fix: Drop a file path listed more than once instead of treating its entries as duplicate keys

A path listed more than once would make every entry of that file a duplicate of itself: under the
default fail handling the load failed, and under ignore the first listing won by accident. The
builder now drops the repeats, keeps the first occurrence in its place in the order, and logs a
warning once at build time. This is the behavior the Node SDK has.

Author: kinyoklion

ldclient/integrations/overrides.py

@@ -157,7 +157,14 @@ def build(self, config: DataSourceBuilderConfig) -> OverrideSource:  # pylint: d
             log.warning("Poll interval %s is below the minimum for the file-based override source; using %s", poll_interval, self.MINIMUM_POLL_INTERVAL)
             poll_interval = self.MINIMUM_POLL_INTERVAL
 
-        return _FileOverrideSource(abs_file_paths(self.__paths), self.__duplicate_keys_handling, self.__change_detection, poll_interval)
+        # A path listed more than once would make every entry of that file a duplicate of itself,
+        # so the repeats are dropped and the first occurrence keeps its place in the order.
+        all_paths = abs_file_paths(self.__paths)
+        paths = list(dict.fromkeys(all_paths))
+        if len(paths) != len(all_paths):
+            log.warning("The file-based override source lists the same file more than once; the repeats are ignored")
+
+        return _FileOverrideSource(paths, self.__duplicate_keys_handling, self.__change_detection, poll_interval)
 
 
 __all__ = ['ChangeDetection', 'DuplicateKeysHandling', 'FileOverrideSourceBuilder']

ldclient/testing/integrations/test_file_override_source.py

@@ -125,6 +125,16 @@ def test_builder_resolves_relative_paths():
     assert source._paths == [os.path.abspath('relative/overrides.json')]
 
 
+def test_builder_drops_a_repeated_path_and_warns(caplog):
+    # Under the default fail handling the same file listed twice would make every one of its keys
+    # a duplicate of itself. The repeat is ignored and the first occurrence keeps its place.
+    with caplog.at_level(logging.WARNING):
+        source = FileOverrideSourceBuilder(['/a.json', '/b.json', '/a.json']).build(SDKConfig('SDK_KEY'))
+    assert isinstance(source, _FileOverrideSource)
+    assert source._paths == [os.path.abspath('/a.json'), os.path.abspath('/b.json')]
+    assert any('more than once' in r.getMessage() for r in caplog.records)
+
+
 def test_builder_polls_by_default(tmp_path):
     path = os.path.join(str(tmp_path), 'overrides.json')
     source = FileOverrideSourceBuilder([path]).build(SDKConfig('SDK_KEY'))