This is a rough plan for the future of the tikzDevice up to version 1.0
See milestone v0.6 for development status.
Support for new graphics primitives that were added after R 2.10.x
Initial UTF8 Support
Refactor package support systems
See milestone v0.7 for development status.
xelatex(native UTF8 support)
lualatex(native UTF8 support) See issue #28.
As much ~50% of the time involved in producing a figure using
tikz can be spent inside the metrics calculation routines. Some ideas for speeding this up:
.fmt file (a.k.a. precompiled preamble) for each graphics device. This may drop the amount of time it takes a TeX engine to start up.
Instead of making a system call to TeX each time a calculation needs to be done, start one instance of the TeX engine per graphics device and use interprocess communication to extract metrics from it.
This probably won't work. The problem is that in order to get real call and response communication with a subprogram, the caller has to communicate with it via a pseudo-terminal or some sort of console device on Windows. The logistics of doing this reliably in a cross-platform manner are mind-bogglingly difficult. As far as I know, noone has even made a cross platform library that solves this problem.
Some detailed profiling is needed to determine the best course of action.
Two step process for adding support for
tikzCoord in Grid-based graphics:
Add functionality for converting Grid coordinates to device coordinates and for representing TikZ annotations as Grid
grob objects. Work progressing in issue #37.
Add functionality for adding annotations to GGplot graphics. This is difficult as GGplot2 has an additional level of coordinate system manipulation that exists above the Grid coordinate system.
See milestone v0.8 for development status.
In addition to LaTeX, add support for:
Convince Kanye West to do a rap about us.
Ideas that don't have a home yet.
Asymptote: Would be fun because it can embed 3D imagry into TeX documents.
PSTricks: Possibly. But I'm not getting a warm fuzzy feeling about this one at the moment.
Would be great for authors who want to take advantage of two-color printing. Experimental implementation in the METAPOST branch. Difficult because it requires dodging around the R color handling system.