Chapter 05Stacks

When having multivariate data, it may be useful to show the data depicted as areas stacked on top of each other. Examples of this include stacked bar charts, stacked area graphs, and streamgraphs.

Consider the following array containing sales data for three fruits over three months.

var data = [
  {month: new Date(2018, 1, 1), apples: 10, bananas: 20, oranges: 15},
  {month: new Date(2018, 2, 1), apples: 15, bananas: 15, oranges: 15},
  {month: new Date(2018, 3, 1), apples: 20, bananas: 25, oranges: 15}
];

This dataset has a form that is typically visualized using a stacked chart or graph. The data for each fruit is referred to as a series. For example, the data for apples is the series (10, 15, 20). Each series can be displayed as an area, with one stacked on top of the other as shown in the visualization below. The number of objects in each series determines the width of the areas.

<script>
var data = [
 {month: new Date(2018, 1, 1), apples: 10, bananas: 20, oranges: 15},
 {month: new Date(2018, 2, 1), apples: 15, bananas: 15, oranges: 15},
 {month: new Date(2018, 3, 1), apples: 20, bananas: 25, oranges: 15}
];

var stackGen = d3.stack()
 .keys(["apples", "bananas", "oranges"]);
   
var stackedSeries = stackGen(data); 

var xScale = d3.scaleTime().domain([data[0].month, data[2].month]).range([50, 275]);
var yScale = d3.scaleLinear().domain([0,60]).range([275, 25]);

addAxis(d3.select("#demo0"), data, xScale, yScale, true); 

var colorScale = d3.scaleOrdinal()
 .domain(["apples", "bananas", "oranges"])
 .range(["red", "yellow", "orange"]);

var areaGen = d3.area()
 .x((d) => xScale(d.data.month))
 .y0((d) => yScale(d[0]))
 .y1((d) => yScale(d[1]));
   
d3.select("#demo0")
 .selectAll(".areas")
 .data(stackedSeries)
 .join("path")
 .attr("d", areaGen)
 .attr("fill", (d) => colorScale(d.key));
   
addLabels(d3.select("#demo0"), stackedSeries, areaGen);
</script>

<svg id="demo0" width= "300" height="300"></svg>
Figure 1. A stacked area graph.

In order to create stacked areas we need to compute cumulative y-values for each data point in each series. Let’s consider the stacked area graph shown above to see what we mean by cumulative values.

Notice that the visualization shows apples on the bottom, bananas in the middle, and oranges on top. Also notice that the bottom line of the apple area is the x-axis, and that the bottom line of the other two areas is the top line of the area below it. Therefore in order to define the areas we only need to compute the top lines of each area. Since the top lines are line segments drawn from points at each month mark on the x-axis, we only need to compute the value of the top line at these points.

Let’s start at the bottom of the graph and work our way up. Since the number of apples sold each month is depicted in the lower area, the y-coordinates of the points on the top line of the apple area are simply the values in the apples series. For the next layer, bananas, at each month point, the y-coordinate of the point on the top line of the bananas area is the y-coordinate of the point on the top line of the apples area plus the respective value in the bananas series. Similarly, at each time interval, the y-coordinate of the point on the top line of the oranges area is the y-coordinate of the point on the top line of the bananas area plus the respective value in the oranges series.

Thankfully, D3.js provides a stack generator that does this computation for us.

Creating Stacks

To create a stacked visualization using D3.js, we can use d3.stack, a method that returns a stack generator.

We may not want to create an area/layer for each series in the original dataset. Therefore, when using d3.stack, we need to specify which properties (series) in the dataset to use. We do so by using stack.keys. When we call d3.stack to create the generator, we’ll chain a call to stack.keys passing to it an array of the strings, where each string is the name of a property that we want to create a layer for.

For the dataset used in Figure 1, the keys are apples, bananas, and oranges, therefore to create a stack generator and assign the keys, we’ll do the following.

var stackGen = d3.stack()
  .keys(["apples", "bananas", "oranges"]);

After setting up the stack generator we invoke it, passing to it the dataset.

var stackedSeries = stackGen(data); 

When the stack generator is invoked with a dataset, it maps each series specified in stack.keys to a new series. Each new series contains the bottom and top y-coordinates for each data point in the original series, as well as reference to the object in the original dataset from which the top y-coordinate is computed.

The array referenced by stackedSeries holds the following data:

[
  [[0,  10, data[0]], [0,  15, data[1]], [0,  20, data[2]]],  //apples
  [[10, 30, data[0]], [15, 30, data[1]], [20, 45, data[2]]],  //bananas
  [[30, 45, data[0]], [30, 45, data[1]], [45, 60, data[2]]]   //oranges
]   

To render this data into stacked areas we need an area generator and some scales:

var xScale = d3.scaleTime()
  .domain([data[0].month, data[2].month])
  .range([50, 275]);
  
var yScale = d3.scaleLinear()
  .domain([0,60])
  .range([275, 25]);

var colorScale = d3.scaleOrdinal()
  .domain(["apples", "bananas", "oranges"])
  .range(["red", "yellow", "orange"]);

var areaGen = d3.area()
  .x((d) => xScale(d.data.month))
  .y0((d) => yScale(d[0]))
  .y1((d) => yScale(d[1]));

Now we can use the stackedSeries data with our areaGen to create multiple SVG paths which are filled with different colors.

d3.select("#demo1")
  .selectAll(".areas")
  .data(stackedSeries)
  .join("path")
  .attr("d", areaGen)
  .attr("fill", (d) => colorScale(d.key));
<script>
var data = [
 {month: new Date(2018, 1, 1), apples: 10, bananas: 20, oranges: 15},
 {month: new Date(2018, 2, 1), apples: 15, bananas: 15, oranges: 15},
 {month: new Date(2018, 3, 1), apples: 20, bananas: 25, oranges: 15}
];

var stackGen = d3.stack()
 .keys(["apples", "bananas", "oranges"]);
   
var stackedSeries = stackGen(data);

var xScale = d3.scaleTime().domain([data[0].month, data[2].month]).range([50, 275]);
var yScale = d3.scaleLinear().domain([0,60]).range([275, 25]);

var colorScale = d3.scaleOrdinal()
 .domain(["apples", "bananas", "oranges"])
 .range(["red", "yellow", "orange"]);

var areaGen = d3.area()
 .x((d) => xScale(d.data.month))
 .y0((d) => yScale(d[0]))
 .y1((d) => yScale(d[1]));
   
d3.select("#demo1")
 .selectAll(".areas")
 .data(stackedSeries)
 .join("path")
 .attr("d", areaGen)
 .attr("fill", (d) => colorScale(d.key));
   
</script>

<svg id="demo1" width= "300" height="300"></svg>
Figure 2. A basic stacked area graph.

Stack.value()

The stack.value method is used to retrieve the values associated with the keys from the data passed to the stack generator. By default, stack.value does not need to be called, as the keys are assumed to be properties of the objects in the array passed to the stack generator.

To see how stack.value can be used, consider a dataset formatted as follows:

var data = [
  {month: new Date(2018, 1, 1), fruitSales: {apples: 10, bananas: 20, oranges: 15}},
  ...

To create a stack generator with this dataset we call d3.stack and then set the keys to the names of the fruits as we did in the previous examples. We then call stack.value, passing it a lambda expression. The lambda expression has two parameters, obj and key, which correspond to an object in data and a key passed to stack.keys, respectively, and returns the value associated with the key in the object obj.

var stack = d3.stack()
  .keys(["apples", "bananas", "oranges"])
  .value((obj, key) => obj.fruitSales[key]);
<script>
var data = [
  {month: new Date(2018, 1, 1), fruitSales: {apples: 10, bananas: 20, oranges: 15}},
  {month: new Date(2018, 2, 1), fruitSales: {apples: 15, bananas: 15, oranges: 15}},
  {month: new Date(2018, 3, 1), fruitSales: {apples: 20, bananas: 25, oranges: 15}}
];

var stack = d3.stack()
  .keys(["apples", "bananas", "oranges"])
  .value((obj, key) => obj.fruitSales[key]);

var stackedSeries = stack(data);

var xScale = d3.scaleTime().domain([data[0].month, data[2].month]).range([50, 275]);
var yScale = d3.scaleLinear().domain([0,60]).range([275, 25]);

var colorScale = d3.scaleOrdinal()
  .domain(["apples", "bananas", "oranges"])
  .range(["red", "yellow", "orange"]);

var areaGen = d3.area()
  .x((d) => xScale(d.data.month))
  .y0((d) => yScale(d[0]))
  .y1((d) => yScale(d[1]));
    
d3.select("#demo3")
  .selectAll(".areas")
  .data(stackedSeries)
  .join("path")
  .attr("d", areaGen)
  .attr("fill", (d) => colorScale(d.key));
</script>

<svg id="demo3" width="300" height="300"></svg>
Figure 3. A stacked graph using stack.value().

Creating Bar Graphs

Instead of appending areas we can append SVG rects to create stacked bar graphs.

To do this, we will first bind each of the series in stackedSeries to a new g elements in our SVG. This gives us a g element for each fruit.

var sel = d3.select("#demo2")
  .select('g')
  .selectAll('g.series')
  .data(stackedSeries)
  .join('g')
  .classed('series', true)
  .style('fill', (d) => colorScale(d.key));

Recall that each series in stackedSeries is an array containing arrays where each inner array contains low and high y-coordinates.

Then, for each g element, we append multiple rect element, one for each pair of low and high y-coordinates.

sel.selectAll('rect')
  .data((d) => d)
  .join('rect')
  .attr('width', 40)
  .attr('y', (d) => yScale(d[1]))
  .attr('x', (d) => xScale(d.data.month) - 20)
  .attr('height', (d) => yScale(d[0]) -  yScale(d[1]));

Note that we use functions to generate our axis, labels, and areas in the example below and in many other examples on this page. The definitions of these functions can be found at the bottom of this page.

<script>
var data = [
  {month: new Date(2018, 1, 1), apples: 400, bananas: 200, cherries: 96, dates: 40},
  {month: new Date(2018, 2, 1), apples: 160, bananas: 150, cherries: 96, dates: 40},
  {month: new Date(2018, 3, 1), apples:  64, bananas:  96, cherries: 64, dates: 40},
  {month: new Date(2018, 4, 1), apples:  32, bananas:  48, cherries: 64, dates: 40}
];

var stack = d3.stack()
  .keys(["apples", "bananas", "cherries", "dates"]);
  
var stackedSeries = stack(data);

var xScale = d3.scaleTime().domain([data[0].month, data[3].month]).range([50,235]);
var yScale = d3.scaleLinear().domain([0, 650]).range([275,25]);
var colorScale = d3.scaleOrdinal()
  .domain(["apples", "bananas", "oranges", "cherries", "grapes", "dates"])
  .range(["red", "yellow", "orange", "pink", "purple", "brown"]);
        
//See end of page for addAxis() function definition
addAxis(d3.select("#demo2").append("g")
  .attr("transform", "translate(20,0)"), data, xScale, null, true);    //Adds in the X axis with ticks
addAxis(d3.select("#demo2"), data, null, yScale, true);                  //Adds in the Y axis
addAxis(d3.select("#demo2"), null, d3.scaleLinear().range([50,275]), null, true);  //Adds in a blank X axis
        
// Create a g element for each series
var sel = d3.select("#demo2")
  .select('g')
  .selectAll('g.series')
  .data(stackedSeries)
  .join('g')
  .classed('series', true)
  .style('fill', (d) => colorScale(d.key));

// For each series create a rect element for each month
sel.selectAll('rect')
  .data((d) => d)
  .join('rect')
  .attr('width', 40)
  .attr('y', (d) => yScale(d[1]))
  .attr('x', (d) => xScale(d.data.month) - 20)
  .attr('height', (d) => yScale(d[0]) -  yScale(d[1]));
</script>

<svg id="demo2" width="300" height="300"></svg>
Figure 4. A stacked bar graph.

Stack Ordering

By calling stack.order([order]) on a stack generator we can change the order of the series in the array that is returned by the stack generator. The default ordering is d3.stackOrderNone, which orders the series in the same order as that of the keys.

If the stack generator has a non-default ordering, we can change the ordering to the default by using d3.stackOrderNone.

If we want an ordering that is the reverse of the default, we can use d3.stackOrderReverse.

<script>
    var data = [
        {month: new Date(2018, 0, 1), fruitSales: {apples: 400, bananas: 200, cherries: 96,  dates: 40, oranges: 250, grapes: 20}},
        {month: new Date(2018, 1, 1), fruitSales: {apples: 160, bananas: 150, cherries: 96,  dates: 40, oranges: 200, grapes: 25}},
        {month: new Date(2018, 2, 1), fruitSales: {apples:  64, bananas:  96, cherries: 64,  dates: 40, oranges: 150, grapes: 30}},
        {month: new Date(2018, 3, 1), fruitSales: {apples:  32, bananas:  48, cherries: 64,  dates: 40, oranges: 100, grapes: 20}},
        {month: new Date(2018, 4, 1), fruitSales: {apples:  40, bananas: 100, cherries: 64,  dates: 40, oranges: 115, grapes: 45}},
        {month: new Date(2018, 5, 1), fruitSales: {apples: 100, bananas: 250, cherries: 86,  dates: 40, oranges: 225, grapes: 50}},
        {month: new Date(2018, 6, 1), fruitSales: {apples: 150, bananas: 125, cherries: 96,  dates: 40, oranges: 200, grapes: 15}},
        {month: new Date(2018, 7, 1), fruitSales: {apples: 100, bananas:  75, cherries: 106, dates: 40, oranges: 210, grapes: 10}}
    ];

    var xScale = d3.scaleTime().domain([data[0].month, data[data.length - 1].month]).range([50,275]);
    var yScale = d3.scaleLinear().domain([0,1000]).range([275,25]);
    var colorScale = d3.scaleOrdinal()
        .domain(["apples", "bananas", "cherries", "dates", "oranges", "grapes"])
        .range(["red", "yellow", "pink", "brown", "orange", "purple"]);

    var stack1 = d3.stack() 
            .keys(["apples", "bananas", "cherries", "dates", "oranges", "grapes"])
            .value((d, key) => d.fruitSales[key])
            .order(d3.stackOrderNone)
            .offset(d3.stackOffsetNone);
            
    var stackedSeries1 = stack1(data);

    var stack2 = d3.stack()
            .keys(["apples", "bananas", "cherries", "dates", "oranges", "grapes"])
            .value((d, key) => d.fruitSales[key])
            .order(d3.stackOrderReverse)
            .offset(d3.stackOffsetNone);
    var stackedSeries2 = stack2(data);

    var area = d3.area()
        .x((d) => xScale(d.data.month))
        .y0((d) => yScale(d[0]))
        .y1((d) => yScale(d[1]))
        .curve(d3.curveBasis);
                
    addAreas(d3.select("#demo4n"), stackedSeries1, area); // Areas to stackOrderNone
    addAreas(d3.select("#demo4r"), stackedSeries2, area); // Areas to stackOrderReverse
     
    addLabels(d3.select("#demo4n"), stackedSeries1, area) // Labels to stackOrderNone
    addLabels(d3.select("#demo4r"), stackedSeries2, area) // Labels to stackOrderReverse
</script>

<svg id="demo4n" width="300" height="300"></svg>
<svg id="demo4r" width="300" height="300"></svg>
Figure 5. Stacks using d3.stackOffsetNone (left) and d3.stackOrderReverse (right).

Ordering based on the sum of the values in each series

D3.js provides two orderings that are based on the sum of the values in each series.

<script>
        var data = [
            {month: new Date(2018, 0, 1), fruitSales: {apples: 400, bananas: 200, cherries: 96,  dates: 40, oranges: 250, grapes: 20}},
            {month: new Date(2018, 1, 1), fruitSales: {apples: 160, bananas: 150, cherries: 96,  dates: 40, oranges: 200, grapes: 25}},
            {month: new Date(2018, 2, 1), fruitSales: {apples:  64, bananas:  96, cherries: 64,  dates: 40, oranges: 150, grapes: 30}},
            {month: new Date(2018, 3, 1), fruitSales: {apples:  32, bananas:  48, cherries: 64,  dates: 40, oranges: 100, grapes: 20}},
            {month: new Date(2018, 4, 1), fruitSales: {apples:  40, bananas: 100, cherries: 64,  dates: 40, oranges: 115, grapes: 45}},
            {month: new Date(2018, 5, 1), fruitSales: {apples: 100, bananas: 250, cherries: 86,  dates: 40, oranges: 225, grapes: 50}},
            {month: new Date(2018, 6, 1), fruitSales: {apples: 100, bananas: 125, cherries: 96,  dates: 40, oranges: 200, grapes: 15}},
            {month: new Date(2018, 7, 1), fruitSales: {apples: 100, bananas:  75, cherries: 106, dates: 40, oranges: 210, grapes: 10}}
        ];
       
        var xScale = d3.scaleTime().domain([data[0].month,data[data.length -1].month]).range([50,275]);
        var yScale = d3.scaleLinear().domain([0,1000]).range([275,25]);
        var colorScale = d3.scaleOrdinal()
            .domain(["apples", "bananas", "cherries", "dates", "oranges", "grapes"])
            .range(["red", "yellow", "pink", "brown", "orange", "purple"]);
            
        var stack1 = d3.stack()
                .keys(["apples", "bananas", "cherries", "dates", "oranges", "grapes"])
                .value((d, key) => d.fruitSales[key])
                .order(d3.stackOrderAscending)
                .offset(d3.stackOffsetNone);
        var stackedSeries1 = stack1(data);

        var stack2 = d3.stack()
                .keys(["apples", "bananas", "cherries", "dates", "oranges", "grapes"])
                .value((d, key) => d.fruitSales[key])
                .order(d3.stackOrderDescending)
                .offset(d3.stackOffsetNone);
        var stackedSeries2 = stack2(data);

        var area = d3.area()
            .x((d) => xScale(d.data.month))
            .y0((d) => yScale(d[0]))
            .y1((d) => yScale(d[1]))
            .curve(d3.curveBasis);
        
        addAreas(d3.select("#demo5a"), stackedSeries1, area); // Areas to stackOrderAscending
        addAreas(d3.select("#demo5d"), stackedSeries2, area); // Areas to stackOrderDescending 
        
        addLabels(d3.select("#demo5a"), stackedSeries1, area); // Add Labels to stackOrderAcsending
        addLabels(d3.select("#demo5d"), stackedSeries2, area); // Add Labels to stackOrderDescending
   </script>

   <svg id="demo5a" width="300" height="300"></svg>
   <svg id="demo5d" width="300" height="300"></svg>
Figure 6. Stacks using d3.stackOrderAscending (left) and d3.stackOrderDescending (right).

Ordering based on the maximum value of each series

When creating stacks to visualize large amounts of data, readability is important. d3.stackOrderAppearance and d3.stackOrderInsideOut can be used to improve readability of a stack visualization by ordering the series based on the maximum value of each series or more specifically, the index of the maximum value for each series.

Take for example the dataset shown below containing sales data for 4 fruits over 14 months. When d3.stackOrderAppearance and d3.stackOrderInsideOut order the fruit series, they find the index of the maximum value for each series. We’ve highlighted these indices and values in the table below.

Index Apples Bananas Oranges Grapes
01111
110111
225111
310111
411110
511125
611110
711101
811251
911101
1011011
1112511
1211011
131111
Index at
Max Value
Apples: 2 Bananas: 11 Oranges: 8 Grapes: 5

Next, the orderings create an array containing the indices of the maximum values.

Apples Bananas Oranges Grapes
Indices 2 11 8 5

Finally, the series are sorted according to the indices of their maximum values.

Here we see that the series are sorted in the following order: apples, grapes, oranges, and bananas.

Apples Grapes Oranges Bananas
Indices 2 5 8 11
------------------------------------------------------->
<script>
    var data = [
   		{month: new Date(2018, 0, 1),  fruitSales: {apples: 1,    bananas: 1,   oranges: 1,   grapes: 1}},
        {month: new Date(2018, 1, 1),  fruitSales: {apples: 10,   bananas: 1,   oranges: 1,   grapes: 1}},
        {month: new Date(2018, 2, 1),  fruitSales: {apples: 25,   bananas: 1,   oranges: 1,   grapes: 1}},
        {month: new Date(2018, 3, 1),  fruitSales: {apples: 10,   bananas: 1,   oranges: 1,   grapes: 1}},
        {month: new Date(2018, 4, 1),  fruitSales: {apples: 1,    bananas: 1,  oranges: 1,   grapes: 10}},
        {month: new Date(2018, 5, 1),  fruitSales: {apples: 1,    bananas: 1,  oranges: 1,   grapes: 25}},
        {month: new Date(2018, 6, 1),  fruitSales: {apples: 1,    bananas: 1,  oranges: 1,   grapes: 10}},
        {month: new Date(2018, 7, 1),  fruitSales: {apples: 1,    bananas: 1,   oranges: 10,  grapes: 1}},
        {month: new Date(2018, 8, 1),  fruitSales: {apples: 1,    bananas: 1,   oranges: 25,  grapes: 1}},
        {month: new Date(2018, 9, 1),  fruitSales: {apples: 1,    bananas: 1,   oranges: 10,  grapes: 1}},
        {month: new Date(2018, 10, 1), fruitSales: {apples: 1,    bananas: 10,   oranges: 1,   grapes: 1}},
        {month: new Date(2018, 11, 1), fruitSales: {apples: 1,    bananas: 25,   oranges: 1,   grapes: 1}},
        {month: new Date(2019, 0, 1),  fruitSales: {apples: 1,    bananas: 10,   oranges: 1,   grapes: 1}},
        {month: new Date(2019, 1, 1),  fruitSales: {apples: 1,    bananas: 1,   oranges: 1,   grapes: 1}}
    ];

    var xScale = d3.scaleLinear().domain([data[0].month, data[data.length-1].month]).range([10,290]);
    var yScale = d3.scaleLinear().domain([0,30]).range([175,25]);
    var colorScale = d3.scaleOrdinal()
        .domain(["apples", "bananas", "oranges", "grapes"])
        .range(["red", "yellow", "orange", "purple"]);

    var stack1 = d3.stack()
            .keys(["apples", "bananas", "oranges", "grapes"])
            .value((d, key) => d.fruitSales[key]);
            
    var stackedSeries1 = stack1(data);

    var stack2 = d3.stack()
            .keys(["apples", "bananas", "oranges", "grapes"])
            .value((d, key) => d.fruitSales[key])
            .order(d3.stackOrderAppearance);
            
    var stackedSeries2 = stack2(data);

    var area = d3.area()
        .x((d) => xScale(d.data.month))
        .y0((d) => yScale(d[0]))
        .y1((d) => yScale(d[1]))
        .curve(d3.curveBasis);

    addAreas(d3.select("#demo6"), stackedSeries1, area);  // Areas to default order
    addAreas(d3.select("#demo6a"), stackedSeries2, area); // Areas to stackOrderAppearance
        
    addLabels(d3.select("#demo6"), stackedSeries1, area); // Labels to default order
    addLabels(d3.select("#demo6a"), stackedSeries2, area); // Labels to stackOrderAppearance
</script>

<svg id="demo6" class="svgClass" width="300" height="200"></svg>
<svg id="demo6a" class="svgClass" width="300" height="200"></svg>
Figure 7. Stacks using the default order (left) and d3.stackOrderAppearance (right).

Here we see that the series are sorted in the following order: oranges, apples, grapes, and bananas.

Oranges Apples Grapes Bananas
Indices 8 2 5 11
<-------------------------- -------------------------->
<script>
    var data = [
   		{month: new Date(2018, 0, 1),  fruitSales: {apples: 1,    bananas: 1,   oranges: 1,   grapes: 1}},
        {month: new Date(2018, 1, 1),  fruitSales: {apples: 10,   bananas: 1,   oranges: 1,   grapes: 1}},
        {month: new Date(2018, 2, 1),  fruitSales: {apples: 25,   bananas: 1,   oranges: 1,   grapes: 1}},
        {month: new Date(2018, 3, 1),  fruitSales: {apples: 10,   bananas: 1,   oranges: 1,   grapes: 1}},
        {month: new Date(2018, 4, 1),  fruitSales: {apples: 1,    bananas: 1,  oranges: 1,   grapes: 10}},
        {month: new Date(2018, 5, 1),  fruitSales: {apples: 1,    bananas: 1,  oranges: 1,   grapes: 25}},
        {month: new Date(2018, 6, 1),  fruitSales: {apples: 1,    bananas: 1,  oranges: 1,   grapes: 10}},
        {month: new Date(2018, 7, 1),  fruitSales: {apples: 1,    bananas: 1,   oranges: 10,  grapes: 1}},
        {month: new Date(2018, 8, 1),  fruitSales: {apples: 1,    bananas: 1,   oranges: 25,  grapes: 1}},
        {month: new Date(2018, 9, 1),  fruitSales: {apples: 1,    bananas: 1,   oranges: 10,  grapes: 1}},
        {month: new Date(2018, 10, 1), fruitSales: {apples: 1,    bananas: 10,   oranges: 1,   grapes: 1}},
        {month: new Date(2018, 11, 1), fruitSales: {apples: 1,    bananas: 25,   oranges: 1,   grapes: 1}},
        {month: new Date(2019, 0, 1),  fruitSales: {apples: 1,    bananas: 10,   oranges: 1,   grapes: 1}},
        {month: new Date(2019, 1, 1),  fruitSales: {apples: 1,    bananas: 1,   oranges: 1,   grapes: 1}}
    ];

    var xScale = d3.scaleLinear().domain([data[0].month, data[data.length-1].month]).range([10,290]);
    var yScale = d3.scaleLinear().domain([0,30]).range([175,25]);
    var colorScale = d3.scaleOrdinal()
        .domain(["apples", "bananas", "oranges", "grapes"])
        .range(["red", "yellow", "orange", "purple"]);

    var stack1 = d3.stack()
            .keys(["apples", "bananas", "oranges", "grapes"])
            .value((d, key) => d.fruitSales[key]);
            
    var stackedSeries1 = stack1(data);

    var stack2 = d3.stack()
            .keys(["apples", "bananas", "oranges", "grapes"])
            .value((d, key) => d.fruitSales[key])
            .order(d3.stackOrderInsideOut)
            .offset(d3.stackOffsetWiggle);
            
    var stackedSeries2 = stack2(data);

    var area1 = d3.area()
        .x((d) => xScale(d.data.month))
        .y0((d) => yScale(d[0]))
        .y1((d) => yScale(d[1]))
        .curve(d3.curveBasis);
        
    var area2 = d3.area()
        .x((d) => xScale(d.data.month))
        .y0((d) => yScale(d[0] + 25/2))
        .y1((d) => yScale(d[1] + 25/2))
        .curve(d3.curveBasis);

    addAreas(d3.select("#demo6b"), stackedSeries1, area1); // Areas to stackOrderAppearance
    addAreas(d3.select("#demo6i"), stackedSeries2, area2); // Areas to stackOrderInsideOut
        
    addLabels(d3.select("#demo6b"), stackedSeries1, area1); // Labels to stackOrderAppearance
    addLabels(d3.select("#demo6i"), stackedSeries2, area2); // Labels to stackOrderInsideOut
</script>

<svg id="demo6b" class="svgClass" width="300" height="200"></svg>
<svg id="demo6i" class="svgClass" width="300" height="200"></svg>
Figure 8. Stacks using the default order (left) and d3.stackOrderInsideOut (right).

Note that d3.stackOrderInsideOut can be used to make steamgraphs as we do in Figure 8. When doing so we also set the offset to d3.stackOffsetWiggle. For more information, please see the section on stack offsets below.

For more information on the orderings used by d3.stackOrderAppearance and d3.stackOrderInsideOut please see Stacked Graphs—Geometry & Aesthetics by Byron & Wattenberg.

Stack Offsets

The baseline of a stack is the line y=0. By default, the series are positioned above the baseline with the first series’ lower y-coordinates set at the baseline. Often, we’ll want to change how the series are positioned relative to the baseline. For example, we may want positive data points positioned above the baseline and lower data points positioned below the baseline.

To change how the series are positioned relative to the baseline we call stack.offset and pass to it an offset function.

If the stack generator has an offset that is not the default, we can change it to the default using d3.stackOffsetNone.

<script>
var data = [
    {month: new Date(2018, 0, 1), fruitSales: {apples: 400, bananas: 200, cherries: 96,  dates: 40, oranges: 250, grapes: 20}},
    {month: new Date(2018, 1, 1), fruitSales: {apples: 160, bananas: 150, cherries: 96,  dates: 40, oranges: 200, grapes: 25}},
    {month: new Date(2018, 2, 1), fruitSales: {apples:  64, bananas:  96, cherries: 64,  dates: 40, oranges: 150, grapes: 30}},
    {month: new Date(2018, 3, 1), fruitSales: {apples:  32, bananas:  48, cherries: 64,  dates: 40, oranges: 100, grapes: 20}},
    {month: new Date(2018, 4, 1), fruitSales: {apples:  40, bananas: 100, cherries: 64,  dates: 40, oranges: 115, grapes: 45}},
    {month: new Date(2018, 5, 1), fruitSales: {apples: 100, bananas: 250, cherries: 86,  dates: 40, oranges: 225, grapes: 50}},
    {month: new Date(2018, 6, 1), fruitSales: {apples: 150, bananas: 125, cherries: 96,  dates: 40, oranges: 200, grapes: 15}},
    {month: new Date(2018, 7, 1), fruitSales: {apples: 100, bananas:  75, cherries: 106, dates: 40, oranges: 210, grapes: 10}}
];

var xScale = d3.scaleTime().domain([data[0].month, data[data.length - 1].month]).range([50,275]);
var yScaleNone = d3.scaleLinear().domain([0,1000]).range([275,25]);
var yScaleExpand = d3.scaleLinear().domain([0,1]).range([275,25]);
var colorScale = d3.scaleOrdinal()
  .domain(["apples", "bananas", "cherries", "dates", "oranges", "grapes"])
  .range(["red", "yellow", "pink", "brown", "orange", "purple"]);

var stack = d3.stack() 
  .keys(["apples", "bananas", "cherries", "dates", "oranges", "grapes"])
  .value((d, key) => d.fruitSales[key])
  .order(d3.stackOrderNone)
  .offset(d3.stackOffsetNone);
        
var stackedSeries = stack(data);

var areaNone = d3.area()
  .x((d) => xScale(d.data.month))
  .y0((d) => yScaleNone(d[0]))
  .y1((d) => yScaleNone(d[1]))
  .curve(d3.curveBasis);

addAreas(d3.select("#demo7").select("#stack"), stackedSeries, areaNone);
addLabels(d3.select("#demo7").select("#stack"), stackedSeries, areaNone);
</script>

<svg id="demo7" width="300" height="300">
  <g id="stack"></g>
  <g id="baseline">
      <text y="290">Baseline</text>
      <path d="M 25 275 l 325 0" stroke="black" stroke-dasharray="10,5" stroke-width="2px"></path>
  </g>
</svg>
Figure 9. Stack with the default offset.

If we wish to position all positive data points above the baseline and negative data points below the baseline we can use d3.stackOffsetDiverging.

<script>
var data = [
    {month: new Date(2018, 0, 1), fruitSales: {apples: 400, bananas: 200, cherries: 96,  dates: 40, oranges: 250, grapes: 20}},
    {month: new Date(2018, 1, 1), fruitSales: {apples: 160, bananas: 150, cherries: 96,  dates: 40, oranges: 200, grapes: 25}},
    {month: new Date(2018, 2, 1), fruitSales: {apples:  64, bananas:  96, cherries: 64,  dates: 40, oranges: 150, grapes: -30}},
    {month: new Date(2018, 3, 1), fruitSales: {apples:  32, bananas:  48, cherries: 64,  dates: 40, oranges: 100, grapes: -20}},
    {month: new Date(2018, 4, 1), fruitSales: {apples:  40, bananas: 100, cherries: 64,  dates: 40, oranges: 115, grapes: -45}},
    {month: new Date(2018, 5, 1), fruitSales: {apples: 100, bananas: -25, cherries: 86,  dates: 40, oranges: -225, grapes: 50}},
    {month: new Date(2018, 6, 1), fruitSales: {apples: 150, bananas: -125, cherries: 96,  dates: 40, oranges: -200, grapes: 15}},
    {month: new Date(2018, 7, 1), fruitSales: {apples: 100, bananas:  -75, cherries: 106, dates: 40, oranges: -210, grapes: 10}}
];

var xScale = d3.scaleTime().domain([data[0].month, data[data.length - 1].month]).range([50+225/data.length/2, 275 - 225/data.length/2]);
var yScaleDiv = d3.scaleLinear().domain([-1000,1000]).range([275,25]);
var colorScale = d3.scaleOrdinal()
  .domain(["apples", "bananas", "cherries", "dates", "oranges", "grapes"])
  .range(["red", "yellow", "pink", "brown", "orange", "purple"]);

var stack = d3.stack() 
  .keys(["apples", "bananas", "cherries", "dates", "oranges", "grapes"])
  .value((d, key) => d.fruitSales[key])
  .order(d3.stackOrderNone)
  .offset(d3.stackOffsetDiverging);
  
var stackedSeries = stack(data);

addAxis(d3.select("#demo8d"), data, xScale, yScaleDiv);
addAxis(d3.select("#demo8d"), null, d3.scaleTime().range([50,275]), null);

// Create a g element for each series
var g = d3.select("#demo8d")
  .select('#stack')
  .selectAll('g.series')
  .data(stackedSeries)
  .join('g')
  .classed('series', true)
  .style('fill', (d) => colorScale(d.key));

// For each series create a rect element for each month
g.selectAll('rect')
  .data((d) => d)
  .join('rect')
  .attr('width', 225/data.length)
  .attr('y', (d) => yScaleDiv(d[1]))
  .attr('x', (d, i) => i * (225/data.length) + 50)
  .attr('height', (d) => yScaleDiv(d[0]) -  yScaleDiv(d[1])); 
    
//Adds in the baseline
d3.select("#baseline8")
  .append("path")
  .attr("d", "M 25 " + yScaleDiv(0) + " l 325 0")
  .attr("stroke", "black")
  .attr("stroke-dasharray", "10,5")
  .attr("stroke-width", "2px");
  
d3.select("#baseline8")
  .append("text")
  .attr("x", 0)
  .attr("y", yScaleDiv(0) - 10)
  .text("Baseline"); 
</script>

<svg id="demo8d" width="300" height="300">
    <g id="stack"></g>
    <g id="baseline8"></g>
</svg>
Figure 10. A bar graph using d3.stackOffsetDiverging.

The d3.stackOffsetSilhouette and d3.stackOffsetWiggle offset functions create streamgraphs.

<script>
var data = [
    {month: new Date(2018, 0, 1), fruitSales: {apples: 400, bananas: 200, cherries: 96,  dates: 40, oranges: 250, grapes: 20}},
    {month: new Date(2018, 1, 1), fruitSales: {apples: 160, bananas: 150, cherries: 96,  dates: 40, oranges: 200, grapes: 25}},
    {month: new Date(2018, 2, 1), fruitSales: {apples:  64, bananas:  96, cherries: 64,  dates: 40, oranges: 150, grapes: 30}},
    {month: new Date(2018, 3, 1), fruitSales: {apples:  32, bananas:  48, cherries: 64,  dates: 40, oranges: 100, grapes: 20}},
    {month: new Date(2018, 4, 1), fruitSales: {apples:  40, bananas: 100, cherries: 64,  dates: 40, oranges: 115, grapes: 45}},
    {month: new Date(2018, 5, 1), fruitSales: {apples: 100, bananas: 250 , cherries: 86,  dates: 40, oranges: 225, grapes: 50}},
    {month: new Date(2018, 6, 1), fruitSales: {apples: 150, bananas: 125, cherries: 96,  dates: 40, oranges: 200, grapes: 15}},
    {month: new Date(2018, 7, 1), fruitSales: {apples: 100, bananas:  75, cherries: 106, dates: 40, oranges: 210, grapes: 10}}
];

var xScale = d3.scaleTime().domain([data[0].month, data[data.length - 1].month]).range([50,275]);
var yScale = d3.scaleLinear().domain([0,1000]).range([275,25]);
var colorScale = d3.scaleOrdinal()
  .domain(["apples", "bananas", "cherries", "dates", "oranges", "grapes"])
  .range(["red", "yellow", "pink", "brown", "orange", "purple"]);

var stack1 = d3.stack() 
  .keys(["apples", "bananas", "cherries", "dates", "oranges", "grapes"])
  .value((d, key) => d.fruitSales[key])
  .order(d3.stackOrderNone)
  .offset(d3.stackOffsetSilhouette);
        
var stackedSeries1 = stack1(data);

var stack2 = d3.stack()
  .keys(["apples", "bananas", "cherries", "dates", "oranges", "grapes"])
  .value((d, key) => d.fruitSales[key])
  .order(d3.stackOrderInsideOut)
  .offset(d3.stackOffsetWiggle);
  
var stackedSeries2 = stack2(data);

var area = d3.area()
  .x((d) => xScale(d.data.month))
  .y0((d) => yScale(d[0]))
  .y1((d) => yScale(d[1]))
  .curve(d3.curveBasis);

addAreas(d3.select("#demo9s")
  .select("#stack9s"), stackedSeries1, area, "translate(0, -150)"); // Areas to stackOffsetSilhouette
  
addAreas(d3.select("#demo9w")
  .select("#stack9w"), stackedSeries2, area); // Areas to stackOrderReverse // Areas to stackOffsetWiggle

addLabels(d3.select("#demo9s")
  .select("#stack9s")
  .append("g")
  .attr("transform", "translate(0, -150)"), stackedSeries1, area); 

addLabels(d3.select("#demo9w")
  .select("#stack9w"), stackedSeries2, area);

//Adds the baseline the Silhouette
d3.select("#baseline9s")
  .append("path")
  .attr("d", "M 25 " + yScale(0) + " l 325 0")
  .attr("stroke-dasharray", "10,5")
  .attr("stroke", "black")
  .attr("stroke-width", "2px");

d3.select("#baseline9s")
  .append("text")
  .attr("x", 0)
  .attr("y", yScale(0) - 10)
  .text("Baseline");
    
  d3.select("#baseline9w")
  .append("path")
  .attr("d", "M 25 " + yScale(0) + " l 325 0")
  .attr("stroke-dasharray", "10,5")
  .attr("stroke", "black")
  .attr("stroke-width", "2px");
  
  d3.select("#baseline9w")
  .append("text")
  .attr("x", 0)
  .attr("y", yScale(0) - 10)
  .text("Baseline");
</script>

<svg id="demo9s" width="300" height="300">
  <g id="stack9s"></g>
  <g id="baseline9s" transform="translate(0,-150)")></g>
</svg>
<svg id="demo9w" width="300" height="300">
  <g id="stack9w"></g>
  <g id="baseline9w")></g>
</svg>
Figure 11. Stacks using d3.stackOffsetSilhouette (left) and d3.stackOffsetWiggle (right).

The last offset function is d3.stackOffsetExpand which positions the series so that the bottom series is bounded below by y=0 and the top series is bounded above by y=1.

<script>
var data = [
    {month: new Date(2018, 0, 1), fruitSales: {apples: 400, bananas: 200, cherries: 96,  dates: 40, oranges: 250, grapes: 20}},
    {month: new Date(2018, 1, 1), fruitSales: {apples: 160, bananas: 150, cherries: 96,  dates: 40, oranges: 200, grapes: 25}},
    {month: new Date(2018, 2, 1), fruitSales: {apples:  64, bananas:  96, cherries: 64,  dates: 40, oranges: 150, grapes: 30}},
    {month: new Date(2018, 3, 1), fruitSales: {apples:  32, bananas:  48, cherries: 64,  dates: 40, oranges: 100, grapes: 20}},
    {month: new Date(2018, 4, 1), fruitSales: {apples:  40, bananas: 100, cherries: 64,  dates: 40, oranges: 115, grapes: 45}},
    {month: new Date(2018, 5, 1), fruitSales: {apples: 100, bananas: 250, cherries: 86,  dates: 40, oranges: 225, grapes: 50}},
    {month: new Date(2018, 6, 1), fruitSales: {apples: 150, bananas: 125, cherries: 96,  dates: 40, oranges: 200, grapes: 15}},
    {month: new Date(2018, 7, 1), fruitSales: {apples: 100, bananas:  75, cherries: 106, dates: 40, oranges: 210, grapes: 10}}
];

var xScale = d3.scaleTime().domain([data[0].month, data[data.length - 1].month]).range([50,275]);
var yScaleNone = d3.scaleLinear().domain([0,1000]).range([275,25]);
var yScaleExpand = d3.scaleLinear().domain([0,1]).range([275,25]);
var colorScale = d3.scaleOrdinal()
  .domain(["apples", "bananas", "cherries", "dates", "oranges", "grapes"])
  .range(["red", "yellow", "pink", "brown", "orange", "purple"]);

var stack2 = d3.stack()
  .keys(["apples", "bananas", "cherries", "dates", "oranges", "grapes"])
  .value((d, key) => d.fruitSales[key])
  .order(d3.stackOrderNone)
  .offset(d3.stackOffsetExpand);
        
var stackedSeries2 = stack2(data);
    
var areaExpanding = d3.area()
  .x((d) => xScale(d.data.month))
  .y0((d) => yScaleExpand(d[0]))
  .y1((d) => yScaleExpand(d[1]))
  .curve(d3.curveBasis);

addAreas(d3.select("#demo12").select("#stack"), stackedSeries2, areaExpanding);
addLabels(d3.select("#demo12").select("#stack"), stackedSeries2, areaExpanding);
</script>

<svg id="demo12" width="300" height="300">
  <g id="stack"></g>
  <g id="baseline">
  		<text x="" y="290">Baseline</text>
        <text x="290" y="270">0</text>
        <text x="290" y="40">1</text>
        <path d="M 25 25 l 325 0" stroke="black" stroke-dasharray="10,5" stroke-width="2px"></path>
        <path d="M 25 275 l 325 0" stroke="black" stroke-dasharray="10,5" stroke-width="2px"></path>
  </g>
</svg>
Figure 12. A stack using d3.stackOffsetExpand.