pixelworks
Utilities for working with ImageData
in Workers
.
Install
The pixelworks
package is meant to be used in a browser with a CommonJS module loader (e.g. Browserify or Webpack). Add it as a dependency to your project with npm
:
npm install pixelworks
Use
var pixelworks = require('pixelworks');
API
new Processor(options)
A processor runs pixel or image operations in workers.
var processor = new pixelworks.Processor(options);
Supported options
-
imageOps : boolean
- By default, operations will be called for each pixel. By settingimageOps: true
, operations will be called with anImageData
object. -
operation : Function
- A function that processes input data and returns output data. The operation will be called with two arguments: an array of inputs, and a user storage object. By default, operations will be called for each pixel in the input data, and the first argument is an array of input pixels (each a[R, G, B, A]
array). IfimageOps
istrue
, the first argument will be an array ofImageData
objects. The second object is the user storage object passed to theprocess
method.
Operations return processed output data. For pixel-wise operations, this must be an output pixel (a [R, G, B, A]
array). For image operations, this must be an ImageData
object.
Because operations run in workers, they must only operate on the arguments they are given.
-
lib : Object
- An optional lookup of functions that can be accessed by an operation run in a worker. Because operations are run in workers, they cannot access functions from the scope where they are authored. Thelib
object can be used to pass additional library functions that are made available in the worker scope. For example, if{lib: {someFunc: function() {/* do something */}}}
were provided, the operation could callsomeFunc()
. -
threads : number
- Pixel-wise operations can be run in parallel in multiple worker threads. By default, a single worker thread is created for running operations. Settingthreads: 2
would process half of the input pixels in one thread and half in another. For image type operations,threads
cannot be greater than1
. If you want to force operations to run in the main (UI) thread, setthreads: 0
. -
queue : number
- Maximum queue length. This limits the number of pending workers whenprocess
is called multiple times before work completes. If you want to callprocess
many times (in response to user generated events for example), setqueue: 1
, and only one worker will be pending at a time.
processor.process(inputs, meta, callback)
Run the operation on an array of input image data.
inputs : Array.<ImageData>
- Array of pixels or image data (depending on the operation type).meta : Object
- A user data object. This is passed to all operations and must be serializable.callback : function(Error, ImageData, Object)
- Called when work completes. The first argument is any error. The second is theImageData
generated by the operation. The third is the user data object. Whenprocess
is called repeatedly, a queue of pending workers will be generated. If this queue exceeds the maximumqueue
length, workers will be removed from the queue and the callback will be called withnull
for the secondImageData
argument.
processor.destroy()
Stop responding to any completed work and destroy the processor.