Align binary outcome dictionaries by population key

src/rtichoke/performance_data/performance_data.py

@@ -44,6 +44,47 @@
 ]
 
 
+def _validate_and_align_binary_inputs(
+    probs: Dict[str, np.ndarray],
+    reals: Union[np.ndarray, Dict[str, np.ndarray]],
+) -> Union[np.ndarray, Dict[str, np.ndarray]]:
+    """Validate binary input alignment and normalize outcome dictionaries."""
+    if not isinstance(probs, dict) or not probs:
+        raise ValueError("`probs` must be a non-empty dictionary of probability arrays.")
+
+    groups = list(probs)
+
+    if isinstance(reals, dict):
+        expected_keys = set(groups)
+        if set(reals) != expected_keys:
+            raise ValueError("`reals` dictionary keys must exactly match `probs`.")
+
+        for group in groups:
+            n_probs = len(np.asarray(probs[group]))
+            n_reals = len(np.asarray(reals[group]))
+            if n_probs != n_reals:
+                raise ValueError(
+                    f"Input lengths must match within group {group!r}: "
+                    f"len(probs)={n_probs}, len(reals)={n_reals}."
+                )
+
+        if len(groups) == 1:
+            return np.asarray(reals[groups[0]])
+
+        return {group: np.asarray(reals[group]) for group in groups}
+
+    n_reals = len(np.asarray(reals))
+    for group in groups:
+        n_probs = len(np.asarray(probs[group]))
+        if n_probs != n_reals:
+            raise ValueError(
+                f"Shared outcome length must match probabilities for group {group!r}: "
+                f"len(probs)={n_probs}, len(reals)={n_reals}."
+            )
+
+    return reals
+
+
 def _probs_with_r_binary_cutoff_semantics(
     probs: Dict[str, np.ndarray], by: float
 ) -> Dict[str, np.ndarray]:
@@ -113,6 +154,7 @@ def prepare_binned_classification_data(
         any other stratification variables. It forms the basis for subsequent
         performance calculations.
     """
+    reals = _validate_and_align_binary_inputs(probs=probs, reals=reals)
     breaks = create_breaks_values(None, "probability_threshold", by)
 
     aj_data_combinations = _create_aj_data_combinations_binary(

tests/test_binary_input_validation.py

@@ -0,0 +1,69 @@
+import numpy as np
+import pytest
+from polars.testing import assert_frame_equal
+
+from rtichoke import prepare_performance_data
+
+
+def test_binary_reals_dict_is_aligned_by_group_key_not_insertion_order():
+    probs = {
+        "group_a": np.array([0.1, 0.8, 0.7, 0.2]),
+        "group_b": np.array([0.9, 0.2, 0.3, 0.8]),
+    }
+    reals_a = np.array([0, 1, 1, 0])
+    reals_b = np.array([1, 0, 0, 1])
+
+    aligned = prepare_performance_data(
+        probs=probs,
+        reals={"group_a": reals_a, "group_b": reals_b},
+        by=0.5,
+    )
+    reversed_order = prepare_performance_data(
+        probs=probs,
+        reals={"group_b": reals_b, "group_a": reals_a},
+        by=0.5,
+    )
+
+    assert_frame_equal(aligned, reversed_order)
+
+
+def test_binary_single_group_reals_dict_matches_array_input():
+    probs = {"group_a": np.array([0.1, 0.8, 0.7, 0.2])}
+    reals = np.array([0, 1, 1, 0])
+
+    from_array = prepare_performance_data(probs=probs, reals=reals, by=0.5)
+    from_dict = prepare_performance_data(
+        probs=probs,
+        reals={"group_a": reals},
+        by=0.5,
+    )
+
+    assert_frame_equal(from_array, from_dict)
+
+
+def test_binary_reals_dict_keys_must_match_probability_groups():
+    probs = {
+        "group_a": np.array([0.1, 0.8]),
+        "group_b": np.array([0.9, 0.2]),
+    }
+
+    with pytest.raises(ValueError, match="keys must exactly match"):
+        prepare_performance_data(
+            probs=probs,
+            reals={
+                "group_a": np.array([0, 1]),
+                "wrong_group": np.array([1, 0]),
+            },
+            by=0.5,
+        )
+
+
+def test_binary_input_lengths_must_match_within_group():
+    probs = {"group_a": np.array([0.1, 0.8, 0.7])}
+
+    with pytest.raises(ValueError, match="Input lengths must match"):
+        prepare_performance_data(
+            probs=probs,
+            reals={"group_a": np.array([0, 1])},
+            by=0.5,
+        )