Skip to content

Instantly share code, notes, and snippets.

Last active November 22, 2022 23:32
What would you like to do?
Line Transition
license: gpl-3.0

This example is the second of three in the Path Transitions tutorial; see the previous example for context.

The desired pairing of numbers for path interpolation is like this:

M x0, y0 L x1, y1 L x2, y2 L x3, y3 L xR, y4
   ↓   ↓    ↓   ↓    ↓   ↓    ↓   ↓
M xl, y0 L x0, y1 L x1, y2 L x2, y3 L x3, y4

Where xl is some negative value off the left side, and xr is some positive value off the right side. This way, the first point ⟨x0,y0⟩ is interpolated to ⟨xl,y0⟩; meaning, the x-coordinate is interpolated rather than the y-coordinate, and so the path appears to slide off to the left. Likewise, the incoming point ⟨xr,y4⟩ is interpolated to ⟨x3,y4⟩.

While you could write a custom interpolator and use transition.attrTween to achieve this, a much simpler solution is to interpolate the transform attribute rather than the path. This way, the shape of the path remains static while the it translates left during the transition.

Immediately prior to the transition, the path is redrawn as follows:

M x0, y0 L x1, y1 L x2, y2 L x3, y3 L xr, y4

Then, a transform transition is applied:


This causes the path to slide left. A clip path is used so the path is not visible outside of the chart body.

Note that for charts with spline interpolation, you’ll need to crop the visible part of the line by an extra point, so that the change in tangent is not visible; see the next example.

<!DOCTYPE html>
<meta charset="utf-8">
.line {
fill: none;
stroke: #000;
stroke-width: 1.5px;
<svg width="960" height="500"></svg>
<script src="//"></script>
var n = 40,
random = d3.randomNormal(0, .2),
data = d3.range(n).map(random);
var svg ="svg"),
margin = {top: 20, right: 20, bottom: 20, left: 40},
width = +svg.attr("width") - margin.left - margin.right,
height = +svg.attr("height") - - margin.bottom,
g = svg.append("g").attr("transform", "translate(" + margin.left + "," + + ")");
var x = d3.scaleLinear()
.domain([0, n - 1])
.range([0, width]);
var y = d3.scaleLinear()
.domain([-1, 1])
.range([height, 0]);
var line = d3.line()
.x(function(d, i) { return x(i); })
.y(function(d, i) { return y(d); });
.attr("id", "clip")
.attr("width", width)
.attr("height", height);
.attr("class", "axis axis--x")
.attr("transform", "translate(0," + y(0) + ")")
.attr("class", "axis axis--y")
.attr("clip-path", "url(#clip)")
.attr("class", "line")
.on("start", tick);
function tick() {
// Push a new data point onto the back.
// Redraw the line.
.attr("d", line)
.attr("transform", null);
// Slide it to the left.
.attr("transform", "translate(" + x(-1) + ",0)")
.on("start", tick);
// Pop the old data point off the front.
Copy link

canimus commented Jan 20, 2012

Very nice example!
Quick question: How does your CPU look when running in Google Chrome?
Do you experience a resource intense behavior?

Copy link

thanks for explaining this further here.

Copy link

kfrajtak commented Apr 8, 2013

Can you please help me with my problem ( Thank you, Karel

Copy link

sgizm commented May 10, 2013

Hello, I have a question: I want to use some external data with this graph. Can i learn which kind of data format i should have?

Copy link

Hi Mike, Thanks for this nice example. But when I try it in my browser,There is one error which keeps appearing"Unexpected value null parsing transform attribute.", as it running,this error repeats again and again. Do you notice this error and have any solution for it ?

Copy link

< .attr("transform", null)

.attr("transform", "")
then works in ie

Copy link

ghost commented Jun 5, 2014

To canimus

Probably too late but it only runs the cpu hard if you require that it preform the ticks faster. He only has them every 500 milliseconds. I'm using a similar model for tracking mouse movements and it takes up quite a bit of cpu.

Copy link

It is awesome!!!

Copy link

      .attr("transform", "translate(" + x(-1) + ",0)")
      .each("end", tick);

.each(,) runs a new function "tick" for each point after each iteration, so it'll exponentially grow per cycle. I left this open on a second tab and it ran my CPU into the ground.

Copy link

base698 commented Jul 14, 2015

Anyone know how to animate the x axis? I have a data stream that should increase and change in the x and y. Also the domain will change. Haven't found any good examples.

Copy link

sunking commented Oct 30, 2015

I also want to animate the x-axis like it was in the last example. Unfortunately, there is no sample code given, I am unable to reproduce a graph with a data stream along a moving x-axis.

Copy link

bjorngi commented Dec 9, 2015

@sunking @base698 check out the source code for the site.
Pulled out code for last example

Copy link

oller commented Feb 4, 2016

Thanks for this @mbostock.

I'm trying to get my head around these magic numbers and how they interact with each other. This pertains to the gist @bjorngi posted above...

n = 243,
duration = 750

Is it just the case that :
n the length of the data array
duration the time in ms between iterating over these

So 243 * 750ms ~= 180s which is the 3 minute width of the x axis.

My scenario is I'll have an unknown length of data coming (via websockets) but will know the length of time I'd like the x axis to represent. Unsure If I'll have to buffer this data to update the chart on a more regular basis.

Any insight appreciated!

Copy link

Jaapze commented Feb 26, 2016

I made a test with the last example (scroll graph) on the page:

Copy link


The graph is working perfectly on my computer. However, if I leave the screen (such as changing tabs) and come back to it, the x-axis will speed up significantly to catch back up to the current time. Is there a way I can make it run smoothly throughout? Thank you!

Copy link


Quick addition for people who might also have problems with the translate reset not being executed properly, because the translate transition is still in progress (might happen when you are dealing with varying frame rates or data intervals). Took me a fair while and significant debugging to figure it out.

Make sure you reset the transition on the selection when resetting the transform:
  .attr("d", line)
  .attr("transform", null);

Copy link

xmesaj2 commented Apr 9, 2017

How to modify this example to get multi-series chart, for example by adding a 2nd data array?
Now it takes let's say [11,15,30,1,10,20] and draws a line with it and how to add an example array of [33,35,21,40,22,2] that would draw another line?

Copy link

I forked the code pen from @Atticweb and updated to v4. Everything seems to be working properly, but I'm a little unhappy with the transition of the axis text, it doesn't respect the easing (jumpy), but the ticks transition smoothly. Any ideas?

I'm pretty new to d3, but wanted to use this same idea in a similar project that requires v4.

(sorry if you get this twice I should have posted here in the first place probably not on bjorngi's gist)

Copy link

Has anyone implemented something similar in react-native? If so, would you mind sharing your experience?

Copy link

Yaweii commented May 30, 2018

@alexsmartens I am trying to build something similar in react-native as well. Did you get any chance to figure this out yet?

Copy link

How can people use this knowing that the tick function is called by an exponentially number of growing of callees ? It will just grind user's cpu/browser into the ground.

Copy link

      .attr("transform", "translate(" + x(-1) + ",0)")
      .each("end", tick);

.each(,) runs a new function "tick" for each point after each iteration, so it'll exponentially grow per cycle. I left this open on a second tab and it ran my CPU into the ground.

which therefore makes this example completely invalid without a significant refactoring, which everyone else seems to be happily ignoring ... I don't understand.

Copy link

@NimChimpsky I agree.

Copy link

thanhtschoepe commented Sep 3, 2021

I made a notebook using this example. Check it out. 😊. Please give me feedback if you have any.

Thank you @mbostock for the amazing guide & @pepijnverburg for the duration 0 trick.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment