Question

I’m working on getting knitr setup. I installed the latest version of R (2.15.1), Lyx 2.0 including the MiKTex 2.9 distribution, and RStudio 0.96.304 on a Windows 7 Enterprise box. I can get pdflatex output if I open up Lyx and simply view the tutorial, so the basic system is working. I then downloaded the minimal Rnw example saved it in my working directory as testknitr.Snw, opened that file in RStudio, and pressed the compile PDF button. The knitr output completes with a single warning about the parser package, and produces a file testknitr.tex. pdflatex.exe then runs, but fails, and the particular error in the log file seems to be

 ! LaTeX Error: Environment alltt undefined.  

I received the same error when attempting to compile testknitr.tex using TexWorks. I created that file from inside R using knit(“testknitr.Rnw”) – different extension because of the default in RStudio. I did tell MikTek to update packages automatically when compiling. A bit of research on CTAN suggests that the alltt environment is part of the ltxbase package, which is installed when I look at the package manager. In fact it seems like a pretty core part of the whole thing!

OK – I was reading the comments on the minimal example page, and discovered a workaround to that problem, and that it is a known bug as of 19 hours ago. And yes, I doublechecked that RStudio is set to use knitr, not sweave. I now get a new error:

! LaTeX Error: Command \textquotesingle unavailable in encoding T1. 

EDIT: OK! it turns out that error isn’t fatal – there’s a lovely pdf of the minimal example sitting in the working directory if one takes the time to look. Somewhat alarming that RStudio thinks compilation failed when it didn’t?

Answer 1

The workaround, perhaps obvious to an experience LaTex-ie, is to add

\usepackage{alltt}

to the file. According to Yihui’s comment this will be fixed in future versions, or now if you want to get the development version from github. I also tested @Yihui’s comment above that the line

\usepackage[T1]{fontenc}

could be commented out. This fixes the 2nd issue in the question, and RStudio now treats the compilation as successful, cleaning up (some?) intermediate files and immediately previewing the pdf.