howto

How to Adapt a D3.js Template in a Web App

May 24, 2017

D3.js is a state-of-the-art library for data visualization. Check out the D3 gallery for stunning and beautiful examples. Happily, many of these visualizations include their source code, so that you can easily duplicate them.

For example, the parallel coordinates chart, created by Mike Bostock, is given with the generating D3 code and data! This is a cool and useful data viz that allows you to quickly visualize a multi-dimensional (but relatively small) dataset: you can immediately spot correlations across dimensions and uncover clusters. In this interactive viz, you can explore the data in depth by filtering values on each dimension with a brush tool.

Let’s try to replicate the parallel coordinates chart in a web app.


Example of a parallel coordinates chart


Upload the data and create a new web app

The parallel coordinates chart is illustrated on a dataset of car specifications. In your project, create the cars dataset from this CSV file.

Create a new “standard”, empty webapp. To access the data in your web app, open the Settings tab and click on “Configure” in the Security section. In the dataset list, find the cars dataset and allow your webapp to read it. Then, import the D3.js library on the main page of the Settings tab. As a reminder, this post explains how to do this.

Understand the overall code structure

Many D3 code samples, given in the gallery or bl.ocks.org, have the same overall structure.

<!DOCTYPE html>
<meta charset="utf-8">
<style>
    /* CSS  code */
</style>
<body>
    <!-- HTML code -->
    <script src="http://d3js.org/d3.v3.min.js"></script>
    <script>
        // JS code
    </script>
<body>

To replicate the D3 visualizations in your web app, you will simply need to copy the CSS and HTML code in the corresponding panels of the web app editor. For the JavaScript code, it requires a little more work, as we will see promptly.

Copying the HTML code

In the parallel coordinates chart example, there is no HTML code written within the <body> tags, and thus the HTMl panel of your web app code editor should be blank.

Copying the CSS code

Here’s the CSS code, defined within the <style> tags, that you should copy into the CSS panel of your editor.

svg {
  font: 10px sans-serif;
}

.background path {
  fill: none;
  stroke: #ccc;
  stroke-opacity: .4;
  shape-rendering: crispEdges;
}

.foreground path {
  fill: none;
  stroke: steelblue;
  stroke-opacity: .7;
}

.brush .extent {
  fill-opacity: .3;
  stroke: #fff;
  shape-rendering: crispEdges;
}

.axis line,
.axis path {
  fill: none;
  stroke: #000;
  shape-rendering: crispEdges;
}

.axis text {
  text-shadow: 0 1px 0 #fff;
  cursor: move;
}

Adapting the JS code

The trickiest part in adapting a D3 template is always to shape the data in the format required by the data viz. In the parallel coordinates charts, the data in the D3 code is represented as the cars JSON array.

But generally your source data is not in JSON format. In many D3 templates, the data is given as a CSV file which is converted to JSON.

In the D3 original code, the data is thus read from the cars.csv file:

// D3 code
d3.csv("cars.csv", function(error, cars) {
    // D3 code
});
// D3 code

Then the D3 code defined inside the d3.csv function is applied on the cars JSON array.

In our web app in Dataiku DSS, you will have to connect to your dataset (which can be stored in a great variety of formats and database systems) through the Dataiku JavaScript API. In order to do this, we need to slightly modify the JS code. Replace the original JS code above by

// same D3 code
function parallelCoordinatesChart(cars) {
    // same D3 code
}
// same D3 code

and copy it into the JS panel of the web app editor. In other words, keep the entire D3 code unchanged, except for the call to the d3.csv function, which is replaced by defining the parallelCoordinatesChart function, which takes the cars JSON array as input.

Note: be sure to remove the parenthesis and semicolon that were at the end of the d3.csv function; they are not needed (and the parenthesis will indeed cause an error).

Now, we only need to connect to the cars dataset through the dataiku JS API, in order to create the corresponding cars JSON array. Notice that, when you gave permission for your web app to read the cars datsaset, a call to the dataiku.fetch function was automatically added. You finally need to copy the JS code defined below into the dataiku.fetch function. This code creates the cars JSON array and calls the parallelCoordinatesChart function to create the chart.

dataiku.fetch('cars', function(dataFrame) {
    var columnNames = dataFrame.getColumnNames();
    function formatData(row) {
        var out = {};
        columnNames.forEach(function (col) {
            out[col]= col==='name' ? row[col] : +row[col];
        });
        return out;
    }
  var cars = dataFrame.mapRecords(formatData);
  parallelCoordinatesChart(cars);
});

That’s it, you have a running D3 data viz in your web app!

Troubleshooting

If you’re having trouble, be sure you’ve carefully followed all the steps. The best way to debug is to use the JS console in your browser with the web app editor open.