created 2005 · complexity intermediate · author Wu Yongwei · version 6.0
Currently fileencoding does not work well in a modeline. When Vim sees the fileencoding setting, the file has already been read, so setting 'fileencoding' in a modeline normally has no effect until you write the file. That will cause even more trouble as, for example, if the default encoding is latin1 and the modeline has
fileencoding=utf-8, then the resulting file will be a utf-8 file converted as latin1 to utf-8 again.
Vim does as best as it can to detect your encoding, if you place its name in your 'fileencodings' option. However, Vim cannot distinguish between multiple 8-bit encodings, so only the first one of those will be used. You can install a plugin like script#2721 to automatically detect the correct encoding in many cases where 'fileencodings' fails, but you can also just add a few lines to your .vimrc to allow modelines to work as intended:
function! CheckFileEncoding() if exists('b:fenc_at_read') && &fileencoding != b:fenc_at_read exec 'e! ++enc=' . &fileencoding unlet b:fenc_at_read endif endfunction au BufRead *.txt let b:fenc_at_read=&fileencoding au BufWinEnter *.txt call CheckFileEncoding()
Change the "*.txt" part if wanted. See Working with Unicode if your Vim is not yet configured for working with multiple encodings.
This script will store off the fileencoding which Vim detects after reading the file, but before processing modelines, and then compare it to the fileencoding after processing modelines. If they differ, it will re-read the file with the new fileencoding value. For this step
:e! must be used, because when the fileencoding changes the buffer is marked as modified. This is made less dangerous because we unlet the stored fileencoding value, so that we only issue the
e! command if this is the first time BufWinEnter fired after reading the file.