Chart.js 2.0 introduces the concept of controllers for each dataset. Like scales, new controllers can be written as needed.
Chart.controllers.MyType = Chart.DatasetController.extend({
});
// Now we can create a new instance of our chart, using the Chart.js API
new Chart(ctx, {
// this is the string the constructor was registered at, ie Chart.controllers.MyType
type: 'MyType',
data: data,
options: options
});
Dataset controllers must implement the following interface.
{
// Create elements for each piece of data in the dataset. Store elements in an array on the dataset as dataset.metaData
addElements: function() {},
// Create a single element for the data at the given index and reset its state
addElementAndReset: function(index) {},
// Draw the representation of the dataset
// @param ease : if specified, this number represents how far to transition elements. See the implementation of draw() in any of the provided controllers to see how this should be used
draw: function(ease) {},
// Remove hover styling from the given element
removeHoverStyle: function(element) {},
// Add hover styling to the given element
setHoverStyle: function(element) {},
// Update the elements in response to new data
// @param reset : if true, put the elements into a reset state so they can animate to their final values
update: function(reset) {},
}
The following methods may optionally be overridden by derived dataset controllers
{
// Initializes the controller
initialize: function(chart, datasetIndex) {},
// Ensures that the dataset represented by this controller is linked to a scale. Overridden to helpers.noop in the polar area and doughnut controllers as these
// chart types using a single scale
linkScales: function() {},
// Called by the main chart controller when an update is triggered. The default implementation handles the number of data points changing and creating elements appropriately.
buildOrUpdateElements: function() {}
}
Extending or replacing an existing controller type is easy. Simply replace the constructor for one of the built in types with your own.
The built in controller types are:
Chart.controllers.line
Chart.controllers.bar
Chart.controllers.radar
Chart.controllers.doughnut
Chart.controllers.polarArea
Chart.controllers.bubble
For example, to derive a new chart type that extends from a bubble chart, you would do the following.
// Sets the default config for 'derivedBubble' to be the same as the bubble defaults.
// We look for the defaults by doing Chart.defaults[chartType]
// It looks like a bug exists when the defaults don't exist
Chart.defaults.derivedBubble = Chart.defaults.bubble;
// I think the recommend using Chart.controllers.bubble.extend({ extensions here });
var custom = Chart.controllers.bubble.extend({
draw: function(ease) {
// Call super method first
Chart.controllers.bubble.prototype.draw.call(this, ease);
// Now we can do some custom drawing for this dataset. Here we'll draw a red box around the first point in each dataset
var meta = this.getMeta();
var pt0 = meta.data[0];
var radius = pt0._view.radius;
var ctx = this.chart.chart.ctx;
ctx.save();
ctx.strokeStyle = 'red';
ctx.lineWidth = 1;
ctx.strokeRect(pt0._view.x - radius, pt0._view.y - radius, 2 * radius, 2 * radius);
ctx.restore();
}
});
// Stores the controller so that the chart initialization routine can look it up with
// Chart.controllers[type]
Chart.controllers.derivedBubble = custom;
// Now we can create and use our new chart type
new Chart(ctx, {
type: 'derivedBubble',
data: data,
options: options,
});
The bar controller has a special property that you should be aware of. To correctly calculate the width of a bar, the controller must determine the number of datasets that map to bars. To do this, the bar controller attaches a property bar
to the dataset during initialization. If you are creating a replacement or updated bar controller, you should do the same. This will ensure that charts with regular bars and your new derived bars will work seamlessly.