Vim Tips Wiki
Tip 1074 Printable Monobook Previous Next

created 2005 · complexity advanced · author Wu Yongwei · version 6.0

If one needs to edit files encoded in multiple legacy encodings, then the Vim fileencodings option cannot help much. Some hacks can be used to put the file encoding in the file (see VimTip911). However, in the case of HTML files, the encoding information is often in the HTML file already, especially for non-Latin1 Web pages, for example:

<meta http-equiv="Content-Type" content="text/html; charset=gb2312" >

The following code can be put in vimrc to detect and use such an encoding specification:

if has('autocmd')
  function! ConvertHtmlEncoding(encoding)
    if a:encoding ==? 'gb2312'
      return 'cp936' " GB2312 imprecisely means CP936 in HTML
    elseif a:encoding ==? 'iso-8859-1'
      return 'latin1' " The canonical encoding name in Vim
    elseif a:encoding ==? 'utf8'
      return 'utf-8' " Other encoding aliases should follow here
      return a:encoding

  function! DetectHtmlEncoding()
    if &filetype != 'html'
    normal m`
    normal gg
    if search('\c<meta[ \t\n]\+http-equiv=\("\?\)Content-Type\1[ \t\n]\+content="text/html;[ \t\n]*charset=[-A-Za-z0-9_]\+"[ \t\n]*>') != 0
      let reg_bak=@"
      normal y$
      let charset=matchstr(@", 'text/html; charset=\zs[-A-Za-z0-9_]\+')
      let charset=ConvertHtmlEncoding(charset)
      normal ``
      let @"=reg_bak
      if &fileencodings == ''
        let auto_encodings=',' . &encoding . ','
        let auto_encodings=',' . &fileencodings . ','
      if charset !=? &fileencoding &&
            \auto_encodings =~ ',' . &fileencoding . ','
        silent! exec 'e ++enc=' . charset
      normal ``

  " Detect charset encoding in an HTML file
  au BufReadPost *.htm* nested call DetectHtmlEncoding()

Please notice that the nested autocommand is used to ensure the syntax highlighting is OK and the remembered cursor position is still kept.

It is recommended to use set encoding=utf-8 in order to ensure successful encoding conversion.



The following source code form is common for generated pages:

<meta content="text/html …" http-equiv="Content-Type" >

This form will not be recognised.

It would be reasonable to limit the search to the document head, expressed as an absolute characters to scan. This restriction will cause pages containing lots of comments and white space in head to be left alone. I do not think this is much of a problem.

Version vim7.3_v7 or higher of the :TOhtml plugin distributed with Vim includes an autoload function you could call that does a much more complete HTML-charset to Vim encoding conversion. --Fritzophrenic 16:30, November 15, 2010 (UTC)

This is now done in the AutoFenc.vim plugin mentioned above. For an example, see the plugin code. --Fritzophrenic 22:24, April 4, 2011 (UTC)