Allow each metric to return multiple values instead of only a single …

…scalar

ieegprep/bids/data_epoch.py

@@ -651,10 +651,10 @@ def _load_data_epoch_averages__by_condition_trials(data_reader, retrieve_channel
         if metric_callbacks is not None:
             if callable(metric_callbacks):
                 metric_values = allocate_array((len(retrieve_channels), len(conditions_onsets)),
-                                               fill_value=np.nan, dtype=np.float64)
+                                               fill_value=np.nan, dtype=np.ndarray)
             elif type(metric_callbacks) is tuple and len(metric_callbacks) > 0:
                 metric_values = allocate_array((len(retrieve_channels), len(conditions_onsets), len(metric_callbacks)),
-                                               fill_value=np.nan, dtype=np.float64)
+                                               fill_value=np.nan, dtype=np.ndarray)
     except MemoryError:
         raise MemoryError('Not enough memory create metric output matrix')
 
@@ -811,9 +811,6 @@ def _load_data_epoch_averages__by_condition_trials(data_reader, retrieve_channel
                                                     None if baseline_data is None else baseline_data[channel_idx, :, :])
 
                     if metric_value is not None:
-                        if not np.isscalar(metric_value):
-                            logging.error('Return metric is not scalar')
-                            raise RuntimeError('Return metric is not scalar')
                         metric_values[channel_idx, condition_idx] = metric_value
 
                 elif type(metric_callbacks) is tuple and len(metric_callbacks) > 0:
@@ -825,9 +822,6 @@ def _load_data_epoch_averages__by_condition_trials(data_reader, retrieve_channel
                                                                        condition_data[channel_idx, :, :],
                                                                        None if baseline_data is None else baseline_data[channel_idx, :, :])
                             if metric_value is not None:
-                                if not np.isscalar(metric_value):
-                                    logging.error('Return metric is not scalar')
-                                    raise RuntimeError('Return metric is not scalar')
                                 metric_values[channel_idx, condition_idx, iCallback] = metric_value
 
             # the callback has been made, perform -if needed- the (postponed) normalization with the baseline values
@@ -1223,10 +1217,10 @@ def _load_data_epoch_averages__by_channel_condition_trial(data_reader, channels,
         if metric_callbacks is not None:
             if callable(metric_callbacks):
                 metric_values = allocate_array((len(channels), len(conditions_onsets)),
-                                               fill_value=np.nan, dtype=np.float64)
+                                               fill_value=np.nan, dtype=np.ndarray)
             elif type(metric_callbacks) is tuple and len(metric_callbacks) > 0:
                 metric_values = allocate_array((len(channels), len(conditions_onsets), len(metric_callbacks)),
-                                               fill_value=np.nan, dtype=np.float64)
+                                               fill_value=np.nan, dtype=np.ndarray)
     except MemoryError:
         raise MemoryError('Not enough memory create a metric output matrix')
 
@@ -1321,10 +1315,10 @@ def _load_data_epochs__by_channels__withPrep(average, data_reader, retrieve_chan
             if metric_callbacks is not None:
                 if callable(metric_callbacks):
                     metric_values = allocate_array((len(retrieve_channels), len(onsets)),
-                                                   fill_value=np.nan, dtype=np.float64)
+                                                   fill_value=np.nan, dtype=np.ndarray)
                 elif type(metric_callbacks) is tuple and len(metric_callbacks) > 0:
                     metric_values = allocate_array((len(retrieve_channels), len(onsets), len(metric_callbacks)),
-                                                   fill_value=np.nan, dtype=np.float64)
+                                                   fill_value=np.nan, dtype=np.ndarray)
         except MemoryError:
             raise MemoryError('Not enough memory create a metric output matrix')