# Programmatically Making Gradients in JavaScript

In this post we will make gradients without using any of the helper functions built into the canvas context for making them. In other words, we’ll make a gradient without using `createLinearGradient`

or `createRadialGradient`

. Whilst those two functions are satisfactory for most applications, you can do much more with these gradients, as you can interpolate between colours in any way you like. For example, you can use cosine interpolation rather than linear for a smoother, eased gradient. Also, if you understand how to perform colour interpolation, you will better understand how to make metaballs and similar graphical effects. (I’ll probably write a post on them soon, they’re really cool.)

## Making a linear gradient

Below is an example of a linear gradient using linear interpolation (LERP.) Notice how (at least for me in Firefox and Chrome) the two gradients are identical.

See the Pen by OliverBalfour on CodePen.

To make our gradient, we need to have two colours: the start and end colours. We’ll have them in an object format, with `r`

, `g`

and `b`

properties.

1
2

var startColour = {r: 0, g: 255, b: 50}, // green
endColour = {r: 255, g: 0, b: 100}; // pink

Then, we simply need to go across the screen, drawing values from an interpolation function. For now, we will use a simple linear interpolation (LERP) function. All it does is, given the decimal `p`

, get the number between `a`

and `b`

that falls along a straight line between `a`

and `b`

on a graph. So, if you were to plot the values of the function from `p = 0.0`

to `p = 1.0`

on a graph, you would get a straight line going from `a`

to `b`

. Instead, though, we will use the value for each of the three colour values (or four, if you wanted a fading gradient) from the start and end colours (as `a`

and `b`

,) and plot the interpolated values as the red, green and blue components of the gradient. We will do this for every column of pixels along the gradient.

1
2
3
4
5

function interpolate (a, b, p) {
// The higher p is, the less a is weighted and the more b is weighted in the result
// The inverse is also true
return a * (1 - p) + b * p;
}

It’s that easy. Then, we just pass each element of the colour into that function. We do this for every pixel along the length of the gradient. All in all, it’s very simple.

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16

var currentColour = {r: startColour.r, g: startColour.g, b: startColour.b},
x = 0, // x position along the canvas
percent = 0; // p
while (x < w) { // w is the width of the canvas in pixels
percent = x / w;
// Notice we aren't actually calling the function here, although we are still using the same algorithm
currentColour.r = Math.round(startColour.r * (1 - percent) + endColour.r * percent);
currentColour.g = Math.round(startColour.g * (1 - percent) + endColour.g * percent);
currentColour.b = Math.round(startColour.b * (1 - percent) + endColour.b * percent);
ctx.fillStyle = 'rgb('+currentColour.r+', '+currentColour.g+', '+currentColour.b+')';
ctx.fillRect(x, 0, 1, h);
x++;
}

Remember when setting the fill style to make sure the `r`

, `g`

and `b`

values are all whole numbers using `Math.round`

!

See the Pen by OliverBalfour on CodePen.

You can easily make a function that interfaces like `createLinearGradient`

using the above code as a starting point. However, that’s out of the scope of this tutorial. But first, we will adapt our above code to produce a prettier gradient and use cosine interpolation instead of linear interpolation.

### Removing ‘sludge’

If you look at the gradient below, you’ll notice that the gradient between the two colours is like a horrible brown sludge. You may recognise this kind of sludge, as blurring operations in photo editors produce it when incorrectly blurring adjacent bright colours together. This sludge occurs because the way we perceive brightness is different to that of a computer; while computers register brightness along a strictly linear scale, humans perceive brightness on an exponential scale. This is why turning on a light where there is no light has a larger percieved difference in brightness than turning on a light in an already lit area. This video is a great resource about how and why sludge exists (Thanks @osublake!)

See the Pen by OliverBalfour on CodePen.

To avoid creating sludge, we simply need to square the RGB components of the start and end colours, and then square root the return value of the interpolation function. Below is an example comparing our original (naive) approach and the new one with proper blending:

See the Pen by OliverBalfour on CodePen.

If you look in the above code, all we’ve done is square each component of the start and end colours, and inside the loop, we square root the value before rounding it to calculate the color at each step along the way:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17

// at the beginning:
startColour = {
r: 0,
g: 255 * 255,
b: 50 * 50
};
endColour = {
r: 255 * 255,
g: 0,
b: 100 * 100
};
// inside the loop:
currentColour.r = Math.floor(Math.sqrt(startColour.r * (1 - percent) + endColour.r * percent));
currentColour.g = Math.floor(Math.sqrt(startColour.g * (1 - percent) + endColour.g * percent));
currentColour.b = Math.floor(Math.sqrt(startColour.b * (1 - percent) + endColour.b * percent));

### Cosine interpolation

The code for cosine interpolation looks like this:

1
2
3
4
5

function interpolate(a, b, px){
var ft = px * Math.PI,
f = (1 - Math.cos(ft)) * 0.5;
return a * (1 - f) + b * f;
}

I won’t go into the math, but If you use it as a replacement for the linear interpolation function, you get this result:

See the Pen by OliverBalfour on CodePen.

There isn’t much difference, but if you compare it closely with the last one, you’ll find it eases in and out, whereas linear interpolation doesn’t. Cosine interpolation is much prettier for blending bright colour blocks together at the edges, but a very subtle difference and not as efficient.

## Making a radial gradient

By simply drawing progressively smaller circles on the same spot, we can effectively use the `arc`

function to draw a radial gradient. All we have to do to modify our linear code is change the drawing code to draw circles instead of lines:

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16

var radius = 100,
x = radius,
percent = 0;
while(x > 0){
percent = x / radius;
currentColour.r = Math.floor(Math.sqrt(startColour.r * (1 - percent) + endColour.r * percent));
currentColour.g = Math.floor(Math.sqrt(startColour.g * (1 - percent) + endColour.g * percent));
currentColour.b = Math.floor(Math.sqrt(startColour.b * (1 - percent) + endColour.b * percent));
ctx.fillStyle = 'rgb('+currentColour.r+', '+currentColour.g+', '+currentColour.b+')';
ctx.beginPath();
ctx.arc(w / 2, h / 2, x, 0, Math.PI * 2);
ctx.fill();
x--;
}

Combine that with the beginning of the linear gradient’s code (the start and end colour variables, the canvas setup etc.) you have a fully functioning radial gradient maker!

See the Pen by OliverBalfour on CodePen.

We can also draw a slightly less attractive and efficient radial gradient using trigonometry, just because it was pi day yesterday as of writing (3/14/15.) The way we do it is, instead of drawing a circle with the `arc`

function, we loop an amount of iterations equal to the circumference of the circle at the current radius. (Equal to `radius * Math.PI * 2`

.) Then, we calculate the angle that we’re up to, which is a fraction of `Math.PI * 2`

. Lastly we draw a small dot at the position we’re up to. Seeing as we only have the angle and radius, we need trig. So, `x = Math.cos(angle) * radius`

and `y = Math.sin(angle) * radius`

.

See the Pen by OliverBalfour on CodePen.

The above gradient isn’t particularly attractive with its rough edges; those are because each dot is drawn as a 3x3 pixel square, as any smaller leaves gaps in the gradient. The previous example is certainly faster, and prettier. Not as fast as a generic radial gradient made with CSS or the canvas’s context, but good all the same. Anyway, that brings us to the end of making radial gradients programmatically.

Thanks for reading. See you next time!

*This post was originally posted on my CodePen blog here.*