make sure the path is UTF-8

[randy3k]

it should fix some of the issues in #277.

We still need to fix the round trip issue

r$> a = "file:///foo/bar/中文.R"

r$> (b = languageserver:::path_from_uri(a))
[1] "/foo/bar/中文.R"

r$> (languageserver:::path_to_uri(b))
[1] "file:///foo/bar/%E4%B8%AD%E6%96%87.R"

[randy3k]

One solution is to always escape the uri from textDocument$uri. Say for example,

uri <- URLencode(textDocument$uri)

[renkun-ken]

Do you mean always URLencode the incoming URI from editor and URLdecode the outcoming URI to editor? I think it should work with VSCode but if an editor uses encoded URI that escapes all unicode characters, then returning a decoded URI seems undesired by such editors?

[renkun-ken]

I don't know what other editors do but VSCode seems to only encode a minimal set of reserved characters. For example, on macOS:

测试 copy !$&'()*+,;=:?@#.R

becomes

测试%20copy%20!$&%27()*+,;=:%3F@%23.R

[randy3k]

URLdecode the outcoming URI to editor?

No. We still need to escape special characters like spaces.

But it raises another issue: does vscode understand both "测试%20copy%20!$&%27()*+,;=:%3F@%23.R" and "%E6%B5%8B%E8%AF%95%20copy%20!$&'()*+,;=:?@#.R"?

[randy3k]

Oops,

I just notice that vscode escapes ' but R does not so URLencode might not work.

[renkun-ken]

I tried the latest commit in this PR and I don't observe any problem working with 测试 copy !$&'()*+,;=:?@#.R.

[renkun-ken]

Always encoding incoming URIs and not decoding outcoming URIs assumes the editor always does full decoding (with all %xx) and thus understands the encoded URI, which I think is quite a safe assumption.

[randy3k]

It also fixes #279 (but I don't understand why). It was a false positive fix.

[renkun-ken]

Thanks for working on this. LGTM.