def test_get_value_returns_expected_value(self):
         document = {"one": {"two": "value"}, "ten": {"eleven": "another value"}}