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

[Otis0408]

Summary

AnswerExactMatchEvaluator, DocumentMAPEvaluator, DocumentMRREvaluator and DocumentRecallEvaluator average their per-query scores by dividing by the number of input queries, without guarding against an empty batch. Two empty lists satisfy the equal-length check these evaluators perform, so the code proceeds to sum(...) / len(...) and crashes with an opaque ZeroDivisionError: division by zero.

DocumentNDCGEvaluator already handles this case: its validate_inputs rejects empty input with a descriptive ValueError ("ground_truth_documents and retrieved_documents must be provided."). This PR makes the other four evaluators consistent with that contract, so an empty batch surfaces a clear, actionable error instead of a bare arithmetic failure.

Before

from haystack.components.evaluators import DocumentRecallEvaluator

DocumentRecallEvaluator().run(ground_truth_documents=[], retrieved_documents=[])
# ZeroDivisionError: division by zero

After

from haystack.components.evaluators import DocumentRecallEvaluator

DocumentRecallEvaluator().run(ground_truth_documents=[], retrieved_documents=[])
# ValueError: ground_truth_documents and retrieved_documents must be provided.

The same change applies to AnswerExactMatchEvaluator (message: ground_truth_answers and predicted_answers must be provided.), DocumentMAPEvaluator and DocumentMRREvaluator.

Tests

Added test/components/evaluators/test_evaluators_empty_input.py, a parametrized regression test asserting all four evaluators raise ValueError on empty input. It fails on the pre-fix source (ZeroDivisionError) and passes with the fix. All existing evaluator test suites continue to pass.

[vercel]

@Otis0408 is attempting to deploy a commit to the deepset Team on Vercel.

A member of the Team first needs to authorize it.

[github-actions]

Coverage report

Click to see where and how coverage changed

File	Statements	Missing	Coverage	Coverage (new stmts)	Lines missing
haystack/components/evaluators
answer_exact_match.py
document_map.py
document_mrr.py
document_recall.py
Project Total

This report was generated by python-coverage-comment-action

[bogdankostic]

Looks already good in principle, let's just place the tests in the existing evaluator test files.

[bogdankostic]

Let's not add this test file containing only a single test function. Instead, let's add this test to the existing evaluator test files.