Mercurial > hg > svcore
view data/model/AlignmentModel.cpp @ 297:c022976d18e8
* Merge from sv-match-alignment branch (excluding alignment-specific document).
- add aggregate wave model (not yet complete enough to be added as a true
model in a layer, but there's potential)
- add play solo mode
- add alignment model -- unused in plain SV
- fix two plugin leaks
- add m3u playlist support (opens all files at once, potentially hazardous)
- fix retrieval of pre-encoded URLs
- add ability to resample audio files on import, so as to match rates with
other files previously loaded; add preference for same
- add preliminary support in transform code for range and rate of transform
input
- reorganise preferences dialog, move dark-background option to preferences,
add option for temporary directory location
author | Chris Cannam |
---|---|
date | Fri, 28 Sep 2007 13:56:38 +0000 |
parents | |
children | df707a61b23f |
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 2007 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 "AlignmentModel.h" #include "SparseTimeValueModel.h" AlignmentModel::AlignmentModel(Model *reference, Model *aligned, Model *inputModel, SparseTimeValueModel *path) : m_reference(reference), m_aligned(aligned), m_inputModel(inputModel), m_path(path), m_reversePath(0), m_pathComplete(false) { connect(m_path, SIGNAL(modelChanged()), this, SLOT(pathChanged())); connect(m_path, SIGNAL(modelChanged(size_t, size_t)), this, SLOT(pathChanged(size_t, size_t))); connect(m_path, SIGNAL(completionChanged()), this, SLOT(pathCompletionChanged())); constructReversePath(); } AlignmentModel::~AlignmentModel() { delete m_inputModel; delete m_path; delete m_reversePath; } bool AlignmentModel::isOK() const { return m_path->isOK(); } size_t AlignmentModel::getStartFrame() const { //!!! do we care about distinct rates? size_t a = m_reference->getStartFrame(); size_t b = m_aligned->getStartFrame(); return std::min(a, b); } size_t AlignmentModel::getEndFrame() const { //!!! do we care about distinct rates? size_t a = m_reference->getEndFrame(); size_t b = m_aligned->getEndFrame(); return std::max(a, b); } size_t AlignmentModel::getSampleRate() const { return m_reference->getSampleRate(); } Model * AlignmentModel::clone() const { return new AlignmentModel (m_reference, m_aligned, m_inputModel ? m_inputModel->clone() : 0, m_path ? static_cast<SparseTimeValueModel *>(m_path->clone()) : 0); } bool AlignmentModel::isReady(int *completion) const { return m_path->isReady(completion); } const ZoomConstraint * AlignmentModel::getZoomConstraint() const { return m_path->getZoomConstraint(); } const Model * AlignmentModel::getReferenceModel() const { return m_reference; } const Model * AlignmentModel::getAlignedModel() const { return m_aligned; } size_t AlignmentModel::toReference(size_t frame) const { // std::cerr << "AlignmentModel::toReference(" << frame << ")" << std::endl; if (!m_reversePath) constructReversePath(); return align(m_reversePath, frame); } size_t AlignmentModel::fromReference(size_t frame) const { // std::cerr << "AlignmentModel::fromReference(" << frame << ")" << std::endl; return align(m_path, frame); } void AlignmentModel::pathChanged() { } void AlignmentModel::pathChanged(size_t, size_t) { if (!m_pathComplete) return; constructReversePath(); } void AlignmentModel::pathCompletionChanged() { if (!m_pathComplete) { int completion = 0; m_path->isReady(&completion); std::cerr << "AlignmentModel::pathCompletionChanged: completion = " << completion << std::endl; m_pathComplete = (completion == 100); //!!! a bit of a hack if (m_pathComplete) { constructReversePath(); delete m_inputModel; m_inputModel = 0; } } emit completionChanged(); } void AlignmentModel::constructReversePath() const { if (!m_reversePath) { m_reversePath = new SparseTimeValueModel (m_path->getSampleRate(), m_path->getResolution(), false); } m_reversePath->clear(); SparseTimeValueModel::PointList points = m_path->getPoints(); for (SparseTimeValueModel::PointList::const_iterator i = points.begin(); i != points.end(); ++i) { long frame = i->frame; float value = i->value; long rframe = lrintf(value * m_aligned->getSampleRate()); float rvalue = (float)frame / (float)m_reference->getSampleRate(); m_reversePath->addPoint (SparseTimeValueModel::Point(rframe, rvalue, "")); } std::cerr << "AlignmentModel::constructReversePath: " << m_reversePath->getPointCount() << " points" << std::endl; } size_t AlignmentModel::align(SparseTimeValueModel *path, size_t frame) const { // The path consists of a series of points, each with x (time) // equal to the time on the source model and y (value) equal to // the time on the target model. Times and values are both // monotonically increasing. const SparseTimeValueModel::PointList &points = path->getPoints(); if (points.empty()) { // std::cerr << "AlignmentModel::align: No points" << std::endl; return frame; } SparseTimeValueModel::Point point(frame); SparseTimeValueModel::PointList::const_iterator i = points.lower_bound(point); if (i == points.end()) --i; float time = i->value; size_t rv = lrintf(time * getSampleRate()); //!!! interpolate! // std::cerr << "AlignmentModel::align: rv = " << rv << std::endl; return rv; }