ls() method parses incorrectly when a filename contain a whitespace (FTP)

[ArnauCForgeFlow]

I encountered an issue with FTPFileSystem regarding whitespace handling. When a file or directory name on the server contains spaces, the ls() method fails to parse the full name correctly, often truncating the result.

Assume a directory /data exists on the FTP server containing a file named "my file.txt".

Actual Output: The filename is truncated (split by space):
['/data/file.txt']

Expected Output: It should return the full filename:
['/data/my file.txt']

[martindurant]

Can you check with detail=True, please?

This line upstream looks suspicious to me.

[ArnauCForgeFlow]

Output still truncated with detail=True

[martindurant]

I was wondering if the rest of the fields look ok, or if some of the name ended up elsewhere

[ArnauCForgeFlow]

The other fields look correct. I checked them and I don't see the missing part of the name in any of them, it seems to be simply lost.

[martindurant]

The following test passes successfully for me:

def test_list_space(ftp_writable):
    host, port, user, pw = ftp_writable
    fs = FTPFileSystem(host, port, user, pw)
    fs.pipe("/space file", b"hello")
    assert fs.cat("/space file") == b"hello"
    assert "space file" in fs.ls("/", False)

[ArnauCForgeFlow]

I've been investigating and I suspect the issue is that my FTP doesn't support MLSD. Could this be the reason why your output is correct and mine isn't?

[martindurant]

Could well be. In that case, the implementation falls back to DIR and then line.split(), see fsspec.implementations.ftp._mlsd2. Can you debug what happens within that function?

[ArnauCForgeFlow]

The issue is indeed caused by line.split() when handling filenames with spaces.

For a file named "my file.pdf", the method line.split() produces the following output:

['-rw-r--r--', '1', 'owner', 'group', '38283', 'Jan', '12', '07:08', 'my', 'file.pdf']

Since the implementation uses split_line[-1], it only retrieves 'file.pdf' and loses the rest of the name.

[martindurant]

OK, so the question becomes: what can we do about this? DIR does not promise any particular format. For your case, this looks like posix, so maybe we can at least make this one work, like " ".join(split_line[8:])

[martindurant]

@ArnauCForgeFlow , do you think you have enough information to propose a PR fix?