tikzDevice produces a non-UTF-8 output file on Windows

[julianre]

The TeX/TikZ files generated by tikzDevice are not encoded in UTF-8 but in the Windows 1252 file encoding (under Windows, see Session Info). This results in a problem, if the plot is included in the main LaTeX document (UTF-8 encoding). In my case (MWE), I receive the following error message:

pandoc.exe: Cannot decode byte '\xd6': Data.Text.Internal.Encoding.streamDecodeUtf8With: Invalid UTF-8 stream
Error: pandoc document conversion failed with error 1

Minimal working example (RMarkdown file):

---
title: "test"
author: "test"
date: "Today"
output: 
  pdf_document: 
header-includes:
   - \usepackage{tikz}
---
```{r setup, include=FALSE}
library(tikzDevice)
options(tikzDefaultEngine = "xetex")
```

```{r plot, dev="tikz", external=FALSE, echo=FALSE}
x <- rnorm(50)
y <- rnorm(50)

plot(x, y, xlab = "ÖÄÜ", ylab = "öäü")
```

I tried the MWE under Ubuntu and the generated file is encoded in UTF-8 (no problem).

xfun::session_info()

R version 3.6.1 (2019-07-05)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows 10 x64 (build 18362), RStudio 1.2.1335

Locale:
  LC_COLLATE=German_Germany.1252 
  LC_CTYPE=German_Germany.1252   
  LC_MONETARY=German_Germany.1252
  LC_NUMERIC=C                   
  LC_TIME=German_Germany.1252    

Package version:
  compiler_3.6.1    evaluate_0.14    
  filehash_2.4-2    glue_1.3.1       
  graphics_3.6.1    grDevices_3.6.1  
  grid_3.6.1        highr_0.8        
  knitr_1.24        magrittr_1.5     
  markdown_1.1      methods_3.6.1    
  mime_0.7          png_0.1.7        
  rstudioapi_0.10   stats_3.6.1      
  stringi_1.4.3     stringr_1.4.0    
  tikzDevice_0.12.3 tools_3.6.1      
  utils_3.6.1       xfun_0.9         
  yaml_2.2.0  

[rstub]

Thanks for the report. Besides the workarounds mentioned on SO, you could also try to tell LaTeX to read the file using Windows 1252 encoding instead of the default UTF-8:

options( tikzLatexPackages = c(getOption( "tikzLatexPackages" ),
                               "\\usepackage[cp1252]{inputenc}" ))

(Completely untested since I am not on Windows right now ...)

[rstub]

Maybe related: Encoding problems on Linux at https://stackoverflow.com/q/58401825/8416610