rockbox/utils/rbutilqt/base/ttsfestival.cpp

/***************************************************************************
*             __________               __   ___.
*   Open      \______   \ ____   ____ |  | _\_ |__   _______  ___
*   Source     |       _//  _ \_/ ___\|  |/ /| __ \ /  _ \  \/  /
*   Jukebox    |    |   (  <_> )  \___|    < | \_\ (  <_> > <  <
*   Firmware   |____|_  /\____/ \___  >__|_ \|___  /\____/__/\_ \
*                     \/            \/     \/    \/            \/
*
*   Copyright (C) 2007 by Dominik Wenger
*
* All files in this archive are subject to the GNU General Public License.
* See the file COPYING in the source tree root for full license agreement.
*
* This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY
* KIND, either express or implied.
*
****************************************************************************/

#include <QtCore>
#include <QTcpSocket>

#include "ttsfestival.h"
#include "utils.h"
#include "rbsettings.h"
#include "Logger.h"

TTSFestival::~TTSFestival()
{
    LOG_INFO() << "Destroying instance";
    stop();
}

TTSBase::Capabilities TTSFestival::capabilities()
{
    return RunInParallel;
}

void TTSFestival::generateSettings()
{
    // server path
    QString exepath = RbSettings::subValue("festival-server",
                        RbSettings::TtsPath).toString();
    if(exepath == "" ) exepath = Utils::findExecutable("festival");
    insertSetting(eSERVERPATH,new EncTtsSetting(this,
                        EncTtsSetting::eSTRING, "Path to Festival server:",
                        exepath,EncTtsSetting::eBROWSEBTN));

    // client path
    QString clientpath = RbSettings::subValue("festival-client",
                        RbSettings::TtsPath).toString();
    if(clientpath == "" ) clientpath = Utils::findExecutable("festival_client");
    insertSetting(eCLIENTPATH,new EncTtsSetting(this,EncTtsSetting::eSTRING,
                        tr("Path to Festival client:"),
                        clientpath,EncTtsSetting::eBROWSEBTN));

    // voice
    EncTtsSetting* setting = new EncTtsSetting(this,
                        EncTtsSetting::eSTRINGLIST, tr("Voice:"),
                        RbSettings::subValue("festival", RbSettings::TtsVoice),
                        getVoiceList(), EncTtsSetting::eREFRESHBTN);
    connect(setting, &EncTtsSetting::refresh,
            this, &TTSFestival::updateVoiceList);
    connect(setting, &EncTtsSetting::dataChanged,
            this, &TTSFestival::clearVoiceDescription);
    insertSetting(eVOICE,setting);

    //voice description
    setting = new EncTtsSetting(this,EncTtsSetting::eREADONLYSTRING,
        tr("Voice description:"),"",EncTtsSetting::eREFRESHBTN);
    connect(setting, &EncTtsSetting::refresh,
            this, &TTSFestival::updateVoiceDescription);
    insertSetting(eVOICEDESC,setting);
}

void TTSFestival::saveSettings()
{
    //save settings in user config
    RbSettings::setSubValue("festival-server",
            RbSettings::TtsPath,getSetting(eSERVERPATH)->current().toString());
    RbSettings::setSubValue("festival-client",
            RbSettings::TtsPath,getSetting(eCLIENTPATH)->current().toString());
    RbSettings::setSubValue("festival",
            RbSettings::TtsVoice,getSetting(eVOICE)->current().toString());

    RbSettings::sync();
}

void TTSFestival::updateVoiceDescription()
{
    // get voice Info with current voice and path
    currentPath = getSetting(eSERVERPATH)->current().toString();
    QString info = getVoiceInfo(getSetting(eVOICE)->current().toString());
    currentPath = "";

    getSetting(eVOICEDESC)->setCurrent(info);
}

void TTSFestival::clearVoiceDescription()
{
    getSetting(eVOICEDESC)->setCurrent("");
}

void TTSFestival::updateVoiceList()
{
   currentPath = getSetting(eSERVERPATH)->current().toString();
   QStringList voiceList = getVoiceList();
   currentPath = "";

   getSetting(eVOICE)->setList(voiceList);
   if(voiceList.size() > 0) getSetting(eVOICE)->setCurrent(voiceList.at(0));
   else getSetting(eVOICE)->setCurrent("");
}

void TTSFestival::startServer()
{
    if(!configOk())
        return;

    if(serverProcess.state() != QProcess::Running)
    {
        QString path;
        /* currentPath is set by the GUI - if it's set, it is the currently set
         path in the configuration GUI; if it's not set, use the saved path */
        if (currentPath.isEmpty())
            path = RbSettings::subValue("festival-server",RbSettings::TtsPath).toString();
        else
            path = currentPath;

        serverProcess.start(path, QStringList("--server"));
        serverProcess.waitForStarted();

        if(serverProcess.state() == QProcess::Running)
            LOG_INFO() << "Server is up and running";
        else
            LOG_ERROR() << "Server failed to start, state:"
                        << serverProcess.state();
    }
}

bool TTSFestival::ensureServerRunning()
{
    if(serverProcess.state() != QProcess::Running)
    {
        startServer();
    }
    return serverProcess.state() == QProcess::Running;
}

bool TTSFestival::start(QString* errStr)
{
    LOG_INFO() << "Starting server with voice"
               << RbSettings::subValue("festival", RbSettings::TtsVoice).toString();

    bool running = ensureServerRunning();
    if (!RbSettings::subValue("festival",RbSettings::TtsVoice).toString().isEmpty())
    {
        /* There's no harm in using both methods to set the voice .. */
        QString voiceSelect = QString("(voice.select '%1)\n")
        .arg(RbSettings::subValue("festival", RbSettings::TtsVoice).toString());
        queryServer(voiceSelect, 3000);

        if(prologFile.open())
        {
          prologFile.write(voiceSelect.toLatin1());
          prologFile.close();
          prologPath = QFileInfo(prologFile).absoluteFilePath();
          LOG_INFO() << "Prolog created at" << prologPath;
        }

    }

    if (!running)
      (*errStr) = tr("Festival could not be started");
    return running;
}

bool TTSFestival::stop()
{
    serverProcess.terminate();
    serverProcess.kill();

    return true;
}

TTSStatus TTSFestival::voice(const QString& text, const QString& wavfile, QString* errStr)
{
    LOG_INFO() << "Voicing" << text << "->" << wavfile;

    QString path = RbSettings::subValue("festival-client",
            RbSettings::TtsPath).toString();
    QStringList cmd;
    cmd << "--server" << "localhost" << "--otype" << "riff" << "--ttw"
        << "--withlisp" << "--output" << wavfile << "--prolog" << prologPath << "-";
    LOG_INFO() << "Client cmd:" << path << cmd;

    QProcess clientProcess;
    clientProcess.start(path, cmd);
    clientProcess.write(QString("%1.\n").arg(text).toLatin1());
    clientProcess.waitForBytesWritten();
    clientProcess.closeWriteChannel();
    clientProcess.waitForReadyRead();
    QString response = clientProcess.readAll();
    response = response.trimmed();
    if(!response.contains("Utterance"))
    {
        LOG_WARNING() << "Could not voice string: " << response;
        *errStr = tr("engine could not voice string");
        return Warning;
        /* do not stop the voicing process because of a single string
        TODO: needs proper settings */
    }
    clientProcess.closeReadChannel(QProcess::StandardError);
    clientProcess.closeReadChannel(QProcess::StandardOutput);
    clientProcess.terminate();
    clientProcess.kill();

    return NoError;
}

bool TTSFestival::configOk()
{
    bool ret;
    if (currentPath == "")
    {
        QString serverPath = RbSettings::subValue("festival-server",
                                    RbSettings::TtsPath).toString();
        QString clientPath = RbSettings::subValue("festival-client",
                                    RbSettings::TtsPath).toString();

        ret = QFileInfo(serverPath).isExecutable() &&
            QFileInfo(clientPath).isExecutable();
        if(RbSettings::subValue("festival",RbSettings::TtsVoice).toString().size() > 0
                && voices.size() > 0)
            ret = ret && (voices.indexOf(RbSettings::subValue("festival",
                            RbSettings::TtsVoice).toString()) != -1);
    }
    else /* If we're currently configuring the server, we need to know that
            the entered path is valid */
        ret = QFileInfo(currentPath).isExecutable();

    return ret;
}

QStringList TTSFestival::getVoiceList()
{
    if(!configOk())
        return QStringList();

    if(voices.size() > 0)
    {
        LOG_INFO() << "Using voice cache";
        return voices;
    }

    QString response = queryServer("(voice.list)", 10000);

    // get the 2nd line. It should be (<voice_name>, <voice_name>)
    response = response.mid(response.indexOf('\n') + 1, -1);
    response = response.left(response.indexOf('\n')).trimmed();

    voices = response.mid(1, response.size()-2).split(' ');

    voices.sort();
    if (voices.size() == 1 && voices[0].size() == 0)
        voices.removeAt(0);
    if (voices.size() > 0)
        LOG_INFO() << "Voices:" << voices;
    else
        LOG_WARNING() << "No voices. Response was:" << response;

    return voices;
}

QString TTSFestival::getVoiceInfo(QString voice)
{
    if(!configOk())
        return "";

    if(!getVoiceList().contains(voice))
        return "";

    if(voiceDescriptions.contains(voice))
        return voiceDescriptions[voice];

    QString response = queryServer(QString("(voice.description '%1)").arg(voice),
                            10000);

    if (response == "")
    {
        voiceDescriptions[voice]=tr("No description available");
    }
    else
    {
        response = response.remove(QRegularExpression("(description \".*\")",
                    QRegularExpression::CaseInsensitiveOption));
        LOG_INFO() << "voiceInfo w/o descr:" << response;
        response = response.remove(')');
#if QT_VERSION >= 0x050e00
        QStringList responseLines = response.split('(', Qt::SkipEmptyParts);
#else
        QStringList responseLines = response.split('(', QString::SkipEmptyParts);
#endif
        responseLines.removeAt(0); // the voice name itself

        QString description;
        foreach(QString line, responseLines)
        {
            line = line.remove('(');
            line = line.simplified();

            line[0] = line[0].toUpper(); // capitalize the key

            int firstSpace = line.indexOf(' ');
            if (firstSpace > 0)
            {
                // add a colon between the key and the value
                line = line.insert(firstSpace, ':');
                // capitalize the value
                line[firstSpace+2] = line[firstSpace+2].toUpper();
            }

            description += line + "\n";
        }
        voiceDescriptions[voice] = description.trimmed();
    }

    return voiceDescriptions[voice];
}

QString TTSFestival::queryServer(QString query, int timeout)
{
    // make sure we always abort at some point.
    if(timeout == 0)
        timeout = 60000;
    if(!configOk())
        return "";

    // this operation could take some time
    emit busy(true);

    LOG_INFO() << "queryServer with" << query;

    if (!ensureServerRunning())
    {
      LOG_ERROR() << "queryServer: ensureServerRunning failed";
      emit busy(false);
      return "";
    }


    QDateTime endTime = QDateTime::currentDateTime().addMSecs(timeout);

    /* Festival is *extremely* unreliable. Although at this
     * point we are sure that SIOD is accepting commands,
     * we might end up with an empty response. Hence, the loop.
     */
    QTcpSocket socket;
    QString response;
    while(QDateTime::currentDateTime() < endTime)
    {
        QCoreApplication::processEvents(QEventLoop::AllEvents, 50);

        if(socket.state() != QAbstractSocket::ConnectedState)
        {
            LOG_INFO() << "socket not (yet) connected, trying again.";
            socket.connectToHost("localhost", 1314);
            // appears we need to recheck the state still.
            socket.waitForConnected();
        }
        else
        {
            // seems to be necessary to resend the request at times.
            socket.write(QString("%1\n").arg(query).toLatin1());
            socket.waitForBytesWritten();
            socket.waitForReadyRead();

            // we might not get the complete response on the first read.
            // Concatenate until we got a full response.
            response += socket.readAll();

            // The query response ends with this.
            if (response.contains("ft_StUfF_keyOK"))
            {
                break;
            }
        }

        /* make sure we wait a little as we don't want to flood the server
         * with requests */
        QDateTime tmpEndTime = QDateTime::currentDateTime().addMSecs(500);
        while(QDateTime::currentDateTime() < tmpEndTime)
            QCoreApplication::processEvents(QEventLoop::AllEvents);
    }
    emit busy(false);
    socket.disconnectFromHost();

    if(response == "nil")
    {
        return "";
    }

    /* The response starts with "LP\n", and ends with "ft_StUfF_keyOK", but we
     * could get trailing data -- we might have sent the request more than
     * once. Use a regex to get the actual response part.
     */
    QRegularExpression regex("LP\\n(.*?)\\nft_StUfF_keyOK",
                             QRegularExpression::MultilineOption
                             | QRegularExpression::DotMatchesEverythingOption);
    QRegularExpressionMatch match = regex.match(response);
    if(match.hasMatch())
    {
        response = match.captured(1);
    }
    else {
        LOG_WARNING() << "Invalid Festival response." << response;
    }

    return response.trimmed();
}