Please help me test the current development version of tinytex

rmarkdown
latex
pdf
tinytex

#1

Since several users have run into the issue of failing to compile R Markdown to PDF (e.g., this post, and the Github issue rstudio/rmarkdown#1285, with detailed solutions here), I hope some brave users could help me test the current development version of tinytex for a few days:

devtools::install_github('yihui/tinytex')
# although not required, you may also test the dev version of rmarkdown
devtools::install_github('rstudio/rmarkdown')

Although I'm pretty sure it should fix the issue, I need to be careful about releasing it to CRAN (for reasons explained here).

Note that the issue only affects those who compile R Markdown documents to PDF (1) interactively (i.e., call rmarkdown::render() in the R console), (2) on Windows, (3) using MiKTeX, and (4) in the R console of RStudio. This is a rather weird bug. You won't have problems if you don't satisfy one of these conditions. For example, if you hit the Knit button in RStudio, you should be fine (because the R session launched is not interactive).

BTW, you don't have to uninstall MiKTeX and install TinyTeX (tinytex::install_tinytex()), although I believe TinyTeX will make your life much easier (Note that TinyTeX is a LaTeX distribution, and tinytex is the R package; they are not really tied to each other).

Thanks!


#2

I've read a bit about the motivation for tinytex, and agree with you for it's use case. I was unaware of the development until this week when I finally attacked upgrading my Mac work environment (Update to High Sierra, update RStudio, update R, migrate current projects over and check for issues/bugs, resnapshot packrat etc).

Once I realised what had happened by finding the issues you refer to above because of the new message fed to R when a .pdf is compiled about tinytex, I think the solution you've put out is great. I would suggest that, if possible though, tinytex becomes a proper dependency to rmarkdown and knitr rather than a suggests. It seems that expected functionality now (in my case rmarkdown::render(...., output_dir = "./my_folder"/)) depends on the tinytex package, plus the installation is "tiny" so what's the harm :slight_smile:


#3

Absolutely. Done two months ago: https://github.com/rstudio/rmarkdown/commit/0031859 :slight_smile: Still in the dev version of rmarkdown, and I don't have a specific plan about the next CRAN release (1.10) yet.


#4

Shoulda known you'd have been on top of that :slight_smile: