Skip to content

Latest commit

 

History

History
66 lines (47 loc) · 2.46 KB

ROADMAP.md

File metadata and controls

66 lines (47 loc) · 2.46 KB
Raw

Roadmap

Timeline

I have about two months remaining in the Bay Area to get this package into shape The goal is to have a subset of the API working + a draft of the paper before I finish up here. I also need a better name for the package...

The overall design of the grammar is in place, in that it is possible to express a lot of the graphics we want to create, now we have to focus on its implementation. As far as a subset of the features go: the aim is to have the following working:

  • brush and highlight (transient)
  • brush and hightlight (persistent)
  • wheel zoom
  • linked brushing (example, histogram linked to scatterplot)

Implementation

There is the user facing interface (computations done in R) and and client side interface which is JS.

R

At the moment we have the idea of having different "backends", one for static plots rendered with ggplot2 and other backends for interactivity Vega + shiny or just Vega. On the R side, there are the following conceptual issues that I need to sort out:

  • triggering layers via draw and or control, is a list of data frames a good way of doing this?

  • representing the output of a control_*, we want to make it explicit that the output of a layer this is probably best represnted as a reactive data frame using shiny.

  • modifiying aesthetics and graphical options via mutate (how to make it clear to a user which layer they are updating) + in the interactive context expressing relationships between layers.

  • order of evaluation for graphics pipeline: some ops i.e. scaling should be performed at the end of pipeline?

  • compiling down to a spec for vega (mostly sorted but need to think about signal representation in relation to the graphics pipeline)

  • how to represent linking in the grammar (natural to think of it as a join?)

JS

htmlwidgets + Vega (+ shiny), currently kinda hacking at this but could be better integreated with vegawidgets.

For this work, need to use the View API so we can add listeners to signals + dynamically insert/update/transform data. Where possible we could push everything client side essentially using JS like we would use C/C++ for computations in R.

  • we need to first parse the spec before creating the view object
  • once we have a view object we can add signal listeners if there are signals in the spec
  • have the data represented as a url if not local or large?
  • at runtime it would be great to perfom any of the transformations/summaries prior to rendering the plot