feat: apply copilot review

Author: dawidlinek

epftoolbox2/evaluators/rmae.py

@@ -1,3 +1,4 @@
+import math
 from typing import Dict
 
 import pandas as pd
@@ -12,12 +13,15 @@ def __init__(self, base_model: str):
 
     def compute(self, df: pd.DataFrame, **kwargs) -> float:
         model_dfs: Dict[str, pd.DataFrame] = kwargs.get("model_dfs", {})
-        if self.base_model not in model_dfs:
+        if not model_dfs:
             raise ValueError(
-                f"rMAE base model '{self.base_model}' not found in pipeline models. "
-                f"Available: {list(model_dfs)}"
+                f"rMAE base model '{self.base_model}' not found in pipeline models."
             )
+        if self.base_model not in model_dfs:
+            return math.nan
         base_df = model_dfs[self.base_model]
+        if base_df.empty:
+            return math.nan
         base_mae = (base_df["prediction"] - base_df["actual"]).abs().mean()
         if base_mae == 0:
             return float("inf")

epftoolbox2/exporters/csv.py

@@ -6,6 +6,8 @@
 from .base import Exporter
 from ..results.report import EvaluationReport
 
+_MERGE_KEYS = ["run_date", "target_date", "hour", "horizon"]
+
 
 class CsvExporter(Exporter):
     def __init__(self, path: str, extra_columns: Optional[List[str]] = None):
@@ -25,21 +27,25 @@ def export(self, report: EvaluationReport) -> None:
 
         self.path.parent.mkdir(parents=True, exist_ok=True)
 
-        sort_keys = ["target_date", "hour", "horizon"]
         base_cols = ["run_date", "target_date", "hour", "horizon", "day_in_test", "actual"]
-
         base_df: Optional[pd.DataFrame] = None
-        model_names: List[str] = []
 
         for model_name, model_df in report.iter_details():
-            model_df = model_df.sort_values(by=sort_keys).reset_index(drop=True)
-            if base_df is None:
-                base_df = model_df[base_cols].copy()
-            base_df[f"{model_name}_prediction"] = model_df["prediction"].values
-            base_df[f"{model_name}_error"] = (
-                model_df["prediction"].values - base_df["actual"].values
+            model_df = model_df.rename(columns={
+                "prediction": f"{model_name}_prediction",
+            })
+            model_df[f"{model_name}_error"] = (
+                model_df[f"{model_name}_prediction"] - model_df["actual"]
             )
-            model_names.append(model_name)
+            keep = _MERGE_KEYS + [f"{model_name}_prediction", f"{model_name}_error"]
+            if base_df is None:
+                base_df = model_df[base_cols + [f"{model_name}_prediction", f"{model_name}_error"]].copy()
+            else:
+                base_df = base_df.merge(
+                    model_df[keep],
+                    on=_MERGE_KEYS,
+                    how="outer",
+                )
             del model_df
 
         if base_df is None:
@@ -48,6 +54,7 @@ def export(self, report: EvaluationReport) -> None:
         if self.extra_columns and report.source_data is not None:
             base_df = self._join_extra_columns(base_df, report.source_data)
 
+        sort_keys = ["target_date", "hour", "horizon"]
         base_df = base_df.sort_values(by=sort_keys).reset_index(drop=True)
         base_df.to_csv(self.path, index=False)
 

epftoolbox2/results/report.py

@@ -1,6 +1,6 @@
 from __future__ import annotations
 
-from typing import Dict, Iterator, List, Tuple, Union
+from typing import Dict, Iterator, List, Optional, Tuple, Union
 
 import pandas as pd
 
@@ -13,7 +13,7 @@ def __init__(
         self,
         results_or_refs: Dict[str, Union[ModelResultRef, List[Dict]]],
         evaluators: List[Evaluator],
-        source_data: pd.DataFrame = None,
+        source_data: Optional[pd.DataFrame] = None,
     ):
         self.evaluators = evaluators
         self.source_data = source_data