Skip to contents

Get a plot scale

Source: R/message_senders.R
get_scale.Rd

This function returns a specific scale from a specifc plot in a plotscaper scene.

Usage

get_scale(x, id = NULL, scale = NULL)

Arguments

x

A plotscaper scene

id

A string id of the plot. See id

scale

A string id of the scale (x, y, width, height, area, or size)

Value

A list of scale properties

Details

This function is primarily meant for internal use, however, you can use it to learn how plotscaper implements scales. The output can look a bit overwhelming, however, it's not too complicated once you understand how plotscaper scales work.

Each scale has two important properties:

  • Domain: The space values are translated from

  • Codomain: The space values are translated to

For example, in a typical scatterplot, the x scale might have the range of the data (e.g. [1, 10]) as its domain and the width of the plotting region as its codomain (e.g. [0, 800] pixels).

The scale's job is to link the domain and codomain, such that we can push values forward through the scale, first through the domain and then the codomain. This is done by translating to an intermediate range [0, 1]. For example, using the x scale above, we might first translate the value 5.5 to 0.5 (midpoint of the domain) and then translate 0.5 to 400 (midpoint of the codomain). We may also be able to reverse the process and pull values back through the scale, first through the codomain and then through the domain.

Scale, domain, and codomain each have props and defaults properties which store the relevant values. For example, for a continuous scale, props and defaults store the min and max as well as a transformation function and its inverse (trans, inv), for a discrete point scale, they store the vector of labels, their order, etc...

On scale, the props and defaults store the following properties: zero, one, scale, mult. The zero and one properties modify where the normalized domain values get placed in the codomain, and vice versa. Suppose our x ([1, 10], [0, 800] px) scale had zero = 0.1 and one = 0.9. Then data values get pushed to the following intermediate values:

  • The value 1 to 0.1 since 0.1 + (1 - 1) / (10 - 1) * (0.9 - 0.1) = 0.1

  • The value 2 to 0.1889 since 0.1 + (2 - 1) / (10 - 1) * (0.9 - 0.1) = 0.1889

  • The value 3 to 0.2778 since 0.1 + ((3 - 1) / (10 - 1)) * (0.9 - 0.1) = 0.2778

  • ...

  • The value 10 to 0.9 since 0.1 + ((10 - 1) / (10 - 1)) * (0.9 - 0.1) = 0.9

When those values get translated to the space of the codomain, we end up with 10% margins on each side, i.e.

  • The value 1 gets pushed to 80 pixels

  • ...

  • The value 10 gets pushed to 720 pixels

The scale and mult properties both multiply the normalized domain values. They work the same way, however, they are different semantically: scale is meant to be constant whereas mult may change dynamically, through interaction. For example, by default, in a barplot, the width scale gets assigned the scale value of 1 / k, where k is the number of categories/bars, and a mult value of 0.9. This means that each bar is 1 / k * 0.9 * [plot width in pixels] wide, and we can dynamically make it wider or narrower by pressing the +/- keys to modify the mult property (but not the scale property).