Mercurial > hg > svcore
view transform/ModelTransformerFactory.cpp @ 641:fd268881d1ea
* Merge
author | Chris Cannam |
---|---|
date | Fri, 17 Sep 2010 13:13:23 +0100 |
parents | 5ab561f664f2 |
children | 12578237b99c |
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 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" #ifndef NO_SV_GUI #include "widgets/PluginParameterDialog.h" #endif #include "data/model/DenseTimeValueModel.h" #include <vamp-hostsdk/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); #ifndef NO_SV_GUI 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); dialog->setMoreInfoUrl(TransformFactory::getInstance()-> getTransformInfoUrl(transform.getIdentifier())); 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); #endif TransformFactory::getInstance()-> makeContextConsistentWithPlugin(transform, plugin); configurationXml = PluginXml(plugin).toXmlString(); #ifndef NO_SV_GUI delete dialog; #endif 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; } 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. } }