fix setting-merge bug in _get_optuna_settings

Author: JanTeichertKluge

doubleml/utils/_tune_optuna.py

@@ -256,7 +256,7 @@ def _check_tuning_inputs(
     return resolve_optuna_cv(cv)
 
 
-def _get_optuna_settings(optuna_settings, params_name=None):
+def _get_optuna_settings(optuna_settings, params_name):
     """
     Get Optuna settings, considering defaults, user-provided values, and learner-specific overrides.
 
@@ -265,7 +265,7 @@ def _get_optuna_settings(optuna_settings, params_name=None):
     optuna_settings : dict or None
         User-provided Optuna settings.
     params_name : str
-        Name of the learner to check for specific setting, e.g. `ml_g0` or `ml_g1` for `DoubleMLIRM`.
+        Name of the nuisance params to check for specific setting, e.g. `ml_g0` or `ml_g1` for `DoubleMLIRM`.
 
     Returns
     -------
@@ -286,18 +286,25 @@ def _get_optuna_settings(optuna_settings, params_name=None):
 
     # Find matching learner-specific settings, handles the case to match ml_g to ml_g0, ml_g1, etc.
     learner_specific_settings = {}
-    if any(params_name in key for key in learner_or_params_keys):
-        for k in learner_or_params_keys:
-            if params_name in k and params_name != k:
-                learner_specific_settings = optuna_settings[k]
+    prefix_matches = [key for key in learner_or_params_keys if key != params_name and params_name.startswith(key)]
+    if prefix_matches:
+        learner_key = max(prefix_matches, key=len)
+        learner_specific_settings = optuna_settings[learner_key]
+        if not isinstance(learner_specific_settings, dict):
+            raise TypeError(f"Optuna settings for '{learner_key}' must be a dict.")
 
     # set params specific settings
     params_specific_settings = {}
     if params_name in learner_or_params_keys:
         params_specific_settings = optuna_settings[params_name]
+        if not isinstance(params_specific_settings, dict):
+            raise TypeError(f"Optuna settings for '{params_name}' must be a dict.")
 
     # Merge settings: defaults < base < learner-specific < params_specific
-    resolved = default_settings.copy() | base_settings | learner_specific_settings | params_specific_settings
+    resolved = default_settings.copy()
+    resolved |= base_settings
+    resolved |= learner_specific_settings
+    resolved |= params_specific_settings
 
     # Validate types
     if not isinstance(resolved["study_kwargs"], dict):