view tests/test_util.py @ 403:996948ada980

Update tests

import math
from array import array as array_
from unittest.mock import patch

import pytest

from auditok import signal as signal_
from auditok import signal_numpy
from auditok.exceptions import TimeFormatError
from auditok.util import (
    AudioEnergyValidator,
    make_channel_selector,
    make_duration_formatter,
)


def _sample_generator(*data_buffers):
    """
    Takes a list of many mono audio data buffers and makes a sample generator
    of interleaved audio samples, one sample from each channel. The resulting
    generator can be used to build a multichannel audio buffer.
    >>> gen = _sample_generator("abcd", "ABCD")
    >>> list(gen)
    ["a", "A", 1, 1, "c", "C", "d", "D"]
    """
    frame_gen = zip(*data_buffers)
    return (sample for frame in frame_gen for sample in frame)


def _generate_pure_tone(
    frequency, duration_sec=1, sampling_rate=16000, sample_width=2, volume=1e4
):
    """
    Generates a pure tone with the given frequency.
    """
    assert frequency <= sampling_rate / 2
    max_value = (2 ** (sample_width * 8) // 2) - 1
    if volume > max_value:
        volume = max_value
    fmt = signal_.FORMAT[sample_width]
    total_samples = int(sampling_rate * duration_sec)
    step = frequency / sampling_rate
    two_pi_step = 2 * math.pi * step
    data = array_(
        fmt,
        (int(math.sin(two_pi_step * i) * volume) for i in range(total_samples)),
    )
    return data


PURE_TONE_DICT = {
    freq: _generate_pure_tone(freq, 1, 16000, 2) for freq in (400, 800, 1600)
}
PURE_TONE_DICT.update(
    {
        freq: _generate_pure_tone(freq, 0.1, 16000, 2)
        for freq in (600, 1150, 2400, 7220)
    }
)


class TestFunctions:
    def setup_method(self):
        self.data = b"012345679ABC"

    @pytest.mark.parametrize(
        "fmt, duration, expected",
        [
            ("%S", 5400, "5400.000"),  # only_seconds
            ("%I", 5400, "5400000"),  # only_millis
            ("%h:%m:%s.%i", 3725.365, "01:02:05.365"),  # full
            ("%h:%m:%s.%i", 1925.075, "00:32:05.075"),  # full_zero_hours
            ("%h:%m:%s.%i", 3659.075, "01:00:59.075"),  # full_zero_minutes
            ("%h:%m:%s.%i", 3720.075, "01:02:00.075"),  # full_zero_seconds
            ("%h:%m:%s.%i", 3725, "01:02:05.000"),  # full_zero_millis
            (
                "%h %h:%m:%s.%i %s",
                3725.365,
                "01 01:02:05.365 05",
            ),  # duplicate_directive
            ("%h:%m:%s", 3725, "01:02:05"),  # no_millis
            ("%h:%m", 3725, "01:02"),  # no_seconds
            ("%h", 3725, "01"),  # no_minutes
            ("%m:%s.%i", 3725, "02:05.000"),  # no_hours
        ],
        ids=[
            "only_seconds",
            "only_millis",
            "full",
            "full_zero_hours",
            "full_zero_minutes",
            "full_zero_seconds",
            "full_zero_millis",
            "duplicate_directive",
            "no_millis",
            "no_seconds",
            "no_minutes",
            "no_hours",
        ],
    )
    def test_make_duration_formatter(self, fmt, duration, expected):
        formatter = make_duration_formatter(fmt)
        result = formatter(duration)
        assert result == expected

    @pytest.mark.parametrize(
        "fmt",
        [
            "%S %S",  # duplicate_only_seconds
            "%I %I",  # duplicate_only_millis
            "%x",  # unknown_directive
        ],
        ids=[
            "duplicate_only_seconds",
            "duplicate_only_millis",
            "unknown_directive",
        ],
    )
    def test_make_duration_formatter_error(self, fmt):
        with pytest.raises(TimeFormatError):
            make_duration_formatter(fmt)

    @pytest.mark.parametrize(
        "sample_width, channels, selected, expected",
        [
            (
                1,
                1,
                0,
                [48, 49, 50, 51, 52, 53, 54, 55, 57, 65, 66, 67],
            ),  # int8_1channel_select_0
            (1, 2, 0, [48, 50, 52, 54, 57, 66]),  # int8_2channel_select_0
            (1, 3, 0, [48, 51, 54, 65]),  # int8_3channel_select_0
            (1, 3, 1, [49, 52, 55, 66]),  # int8_3channel_select_1
            (1, 3, 2, [50, 53, 57, 67]),  # int8_3channel_select_2
            (1, 4, 0, [48, 52, 57]),  # int8_4channel_select_0
            (
                2,
                1,
                0,
                [12592, 13106, 13620, 14134, 16697, 17218],
            ),  # int16_1channel_select_0
            (2, 2, 0, [12592, 13620, 16697]),  # int16_2channel_select_0
            (2, 2, 1, [13106, 14134, 17218]),  # int16_2channel_select_1
            (2, 3, 0, [12592, 14134]),  # int16_3channel_select_0
            (2, 3, 1, [13106, 16697]),  # int16_3channel_select_1
            (2, 3, 2, [13620, 17218]),  # int16_3channel_select_2
            (
                4,
                1,
                0,
                [858927408, 926299444, 1128415545],
            ),  # int32_1channel_select_0
            (4, 3, 0, [858927408]),  # int32_3channel_select_0
            (4, 3, 1, [926299444]),  # int32_3channel_select_1
            (4, 3, 2, [1128415545]),  # int32_3channel_select_2
        ],
        ids=[
            "int8_1channel_select_0",
            "int8_2channel_select_0",
            "int8_3channel_select_0",
            "int8_3channel_select_1",
            "int8_3channel_select_2",
            "int8_4channel_select_0",
            "int16_1channel_select_0",
            "int16_2channel_select_0",
            "int16_2channel_select_1",
            "int16_3channel_select_0",
            "int16_3channel_select_1",
            "int16_3channel_select_2",
            "int32_1channel_select_0",
            "int32_3channel_select_0",
            "int32_3channel_select_1",
            "int32_3channel_select_2",
        ],
    )
    def test_make_channel_selector_one_channel(
        self, sample_width, channels, selected, expected
    ):

        # force using signal functions with standard python implementation
        with patch("auditok.util.signal", signal_):
            selector = make_channel_selector(sample_width, channels, selected)
            result = selector(self.data)

        fmt = signal_.FORMAT[sample_width]
        expected = array_(fmt, expected)
        if channels == 1:
            expected = bytes(expected)
        assert result == expected

        # Use signal functions with numpy implementation
        with patch("auditok.util.signal", signal_numpy):
            selector = make_channel_selector(sample_width, channels, selected)
            result_numpy = selector(self.data)

        expected = array_(fmt, expected)
        if channels == 1:
            expected = bytes(expected)
            assert result_numpy == expected
        else:
            assert all(result_numpy == expected)

    @pytest.mark.parametrize(
        "sample_width, channels, selected, expected",
        [
            (1, 2, "avg", [48, 50, 52, 54, 61, 66]),  # int8_2channel
            (1, 4, "average", [50, 54, 64]),  # int8_4channel
            (
                2,
                1,
                "mix",
                [12592, 13106, 13620, 14134, 16697, 17218],
            ),  # int16_1channel
            (2, 2, "avg", [12849, 13877, 16957]),  # int16_2channel
            (4, 3, "average", [971214132]),  # int32_3channel
        ],
        ids=[
            "int8_2channel",
            "int8_4channel",
            "int16_1channel",
            "int16_2channel",
            "int32_3channel",
        ],
    )
    def test_make_channel_selector_average(
        self, sample_width, channels, selected, expected
    ):
        # force using signal functions with standard python implementation
        with patch("auditok.util.signal", signal_):
            selector = make_channel_selector(sample_width, channels, selected)
            result = selector(self.data)

        fmt = signal_.FORMAT[sample_width]
        expected = array_(fmt, expected)
        if channels == 1:
            expected = bytes(expected)
        assert result == expected

        # Use signal functions with numpy implementation
        with patch("auditok.util.signal", signal_numpy):
            selector = make_channel_selector(sample_width, channels, selected)
            result_numpy = selector(self.data)

        if channels in (1, 2):
            assert result_numpy == expected
        else:
            assert all(result_numpy == expected)

    @pytest.mark.parametrize(
        "sample_width, channels, selected, expected",
        [
            (
                1,
                1,
                "any",
                [[48, 49, 50, 51, 52, 53, 54, 55, 57, 65, 66, 67]],
            ),  # int8_1channel
            (
                1,
                2,
                None,
                [[48, 50, 52, 54, 57, 66], [49, 51, 53, 55, 65, 67]],
            ),  # int8_2channel
            (
                1,
                4,
                "any",
                [[48, 52, 57], [49, 53, 65], [50, 54, 66], [51, 55, 67]],
            ),  # int8_4channel
            (
                2,
                2,
                None,
                [[12592, 13620, 16697], [13106, 14134, 17218]],
            ),  # int16_2channel
            (
                4,
                3,
                "any",
                [[858927408], [926299444], [1128415545]],
            ),  # int32_3channel
        ],
        ids=[
            "int8_1channel",
            "int8_2channel",
            "int8_4channel",
            "int16_2channel",
            "int32_3channel",
        ],
    )
    def test_make_channel_selector_any(
        self, sample_width, channels, selected, expected
    ):

        # force using signal functions with standard python implementation
        with patch("auditok.util.signal", signal_):
            selector = make_channel_selector(sample_width, channels, selected)
            result = selector(self.data)

        fmt = signal_.FORMAT[sample_width]
        expected = [array_(fmt, exp) for exp in expected]
        if channels == 1:
            expected = bytes(expected[0])
        assert result == expected

        # Use signal functions with numpy implementation
        with patch("auditok.util.signal", signal_numpy):
            selector = make_channel_selector(sample_width, channels, selected)
            result_numpy = selector(self.data)

        if channels == 1:
            assert result_numpy == expected
        else:
            assert (result_numpy == expected).all()


class TestAudioEnergyValidator:
    @pytest.mark.parametrize(
        "data, channels, use_channel, expected",
        [
            ([350, 400], 1, None, True),  # mono_valid_uc_None
            ([350, 400], 1, "any", True),  # mono_valid_uc_any
            ([350, 400], 1, 0, True),  # mono_valid_uc_0
            ([350, 400], 1, "mix", True),  # mono_valid_uc_mix
            ([300, 300], 1, None, False),  # mono_invalid_uc_None
            ([300, 400, 350, 300], 2, None, True),  # stereo_valid_uc_None
            ([300, 400, 350, 300], 2, "any", True),  # stereo_valid_uc_any
            ([300, 400, 350, 300], 2, "mix", True),  # stereo_valid_uc_mix
            ([300, 400, 350, 300], 2, "avg", True),  # stereo_valid_uc_avg
            (
                [300, 400, 300, 300],
                2,
                "average",
                True,
            ),  # stereo_valid_uc_average
            (
                [634, 0, 634, 0],
                2,
                "mix",
                True,
            ),  # stereo_valid_uc_mix_with_null_channel
            ([320, 100, 320, 100], 2, 0, True),  # stereo_valid_uc_0
            ([100, 320, 100, 320], 2, 1, True),  # stereo_valid_uc_1
            ([280, 100, 280, 100], 2, None, False),  # stereo_invalid_uc_None
            ([280, 100, 280, 100], 2, "any", False),  # stereo_invalid_uc_any
            ([400, 200, 400, 200], 2, "mix", False),  # stereo_invalid_uc_mix
            ([300, 400, 300, 400], 2, 0, False),  # stereo_invalid_uc_0
            ([400, 300, 400, 300], 2, 1, False),  # stereo_invalid_uc_1
            ([0, 0, 0, 0], 2, None, False),  # zeros
        ],
        ids=[
            "mono_valid_uc_None",
            "mono_valid_uc_any",
            "mono_valid_uc_0",
            "mono_valid_uc_mix",
            "mono_invalid_uc_None",
            "stereo_valid_uc_None",
            "stereo_valid_uc_any",
            "stereo_valid_uc_mix",
            "stereo_valid_uc_avg",
            "stereo_valid_uc_average",
            "stereo_valid_uc_mix_with_null_channel",
            "stereo_valid_uc_0",
            "stereo_valid_uc_1",
            "stereo_invalid_uc_None",
            "stereo_invalid_uc_any",
            "stereo_invalid_uc_mix",
            "stereo_invalid_uc_0",
            "stereo_invalid_uc_1",
            "zeros",
        ],
    )
    def test_audio_energy_validator(
        self, data, channels, use_channel, expected
    ):

        data = array_("h", data)
        sample_width = 2
        energy_threshold = 50
        validator = AudioEnergyValidator(
            energy_threshold, sample_width, channels, use_channel
        )

        if expected:
            assert validator.is_valid(data)
        else:
            assert not validator.is_valid(data)