Not supporting UTF-8 characters in file paths when generating documentation

[JantekM]

When making a custom template I noticed that after adding a custom format for the template I couldn't generate roxygen documentation with devtools::document() from RStudio which gave this error:

> devtools::document()
i Updating JantoTools documentation
Error: Failed to open G:\MĂłj dysk\Zdalne\Dokumenty zdalne\Statystyka\JantoTools/R/templates.R

Which, after some digging, led me to this piece of code throwing the error:
roxygen2::roxygenise("G:/Mój dysk/Zdalne/Dokumenty zdalne/Statystyka/JantoTools", NULL)

Note the change in the file path from G:/Mój dysk/... to G:\MĂłj dysk\...

Running the whole thing from another location with no ó in file path succeeded with no errors.

I tried (or checked if it is set) with no effect:

• setting global default encoding to UTF-8
• same with project settings
• reopening the file with UTF-8 encoding
• options(encoding = "UTF-8")
• using Sys.setlocale()

The whole code is available here.

[kleinbub]

I have exactly the same issue (on windows 10). I'm forced by my institution to work in a path that has a "Università" with a unicode letter in the path. I used to be able to overcome this using symbolic links in older versions. But now I get
"Error: Failed to open C:\Users\Kleinbub\OneDrive - Università degli Studi di Padova" (note the conversion from UTF8 'à' to ANSI 'Ã' )
I have pinpointed the issue up to the update_collate() function. Indeed when I run roxygenize without the "collets" roclets it runs with no error.

[jacobmaugoust]

Hello,
Same issue here. It works fine with the 7.1.1 version but not with the 7.1.2 one; I changed absolutely nothing in the path or project enconding options. Had to roll back to the 7.1.1 version.
I tried to run the code of the roxygenize function; no particular errors, but at the end, the various lapply's on the blocks object lead to empty things, so maybe it's somewhere there...
Hope this issue will be solved, good luck with that!

[hadley]

Hmmmm, given that it works in 7.1.1 and not 7.1.2, the most likely cause would be the conversion from Rcpp to cpp11.

Could any of you give me a traceback() to confirm exactly where the error is coming from?

[hadley]

Looks like readxl had the same problem: https://github.com/tidyverse/readxl/blob/330805693a0be798458dc3a919effa744429abea/src/XlsWorkBook.h#L29-L34

[kleinbub]

Is this what you need @hadley?
It should be noted that the correct path would be ...Università... not ...UniversitÃ... as reported in the error.
Also, if relevant, this is a windows 10 system.

Error: Failed to open A:\OneDrive - Università degli Studi di Padova\rMEA development\rMEA/R/rMEA_IO.R 

9.
FUN(X[[i]], ...) 
8.
lapply(paths, find_includes) 
7.
generate_collate(file.path(base_path, "R")) 
6.
update_collate(base_path) 
5.
roxygen2::roxygenise(pkg$path, roclets) 
4.
devtools::document(roclets = c("rd", "collate", "namespace")) 
3.
.is.positive.intlike(x) 
2.
.traceback(x, max.lines = max.lines) 
1.
traceback({
    devtools::document(roclets = c("rd", "collate", "namespace"))
})