tests/test_snac_codec.py

"""Tests for the SNAC codec wrapper."""

from __future__ import annotations

from pathlib import Path

import numpy as np
import pytest
import torch
import torchaudio


@pytest.fixture()
def wav_file(tmp_path: Path) -> Path:
    """Create a short mono WAV at 24 kHz."""
    sr = 24_000
    samples = torch.randn(1, sr * 2)  # 2 seconds
    path = tmp_path / "test.wav"
    torchaudio.save(str(path), samples, sr)
    return path


def test_snac_name():
    from compare_codec.snac_codec import SNACCodec

    codec = SNACCodec()
    assert codec.name == "SNAC"


def test_snac_sample_rate():
    from compare_codec.snac_codec import SNACCodec

    codec = SNACCodec()
    assert codec.sample_rate == 24_000


def test_snac_configs_not_empty():
    from compare_codec.snac_codec import SNACCodec

    codec = SNACCodec()
    configs = codec.configs()
    assert len(configs) >= 3  # at least one per model variant


def test_snac_configs_have_sample_rate():
    from compare_codec.snac_codec import SNACCodec

    codec = SNACCodec()
    for cfg in codec.configs():
        assert "sample_rate" in cfg.params
        assert "model_id" in cfg.params


def test_snac_encode_decode_returns_float32_array(wav_file: Path):
    from compare_codec.snac_codec import SNACCodec

    codec = SNACCodec()
    cfg = [c for c in codec.configs() if c.params["sample_rate"] == 24_000][0]
    result = codec.encode_decode(wav_file, cfg)

    assert isinstance(result, np.ndarray)
    assert result.dtype == np.float32
    assert result.ndim == 1
    assert len(result) > 0