import sys
from pathlib import Path
import pytest
# Add project root to path for benchmark imports
sys.path.insert(0, str(Path(__file__).parent.parent.parent))
from benchmark.metrics import compute_f_score, compute_joint_f_score
class TestComputeFScore:
def test_perfect_scores(self) -> None:
result = compute_f_score(1.0, 1.0, beta=1.0)
assert result == pytest.approx(1.0)
def test_zero_precision(self) -> None:
assert compute_f_score(0.0, 1.0, beta=1.0) == 0.0
def test_zero_recall(self) -> None:
assert compute_f_score(1.0, 0.0, beta=1.0) == 0.0
def test_f1_balanced(self) -> None:
# F1 with equal precision and recall should equal them
result = compute_f_score(0.8, 0.8, beta=1.0)
assert result == pytest.approx(0.8)
def test_f1_unbalanced(self) -> None:
# F1 = 2 * P * R / (P + R) = 2 * 0.8 * 0.6 / 1.4 ≈ 0.6857
result = compute_f_score(0.8, 0.6, beta=1.0)
expected = 2 * 0.8 * 0.6 / (0.8 + 0.6)
assert result == pytest.approx(expected)
def test_beta_half_emphasizes_precision(self) -> None:
# β=0.5: precision weighted more heavily
# Fβ = (1 + β²) * P * R / (β² * P + R)
# F0.5 = 1.25 * 0.9 * 0.5 / (0.25 * 0.9 + 0.5)
result = compute_f_score(0.9, 0.5, beta=0.5)
expected = 1.25 * 0.9 * 0.5 / (0.25 * 0.9 + 0.5)
assert result == pytest.approx(expected)
# With β=0.5, high precision + low recall should score well
assert result > 0.6
def test_beta_two_emphasizes_recall(self) -> None:
# β=2: recall weighted more heavily
result_high_recall = compute_f_score(0.5, 0.9, beta=2.0)
result_high_prec = compute_f_score(0.9, 0.5, beta=2.0)
# High recall should score better with β=2
assert result_high_recall > result_high_prec
class TestComputeJointFScore:
def test_empty_returned(self) -> None:
result = compute_joint_f_score({}, {"a.py": [(1, 10)]})
assert result["file_precision"] == 0.0
assert result["file_recall"] == 0.0
assert result["joint_f"] == 0.0
def test_perfect_match(self) -> None:
files = {"a.py": [[1, 10]]}
gt = {"a.py": [(1, 10)]}
result = compute_joint_f_score(files, gt, beta=1.0)
assert result["file_precision"] == pytest.approx(1.0)
assert result["file_recall"] == pytest.approx(1.0)
assert result["file_f"] == pytest.approx(1.0)
assert result["line_precision"] == pytest.approx(1.0)
assert result["line_recall"] == pytest.approx(1.0)
assert result["line_f"] == pytest.approx(1.0)
assert result["joint_f"] == pytest.approx(1.0)
def test_partial_file_match(self) -> None:
files = {"a.py": [[1, 10]], "b.py": [[1, 5]]}
gt = {"a.py": [(1, 10)]}
result = compute_joint_f_score(files, gt, beta=1.0)
# File precision = 1/2 = 0.5, recall = 1/1 = 1.0
assert result["file_precision"] == pytest.approx(0.5)
assert result["file_recall"] == pytest.approx(1.0)
def test_beta_affects_f_scores(self) -> None:
files = {"a.py": [[1, 10]]}
gt = {"a.py": [(1, 5)], "b.py": [(1, 10)]}
result_beta_half = compute_joint_f_score(files, gt, beta=0.5)
result_beta_two = compute_joint_f_score(files, gt, beta=2.0)
# With higher file precision than recall, β=0.5 should give higher score
assert result_beta_half["file_f"] > result_beta_two["file_f"]
def test_file_weight(self) -> None:
files = {"a.py": [[1, 10]]}
gt = {"a.py": [(1, 10)]}
result_file_heavy = compute_joint_f_score(files, gt, file_weight=0.9)
result_line_heavy = compute_joint_f_score(files, gt, file_weight=0.1)
# Perfect match, both should be 1.0
assert result_file_heavy["joint_f"] == pytest.approx(1.0)
assert result_line_heavy["joint_f"] == pytest.approx(1.0)
def test_returns_all_keys(self) -> None:
result = compute_joint_f_score({}, {})
expected_keys = {
"file_precision",
"file_recall",
"file_f",
"line_precision",
"line_recall",
"line_f",
"joint_f",
}
assert set(result.keys()) == expected_keys
