From b93ab6fb5c30c8cbd9990afd646fd435adc1c589 Mon Sep 17 00:00:00 2001
From: rileyh <rileyh@umn.edu>
Date: Wed, 11 Dec 2024 14:53:02 -0600
Subject: [PATCH] [#179] Add hypothesis and some property tests for
 core.model_metrics

---
 hlink/tests/core/model_metrics_test.py | 46 ++++++++++++++++++++++++++
 pyproject.toml                         |  1 +
 2 files changed, 47 insertions(+)

diff --git a/hlink/tests/core/model_metrics_test.py b/hlink/tests/core/model_metrics_test.py
index c8d046d..7f861a8 100644
--- a/hlink/tests/core/model_metrics_test.py
+++ b/hlink/tests/core/model_metrics_test.py
@@ -2,9 +2,15 @@
 # For copyright and licensing information, see the NOTICE and LICENSE files
 # in this project's top-level directory, and also on-line at:
 #   https://github.com/ipums/hlink
+import math
+
+from hypothesis import assume, given
+import hypothesis.strategies as st
 
 from hlink.linking.core.model_metrics import mcc, precision, recall
 
+NonNegativeInt = st.integers(min_value=0)
+
 
 def test_mcc_example() -> None:
     tp = 3112
@@ -26,6 +32,26 @@ def test_precision_example() -> None:
     ), "expected precision to be near 0.9381972"
 
 
+@given(true_pos=NonNegativeInt, false_pos=NonNegativeInt)
+def test_precision_between_0_and_1(true_pos: int, false_pos: int) -> None:
+    """
+    Under "normal circumstances" (there were at least some positive predictions)
+    precision()'s range is the interval [0.0, 1.0].
+    """
+    assume(true_pos + false_pos > 0)
+    precision_score = precision(true_pos, false_pos)
+    assert 0.0 <= precision_score <= 1.0
+
+
+def test_precision_no_positive_predictions() -> None:
+    """
+    When there are no positive predictions, true_pos=0 and false_pos=0, and
+    precision is not well defined. In this case we return NaN.
+    """
+    precision_score = precision(0, 0)
+    assert math.isnan(precision_score)
+
+
 def test_recall_example() -> None:
     tp = 3112
     fn = 1134
@@ -34,3 +60,23 @@ def test_recall_example() -> None:
     assert (
         abs(recall_score - 0.7329251) < 0.0001
     ), "expected recall to be near 0.7329251"
+
+
+@given(true_pos=NonNegativeInt, false_neg=NonNegativeInt)
+def test_recall_between_0_and_1(true_pos: int, false_neg: int) -> None:
+    """
+    Under "normal circumstances" (there is at least one true positive or false
+    negative), the range of recall() is the interval [0.0, 1.0].
+    """
+    assume(true_pos + false_neg > 0)
+    recall_score = recall(true_pos, false_neg)
+    assert 0.0 <= recall_score <= 1.0
+
+
+def test_recall_no_true_pos_or_false_neg() -> None:
+    """
+    When both true_pos and false_neg are 0, recall is not well defined, and we
+    return NaN.
+    """
+    recall_score = recall(0, 0)
+    assert math.isnan(recall_score)
diff --git a/pyproject.toml b/pyproject.toml
index 2a4b001..5c13c39 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -30,6 +30,7 @@ dependencies = [
 [project.optional-dependencies]
 dev = [
     "pytest>=7.1.0",
+    "hypothesis>=6.0",
     "black>=23.0",
     "flake8>=5.0",
     "pre-commit>=2.0",