..

tests/benchmarks/test_hot_paths.py

@@ -6,10 +6,18 @@ to establish baselines and detect regressions.
 Run with: pytest tests/benchmarks/ --benchmark-enable
 Compare: pytest tests/benchmarks/ --benchmark-compare
 Save baseline: pytest tests/benchmarks/ --benchmark-save=baseline
+
+Note: These tests are marked as slow and excluded from CI unit test runs.
+Run explicitly with: pytest tests/benchmarks/ -m slow
 """
 
 from __future__ import annotations
 
+import pytest
+
+# Mark all tests in this module as slow (excluded from CI unit tests)
+pytestmark = pytest.mark.slow
+
 import numpy as np
 import pytest
 from typing import TYPE_CHECKING
@@ -70,7 +78,9 @@ def typed_benchmark[T](
     return cast(T, _run_benchmark(benchmark, func, *args))
 
 
-def benchmark_array(benchmark: BenchmarkFixture, func: object, *args: object) -> NDArray[np.float32]:
+def benchmark_array(
+    benchmark: BenchmarkFixture, func: object, *args: object
+) -> NDArray[np.float32]:
     """Run benchmark for functions returning float32 arrays.
 
     BenchmarkFixture.__call__ is untyped. This wrapper returns a properly typed NDArray.
@@ -117,9 +127,7 @@ def benchmark_array_list(
     return cast(list[NDArray[np.float32]], _run_benchmark(benchmark, func, *args))
 
 
-def benchmark_float_list(
-    benchmark: BenchmarkFixture, func: object, *args: object
-) -> list[float]:
+def benchmark_float_list(benchmark: BenchmarkFixture, func: object, *args: object) -> list[float]:
     """Run benchmark for functions returning list of floats."""
     from typing import cast
 
@@ -326,7 +334,10 @@ class TestVadBenchmark:
         assert isinstance(result, bool), "VAD should return boolean"
 
     def test_streaming_vad_process_chunk(
-        self, benchmark: BenchmarkFixture, streaming_vad: StreamingVad, audio_chunk: NDArray[np.float32]
+        self,
+        benchmark: BenchmarkFixture,
+        streaming_vad: StreamingVad,
+        audio_chunk: NDArray[np.float32],
     ) -> None:
         """Benchmark StreamingVad.process_chunk() call."""
         result = typed_benchmark(benchmark, bool, streaming_vad.process_chunk, audio_chunk)
@@ -392,14 +403,20 @@ class TestRmsLevelProviderBenchmark:
     """Benchmark tests for RmsLevelProvider methods."""
 
     def test_get_rms(
-        self, benchmark: BenchmarkFixture, rms_provider: RmsLevelProvider, audio_chunk: NDArray[np.float32]
+        self,
+        benchmark: BenchmarkFixture,
+        rms_provider: RmsLevelProvider,
+        audio_chunk: NDArray[np.float32],
     ) -> None:
         """Benchmark get_rms() method."""
         result = typed_benchmark(benchmark, float, rms_provider.get_rms, audio_chunk)
         assert 0 <= result <= 1, "RMS should be normalized"
 
     def test_get_db(
-        self, benchmark: BenchmarkFixture, rms_provider: RmsLevelProvider, audio_chunk: NDArray[np.float32]
+        self,
+        benchmark: BenchmarkFixture,
+        rms_provider: RmsLevelProvider,
+        audio_chunk: NDArray[np.float32],
     ) -> None:
         """Benchmark get_db() method."""
         result = typed_benchmark(benchmark, float, rms_provider.get_db, audio_chunk)
@@ -465,7 +482,9 @@ class TestNumpyOperationsBenchmark:
 class TestBufferOperationsBenchmark:
     """Benchmark tests for list operations used in buffers."""
 
-    def test_list_append(self, benchmark: BenchmarkFixture, audio_chunk: NDArray[np.float32]) -> None:
+    def test_list_append(
+        self, benchmark: BenchmarkFixture, audio_chunk: NDArray[np.float32]
+    ) -> None:
         """Benchmark list append (partial buffer accumulation)."""
         buffer: list[NDArray[np.float32]] = []
 
@@ -475,7 +494,9 @@ class TestBufferOperationsBenchmark:
         benchmark(append)
         assert buffer, "Buffer should have items"
 
-    def test_list_clear(self, benchmark: BenchmarkFixture, audio_chunk: NDArray[np.float32]) -> None:
+    def test_list_clear(
+        self, benchmark: BenchmarkFixture, audio_chunk: NDArray[np.float32]
+    ) -> None:
         """Benchmark list clear (partial buffer clearing)."""
         # Pre-fill buffer
         buffer = [audio_chunk.copy() for _ in range(20)]
@@ -487,7 +508,9 @@ class TestBufferOperationsBenchmark:
 
         benchmark(clear_and_refill)
 
-    def test_sum_lengths_naive(self, benchmark: BenchmarkFixture, audio_chunk: NDArray[np.float32]) -> None:
+    def test_sum_lengths_naive(
+        self, benchmark: BenchmarkFixture, audio_chunk: NDArray[np.float32]
+    ) -> None:
         """Benchmark naive sum of chunk lengths (OLD segmenter pattern)."""
         chunks = [audio_chunk.copy() for _ in range(20)]
 
@@ -515,6 +538,7 @@ class TestPartialBufferComparisonBenchmark:
         self, benchmark: BenchmarkFixture, audio_chunk: NDArray[np.float32]
     ) -> None:
         """Benchmark OLD pattern: list append + np.concatenate (20 chunks = 2s)."""
+
         def old_pattern() -> NDArray[np.float32]:
             buffer: list[NDArray[np.float32]] = []
             for _ in range(TYPICAL_PARTIAL_CHUNKS):
@@ -570,6 +594,7 @@ class TestPartialBufferComparisonBenchmark:
         self, benchmark: BenchmarkFixture, audio_chunk: NDArray[np.float32]
     ) -> None:
         """Benchmark OLD pattern: 10 cycles of accumulate/concat/clear."""
+
         def old_pattern_cycles() -> list[NDArray[np.float32]]:
             results: list[NDArray[np.float32]] = []
             for _ in range(10):