Running R code from latex. (alternative to knitr)

Question

If you want to export the output of an R script to latex and create a pdf document you can use the package Knitr with RStudio. But I guess this is only a good practise for short results.

What if I want to create a whole book, a 1000 pages thesis?

How do yo do the opposite?

I mean, Can I include R code within my latex files and run it directly from say TexStudio+LaTex/LuaTex? (In order to get the numerical results and graphics in the pdf).

How?

NEW: Some people said one can use knitr to generate big docuemnts. Even if it's possible I think it's more comfortable and versatile to create LaTex documents from specific platform such as TexStudio.

So, what is the best practice?. Mixing both methods?: Generating parts of the document with Knitr and later mixing and modifying them with TexStudio?. How?

There are packages like lstlisting designed to show code properly but not to run it.

Answer 1

In ConTeXt, you can use the filter module for such tasks. In fact, the filter module is a generalization of ConTeXt MkII module m-r, which was designed for specifically what you are asking.

However, with the filter module it is easy to replicate the functionality for any language, so there is no MkIV version of m-r.

Here is a minimal example:

\usemodule[filter]

\defineexternalfilter
  [R]
  [
     filtercommand={R CMD BATCH -q --save --restore \externalfilterinputfile\space \externalfilteroutputfile},
     output=\externalfilterbasefile.out,
     readcommand=\typefile,
     read=no,
     cache=yes,
   ]

\starttext
\startR
  ushape <- c(rexp(500000), 12-rexp(500000))
  pdf("ushape.pdf")
  par(mfrow=c(1,2))
  hist(ushape)
  plot(density(ushape), main="Density")
  dev.off()
\stopR

\startplacefigure[location=here, title={Output from R}]
  \externalfigure[ushape.pdf][height=0.5\textheight]
\stopplacefigure

\stoptext

which gives

Note that the files are cahced, so subsequent runs are very quick.

Answer 2

But I guess this is only a good practice for short results.

I believe that this is a false premise. I never make books of 1000 pages, but I made several statistical reports, some near the 100 pages at two columns, with only a few text paragraphs in LaTeX but dozens and dozens of tables (xtables), dozens of various types of plots and dozens of statistical results, all generated by R chunks in a single file, with the same problems that in shorter documents.

With this experience, for me the rule is just the opposiste: the larger size of the document, the higher the convenience of the Sweave/knitr chunks (using Rstudio, LyX or TeXworks, that is not the question).

If the compilation times are too high in the previews, try with cache=TRUE in knitr and consider working mainly with subsets of the final document as far as possible. However, I've never needed any of these workarounds, because the compilation times were were only enough for some eye-rest.