Source github

Selecting and Binding Data

Selecting and appending

The selection methods select() and selectAll() receive a CSS selector expression and return a D3 handle for a node or a set of nodes. You can convert the D3 handle into a DOM object calling node() or nodes().

Add the following HTML to the <body> of your HTML file (or use Selecting/1-select.html):

<p>See results in console log.</p>
<div id="section">
    <p class=first>Paragraph 1</p>
    <p>Paragraph 2</p>
</div>
<p>Paragraph 3</p>

Now you can use the JavaScript console (or a <script> block in your page) to type in the following code, which shows how select() and selectAll() can be used:

const div = d3.select("#section");       // selects element with the 
                                         // "section" ID
const domDiv = div.node();               // converts the object into a 
                                         // DOM node
const firstP = d3.select("p");           // selects the first <p> 
                                         // in the page
const allParagraphs = d3.selectAll("p"); // selects all four <p> nodes 
                                         // in the page
const allPDom = allParagraphs.nodes();   // converts selection into 
                                         // DOM nodelist

const sectionParagraphs =
         d3.select("div")
                 .selectAll("p");       // selects the two <p> nodes 
                                        // inside <div>

A selection object serves as a handle to elements in your page. You can use it to change their styles, properties, attributes, classes, and contents, using methods that can receive static or dynamic values (using functions) as parameters. The following examples demonstrate some of these methods (see Selecting/2-attributes-styles.html):

div.style('border', 'solid blue 2px'); // draws blue border around <div>
firstP.classed('big', true);           // adds class ‘.big’ to 
                                         first paragraph 
firstP.text('This is paragraph zero'); // replaces contents of 
                                         first paragraph
allParagraphs.style('font-weight', 'bold'); // applies style to set 
                                              of paragraphs


// this looks promising, however there is nothing tagged with class "first" so nothing happes
d3.select('.first') // selects first element of class ‘.first’
        .attr('title', 'Tooltip')                  // adds an attribute
        .style('font-variant', 'small-caps')       // applies a style
        .html('This is a <b>bold</b> paragraph.'); // replaces contents with HTML

The select() method will always return only one object (if the selector matches more items, only the first will be returned). The selectAll() method returns a collection, which can be iterated with the each() method, which receives a callback function, which receives up to three parameters. Inside the function, this refers to the current DOM element, and d3.select(this) wraps it in a D3 selection object, which can be manipulated with selection methods, as follows:

sectionParagraphs.each(function(d, i) {     // i is the index
     d3.select(this).classed('red', true);
     console.log(`Paragraph ${i}: `, this); // this refers to the current element
 });
 ```

This code applies the class ```.red``` to each element of a collection. If you open the JavaScript console, it will print each element and its index (starting in zero).

You can also use selection methods to modify the DOM tree. Calling ```remove()``` on a D3 selection removes elements in the selection. You can call ```append()``` on a selection to add an element as its last child, or use ```insert()``` to add the new element in a position determined by a selector passed as the second argument. These methods are demonstrated as follows (see ```Selecting/3-append-remove.html)```:

```js
d3.select("#section")     // current context is element with id #section
  .insert("p", '.first') // add a <p> before the child being of class ‘.first’
  .append('a')           // context is now <p>; inserts an <a> in <p>
  .attr('href', 'https://d3js.org') // context is now <a> 
  .text("D3 website");   // contents of the <a> element

d3.select('div')            // context is first <div> in page
   .select('p:last-of-type') // context is now last <p> in <div>
   .remove(); // context is detached <p> (parentNode is null)

const div = d3.select('div')

div.append("p")             // adds a new <p> as last child of selected <div>
    .text("New paragraph");  // sets the text contents of <p>

d3.select("body")     // selects the <body> element of the page
    .insert("h1", "*")  // adds a new <h1> as the first child (before *)
    .text("New title"); // sets the text contents of <h1>

d3.select("body")
    .append("h2")       // adds a new <h2> as the last child
    .text("Footer");

The append() and insert() methods can also be chained, since they return the nodes they added. You just have to pay attention to the current so you don't modify or add data to the wrong elements. The remove() method returns the removed selection (a node or a set of nodes) after it's detached from the DOM (null parent).

D3 selections are not limited to HTML. In fact, they are most commonly applied to SVG elements, but can also be used with any embedded XML. This was a brief introduction. In Chapter 4, Data Binding, we will explore these methods in detail.

Binding Data

D3 makes it almost trivial to bind data to DOM elements. Data binding requires calling at least two of the following four methods in the context of a selection:

The exit() method returns a selection of unbound elements. After an exit(), remove() is usually called, which removes the entire selection. The enter() method returns a selection of placeholders for new elements that need to be created. After an enter(), either append() or insert() is called, which connects the elements to the DOM tree. The join() method is a magic method that replaces enter() and exit(), automatically updating, removing, or adding elements, as necessary.

Once data is connected to an element, the attributes, style, and contents of each element can be modified using the attr(),style(),text(), and html() methods that take a callback function as the second argument. This binding process is illustrated as follows:

architecture

Let's try an example. Create an HTML page with this code in the <body> and open it in a browser (or use Binding/1-binding.html):

In a <script> block, create this array of numbers (or type it in your JavaScript console):

const numbers = [6, 2, 5, 7, 9];

The following code will bind each number in the array to a paragraph. The data will be copied to an internal array and can be used to change the attributes, styles, or contents of the existing paragraphs:

const selection = d3.select("#section")
                    .selectAll("p")
                    .data(numbers);

The data() method returns the current selection. The value of each item in the array is received as the first parameter of a callback function in attr(),style(),text(), and so on. The text() method is used in the following code that replaces the contents of each paragraph with the first two values of the array:

selection.text(d => d); // binds to existing paragraphs (update)

But we need more paragraphs, since we have five elements of data. By calling enter(), a new selection is created with three more placeholder nodes, containing the remaining data, but not yet mapped to any element:

const newSelection = 
     selection.enter(); // binds data to a selection of placeholders

We can't just call text() on the selection yet. First, we need to append or insert those elements into the DOM tree. You call append() only once, and it will be executed three times. Now, you can use text() to print the paragraph with the data:

newSelection.append("p")
             .text(d => d); // binds to new paragraphs (enter)

Normally, the entire selecting-binding-appending process is written as a chain of commands. The following code achieves the same result, using the same array and paragraphs (see Binding/3-chain-text.html):

d3.select("body")         // select the body element
    .select("div")        // select the first div element inside <body>
    .selectAll("p")       // select all <p> elements (there are two)
    .data(numbers)        // load the data from the numbers array
    .text(d => d)         // set contents of existing paragraphs
    .enter()              // bind remaining data to array with 3 elements
    .append("p")          // add 3 new <p> elements to end of <div>
    .text(d => d);        // set the contents of the new paragraphs

It's simpler (and more common, when using D3) to use JavaScript (and not HTML) to create all of the elements you need. After all, data is dynamic and you won't always know how many items you have to display. Most of the time, HTML is just used to provide the basic structure, such as the <body> tag, a container <div>, or <svg>.

The following code achieves the same result starting with an empty <body> tag (and the same data array). The <div> and all the <p> are created using D3 commands (see Binding/4-empty-binding.html):

d3.select("body")        // select the body element
    .append("div")       // append a div element inside <body>
    .selectAll("p")      // select all <p> elements (there are none)
    .data(numbers)       // load the data from the numbers array
    .enter()             // bind the data to enter array with 5 placeholders
    .append("p")         // create 5 new <p> elements and add to end of <div>
    .text(d => d);       // set the contents of the new paragraphs

Note that even though there are no <p> elements in the page, the selectAll("p") command is still necessary, since it provides the selection context for the data binding. The selectAll() command can use any CSS selectors to locate its elements, but it should return a selection containing the same type of elements added by the append() command.

As you have seen, once you have data bound to DOM elements, you can use the data to change attributes, styles, classes, and properties, using callbacks. The callback function has a second parameter that contains the index of the data array, which is used in the following example to change the contents and style of list items:

d3.select("body")         // select the body element
    .append("ul")         // append an <ul> element inside <body>
    .selectAll("li")      // select all <li> elements (there are none)
    .data(numbers)        // load the data from the numbers array
    .enter()              // create an enter array with 5 objects
    .append("li")         // create 5 new <li> elements and append to <ul>
    .text(function(d, i) {
        return "Item " + (i+1) + ": " + d;
    })
    .style("font-size", function(d, i) {
    return ((i+2) * 5) + "pt";
    });

The result is shown in the following screenshot: list_example

D3 used to create and style new elements from data. Code: Binding/5-callbacks.html.

Callback functions also have a third parameter that contains the array of elements in the selection. You might need it if you want to obtain the current element when using arrow functions, since the this reference in these functions doesn't refer to the current element:

.text( (d, i, nodes) => console.log("Current element: " + nodes[i]) );

You can easily reuse the preceding code with different types of data. If you have an array of objects containing values of different types, you don't need to convert it to an array of numbers. You can directly pass the entire object array to the data() method, and later select which properties of each object you wish to use. For example, you can use practically the same code that was shown previously to display the contents of the following array:

const distances = [
    {name: "Mercury", distance: 0.387},
    {name: "Venus", distance: 0.723},
    {name: "Earth", distance: 1},
    {name: "Mars", distance: 1.52},
    {name: "Jupiter", distance: 5.2},
    {name: "Saturn", distance: 9.54},
    {name: "Uranus", distance: 19.2},
    {name: "Neptune", distance: 30.1},
    {name: "Ceres", distance: 2.765},
    {name: "Pluto", distance: 39.481},
    {name: "Eris", distance: 67.67},
    {name: "Haumea", distance: 43},
    {name: "Makemake", distance: 45.346}
];

The object array is received by the data() command. Each array item is available as the first parameter of the callback, which can be used to extract the property that contains the data (see Binding/6-object-array.html):

d3.select("body")    // select the body element
    .append("ul")     // append an <ul> element inside <body>
    .selectAll("li")  // select all <li> elements (there are none)
    .data(distances)  // load the data from the numbers array
    .enter()          // bind the data to enter array with 5 placeholders
    .append("li")     // create <li> from enter selection and append to <ul>
    .text(function(d, i) {
        return d.name + ": " + d.distance;
    });

Since D3 Version 5.8, you can replace enter().append(element) in simple binding operations, as shown with the join(element) method (see Binding/7-join.html):

d3.select("body") 
   .append("ul") 
   .selectAll("li") 
   .data(distances) 
   .join("li")  // obtains enter selection and appends 5 new <li> elements
   .text(function(d, i) {
       return d.name + ": " + d.distance;
   });

The preceding code will generate an HTML list with the data.

Creating a bar chart with D3 and HTML

Now that you have had a quick introduction to D3's basic data-binding mechanisms, let's create some full data visualization examples that will demonstrate the power of D3. You should know HTML very well, so we will start with a simple bar chart using HTML and CSS. Later, we will repeat the exact same procedure using SVG, the main standard used by D3 to create visualizations.

All the code for these examples is available in the GitHub repository for this chapter.

Binding data to HTML

In this example, we will draw a horizontal bar chart using HTML <div> elements. The colors and length of each bar are controlled by CSS, which we can configure using the selection.style() command. We will use the object array containing planetary distances from the last example. First, let's create a style sheet with classes for the bar-chart and each individual bar:

<style>
    .bar-chart {
        border: solid 1px gray; /* a gray border around the container */
        position: relative;
    }
    .bar {
        height: 20px;
        background-color: orange;
        position: absolute;
    }
</style>

The bars are positioned absolutely. Since each bar has an equal height (20px), we need to position each bar slightly below the previous one. This can be achieved by setting the value for the CSS top property (in pixels) so that it contains the sum of the heights of all the previous bars, plus an extra pixel to keep them slightly apart. The top property can be dynamically calculated as proportional to each entry's array index, using a callback in selection.style() (see HTML_Bar/1-bar-chart.html):

d3.select("body")
    .append("div").attr("class", "bar-chart")                 // container div for the chart
    .style("height", () => distances.length * 21 + "px")      // set chart height
    .selectAll("div").data(distances)                         // binds data
    .enter().append("div")                                    // appends a div for each data element
    .attr("class", "bar")                                     // these divs are the bars of  the chart
    .style("top", (d,i) => i * 21 + "px")                     // stacks bars 
    .style("width", "100px")                                  // fixed width

output:

bar_chart_0

This prints all the bars, one on top of the other, but they all have the same static width! We need to use a callback to change the width of each individual bar. In a bar chart, the actual width should be proportional to the distance. A solution would be to multiply each value by 10. This will make the smallest bar three pixels wide, and the largest over 670 pixels:

.style("width", d => (d.distance * 10) + "px")

Now, the lengths of the bars are proportional to the data values. This code is in HTML_Bar/2-bar-width.html. The result is shown as follows:

bar_chart_1

HTML bar chart with D3: Drawing DIVs and using the data to change the CSS width. Code: HTML_Bar/2-bar-width.html

Scales

What if a new planet is discovered that is 10 times the largest distance? The bars wouldn't fit and their widths would need to be recalculated! D3 also provides a solution for this: scales.

Scales are mappings between different dimensions. In our case, we have to deal with two dimensions: the dimension of the data, called the domain, and the dimension of the graphics context where the data will be displayed, called the range. Before using a scale, you have to configure its domain and range. There are several different kinds of scales in D3, including linear and logarithmic scales, with many configuration options.

A scale is created with a special generator function available in the d3-scale module. To use this module, you need to load several dependencies, so let's replace the <script> tag with the default bundle:

<script src="https://d3js.org/d3.v5.min.js"></script>

To create a scale function, you call a special generator function. A linear scale function can be created using the following:

const barScale = d3.scaleLinear();

To use the scale function you created, you pass a value as the argument and receive the converted result:

const result = barScale(45); // returns 45

This will return 45, because the scale hasn't been configured yet (the default scale is 1:1). To configure the scale, you need to call two methods to set up the domain (data dimensions) and range (graphical dimensions). Each method receives an array. For example, we can set up the domain as follows:

barScale.domain([0, 100]); // input domain from 0 to 100

This fits all the distances in our data. The range is set up in a similar fashion:

barScale.range([0, 600]); // output domain from 0 to 600

This means that zero is mapped to zero, and 100 astronomical units (AU) is mapped to 600 (pixels). Intermediate values will be interpolated.

If you now call barScale(45), you will get 270. We can use the scale to convert the distances in AU to pixels, replacing the previous expression with a scale conversion (see HTML_Bar/3-scales.html):

.style("width", d => barScale(d.distance) + "px");

Array utilities

We chose 100 as the upper limit in our scale domain because it's larger than any of the distances in the data, but the choice was rather arbitrary. We could have chosen the largest value in the array. If you have hundreds of lines of data, you can use JavaScript array functions (see Chapter 1, Introduction) to find out the maximum value.

D3 also includes a collection of array manipulation functions that extend JavaScript's native functions. Some look similar but may be more efficient for data manipulation (for example, by ignoring null, NaN, and undefined values). In order to use them, you need the d3-array module (which is also part of the default bundle).

Here, we changed the configuration of our barScale() function so that it uses the largest distance from the distances object array as the upper value for the domain. This is achieved with by calling the d3.max() function, which receives an array and an accessor function for each array element, as follows:

const barScale = d3.scaleLinear()
                   .domain([0, d3.max(distances, d => d.distance)]) 
                   .range([0, 600]);

The d3.max() function will scan the distances array and compare the distance property of each object, returning the largest one.

There are many more useful functions in d3-array, which we will cover in the next chapter. Two of them, d3.descending(a,b) and d3.ascending(a,b), are used to provide a sorting rule for JavaScript's native sort() method. We can use it to sort the array by the distance:

distances.sort((a,b) => d3.ascending(a.distance, b.distance));

See and run the code after these transformations in HTML_Bar/4-max-sort.html.

Adding labels

Since the <div> elements in HTML can contain text, you can call the text() method for each <div> and set its contents based on the data:

d3.select("body")
        .append("div").attr("class", "bar-chart")
        // ...all previous code to build chart
        .text(d => d.distance); // display distance in the <div>

With CSS, we can right-align the text and adjust the fonts. Since these styles are static and don't change with the data, instead of calling style() for each property, you should use a style sheet:

<style>
    /* ... */
    .bar {
        height: 20px;
        left: 100px;
        background-color: orange;
        position: absolute;
        text-align: right;
        padding: 0 5px;
        font-family: sans-serif;
        font-size: 9pt;
    }
</style>

As a result, the labels are placed inside the bars, as follows: bar_chart_2

Adding labels to the bars. Code: HTML_Bar/5-labels.html.

More labels, formatting, and colors

Each object in our data array also contains a name property that can be used to label each bar. To position text outside the bar, we will need to refactor the code so that each data entry contains a container <div>, which will be bound to the data values. This entry <div> will then contain three other <div> elements: a category label (the name of the planet), the bar, and a value label (the distance). Classes will be used to identify each <div>. The following style sheet contains the static properties for these elements:

<style>
    .bar-chart {         /* The container <div> for the entire chart */
        border: solid 1px gray;
        position: relative;
        width: 800px;
    }
    .entry {             /* a container <div> for each data entry */
        position: absolute;
        width: 100%;
    }
    .bar {               /* the colored rectangle */
        height: 20px;
        top: 1px;
        left: 100px;
        background-color: orange;
        position: absolute;
    }
    .label {             /* a text label */ 
        padding: 4px 5px;
        font-family: sans-serif;
        font-size: 9pt;
        position: absolute;
        height: 20px;
    } 
    .category {           /* the category text label at left (name) */
        text-align: right;
        width: 80px;
    }
    .value {              /* the value text label at right (distance) */
        text-align: left;
    }
</style>

Since each entry <div> has three children, we need to keep a reference to its selection so that the child elements can be appended. The following code saves references for each one of the container <div> elements. The chart constant contains a selection of the root <div> element, and the entries constant contains the selection of all entry <div> elements:

// selects the entire chart (one node)
 const chart = d3.select("body")
                 .append("div").attr("class", "bar-chart")
                 .style("height", distances.length * 21 + "px");

 // selects each entry (a nodelist)
 const entries = chart.selectAll("div").data(distances)
         .enter().append("div")
         .attr("class", "entry")
         .style("top", (d,i) => i * 21 + "px");

You can now use the entries constant to append the child elements to each entry <div> (see HTML_Bar/6-entries.html):

entries.append("div").attr("class", "label category")
        .text(d => d.name);

entries.append("div").attr("class", "bar")
        .style("width", d => barScale(d.distance) + "px");

entries.append("div").attr("class", "label value")
        .style("left", d => (barScale(d.distance) + 100) + "px")
        .text(d => d.distance + " AU");

Each append() shown is called once for each entry. The attributes, style, and text are set using the data that was bound to the parent container.

You can add child elements without having to break the selection chain with the selection.each() method, which calls a function for each entry. Inside it, you can obtain a selection to the current element using d3.select(this). The following code produces the same result (see HTML_Bar/7-entries-each.html):

entries.each(function(d) { 
    const entry = d3.select(this); // the current entry
    entry.append("div").attr("class", "label category")
         .text(d.name);

    entry.append("div").attr("class", "bar")
         .style("width", barScale(d.distance) + "px");

    entry.append("div").attr("class", "label value")
         .style("left", (barScale(d.distance) + 100) + "px")
         .text(d.distance + " AU");
});

We can improve the rendering of the labels by formatting the numbers to display only two decimal places, using the d3.format() generator function (from the d3-format module). The following code creates a function fmt() that can be used to format numbers:

const fmt = d3.format(".2f");

Now, we can use it to format the distances:

.text(d => fmt(d.distance) + " AU");

The final result is shown as follows: bar_chart_3

Adding category names and placing the formatted label values outside the bars. Code: HTML_Bar/7-entries-each.html

Changing colors

The d3-colors module contains functions to generate, convert, and transform colors. Passing any CSS-compatible color representation to d3.color() generates an object with methods that can be used to modify the color. The darker() and brighter() methods receive a value between 0 (no change) and 1 (maximum change) to adjust the lightness component of a color and return a hexadecimal color string.

Let's use this feature to darken the bars when the distance from the sun increases. We will need a scale that maps the distance domain to the [0-1] range, which contains the values accepted by the darker() function:

colorScale = d3.scaleLinear()
        .domain([0, d3.max(distances, d => d.distance)])
        .range([0,1])

Now, this colorScale can be used to generate different tones for the bars:

entry.append("div").attr("class", "bar")
    .style("width", barScale(d.distance) + "px")
    .style("background-color", d3.color('orange')
    .darker(colorScale(d.distance)))

See the code in HTML_Bar/8-colors.html. The result is shown as follows: bar_chart_4

Changing bar colors with the distance. Code: HTML_Bar/8-colors.html

Loading external files

Usually, your data will come from external files that need to be loaded with an Ajax request. You can, of course, use jQuery or ECMAScript fetch() commands (and promises, if you need to load multiple files). After loading, you will also need to parse the data and convert it into JavaScript arrays and objects. D3 again provides a more efficient solution in the d3-fetch module: a set of convenient methods for loading and parsing files in popular formats, such as XML, JSON, or CSV. The d3-fetch module is also included in the default bundle.

The planetary data we used in the previous examples is actually part of a larger JSON file containing several properties for planets, asteroids, and satellites. The basic structure of this file (Data/sol_2016.json) is shown as follows:

{
     "star":{…},
     "planets":[
         {
             "id":"p1",
             "name":"Mercury",
             "diameterKm":4879, 
             "semiMajorAxisAU":0.387, 
             …
            },{
             "id":"p2",
             "name":"Venus",
             "diameterKm":12104, 
             "semiMajorAxisAU":0.723, 
               …
            }, …
 }
 ```

You can load the data using the ```d3.json()``` function. After it loads and parses the file, it will become available in a callback provided as a parameter to the ```then()``` method (which is a JavaScript promise).

We don't need all the loaded data. The distance used in our examples is stored in the property called ```semiMajorAxisAU```. After loading the file, we can filter the dataset to only save the ```semiMajorAxisAU``` and ```name``` properties. The following code demonstrates this. It loads the file, and it then uses the data obtained in the ```then()``` callback to loop through the planets array, adding only the chosen properties to a new object and pushing it into an array. The array is used to create an HTML list with these properties (see``` Loading/1-loading-json.html```):

```js
//d3.json("../Data/sol_2016.json") // if you down load it
d3.json("https://raw.githubusercontent.com/PacktPublishing/Learn-D3.js/master/Chapter03/Data/sol_2016.json") // if you use the git hub data. This is so cool!
        .then(function(data) {
            const planets = [];
            data.planets.forEach(function(obj) {
                planets.push({
                    name: obj.name,
                    distance: obj.semiMajorAxisAU
                });
            });
            draw(planets);
        });

function draw(distances) {
    d3.select("body").append("ol")
      .selectAll("li")
      .data(distances)
      .join("li")
      .text(d => d.name + " (" + d.distance + " AU)");
}

To apply this to our chart, you just need to replace the contents of the draw() method shown with the entire code used in the previous examples (except for the distances array). See an example in HTML_Bar/9-load-json.html. As an exercise, try feeding the bar chart with a list of Jupiter's moons, instead of the planets (use data.planets[4].satellites to obtain the array, and the semiMajorAxisKm property as the distance, and d3.format(",.0f") to format the distance values).

Full Example:

d3.json("https://raw.githubusercontent.com/PacktPublishing/Learn-D3.js/master/Chapter03/Data/sol_2016.json") // if you use the git hub data. This is so cool!
        .then(function(data) {
            const planets = [];
            data.planets.forEach(function(obj) {
                planets.push({name: obj.name, distance: obj.semiMajorAxisAU});
            });
            draw(planets);
        });

function draw(distances) {
    const barScale = d3.scaleLinear()
    .domain([0, d3.max(distances, d => d.distance)]) 
    .range([0, 600]);

    const colorScale = d3.scaleLinear()
        .domain([0, d3.max(distances, d => d.distance)])
        .range([0,1])

    // selects the entire chart (one node)
    const chart = d3.select("body")
    .append("div").attr("class", "bar-chart")
    .style("height", distances.length * 21 + "px");

    // selects each entry (a nodelist)
    const entries = chart.selectAll("div").data(distances.sort((a,b) => d3.ascending(a.distance, b.distance)))
    .enter().append("div")
    .attr("class", "entry")
    .style("top", (d,i) => i * 21 + "px");


    entries.each(function(d) { 
    const entry = d3.select(this); // the current entry
    const fmt = d3.format(".2f"); // format for each entry

    entry.append("div").attr("class", "label category")
    .text(d.name);

    entry.append("div").attr("class", "bar")
    .style("width", barScale(d.distance) + "px")
    .style("background-color", d3.color('orange')
                        .darker(colorScale(d.distance)))

    entry.append("div").attr("class", "label value")
    .style("left", (barScale(d.distance) + 100) + "px")
    .text(d => fmt(d.distance) + " AU");
    });
}

Creating a bar chart with D3 and SVG

It's easy to create a horizontal bar chart in HTML. It's a bit trickier to make a vertical one, and quite challenging to create other charts, such as line charts, pie charts, and network diagrams, since HTML wasn't intended for vector graphics. While you can create such charts using HTML and CSS, it's not at all simple, nor is it recommended. You can, of course, use Canvas, but there are no graphical DOM elements to bind in Canvas. Canvas is used in D3 charts, but mostly for performance optimization. D3 works best with SVG.

The same bar chart that we created with HTML can be created with SVG. Do you think you can do it? Why not try it as an exercise before proceeding? You already learned how to use D3 and HTML, and you know the basics of SVG rectangles, attributes, and styles. That's all you need!

Most of the code is the same. You can reuse the scales, formatting function, sorting function, and JSON parsing code. The CSS will be simpler, since positioning is done in SVG. You only need to change the selection code, replacing the bar <div> element with <rect>, text label <div> elements with <text>, and container <div> elements with <g>. Remember to use fill (and not background-color) to fill the bars.

The same step-by-step files that were used in the HTML example are available for the SVG example in the SVG_Bar/ folder. The following, final code produces exactly the same chart as the HTML version.

This is the CSS style sheet that's used. Since all positioning is done in SVG, it's much smaller:

<style>
     * {
         font-family: sans-serif;
     }
     .bar-chart {
         border: solid 1px gray;
         width: 800px;
     }
     .bar {
         height: 20px;
         fill: orange;
     }
     .label {
         font-size: 9pt;
     }
     .category {
         text-anchor: end;
     }
 </style>

The main container is the <svg> element, which is appended to the body, as shown in the following code snippet. We renamed the chart object svg:


// selects the entire chart (one node)
 const svg = d3.select("body")
               .append("svg").attr("class", "bar-chart")
               .style("height", distances.length * 21);

The containers used for each entry are <g> elements, which group several child elements and can have their own coordinate system configured using a transform. Although the transform property is supported in both CSS and SVG, you should use the SVG version (selected with the attr() method), because it considers pixel values and degrees as the default. You don't have to append deg or px to any values:

// selects each entry (a nodelist)
 const entries = svg.selectAll("g").data(distances)
                    .enter().append("g")
                    .attr("class", "entry")
                    .attr("transform", (d,i) => `translate(0, ${i * 21})`);

The entries are appended to the container <g> in the each() method, as follows. Compare this code to the HTML version:


// sort distances
distances = distances.sort((a,b) => d3.ascending(a.distance, b.distance));

// selects the entire chart (one node)
const svg = d3.select("body")
.append("svg").attr("class", "bar-chart")
.style("height", distances.length * 21);

// barScales
const barScale = d3.scaleLinear()
.domain([0, d3.max(distances, d => d.distance)]) 
.range([0, 600]);

// colorScale
const colorScale = d3.scaleLinear()
.domain([0, d3.max(distances, d => d.distance)])
.range([0,1])

// decimal format
const fmt = d3.format(".2f");

// selects each entry (a nodelist)
const entries = svg.selectAll("g").data(distances)
.enter().append("g")
.attr("class", "entry")
.attr("transform", (d,i) => `translate(0, ${i * 21})`);

entries.each(function(d) {
    const entry = d3.select(this); // the current entry

    entry.append("text").attr("class", "label category")
         .attr("y", 15)
         .attr("x", 90)
         .text(d.name);

    entry.append("rect").attr("class", "bar")
         .attr("x", 100)
         .attr("width", barScale(d.distance) + "px")
         .style("fill", d3.color('orange')
                          .darker(colorScale(d.distance)))

    entry.append("text").attr("class", "label value")
        .attr("y", 15)
        .attr("x", barScale(d.distance) + 105)
        .text(fmt(d.distance) + " AU");
});

Try the full code from SVG_Bar/9-load-json.html. If you open it in your browser, you will notice that the result is identical to the one you get with HTML_Bar/9-load-json.html. To explore more D3 features, in the next sections, we will use the SVG version of this bar chart.

Updating Data Visualizations

Once the data is bound to SVG elements, you can change the original data values and reflect the changes in the chart. Values can change immediately or transition smoothly. This section will provide an introduction on how to trigger data updates and configure smooth transitions using D3.

First, we need more data. Let's change the code where we create our data object and include two more values that use the same dimensions: the aphelium (the longest distance between a planet and the sun) in the max property, and the perihelium (the shortest distance) in the min property. The distance is now stored in the avg property (see Updating/1-three-charts.html):

const planets = []; // this array will store all the data when loaded

 d3.json("../Data/sol_2016.json")
   .then(function(data) {
       data.planets.forEach(function(obj) {
           planets.push({
             name: obj.name,
             avg: obj.semiMajorAxisAU,
             max: obj.apheliumAU,
             min: obj.periheliumAU});
       });
       init();
   });

You can create bar charts with any of these three values, or all of them. Let's create an application where the user can choose which chart to display using HTML buttons. The following is the HTML code for the page; it includes an <svg> element and some buttons:

<body>
<h1><span id="chart">Average</span> distance from the Sun</h1>
<svg class="bar-chart"></svg>
<form>
    <button type="button" id="avg">Average</button>
    <button type="button" id="max">Maximum</button>
    <button type="button" id="min">Minimum</button>
</form>

<script>...</script>
</body>

These buttons will be attached to event handlers. They will select the data that should be displayed in the bar charts. The following array relates a key to a title and a color. The key contains the name of a property from each element in the planets array. It's also used for the button IDs:

const charts = [
    {key: "avg", title: "Average", color: "orange"},
    {key: "max", title: "Maximum", color: "blue"},
    {key: "min", title: "Minimum", color: "red"},
];

The chart object stores the dimensions of the current chart (which may have a variable height) and the chart that is currently displayed:

const chart = {
    width: 800,
    height: 0,          // the height is set after data is loaded
    current: charts[0]  // chart to display first
}

These other global constants initialize scales, a formatting function, and a selection of the svg object:

const barScale = d3.scaleLinear().range([0, 600]);
const colorScale = d3.scaleLinear().range([0, 1]);
const format = d3.format(".2f");

const svg = d3.select("svg.bar-chart"); // the container SVG

The init() function that is called right after the data is loaded and the planets array is populated performs basic initialization that requires the loaded data. In this case, it sets the height of the chart:

function init() {   // runs once
    chart.height = planets.length * 21;
    svg.attr("width", chart.width)
       .attr("height", chart.height);

    setupView(); // sets up scales; Not defined yet...see below

    // ...
}

The setupView() function called inside init() configures the current view. It disables the button that refers to the currently displayed view, replaces the <span> element in the title with the title of the current chart, sorts the planets array, and initializes the domains of the scales, based on the current data:

function setupView() {
    // disable all buttons
    d3.selectAll("button").property("disabled", false);

    // enable only buttons that are not current chart
    d3.select("#" + chart.current.key).property("disabled", true);

    // update page title
    d3.select("#chart").text(chart.current.title);

    // sort the planets using current data
    planets.sort((a,b) => d3.ascending(a[chart.current.key], 
                                        b[chart.current.key]));

    // update scale domain with current data
    const maxValue = d3.max(planets, d => d[chart.current.key]);
    barScale.domain([0, maxValue]);
    colorScale.domain([0, maxValue]);
}

After setting up the view, the init() function uses the current data to render the chart. This code is identical to the code we used in the previous examples. The only difference is that it uses a key reference to access the data property: instead of d.avg, it uses d[chart.current.key]. This will allow the chart to reference other properties when the current key changes:

function init() {
     // ...
     setupView();

     svg.selectAll("g")
        .data(planets)
        .enter().append("g").attr("class", "entry")
        .attr("transform", (d,i) => `translate(0,${i * 21})`)
        .each(function(d) {
            const entry = d3.select(this); // the current entry

            entry.append("text").attr("class", "label category")
                .attr("y", 15)
                .attr("x", 90)
                .text(d.name);

            entry.append("rect").attr("class", "bar")
                .attr("x", 100)
                .attr("height", 20)
                .attr("width", barScale(d[chart.current.key]) )
                .style("fill", d3.color(chart.current.color)
                                 .darker(colorScale(d[chart.current.key])) )

            entry.append("text").attr("class", "label value")
                .attr("y", 15)
                .attr("x", barScale(d[chart.current.key]) + 105)
                .text(format(d[chart.current.key]) + " AU");
        });
}

The result is shown as follows. It's the same chart we created before, with a title and some buttons that don't work yet:

bar_chart_5

The previous example adapted to show three different charts. Code: Updating/1-three-charts.html.

Handling events

The on() method is used to handle events, and it can be called from any selection. The first parameter is a standard JavaScript event name string (such as click or mouseover), and the second parameter is the handler function that will execute when the event happens.

The following code obtains a selection containing all button objects and attaches an event handler to all of them. It obtains the id of the button that was clicked and uses it to change the current chart by assigning a corresponding object from the charts array. After changing the current chart, it calls the draw() function, which will update the chart. This code should be placed in a global context, since it only needs to run once:

d3.selectAll("button")
     .on("click", function() {
         chart.current = charts.filter(c => c.key == this.id)[0];
         draw();
     });

The draw function in this example only prints the current array. You can use it to test whether the buttons are selecting the correct chart, as expected. The call to setupView() will disable/enable the buttons and update the chart's title according to the current view (see Updating/2-events.html):

function draw() {
     console.log(chart.current.key);
     setupView();
 }

Data updates

To update the data on a selection, you just need to update the styles and attributes. If the data has changed, you should call the selection.data() method with the new data, and then update everything that depends on it, such as attributes and styles, and any functions called by them, such as scales.

In our case, the data bound to the container g.entry object is the entire planets array, which may have been sorted in a different order (in the setupView() function). It can be updated simply by reassigning it to the selection:

svg.selectAll("g.entry").data(planets)

Then, you need to update the attributes and styles, which depend on properties from this array that have changed, but before doing that, you need to update the scales' domains, since they are called from the style() and attr() methods, and their maximum value depends on the new data. The scales were updated in the setupView() function. The rest of the draw() function contains the data updates:

function draw() {
    setupView(); // sorts data and updates scales

    svg.selectAll("g.entry").data(planets)
        .each(function (d) {
            d3.select(this).select(".label.category")
                .text(d.name); // the order may have changed
            d3.select(this).select(".bar")
                .attr("width", barScale(d[chart.current.key]))
                .style("fill", d3.color(chart.current.color)
                                .darker(colorScale(d[chart.current.key])));
            d3.select(this).select(".label.value")
                .attr("x", barScale(d[chart.current.key]) + 105)
                .text(format(d[chart.current.key]) + " AU");
        });
}

Now, when you click any button, a new bar chart will be displayed. The chart reuses the same graphical elements, changing its colors, dimensions, and text contents. Bars are always sorted in ascending order (see Updating/3-updates.html):

bar_chart_6

Each button replaces the current chart with a new chart, reusing the same chart elements. Code: Updating/3-updates.html.

Smooth transitions

Instead of immediately replacing one chart with another, you can add transitions, so that they occur in smooth animations. Adding a transition is very easy. You just need to call the transition() method on a selection before setting the new properties and styles. Instead of changing the old data with the new data, it will interpolate intermediate values during a quarter of a second. Add it to your update selections, and you will notice that when you change the chart, the bars and labels will animate to their new sizes, colors, and positions.

In our examples, transitions were added before changing the color and width of each bar, and before changing the position of each value label (see Updating/4-transitions.html):

svg.selectAll("g.entry").data(planets)
     .each(function (d) {
         d3.select(this).select(".label.category").text(d.name); 
         d3.select(this).select(".bar")
           .transition()                 // 1) transition fill and width
             .attr("width", barScale(d[chart.current.key]))
             .style("fill", d3.color(chart.current.color)
                              .darker(colorScale(d[chart.current.key])));
         d3.select(this).select(".label.value")
           .transition()                 // 2) transition x position
             .attr("x", barScale(d[chart.current.key]) + 105)
             .text(format(d[chart.current.key]) + " AU");
     });

If you want a slower transition, just add duration(value) after the transition() command, with a value in milliseconds. The following code will make the animation last one second (see Updating/5-durations.html):

d3.select(this).select(".bar")
   .transition()
   .duration(1000)               // animate during a second
     .attr("width", barScale(d[chart.current.key]))
     .style("fill", d3.color(chart.current.color)
                      .darker(colorScale(d[chart.current.key])));

You can also configure a delay before the transition starts with the delay() method. It can receive a fixed value in milliseconds, or a callback function that will apply a different delay for each object. In the following example, each bar will wait an amount of milliseconds proportional to its array index (the second parameter in the each() function), making each bar animate in a sequence:

.each(function (d,i) {     // include the i (index) parameter
     // ...
     d3.select(this).select(".bar")
       .transition()
       .duration(1000)         // animate during a second
       .delay(50 * i)          // longer animations for each bar
         .attr("width", barScale(d[chart.current.key]))
         .style("fill", d3.color(chart.current.color)
                          .darker(colorScale(d[chart.current.key])));
     // ...
 });

The following diagram shows a snapshot of the animation captured in the middle of a transition. Try it out and see the effects for yourself (see the code in Updating/6-delays.html):

bar_chart_7

Transition with delay. Code: Updating/6-delays.html.

So far, you've used the most common features of D3, rendering charts in HTML and SVG. We will finish this chapter with a D3 visualization of a map of the world, using the same methods you have used in the previous examples.