Environment variable LANG

[randy3k]

Recently, I notice that languageserver fails in some subtle way when the LANG variable is not set.
More precisely, the json returns by the server will be garbled if it contains any non-ascii chars.

It is caused by the fact that R will dump non-ascii chars in their literal form. It causes a dismatch in their number of bytes of the json and the calculated length of the json message.

(=)-~$ Rscript -e "cat('\U2018\n')"
‘
(=)-~$ LANG= Rscript -e "cat('\U2018\n')"
<U+2018>

I observed this bug in the recent version of languageserver when I am hovering on some text on Sublime Text and the doc rendered by html_to_markdown contains the char ‘. It is particularly an issue of Sublime Text because the variable LANG is not set in default.

Any suggestion on how to handle it? It could be an ongoing reason of issues.

[randy3k]

I haven't checked how it affects Windows users. But I assume that the special treatment has already handled it.

[randy3k]

One solution is to check on unix system that isTRUE(l10n_info()[["UTF-8"]]) and warn user to set the LANG variable. Another solution is to always use

writeLines(text, self$outputcon, sep = "", useBytes = TRUE)

as what we have done on Windows.