fix: raise clear error for empty inputs in retrieval/answer evaluators

AnswerExactMatchEvaluator, DocumentMAPEvaluator, DocumentMRREvaluator and
DocumentRecallEvaluator averaged their per-query scores by dividing by the
input length without guarding against an empty batch. Passing two empty
lists (which satisfies the equal-length check) therefore raised an opaque
ZeroDivisionError. They now raise a descriptive ValueError, matching the
existing contract of DocumentNDCGEvaluator.

Author: otiscuilei

haystack/components/evaluators/answer_exact_match.py

@@ -53,6 +53,9 @@ def run(self, ground_truth_answers: list[str], predicted_answers: list[str]) ->
             - `score` - A number from 0.0 to 1.0 that represents the proportion of questions where any predicted
                          answer matched one of the ground truth answers.
         """
+        if len(ground_truth_answers) == 0 or len(predicted_answers) == 0:
+            raise ValueError("ground_truth_answers and predicted_answers must be provided.")
+
         if not len(ground_truth_answers) == len(predicted_answers):
             raise ValueError("The length of ground_truth_answers and predicted_answers must be the same.")
 

haystack/components/evaluators/document_map.py

@@ -108,6 +108,10 @@ def run(
             - `individual_scores` - A list of numbers from 0.0 to 1.0 that represents how high retrieved documents
                 are ranked.
         """
+        if len(ground_truth_documents) == 0 or len(retrieved_documents) == 0:
+            msg = "ground_truth_documents and retrieved_documents must be provided."
+            raise ValueError(msg)
+
         if len(ground_truth_documents) != len(retrieved_documents):
             msg = "The length of ground_truth_documents and retrieved_documents must be the same."
             raise ValueError(msg)

haystack/components/evaluators/document_mrr.py

@@ -106,6 +106,10 @@ def run(
             - `individual_scores` - A list of numbers from 0.0 to 1.0 that represents how high the first retrieved
                 document is ranked.
         """
+        if len(ground_truth_documents) == 0 or len(retrieved_documents) == 0:
+            msg = "ground_truth_documents and retrieved_documents must be provided."
+            raise ValueError(msg)
+
         if len(ground_truth_documents) != len(retrieved_documents):
             msg = "The length of ground_truth_documents and retrieved_documents must be the same."
             raise ValueError(msg)

haystack/components/evaluators/document_recall.py

@@ -158,6 +158,10 @@ def run(
             - `individual_scores` - A list of numbers from 0.0 to 1.0 that represents the proportion of matching
                 documents retrieved. If the mode is `single_hit`, the individual scores are 0 or 1.
         """
+        if len(ground_truth_documents) == 0 or len(retrieved_documents) == 0:
+            msg = "ground_truth_documents and retrieved_documents must be provided."
+            raise ValueError(msg)
+
         if len(ground_truth_documents) != len(retrieved_documents):
             msg = "The length of ground_truth_documents and retrieved_documents must be the same."
             raise ValueError(msg)

releasenotes/notes/evaluators-empty-input-value-error-c2e8b41d7a56f309.yaml

@@ -0,0 +1,7 @@
+---
+fixes:
+  - |
+    ``AnswerExactMatchEvaluator``, ``DocumentMAPEvaluator``, ``DocumentMRREvaluator`` and
+    ``DocumentRecallEvaluator`` now raise a descriptive ``ValueError`` when called with
+    empty input lists, instead of failing with an opaque ``ZeroDivisionError``. This
+    matches the existing behavior of ``DocumentNDCGEvaluator``.

test/components/evaluators/test_evaluators_empty_input.py

@@ -0,0 +1,30 @@
+# SPDX-FileCopyrightText: 2022-present deepset GmbH <info@deepset.ai>
+#
+# SPDX-License-Identifier: Apache-2.0
+
+import pytest
+
+from haystack.components.evaluators import (
+    AnswerExactMatchEvaluator,
+    DocumentMAPEvaluator,
+    DocumentMRREvaluator,
+    DocumentRecallEvaluator,
+)
+
+
+@pytest.mark.parametrize(
+    "evaluator, kwargs",
+    [
+        (AnswerExactMatchEvaluator(), {"ground_truth_answers": [], "predicted_answers": []}),
+        (DocumentMAPEvaluator(), {"ground_truth_documents": [], "retrieved_documents": []}),
+        (DocumentMRREvaluator(), {"ground_truth_documents": [], "retrieved_documents": []}),
+        (DocumentRecallEvaluator(), {"ground_truth_documents": [], "retrieved_documents": []}),
+    ],
+)
+def test_run_with_empty_inputs_raises_value_error(evaluator, kwargs):
+    # Empty (equal-length) inputs previously fell through the length check and
+    # crashed with a bare ZeroDivisionError when averaging over zero items.
+    # They must instead raise a descriptive ValueError, matching
+    # DocumentNDCGEvaluator's contract.
+    with pytest.raises(ValueError, match="must be provided"):
+        evaluator.run(**kwargs)