Add test for preprocessing class with purest and remove unnecessary lib in project.toml

pyproject.toml

@@ -34,7 +34,6 @@ dev = [
     "tensorflow==2.20.0",
     "tensorflow-io==0.37.1",
     "tensorflow-hub==0.16.1",
-    "pydub==0.25.1",
     "librosa==0.11.0"
 ]
 test = [

tests/unit/nightingale/test_audio_preprocessor.py

@@ -0,0 +1,57 @@
+import os
+from pathlib import Path
+
+import numpy as np
+import soundfile as sf
+import pytest
+
+from nightingale.data_pipeline.audio_preprocessor import AudioPreprocessor
+
+def _write_test_wav(path: Path, sr: int = 16000, channels: int = 1, duration=0.5):
+    t = np.linspace(0, duration, int(sr * duration), endpoint=False)
+    tone = 0.1 * np.sin(2 * np.pi * 440 * t)
+    if channels == 1:
+        data = tone.astype(np.float32)
+    else:
+        data = np.stack([tone, tone], axis=-1).astype(np.float32)
+    path.parent.mkdir(parents=True, exist_ok=True)
+    sf.write(str(path), data, sr)
+
+def test_is_yamnet_ready_true_for_matching_wav(tmp_path):
+    ap = AudioPreprocessor(target_sr=16000, mono=True)
+    p = tmp_path / "mono_16k.wav"
+    _write_test_wav(p, sr=16000, channels=1)
+    assert ap.is_yamnet_ready(str(p)) is True
+
+def test_load_audio_resamples_and_converts_to_mono(tmp_path):
+    ap = AudioPreprocessor(target_sr=16000, mono=True)
+    p = tmp_path / "stereo_22k.wav"
+    _write_test_wav(p, sr=22050, channels=2)
+    wav, sr = ap.load_audio(str(p))
+    assert sr == ap.target_sr
+    assert wav.dtype == np.float32
+    assert wav.ndim == 1  # mono
+
+def test_process_file_creates_output_and_is_ready(tmp_path):
+    ap = AudioPreprocessor(target_sr=16000, mono=True)
+    inp = tmp_path / "in" / "stereo_src.wav"
+    out = tmp_path / "out" / "converted.wav"
+    _write_test_wav(inp, sr=22050, channels=2)
+    # ensure output does not exist initially
+    assert not out.exists()
+    ap.process_file(str(inp), str(out))
+    assert out.exists()
+    assert ap.is_yamnet_ready(str(out)) is True
+
+def test_process_folder_creates_mirrored_structure(tmp_path):
+    ap = AudioPreprocessor(target_sr=16000, mono=True)
+    input_root = tmp_path / "input_root"
+    subfolder = input_root / "sub1"
+    file_in = subfolder / "sound.wav"
+    _write_test_wav(file_in, sr=22050, channels=2)
+
+    output_root = tmp_path / "output_root"
+    ap.process_folder(str(input_root), str(output_root))
+    expected_out = output_root / "sub1" / "sound.wav"
+    assert expected_out.exists()
+    assert ap.is_yamnet_ready(str(expected_out)) is True