Loon: An Interactive Statistical Visualization Toolkit

Introduction

loon is a toolkit for interactive data visualization and exploration. Currently, loon is embedded in R and Tcl. Please see the beta section for the current state of the software.

Installation

Statistical Displays

Currently we provide an interactive display for histograms, scatterplots, serialaxes plots (i.e. star glyphs or parallel coordinates) and navigation graphs.

User Interface

The user interface of loon has two elements: the display and the inspector.

loon screenshot. Left: Inspector and Right: Scatterplot display.

loon screenshot. Left: Inspector and Right: Scatterplot display.

The inspector lays out most of the functionality of the display. Displays support further actions via mouse gestures and modifier keys.

Point Glyph Types

Points in scatterplot can be displayed with different glyphs. loon distinguishes between primitive and non-primitive glyphs. The primitive glyphs are available in all scatterplots whereas the non-primitive ones need to be added to a scatterplot before they can be used. Read more.

Zoom and Pan

When zooming and panning the plot region gets shown in relation to all data on the worldview in the inspector.

Linked Displays

Selecting points in one display will also select them in all linked displays.

Selecting points in one display will also select them in all linked displays.

Multiple displays can be linked such that the linked points share their color, size, selected and active states.

Scatterplot displays can be linked ad-hoc with the inspector by changing the Linking Group of the scatterplot displays to be the same. loon will always suggest a linking group that has no display associated to it. The linking group none keeps the display unlinked.

Displays with different numbers of points can be linked too. The only linking constraint is that a point can be linked with no more than one point on another display.

Selection

loon provides multiple interactive selection tools.

Click Select: individual points can be selected or deselected by simply clicking with the mouse cursor on them.

Sweep Select: a sweep selection selects all points below a rectangular area which is defined by a left-click drag gesture. That is, the upper left corner of the sweep rectangle is at the location of the left-button press, and the lower right corner is at the current cursor location while the left button is pressed.

Brush Select: a brush selection also selects all points below a rectangular area. However, in contrast to the sweep selection, the rectangular brush area has a fixed size and a left-button press moves the lower right corner to the current mouse location, and a left-click dragging gesture moves the brush area along the mouse pointer.

Additionally, we provide selection by point color and via the command line.

Modify

The state of the selected points can be modified with the tools and actions in the Modify panel of the inspector. The following Modify panel is from the scatterplot inspector.

Move Points

On the scatterplot display selected points can be temporarily moved to a new location.

Command Line Control

Displays can be completely controlled via the command line. The R API is discussed in learn R section. It is similar to the tcltk package API. For example, in R for a plot with handle p once can query a plot state as follows

# Query State
p['selected']
l_cget(p, 'selected')

# Modify State(s)
p['selected'] <- TRUE
l_configure(p, zoomX=2, zoomY=3, selected=FALSE)

The Tcl API follows the Tk widget convention and provides a cget and configure widget function to query and modify plot states. The following code demonstrates this where the variable p holds the widget path name:

$p cget -x
$p configure -zoomX 2 -zoomY 3 -selected FALSE

Event Bindings

loon's event bindings provide the facility to have custom code (i.e. callbacks) evaluated at specific events. We support a number of different event types which, for example, include such changes in point color, zoom, selection, moving the mouse cursor over a visual element and window resizing.

$p bind state add {zoom pan} {
    puts "Zoomed or Panned"
}

or in R

l_bind_state(p, c('zoom', 'pan'), function() {cat('Zoomed or Panned')})

We provide various binding types. Read more.

Layers

The histogram, scatterplot and graph displays support layering of lines, polygons, text, rectangles and more. For example, the following picture shows a scatterplot of international airports with map information layered underneath the scatterplot points. The data is from the naturalearth project.

Layers can be arranged with loon's layers inspector.