Use the shared ResourcePool also for ad-hoc EffectChains.

[mlt] / docs / mlt++.txt

MLT++
-----

        This mlt sub-project provides a C++ wrapping for the MLT library.

Usage
-----

        Use the following definitions in a Makefile to compile and link with mlt++:

                CXXFLAGS=`pkg-config --cflags mlt-framework` -Wall
                LDFLAGS=-lmlt++ `pkg-config --libs mlt-framework`

        Include files for the classes can either be explicitly included, ie:

                #include <mlt++/MltProducer.h>
                etc

        Or you can include all using:

                #include <mlt++/Mlt.h>

        All definitions are placed in an Mlt namespace, and adhere closely to the C
        naming convention. Mappings always follow the pattern:

        Factory methods:

                mlt_factory_init        ==> Mlt::Factory::init
                mlt_factory_producer    ==> Mlt::Factory::producer
                mlt_factory_filter      ==> Mlt::Factory::filter
                mlt_factory_transition  ==> Mlt::Factory::transition
                mlt_factory_consumer    ==> Mlt::Factory::consumer
                mlt_factory_close       ==> Mlt::Factory::close

        NB: Factory usage for service construction is optional.

        Types:

                mlt_properties          ==> Mlt::Properties
                mlt_frame               ==> Mlt::Frame
                mlt_service             ==> Mlt::Service
                mlt_producer            ==> Mlt::Producer
                mlt_filter              ==> Mlt::Filter
                mlt_transition          ==> Mlt::Transition
                mlt_consumer            ==> Mlt::Consumer

        Methods:

                mlt_type_method         ==> Mlt::Type.method
        ie:     mlt_playlist_append     ==> Mlt::Playlist.append

        Parent methods are available directly on children.

        Additionally, you can specify:

                using namespace Mlt;

        To avoid the enforced use of the Mlt:: prefix.

        Enumerators and macros are reused directly from the C library.

Class Hierarchy
---------------

        The currently mapped objects are shown in the following hierarchy:

                Factory
                Properties
                        Frame
                        Service
                                Consumer
                                Field
                                Filter
                                Multitrack
                                Producer
                                        Playlist
                                Tractor
                                Transition

Special Cases
-------------

        Care should be taken with wrapper objects. 

        Taking, as an example, the C function that returns the immediate consumer of
        a service:

                mlt_service mlt_service_consumer( mlt_service );

        This maps to:

                Mlt::Service *Mlt::Service.consumer( );

        Note that you get an object back - it is never the original c++ object, but 
        a wrapping object. This is done to keep consistency with the C api which may 
        instantiate C instances - therefore it cannot be assumed that a C++ object 
        exists for all mlt service instances.

        As such, it is mandatory that you delete these objects. The original will 
        not be affected. However, all other modifications (to properties or its
        state of connection) will be reflected in the original object.

        This approach excludes the use of RTTI to determine the real type of the 
        object - this can only be done by parsing the objects properties.

        Objects may be invalid - always use the is_valid method to check validity 
        before use.

Limitations
-----------

        The mechanisms for the definition of new services are deliberately 
        excluded from the C++ wrappings - this is done to ensure that service 
        networks constructed can be serialised and used by existing applications
        which are based on the C API (such as melted).

Hello World
-----------

        The mlt++ wrapper is a c++ wrapper for the mlt C library. As such, it 
        provides clean C++ access to the underlying library.

        An example of use is as follows:

            #include <mlt++/Mlt.h>
            using namespace Mlt;

            int main( void )
            {
                Factory::init( );
                Producer p( "pango:", "Hello World" );
                Consumer c( "sdl" );
                c.connect( p );
                c.run( );
                return 0;
            }

        This is a fairly typical example of mlt++ usage - create a 'producer' (an
        object which produces 'frames'), create a 'consumer' (an object which consumes
        frames), connect them together, start the consumer and wait until done (here
        we just wait for the user to close the window).

        In this case, we construct a window as a consumer using the 'sdl' consumer
        (SDL is a standard portable library which provides platform independent
        access to accelerated video display and audio) and use the 'pango' 
        producer to generate frames with the words 'Hello World' (pango is a 
        library from the gtk toolkit).

        The main point of this example is to show that mlt uses existing libraries
        to provide its functionality - this keeps the framework itself very small.

        Note that mlt is designed to be housed in GUI or server type applications -
        typically, applications don't wait around for the consumer to be stopped in
        the manner shown.

        So far, we've introduced the Producer and Consumer mlt classes. We'll cover
        each of these in more detail later in the tutorial, but for now, we'll 
        briefly cover the remaining classes.


Playlists
---------

        Another simple class is the Playlist - this is direct extension of Producer
        and it allows you to maintain a list of producer objects.

        As a simple example of the Playlist in action, we'll convert the example
        above into an application which plays multiple video or audio files.

            #include <mlt++/Mlt.h>
            using namespace Mlt;

            int main( int argc, char **argv )
            {
                Factory::init( );
                Playlist list;
                for ( int i = 1; i < argc; i ++ )
                {
                    Producer p( argv[i] );
                    if ( p.is_valid( ) )
                        list.append( p );
                }
                Consumer c( "sdl" );
                c.connect( list );
                c.run( );
                return 0;
            }

        Now you can run the program as:

            ./player *.avi *.mp3 *.jpg etc

        In this case, we construct a playlist by simply appending producers to it.
        Notice that although the scope of the Producer is limited to the inner 
        for loop, we can safely add it to the playlist - this is due to the fact
        that all mlt objects maintain reference counts and no object is really
        destroyed until all the references are gone. In this case, when the list
        object goes out of scope, all the producers we created will automatically
        be destroyed.


Filters
-------

        So far, we've shown how you can load and play media. We've given a brief
        intro to the Playlist container, now it's time to start manipulating 
        things...

        For the next example, I'll add a 'watermark' to the video - a watermark
        is used by broadcasters to brand the channel and normally consists of a 
        logo of some sort. We'll just use some black text on a partially 
        transparent red background.

        #include <mlt++/Mlt.h>
        using namespace Mlt;

        int main( int argc, char **argv )
        {
            Factory::init( );
            Playlist list;
            for ( int i = 1; i < argc; i ++ )
            {
                Producer p( argv[i] );
                if ( p.is_valid( ) )
                    list.append( p );
            }
            Filter f( "watermark", "pango:" );
            f.set( "producer.text", "MLT++" );
            f.set( "producer.fgcolour", "0x000000ff" );
            f.set( "producer.bgcolour", "0xff000080" );
            list.attach( f );
            Consumer c( "sdl" );
            c.connect( list );
            c.run( );
            return 0;
        }

        Notice that the watermark filter reuses the 'pango' producer we showed in the
        first example. In fact, you could use any producer here - if you wanted to
        use a graphic or a video, you would just construct the filter with a full path
        to that as the second argument.

        We manipulate the filter using the set method - this method was also shown
        in the first example. 

        Finally, we attach the filter to the playlist. This ensure that all frames 
        that are obtained from the playlist are watermarked. 


Cuts
----

        When you add a clip to a playlist, the a cut object is created - this is merely a
        wrapper for the producer, spanning the specified in and out points. 

        Whenever you retrieve a clip from a playlist, you will always get a cut object.
        This allows you to attach filters to a specific part of a producer and should
        the position of the cut in the playlist change, then the filter will remain
        correctly associated to it.

        A producer and a cut are generally identical in behaviour, but should you need to
        distinguish between them, you can use:

            if ( producer.is_cut( ) )

        and to retrieve the parent of a cut, you can use:

            Producer parent = producer.parent_cut( );

        Filters that are attached directly to a parent are executed before any filters
        attached to the cut.


Tractor
-------

        A tractor is an object that allows the manipulation of multiple video and audio
        tracks. 

        Stepping away from the player example we've been tinkering with for a minute,
        let's assume we want to do something like dub a video with some audio. This
        a very trivial thing to do:

        Tractor *dub( char *video_file, char *audio_file )
        {
            Tractor *tractor = new Tractor( );
            Producer video( video_file );
            Producer audio( audio_file );
            tractor->set_track( video, 0 );
            tractor->set_track( audio, 1 );
            return tractor;
        }

        That's all that needs to be done - you can now connect the returned object to a
        consumer, or add it to a playlist, or even apply it as a track to another tractor.


Transition
----------

        Let's now assume we want to mix the audio between two tracks - to do this, we 
        need to introduce the concept of a transition. A transition in mlt is a service
        which combines frames from two producers to produce a new frame.

        Tractor *mix( char *video_file, char *audio_file )
        {
            Tractor *tractor = new Tractor( );
            Transition mix( "mix" );
            Producer video( video_file );
            Producer audio( audio_file );
            tractor.set_track( video, 0 );
            tractor.set_track( audio, 1 );
            tractor.field.plant_transition( mix, 0, 1 );
            return tractor;
        }

        The tractor returned will now mix the audio from the original video and the 
        audio.


Mix
---

        There is a convenience function which simplifies the process of applying 
        transitions betwee adjacent cuts on a playlist. This is often preferable 
        to use over the constuction of your own tractor and transition set up.

        To apply a 25 frame luma transition between the first and second cut on 
        the playlist, you could use:

            Transition luma;
            playlist.mix( 0, 25, luma );


Events
------

        Typically, applications need to be informed when changes occur in an mlt++ object.
        This facilitates application services such as undo/redo management, or project
        rendering in a timeline type widget and many other types of operations which an
        application needs.

        As an example, consider the following:

            class Xml
            {
                private:
                    Consumer consumer;
                    Tractor &tractor;
                public:
                    Xml( MltTractor &tractor ) :
                        tractor( tractor ),
                        consumer( "xml" )
                    {
                        consumer.connect( tractor );
                        tractor.listen( tractor, "producer-changed", 
                                        ( mlt_listener )Xml::listener );
                    }
                    
                    static void listener( Properties *tractor, Xml *object )
                    {
                        object->activate( );
                    }
        
                    void activate( )
                    {
                        consumer.start( );
                    }
            };

        Now, each time the tractor is changed, the XML representation is output to 
        stderr.


That's All Folks...
-------------------

        And that, believe it or not, is a fairly complete summary of the classes you'll 
        typically be interfacing with in mlt++. Obviously, there's a little more to it 
        than this - a couple of intrisinc classes have been glossed over (notably, the 
        Properties and Service base classes). The next section will cover all of the 
        above, but in much more detail...


DIGGING DEEPER
--------------

        The previous section was designed to give you a whistle stop tour through the major
        framework classes. This section will take you through the scenic route.


Introducing Base Classes
------------------------

        Services in mlt are the collective noun for Producers, Filters, Transitions and 
        Consumer. A Service is also the base class from which all of these classes 
        extend. It provides the basic connectivity which has been shown throughout the
        examples in the previous section.

        Properties are the main way in which we communicate with the Services - 
        essentially, it provides get/set methods for named values. All services extend
        Properties.


Properties
----------

        Properties provide the general mechanism for communicating with Services - 
        through the Properties interface, we are able to manipulate and serialise 
        a services state.

        For example, to dump all the properties to stdout, you can use something 
        like:

        void dump( Properties &properties )
        {
            for ( int i = 0; i < properties.count( ); i ++ )
                cout << Properties.get_name( i ) << " = " << Properties.get( i ) << endl;
        }

        Note that the properties object handles type conversion, so the following
        is acceptable:

        properties.set( "hello", "10.5" );
        int hello_int = properties.get_int( "hello" );
        double hello_double = properties.get_double( "hello" );

        A couple of convenience methods are provide to examine or serialise property
        objects. 

        For example:

        properties.debug( );

        will report all serialisable properties on stderr, in the form:

        Object: [ ref=1, in=0, out=0, track=0, u=75, v=150, _unique_id=15, 
        mlt_type=filter, mlt_service=sepia ]


Services
--------

        Typically, all the services are constructed via the specific classes 
        constructor. Often, you will receive Service objects rather than their
        specific type. In order to access the extended classes interface, 
        you will need to create a reference.

        For example, given an arbitrary Service object, you can determine its
        type by using the type method - this will return a 'service_type' which
        has values of producer_type, filter_type etc. Alternatively, you can
        create a wrapping object and check on its validity.

        bool do_we_have_a_producer( Service &service )
        {
            Producer producer( service );
            return producer.is_valid( );
        }


Events
------


Servers and MLT XML Docs
------------------------

        For various reasons, you might want to serialise a producer to a string.
        To do this, you just need to specify a property to write to:

            Consumer xml( "xml", "buffer" );
            xml.connect( producer );
            xml.start( );
            buffer = xml.get( "buffer" );

        You can use any name you want, and you can change it using the "resource"
        property. Any name with a '.' in it is considered to be a file. Hence, you
        can use a xml consumer to store multiple instances of the same MLT 
        object - useful if you want to provide undo/redo capabilities in an
        editing application.

        Should you receive an XML document as a string, and you want to send it
        on to a server, you can use:

            Consumer client( "mvsp", "localhost:5250" );
            client.set( "xml", buffer );
            client.start( );

        If you need to obtain an MLT object from a XML string:

            Producer producer( "xml-string", buffer );
    
        The following shows a working example of an extended server:

            class ShotcutServer : public Melted
            {
                public:
                    ShotcutServer( char *id, int port ) :
                        Melted( id, port )
                    {
                    }
    
                    void set_receive_doc( bool doc )
                    {
                        set( "push-parser-off", doc );
                    }

                    // Reject all commands other than push/receive
                    Response *execute( char *command )
                    {
                        mvsp_response response = mvsp_response_init( );
                        mvsp_response_set_error( response, 400, "Not OK" );
                        return new Response( response );
                    }
        
                    // Push document handler
                    Response *received( char *command, char *doc )
                    {
                        mvsp_response response = mvsp_response_init( );
                        // Use doc in some way and assign Response
                        if ( doc != NULL )
                            mvsp_response_set_error( response, 200, "OK" );
                        return new Response( response );
                    }

                    // Push service handler
                    Response *push( char *command, Service *service )
                    {
                        mvsp_response response = mvsp_response_init( );
                        // Use service in some way and assign Response
                        if ( service != NULL )
                            mvsp_response_set_error( response, 200, "OK" );
                        return new Response( response );
                    }
            };

        NB: Should you be incorporating this into a GUI application, remember that the
        execute, received and push methods are invoked from a thread - make sure that 
        you honour the locking requirements of your GUI toolkit before interacting with
        the UI.