More on graphics translators

[cpsievert]

https://github.com/duncantl/RD3Device

• Inspects the graphics device and attempts to map R's graphical primitives to d3js code.

https://github.com/pmur002/gridgraphics

• Translates base to grid graphics. Thus, a low-level grid translator could automatically translate base as well.

[cpsievert]

More modern approaches

https://github.com/hadley/svglite

• Drop-in replacement for svg() which is faster and produces smaller files.

https://github.com/davidgohel/rvg

• Similar functionality to svglite and gridSVG, but also offers a graphics device for DrawingML files which allows one to embed and edit graphics inside Microsoft Word and Powerpoint.

https://github.com/davidgohel/ggiraph

• Uses rvg to produce web-based interactive ggplot2 graphics. Similar to animint, it provides tooltip and onclick aesthetics to trigger actions on hover or click of a graphical object. It does not, hover, provide any semantics for linked selections.

[cpsievert]

Why our approach is better than translating grid objects directly:

Given a grid (grob) object, it's impossible to know other what other data is attached to that object. However, in an exploratory analysis, we often want to obtain more information about outliers or interesting points, a technique known as identification. ggiraph addresses this problem by implementing new grid (grob) objects that essentially add attributes (e.g. tooltip/onclick) to existing grobs which are the output of its custom ggplot2 layers (e.g., geom_point_interactive())