Mercurial > hg > svcore
view data/model/test/TestWaveformOversampler.h @ 1881:b504df98c3be
Ensure completion on output model is started at zero, so if it's checked before the input model has become ready and the transform has begun, it is not accidentally reported as complete (affected re-aligning models in Sonic Lineup when replacing the session)
author | Chris Cannam |
---|---|
date | Fri, 26 Jun 2020 11:45:39 +0100 |
parents | 074b860a7828 |
children |
line wrap: on
line source
/* -*- c-basic-offset: 4 indent-tabs-mode: nil -*- vi:set ts=8 sts=4 sw=4: */ /* Sonic Visualiser An audio file viewer and annotation editor. Centre for Digital Music, Queen Mary, University of London. This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version. See the file COPYING included with this distribution for more information. */ #ifndef TEST_WAVEFORM_OVERSAMPLER_H #define TEST_WAVEFORM_OVERSAMPLER_H #include "../WaveformOversampler.h" #include "../WritableWaveFileModel.h" #include "../../../base/BaseTypes.h" #include <QObject> #include <QtTest> class TestWaveformOversampler : public QObject { Q_OBJECT public: TestWaveformOversampler() { m_source = floatvec_t(5000, 0.f); m_source[0] = 1.f; m_source[2500] = 0.5f; m_source[2501] = -0.5f; m_source[4999] = -1.f; for (int i = 3000; i < 3900; ++i) { m_source[i] = float(sin(double(i - 3000) * M_PI / 50.0)); } m_sourceModel = new WritableWaveFileModel(8000, 1); const float *d = m_source.data(); QVERIFY(m_sourceModel->addSamples(&d, m_source.size())); m_sourceModel->writeComplete(); } ~TestWaveformOversampler() { delete m_sourceModel; } private: floatvec_t m_source; WritableWaveFileModel *m_sourceModel; void compareStrided(floatvec_t obtained, floatvec_t expected, int stride) { QCOMPARE(obtained.size(), expected.size() * stride); float threshold = 1e-10f; for (int i = 0; in_range_for(expected, i); ++i) { if (fabsf(obtained[i * stride] - expected[i]) > threshold) { std::cerr << "At position " << i * stride << ": " << obtained[i * stride] << " != " << expected[i] << std::endl; QCOMPARE(obtained, expected); } } } void compareVecs(floatvec_t obtained, floatvec_t expected) { compareStrided(obtained, expected, 1); } floatvec_t get(sv_frame_t sourceStartFrame, sv_frame_t sourceFrameCount, int oversampleBy) { return WaveformOversampler::getOversampledData (*m_sourceModel, 0, sourceStartFrame, sourceFrameCount, oversampleBy); } void testVerbatim(sv_frame_t sourceStartFrame, sv_frame_t sourceFrameCount, int oversampleBy, floatvec_t expected) { floatvec_t output = get(sourceStartFrame, sourceFrameCount, oversampleBy); compareVecs(output, expected); } void testStrided(sv_frame_t sourceStartFrame, sv_frame_t sourceFrameCount, int oversampleBy, floatvec_t expected) { // check only the values that are expected to be precisely the // original samples floatvec_t output = get(sourceStartFrame, sourceFrameCount, oversampleBy); compareStrided(output, expected, oversampleBy); } floatvec_t sourceSubset(sv_frame_t start, sv_frame_t length) { return floatvec_t(m_source.begin() + start, m_source.begin() + start + length); } private slots: void testWholeVerbatim() { testVerbatim(0, 5000, 1, m_source); } void testSubsetsVerbatim() { testVerbatim(0, 500, 1, sourceSubset(0, 500)); testVerbatim(4500, 500, 1, sourceSubset(4500, 500)); testVerbatim(2000, 1000, 1, sourceSubset(2000, 1000)); } void testOverlapsVerbatim() { // overlapping the start -> result should be zero-padded to // preserve start frame floatvec_t expected = sourceSubset(0, 400); expected.insert(expected.begin(), 100, 0.f); testVerbatim(-100, 500, 1, expected); // overlapping the end -> result should be truncated to // preserve source length expected = sourceSubset(4600, 400); testVerbatim(4600, 500, 1, expected); } void testWhole2x() { testStrided(0, 5000, 2, m_source); // check for windowed sinc values between the original samples floatvec_t output = get(0, 5000, 2); QVERIFY(output[1] - 0.6358 < 0.0001); QVERIFY(output[3] + 0.2099 < 0.0001); } void testWhole3x() { testStrided(0, 5000, 3, m_source); // check for windowed sinc values between the original samples floatvec_t output = get(0, 5000, 3); QVERIFY(output[1] > 0.7); QVERIFY(output[2] > 0.4); QVERIFY(output[4] < -0.1); QVERIFY(output[5] < -0.1); } void testWhole4x() { testStrided(0, 5000, 4, m_source); // check for windowed sinc values between the original samples floatvec_t output = get(0, 5000, 4); QVERIFY(output[1] - 0.9000 < 0.0001); QVERIFY(output[2] - 0.6358 < 0.0001); QVERIFY(output[3] - 0.2993 < 0.0001); QVERIFY(output[5] + 0.1787 < 0.0001); QVERIFY(output[6] + 0.2099 < 0.0001); QVERIFY(output[7] + 0.1267 < 0.0001); // alternate values at 2n should equal all values at n output = get(0, 5000, 4); floatvec_t half = get(0, 5000, 2); compareStrided(output, half, 2); } void testWhole8x() { testStrided(0, 5000, 8, m_source); // alternate values at 2n should equal all values at n floatvec_t output = get(0, 5000, 8); floatvec_t half = get(0, 5000, 4); compareStrided(output, half, 2); } void testWhole10x() { testStrided(0, 5000, 10, m_source); // alternate values at 2n should equal all values at n floatvec_t output = get(0, 5000, 10); floatvec_t half = get(0, 5000, 5); compareStrided(output, half, 2); } void testWhole16x() { testStrided(0, 5000, 16, m_source); // alternate values at 2n should equal all values at n floatvec_t output = get(0, 5000, 16); floatvec_t half = get(0, 5000, 8); compareStrided(output, half, 2); } void testSubsets4x() { testStrided(0, 500, 4, sourceSubset(0, 500)); testStrided(4500, 500, 4, sourceSubset(4500, 500)); testStrided(2000, 1000, 4, sourceSubset(2000, 1000)); // check for windowed sinc values between the original // samples, even when the original sample that was the source // of this sinc kernel is not within the requested range floatvec_t output = get(1, 10, 4); QVERIFY(output[0] < 0.0001); QVERIFY(output[1] + 0.1787 < 0.0001); QVERIFY(output[2] + 0.2099 < 0.0001); QVERIFY(output[3] + 0.1267 < 0.0001); // and again at the end output = get(4989, 10, 4); QVERIFY(output[39] + 0.9000 < 0.0001); QVERIFY(output[38] + 0.6358 < 0.0001); QVERIFY(output[37] + 0.2993 < 0.0001); QVERIFY(output[35] - 0.1787 < 0.0001); QVERIFY(output[34] - 0.2099 < 0.0001); QVERIFY(output[33] - 0.1267 < 0.0001); } void testOverlaps4x() { // overlapping the start -> result should be zero-padded to // preserve start frame floatvec_t expected = sourceSubset(0, 400); expected.insert(expected.begin(), 100, 0.f); testStrided(-100, 500, 4, expected); // overlapping the end -> result should be truncated to // preserve source length expected = sourceSubset(4600, 400); testStrided(4600, 500, 4, expected); } void testSubsets15x() { testStrided(0, 500, 15, sourceSubset(0, 500)); testStrided(4500, 500, 15, sourceSubset(4500, 500)); testStrided(2000, 1000, 15, sourceSubset(2000, 1000)); } void testOverlaps15x() { // overlapping the start -> result should be zero-padded to // preserve start frame floatvec_t expected = sourceSubset(0, 400); expected.insert(expected.begin(), 100, 0.f); testStrided(-100, 500, 15, expected); // overlapping the end -> result should be truncated to // preserve source length expected = sourceSubset(4600, 400); testStrided(4600, 500, 15, expected); } }; #endif