D3: Data Driven Documents
D3 can be used to manipulate pure HTML, but most commonly it’s used in combination with SVG (i.e., we will be producing SVG charts using D3).
In addition to the introduction in Scott Murray’s book (the mandatory reading) you should work with the D3 API Reference to look up particulars of all the features of D3. And of course, you should be learning by examples. A great collection are Mike Bostock’s blocks which contain simple examples, such as a bar chart to complex examples such as this calendar view.
You can download the library to run locally on your computer from the D3 Website, or you can link directly to the latest release with this snippet:See output in new page.
You should also note that D3 underwent a major version change from D3 v3 to D3 v4. Unfortunately, many examples you’ll find on the web will be using version 3. We’ll be using D3 v5, which is a minor update from D3 v4, in this class. As such, you can’t necessarily expect to just copy and paste example code from the web and have it work. Check out the release notes to learn about the changes across the D3 versions.
Here is a minimal D3 example taken from the D3 website:See output in new page.
You can see that we achieve a similar result to the DOM manipulation examples we had before. We select an existing element in the DOM, here the first
<p> element, and apply a style. However, you can also see differences: instead of having to use the API standard
document.getElementsByTagName we use
d3.select; and instead of using
setAttribute("style", "color: steelblue;") we use D3’s
See output in new page.
d3.select selects the first element that matches a selector. Selectors can specify tags (
p in our example above), classes, and IDs, all through the same interface:
Notice, however, that as mentioned previously, only the first element that matches is selected. Of course, it is more practical to select all elements of a certain type, which we can achieve with
The last example illustrates the declarative approach of D3: we don’t have to iterate over a list of elements and apply the style. Instead we select a set of elements through rules and declare properties.
Once you have a selection, you can bulk-modify it’s content, not only in terms of style, but we can modify arbitrary properties using
selection.property(name[, value]), the textual content of the elements with
selection.text([value]), etc. We can also append elements:
Here is a code snippet that we will use in future examples, included as run.js:
1 2 3 var button = d3.select("body").append("button"); button.text("Run!"); button.on("click", execute);
Here we have three SVG rectangles and use selectAll to apply a new style to them. Notice two things:
- the method chaining used to make this convenient to write,
- the use of anonymous functions and the parameters d and i.
- the comparison of explicit anonymous function and the arrow function.
THE key feature of D3 is mapping DOM elements with data. We can do this by calling the
data() method on a selection:
D3 binds the data that is passed to a selection directly to the DOM elements. We can see this by logging and/or inspecting the selection.
Here we’ve also used data-driven styling for the first time! By setting the width of the rectangle dynamically to the size of the data item we get a data driven bar chart!
What happens if we have more data points than elements? Let’s try it:See output in new page.
There are still only three elements, it doesn’t matter how many data points we have - we can’t use more than there are elements to select. What to do?
enter()! The enter selection holds placeholders for all the data elements that had no corresponding DOM element. We can use this selection to append new elements.
Progress - we have one element for each data item, but it doesn’t look good and it’s not drawn based on data. To fix this, we can apply data-driven attributes to the enter selection as well:See output in new page.
Now, that works! But we’re duplicating code. So instead we can do this shorter version:See output in new page.
So what if we don’t have initialized svg elements at all?See output in new page.
That’s great! D3 can also select things that aren’t there. Sounds strange, but that is very practical. We rarely start with existing DOM elements but mostly want to create them completely in code.
Now let’s explore what happens when we have fewer data items than DOM elements:See output in new page.
We have a similar problem as before. Here, the element that was not bound to data just remains as it was. Fortunately, there is a solution, which is the
Here is a an illustration showing the differences between the enter, update and exit selection:
We now have all the knowledge to select, bind data, react to changes in data. Next, we’ll look at how we can make state changes look great.
One of the cool features of D3 is that transitions are baked in!See output in new page.
Cool, but we’re missing the enter again:See output in new page.
Here, we’ve handled the enter and chained two transitions.
Here’s one way to draw a line:See output in new page.
But we already know that. There must be something better, and there is!See output in new page.
Here is an exercise to practice your D3 skills.