[Minor] Reduce DataFrame copies (#1648)

* don't copy return df

* explicity copy df before calling merge

* remove copy from normalize

* fix test

* remove cv fold copy

* do not drop ID when normalizing

* do not copy when init data params

* cleanup

* remove copy from global cv intersect

* remove copy from create_dict_for_events_or_regressors

* handle None events regressors

* fix regressors

* remove copy from _normalize

* remove copy from _make_future_dataframe

* remove copy from _prepare_dataframe_to_predict

* remove double copy in plotting func

* fix typo

* remove split deepcopy

* retain comments of former copies in split

Author: ourownstory

neuralprophet/data/process.py

@@ -251,7 +251,6 @@ def _prepare_dataframe_to_predict(model, df: pd.DataFrame, max_lags: int, freq:
     # Receives df with ID column
     df_prepared = pd.DataFrame()
     for df_name, df_i in df.groupby("ID"):
-        df_i = df_i.copy(deep=True)
         _ = df_utils.infer_frequency(df_i, n_lags=max_lags, freq=freq)
         # check if received pre-processed df
         if "y_scaled" in df_i.columns or "t" in df_i.columns:
@@ -283,7 +282,7 @@ def _prepare_dataframe_to_predict(model, df: pd.DataFrame, max_lags: int, freq:
                 config_seasonality=model.config_seasonality,
                 predicting=True,
             )
-        df_prepared = pd.concat((df_prepared, df_i.copy(deep=True).reset_index(drop=True)), ignore_index=True)
+        df_prepared = pd.concat((df_prepared, df_i.reset_index(drop=True)), ignore_index=True)
     return df_prepared
 
 
@@ -399,8 +398,6 @@ def _check_dataframe(
             "Dataframe has less than n_forecasts + n_lags rows. "
             "Forecasting not possible. Please either use a larger dataset, or adjust the model parameters."
         )
-    # df = df.copy(deep=True)
-    # df, _, _, _ = df_utils.check_multiple_series_id(df)
     df, regressors_to_remove, lag_regressors_to_remove = df_utils.check_dataframe(
         df=df,
         check_y=check_y,
@@ -475,9 +472,6 @@ def _handle_missing_data(
         The pre-processed DataFrame, including imputed missing data, if applicable.
 
     """
-    # df = df.copy(deep=True)
-    # df, _, _, _ = df_utils.check_multiple_series_id(df)
-
     if n_lags == 0 and not predicting:
         # drop rows with NaNs in y and count them
         df_na_dropped = df.dropna(subset=["y"])

neuralprophet/data/split.py

@@ -65,7 +65,7 @@ def _maybe_extend_df(
             future_df["ID"] = df_name
             df_i = pd.concat([df_i, future_df])
             df_i.reset_index(drop=True, inplace=True)
-        extended_df = pd.concat((extended_df, df_i.copy(deep=True)), ignore_index=True)
+        extended_df = pd.concat((extended_df, df_i), ignore_index=True)
     return extended_df, periods_add
 
 
@@ -126,8 +126,8 @@ def _get_maybe_extend_periods(
 def _make_future_dataframe(
     model,
     df: pd.DataFrame,
-    events_df: pd.DataFrame,
-    regressors_df: pd.DataFrame,
+    events_df: Optional[pd.DataFrame],
+    regressors_df: Optional[pd.DataFrame],
     periods: Optional[int],
     n_historic_predictions: int,
     n_forecasts: int,
@@ -174,13 +174,12 @@ def _make_future_dataframe(
         log.warning(
             "Not extending df into future as no periods specified. You can skip this and predict directly instead."
         )
-    df = df.copy(deep=True)
     _ = df_utils.infer_frequency(df, n_lags=max_lags, freq=freq)
     last_date = pd.to_datetime(df["ds"].copy(deep=True).dropna()).sort_values().max()
     if events_df is not None:
-        events_df = events_df.copy(deep=True).reset_index(drop=True)
+        events_df = events_df.reset_index(drop=True)
     if regressors_df is not None:
-        regressors_df = regressors_df.copy(deep=True).reset_index(drop=True)
+        regressors_df = regressors_df.reset_index(drop=True)
     if periods is None:
         periods = 1 if max_lags == 0 else n_forecasts
     else:

neuralprophet/data/transform.py

@@ -24,13 +24,11 @@ def _normalize(df: pd.DataFrame, config_normalization: Normalization) -> pd.Data
     -------
         df: pd.DataFrame, normalized
     """
-    # df = df.copy(deep=True)
-    # df, _, _, _ = df_utils.check_multiple_series_id(df)
     df_norm = pd.DataFrame()
     for df_name, df_i in df.groupby("ID"):
         data_params = config_normalization.get_data_params(df_name)
         df_i.drop("ID", axis=1, inplace=True)
-        df_aux = df_utils.normalize(df_i, data_params).copy(deep=True)
+        df_aux = df_utils.normalize(df_i, data_params)
         df_aux["ID"] = df_name
         df_norm = pd.concat((df_norm, df_aux), ignore_index=True)
     return df_norm

neuralprophet/df_utils.py

@@ -77,12 +77,11 @@ def return_df_in_original_format(df, received_ID_col=False, received_single_time
         pd.Dataframe
             original input format
     """
-    new_df = df.copy(deep=True)
     if not received_ID_col and received_single_time_series:
-        assert len(new_df["ID"].unique()) == 1
-        new_df.drop("ID", axis=1, inplace=True)
+        assert len(df["ID"].unique()) == 1
+        df.drop("ID", axis=1, inplace=True)
         log.info("Returning df with no ID column")
-    return new_df
+    return df
 
 
 def merge_dataframes(df: pd.DataFrame) -> pd.DataFrame:
@@ -102,7 +101,7 @@ def merge_dataframes(df: pd.DataFrame) -> pd.DataFrame:
         raise ValueError("Can not join other than pd.DataFrames")
     if "ID" not in df.columns:
         raise ValueError("df does not contain 'ID' column")
-    df_merged = df.copy(deep=True).drop("ID", axis=1)
+    df_merged = df.drop("ID", axis=1)
     df_merged = df_merged.sort_values("ds")
     df_merged = df_merged.drop_duplicates(subset=["ds"])
     df_merged = df_merged.reset_index(drop=True)
@@ -282,11 +281,8 @@ def init_data_params(
             ShiftScale entries containing ``shift`` and ``scale`` parameters for each column
     """
     # Compute Global data params
-    # df = df.copy(deep=True)
-    # df, _, _, _ = check_multiple_series_id(df)
-    df_merged = df.copy(deep=True).drop("ID", axis=1)
     global_data_params = data_params_definition(
-        df_merged, normalize, config_lagged_regressors, config_regressors, config_events, config_seasonality
+        df, normalize, config_lagged_regressors, config_regressors, config_events, config_seasonality
     )
     if global_normalization:
         log.debug(
@@ -296,7 +292,6 @@ def init_data_params(
     local_data_params = OrderedDict()
     local_run_despite_global = True if global_normalization else None
     for df_name, df_i in df.groupby("ID"):
-        df_i.drop("ID", axis=1, inplace=True)
         local_data_params[df_name] = data_params_definition(
             df=df_i,
             normalize=normalize,
@@ -378,7 +373,6 @@ def normalize(df, data_params):
         pd.DataFrame
             normalized dataframes
     """
-    df = df.copy(deep=True)
     for name in df.columns:
         if name == "ID":
             continue
@@ -428,8 +422,7 @@ def check_dataframe(
         pd.DataFrame or dict
             checked dataframe
     """
-    # df = df.copy(deep=True)
-    # df, _, _, _ = check_multiple_series_id(df)
+    # TODO: move call to check_multiple_series_id here
     if df.groupby("ID").size().min() < 1:
         raise ValueError("Dataframe has no rows.")
     if "ds" not in df:
@@ -542,7 +535,7 @@ def _crossvalidation_split_df(df, n_lags, n_forecasts, k, fold_pct, fold_overlap
     min_train = total_samples - samples_fold - (k - 1) * (samples_fold - samples_overlap)
     assert min_train >= samples_fold
     folds = []
-    df_fold = df.copy(deep=True)
+    df_fold = df
     for i in range(k, 0, -1):
         df_train, df_val = split_df(df_fold, n_lags, n_forecasts, valid_p=samples_fold, inputs_overbleed=True)
         folds.append((df_train, df_val))
@@ -635,33 +628,30 @@ def _crossvalidation_with_time_threshold(df, n_lags, n_forecasts, k, fold_pct, f
 
             validation data
     """
-    df_merged = merge_dataframes(df)
+    df_merged = merge_dataframes(df.copy(deep=True))
     total_samples = len(df_merged) - n_lags + 2 - (2 * n_forecasts)
     samples_fold = max(1, int(fold_pct * total_samples))
     samples_overlap = int(fold_overlap_pct * samples_fold)
     assert samples_overlap < samples_fold
     min_train = total_samples - samples_fold - (k - 1) * (samples_fold - samples_overlap)
     assert min_train >= samples_fold
     folds = []
-    df_fold = df
-    # df_fold = df.copy(deep=True)
-    # df_fold, _, _, _ = check_multiple_series_id(df_fold)
     for i in range(k, 0, -1):
-        threshold_time_stamp = find_time_threshold(df_fold, n_lags, n_forecasts, samples_fold, inputs_overbleed=True)
+        threshold_time_stamp = find_time_threshold(df, n_lags, n_forecasts, samples_fold, inputs_overbleed=True)
         df_train, df_val = split_considering_timestamp(
-            df_fold, n_lags, n_forecasts, inputs_overbleed=True, threshold_time_stamp=threshold_time_stamp
+            df, n_lags, n_forecasts, inputs_overbleed=True, threshold_time_stamp=threshold_time_stamp
         )
         folds.append((df_train, df_val))
         split_idx = len(df_merged) - samples_fold + samples_overlap
         df_merged = df_merged[:split_idx].reset_index(drop=True)
         threshold_time_stamp = df_merged["ds"].iloc[-1]
         df_fold_aux = pd.DataFrame()
-        for df_name, df_i in df_fold.groupby("ID"):
-            df_aux = (
-                df_i.copy(deep=True).iloc[: len(df_i[df_i["ds"] < threshold_time_stamp]) + 1].reset_index(drop=True)
-            )
+        for df_name, df_i in df.groupby("ID"):
+            # df_i = df_i.copy(deep=True)
+            df_aux = df_i.iloc[: len(df_i[df_i["ds"] < threshold_time_stamp]) + 1].reset_index(drop=True)
             df_fold_aux = pd.concat((df_fold_aux, df_aux), ignore_index=True)
-        df_fold = df_fold_aux.copy(deep=True)
+        df = df_fold_aux
+        # df = df.copy(deep=True)
     folds = folds[::-1]
     return folds
 
@@ -707,7 +697,6 @@ def crossvalidation_split_df(
 
             validation data
     """
-    # df = df.copy(deep=True)
     df, _, _, _ = check_multiple_series_id(df)
     folds = []
     if len(df["ID"].unique()) == 1:
@@ -733,7 +722,7 @@ def crossvalidation_split_df(
             start_date, end_date = find_valid_time_interval_for_cv(df)
             for df_name, df_i in df.groupby("ID"):
                 mask = (df_i["ds"] >= start_date) & (df_i["ds"] <= end_date)
-                df_i = df_i[mask].copy(deep=True)
+                df_i = df_i[mask]
                 folds_dict[df_name] = _crossvalidation_split_df(
                     df_i, n_lags, n_forecasts, k, fold_pct, fold_overlap_pct
                 )
@@ -768,8 +757,6 @@ def double_crossvalidation_split_df(df, n_lags, n_forecasts, k, valid_pct, test_
         tuple of k tuples [(folds_val, folds_test), …]
             elements same as :meth:`crossvalidation_split_df` returns
     """
-    # df = df.copy(deep=True)
-    # df, _, _, _ = check_multiple_series_id(df)
     if len(df["ID"].unique()) > 1:
         raise NotImplementedError("double_crossvalidation_split_df not implemented for df with many time series")
     fold_pct_test = float(test_pct) / k
@@ -800,7 +787,7 @@ def find_time_threshold(df, n_lags, n_forecasts, valid_p, inputs_overbleed):
         str
             time stamp threshold defines the boundary for the train and validation sets split.
     """
-    df_merged = merge_dataframes(df)
+    df_merged = merge_dataframes(df.copy(deep=True))
     n_samples = len(df_merged) - n_lags + 2 - (2 * n_forecasts)
     n_samples = n_samples if inputs_overbleed else n_samples - n_lags
     if 0.0 < valid_p < 1.0:
@@ -842,11 +829,14 @@ def split_considering_timestamp(df, n_lags, n_forecasts, inputs_overbleed, thres
     df_val = pd.DataFrame()
     for df_name, df_i in df.groupby("ID"):
         if df[df["ID"] == df_name]["ds"].max() < threshold_time_stamp:
-            df_train = pd.concat((df_train, df_i.copy(deep=True)), ignore_index=True)
+            # df_i = df_i.copy(deep=True)
+            df_train = pd.concat((df_train, df_i), ignore_index=True)
         elif df[df["ID"] == df_name]["ds"].min() > threshold_time_stamp:
-            df_val = pd.concat((df_val, df_i.copy(deep=True)), ignore_index=True)
+            # df_i = df_i.copy(deep=True)
+            df_val = pd.concat((df_val, df_i), ignore_index=True)
         else:
-            df_aux = df_i.copy(deep=True)
+            df_aux = df_i
+            # df_i = df_i.copy(deep=True)
             n_train = len(df_aux[df_aux["ds"] < threshold_time_stamp])
             split_idx_train = n_train + n_lags + n_forecasts - 1
             split_idx_val = split_idx_train - n_lags if inputs_overbleed else split_idx_train
@@ -890,8 +880,6 @@ def split_df(
         pd.DataFrame, dict
             validation data
     """
-    # df = df.copy(deep=True)
-    # df, _, _, _ = check_multiple_series_id(df)
     df_train = pd.DataFrame()
     df_val = pd.DataFrame()
     if local_split:
@@ -1373,8 +1361,6 @@ def infer_frequency(df, freq, n_lags, min_freq_percentage=0.7):
             Valid frequency tag according to major frequency.
 
     """
-    # df = df.copy(deep=True)
-    # df, _, _, _ = check_multiple_series_id(df)
     freq_df = list()
     for df_name, df_i in df.groupby("ID"):
         freq_df.append(_infer_frequency(df_i, freq, min_freq_percentage))
@@ -1396,6 +1382,7 @@ def create_dict_for_events_or_regressors(
     df: pd.DataFrame,
     other_df: Optional[pd.DataFrame],
     other_df_name: str,
+    received_ID_col: bool,
 ) -> dict:  # Not sure about the naming of this function
     """Create a dict for events or regressors according to input df.
 
@@ -1417,12 +1404,10 @@ def create_dict_for_events_or_regressors(
     if other_df is None:
         # if other_df is None, create dictionary with None for each ID
         return {df_name: None for df_name in df_names}
-    other_df = other_df.copy(deep=True)
-    other_df, received_ID_col, _, _ = check_multiple_series_id(other_df)
     # if other_df does not contain ID, create dictionary with original ID with the same other_df for each ID
     if not received_ID_col:
         other_df = other_df.drop("ID", axis=1)
-        return {df_name: other_df.copy(deep=True) for df_name in df_names}
+        return {df_name: other_df for df_name in df_names}
 
     # else, other_df does contain ID, create dict with respective IDs
     df_unique_names, other_df_unique_names = list(df["ID"].unique()), list(other_df["ID"].unique())
@@ -1438,7 +1423,7 @@ def create_dict_for_events_or_regressors(
     df_other_dict = {}
     for df_name in df_unique_names:
         if df_name in other_df_unique_names:
-            df_aux = other_df[other_df["ID"] == df_name].reset_index(drop=True).copy(deep=True)
+            df_aux = other_df[other_df["ID"] == df_name].reset_index(drop=True)
             df_aux.drop("ID", axis=1, inplace=True)
         else:
             df_aux = None