Consider adding a couple of benchmarks with plotting code?

[jimhester]

As one of the primary use cases of R is plotting I think having a benchmark that measures this might be useful.

It is also is the type of code (a lot of single calls) that compilers might not be intrinsically that great at, so it might be worth investigating strategies to improve performance.

Ideally we would have a base R plot, a lattice plot and a ggplot2 plot, but as the current benchmarks only seem to use packages shipped in the main R distribution we might have to settle for only base R plotting for now.

I would be happy to write some simple code and open a PR if you would be interested.

[o-]

That would be great. Currently all our benchmarks run without dependencies. But maybe it is about time to drop that limitation. We can just pre-install the dependencies in our benchmark-container job.

The whole setup for running the benchmarks is honestly quite convoluted. If you have a benchmark, I can add it to the suite.

[jimhester]

Here is some basic plotting code for all three types. They all use the diamonds data which is part of the ggplot2 package.

## ggplot2 plot
library(ggplot2)

p <- ggplot(diamonds) + geom_point(aes(carat, price, color = cut)) + facet_wrap(vars(color))

ggsave(p, filename = "ggplot.png")

## base R plot - this doesn't currently do any facetting / lattices, but I could probably do so if we want.
png("base.png")
plot(diamonds$carat, diamonds$price, col = diamonds$cut, pch=19)
dev.off()

## Lattice plot
library(lattice)
png("lattice.png")
xyplot(carat ~ price | color, group = cut, data = diamonds, pch = 19)
dev.off()