Chris@756: /* -*- c-basic-offset: 4 indent-tabs-mode: nil -*-  vi:set ts=8 sts=4 sw=4: */
Chris@756: 
Chris@756: /*
Chris@756:     Sonic Visualiser
Chris@756:     An audio file viewer and annotation editor.
Chris@756:     Centre for Digital Music, Queen Mary, University of London.
Chris@756:     This file copyright 2013 Chris Cannam.
Chris@756:     
Chris@756:     This program is free software; you can redistribute it and/or
Chris@756:     modify it under the terms of the GNU General Public License as
Chris@756:     published by the Free Software Foundation; either version 2 of the
Chris@756:     License, or (at your option) any later version.  See the file
Chris@756:     COPYING included with this distribution for more information.
Chris@756: */
Chris@756: 
Chris@756: #ifndef TEST_AUDIO_FILE_READER_H
Chris@756: #define TEST_AUDIO_FILE_READER_H
Chris@756: 
Chris@756: #include "../AudioFileReaderFactory.h"
Chris@756: #include "../AudioFileReader.h"
Chris@1313: #include "../WavFileWriter.h"
Chris@756: 
Chris@756: #include "AudioTestData.h"
Chris@756: 
Chris@756: #include <cmath>
Chris@756: 
Chris@756: #include <QObject>
Chris@756: #include <QtTest>
Chris@756: #include <QDir>
Chris@756: 
Chris@756: #include <iostream>
Chris@756: 
Chris@756: using namespace std;
Chris@756: 
Chris@1263: static QString audioDir = "svcore/data/fileio/test/testfiles";
Chris@1313: static QString diffDir  = "svcore/data/fileio/test/diffs";
Chris@756: 
Chris@756: class AudioFileReaderTest : public QObject
Chris@756: {
Chris@756:     Q_OBJECT
Chris@756: 
Chris@756:     const char *strOf(QString s) {
Chris@756:         return strdup(s.toLocal8Bit().data());
Chris@756:     }
Chris@756: 
Chris@1313:     void getFileMetadata(QString filename,
Chris@1313:                          QString &extension,
Chris@1313:                          sv_samplerate_t &rate,
Chris@1313:                          int &channels,
Chris@1313:                          int &bitdepth) {
Chris@1313: 
Chris@1313:         QStringList fileAndExt = filename.split(".");
Chris@1313:         QStringList bits = fileAndExt[0].split("-");
Chris@1313: 
Chris@1313:         extension = fileAndExt[1];
Chris@1313:         rate = bits[0].toInt();
Chris@1313:         channels = bits[1].toInt();
Chris@1313:         bitdepth = 16;
Chris@1313:         if (bits.length() > 2) {
Chris@1313:             bitdepth = bits[2].toInt();
Chris@1313:         }
Chris@1313:     }
Chris@1313:     
Chris@1313:     void getExpectedThresholds(QString filename,
Chris@1313:                                bool resampled,
Chris@1313:                                bool gapless,
Chris@1313:                                bool normalised,
Chris@1313:                                double &maxLimit,
Chris@1313:                                double &rmsLimit) {
Chris@1313: 
Chris@1313:         QString extension;
Chris@1313:         sv_samplerate_t fileRate;
Chris@1313:         int channels;
Chris@1313:         int bitdepth;
Chris@1313:         getFileMetadata(filename, extension, fileRate, channels, bitdepth);
Chris@1313:         
Chris@1313:         if (normalised) {
Chris@1313: 
Chris@1313:             if (extension == "ogg") {
Chris@1313: 
Chris@1313:                 // Our ogg is not especially high quality and is
Chris@1313:                 // actually further from the original if normalised
Chris@1313: 
Chris@1313:                 maxLimit = 0.1;
Chris@1313:                 rmsLimit = 0.03;
Chris@1313: 
Chris@1313:             } else if (extension == "m4a" || extension == "aac") {
Chris@1313: 
cannam@1314:                 // Like ogg but more so, quite far off in signal terms
cannam@1314:                 // and even worse if normalised
cannam@1314:                 maxLimit = 0.1;
cannam@1314:                 rmsLimit = 0.1;
Chris@1313: 
Chris@1313:             } else if (extension == "mp3") {
Chris@1313: 
Chris@1313:                 if (resampled && !gapless) {
Chris@1313: 
Chris@1313:                     // We expect worse figures here, because the
Chris@1313:                     // combination of uncompensated encoder delay +
Chris@1313:                     // resampling results in a fractional delay which
Chris@1313:                     // means the decoded signal is slightly out of
Chris@1313:                     // phase compared to the test signal
Chris@1313: 
Chris@1313:                     maxLimit = 0.1;
Chris@1313:                     rmsLimit = 0.05;
Chris@1313: 
Chris@1313:                 } else {
Chris@1313: 
Chris@1313:                     maxLimit = 0.05;
Chris@1313:                     rmsLimit = 0.01;
Chris@1313:                 }
Chris@1313: 
Chris@1313:             } else {
Chris@1313: 
Chris@1313:                 // supposed to be lossless then (wav, aiff, flac)
Chris@1313:                 
Chris@1313:                 if (bitdepth >= 16 && !resampled) {
Chris@1313:                     maxLimit = 1e-3;
Chris@1313:                     rmsLimit = 3e-4;
Chris@1313:                 } else {
Chris@1313:                     maxLimit = 0.01;
Chris@1313:                     rmsLimit = 5e-3;
Chris@1313:                 }
Chris@1313:             }
Chris@1313:             
Chris@1313:         } else { // !normalised
Chris@1313:             
Chris@1313:             if (extension == "ogg") {
Chris@1313: 
Chris@1313:                 maxLimit = 0.06;
Chris@1313:                 rmsLimit = 0.03;
Chris@1313: 
Chris@1313:             } else if (extension == "m4a" || extension == "aac") {
Chris@1313: 
cannam@1314:                 maxLimit = 0.06;
cannam@1314:                 rmsLimit = 0.03;
Chris@1313: 
Chris@1313:             } else if (extension == "mp3") {
Chris@1313: 
Chris@1313:                 // all mp3 figures are worse when not normalising
Chris@1313:                 maxLimit = 0.1;
Chris@1313:                 rmsLimit = 0.05;
Chris@1313: 
Chris@1313:             } else {
Chris@1313: 
Chris@1313:                 // supposed to be lossless then (wav, aiff, flac)
Chris@1313:                 
Chris@1313:                 if (bitdepth >= 16 && !resampled) {
Chris@1313:                     maxLimit = 1e-3;
Chris@1313:                     rmsLimit = 3e-4;
Chris@1313:                 } else {
Chris@1313:                     maxLimit = 0.02;
Chris@1313:                     rmsLimit = 0.01;
Chris@1313:                 }
Chris@1313:             }
Chris@1313:         }
Chris@1313:     }
Chris@1313: 
Chris@1313:     QString testName(QString filename, int rate, bool norm, bool gapless) {
Chris@1313:         return QString("%1 at %2%3%4")
Chris@1313:             .arg(filename)
Chris@1313:             .arg(rate)
Chris@1313:             .arg(norm ? " normalised": "")
Chris@1313:             .arg(gapless ? "" : " non-gapless");
Chris@1313:     }
Chris@1313: 
Chris@756: private slots:
Chris@756:     void init()
Chris@756:     {
Chris@756:         if (!QDir(audioDir).exists()) {
Chris@756:             cerr << "ERROR: Audio test file directory \"" << audioDir << "\" does not exist" << endl;
Chris@756:             QVERIFY2(QDir(audioDir).exists(), "Audio test file directory not found");
Chris@756:         }
Chris@1313:         if (!QDir(diffDir).exists() && !QDir().mkpath(diffDir)) {
Chris@1313:             cerr << "ERROR: Audio diff directory \"" << diffDir << "\" does not exist and could not be created" << endl;
Chris@1313:             QVERIFY2(QDir(diffDir).exists(), "Audio diff directory not found and could not be created");
Chris@1313:         }
Chris@756:     }
Chris@756: 
Chris@756:     void read_data()
Chris@756:     {
Chris@756:         QTest::addColumn<QString>("audiofile");
Chris@1313:         QTest::addColumn<int>("rate");
Chris@1313:         QTest::addColumn<bool>("normalised");
Chris@1313:         QTest::addColumn<bool>("gapless");
Chris@756:         QStringList files = QDir(audioDir).entryList(QDir::Files);
Chris@1313:         int readRates[] = { 44100, 48000 };
Chris@1313:         bool norms[] = { false, true };
Chris@1313:         bool gaplesses[] = { true, false };
Chris@756:         foreach (QString filename, files) {
Chris@1313:             for (int rate: readRates) {
Chris@1313:                 for (bool norm: norms) {
Chris@1313:                     for (bool gapless: gaplesses) {
Chris@1313: 
Chris@1313:                         if (QFileInfo(filename).suffix() != "mp3" &&
Chris@1313:                             !gapless) {
Chris@1313:                             continue;
Chris@1313:                         }
Chris@1313:                         
Chris@1313:                         QString desc = testName(filename, rate, norm, gapless);
Chris@1313: 
Chris@1313:                         QTest::newRow(strOf(desc))
Chris@1313:                             << filename << rate << norm << gapless;
Chris@1313:                     }
Chris@1313:                 }
Chris@1313:             }
Chris@756:         }
Chris@756:     }
Chris@756: 
Chris@756:     void read()
Chris@756:     {
Chris@756:         QFETCH(QString, audiofile);
Chris@1313:         QFETCH(int, rate);
Chris@1313:         QFETCH(bool, normalised);
Chris@1313:         QFETCH(bool, gapless);
Chris@756: 
Chris@1313:         sv_samplerate_t readRate(rate);
Chris@1313:         
Chris@1313:         cerr << "\naudiofile = " << audiofile << endl;
Chris@1313: 
Chris@1313:         AudioFileReaderFactory::Parameters params;
Chris@1313:         params.targetRate = readRate;
Chris@1313:         params.normalisation = (normalised ?
Chris@1313:                                 AudioFileReaderFactory::Normalisation::Peak :
Chris@1313:                                 AudioFileReaderFactory::Normalisation::None);
Chris@1313:         params.gaplessMode = (gapless ?
Chris@1313:                               AudioFileReaderFactory::GaplessMode::Gapless :
Chris@1313:                               AudioFileReaderFactory::GaplessMode::Gappy);
Chris@757: 
Chris@756: 	AudioFileReader *reader =
Chris@756: 	    AudioFileReaderFactory::createReader
Chris@1313: 	    (audioDir + "/" + audiofile, params);
Chris@1313:         
Chris@756: 	if (!reader) {
Chris@820: #if ( QT_VERSION >= 0x050000 )
Chris@763: 	    QSKIP("Unsupported file, skipping");
Chris@820: #else
Chris@820: 	    QSKIP("Unsupported file, skipping", SkipSingle);
Chris@820: #endif
Chris@756: 	}
Chris@756: 
Chris@1313:         QString extension;
Chris@1313:         sv_samplerate_t fileRate;
Chris@1313:         int channels;
Chris@1313:         int fileBitdepth;
Chris@1313:         getFileMetadata(audiofile, extension, fileRate, channels, fileBitdepth);
Chris@1313: 
Chris@1313:         QString diffFile = testName(audiofile, rate, normalised, gapless);
Chris@1313:         diffFile.replace(".", "_");
Chris@1313:         diffFile.replace(" ", "_");
Chris@1313:         diffFile += ".wav";
Chris@1313:         diffFile = QDir(diffDir).filePath(diffFile);
Chris@1313:         WavFileWriter diffWriter(diffFile, readRate, channels,
Chris@1313:                                  WavFileWriter::WriteToTarget); //!!! NB WriteToTemporary not working, why?
Chris@1313:         QVERIFY(diffWriter.isOK());
Chris@1313:         
Chris@1313:         QCOMPARE((int)reader->getChannelCount(), channels);
Chris@1313:         QCOMPARE(reader->getNativeRate(), fileRate);
Chris@1040:         QCOMPARE(reader->getSampleRate(), readRate);
Chris@757: 
Chris@757: 	AudioTestData tdata(readRate, channels);
Chris@756: 	
Chris@756: 	float *reference = tdata.getInterleavedData();
Chris@1040:         sv_frame_t refFrames = tdata.getFrameCount();
Chris@756: 	
Chris@756: 	// The reader should give us exactly the expected number of
Chris@759: 	// frames, except for mp3/aac files. We ask for quite a lot
Chris@759: 	// more, though, so we can (a) check that we only get the
Chris@759: 	// expected number back (if this is not mp3/aac) or (b) take
Chris@759: 	// into account silence at beginning and end (if it is).
Chris@1041: 	vector<float> test = reader->getInterleavedFrames(0, refFrames + 5000);
Chris@1040: 	sv_frame_t read = test.size() / channels;
Chris@756: 
Chris@1313:         bool perceptual = (extension == "mp3" ||
Chris@1313:                            extension == "aac" ||
Chris@1313:                            extension == "m4a");
Chris@1313:         
Chris@1313:         if (perceptual && !gapless) {
Chris@1313:             // allow silence at start and end
Chris@759:             QVERIFY(read >= refFrames);
Chris@757:         } else {
Chris@759:             QCOMPARE(read, refFrames);
Chris@757:         }
Chris@757: 
Chris@1313:         bool resampled = readRate != fileRate;
Chris@1313:         double maxLimit, rmsLimit;
Chris@1313:         getExpectedThresholds(audiofile,
Chris@1313:                               resampled,
Chris@1313:                               gapless,
Chris@1313:                               normalised,
Chris@1313:                               maxLimit, rmsLimit);
Chris@1313:         
Chris@1313:         double edgeLimit = maxLimit * 3; // in first or final edgeSize frames
Chris@1313:         if (resampled && edgeLimit < 0.1) edgeLimit = 0.1;
Chris@759:         int edgeSize = 100; 
Chris@759: 
Chris@759:         // And we ignore completely the last few frames when upsampling
Chris@1313:         int discard = 1 + int(round(readRate / fileRate));
Chris@759: 
Chris@759:         int offset = 0;
Chris@759: 
Chris@1313:         if (perceptual) {
Chris@759: 
cannam@1314:             // Look for an initial offset.
cannam@1314:             //
cannam@1314:             // We know the first channel has a sinusoid in it. It
cannam@1314:             // should have a peak at 0.4ms (see AudioTestData.h) but
cannam@1314:             // that might have been clipped, which would make it
cannam@1314:             // imprecise. We can tell if it's clipped, though, as
cannam@1314:             // there will be samples having exactly identical
cannam@1314:             // values. So what we look for is the peak if it's not
cannam@1314:             // clipped and, if it is, the first zero crossing after
cannam@1314:             // the peak, which should be at 0.8ms.
cannam@1314: 
Chris@1296:             int expectedPeak = int(0.0004 * readRate);
cannam@1314:             int expectedZC = int(0.0008 * readRate);
cannam@1314:             bool foundPeak = false;
cannam@1314:             for (int i = 1; i+1 < read; ++i) {
cannam@1314:                 float prevSample = test[(i-1) * channels];
cannam@1314:                 float thisSample = test[i * channels];
cannam@1314:                 float nextSample = test[(i+1) * channels];
cannam@1314:                 if (thisSample > 0.8 && nextSample < thisSample) {
cannam@1314:                     foundPeak = true;
cannam@1314:                     if (thisSample > prevSample) {
cannam@1314:                         // not clipped
cannam@1314:                         offset = i - expectedPeak - 1;
cannam@1314:                         break;
cannam@1314:                     }
cannam@1314:                 }
cannam@1314:                 if (foundPeak && (thisSample >= 0.0 && nextSample < 0.0)) {
cannam@1314:                     cerr << "thisSample = " << thisSample << ", nextSample = "
cannam@1314:                          << nextSample << endl;
cannam@1314:                     offset = i - expectedZC - 1;
Chris@759:                     break;
Chris@759:                 }
Chris@759:             }
Chris@1313: 
cannam@1314:             int fileRateEquivalent = int((offset / readRate) * fileRate);
cannam@1314:             
Chris@1313:             std::cerr << "offset = " << offset << std::endl;
cannam@1314:             std::cerr << "at file rate would be " << fileRateEquivalent << std::endl;
Chris@1313: 
Chris@1313:             // Previously our m4a test file had a fixed offset of 1024
Chris@1313:             // at the file sample rate -- this may be because it was
Chris@1313:             // produced by FAAC which did not write in the delay as
Chris@1313:             // metadata? We now have an m4a produced by Core Audio
Chris@1313:             // which gives a 0 offset. What to do...
Chris@1313: 
Chris@1313:             // Anyway, mp3s should have 0 offset in gapless mode and
Chris@1313:             // "something else" otherwise.
Chris@1313:             
Chris@1313:             if (gapless) {
Chris@1313:                 QCOMPARE(offset, 0);
Chris@1313:             }
Chris@759:         }
Chris@756: 
Chris@1313:         vector<vector<float>> diffs(channels);
Chris@1313:             
Chris@756: 	for (int c = 0; c < channels; ++c) {
Chris@1313: 
Chris@1313:             double maxDiff = 0.0;
Chris@1313:             double totalDiff = 0.0;
Chris@1313:             double totalSqrDiff = 0.0;
Chris@1313: 	    int maxIndex = 0;
Chris@1313: 
Chris@1296: 	    for (int i = 0; i < refFrames; ++i) {
Chris@1296:                 int ix = i + offset;
Chris@1296:                 if (ix >= read) {
cannam@1308:                     cerr << "ERROR: audiofile " << audiofile << " reads truncated (read-rate reference frames " << i << " onward, of " << refFrames << ", are lost)" << endl;
Chris@1296:                     QVERIFY(ix < read);
Chris@1296:                 }
Chris@1313: 
Chris@1313:                 float signeddiff =
Chris@1313:                     test[ix * channels + c] -
Chris@1313:                     reference[i * channels + c];
Chris@1313:                     
Chris@1313:                 diffs[c].push_back(signeddiff);
Chris@1313: 
Chris@1296:                 if (ix + discard >= read) {
Chris@1296:                     // we forgive the very edge samples when
Chris@1296:                     // resampling (discard > 0)
Chris@1296:                     continue;
Chris@1296:                 }
Chris@1313:                 
Chris@1313: 		double diff = fabs(signeddiff);
Chris@1313: 
Chris@1313: 		totalDiff += diff;
Chris@1313:                 totalSqrDiff += diff * diff;
Chris@1313:                 
Chris@757:                 // in edge areas, record this only if it exceeds edgeLimit
Chris@1313:                 if (i < edgeSize || i + edgeSize >= refFrames) {
Chris@1313:                     if (diff > edgeLimit && diff > maxDiff) {
Chris@1313:                         maxDiff = diff;
Chris@1313:                         maxIndex = i;
Chris@757:                     }
Chris@757:                 } else {
Chris@1313:                     if (diff > maxDiff) {
Chris@1313:                         maxDiff = diff;
Chris@1313:                         maxIndex = i;
Chris@757:                     }
Chris@756: 		}
Chris@756: 	    }
Chris@1313:                 
Chris@1313: 	    double meanDiff = totalDiff / double(refFrames);
Chris@1313:             double rmsDiff = sqrt(totalSqrDiff / double(refFrames));
cannam@1308: 
cannam@1314:             /*
Chris@1313: 	    cerr << "channel " << c << ": mean diff " << meanDiff << endl;
Chris@1313: 	    cerr << "channel " << c << ":  rms diff " << rmsDiff << endl;
Chris@1313: 	    cerr << "channel " << c << ":  max diff " << maxDiff << " at " << maxIndex << endl;
cannam@1314:             */            
Chris@1313:             if (rmsDiff >= rmsLimit) {
Chris@1313: 		cerr << "ERROR: for audiofile " << audiofile << ": RMS diff = " << rmsDiff << " for channel " << c << " (limit = " << rmsLimit << ")" << endl;
Chris@1313:                 QVERIFY(rmsDiff < rmsLimit);
Chris@1313:             }
Chris@1313: 	    if (maxDiff >= maxLimit) {
Chris@1313: 		cerr << "ERROR: for audiofile " << audiofile << ": max diff = " << maxDiff << " at frame " << maxIndex << " of " << read << " on channel " << c << " (limit = " << maxLimit << ", edge limit = " << edgeLimit << ", mean diff = " << meanDiff << ", rms = " << rmsDiff << ")" << endl;
Chris@1313: 		QVERIFY(maxDiff < maxLimit);
Chris@1313: 	    }
Chris@1313: 
Chris@1313:             // and check for spurious material at end
Chris@1313:             
Chris@1309:             for (sv_frame_t i = refFrames; i + offset < read; ++i) {
Chris@1309:                 sv_frame_t ix = i + offset;
Chris@1313:                 float quiet = 0.1; //!!! allow some ringing - but let's come back to this, it should tail off
cannam@1308:                 float mag = fabsf(test[ix * channels + c]);
cannam@1308:                 if (mag > quiet) {
Chris@1313:                     cerr << "ERROR: audiofile " << audiofile << " contains spurious data after end of reference (found sample " << test[ix * channels + c] << " at index " << ix << " of channel " << c << " after reference+offset ended at " << refFrames+offset << ")" << endl;
cannam@1308:                     QVERIFY(mag < quiet);
cannam@1308:                 }
cannam@1308:             }
Chris@756: 	}
Chris@1313: 
Chris@1313:         float **ptrs = new float*[channels];
Chris@1313:         for (int c = 0; c < channels; ++c) {
Chris@1313:             ptrs[c] = diffs[c].data();
Chris@1313:         }
Chris@1313:         diffWriter.writeSamples(ptrs, refFrames);
Chris@1313:         delete[] ptrs;
Chris@756:     }
Chris@756: };
Chris@756: 
Chris@756: #endif