Add multi_index configuration option

Used to decide to apply reset_index() to the features dataframes

src/blueetl/config/analysis_model.py

@@ -151,6 +151,7 @@ class FeaturesConfig(BaseModel):
     params_product: dict[str, Any] = {}
     params_zip: dict[str, Any] = {}
     suffix: str = ""
+    multi_index: bool = True
 
 
 class SingleAnalysisConfig(BaseModel):

src/blueetl/features.py

@@ -417,7 +417,11 @@ def _func_wrapper(
         # ignore the index if it's unnamed and with one level; this can be useful
         # for example when the returned DataFrame has a RangeIndex to be dropped
         drop = result_df.index.names == [None]
-        result_df = result_df.etl.add_conditions(conditions=key._fields, values=key, drop=drop)
+        if features_config.multi_index:
+            result_df = result_df.etl.add_conditions(conditions=key._fields, values=key, drop=drop)
+        else:
+            result_df.reset_index(drop=drop, inplace=True)
+            result_df.etl.insert_columns(loc=0, columns=key._fields, values=key)
         features_records[feature_group + features_config.suffix] = result_df
     return features_records
 

src/blueetl/schemas/analysis_config.yaml

@@ -444,6 +444,15 @@ $defs:
           A numeric suffix is automatically added when any of ``params_product`` or ``params_zip`` is specified.
         default: "''"
         type: string
+      multi_index:
+        title: MultiIndex
+        description: |
+          - If True, do not reset the index of the resulting DataFrames of features, and add the values specified in ``groupby`` to the MultiIndex.
+          - If False, reset the index, returning columnar DataFrames.
+
+          The DataFrames with MultiIndex should use less memory then the columnar DataFrames, but they take more time to load and dump to disk.
+        type: boolean
+        default: "true"
     required:
     - type
     - groupby