annotate base/Layer.h @ 20:742e6882e187

* Refactor sparse models.  Previously the 1D and time-value models duplicated
  a lot of code; now there is a base class (SparseModel) templated on the
  stored point type, and the subclasses define point types with the necessary
  characteristics.

* Add NoteModel, a new SparseModel subclass.

* Reorganise local feature description display.  Instead of asking the layer
  to draw its own, just query it for a textual description and draw that in
  Pane.  Greatly simplifies this part of the layer code.

* Add local feature descriptions to colour 3D plot and waveform layers.

* Add pitch in MIDI-pitch-and-cents to spectrogram layer.

* Give AudioGenerator its own mutex to shorten lock times in CallbackPlaySource.

* Minor adjustments to layers menu &c

/* -*- c-basic-offset: 4 -*-  vi:set ts=8 sts=4 sw=4: */

/*
    A waveform viewer and audio annotation editor.
    Chris Cannam, Queen Mary University of London, 2005-2006
    
    This is experimental software.  Not for distribution.
*/

#ifndef _VIEWER_H_
#define _VIEWER_H_

#include "PropertyContainer.h"
#include "XmlExportable.h"

#include <QObject>
#include <QRect>
#include <QXmlAttributes>

class ZoomConstraint;
class Model;
class QPainter;
class View;
class QMouseEvent;

/**
 * The base class for visual representations of the data found in a
 * Model.  Layers are expected to be able to draw themselves onto a
 * View, and may also be editable.
 */

class Layer : public QObject,
	      public PropertyContainer,
	      public XmlExportable
{
    Q_OBJECT

public:
    Layer(View *w);
    virtual ~Layer();

    virtual const Model *getModel() const = 0;
    virtual const ZoomConstraint *getZoomConstraint() const { return 0; }
    virtual void paint(QPainter &, QRect) const = 0;   

    enum VerticalPosition {
	PositionTop, PositionMiddle, PositionBottom
    };
    virtual VerticalPosition getPreferredTimeRulerPosition() const {
	return PositionMiddle;
    }
    virtual VerticalPosition getPreferredFrameCountPosition() const {
	return PositionBottom;
    }

    virtual QString getPropertyContainerIconName() const;

    virtual QString getPropertyContainerName() const {
	return objectName();
    }

    virtual int getVerticalScaleWidth(QPainter &) const { return 0; }
    virtual void paintVerticalScale(QPainter &, QRect) const { }

    virtual QString getFeatureDescription(QPoint &) const {
	return "";
    }

    //!!! We also need a method (like the vertical scale method) for
    //drawing additional scales like a colour scale.  That is, unless
    //all applicable layers can actually do this from
    //paintVerticalScale as well?

    // Select mode:
    //
    //!!! Next, a method that gets the frame of the nearest feature in
    //a particular snap direction.  This would be used for selection
    //mode, where we're selecting from the waveform based on feature
    //location.  Do we need multi-select on features as well?  This is
    //an issue; if you select a feature are you selecting that feature
    //(in which case what do you do with it?) or a section of the
    //underlying waveform?

    virtual int getNearestFeatureFrame(int frame,
				       size_t &resolution,
				       bool snapRight = true) const {
	resolution = 1;
	return frame;
    }

    // Draw and edit modes:
    //
    // Layer needs to get actual mouse events, I guess.  Draw mode is
    // probably the easier.

    virtual void drawStart(QMouseEvent *) { }
    virtual void drawDrag(QMouseEvent *) { }
    virtual void drawEnd(QMouseEvent *) { }

    virtual void editStart(QMouseEvent *) { }
    virtual void editDrag(QMouseEvent *) { }
    virtual void editEnd(QMouseEvent *) { }

    // Text mode:
    //
    // Label nearest feature.  We need to get the feature coordinates
    // and current label from the layer, and then the pane can pop up
    // a little text entry dialog at the right location.  Or we edit
    // in place?  Probably the dialog is easier.

    /**
     * This should return true if the view can safely be scrolled
     * automatically by the widget (simply copying the existing data
     * and then refreshing the exposed area) without altering its
     * meaning.  For the widget as a whole this is usually not
     * possible because of invariant (non-scrolling) material
     * displayed over the top, but the widget may be able to optimise
     * scrolling better if it is known that individual views can be
     * scrolled safely in this way.
     */
    virtual bool isLayerScrollable() const { return true; }

    /**
     * This should return true if the layer completely obscures any
     * underlying layers.  It's used to determine whether the view can
     * safely draw any selection rectangles under the layer instead of
     * over it, in the case where the layer is not scrollable and
     * therefore needs to be redrawn each time (so that the selection
     * rectangle can be cached).
     */
    virtual bool isLayerOpaque() const { return false; }

    /**
     * This should return true if the layer can be edited by the user.
     * If this is the case, the appropriate edit tools may be made
     * available by the application and the layer's drawStart/Drag/End
     * and editStart/Drag/End methods should be implemented.
     */
    virtual bool isLayerEditable() const { return false; }

    /**
     * Return the proportion of background work complete in drawing
     * this view, as a percentage -- in most cases this will be the
     * value returned by pointer from a call to the underlying model's
     * isReady(int *) call.  The widget may choose to show a progress
     * meter if it finds that this returns < 100 at any given moment.
     */
    virtual int getCompletion() const { return 100; }

    virtual void setObjectName(const QString &name);

    /**
     * Return the pixel x-coordinate corresponding to a given sample
     * frame (which may be negative).
     */
    int getXForFrame(long frame) const;

    /**
     * Return the closest frame to the given pixel x-coordinate.
     */
    long getFrameForX(int x) const;

    /**
     * Convert the layer's data (though not those of the model it
     * refers to) into an XML string for file output.  This class
     * implements the basic name/type/model-id output; subclasses will
     * typically call this superclass implementation with extra
     * attributes describing their particular properties.
     */
    virtual QString toXmlString(QString indent = "",
				QString extraAttributes = "") const;

    /**
     * Set the particular properties of a layer (those specific to the
     * subclass) from a set of XML attributes.  This is the effective
     * inverse of the toXmlString method.
     */
    virtual void setProperties(const QXmlAttributes &) = 0;

signals:
    void modelChanged();
    void modelCompletionChanged();
    void modelChanged(size_t startFrame, size_t endFrame);
    void modelReplaced();

    void layerParametersChanged();
    void layerNameChanged();

protected:
    View *m_view;
};

#endif