d3js grouped bar chart, is this possible? - d3.js

I want to create a grouped bar chart where each bar is unique and not part of a series.
For example, imagine a bar chart showing the population of each major city, grouped by state.
Is this possible with d3js and any pointers on how to to get started?
Thanks

Yes it is possible!
As with most visualisation the process is one of working out the structure of the chart and then working out how to get the data in to that structure. In this case I think the data in it's simplest form is going to look something like this ...
let groups = [
{ name:'one', values:[1,3,6,2] },
{ name:'two', values:[3,5,7,3,2,5] },
{ name:'three', values:[9,2,5] },
{ name:'four', values:[6] }
];
An array of groups, where each group has some properties including an array of values to be represented as individual bars.
The tricky bit is working out the position of each group and each bar along the horizontal axis. The way I think about this is as follows: If we say each bar has a width of 1 unit, how much space do we want between the groups? Maybe a bar and a half, so 1.5 units. Now each group is going to take up space equal to the number of bars in it (the length of the values array) and the total width required by the chart will be the sum of those values plus the spaces between them.
i.e.
let dataWidth = d3.sum(groups, d=>d.values.length) + (groups.length-1) * groupPadding;
We also want to go through the groups and work out their "start position", where the group is placed horizontally in terms of bar units, like this:
let groupPadding = 1.5;
let currentWidth = 0;
groups = groups.map(group=>{
group.width = group.values.length;
group.startPosition = currentWidth;
currentWidth += group.width+groupPadding;
return group;
});
Next we need to make an x and a y-scale using d3-scale.
For the y-scale let's keep it simple and hard code the maximum value
const yScale = d3.scaleLinear()
.domain([0,10])
.range([chartHeight, 0]); // chart height is just your charts height in pixels
and for the x scale we use the dataWidth we calculated above as the maximum for the domain, we'll use this scale to convert bar units into screen pixels
const xScale = d3.scaleLinear()
.domain([0, barWidth)
.range([0, chartWidth]); //chartWidth, the width of your chart in pixels
OK. Now we have everything we need to know to draw the chart. The structure of the chart as I see envisage it...
Each group in groups is an svg g element positioned according to its startPosition. Each of those g elements contains a set of rect elements, one for each value. Inside the g element you may also want to put stuff like a group label.
Broadly this would look something like this...
Create the groups
const barGroups = chart.selectAll('g.bar-group')
.data(groups)
.enter()
.append('g')
.classed('bar-group', true)
.attr('transform', group=>`translate(${xScale(group.startPosition)}, 0)`);
Add the rectangles
barGroups.each(function(group){
const barGroup = d3.select(this);
barGroup.selectAll('rect')
.data(group.values)
.enter()
.append('rect')
.attr('width', xScale(1))
.attr('height', value => yScale(value))
.attr('x', (value, i) => xScale(i))
.attr('y', value => chartHeight - yScale(value));
Here's a bl.ock that puts it all together and fills in the gaps to give a working example: Grouped Bars

Related

Unable to display x axis categories and make y axis start at 0

In this D3.js version 6.7 bar chart I am trying to align the x axis to show the categories and show the y axis to start at 0. Extending the height of the svg and changing the transforms does not appear to be working. How can I make the x axis categories appear under the bars and make the y axis start at 0? Thank you.
async function barChart() {
const dataset = await d3.csv("https://assets.codepen.io/417105/bodypart-injury-clean.csv");
console.log(dataset);
const width = 400;
const height = 400;
const margin = {top:20, right:30, bottom:30, left:40};
const canvas = d3.select("#viz")
.append("svg")
.attr("width", width)
.attr("height", height);
const wrapper = canvas.append("g").style("transform",`translate(${margin.left}px,${margin.top}px)`);
const xScale = d3.scaleBand()
.domain(["Arm","Eye","Head","Hand","Leg","Other"])
.range([0,width - margin.left])
.padding(0.2);
const yScale = d3.scaleLinear()
.domain(d3.extent(dataset, d => +d.Total))
.range([height,0]);
console.log(xScale("Leg"));
console.log(yScale(1700));
const barRect = wrapper.append("g")
.selectAll('rect')
.data(dataset)
.join('rect')
.attr('x', d => xScale(d.BodyRegion))
.attr('y', d => yScale(+d.Total))
.attr('width', xScale.bandwidth())
.attr('height', d => height - yScale(+d.Total))
.attr('fill', 'teal');
const yAxis = d3.axisLeft().scale(yScale);
wrapper.append("g").call(yAxis);
const xAxis = d3.axisBottom().scale(xScale);
wrapper.append("g").attr('transform', `translate(0,${height-margin.bottom})`).call(xAxis);
}
barChart();
The Y scale
The scale's domain sets the extent of the scale in your data's units, the scale's range sets the scale's extent in scaled units (pixels here). The first value in the domain is mapped to the first value in the range.
Your domain is set to:
.domain(d3.extent(dataset, d => +d.Total))
d3.extent returns the minimum and maximum matching values, as your minimum value is not zero, your scale's domain does not start at 0. If you want to set the scale's domain's lower bounds to zero, you need to set that, like so:
.domain([0,d3.max(dataset,d=> +d.Total)])
.domain/.range take arrays, these arrays for a linear scale must have the same number of elements
But you also don't want your scale's range to be [height,0] because of margins:
.range([height-margin.bottom,margin.top])
You want the data to be scaled from between within the two margins, height-margin.bottom is the furthest down the page you want to plot data, and margin.top is the furthest to the top of the SVG you want to plot data.
Now your bars are a bit off, that's because you aren't accounting for the margin in the height attribute:
.attr('height', d => height - yScale(+d.Total))
you need:
.attr('height', d => height - margin.bottom - yScale(+d.Total))
Note, a common approach to avoid having to constantly reference the margin is to apply the margin to a parent g and have width height reflect the size of the plot area within that g (not the entire SVG).
The X Axis
Now that the y scale is configured, let's look at the x axis. All you need to do here is boost the bottom margin: the text is appended (you can inspect the page to see it is there, just not visible). Try margin.bottom = 50.

The axis label at x=0 does not show up

I am using D3 to draw a line chart. The value at x=0 does not show up.
The code for the axis is shown below.
const xScale = d3
.scaleTime()
.domain(d3.extent(data[0].series, d => d.time))
.range([xPadding, width - xPadding]);
const xAxis = d3
.axisBottom(xScale)
.ticks(4)
.tickSizeOuter(0)
.tickSizeInner(0)
.tickFormat(d3.timeFormat('%Y'));
I am not sure why it is not showing up the label at x=0, which is 2014. On checking the SVG, only three tick marks are displayed, but the one at x=0 is not in the SVG element.
CodePen for this: https://codepen.io/vijayst/pen/bLJYoK?editors=1111
I see different solutions which have their pros and cons. The third solution should be the cleanest and most generic.
Add the left tick manually:
Since d3 handles itself the location of x-axis ticks, one way of doing so would (if the data set is fixed) would be to manually add the missing tick:
svg
.append("g")
.append("text")
.text("2014-02-01") // can be retrieved from data instead of being harcoded
.style("font-size", 10)
.style("font-family", "sans-serif")
.attr("transform", "translate(0," + (height - yPadding + 10) + ")")
which looks great, but in this case you might have problems if for a given dataset, d3 chooses to display a tick close to the left edge of the axis. Both d3's tick and the label we've included could overlap.
Modify the x-scale to start before the first day of the year:
An other solution would be to increase the x-axis range on the left to make it start one month before the first point's date. To try this out, we can replace:
.domain(d3.extent(data[0].series, d => d.time))
with
.domain(d3.extent([new Date(2013, 12), new Date(2019, 1)]))
which allow d3 to legitimately include a "year-tick" for 2014 at the begin of the x-axis.
but in this case, the first point will have an offset with the begin of the x-axis range.
Push a specific tick to ticks auto-generated by d3:
An other solution: we can push a specific tick to the ticks auto-generated by d3. But this requires to modify the format of ticks to "%Y-%m".
In order to do this, replace:
.tickFormat(d3.timeFormat("%Y"));
with
.tickFormat(d3.timeFormat("%Y-%m"));
we can then push a new specific tick to the set of ticks generated by d3:
var ticks = xScale.ticks();
ticks.push(new Date(2014, 1, 1));
xAxis.tickValues(ticks);
and include some padding in the left and the right of the chart since now tick labels have a part displayed outside the graph:
const svg = d3
.select("#chart")
.append("svg")
.attr("width", width)
.attr("height", height)
.style("padding-left", 15)
.style("padding-right", 15);

How can I get a time axis onto an object, using D3?

I'm very new to d3 and trying to learn by building a visualization.
My goal right now is to make a circle and color the circle based on some temporal data. I've made the circle, and want to add a timescale to it. The circle I have created fine using d3.arc() on an svg element. I have also created a time scale (seen below). My question is, how can I "attach" this time scale to the circle? I want to be able to say that at xyz point in time, my data holds this value, so now color the circle based on a color scale.
Or...am I going about this wrong?
var time = d3.scale.ordinal()
.domain(d3.extent(data, function(d) {
return d.date;
}))
I think you may need to use a quantitative scale instead of ordinal.
https://github.com/mbostock/d3/wiki/Ordinal-Scales says -
Ordinal scales have a discrete domain, such as a set of names or categories
and in your code, you use the "extent" of the date property, which only gives you 2 values - the earliest and most recent date in your data. That is a discrete domain, but a very limited one, and wouldn't represent your data very well. The scale will only output at most 2 values.
var now = Date.now();
var then = now - 1000;
var colors = d3.scale.ordinal()
.domain([then, now])
.range(['#ff0000','#0000ff']);
colors(then); // red
colors(now); // blue
colors(now - 500); // red... expecting violet
change 'ordinal' to 'linear' and leave the rest as is.
var now = Date.now();
var then = now - 1000;
var colors = d3.scale.linear()
.domain([then, now])
.range(['#ff0000','#0000ff']);
colors(then); // red
colors(now); // blue
colors(now - 500); // violet
The tricky part (at least for me) was remembering that the output of d3.scale.linear() (the 'colors' variable above) is a function. It can be called just like any other function.
var fakeData = d3.range(then, now, 10);
var svg = d3.select('body')
.append('svg')
.attr({ height: 500, width: 500 });
var circle = svg.append('circle')
.attr({ r: 100, cx: 250, cy: 250 });
function changeTime(time){
circle.attr('fill', colors(time));
}

d3.js: can't get padding working on rangeRoundBands?

I'm trying to draw a bar chart with padding between the bars, but I can't get rangeRoundBands to insert padding - though the docs tell me this is possible, so I'm doing something wrong.
This is my code:
var x0 = d3.scale.ordinal()
.rangeRoundBands([margin.left, width], 0.1, 0.1);
...
ap_bars.transition().duration(1000)
.attr("x", function(d, i) {
return i * x0.rangeBand();
})
.attr("width", x0.rangeBand());
But the bars are all stacked together.
JSFiddle here: http://jsfiddle.net/6pnem/5/
You're changing the size of the band with padding correctly, but the way you're plotting the bars will also put them side by side right next to each other. As you increase the padding, you'll see the that the bars just get skinnier and skinnier but stay pinched together.
Instead of using a constant based on the data index (i * x0.rangeBand()) you need to position the bar based on scaled data: (x0(d)). You'll probably want to adjust the position of the bar so that center of the bar is the center of each band: (x0(d) - x0.rangeBand() * 0.5).

D3: Create a continuous color scale with many strings/inputs for the range and dynamically changing values of the domain

I am trying to create a linear color scale for a heatmap. I want to color scale to go through a large set of specific colors, where the first color corresponds to the min of the data and the last color should be given to the max of the data.
I know that I can do this by also giving the domain 17 values in between the min and max, but I do not know how to do this dynamically if the user is able to change the dataset (and thus change the coloring of the heatmap)
In essence I would like to following, but I know it does not work
var colorScale = d3.scale.linear()
.range(["#6363FF", "#6373FF", "#63A3FF", "#63E3FF", "#63FFFB", "#63FFCB",
"#63FF9B", "#63FF6B", "#7BFF63", "#BBFF63", "#DBFF63", "#FBFF63",
"#FFD363", "#FFB363", "#FF8363", "#FF7363", "#FF6364"])
.domain([d3.min(dataset, function(d) {return d;}),
d3.max(dataset, function(d) {return d;})]);
Can anybody please tell me what I need to put into 'domain' to make it work?
EDIT:
I did find something that does what I want. Using R I calculated 256 colors in between the 17 from above with the designer.colors functions and put this into the range. This does give the feeling of a continous color scale
var colorScale = d3.scale.linear()
.range(["#6363FF", "#6364FF", "#6364FF", "#6365FF",
"... several other lines with color codes ..."
"#FF6764", "#FF6564", "#FF6464", "#FF6364"])
.domain(d3.range(1,257));
var quantize = d3.scale.quantile()
.range(d3.range(1,257))
.domain([d3.min(dataset, function(d) {return d;}),
d3.max(dataset, function(d) {return d;})]);
Now I can use the color in this fashion
colorScale(quantize(dataset))
But I'm wondering if this can also be done in less lines of code?
You want to split the problem up. First define a scale for your heatmap that maps 0-1 to your colours. Then define a second (dynamic) scale that maps your dataset to 0-1. You can then combine the scales to paint your shapes.
var colours = ["#6363FF", "#6373FF", "#63A3FF", "#63E3FF", "#63FFFB", "#63FFCB",
"#63FF9B", "#63FF6B", "#7BFF63", "#BBFF63", "#DBFF63", "#FBFF63",
"#FFD363", "#FFB363", "#FF8363", "#FF7363", "#FF6364"];
var heatmapColour = d3.scale.linear()
.domain(d3.range(0, 1, 1.0 / (colours.length - 1)))
.range(colours);
// dynamic bit...
var c = d3.scale.linear().domain(d3.extent(dataset)).range([0,1]);
// use the heatmap to fill in a canvas or whatever you want to do...
canvas.append("svg:rect")
.data(dataset)
.enter()
// snip...
.style("fill", function(d) {
return heatmapColour(c(d));
Plus you can use the d3.extent function to get the min and max of the dataset in one go.
Use a Quantitative Scale plus Color Brewer
// pick any number [3-9]
var numColors = 9;
var heatmapColour = d3.scale.quantize()
.domain(d3.extent(dataset))
.range(colorbrewer.Reds[numColors]);
// use the heatmap to fill in a canvas or whatever you want to do...
canvas.append("svg:rect")
.data(dataset)
.enter()
// snip...
.style("fill", function(d) {return heatmapColour(d);})
Use threshold scales. Here is a quick example:
coffee> d3 = require 'd3'
coffee> color = d3.scale.threshold().domain([5,30,100]).range(["red","orange","green"]);
coffee> color 6
'orange'
coffee> color 3
'red'
coffee> color 33
'green'

Resources