Fix NaN upper-bound check in strict metrics NotIn

pyiceberg/expressions/visitors.py

@@ -1784,6 +1784,11 @@ def visit_not_in(self, term: BoundTerm, literals: set[L]) -> bool:
         if upper_bytes is not None:
             upper = _from_byte_buffer(field.field_type, upper_bytes)
 
+            if self._is_nan(upper):
+                # NaN indicates unreliable bounds.
+                # See the StrictMetricsEvaluator docs for more.
+                return ROWS_MIGHT_NOT_MATCH
+
             literals = {val for val in literals if upper >= val}
 
             if len(literals) == 0:

tests/expressions/test_evaluator.py

@@ -1358,6 +1358,31 @@ def test_strict_not_equal_and_not_in_with_mixed_nans_and_matching_bounds(field_t
     assert should_read == ROWS_MIGHT_NOT_MATCH, "Should not match: bounds prove the non-NaN value is 5.0"
 
 
+@pytest.mark.parametrize("field_type", [FloatType(), DoubleType()])
+def test_strict_not_in_with_nan_upper_bound(field_type: PrimitiveType) -> None:
+    schema = Schema(NestedField(1, "x", field_type, required=False))
+    # Column contains {1.0, NaN}: min is 1.0, max is NaN (NaN sorts greatest).
+    # The non-NaN row 1.0 is in the literal set, so the file cannot be proven
+    # to fully match NotIn, even though the NaN count is known.
+    data_file = DataFile.from_args(
+        file_path="file.parquet",
+        file_format=FileFormat.PARQUET,
+        partition={},
+        record_count=2,
+        file_size_in_bytes=1,
+        value_counts={1: 2},
+        null_value_counts={1: 0},
+        nan_value_counts={1: 1},
+        lower_bounds={1: to_bytes(field_type, 1.0)},
+        upper_bounds={1: to_bytes(field_type, float("nan"))},
+    )
+
+    should_read = _StrictMetricsEvaluator(schema, NotIn("x", {1.0, 2.0})).eval(data_file)
+    assert should_read == ROWS_MIGHT_NOT_MATCH, (
+        "NaN upper bound makes the bounds unusable: the non-NaN row 1.0 is in the literal set"
+    )
+
+
 @pytest.mark.parametrize("field_type", [FloatType(), DoubleType()])
 def test_strict_not_equal_and_not_in_with_all_nans(field_type: PrimitiveType) -> None:
     schema = Schema(NestedField(1, "x", field_type, required=False))