Question about UTF-8 and file:getFilename

[easy82]

Nevermind, Google is my friend.

main.lua

Code: Select all

local filename = ""

function unescape (s)
  s = string.gsub(s, "+", " ")
  s = string.gsub(s, "%%(%x%x)", function (h)
    return string.char(tonumber(h, 16))
  end)
  return s
end

function love.filedropped(file)
  filename = file:getFilename()
end

function love.draw()
  love.graphics.print("Original: " .. filename, 10, 10)
  love.graphics.print("Unescaped: " .. unescape(filename), 10, 30)
end

Source: https://www.lua.org/pil/20.3.html

[Positive07]

Well your solution is good, but this is a problem with your file manager, and you should be careful with this solution, because I could have a file called Let%C3%B6lt%C3%A9sek and your solution would turn this into Letöltések (I think) which is something I don't expect. Basically the inverse to your problem.

By covering an edge case you create another, so the best solution would either be to make this a configuration option or don't depend on it... Also reporting the "error" (may be intended behaviour though) to the developers of the file manager may be a good idea, I don't know

[zorg]

Does it escapes any unicode strings or just file paths?

I don't know, I was testing this through love.filedropped() and file:getFilename().

Test it then, just to be sure.

To me, what's interesting is that not only the text is encoded as UTF-8, but it's not plainly UTF-8, but rather, escaped per-character with %-s like html; very strange.

Do you know any good way to unescape this string in LUA?

It can be done, but i don't know whether that's a solution, and not just a workaround...

Code: Select all

string.gsub(filename, "(%%)([0-9a-fA-F])", "%2") -- untested, but should work.

[easy82]

Well your solution is good, but this is a problem with your file manager, and you should be careful with this solution, because I could have a file called Let%C3%B6lt%C3%A9sek and your solution would turn this into Letöltések (I think) which is something I don't expect. Basically the inverse to your problem.

By covering an edge case you create another, so the best solution would either be to make this a configuration option or don't depend on it... Also reporting the "error" (may be intended behaviour though) to the developers of the file manager may be a good idea, I don't know

I see your point...!

It can be done, but i don't know whether that's a solution, and not just a workaround...

Yes it's just a workaround.

Do you guys think this is a bug that I should report?

[pgimeno]

The URL encoding is probably not LÖVE's fault. I'm unable to reproduce it.[Edit: see below] So the question is whom should you report it against?

Well, there's the possibility that there's some kind of flag indicating whether the input name is URL-encoded and LÖVE is not honouring it.

Edit: I could not reproduce it by dragging from xfe, but I could by dragging from the GTK2 file dialog. Now I have a basis to investigate further.

[raidho36]

So some file managers escape Unicode and some don't. I assume that has to do with file:// protocol which is allowed to escape non ASCII characters, or in fact any characters. What I think is proper solution is for LÖVE to unescape file paths internally before output.

[pgimeno]

I've traced the problem to the SDL library. It seems fixed in a newer version.

[easy82]

I've traced the problem to the SDL library. It seems fixed in a newer version.

Cool, thanks for the investigation!

The only problem with this is that the following code will not read the file with special characters in its full path:
main.lua

Code: Select all

function love.filedropped(file)
  local ok, err = file:open("r")
  local data = nil
  
  if not ok then
    print(err)
  else
    data = file:read()
    file:close()
  end
end

So you have to make a new file object or use the LUA io.* functions instead.

[pgimeno]

You can always upgrade your SDL

[zorg]

True, but did something like this work?

Code: Select all

function love.filedropped(file)
  local file = love.filesystem.newFile(file:getFilename():gsub("+"," "):gsub("%%(%x%x)",function (h)
    return string.char(tonumber(h, 16))
  end))
  local ok, err = file:open("r")
  if not ok then
    print(err)
    return
  end
  local data = file:read()
  file:close()
end

Also, earlier you wrote "/home/user/Let%F6lt%E9sek/test.txt", using singular escaped chars (meaning the encoding was not utf-8 but some 8-bit codepage (Latin-2 or equivalent), but later, positive07 wrote "Let%C3%B6lt%C3%A9sek", which was the escaped utf-8 encoding. Technically those two are separate issues, would be nice to know whether the second case also happens, or not.