[Caml-list] Ocaml and Windows' notion of Unicode file names

[Andreas Rossberg <rossberg@mpi-sws.org>]

Does anybody have advice for dealing with unicode file names under Windows? In particular, my problem is that the following invariant

  let files = Array.to_list (Sys.readdir “.”) in
  assert (List.for_all Sys.file_exists files)

does not hold under Windows, because some of the file names returned by readdir are not valid.

The Unix emulation module has the same issue.

This is only partially Ocaml’s fault — I see that it implements Sys.readdir and Unix.readdir using Windows’ ASCII _findfirst and _findnext functions, which apparently just replace large code points with ‘?’. And I’m not sure how realistic it would be for Ocaml to switch to Windows' wchar APIs and, say, UTF8-encode/decode file names on its side.

But anyway, I’m wondering who else has run into this, and has an idea how to cope. Because I came up blank.

Thanks,
/Andreas