The Media Kit Table of Contents | The Media Kit Index |
Derived from: virtual BMediaNode
Declared in: be/media/BufferProducer.h
Library: libmedia.so
Allocation: Constructor only
A BBufferProducer is a BMediaNode that emits buffers containing media data that other nodes (BBufferConsumers in particular) will receive and, potentially, process. If your node wants to emit buffers, it must be derived from BBufferProducer and override the hook functions to implement the BBufferProducer protocol.
Currently, the only video clipping format supported by the Media Kit is B_CLIP_SHORT_RUNS, although there is a function in this class for converting between this format and BRegions.
This format begins with a header, consisting of two int16 values:
offsetX | X offset for all following coordinates. |
offsetY | Y offset for all following coordinates. |
These values indicate the offset for the X and Y coordinates indicated throughout the rest of the clipping data.
The remainder of the clipping data consists of entries indicating each line of video data, as follows:
numShorts | The number of values in the coordList. Always an even number. If negative, repeats the previous entry |numShorts| times. |
coordList... | List of coordinates. Even entries are left-edge X coordinates, odd entries are right-edge X coordinates. |
The clipping data contains one of these entries for each time the clipping information changes.
For example, if the clipping is a rectangle with the left edge at 100, top edge at 50, right edge at 300, and bottom edge at 200, the clipping data for a 640x480 display might be:
header
|
entry 1
|
entry 2
|
entry 3
|
entry 4
|
The header indicates that the clipping data begins at row 50.
The first entry indicates that clipping should span from column 100 to column 300 on the first row of clipping (row 50). The second entry says to repeat this 150 times.
Entry 3 indicates that clipping from that point on should be from column 0 to column 639 (the entire width of the display). Entry 4 causes this to repeat 280 times, to the bottom of the display.
In order to support media formats that don't provide timing information in their outer encapsulation layer, or to provide enhanced seeking performance for media formats that support key frames, the Media Kit supports the concept of seek tags. Producers that know their data doesn't have timing information, or that can provide enhanced seeking using special tags, should put a tag in the user_data field of the buffer headers it sends. This tag can contain any data the producer wants.
Consumers that can derive good timing information from these packets after decoding them should then choose appropriate seek points (usually key frames) and cache the performance time and tag values of the first buffer that arrives at that seek point.
Producers that can't seek without help from the decoder can then query the consumer by calling FindSeekTag(). This causes the consumer's SeekTagRequested() function to be called. This returns the seek tag and time that are closest to the requested time. The producer can then use this information locate the appropriate point in the media data.
The easiest way to use this is to use the file offset as the tag data, but any value that makes sense to the producer can be used, since the consumer just saves a copy of the data and passes it back without looking at it.
Time Seek Tag |
0.0 seconds 0 |
0.1 seconds <none> |
0.2 seconds 2 |
0.3 seconds <none> |
In this simple example, we have four buffers, two of which have seek tags recorded (at 0.0 seconds and 0.3 seconds). If the producer is seeking to 0.2 seconds, it would call FindSeekTag(), like this:
media_seek_tag tag; bigtime_t time; FindSeekTag(&destination, 0.2*1000000, &tag, &time); /* now we can use the tag to seek */
If the tag contains a file offset, we can simply seek to that offset in the file and we're ready to go.
In this example, the returned tag is "2" and the time is 0.2 seconds, because there's a seek tag located precisely at the requested time. However, if we look for a seek tag for 0.1 seconds, we get a returned tag of "0" and a time of 0.0 seconds, because that's the closest matching tag to the requested time.
protected:
|
Constructs the BBufferProducer object. The producerType specifies the type of media data that will be output by the node.
If the node will produce more than one type of data, your BBufferProducer subclass should set the kind to the default (which is a wildcard value).
If your node has additional latency on startup, you should call SetInitialLatency() to record this information. This might be the case if the buffers your node produces are created from an input signal which refreshes infrequently, such as a television signal.
|
|
When a consumer calls BBufferConsumer::RequestAdditionalBuffer(), this function is called as a result. Its job is to call SendBuffer() to immediately send the next buffer to the consumer.
The previousBufferID, previousTime, and previousTag arguments identify the last buffer the consumer received. Your node should respond by sending the next buffer after the one described.
|
Return B_OK if all is well; otherwise return an appropriate error code.
|
Informs the destination that the data flowing between source and destination is immediately changing to the format specified by newFormat.
|
RETURN CODES
B_OK. The format change request has been sent without error.
|
Given byteCount bytes of clipping data clipDatain the specified format, makes the specified region match the clipping region.
The region you specify must already exist.
The only format currently supported is B_CLIP_SHORT_RUNS.
RETURN CODES
B_OK. The clip data was converted without error.
See also: BBufferConsumer::RegionToClipData()
|
Implement this hook function to establish a connection between the source and the destination. The format negotiation is already complete by the time Connect() is called, so you have to accept the specified format.
The status argument indicates whether or not the connection actually took place; this is the result code returned by the BBufferConsumer::Connected() function or an error code indicating an error that has occurred during other preparation for the connection.
If status isn't B_OK, you should release the media_source that was reserved for this connection by PrepareToConnect(); this lets it be used by other connection attempts.
On entry, ioName contains the connection name specified by the consumer (this may be different from the name specified by the BBufferProducer::PrepareToConnect() function). On return, ioName should point to a name for the connection; if the name really matters to you, copy the name you want the connection to have back into ioName; otherwise, you can leave it alone.
|
Your implementation of Disconnect() should terminate the connection between the specified source and destination. Once you return from this function, you shouldn't send any further buffers on the connection.
If a BBufferGroup has been specified for your producer (via the SetBufferGroup() function), you should delete it here.
|
Once a client has finished iterating through your outputs via GetNextOutput() calls, it will call this function with the last value you returned as a cookie. This gives you the opportunity to dispose of any memory you may have allocated for the iteration process.
Return B_OK if the cookie is successfully disposed of (or if nothing needs to be done); otherwise, return an appropriate error code.
|
This hook function is called when a consumer's SetOutputEnabled() function is called. This indicates whether or not the output specified by whichOutput needs to be sent buffers. You must implement this function so that you don't send buffers to outputs that don't need them. The _depreciated_ argument is no longer used.
By default, output is enabled.
|
GetLatencyFor() returns the latency introduced by sending data to the destination forDestination. On return, the latency will be stored in outLatency, and the time source used by forDestination will be available in outTimeSource (unless an error is returned, in which case these values are undetermined).
The latency of sending a buffer from one time source to another should always be assumed to be zero, since there may be no relationship between the progress of time of two different time sources.
RETURN CODES
B_OK. The latency was returned without error.
|
In order to improve seek performance, the Media Kit provides the concept of seek tags. These are special tags that identify easily-located points in media data (such as key frames in Cinepak video). The FindSeekTag() function asks the consumer specified by forDestination for the nearest seek tag to the time specified by inTargetTime, and returns the tag in outTag and the time corresponding to that tag in outTaggedTime. On return, outFlags (if the pointer isn't NULL) contains flags giving further details about the tag.
There are currently no defined values for inFlags or outFlags.
RETURN CODES
B_OK. No error.
|
Implement FormatChangeRequested() to change the format of the media data flowing from the given source to the specified destination to the format specified by ioFormat. If there are wildcards specified in ioFormat, fill them in to match the format you prefer before returning from this call. You should ignore the _depreciated_ argument; it's no longer used.
|
Return B_OK if the change request is processed successfully; otherwise, return an appropriate error code.
See also: FormatSuggestionRequested(), FormatProposal()
|
Your BBufferProducer should implement this function to verify that the proposed media_format is suitable for the specified output. If any fields in the format are wildcards, and you have a specific requirement, adjust those fields to match your requirements before returning.
Return B_OK if the proposed format is acceptable; once you've done so, the Media Kit will assume that any connection request made on output with the specified format (after any changes you may have made) will succeed.
If output isn't available, return B_BAD_SOURCE.
If format isn't reasonable, return B_BAD_MEDIA_FORMAT.
|
You must implement FormatSuggestionRequested() to return fill the buffer pointed to by format that your producer is capable of emitting that meets the desired type and quality requirements.
If your producer can work with a range of possible formats, let the quality argument guide your selection. For example, you might choose to use 10 fps for previews, and 60 fps interlaced 640x480 for full-quality video.
If type is a media class that your producer doesn't want to work with, return B_BAD_MEDIA_FORMAT. If you're preapared to accept a wide range of values for some specific field, set that field to the wildcard value (see media_audio_format::wildcard() and media_video_format::wildcard() for more information.
Return B_OK if the format is successfully returned.
|
Implement this hook function to store, in outLatency, the total amount of latency your BBufferProducer incurs from receiving a buffer of data until it reaches its ultimate destination.
Call FindLatencyFor() on whatever outputs the data is being forwarded to, add your own latency to the largest of those values, and return that value.
The default implementation of GetLatency() finds the maximum latency of your currently-available outputs by iterating over them, and returns that value in outLatency; therefore, your implementation of this function may simply need to call the inherited version of this function, then add your own processing latency to the returned value.
RETURN CODES
B_OK. The latency has been returned successfully.
|
Implement this function to return information about your available outputs. The first time it's called for a new iteration loop, the value pointed to by cookie will be 0. Each time GetNextOutput() is called, you should set it to some value that makes sense to you so you can keep track of where in the iteration process the client is, but never set it to 0.
For each call to GetNextOutput(), including the first, you should return one of your outputs that the client hasn't seen during the iteration loop in outOutput.
Once all outputs have been reported, you should return B_ERROR.
|
When your node derived from BBufferProducer receives a message on its control port, you should handle it yourself if you know how, or dispatch to each ancestor class in turn (starting with BBufferProducer::HandleMessage()) until one of the HandleMessage() implementations returns B_OK. If none of the inherited implementations of this function returns B_OK, you should pass the message to BMediaNode::HandleBadMessage() to be dealt with.
Your port-listening thread should call HandleMessage() to dispatch the received data.
See also: BMediaNode::HandleMessage(), "About Multiple Virtual Inheritance" on page21
|
This hook function is called when a BBufferConsumer that's receiving data from you determines that its latency has changed. It will call its BBufferConsumer::SendLatencyChange() function, and in response, the Media Server will call your LatencyChanged() function.
The source argument indicates your output that's involved in the connection, and destination specifies the input on the consumer to which the connection is linked. newLatency is the consumer's new latency. The flags are currently unused.
Override this function to implement whatever functionality you need to adjust your own latency calculations to keep the data flowing smoothly.
|
This hook function is called when a BBufferConsumer that's receiving data from you determines that data is arriving late (when the BBufferConsumer::NotifyLateProducer() function is called); the exact degree to which your buffers are late is specified by the howLate argument. Your implementation of this function should take whatever steps are necessary to correct the problem, either by asking nodes upstream from you to deliver buffers earlier, dropping buffers, or other appropriate actions, depending on the current run mode.
The performanceTime argument specifies the performance time at which the notification was sent.
See also: BMediaNode::RunMode()
|
The PrepareToConnect() hook is called before a new connection between the source whichSource and the destination whichDestination is established, in order to give your producer one last chance to specialize any wildcards that remain in the format (although by this point there shouldn't be any, you should check anyway).
Your implementation should, additionally, return in outSource the source to be used for the connection, and should fill the outName buffer with the name the connection will be given; the consumer will see this in the outInput->name argument specified to BBufferConsumer::Connected(). If your node doesn't care what the name is, you can leave the outName untouched.
|
Return B_OK if the connection process should proceed, or an appropriate error code if something's wrong.
If you return B_OK, the consumer's Connected() function will be called, to let it know that a new connection is being established. Finally, the producer's Connect() function is called to complete the exchange.
See also: BBufferConsumer::Connected(), Connect()
|
Returns the media_type of the media data produced by the node.
|
Call this function to determine whether or not the destination forDestination is prepared to accept buffers in the specified format. This function can be especially useful if you want to test various formats to select the best compatible format during a hookup request in which the requested format contains wildcards.
RETURN CODES
B_OK. The proposed format is acceptable to the destination.
|
Call this function to send a buffer of media data to the specified destination, which must already be connected to one of your outputs. This is how your BBufferProducer object will send data downstream to BBufferConsumers to which it's connected.
It's your responsibility to ensure that the buffer's header and the data contained in the buffer itself are valid, although SendBuffer() will automatically fill out the following header fields for you:
In particular, be sure that if you're outputting video buffers you set the media_video_buffer to describe the video properly. If you don't, things will go badly for you.
You can obtain a buffer to fill and send by calling BBufferConsumer::RequestBuffer() on a BBufferConsumer that you own (and that's okay to use for buffers going to the specified destination).
RETURN CODES
B_OK. The buffer was sent without error.
|
Call this function to inform the specified destination whether or not there's data available from your producer node. Specify the appropriate status flag as the status argument, and the time at which the status takes effect as the atTime argument.
Possible values for the status argument are:
Constant | Description |
---|---|
B_DATA_NOT_AVAILABLE | There aren't any buffers ready for the destination. |
B_DATA_AVAILABLE | There are buffers ready for the destination. |
B_PRODUCER_STOPPED | The producer has stopped. |
RETURN CODES
B_OK. The status update was sent without error.
|
When a client wants a specific BBufferGroup to be used for a given output forSource, it will call this function. You should remember the group and use it for all requests for buffers to send on the output forSource (and for no other outputs, unless the client explicitly requests you do so by calling SetBufferGroup() for another output source).
If your BBufferProducer goes away, or the connection is broken, delete the BBufferGroup object.
If group is NULL, you should use whatever BBufferGroup you wish after disposing of the previous group.
It's okay to pass group on to another node upstream from your BBufferProducer if your BBufferProducer only passes along buffers it receives in its processing loop; in that case, you're not really the owner of the BBufferGroup, unless you pass true for willReclaim in the call to BMediaRoster::SetOutputBuffers().
Return B_OK if the buffer group is set without incident; otherwise, return an appropriate error code.
|
If your node has additional startup latency imposed by the signal from which its buffers are constructed, you should call SetInitialLatency() to specify the maximum possible latency that can be added by this delay. initialLatency should be the maximum latency, in microseconds, that might occur.
One situation in which this occurs is for TV capture card nodes. An NTSC television signal broadcasts a new field about every sixtieth of a second, which means that if your node is started partway through one field being received, you might have to wait as long as a sixtieth of a second for the first complete frame to arrive. So the maximum latency in this situation is a sixtieth of a second.
Setting the initial latency correctly can prevent consumers from having problems synchronizing with your node, and can improve performance.
flags should be 0 for now; there are no values defined yet.
|
This function is called to tell the producer to resample the data rate by the specified factor. Specifying a value of 1 (ie, numerator/denominator = 1) indicates that the data should be output at the same playback rate that it comes into the node at. The format of the data should be unchanged.
For example, if you're playing a sound at 48 kHz, and you receive a call to SetPlayRate() with a numerator of 2 and a demoninator of 1 (double speed), you should resample so that you move twice as fast through the source data while keeping the output rate constant. You might do this by doing a brute-force resample to 24 kHz (which would result in twice the data rate) or do time-compression (which would retain the pitch).
As another example, if you're playing video at 30 frames per second, and your SetPlayRate() function is called with a ratio of 1:2 specified (half speed), you should continue sending 30 frames per second, but you need to arrange for the playback to look like half-speed. A reasonable way to do this would be to send each frame twice (re-time-stamped and buffered internally, if necessary), which would result in the desired half-speed appearance.
Return B_OK if the sampling rate is changed; otherwise, return an error. It's okay to return an error if you don't support varying sampling rates—the Media Kit won't hold that against you.
|
This hook function is called when a client wants your BBufferProducer to output video data clipped to a particular region. Your producer must remember this clipping region and apply it to all video data you produce, without altering any bytes outside the region in any buffers sent through the source forSource.
Before your implementation of VideoClippingChanged() returns, you should set the value pointed to by outFromChangeTag to the change tag value at which the clipping will take effect, so the client will know what buffers it can expect to have the requested clipping. This can be done easily by adding the following line to your implementation:
*outFromChangeTag = UpdateChangeTag();
You can use the ClipDataToRegion() function to convert the data in clipData into an actual BRegion if that's a better format for you to work with. If you do, keep in mind that numShorts is the actual number of int16 values in the array specified by clipData, while ClipDataToRegion() requires the number of bytes of data in the array; be sure to multiply numShorts by sizeof(int16).
The media_video_display_info structure referred to by display indicates the format of the video display onto which the video is being displayed; this lets you know what color space, screen size, and so forth is in use on the video display, so your producer can render properly. VideoClippingChanged() is called not only when clipping changes, but when the configuration of the display changes as well. Your producer must abide by this starting at the specified change count.
See "Video Clipping" on page85 for information on the format of the clip data.
See also: ClipDataToRegion(), BBufferConsumer::RegionToClipData()
Declared in: <be/media/BufferProducer.h>
Constant | Description |
---|---|
B_CLIP_SHORT_RUNS | Clipping is encoded using runs of shorts. |
Declared in: be/media/BufferProducer.h
Constant | Meaning |
---|---|
B_ANY_QUALITY | Any quality. |
B_LOW_QUALITY | A low quality level (10). |
B_MEDIUM_QUALITY | Medium quality level (50). |
B_HIGH_QUALITY | High quality level (100). |
Quality values you can use when you don't want to have to come up with one on your own.
The Media Kit Table of Contents | The Media Kit Index |
Copyright © 2000 Be, Inc. All rights reserved..