Creating a Yahoo Finance chart with D3 and d3fc

Most charting libraries are monoliths. The more features they support, the more unwieldy their APIs tend to become. With the d3fc project we have been exploring an alternative approach, constructing charts from a set of small components, using the D3 library.

In this post I want to demonstrate the power of both d3fc and D3 by re-creating the rather complex Yahoo Finance chart.

Creating a pixel-perfect recreation of this chart with any monolithic charting library would be a significant challenge (if not impossible). With d3fc it is surprisingly simple!

This post takes a step-by-step approach to demonstrate how the Yahoo Finance chart can be faithfully recreated.

Rendering a simple chart

d3fc and its dependency are available via npm as detailed in the installation instructions.

d3fc creates charts using SVG, so the first step is to add an SVG element to the page:

<svg id='time-series' style='height: 300px; width: 500px;'/>

The Yahoo Finance data is available as CSV data through an unsupported, yet widely used API. D3 has a number of utility functions for fetching and parsing data, including CSV.

The following code performs an XHR request via d3.csv:

  .row(function(d) { = new Date(d.Timestamp * 1000);
    return d;
  .get(function(error, rows) { renderChart(rows); });

Once the data has been fetched and parsed, the following renderChart function is called:

function renderChart(data) {
  var chart = fc.chart.linearTimeSeries()
        .xDomain(fc.util.extent(data, 'date'))
        .yDomain(fc.util.extent(data, ['open', 'close']));

  var area = fc.series.area()
        .yValue(function(d) { return; });


Looking at this code in detail, the first step constructs a fc.chart.linearTimeSeries. This is a relatively high-level d3fc component which renders a chart with a horizontal date axis and a vertical numeric axis. It’s main responsibility is to construct an SVG layout that houses the various parts of the chart (axes, plot area, etc …). The fc.util.extent utility function is used to compute the extents (max and min values) of various properties of the data.

Next a d3fc area series is constructed, where the yValue accessor property is used to select the open value from each datapoint. The default xValue accessor for most components expects a date property, which is the case in this example.

The chart’s plotArea is set to the area series component, ensuring that the area series has the correct scales applied to it.

Finally, the SVG element is selected using, the data is bound, and the chart component called on the selection.

If you’ve already had some experience with D3, this construction pattern should be quite familiar to you. The d3fc components follow the D3 component pattern.

The simple code above results in the following chart:

View the full code for this example via D3 bl.ocks.

Adding gridlines and line

Where other charting libraries might represent line, point and area as a single series type, d3fc prefers a ‘micro component’ approach where each are separate. For this chart an area and line series are required:

var area = fc.series.area()
      .yValue(function(d) { return; });

var line = fc.series.line()
      .yValue(function(d) { return; });

Gridlines are another d3fc component:

var gridlines = fc.annotation.gridline()

The chart plot area only accepts a single series, however multiple series instances (that have x and y scales) can be grouped together using a multi-series:

var multi = fc.series.multi()
      .series([gridlines, area, line]);


The multi-series creates a containing g element for each of the supplied series, sets their x and y scales and propagates the data to each.

With gridlines, area and line series added, and some minor tweaks to the number of ticks, the chart looks like the following:

View the full code for this example via D3 bl.ocks.


d3fc components are styled via CSS. The easiest way to determine the suitable CSS selectors for styling a component is to just look at the rendered output.

The Yahoo chart has a subtle gradient which is applied to the area series. SVG gradients are (rather clumsily) defined in SVG as follows:

<svg id='time-series' style='height: 300px; width: 500px;'>
    <linearGradient id="area-gradient"
                    x1="0%" y1="0%"
                    x2="0%" y2="100%">
      <stop offset="0%" stop-opacity="0.3" stop-color="#fff" />
      <stop offset="100%" stop-opacity="0" stop-color="#1a9af9" />

Notice that calling the linearTimeSeries component on the above SVG does not destroy the defs element. d3fc components are written in such as way that they identify their own elements via CSS class, allowing these elements to live alongside others within the same container.

With some simple CSS the gradient and line styles can be applied to the chart.

path.area {
  fill: url(#area-gradient);
  fill-opacity: 1;

path.line {
  stroke: rgb(26, 154, 249);

Unfortunately it is not possible to re-position D3 axis labels via CSS. The only way to achieve this is to render the axis then use a D3 selection to locate the labels then move them directly:

d3.selectAll('.y-axis text')
    .style('text-anchor', 'end')
    .attr('transform', 'translate(-3, -8)');

d3.selectAll('.x-axis text')
    .attr('dy', undefined)
    .style({'text-anchor': 'start', 'dominant-baseline': 'central'})
    .attr('transform', 'translate(3, -' + (xAxisHeight / 2 + 3) + ' )');

This is not ideal as the above code will be executed each time the charts is rendered, regardless of whether the axes require updates.

With this styling in place the chart looks like the following:

View the full code for this example via D3 bl.ocks.

Adding a moving average

d3fc has a number of financial indicators, these algorithms are applied directly to the chart data, with the default implementation adding new properties to the data (this can be modified by supplying a custom merge function).

The following computes an exponential moving average (EMA) based on the closed price:

var movingAverage = fc.indicator.algorithm.exponentialMovingAverage()
      .value(function(d) { return d.close; })


In order to render an indicator, a suitable renderer is required. An EMA computes a single value for each datapoint, and is rendered via a regular line series, but for more complex indicators (MACD, Bollinger) d3fc supplies dedicated renderers.

var emaLine = fc.series.line()
  .yValue(function(d) { return d.exponentialMovingAverage; })
  .decorate(function(sel) {
    sel.enter().classed('ema', true);

The code above demonstrates the ‘decorate’ pattern that can be found on most d3fc components. Decorate is passed a selection that is created as a result of the components data join. If you are not familiar with this concept, I’d recommend Mike’s Thinking With Joins article.

In the above code, the selection supplied to decorate is the update selection for the component’s root g element. The enter selection is used to add an ema class to this element. Note, that by using the enter selection, this is only done once, at the point the element is initially constructed.

NOTE: This enter selection is not quite the same as the one obtained through a standard data join, in this case the enter selection already has an element appended.

You’ll see decorate being used in quite a few places in this example, it is a powerful and versatile pattern.

With the EMA series added to the multi-series the chart looks like the following:

View the full code for this example via D3 bl.ocks.

Adding a volume chart

The Yahoo Finance chart shows the traded volume in the bottom half of the plot area, this is a pretty standard financial charting layout.

In order to render the volume chart, a secondary y-scale is required, with the domain based on the data’s volume, and the range set to half the height of the plot area. The linearTimeSeries doesn’t have a volume scale as part of its layout, this is something that has to be added manually.

The linearTimeSeries uses the d3fc layout component, which implemented flexbox layout, so it makes sense to use this to create a container for the volume series also.

The following creates a g element which acts as a container for the volume series:

var container ='#time-series');
var volumeContainer = container.selectAll('g.volume')
      'class': 'volume',
      position: 'absolute',
      top: 150,
      bottom: xAxisHeight,
      right: yAxisWidth,
      left: 0

var layout = fc.layout();

Rather than append a g element directly into the container, it is appended within a data join enter selection. This ensures that only a single element is added regardless of how many times the renderChart function is called.

A volume scale is constructed with a domain based on the input data and a range based on the height of the volume container:

var volumeScale = d3.scale.linear()
    .domain([0, d3.max(data, function (d) { return Number(d.volume); })])
    .range([volumeContainer.layout('height'), 0]);

Finally, a volume series is constructed and rendered:

var volume =
      .yValue(function(d) { return d.volume; });


This results in the following chart:

View the full code for this example via D3 bl.ocks.

Colouring the volume bars

The final step in this example is to colour each bar of the volume chart based on whether the price has risen or fallen within the time period represented by the given bar.

Once again decorate is employed:

var volume =
    .yValue(function(d) { return d.volume; })
    .decorate(function(sel) {
          .style('stroke', function(d, i) {
            return d.close > ? 'red' : 'green';

If you look at the SVG elements constructed for a bar series you will find that each bar is constructed from a g element containing a single path. The decorate function above uses the enter selection, which contains these g elements, selects the nested path and applies a suitable stroke colour based on the direction of price movement.

The chart is now starting to look quite like the Yahoo Finance chart:

View the full code for this example via D3 bl.ocks.


In this blog post you’ve seen how d3fc components can be assembled, configured and decorated to recreate a relatively complex financial chart, although it is not complete just yet.

In my next post I’ll show how d3fc legend and crosshair components can be added to provide some interactivity, and how to create a custom discontinuity provider!

Until then, if you have any questions about d3fc or this example, get in touch either via the comments field below, or via the GitHub project.

Update: The second part of this two-part series has been published, so you can see the complete example in action!

Regards, Colin E.


blog comments powered by Disqus