Porting webkitAudioContext code to standards based AudioContext

The Web Audio standard was first implemented in WebKit, and the implementation was built in parallel with the work on the specification of the API.  As the specification evolved and changes were made to the spec, some of the old implementation pieces were not removed from the WebKit (and Blink) implementations due to backwards compatibility reasons.  New engines implementing the Web Audio spec (such as Gecko) will only implement the official, final version of the specification, which means that code using webkitAudioContext or old naming conventions in the Web Audio specification may not immediately work out of the box in a compliant Web Audio implementation.  This article attempts to summarize the areas where developers are likely to encounter these problems and provide examples on how to port such code to standards based AudioContext, which will work across different browser engines.

Note: There is a library called webkitAudioContext monkeypatch, which automatically fixes some of these changes to make most code targetting webkitAudioContext to work on the standards based AudioContext out of the box, but it currently doesn't handle all of the cases below.  Please consult the README file for that library to see a list of APIs that are automatically handled by it.

Changes to the creator methods

Three of the creator methods on webkitAudioContext have been renamed in AudioContext.

These are simple renames that were made in order to improve the consistency of these method names on AudioContext.  If your code uses either of these names, like in the example below :

JavaScript
// Old method names
var gain = context.createGainNode();
var delay = context.createDelayNode();
var js = context.createJavascriptNode(1024);

you can simply rename the methods to look like this:

JavaScript
// New method names
var gain = context.createGain();
var delay = context.createDelay();
var js = context.createScriptProcessor(1024);

The semantics of these methods remain the same in the renamed versions.

Changes to methods used to start/stop AudioBufferSourceNode and OscillatorNode

In webkitAudioContext, there are two ways to start and stop AudioBufferSourceNode and OscillatorNode: the noteOn() and noteOff() methods, and the start() and stop() methods.  (AudioBufferSourceNode has yet another way of starting output: the noteGrainOn() method.)  The noteOn()/noteGrainOn()/noteOff() methods were the original way to start/stop output in these nodes, and in the newer versions of the specification, the noteOn() and noteGrainOn() methods were consolidated into a single start() method, and the noteOff() method was renamed to the stop() method.

In order to port your code, you can just rename the method that you're using.  For example, if you have code like the below:

JavaScript
var osc = context.createOscillator();
osc.noteOn(1);
osc.noteOff(1.5);

var src = context.createBufferSource();
src.noteGrainOn(1, 0.25);
src.noteOff(2);

you can simply change it like this in order to port it to the standard AudioContext API:

JavaScript
var osc = context.createOscillator();
osc.start(1);
osc.stop(1.5);

var src = context.createBufferSource();
src.start(1, 0.25);
src.stop(2);

Removal of the synchronous AudioContext.createBuffer method

In the old WebKit implementation of Web Audio, there were two versions of createBuffer(), one which created an initially empty buffer, and one which took an existing ArrayBuffer containing encoded audio, decoded it and returned the result in the form of an AudioBuffer.  The latter version of createBuffer() was potentially expensive, because it had to decode the audio buffer synchronously, and with the buffer being arbitrarily large, it could take a lot of time for this method to complete its work, and no other part of your web page's code could execute in the mean time.

Because of these problems, this version of the createBuffer() method has been removed, and you should use the asynchronous decodeAudioData() method instead.

The example below shows old code which downloads an audio file over the network, and then decoded it using createBuffer():

JavaScript
var xhr = new XMLHttpRequest();
xhr.open("GET", "/path/to/audio.ogg", true);
xhr.responseType = "arraybuffer";
xhr.send();
xhr.onload = function() {
  var decodedBuffer = context.createBuffer(xhr.response, false);
  if (decodedBuffer) {
    // Decoding was successful, do something useful with the audio buffer
  } else {
    alert("Decoding the audio buffer failed");
  }
};

Converting this code to use decodeAudioData() is relatively simple, as can be seen below:

JavaScript
var xhr = new XMLHttpRequest();
xhr.open("GET", "/path/to/audio.ogg", true);
xhr.responseType = "arraybuffer";
xhr.send();
xhr.onload = function() {
  context.decodeAudioData(xhr.response, function onSuccess(decodedBuffer) {
    // Decoding was successful, do something useful with the audio buffer
  }, function onFailure() {
    alert("Decoding the audio buffer failed");
  });
};

Note that the decodeAudioData() method is asynchronous, which means that it will return immediately, and then when the decoding finishes, one of the success or failure callback functions will get called depending on whether the audio decoding was successful.  This means that you may need to restructure your code to run the part which happened after the createBuffer() call in the success callback, as you can see in the example above.

Renaming of AudioParam.setTargetValueAtTime

The setTargetValueAtTime() method on the AudioParam interface has been renamed to setTargetAtTime().  This is also a simple rename to improve the understandability of the API, and the semantics of the method are the same.  If your code is using setTargetValueAtTime(), you can simply rename it to use setTargetAtTime(). For example, if we have code that looks like this:

JavaScript
var gainNode = context.createGain();
gainNode.gain.setTargetValueAtTime(0.0, 10.0, 1.0);

you can rename the method, and be compliant with the standard, like so:

JavaScript
var gainNode = context.createGain();
gainNode.gain.setTargetAtTime(0.0, 10.0, 1.0);

Changes to some of the enumerated values in the API

The original webkitAudioContext API used C-style number based enumerated values in the API.  Those values have since been changed to use the Web IDL based enumerated values, which should be familiar because they are similar to things like the HTMLInputElement.type property.

OscillatorNode.type

OscillatorNode's type property has been changed to use Web IDL enums.  Old code using webkitAudioContext can be ported to standards based AudioContext like below:

JavaScript
// Old webkitAudioContext code:
var osc = context.createOscillator();
osc.type = osc.SINE;     // sine waveform
osc.type = osc.SQUARE;   // square waveform
osc.type = osc.SAWTOOTH; // sawtooth waveform
osc.type = osc.TRIANGLE; // triangle waveform
osc.setWaveTable(table);
var isCustom = (osc.type == osc.CUSTOM); // isCustom will be true

// New standard AudioContext code:
var osc = context.createOscillator();
osc.type = "sine";       // sine waveform
osc.type = "square";     // square waveform
osc.type = "sawtooth";   // sawtooth waveform
osc.type = "triangle";   // triangle waveform
osc.setPeriodicWave(table);  // Note: setWaveTable has been renamed to setPeriodicWave!
var isCustom = (osc.type == "custom"); // isCustom will be true

BiquadFilterNode.type

BiquadFilterNode's type property has been changed to use Web IDL enums.  Old code using webkitAudioContext can be ported to standards based AudioContext like below:

JavaScript
// Old webkitAudioContext code:
var filter = context.createBiquadFilter();
filter.type = filter.LOWPASS;   // lowpass filter
filter.type = filter.HIGHPASS;  // highpass filter
filter.type = filter.BANDPASS;  // bandpass filter
filter.type = filter.LOWSHELF;  // lowshelf filter
filter.type = filter.HIGHSHELF; // highshelf filter
filter.type = filter.PEAKING;   // peaking filter
filter.type = filter.NOTCH;     // notch filter
filter.type = filter.ALLPASS;   // allpass filter

// New standard AudioContext code:
var filter = context.createBiquadFilter();
filter.type = "lowpass";        // lowpass filter
filter.type = "highpass";       // highpass filter
filter.type = "bandpass";       // bandpass filter
filter.type = "lowshelf";       // lowshelf filter
filter.type = "highshelf";      // highshelf filter
filter.type = "peaking";        // peaking filter
filter.type = "notch";          // notch filter
filter.type = "allpass";        // allpass filter

PannerNode.panningModel

PannerNode's panningModel property has been changed to use Web IDL enums.  Old code using webkitAudioContext can be ported to standards based AudioContext like below:

JavaScript
// Old webkitAudioContext code:
var panner = context.createPanner();
panner.panningModel = panner.EQUALPOWER;  // equalpower panning
panner.panningModel = panner.HRTF;        // HRTF panning

// New standard AudioContext code:
var panner = context.createPanner();
panner.panningModel = "equalpower";       // equalpower panning
panner.panningModel = "HRTF";             // HRTF panning

PannerNode.distanceModel

PannerNode's distanceModel property has been changed to use Web IDL enums.  Old code using webkitAudioContext can be ported to standards based AudioContext like below:

JavaScript
// Old webkitAudioContext code:
var panner = context.createPanner();
panner.distanceModel = panner.LINEAR_DISTANCE;      // linear distance model
panner.distanceModel = panner.INVERSE_DISTANCE;     // inverse distance model
panner.distanceModel = panner.EXPONENTIAL_DISTANCE; // exponential distance model

// Mew standard AudioContext code:
var panner = context.createPanner();
panner.distanceModel = "linear";                    // linear distance model
panner.distanceModel = "inverse";                   // inverse distance model
panner.distanceModel = "exponential";               // exponential distance model

Removal of AudioBufferSourceNode.gain

The gain attribute of AudioBufferSourceNode has been removed.  The same functionality can be achieved by connecting the AudioBufferSourceNode to a gain node.  See the following example:

JavaScript
// Old webkitAudioContext code:
var src = context.createBufferSource();
src.buffer = someBuffer;
src.gain.value = 0.5;
src.connect(context.destination);
src.noteOn(0);

// New standard AudioContext code:
var src = context.createBufferSource();
src.buffer = someBuffer;
var gain = context.createGain();
src.connect(gain);
gain.gain.value = 0.5;
gain.connect(context.destination);
src.start(0);

Removal of AudioBuffer.gain

The gain attribute of AudioBuffer has been removed.  The same functionality can be achieved by connecting the AudioBufferSourceNode that owns the buffer to a gain node.  See the following example:

JavaScript
// Old webkitAudioContext code:
var src = context.createBufferSource();
src.buffer = someBuffer;
src.buffer.gain = 0.5;
src.connect(context.destination);
src.noteOn(0);

// New standard AudioContext code:
var src = context.createBufferSource();
src.buffer = someBuffer;
var gain = context.createGain();
src.connect(gain);
gain.gain.value = 0.5;
gain.connect(context.destination);
src.start(0);

Removal of AudioBufferSourceNode.looping

The looping attribute of AudioBufferSourceNode has been removed.  This attribute was an alias of the loop attribute, so you can just use the loop attribute instead. Instead of having code like this:

JavaScript
var source = context.createBufferSource();
source.looping = true;

you can change it to respect the last version of the specification:

JavaScript
var source = context.createBufferSource();
source.loop = true;

Note, the loopStart and loopEnd attributes are not supported in webkitAudioContext.

Removal of AudioBufferSourceNode.playbackState and OscillatorNode.playbackState

The playbackState attribute of AudioBufferSourceNode and OscillatorNode has been removed.  Depending on why you used this attribute, you can use the following techniques to get the same information:

  • If you need to compare this attribute to UNSCHEDULED_STATE, you can basically remember whether you've called start() on the node or not.
  • If you need to compare this attribute to SCHEDULED_STATE, you can basically remember whether you've called start() on the node or not.  You can compare the value of AudioContext.currentTime to the first argument passed to start() to know whether playback has started or not.
  • If you need to compare this attribute to PLAYING_STATE, you can compare the value of AudioContext.currentTime to the first argument passed to start() to know whether playback has started or not.
  • If you need to know when playback of the node is finished (which is the most significant use case of playbackState), there is a new ended event which you can use to know when playback is finished.  Please see this code example:
JavaScript
// Old webkitAudioContext code:
var src = context.createBufferSource();
// Some time later...
var isFinished = (src.playbackState == src.FINISHED_STATE);

// New AudioContext code:
var src = context.createBufferSource();
function endedHandler(event) {
  isFinished = true;
}
var isFinished = false;
src.onended = endedHandler;

The exact same changes have been applied to both AudioBufferSourceNode and OscillatorNode, so you can apply the same techniques to both kinds of nodes.

Removal of AudioContext.activeSourceCount

The activeSourceCount attribute has been removed from AudioContext.  If you need to count the number of playing source nodes, you can maintain the count by handling the ended event on the source nodes, as shown above.

Code using the activeSourceCount attribute of the AudioContext, like this snippet:

JavaScript
var src0 = context.createBufferSource();
var src1 = context.createBufferSource();
// Set buffers and other parameters...
src0.start(0);
src1.start(0);
// Some time later...
console.log(context.activeSourceCount);

could be rewritten like that:

JavaScript
// Array to track the playing source nodes:
var sources = [];
// When starting the source, put it at the end of the array,
// and set a handler to make sure it gets removed when the
// AudioBufferSourceNode reaches its end.
// First argument is the AudioBufferSourceNode to start, other arguments are
// the argument to the |start()| method of the AudioBufferSourceNode.
function startSource() {
  var src = arguments[0];
  var startArgs = Array.prototype.slice.call(arguments, 1);
  src.onended = function() {
    sources.splice(sources.indexOf(src), 1);
  }
  sources.push(src);
  src.start.apply(src, startArgs);
}
function activeSources() {
  return sources.length;
}
var src0 = context.createBufferSource();
var src0 = context.createBufferSource();
// Set buffers and other parameters...
startSource(src0, 0);
startSource(src1, 0);
// Some time later, query the number of sources...
console.log(activeSources());

Renaming of WaveTable

The WaveTable interface has been renamed to PeriodicWave.  Here is how you can port old code using WaveTable to the standard AudioContext API:

JavaScript
// Old webkitAudioContext code:
var osc = context.createOscillator();
var table = context.createWaveTable(realArray, imaginaryArray);
osc.setWaveTable(table);

// New standard AudioContext code:
var osc = context.createOscillator();
var table = context.createPeriodicWave(realArray, imaginaryArray);
osc.setPeriodicWave(table);

Removal of some of the AudioParam read-only attributes

The following read-only attributes have been removed from AudioParam: name, units, minValue, and maxValue.  These used to be informational attributes.  Here is some information on how you can get these values if you need them:

  • The name attribute is a string representing the name of the AudioParam object.  For example, the name of GainNode.gain is "gain".  You can simply track where the AudioParam object is coming from in your code if you need this information.
  • The minValue and maxValue attributes are read-only values representing the nominal range for the AudioParam.  For example, for GainNode, these values are 0 and 1, respectively.  Note that these bounds are not enforced by the engine, and are merely used for informational purposes.  As an example, it's perfectly valid to set a gain value to 2, or even -1.  In order to find out these nominal values, you can consult the specification.
  • The units attribute as implemented in webkitAudioContext implementations is unused, and always returns 0.  There is no reason why you should need this attribute.

Removal of MediaElementAudioSourceNode.mediaElement

The mediaElement attribute of MediaElementAudioSourceNode has been removed.  You can keep a reference to the media element used to create this node if you need to access it later.

Removal of MediaStreamAudioSourceNode.mediaStream

The mediaStream attribute of MediaStreamAudioSourceNode has been removed.  You can keep a reference to the media stream used to create this node if you need to access it later.

License

© 2016 Mozilla Contributors
Licensed under the Creative Commons Attribution-ShareAlike License v2.5 or later.
https://developer.mozilla.org/en-us/docs/web/api/web_audio_api/porting_webkitaudiocontext_code_to_standards_based_audiocontext

API porting Reference Référence Web Audio API webkitAudioContext