This document does not reflect the existing codebase, but rather a road map for future development. Some of this design has been implemented in the 2008_SOC_BLEWIS branch, and these changes will be gradually merged into trunk.
The goal of the design is to build a UI which supports the following features:
- multi-track editing
- multi-layer editing*
- multiple selection
- noun-verb interaction
- direct manipulation
- edge snapping
- multiple-level undo/redo support
(*) It is important to distinguish between a track, and a layer in application terminology. Existing video editors use the term track to refer to a UI object which represents a stream of video with a sequence of sources. PiTiVi refers to this as a composition. The term track in PiTiVi means a separate channel of output: for example, audio and video are in separate tracks. The Timeline class contains one TimelineComposition for each of its output tracks. Currently these are hard-coded to audiocomp and videocomp, but in the future multiple audio and video output tracks will be supported. This will enable things like multi-language sound tracks, or multi-angle video sequences.
This is distinguished from the concept of a layer which is directly related to the notion of compositing. Within a track, sources have a property called priority which determines what will appear when the play-head reaches a given position in the timeline. By default, the source with the lowest numerical priority is displayed. Adding effects to a composition enables multiple sources to be composited together. Priority is used to determine which sources will be used by an effect as input.
The MVC/Observer Design Pattern
PiTiVi relies heavily on MVC and Observer design patterns to decouple the core of the application from the user interface. Core objects emit signals which prompt changes in the UI. UI elements wrap core objects to manipulate data, which in turn emit signals. The observer pattern allows the user interface to listen for changes in the core without coupling the core to the UI.
We use pygobject to provide support for the observer pattern in the core. The user interface depends on pygtk and pygoocanvas, both of which are based on GObject.
The advanced UI is implemented in several files in the pitivi/ui directory in the source tree:
The util.py file provides a number of convenience functions for working with goocanvas, including an easy way of creating canvas item objects, as well as generic support for drag-and-drop and selections. This file also provides the SmartGroup class, which extends goocanvas.Group with automatic recalculation of size and position.
The code in this file is intended to be as reusable and generic as possible. Its goal is to overcome some limitations of goocanvas which make programming dynamic, reactive interfaces more challenging than necessary. Code in this file is also used by the simple timeline.
SmartGroup is also used to implement HList and VList, which are container classes that enforce positioning constraints on their children. They work more-or-less like HBox and VBox in gtk. What you need to know about smart group is:
- The smart group keeps track of its own position: Setting the x or y properties on a smartgroup will cause all the group's children to move accordingly.
- The smart gropu keeps track of its size: If any of the group's children change size or position, the group recomputes its width and height properties.
This is currently accomplished by using property notification signals. SmartGroup overrides the add_child method and connects to the notify::x, notify::y, notify::width, and notify::height signals for each of its children.
These functions all take an item as input and return the appropriate property. They make expressions involving these properties more readable and more compact.
This function handles getting coordinates from an event object and converting them into the canvas space:
- event_coords(canvas, event)
These functions manipulate object size and position
- set_pos(item, pos), where pos is a tuple (x, y)
- set_size(item, size), where size is a tuple (width, height)
These functions activate draging and selection management:
- manage_selection(canvas, changed_cb)
- make_selectable(canvas, item)
- make_dragable(canvas, item, start_cb, transform, end_cb, moved_cb)
This class is yet to be implemented. Intended to provide internal support for some of the convenience functions described above:
- manage_selection(), activates internal selection managment, deprecates top-level function with same name
- make_selectable(), activates selection management on a given object, deprecates top-level function with same name.
All the interfaces used in the complex UI are kept in complexinterface.py.
This interface allows for sharing a single gtk.Adjustment among multiple client objects. When the adjustment's value changes, the the Zoomable object's zoomratio property is set, and its zoomChanged() method called.
Implementing classes must define a zoomChanged() method. This method should perform any drawing or size adjustments.
Zoomable containers have the option of defining a setChildZoomAdjustment() method, which they can use to set the zoom adjustment on all of their children. In general, however, a container should set the child's zoom adjustment whenever the child is added to the container.
This interface encapsulates handling the mouse events required to make an object dragable with the mouse. Objects can simply extending from this interface to get basic drag-and-drop functionality. If more complex drag-and-drop behavior is required, this interface provides some hook functions which can be overridden by implementing classes.
- dragStart -- a hook which is called to notify the object that dragging is about to begin
- dragEnd -- a hook which is called to notify the object that dragging has ended
- dragMotion -- a hook which is called to notify the object that its position should be updated.
This encapsulates the notion of an object which may be included in the current UI selection. This is kept strictly separate from Dragable, as there is a use case for selectable objects which cannot be moved by the user, as well as a use case for dragable objects which should never become part of the current selection. Objects are selected with Select(), and deslected with Deselect(). Objects are notified of their selection status through the selected() and deselected() method calls.
Being part of the selection implies that the object represents data that can be manipulated. To this end, all Selectable objects provide a core_object property. The list of all selected core objects can be obtained with the getSelectedCoreObjects class method.
The current selection is a set of Selectable objects, and any command which affects the current selection operates on these objects (or the core objects which they represent). To facilitate this, the interface provides class methods to iterate over all instances of selectable objects in various ways.
- (absract) selected -- notifies the object it has been selected
- (absract) deselected -- notifies the object it has been deselected
- select -- places this item in the current selection
- deselect -- removes this item from the current selection
- (abstract) delete -- removes the core object from application data structures
- (abstract) copy -- places a representation of the core object into the application clipboard
- @classmethod getSelected -- returns a list of all selected selectable objects
- @classmethod getSelectedCoreObjects -- returns the pitivi core object for every object that has been selected
- @classmethod deleteSelected -- deletes all selected objects
This is the explicit merging of the Selectable/Dragable interfaces. Objects which are both selectable and dragable should implement this interface, rather than the two ancestors independently. The reason is that this interface provides support for manipulating selections of movable items: i.e., if the user has multiple items selected and moves one of them, all the other items should move in unison.
- dragStart -- relays dragStart message to all other selected SelectableDraggable objects
- dragEnd -- relays dragEnd message to all other selected SelectableDraggable objects
- dragMove -- relays dragmove message to all other selected SelectableDraggable objects
- (abstract) setPos -- implemented by derived objects, sets the position of the core object represented by this object
- @classmethod selectedDragStart
- @classmethod selectedDragEnd
- @classmethod selectedDragMove
Encapsulates the concept of an important point on the timeline to which timestamps should be snapped during mouse operations. The class keeps track of all its instances in a sorted list, and uses binary search to implement the class method snapTime(), and snapObj, which actually implement magnetic edge snaping.
- flags -- when the control point is magnetic to the cursor, values are RESIZE, MOVE, RAZOR, COMMAND, ALL
The flags property is a bit-field defining when a magnetic point will be used.
- RESIZE -- control point is magnetic during resize operations
- MOVE -- control point is magnetic during drag operations
- RAZOR -- control point is magnetic to razor tool, or during trimming
- COMMAND -- control point can be the input to a command which operates on current selection
- ALL -- equal to RESIZE | MOVE | RAZOR | COMMAND
- setTime -- update this magnet's time value
- @classmethod snapTime(time, flags) -- snap the input time to the nearest magnet according to flags
- @classmethod snapObj(start, duration, flags) -- snap start or end time to the nearest magnet, according to flags)
These classes implement the majority of the pitivi's advanced (or complex) user interface, and can be found in complextimeline.py.
This class manages the command history for the user interface. It maintains a stack of actions and their inverses.
- undo_actions -- stack of (function, data, inverse_function, inverse_data) tuples
- redo_actions -- stack of (function, data, inverse_function, inverse_data) tuples
- undoLast -- pop the top of the undo stack, push onto the redo stack and execute the inverse operation
- redoLast -- pop the redo stack, push the undo stack, and execute the non-inverse operation
- pushAction -- add a new tuple to the top of the history stack.
- peek -- return the top of the undo stack
- poke -- update the top of the undo stack in place
- pop -- pop from the undo stack without performing any action or pushing the redo stack (for example, to clear a canceled operation from the undo stack)
- clear -- clears the undo/redo stack
This widget contains the timeline canvas and the ruler. It is also responsible for showing and hiding toolbar actions associated with the complex timeline.
(currently called ComplexLayers, will be renamed before the next release)
This class is the timeline. The canvas creates one ComplexTrack item for ComplexTrack item for each top-level composition within a timeline. PiTiVi core doesn't yet support multi-track editing, but this support is planned. ComplexTracks should be able to handle creating/destroying tracks dynamically.
In addition to the timeline itself, this widget keeps track of a number of important details about the timeline: current edit points, playhead position, current tool (only razor or pointer, at present), and the current selection.
Mouse and pointer events received by this widget are routed to the selection or the current active tool. Keyboard events are handled here directly depending on the current active tool.
The primary goal of the editing canvas is to allow the user to modify the selection as they see fit, and then apply changes to the selected object. The selection consists of a set of objects implementing the Selectable interface.
The canvas keeps track of whether or not objects are selected. Objects in the timeline always pass pointer events up to their parent group. If an event reaches the root item group, a test is performed to determine if the object should be added to the current selection. If this test passes, the objects select() method is called.
The selection also identifies a primary object: this is the object with which the user is directly interacting with, i.e. the source of the pointer event. Certain operations make the most sense in the context of a single active object. For example, if the user selects several sources and then drags one of them, this object will be used as a reference point for edge snapping.
Finally, the user is provided with a few tool-bar commands which manipulate the selection explicitly.
Selection Management Methods
Selection Manipulation Methods
Commands which operate on the selection are sent to this widget, which iterates over the selection and performs operations on every element contained therein.
This class is a container for pitivi tracks.
- Time is represented by horizontal position, in proportion to the current zoom ratio
- Priority is represented by vertical position, with the top of the canvas representing the highest priority.
This class encapsulates an internal view of a TimelineComposition object. Each ComplexTrack manages exactly one TimelineComposition, and connects to the following signals:
These signals are all sent to the same pair of signal handlers, _objectAdded, _objectRemoved, respectively. This function takes an additional parameter, klass, which is a reference to the sublclass of ComplexTimelineObject which should be instantiated.
Handle(Rect, Dragable, Magnetic)
This object is used by ComplexTimelineObject to represent the in/out edit points of the object. A handle is a goocanvas.Rect item which implements the Dragable, and Magnetic interfaces. It is not directly selectable. A handle object does not directly set its position, but instead hands off mouse events to a callback function, motion_callback.
- __init__ -- sets up initial properties, and stores the motion_callback
- dragBegin -- sets item's color to the active color
- dragEnd -- sets item to the the normal color, updates magnet timestamp
- dragMotion -- calls the motion_callback, after performing some transformations
ComplexTimelineObject(Group, Zoomable, SelectableDragable)
Corresponds to pitivi.timeline.objects.TimelineObject. It is a base
class for all objects represented in the ComplexTimeline. When created,
it is given a reference to a TimelineObject, and connects to that
start-duration-changed signal. When the core object's start
and duration change, the UI object's horizontal position and width are
updated. When the core object's layer position changes, the vertical
position is updated.
Every TimelineObject has a reference to a Content object which is displayed inside of the TimelineObject. This object may be audio or video. The Content object can change height or visibility depending on its state. The parent ComplexTimelineObject must keep track of the height of its content region and adjust its height accordingly.
ComplexTimelineObjects have drag handles which allow them to be directly resized. See the Handle class documentation for more information.
- background -- background rectangle
- coreobject -- the core PiTiVi object which this timelineobject represents
- content -- Content object
- inpoint -- Handle, representing the in point of the source
- outpoint -- Handle, representing the out point of the source
- dragMotion -- calls setStartPoint, Magnetic.snapTime(), and selectable.dragMotion() to adjust the object's position.
- (private) startDurationChanged -- handler for the coreobject's start-duration-changed signal
- setStartPoint -- sets coreobject's start property
- setInPoint -- callback given to inpoint as its motion callback, which sets coreobject's start/duration properties
- setOutPoint -- callback given to outpoint as its motion callback which sets coreobject's duration property
This class derives from TimelineObject. It overrides the signal handlers which set the in/out edit points so that they also set the media-start/media-duration points.
Abstract base class for the content region of ComplexTimelineObject. The content region displays a representation of the core object associated with the Content object's parent ComplexTimelineObject. Content regions may be expanded, contracted, or minimized. When expanded, the full preview image is visible, and the widget is expanded to maximum height so that the keyframe editor can be used. When contracted, only the preview image is visible. When minimized, the content region is completely hidden.
- make_content_image -- creates a generic image thumbnail.
Overrides make_content_image to create an audio waveform from audio stream data.
Overrides make_content_image to create a thumbnail sequence.
Marker(goocanvas.Polygon, SelectableDragable, Magnetic)
Similar to a handle, but can be the selected, which implies that it contains a reference to a core object.
This file contains the ScaleRuler class, a zoomable timeline ruler. It should share the same gtk.Adjustment objects for both zooming and horizontal scrolling.
The results of the search are