Merge branch 'master' into issue-528

CHANGELOG.md

@@ -42,9 +42,9 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - Add parameter `missing_mode` into `R2` and `MedAE` metrics ([#537](https://github.com/etna-team/etna/pull/537))
 - Update `analysis.forecast.plots.plot_metric_per_segment` to handle `None` from metrics ([#540](https://github.com/etna-team/etna/pull/540))
 - Add parameter `missing_mode` into `RMSE` and `MSLE` metrics ([#542](https://github.com/etna-team/etna/pull/542))
+- Update `analysis.forecast.plots.metric_per_segment_distribution_plot` to handle `None` from metrics ([#543](https://github.com/etna-team/etna/pull/543))
 - 
-- 
-- 
+- Add example on using custom `params_to_tune` in `Tune` ([#547](https://github.com/etna-team/etna/pull/547))
 
 ### Fixed
 - Fix working with `embedding_sizes` in `202-NN_examples` notebook ([#489](https://github.com/etna-team/etna/pull/489))

etna/analysis/forecast/plots.py

@@ -726,7 +726,7 @@ def plot_metric_per_segment(
     Warnings
     --------
     UserWarning:
-        There are segments without non-missing metric values.
+        There are segments with all missing metric values.
     UserWarning:
         Some segments have different set of folds to be aggregated on due to missing values.
     """
@@ -803,7 +803,12 @@ def metric_per_segment_distribution_plot(
     seaborn_params: Optional[Dict[str, Any]] = None,
     figsize: Tuple[int, int] = (10, 5),
 ):
-    """Plot per-segment metrics distribution.
+    """Plot distribution of metric values over all segments.
+
+    If for some segment all metric values are missing, it isn't plotted, and the warning is raised.
+
+    If some segments have different set of folds with non-missing metrics,
+    it can lead to incompatible values between folds. The warning is raised in such case.
 
     Parameters
     ----------
@@ -831,6 +836,13 @@ def metric_per_segment_distribution_plot(
         if ``metric_name`` isn't present in ``metrics_df``
     NotImplementedError:
         unknown ``per_fold_aggregation_mode`` is given
+
+    Warnings
+    --------
+    UserWarning:
+        There are segments with all missing metric values.
+    UserWarning:
+        Some segments have different set of folds to be aggregated on due to missing values.
     """
     if seaborn_params is None:
         seaborn_params = {}
@@ -844,6 +856,9 @@ def metric_per_segment_distribution_plot(
     if metric_name not in metrics_df.columns:
         raise ValueError("Given metric_name isn't present in metrics_df")
 
+    _check_metrics_df_empty_segments(metrics_df=metrics_df, metric_name=metric_name)
+    _check_metrics_df_same_folds_for_each_segment(metrics_df=metrics_df, metric_name=metric_name)
+
     # draw plot for each fold
     if per_fold_aggregation_mode is None and "fold_number" in metrics_df.columns:
         if plot_type_enum == MetricPlotType.hist:

etna/analysis/forecast/utils.py

@@ -145,6 +145,6 @@ def _check_metrics_df_same_folds_for_each_segment(metrics_df: pd.DataFrame, metr
     df = metrics_df[["segment", "fold_number", metric_name]]
     # we don't take into account segments without any non-missing metrics, they are handled by other check
     df = df.dropna(subset=[metric_name])
-    num_unique = df.groupby("segment")["fold_number"].apply(frozenset).nunique()
+    num_unique = df.groupby("segment", group_keys=False)["fold_number"].apply(frozenset).nunique()
     if num_unique > 1:
         warnings.warn("Some segments have different set of folds to be aggregated on due to missing values.")