How to use knitr to compare performance of different versions of R?

Question

I want to evaluate the performance of some code in different versions of R. This is easy enough in principle:

• Start R
• Use system.time() to measure the time it takes to run a piece of code
• Terminate R
• Rinse and repeat in a different version

Now, I want to use knitr to create a report to do this. So, it seems to me I need a mechanism to start a new session in each chunk.

How do I do this?

---

Some sample knitr markdown code to serve as a demonstration. This code plots a graphic using ggplot, but clearly both versions return identical timings, since I don't know how to start a new version of R for each chunk.

Comparison of R performance
========================================================

# Do analysis in R version 2.14

```{r fig.width=6, fig.height=3}
library(ggplot2)
data(diamonds)

system.time({
  p <- ggplot(diamonds, aes(carat, price/carat, colour=clarity)) + geom_point()
  print(p)
})
```


# Repeat same analysis in R 2.15

```{r fig.width=6, fig.height=3}
library(ggplot2)
data(diamonds)

system.time({
  p <- ggplot(diamonds, aes(carat, price/carat, colour=clarity)) + geom_point()
  print(p)
})
```

Answer 1

Adding the Rscript engine in knitr was easy, but I was held back by an R bug. Anyway, this engine is available since version 1.1.5 and will be on CRAN as version 1.2.

Now you can specify the chunk option engine='Rscript' and engine.path='path/to/the/desired/Rscript'.

For large scale performance comparisons, I think what Ari B. Friedman suggested in the comment above is a better way to go. It will be pretty tedious to type the engine paths if you have many code chunks for comparisons.