48 KiB

Raw Blame History

Creating a Scatter Plot

Let's pretend we've started jogging, and we want to visualize the data regarding our progress as a runner with a scatter plot. We're going to have an array of objects each with date and distance properties. For each object in the array, we're going to create a circle in our SVG. If the distance property of an object is relatively high, its associated circle will be higher up on the graph. If the date property of an object is relatively high (a later date), its associated circle be farther right.

By the end of this lesson, you should be able to:

Add link to d3 library
Add an <svg> tag and size it with D3
Create some fake data for our app
Add SVG circles and style them
Create a linear scale
Attach data to visual elements
Use data attached to a visual element to affect its appearance
Create a time scale
Parse and format times
Set dynamic domains
Dynamically generate svg elements
Create axes
Display data in a table
Create click handler
Remove data
Drag an element
Update data after a drag
Create a zoom behavior that scales elements
Update axes when zooming
Update click points after a transform
Avoid redrawing entire screen during render
Hide elements beyond axis
Use AJAX

Add link to d3 library

The first thing we want to do is create basic index.html file:

<!DOCTYPE html>
<html>
    <head>
        <meta charset="utf-8">
        <title></title>
    </head>
    <body>
    </body>
</html>

Now add a link to D3 at the bottom of your <body> tag in index.html. We'll put it at the bottom so that the script loads after all your other HTML elements have loaded into the browser:

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

Now create app.js in the same folder as your index.html. In it, we will store all of our JS code. For now just put this code in it to see if it works:

console.log('this works');
console.log(d3);

and link to it in index.html at the bottom of the <body> tag. Make sure it comes after the D3 script tag so that D3 loads before your app.js script:

<body>
    <script src="https://d3js.org/d3.v5.min.js"></script>
    <script src="app.js" charset="utf-8"></script>
</body>

Open index.html in Chrome just like we did in the SVG chapter (File->Open File) and check your dev tools (View->Developer->Developer Tools) to see if your javascript files are linked correctly:

Add an `<svg>` tag and size it with D3

In index.html, at the top of the <body>, before your script tags, add an <svg> tag:

<body>
    <svg></svg>
    <script src="https://d3js.org/d3.v5.min.js"></script>
    <script src="app.js" charset="utf-8"></script>
</body>

If we examine the Elements tab of our dev tools, we'll see the svg element has been placed. In Chrome, it has a default width/height of 300px/150px

In app.js, remove your previous console.log statements and create variables to hold the width and height of the <svg> tag:

var WIDTH = 800;
var HEIGHT = 600;

Next, we can use d3.select() to select a single element, in this case the <svg> element:

var WIDTH = 800;
var HEIGHT = 600;

d3.select('svg');

The return value of d3.select('svg') is a D3 version of the svg element (just like in jQuery), so we can "chain" commands onto this. Let's add some styling to adjust the height/width of the element:

d3.select('svg')
    .style('width', WIDTH)
    .style('height', HEIGHT);

Now when we check the dev tools, we'll see the <svg> element has been resized:

Create some fake data for our app

In app.js let's create an array of "run" objects (NOTE: I'm storing the date as a string on purpose. Also, it's important that this be an array of objects, in order to work with D3). Here's what your app.js code should look like so far:

var WIDTH = 800;
var HEIGHT = 600;

var runs = [
    {
        id: 1,
        date: 'October 1, 2017 at 4:00PM',
        distance: 5.2
    },
    {
        id: 2,
        date: 'October 2, 2017 at 5:00PM',
        distance: 7.0725
    },
    {
        id: 3,
        date: 'October 3, 2017 at 6:00PM',
        distance: 8.7
    }
];

d3.select('svg')
    .style('width', WIDTH)
    .style('height', HEIGHT);

Add SVG circles and style them

In index.html, add three circles to your <svg> element (each one will represent a run):

<svg>
    <circle/>
    <circle/>
    <circle/>
</svg>

Create app.css in the same folder as index.html with some styling for the circles and our svg element:

circle {
    r:5;
    fill: black;
}
svg {
    border: 1px solid black;
}

and link to it in the head of index.html:

<head>
    <meta charset="utf-8">
    <title></title>
    <link rel="stylesheet" href="app.css">
</head>

Our page should now look like this:

Note that all three circles are in the upper left corder of the screen. This is because all three are positioned at (0,0) so they overlap each other. It appears as if there is just one circle, but in reality all three are present

Create a linear scale

We currently have three circles in our SVG and three objects in our runs array. One of the best things D3 does is provide the ability to link SVG elements with data so that as the data changes, so do the SVG elements. In this chapter, we're going to link each circle to an object in the runs array. If the distance property of an object is relatively high, its associated circle will be higher up on the graph. If the date property of an object is relatively high (a later date), its associated circle be farther right.

First, let's position the circles vertically, based on the distance property of the objects in our runs array. One of the most important things that D3 does is provide the ability to convert (or "map") data values to visual points and vice versa. It does so using a scale. There are lots of different kinds of scales that handle lots of different data types, but for now we're just going to use a linear scale which will map numeric data values to numeric visual points and vice versa.

At the bottom of app.js, add the following:

var yScale = d3.scaleLinear(); //create the scale

Whenever we create a scale, we need to tell it what are the minimum and maximum possible values that can exist in our data (this is called the "domain"). To do so for our yScale, add the following to the bottom of app.js:

yScale.domain([0, 10]); //minimum data value is 0, max is 10

We also need to tell the scale what visual values correspond to those min/max values in the data (this is called the "range"). To do so, add the following to the bottom of app.js:

//HEIGHT corresponds to min data value
//0 corresponds to max data value
yScale.range([HEIGHT, 0]);

Your last three lines of code in app.js should look like this now:

var yScale = d3.scaleLinear(); //create the scale
yScale.domain([0, 10]); //minimum data value is 0, max is 10
//HEIGHT corresponds to min data value
//0 corresponds to max data value
yScale.range([HEIGHT, 0]);

In the previous snippet, the first (starting) value for the range is HEIGHT (600) and the second (ending) value is 0. The minimum for the data values is 0 and the max is 10. By doing this, we're saying that a data point (distance run) of 0 should map to a visual height value of HEIGHT (600):

This is because the lower the distance run (data value), the more we want to move the visual point down the Y axis. Remember that the Y axis starts with 0 at the top and increases in value as we move down vertically on the screen.

We also say that a data point (distance run) of 10 should map to a visual height of 0:

Again, this is because as the distance run increases, we want to get back a visual value that is lower and lower so that our circles are closer to the top of the screen.

If you ever need to remind yourself what the domain/range are, you can do so by logging yScale.domain() or yScale.range(). Temporarily add the following at the bottom app.js:

console.log(yScale.domain()); //you can get the domain whenever you want like this
console.log(yScale.range()); //you can get the range whenever you want like this

Our Chrome console should look like this:

When declaring range/domain of a linear scale, we only need to specify start/end values for each. Values in between the start/end will be calculated by D3. For instance, to find out what visual value corresponds to the distance value of 5, use yScale(). Remove the previous two console.log() statements and add the following to the bottom of app.js:

console.log(yScale(5)); //get a visual point from a data value

Here's what our dev console should look like in Chrome:

It makes sense that this logs 300 because the data value of 5 is half way between the minimum data value of 0 and the maximum data value of 10. The range starts at HEIGHT (600) and goes to 0, so half way between those values is 300.

So whenever you want to convert a data point to a visual point, call yScale(). We can go the other way and convert a visual point to a data value by calling yScale.invert(). To find out what data point corresponds to a visual value of 450, remove the previous console.log() statement and add the following to the bottom of app.js:

console.log(yScale.invert(450)); //get a data values from a visual point

Here's what Chrome's console looks like:

It makes sense that this logs 2.5 because the visual value of 450 is 25% of the way from the starting visual value of 600 (HEIGHT) to the ending visual value of 0. You can now delete that last console.log() line.

Attach data to visual elements

Now let's attach each of the javascript objects in our "runs" array to a circle in our SVG. Once we do this, each circle can access the data of its associated "run" object in order to determine its position. Add the following to the bottom of app.js:

d3.selectAll('circle').data(runs); //selectAll is like select, but selects all elements that match the query string

If there were more objects in our "runs" array than there are circles, the extra objects are ignored. If there are more circles than objects, then javascript objects are attached to circles in the order in which they appear in the DOM until there are no more objects to attach.

Use data attached to a visual element to affect its appearance

We can change attributes for a selection of DOM elements by passing static values, and all selected elements will have that attribute set to that one specific value. Add the following temporarily to the end of app.js:

d3.selectAll('circle').attr('cy', 300);

But now that each circle has one of our "runs" javascript data objects attached to it, we can set attributes on each circle using that data. We do that by passing the .attr() method a callback function instead of a static value for its second parameter. Remove d3.selectAll('circle').attr('cy', 300); and adjust the last line of app.js from d3.selectAll('circle').data(runs); to the following:

d3.selectAll('circle').data(runs)
    .attr('cy', function(datum, index){
        return yScale(datum.distance);
    });

If we refresh the browser, this is what we should see:

Let's examine what we just wrote. The callback function passed as the second parameter to .attr() runs on each of the visual elements selected (each of the circle elements in this case). During each execution of the callback, the return value of that callback function is then assigned to whatever aspect of the current element is being set (in this case the cy attribute).

The callback function takes two params:

the individual datum object from the runs array that was attached to that particular visual element when we called .data(runs)
the index of that datum in the runs array

In summary what this does is loop through each circle in the SVG. For each circle, it looks at the "run" object attached to that circle and finds its distance property. It then feeds that data value into yScale() which then converts it into its corresponding visual point. That visual point is then assigned to that circle's cy attribute. Since each data object has a different distance value, each circle is placed differently, vertically.

Create a time scale

Let's position the circles horizontally, based on the date that their associated run happened. First, create a time scale. This is like a linear scale, but instead of mapping numeric values to visual points, it maps Dates to visual points. Add the following to the bottom of app.js:

var xScale = d3.scaleTime(); //scaleTime maps date values with numeric visual points
xScale.range([0,WIDTH]);
xScale.domain([new Date('2017-10-1'), new Date('2017-10-31')]);

console.log(xScale(new Date('2017-10-28')));
console.log(xScale.invert(400));

Here's what our console should look like:

You can now remove the two console.log() statements.

Parse and format times

Note that the date properties of the objects in our runs array are strings and not Date objects. This is a problem because xScale, as with all time scales, expects its data values to be Date objects. Fortunately, D3 provides us an easy way to convert strings to dates and vice versa. We'll use a specially formatted string, based on the documentation (https://github.com/d3/d3-time-format#locale_format), to tell D3 how to parse the date String properties of the objects in our runs array into actual JavaScript Date objects. Add the following at the end of app.js:

var parseTime = d3.timeParse("%B%e, %Y at %-I:%M%p"); //this format matches our data in the runs array
console.log(parseTime('October 3, 2017 at 6:00PM'));

var formatTime = d3.timeFormat("%B%e, %Y at %-I:%M%p"); //this format matches our data in the runs array
console.log(formatTime(new Date()));

Let's use this when calculating cx attributes for our circles. Remove the last two console.log() statements and add the following at the bottom of app.js:

d3.selectAll('circle')
    .attr('cx', function(datum, index){
        return xScale(parseTime(datum.date)); //use parseTime to convert the date string property on the datum object to a Date object, which xScale then converts to a visual value
    });

Here's what Chrome should look like:

In summary, this selects all of the circle elements. It then sets the cx attribute of each circle to the result of a callback function. That callback function runs for each circle and takes the "run" data object associated with that circle and finds its date property (remember it's a string, e.g. 'October 3, 2017 at 6:00PM'). It passes that string value to parseTime() which then turns the string into an actual JavaScript Date object. That Date object is then passed to xScale() which converts the date into a visual value. That visual value is then used for the cx attribute of whichever circle the callback function has just run on. Since each date property of the objects in the runs array is different, the circles have different horizontal locations.

Set dynamic domains

At the moment, we're setting arbitrary min/max values for the domains of both distance and date. D3 can find the min/max of a data set, so that our graph displays just the data ranges we need. All we need to do is pass the min/max methods a callback which gets called for each item of data in the runs array. D3 uses the callback to determine which properties of the datum object to compare for min/max

Go to this part of the code:

var yScale = d3.scaleLinear(); //create the scale
yScale.range([HEIGHT, 0]); //set the visual range (e.g. 600 to 0)
yScale.domain([0, 10]); //set the data domain (e.g. 0 to 10)

and change it to this:

var yScale = d3.scaleLinear(); //create the scale
yScale.range([HEIGHT, 0]); //set the visual range (e.g. 600 to 0)
var yMin = d3.min(runs, function(datum, index){
    return datum.distance; //compare distance properties of each item in the data array
})
var yMax = d3.max(runs, function(datum, index){
    return datum.distance; //compare distance properties of each item in the data array
})
yScale.domain([yMin, yMax]); //now that we have the min/max of the data set for distance, we can use those values for the yScale domain
console.log(yScale.domain());

Chrome should look like this:

Let's examine what we just wrote. The following code finds the minimum distance:

var yMin = d3.min(runs, function(datum, index){
    return datum.distance; //compare distance properties of each item in the data array
})

D3 loops through the runs array (the first parameter) and calls the callback function (the second parameter) on each element of the array. The return value of that function is compared the return values of the callback function as it runs on the other elements. The lowest value is assigned to yMin. The same thing happens for d3.max() but with the highest value.

We can combine both the min/max functions into one extent function that returns an array that has the exact same structure as [yMin, yMax]. Change the code we just wrote:

var yScale = d3.scaleLinear(); //create the scale
yScale.range([HEIGHT, 0]); //set the visual range (e.g. 600 to 0)
var yMin = d3.min(runs, function(datum, index){
    return datum.distance; //compare distance properties of each item in the data array
})
var yMax = d3.max(runs, function(datum, index){
    return datum.distance; //compare distance properties of each item in the data array
})
yScale.domain([yMin, yMax]); //now that we have the min/max of the data set for distance, we can use those values for the yScale domain

to this:

var yScale = d3.scaleLinear(); //create the scale
yScale.range([HEIGHT, 0]); //set the visual range (e.g. 600 to 0)
var yDomain = d3.extent(runs, function(datum, index){
    return datum.distance; //compare distance properties of each item in the data array
})
yScale.domain(yDomain);

Much shorter, right? Let's do the same for the xScale's domain. Go to this part of the code:

var xScale = d3.scaleTime(); //scaleTime maps date values with numeric visual points
xScale.range([0,WIDTH]);
xScale.domain([new Date('2017-10-1'), new Date('2017-10-31')]);

var parseTime = d3.timeParse("%B%e, %Y at %-I:%M%p"); //this format matches our data in the runs array
var formatTime = d3.timeFormat("%B%e, %Y at %-I:%M%p"); //this format matches our data in the runs array

and change it to:

var parseTime = d3.timeParse("%B%e, %Y at %-I:%M%p");
var formatTime = d3.timeFormat("%B%e, %Y at %-I:%M%p");
var xScale = d3.scaleTime();
xScale.range([0,WIDTH]);
var xDomain = d3.extent(runs, function(datum, index){
    return parseTime(datum.date);
});
xScale.domain(xDomain);

Notice we moved parseTime and formatTime up so they could be used within the .extent(). Here's what Chrome should look like:

Dynamically generate SVG elements

Currently, we have just enough <circle> elements to fit our data. What if we don't want to count how many elements are in the array? D3 can create elements as needed. First, remove all <circle> elements from index.html. Your <body> tag should look like this now:

<svg></svg>

In app.js, go to this part of the code:

d3.selectAll('circle').data(runs)
    .attr('cy', function(datum, index){
        return yScale(datum.distance);
    });

modify the code to create the circles:

d3.select('svg').selectAll('circle') //since no circles exist, we need to select('svg') so that d3 knows where to append the new circles
    .data(runs) //attach the data as before
    .enter() //find the data objects that have not yet been attached to visual elements
    .append('circle'); //for each data object that hasn't been attached, append a <circle> to the <svg>

d3.selectAll('circle')
    .attr('cy', function(datum, index){
        return yScale(datum.distance);
    });

It should look exactly the same as before, but now circles are being created for each object in the runs array:

Here's a more in depth explanation of what we just wrote. Take a look at the first line of new code from above:

d3.select('svg').selectAll('circle')

This might seem unnecessary. Why not just do d3.selectAll('circle')? Well, at the moment, there are no circle elements. We're going to be appending circle elements dynamically, so d3.select('svg') tells D3 where to append them. We still need .selectAll('circle') though, so that when we call .data(runs) on the next line, D3 knows what elements to bind the various objects in the runs array to. But there aren't any circle elements to bind data to. That's okay: .enter() finds the "run" objects that haven't been bound to any circle elements yet (in this case all of them). We then use .append('circle') to append a circle for each unbound "run" object that .enter() found.

Create axes

D3 can automatically generate axes for you. Add the following to the bottom of app.js:

var bottomAxis = d3.axisBottom(xScale); //pass the appropriate scale in as a parameter

This creates a bottom axis generator that can be used to insert an axis into any element you choose. Add the following code at the bottom of app.js to append a <g> element inside our SVG element and then insert a bottom axis inside of it:

d3.select('svg')
	.append('g') //put everything inside a group
	.call(bottomAxis); //generate the axis within the group

Here's what Chrome should look like:

We want the axis to be at the bottom of the SVG, though. Modify the code we just wrote so it looks like this (NOTE: we removed a ; after .call(bottomAxis) and added .attr('transform', 'translate(0,'+HEIGHT+')');):

var bottomAxis = d3.axisBottom(xScale); //pass the appropriate scale in as a parameter
d3.select('svg')
	.append('g') //put everything inside a group
	.call(bottomAxis) //generate the axis within the group
    .attr('transform', 'translate(0,'+HEIGHT+')'); //move it to the bottom

Currently, our SVG clips the axis:

Let's alter our svg CSS so it doesn't clip any elements that extend beyond its bounds:

svg {
    overflow: visible;    
}

Now it looks good:

The left axis is pretty similar. Add the following to the bottom of app.js:

var leftAxis = d3.axisLeft(yScale);
d3.select('svg')
	.append('g')
	.call(leftAxis); //no need to transform, since it's placed correctly initially

Note we don't need to set a transform attribute since it starts out in the correct place initially:

It's a little tough to see, so let's add the following at the bottom of app.css:

body {
    margin: 20px 40px;
}

Now our axes are complete:

Display data in a table

Just for debugging purposes, let's create a table which will show all of our data. Make your <body> tag in index.html look like this:

<body>
    <svg></svg>
    <table>
        <thead>
            <tr>
                <th>id</th>
                <th>date</th>
                <th>distance</th>
            </tr>
        </thead>
        <tbody>
        </tbody>
    </table>
    <script src="https://d3js.org/d3.v5.min.js"></script>
    <script src="app.js" charset="utf-8"></script>
</body>

D3 can also be used to manipulate the DOM, just like jQuery. Let's populate the <tbody> in that style. Add the following to the bottom of app.js:

var createTable = function(){
    for (var i = 0; i < runs.length; i++) {
        var row = d3.select('tbody').append('tr');
        row.append('td').html(runs[i].id);
        row.append('td').html(runs[i].date);
        row.append('td').html(runs[i].distance);
    }
}

createTable();

Add some styling for the table at the bottom of app.css:

table, th, td {
   border: 1px solid black;
}
th, td {
    padding:10px;
    text-align: center;
}

Adjust the CSS for svg to add a bottom margin. This will create some space between the graph and the table:

svg {
    overflow: visible;
    margin-bottom: 50px;
}

Now the browser should look like this:

Create click handler

Let's say that we want it so that when the user clicks on the <svg> element, it creates a new run. Add the following to the bottom of app.js:

d3.select('svg').on('click', function(){
    var x = d3.event.offsetX; //gets the x position of the mouse relative to the svg element
    var y = d3.event.offsetY; //gets the y position of the mouse relative to the svg element

    var date = xScale.invert(x) //get a date value from the visual point that we clicked on
    var distance = yScale.invert(y); //get a numeric distance value from the visual point that we clicked on

    var newRun = { //create a new "run" object
        id: runs[runs.length-1].id+1, //generate a new id by adding 1 to the last run's id
        date: formatTime(date), //format the date object created above to a string
        distance: distance //add the distance
    }
    runs.push(newRun); //push the new run onto the runs array
    createTable(); //render the table
});

Let's examine what we just wrote. d3.select('svg').on('click', function(){ Sets up a click handler on the svg element. The anonymous function that gets passed in as the second parameter to .on() gets called each time the user clicks on the SVG. Once inside that callback function, we use d3.event.offsetX to get the x position of the mouse inside the SVG and d3.event.offsetY to get the y position. We then use xScale.invert() and yScale.invert() to turn the x/y visual points into data values (date and distance, respectively). We then use those data values to create a new run object. We create an id for the new run by getting the id of the last element in the runs array and adding 1 to it. Lastly, we push the new run onto the runs array and call createTable().

Click on the SVG to create a new run. You might notice that createTable() just adds on all the run rows again

Let's alter the createTable() function so that when it runs, it clears out any rows previously created and re-renders everything. Add d3.select('tbody').html('') to the top of the createTable function in app.js:

var createTable = function(){
    d3.select('tbody').html(''); //clear out all rows from the table
    for (var i = 0; i < runs.length; i++) {
        var row = d3.select('tbody').append('tr');
        row.append('td').html(runs[i].id);
        row.append('td').html(runs[i].date);
        row.append('td').html(runs[i].distance);
    }
}

Now refresh the page, and click on the SVG to create a new run. The table should look like this now:

The only issue now is that circles aren't being created when you click on the SVG. To fix this, let's wrap the code for creating <circle> elements in a render function, and call render() immediately after it's defined:

var render = function(){

    var yScale = d3.scaleLinear();
    yScale.range([HEIGHT, 0]);
    yDomain = d3.extent(runs, function(datum, index){
        return datum.distance;
    })
    yScale.domain(yDomain);

    d3.select('svg').selectAll('circle')
        .data(runs)
        .enter()
        .append('circle');

    d3.selectAll('circle')
        .attr('cy', function(datum, index){
            return yScale(datum.distance);
        });

    var parseTime = d3.timeParse("%B%e, %Y at %-I:%M%p");
    var formatTime = d3.timeFormat("%B%e, %Y at %-I:%M%p");
    var xScale = d3.scaleTime();
    xScale.range([0,WIDTH]);
    xDomain = d3.extent(runs, function(datum, index){
        return parseTime(datum.date);
    });
    xScale.domain(xDomain);

    d3.selectAll('circle')
        .attr('cx', function(datum, index){
            return xScale(parseTime(datum.date));
        });

}
render();

If you refresh the browser, you'll see an error in the console. This is because the bottomAxis and leftAxis use xScale and yScale which are now scoped to exist only inside the render() function. For future use, let's move the xScale and yScale out of the render function along with the code for creating the domains/ranges:

var parseTime = d3.timeParse("%B%e, %Y at %-I:%M%p");
var formatTime = d3.timeFormat("%B%e, %Y at %-I:%M%p");
var xScale = d3.scaleTime();
xScale.range([0,WIDTH]);
xDomain = d3.extent(runs, function(datum, index){
    return parseTime(datum.date);
});
xScale.domain(xDomain);

var yScale = d3.scaleLinear();
yScale.range([HEIGHT, 0]);
yDomain = d3.extent(runs, function(datum, index){
    return datum.distance;
})
yScale.domain(yDomain);
var render = function(){

    d3.select('svg').selectAll('circle') //since no circles exist, we need to select('svg') so that d3 knows where to append the new circles
        .data(runs) //attach the data as before
        .enter() //find the data objects that have not yet been attached to visual elements
        .append('circle'); //for each data object that hasn't been attached, append a <circle> to the <svg>

    d3.selectAll('circle')
        .attr('cy', function(datum, index){
            return yScale(datum.distance);
        });

    d3.selectAll('circle')
        .attr('cx', function(datum, index){
            return xScale(parseTime(datum.date)); //use parseTime to convert the date string property on the datum object to a Date object, which xScale then converts to a visual value
        });

}
render();

Now go to the bottom of app.js and add a line to call render() inside our <svg> click handler:

var newRun = { //create a new "run" object
    id: runs[runs.length-1].id+1, //generate a new id by adding 1 to the last run's id
    date: formatTime(date), //format the date object created above to a string
    distance: distance //add the distance
}
runs.push(newRun);
createTable();
render(); //add this line

Now when you click the SVG, a circle will appear:

Remove data

Let's set up a click handler on all <circle> elements so that when the user clicks on a <circle> D3 will remove that circle and its associated data element from the array. Add the following code at the bottom of the render function declaration we wrote in the last section. We do this so that the click handlers are attached AFTER the circles are created:

//put this at the bottom of the render function, so that click handlers are attached when the circle is created
d3.selectAll('circle').on('click', function(datum, index){
    d3.event.stopPropagation(); //stop click event from propagating to the SVG element and creating a run
    runs = runs.filter(function(run, index){ //create a new array that has removed the run with the correct id.  Set it to the runs var
        return run.id != datum.id;
    });
    render(); //re-render dots
    createTable(); //re-render table
});

Let's examine the above code. The first line selects all <circle> elements and creates a click handler on each of them. d3.event.stopPropagation(); prevents the our click from bubbling up the DOM to the SVG. If we don't add it, the click handler on the SVG will fire in addition, when we click on a circle. This would create an additional run every time we try to remove a run. Next we call:

runs = runs.filter(function(run, index){
    return run.id != datum.id;
});

This loops through the runs array and filters out any objects that have an id property that matches that of the id property of the datum that is associated with the <circle> that was clicked. Notice that the callback function in .on('click', function(datum, index){ takes two parameters: datum, the "run" object associated with that <circle> and the index of the that "run" object in the runs array.

Once we've filtered out the correct "run" object from the runs array, we call render() and createdTable() to re-render the the graph and the table.

But, if we click on the middle circle and examine the Elements tab of the dev tools, we'll see the <circle> element hasn't been removed:

In the image above, it appears as though there are only two circles, but really the middle one has had its cx set to 800 and its cy set to 0. It's overlapping the other circle in the same position. This is because we've removed the 2nd element in the runs array. When we re-render the graph, the runs array only has two objects. The 2nd "run" object used to be the third "run" object before we removed the the middle run. Now that it's the 2nd "run" object, the second <circle> is assigned its data. The third circle still has its old data assigned to it, so both the second and the third circle have the same data and are therefore placed in the same location.

Let's put the circles in a <g> so that it's easy to clear out all the circles and re-render them when we remove a run. This way we won't have any extra <circle> elements laying around when we try to remove them. This approach is similar to what we do when re-rendering the table. Adjust your <svg> element in index.html so it looks like this:

<svg>
    <g id="points"></g>
</svg>

Now we can clear out the <circle> elements each time render() is called. This is a little crude, but it'll work for now. Later on, we'll do things in a more elegant fashion. At the top of the render() function declaration, add d3.select('#points').html(''); and adjust the next line from d3.select('svg').selectAll('circle') to d3.select('#points').selectAll('circle'):

//adjust the code at the top of your render function
d3.select('#points').html(''); //clear out all circles when rendering
d3.select('#points').selectAll('circle') //add circles to #points group, not svg
    .data(runs)
    .enter()
    .append('circle');

Now if we click on the middle circle, the element is removed from the DOM:

If you try to delete all the circles and then add a new one, you'll get an error:

This is because, our code for creating a newRun in the SVG click handler needs some work:

var newRun = { //create a new "run" object
    id: runs[runs.length-1].id+1, //generate a new id by adding 1 to the last run's id
    date: formatTime(date), //format the date object created above to a string
    distance: distance //add the distance
}

This is because when there are no run elements in the runs array,runs[runs.length-1] tries to access an element at index -1 in the array. Inside the <svg> click handler, let's put in a little code to handle when the user has deleted all runs and tries to add a new one:

//inside svg click handler
var newRun = {
    id: ( runs.length > 0 ) ? runs[runs.length-1].id+1 : 1, //add this line
    date: formatTime(date),
    distance: distance
}

Here's what Chrome should look like now if you delete all the runs and then try to add a new one:

Lastly, let's put in some css, so we know we're clicking on a circle. First, add transition: r 0.5s linear, fill 0.5s linear; to the CSS code you've already written for circle:

circle {
    r: 5;
    fill: black;
    transition: r 0.5s linear, fill 0.5s linear; /* add this transition to original code */
}

then add this to the bottom of app.css:

/* add this css for the hover state */
circle:hover {
    r:10;
    fill: blue;
}

Here's what a circle should look like when you hover over it:

Drag an element

We want to be able to update the data for a run by dragging the associated circle. To do this, we'll use a "behavior," which you can think of as a combination of multiple event handlers. For a drag behavior, there are three callbacks:

when the user starts to drag
each time the user moves the cursor before releasing the "mouse" button
when the user releases the "mouse" button

There are two steps whenever we create a behavior:

create the behavior
attach the behavior to one or more elements

Put the following code at the bottom of the render() function declaration:

var drag = function(datum){
    var x = d3.event.x;
    var y = d3.event.y;
    d3.select(this).attr('cx', x);
    d3.select(this).attr('cy', y);
}
var dragBehavior = d3.drag()
    .on('drag', drag);
d3.selectAll('circle').call(dragBehavior);

You can now drag the circles around, but the data doesn't update:

Let's examine how this code works:

var drag = function(datum){
    var x = d3.event.x;
    var y = d3.event.y;
    d3.select(this).attr('cx', x);
    d3.select(this).attr('cy', y);
}

This drag function will be used as a callback anytime the user moves the cursor before releasing the "mouse" button. It gets the x and y coordinates of the mouse and sets the cx and cy values of the element being dragged (d3.select(this)) to those coordinates.

Next we generate a drag behavior that will, at the appropriate time, call the drag function that was just explained:

var dragBehavior = d3.drag()
    .on('drag', drag);

Lastly, we attach that behavior to all <circle> elements:

d3.selectAll('circle').call(dragBehavior);

Update data after a drag

Now we're going to add functionality so that when the user releases the "mouse" button, the data for the "run" object associated with the circle being dragged gets updated.

First lets create the callback function that will get called when the user releases the "mouse" button. Towards the bottom of the render() function declaration, add the following code just above var drag = function(datum){:

var dragEnd = function(datum){
    var x = d3.event.x;
    var y = d3.event.y;

    var date = xScale.invert(x);
    var distance = yScale.invert(y);

    datum.date = formatTime(date);
    datum.distance = distance;
    createTable();
}

Now attach that function to the dragBehavior so that it is called when the user stops dragging a circle. Change the following code:

var dragBehavior = d3.drag()
    .on('drag', drag);

to this:

var dragBehavior = d3.drag()
    .on('drag', drag)
    .on('end', dragEnd);

Now, once you stop dragging a circle around, you should see the data in the table change.

Let's change the color of a circle while it's being dragged too. Add this at the bottom of app.css:

circle:active {
    fill: red;
}

When you drag a circle, it should turn red

Create a zoom behavior that scales elements

Another behavior we can create is the zooming/panning ability. Once this functionality is complete you will be able to zoom in and out on different parts of the graph by doing one of the following:

two finger drag on a trackpad
rotate your mouse wheel
pinch/spread on a trackpad

You will also be able to pan left/right/up/down on the graph by clicking and dragging on the SVG element

Put the following code at the bottom of app.js:

var zoomCallback = function(){
	d3.select('#points').attr("transform", d3.event.transform);
}

This is the callback function that will be called when user attempts to zoom/pan. All it does is take the zoom/pan action and turn it into a transform attribute that gets applied to the <g id="points"></g> element that contains the circles. Now add the following code at the bottom of app.js to create the zoom behavior and attach it to the svg element:

var zoom = d3.zoom()
    .on('zoom', zoomCallback);
d3.select('svg').call(zoom);

Now if we zoom out, the graph should look something like this:

Update axes when zooming/panning

Now when we zoom, the points move in/out. When we pan, they move vertically/horizontally. Unfortunately, the axes don't update accordingly. Let's first add ids to the <g> elements that contain them. Find the following code:

var bottomAxis = d3.axisBottom(xScale);
d3.select('svg')
	.append('g')
	.call(bottomAxis)
    .attr('transform', 'translate(0,'+HEIGHT+')');

var leftAxis = d3.axisLeft(yScale);
d3.select('svg')
	.append('g')
	.call(leftAxis);

Add .attr('id', 'x-axis') after the first .append('g') and .attr('id', 'y-axis') after the second .append('g'):

d3.select('svg')
	.append('g')
    .attr('id', 'x-axis') //add an id
	.call(bottomAxis)
    .attr('transform', 'translate(0,'+HEIGHT+')');

var leftAxis = d3.axisLeft(yScale);
d3.select('svg')
	.append('g')
    .attr('id', 'y-axis') //add an id
	.call(leftAxis);

Now let's use those ids to adjust the axes when we zoom. Find this code:

var zoomCallback = function(){
	d3.select('#points').attr("transform", d3.event.transform);
}

Add the following at the end of it:

d3.select('#x-axis')
    .call(bottomAxis.scale(d3.event.transform.rescaleX(xScale)));
d3.select('#y-axis')
    .call(leftAxis.scale(d3.event.transform.rescaleY(yScale)));

Your zoomCallback should now look like this:

var zoomCallback = function(){
	d3.select('#points').attr("transform", d3.event.transform);
    d3.select('#x-axis')
        .call(bottomAxis.scale(d3.event.transform.rescaleX(xScale)));
    d3.select('#y-axis')
        .call(leftAxis.scale(d3.event.transform.rescaleY(yScale)));
}

Two things to note about the code above:

bottomAxis.scale() tells the axis to redraw itself
d3.event.transform.rescaleX(xScale) returns a value indicating how the bottom axis should rescale

Now when you zoom out, the axes should redraw themselves:

Update click points after a transform

If we zoom/pan and then click on the svg to create new runs, the circles/data created are incorrect
When we zoom, we need to save the transformation information to a variable so that we can use it later to figure out how to properly create circles and runs

var lastTransform = null; //reference to the last transform that happened
var zoomCallback = function(){
    lastTransform = d3.event.transform; //update the transform reference
	d3.select('#points').attr("transform", d3.event.transform);
    d3.select('#x-axis').call(bottomAxis.scale(d3.event.transform.rescaleX(xScale)));
	d3.select('#y-axis').call(leftAxis.scale(d3.event.transform.rescaleY(yScale)));
}

Now use that reference to the last transform when clicking on the svg:

d3.select('svg').on('click', function(){
    var x = lastTransform.invertX(d3.event.offsetX); //invert the transformation so we get a proper x value
    var y = lastTransform.invertY(d3.event.offsetY);

    var date = xScale.invert(x)
    var distance = yScale.invert(y);

    var newRun = {
        id: ( runs.length > 0 ) ? runs[runs.length-1].id+1 : 1,
        date: formatTime(date),
        distance: distance
    }
    runs.push(newRun);
    createTable();
    render();
});

But now clicking before any zoom is broken, since lastTransform will be null:

//at top of click handler, adjust the code:
//set x/y like normal
var x = d3.event.offsetX;
var y = d3.event.offsetY;

//if lastTransform has been updated, overwrite these values
if(lastTransform !== null){
    x = lastTransform.invertX(d3.event.offsetX);
    y = lastTransform.invertY(d3.event.offsetY);
}

var date = xScale.invert(x);
var distance = yScale.invert(y);

add a new run initially:

now pan right and add a new point:

Avoid redrawing entire screen during render

At the top of the render() function, assign the d3.select('#points').selectAll('circle').data(runs) to a variable, so we can use it later. This helps preserve how DOM elements are assigned to data elements in the next sections

//top of render function
d3.select('#points').html('');
var circles = d3.select('#points').selectAll('circle').data(runs); //alter this
circles.enter().append('circle'); //alter this

At the moment, we wipe all <circle> elements in the <svg> each time we call render()
- This is inefficient. Let's just remove the ones we don't want
We'll use .exit() to find the selection of circles that haven't been matched with data
- then we'll use .remove() to remove those circles

//top of render function
var circles = d3.select('#points').selectAll('circle').data(runs);
circles.enter().append('circle');
circles.exit().remove(); //remove all circles not associated with data

This can cause weird side effects, because some circles are being reassigned to a different set of data
- e.g. reload the page, click on the center circle. You'll notice the circle disappears and the one in the upper right gains a hover state
- if we remove a piece of data in the center of the array, the <circle> in the the DOM that was assigned to it gets reassigned to the piece of data that used to be assigned to the next sibling <circle> in the DOM. Each <circle> gets reassigned over one space from there on out
- to avoid these affects, we need to make sure that each circle stays with the data it used to be assigned to
  - to do this, we can tell D3 to map <circles> to datum by id, rather than index in the array

//when redrawing circles, make sure pre-existing circles match with their old data
var circles = d3.select('#points').selectAll('circle').data(runs, function(datum){
      return datum.id
});
circles.enter().append('circle');
circles.exit().remove();

Now clicking on the middle circle should work correctly

Hide elements beyond axis

Check out our graph when you pan:

To remove elements once they get beyond an axis, we can just add an outer SVG:

<svg id="container">
    <svg>
        <g id="points"></g>
    </svg>
</svg>

Now replace all d3.select('svg') with d3.select('#container')

d3.select('#container')
    .style('width', WIDTH)
    .style('height', HEIGHT);
//
// ...
//
d3.select('#container')
	.append('g')
    .attr('id', 'x-axis')
	.call(bottomAxis)
    .attr('transform', 'translate(0,'+HEIGHT+')');

var leftAxis = d3.axisLeft(yScale);
d3.select('#container')
	.append('g')
    .attr('id', 'y-axis')
	.call(leftAxis);
//
// ...
//
d3.select('#container').on('click', function(){
//
// ...
//
});
//
// ...
//
d3.select('#container').call(zoom);

And lastly, adjust css:

/* replace the rule for svg */
#container {
     overflow: visible;
     margin-bottom: 50px;
 }

Conclusion

In this chapter we've learned the basics of D3 and created a fully interactive scatter plot. In the next chapter we'll learn how to use AJAX to make an asynchronous request that will populate a bar graph.

48 KiB Raw Blame History