view transform/ModelTransformerFactory.cpp @ 408:115f60df1e4d

* Speed up spectrogram painting by releasing mutex in FFTDataServer
while calculating data prior to writing it, and by adding whole-column
value query methods to FFT objects
* Add paint cache to Thumbwheel -- repaints of this widget were slowing
down the whole spectrogram repaint
* More uses of MutexLocker (named and with debug) and more profile
points
* Make startup much quicker some of the time, with OSC server in place

/* -*- 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 file copyright 2006 Chris Cannam and QMUL.
   
    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.
*/

#include "ModelTransformerFactory.h"

#include "FeatureExtractionModelTransformer.h"
#include "RealTimeEffectModelTransformer.h"

#include "TransformFactory.h"

#include "base/AudioPlaySource.h"

#include "plugin/FeatureExtractionPluginFactory.h"
#include "plugin/RealTimePluginFactory.h"
#include "plugin/PluginXml.h"

#include "widgets/PluginParameterDialog.h"

#include "data/model/DenseTimeValueModel.h"

#include "vamp-sdk/PluginHostAdapter.h"

#include <iostream>
#include <set>

#include <QRegExp>

ModelTransformerFactory *
ModelTransformerFactory::m_instance = new ModelTransformerFactory;

ModelTransformerFactory *
ModelTransformerFactory::getInstance()
{
    return m_instance;
}

ModelTransformerFactory::~ModelTransformerFactory()
{
}

bool
ModelTransformerFactory::getChannelRange(TransformId identifier,
                                         Vamp::PluginBase *plugin,
                                         int &minChannels, int &maxChannels)
{
    Vamp::Plugin *vp = 0;
    if ((vp = dynamic_cast<Vamp::Plugin *>(plugin)) ||
        (vp = dynamic_cast<Vamp::PluginHostAdapter *>(plugin))) {
        minChannels = vp->getMinChannelCount();
        maxChannels = vp->getMaxChannelCount();
        return true;
    } else {
        return TransformFactory::getInstance()->
            getTransformChannelRange(identifier, minChannels, maxChannels);
    }
}

ModelTransformer::Input
ModelTransformerFactory::getConfigurationForTransform(Transform &transform,
                                                      const std::vector<Model *> &candidateInputModels,
                                                      Model *defaultInputModel,
                                                      AudioPlaySource *source,
                                                      size_t startFrame,
                                                      size_t duration)
{
    ModelTransformer::Input input(0);

    if (candidateInputModels.empty()) return input;

    //!!! This will need revision -- we'll have to have a callback
    //from the dialog for when the candidate input model is changed,
    //as we'll need to reinitialise the channel settings in the dialog
    Model *inputModel = candidateInputModels[0];
    QStringList candidateModelNames;
    QString defaultModelName;
    std::map<QString, Model *> modelMap;
    for (size_t i = 0; i < candidateInputModels.size(); ++i) {
        QString modelName = candidateInputModels[i]->objectName();
        QString origModelName = modelName;
        int dupcount = 1;
        while (modelMap.find(modelName) != modelMap.end()) {
            modelName = tr("%1 <%2>").arg(origModelName).arg(++dupcount);
        }
        modelMap[modelName] = candidateInputModels[i];
        candidateModelNames.push_back(modelName);
        if (candidateInputModels[i] == defaultInputModel) {
            defaultModelName = modelName;
        }
    }

    QString id = transform.getPluginIdentifier();
    QString output = transform.getOutput();
    QString outputLabel = "";
    QString outputDescription = "";
    
    bool ok = false;
    QString configurationXml = m_lastConfigurations[transform.getIdentifier()];

    std::cerr << "last configuration: " << configurationXml.toStdString() << std::endl;

    Vamp::PluginBase *plugin = 0;

    bool frequency = false;
    bool effect = false;
    bool generator = false;

    if (FeatureExtractionPluginFactory::instanceFor(id)) {

        std::cerr << "getConfigurationForTransform: instantiating Vamp plugin" << std::endl;

        Vamp::Plugin *vp =
            FeatureExtractionPluginFactory::instanceFor(id)->instantiatePlugin
            (id, inputModel->getSampleRate());

        if (vp) {

            plugin = vp;
            frequency = (vp->getInputDomain() == Vamp::Plugin::FrequencyDomain);

            std::vector<Vamp::Plugin::OutputDescriptor> od =
                vp->getOutputDescriptors();
            if (od.size() > 1) {
                for (size_t i = 0; i < od.size(); ++i) {
                    if (od[i].identifier == output.toStdString()) {
                        outputLabel = od[i].name.c_str();
                        outputDescription = od[i].description.c_str();
                        break;
                    }
                }
            }
        }

    } else if (RealTimePluginFactory::instanceFor(id)) {

        RealTimePluginFactory *factory = RealTimePluginFactory::instanceFor(id);
        const RealTimePluginDescriptor *desc = factory->getPluginDescriptor(id);

        if (desc->audioInputPortCount > 0 && 
            desc->audioOutputPortCount > 0 &&
            !desc->isSynth) {
            effect = true;
        }

        if (desc->audioInputPortCount == 0) {
            generator = true;
        }

        if (output != "A") {
            int outputNo = output.toInt();
            if (outputNo >= 0 && outputNo < int(desc->controlOutputPortCount)) {
                outputLabel = desc->controlOutputPortNames[outputNo].c_str();
            }
        }

        size_t sampleRate = inputModel->getSampleRate();
        size_t blockSize = 1024;
        size_t channels = 1;
        if (effect && source) {
            sampleRate = source->getTargetSampleRate();
            blockSize = source->getTargetBlockSize();
            channels = source->getTargetChannelCount();
        }

        RealTimePluginInstance *rtp = factory->instantiatePlugin
            (id, 0, 0, sampleRate, blockSize, channels);

        plugin = rtp;

        if (effect && source && rtp) {
            source->setAuditioningEffect(rtp);
        }
    }

    if (plugin) {

        // Ensure block size etc are valid
        TransformFactory::getInstance()->
            makeContextConsistentWithPlugin(transform, plugin);

        // Prepare the plugin with any existing parameters already
        // found in the transform
        TransformFactory::getInstance()->
            setPluginParameters(transform, plugin);
        
        // For this interactive usage, we want to override those with
        // whatever the user chose last time around
        PluginXml(plugin).setParametersFromXml(configurationXml);

        int sourceChannels = 1;
        if (dynamic_cast<DenseTimeValueModel *>(inputModel)) {
            sourceChannels = dynamic_cast<DenseTimeValueModel *>(inputModel)
                ->getChannelCount();
        }

        int minChannels = 1, maxChannels = sourceChannels;
        getChannelRange(transform.getIdentifier(), plugin,
                        minChannels, maxChannels);

        int targetChannels = sourceChannels;
        if (!effect) {
            if (sourceChannels < minChannels) targetChannels = minChannels;
            if (sourceChannels > maxChannels) targetChannels = maxChannels;
        }

        int defaultChannel = -1; //!!! no longer saved! [was context.channel]

        PluginParameterDialog *dialog = new PluginParameterDialog(plugin);

        if (candidateModelNames.size() > 1 && !generator) {
            dialog->setCandidateInputModels(candidateModelNames,
                                            defaultModelName);
        }

        if (startFrame != 0 || duration != 0) {
            dialog->setShowSelectionOnlyOption(true);
        }

        if (targetChannels > 0) {
            dialog->setChannelArrangement(sourceChannels, targetChannels,
                                          defaultChannel);
        }
        
        dialog->setOutputLabel(outputLabel, outputDescription);
        
        dialog->setShowProcessingOptions(true, frequency);

        if (dialog->exec() == QDialog::Accepted) {
            ok = true;
        }

        QString selectedInput = dialog->getInputModel();
        if (selectedInput != "") {
            if (modelMap.find(selectedInput) != modelMap.end()) {
                inputModel = modelMap[selectedInput];
                std::cerr << "Found selected input \"" << selectedInput.toStdString() << "\" in model map, result is " << inputModel << std::endl;
            } else {
                std::cerr << "Failed to find selected input \"" << selectedInput.toStdString() << "\" in model map" << std::endl;
            }
        } else {
            std::cerr << "Selected input empty: \"" << selectedInput.toStdString() << "\"" << std::endl;
        }
        
        // Write parameters back to transform object
        TransformFactory::getInstance()->
            setParametersFromPlugin(transform, plugin);

        input.setChannel(dialog->getChannel());
        
        //!!! The dialog ought to be taking & returning transform
        //objects and input objects and stuff rather than passing
        //around all this misc stuff, but that's for tomorrow
        //(whenever that may be)

        if (startFrame != 0 || duration != 0) {
            if (dialog->getSelectionOnly()) {
                transform.setStartTime(RealTime::frame2RealTime
                                       (startFrame, inputModel->getSampleRate()));
                transform.setDuration(RealTime::frame2RealTime
                                      (duration, inputModel->getSampleRate()));
            }
        }

        size_t stepSize = 0, blockSize = 0;
        WindowType windowType = HanningWindow;

        dialog->getProcessingParameters(stepSize,
                                        blockSize,
                                        windowType);

        transform.setStepSize(stepSize);
        transform.setBlockSize(blockSize);
        transform.setWindowType(windowType);

        TransformFactory::getInstance()->
            makeContextConsistentWithPlugin(transform, plugin);

        configurationXml = PluginXml(plugin).toXmlString();

        delete dialog;

        if (effect && source) {
            source->setAuditioningEffect(0); // will delete our plugin
        } else {
            delete plugin;
        }
    }

    if (ok) {
        m_lastConfigurations[transform.getIdentifier()] = configurationXml;
        input.setModel(inputModel);
    }

    return input;
}
/*!!!
PluginTransformer::ExecutionContext
ModelTransformerFactory::getDefaultContextForTransformer(TransformId identifier,
                                                Model *inputModel)
{
    PluginTransformer::ExecutionContext context(-1);

    QString id = identifier.section(':', 0, 2);

    if (FeatureExtractionPluginFactory::instanceFor(id)) {

        Vamp::Plugin *vp =
            FeatureExtractionPluginFactory::instanceFor(id)->instantiatePlugin
            (id, inputModel ? inputModel->getSampleRate() : 48000);

        if (vp) {
            context = PluginTransformer::ExecutionContext(-1, vp);
            delete vp;
        }
    }

    return context;
}
*/
ModelTransformer *
ModelTransformerFactory::createTransformer(const Transform &transform,
                                           const ModelTransformer::Input &input)
{
    ModelTransformer *transformer = 0;

    QString id = transform.getPluginIdentifier();

    if (FeatureExtractionPluginFactory::instanceFor(id)) {

        transformer =
            new FeatureExtractionModelTransformer(input, transform);

    } else if (RealTimePluginFactory::instanceFor(id)) {

        transformer =
            new RealTimeEffectModelTransformer(input, transform);

    } else {
        std::cerr << "ModelTransformerFactory::createTransformer: Unknown transform \""
                  << transform.getIdentifier().toStdString() << "\"" << std::endl;
        return transformer;
    }

    if (transformer) transformer->setObjectName(transform.getIdentifier());
    return transformer;
}

Model *
ModelTransformerFactory::transform(const Transform &transform,
                                   const ModelTransformer::Input &input,
                                   QString &message)
{
    std::cerr << "ModelTransformerFactory::transform: Constructing transformer with input model " << input.getModel() << std::endl;

    ModelTransformer *t = createTransformer(transform, input);
    if (!t) return 0;

    connect(t, SIGNAL(finished()), this, SLOT(transformerFinished()));

    m_runningTransformers.insert(t);

    t->start();
    Model *model = t->detachOutputModel();

    if (model) {
        QString imn = input.getModel()->objectName();
        QString trn =
            TransformFactory::getInstance()->getTransformFriendlyName
            (transform.getIdentifier());
        if (imn != "") {
            if (trn != "") {
                model->setObjectName(tr("%1: %2").arg(imn).arg(trn));
            } else {
                model->setObjectName(imn);
            }
        } else if (trn != "") {
            model->setObjectName(trn);
        }
    } else {
        t->wait();
    }

    message = t->getMessage();

    return model;
}

void
ModelTransformerFactory::transformerFinished()
{
    QObject *s = sender();
    ModelTransformer *transformer = dynamic_cast<ModelTransformer *>(s);
    
    std::cerr << "ModelTransformerFactory::transformerFinished(" << transformer << ")" << std::endl;

    if (!transformer) {
	std::cerr << "WARNING: ModelTransformerFactory::transformerFinished: sender is not a transformer" << std::endl;
	return;
    }

    if (m_runningTransformers.find(transformer) == m_runningTransformers.end()) {
        std::cerr << "WARNING: ModelTransformerFactory::transformerFinished(" 
                  << transformer
                  << "): I have no record of this transformer running!"
                  << std::endl;
    }

    m_runningTransformers.erase(transformer);

    transformer->wait(); // unnecessary but reassuring
    delete transformer;
}

void
ModelTransformerFactory::modelAboutToBeDeleted(Model *m)
{
    TransformerSet affected;

    for (TransformerSet::iterator i = m_runningTransformers.begin();
         i != m_runningTransformers.end(); ++i) {

        ModelTransformer *t = *i;

        if (t->getInputModel() == m || t->getOutputModel() == m) {
            affected.insert(t);
        }
    }

    for (TransformerSet::iterator i = affected.begin();
         i != affected.end(); ++i) {

        ModelTransformer *t = *i;

        t->abandon();

        t->wait(); // this should eventually call back on
                   // transformerFinished, which will remove from
                   // m_runningTransformers and delete.
    }
}