path: root/plugins/soundsourcemediafoundation/soundsourcemediafoundation.cpp

/**
 * \file soundsourcemediafoundation.cpp
 * \author Bill Good <bkgood at gmail dot com>
 * \author Albert Santoni <alberts at mixxx dot org>
 * \date Jan 10, 2011
 * \note This file uses COM interfaces defined in Windows 7 and later added to
 * Vista and Server 2008 via the "Platform Update Supplement for Windows Vista
 * and for Windows Server 2008" (http://support.microsoft.com/kb/2117917).
 * Earlier versions of Vista (and possibly Server 2008) have some Media
 * Foundation interfaces but not the required IMFSourceReader, and are missing
 * the Microsoft-provided AAC decoder. XP does not include Media Foundation.
 */

/***************************************************************************
 *                                                                         *
 *   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.                                   *
 *                                                                         *
 ***************************************************************************/

#include <QtDebug>
#include <taglib/mp4file.h>
#include <windows.h>
#include <mfapi.h>
#include <mfidl.h>
#include <mfreadwrite.h>
#include <mferror.h>
#include <propvarutil.h>

#include "soundsourcemediafoundation.h"

const int kBitsPerSample = 16;
const int kNumChannels = 2;
const int kSampleRate = 44100;
const int kLeftoverSize = 4096; // in int16's, this seems to be the size MF AAC
// decoder likes to give

const static bool sDebug = false;

/** Microsoft examples use this snippet often. */
template<class T> static void safeRelease(T **ppT)
{
    if (*ppT) {
        (*ppT)->Release();
        *ppT = NULL;
    }
}

SoundSourceMediaFoundation::SoundSourceMediaFoundation(QString filename)
    : SoundSource(filename)
    , m_file(filename)
    , m_pReader(NULL)
    , m_pAudioType(NULL)
    , m_wcFilename(NULL)
    , m_nextFrame(0)
    , m_leftoverBuffer(NULL)
    , m_leftoverBufferSize(0)
    , m_leftoverBufferLength(0)
    , m_leftoverBufferPosition(0)
    , m_mfDuration(0)
    , m_iCurrentPosition(0)
    , m_dead(false)
    , m_seeking(false)
{
    // these are always the same, might as well just stick them here
    // -bkgood
    m_iChannels = kNumChannels;
    m_iSampleRate = kSampleRate;

    // http://social.msdn.microsoft.com/Forums/en/netfxbcl/thread/35c6a451-3507-40c8-9d1c-8d4edde7c0cc
    // gives maximum path + file length as 248 + 260, using that -bkgood
    m_wcFilename = new wchar_t[248 + 260];
}

SoundSourceMediaFoundation::~SoundSourceMediaFoundation()
{
    delete [] m_wcFilename;
    delete [] m_leftoverBuffer;

    safeRelease(&m_pReader);
    safeRelease(&m_pAudioType);
    MFShutdown();
    CoUninitialize();
}

Result SoundSourceMediaFoundation::open()
{
    if (sDebug) {
        qDebug() << "open()" << m_qFilename;
    }

    QString qurlStr(m_qFilename);
    int wcFilenameLength(m_qFilename.toWCharArray(m_wcFilename));
    // toWCharArray does not append a null terminator to the string!
    m_wcFilename[wcFilenameLength] = '\0';

    HRESULT hr(S_OK);
    // Initialize the COM library.
    hr = CoInitializeEx(NULL, COINIT_APARTMENTTHREADED | COINIT_DISABLE_OLE1DDE);
    if (FAILED(hr)) {
        qWarning() << "SSMF: failed to initialize COM";
        return ERR;
    }

    // Initialize the Media Foundation platform.
    hr = MFStartup(MF_VERSION);
    if (FAILED(hr)) {
        qWarning() << "SSMF: failed to initialize Media Foundation";
        return ERR;
    }

    // Create the source reader to read the input file.
    hr = MFCreateSourceReaderFromURL(m_wcFilename, NULL, &m_pReader);
    if (FAILED(hr)) {
        qWarning() << "SSMF: Error opening input file:" << m_qFilename;
        return ERR;
    }

    if (!configureAudioStream()) {
        qWarning() << "SSMF: Error configuring audio stream.";
        return ERR;
    }

    if (!readProperties()) {
        qWarning() << "SSMF::readProperties failed";
        return ERR;
    }

    //Seek to position 0, which forces us to skip over all the header frames.
    //This makes sure we're ready to just let the Analyser rip and it'll
    //get the number of samples it expects (ie. no header frames).
    seek(0);

    return OK;
}

long SoundSourceMediaFoundation::seek(long filepos)
{
    if (sDebug) { qDebug() << "seek()" << filepos; }
    PROPVARIANT prop;
    HRESULT hr(S_OK);
    qint64 seekTarget(filepos / kNumChannels);
    qint64 mfSeekTarget(mfFromFrame(seekTarget) - 1);
    // minus 1 here seems to make our seeking work properly, otherwise we will
    // (more often than not, maybe always) seek a bit too far (although not
    // enough for our calculatedFrameFromMF <= nextFrame assertion in ::read).
    // Has something to do with 100ns MF units being much smaller than most
    // frame offsets (in seconds) -bkgood
    long result = m_iCurrentPosition;
    if (m_dead) {
        return result;
    }

    // this doesn't fail, see MS's implementation
    hr = InitPropVariantFromInt64(mfSeekTarget < 0 ? 0 : mfSeekTarget, &prop);


    hr = m_pReader->Flush(MF_SOURCE_READER_FIRST_AUDIO_STREAM);
    if (FAILED(hr)) {
        qWarning() << "SSMF: failed to flush before seek";
    }

    // http://msdn.microsoft.com/en-us/library/dd374668(v=VS.85).aspx
    hr = m_pReader->SetCurrentPosition(GUID_NULL, prop);
    if (FAILED(hr)) {
        // nothing we can do here as we can't fail (no facility to other than
        // crashing mixxx)
        qWarning() << "SSMF: failed to seek" << (
            hr == MF_E_INVALIDREQUEST ? "Sample requests still pending" : "");
    } else {
        result = filepos;
    }
    PropVariantClear(&prop);

    // record the next frame so that we can make sure we're there the next
    // time we get a buffer from MFSourceReader
    m_nextFrame = seekTarget;
    m_seeking = true;
    m_iCurrentPosition = result;
    return result;
}

unsigned int SoundSourceMediaFoundation::read(unsigned long size,
    const SAMPLE *destination)
{
    if (sDebug) { qDebug() << "read()" << size; }
    SAMPLE *destBuffer(const_cast<SAMPLE*>(destination));
    size_t framesRequested(size / kNumChannels);
    size_t framesNeeded(framesRequested);

    // first, copy frames from leftover buffer IF the leftover buffer is at
    // the correct frame
    if (m_leftoverBufferLength > 0 && m_leftoverBufferPosition == m_nextFrame) {
        copyFrames(destBuffer, &framesNeeded, m_leftoverBuffer,
            m_leftoverBufferLength);
        if (m_leftoverBufferLength > 0) {
            if (framesNeeded != 0) {
                qWarning() << __FILE__ << __LINE__
                           << "WARNING: Expected frames needed to be 0. Abandoning this file.";
                m_dead = true;
            }
            m_leftoverBufferPosition += framesRequested;
        }
    } else {
        // leftoverBuffer already empty or in the wrong position, clear it
        m_leftoverBufferLength = 0;
    }

    while (!m_dead && framesNeeded > 0) {
        HRESULT hr(S_OK);
        DWORD dwFlags(0);
        qint64 timestamp(0);
        IMFSample *pSample(NULL);
        bool error(false); // set to true to break after releasing

        hr = m_pReader->ReadSample(
            MF_SOURCE_READER_FIRST_AUDIO_STREAM, // [in] DWORD dwStreamIndex,
            0,                                   // [in] DWORD dwControlFlags,
            NULL,