ſ (long s) and hyphenation

Question

I'm using UTF-8 so it is – if one is using the right font – no problem to implement the different types of s in older German texts (s, ſ, ß). Unfortunately the hyphenation breaks because LaTeX does not know that ſ has to be dealt with just the same as it would do when dealing with "s".

MWE:

\documentclass{article}
\usepackage[ngerman]{babel}

\begin{document}
XXX Gesellschaft Gesellschaft Gesellschaft Gesellschaft Gesellschaft Gesellschaft Gesellschaft Gesellschaft Gesellschaft

XXX Geſellſchaft Geſellſchaft Geſellſchaft Geſellſchaft Geſellſchaft Geſellſchaft Geſellſchaft Geſellſchaft Geſellſchaft

\hyphenation{Ge-ſell-ſchaft}

XXX Geſellſchaft Geſellſchaft Geſellſchaft Geſellſchaft Geſellſchaft Geſellſchaft Geſellſchaft Geſellſchaft Geſellſchaft

\end{document}

As one can see at the end of the line, I have to add the correct hyphenation manually. Any idea how I can solve this?

Answer 1

For LuaTeX here is an implementation of David Carlisles idea to create a hypenate callback. It works by replacing every ſ with a marked s before hyphenation and then recovering the original characters after hyphenation:

\documentclass{article}
\usepackage[ngerman]{babel}
\usepackage{luacode}
\begin{luacode*}
local sattr = luatexbase.new_attribute("longsattr")
local disc = node.id'disc'
print('DISC', disc)

local function long_to_s(head, tail)
  for n in node.traverse(head) do
    if n == tail then break end
    if n.id == disc then
    print(n)
      long_to_s(n.pre)
      long_to_s(n.post)
      long_to_s(n.replace)
    end
    if n.char == 383 then
      n.char = 115
      node.set_attribute(n, sattr, 383)
    end
  end
end
local function s_to_long(head, tail)
  for n in node.traverse(head) do
    if n == tail then break end
    if n.id == disc then
      s_to_long(n.pre)
      s_to_long(n.post)
      s_to_long(n.replace)
    end
    local a = node.get_attribute(n, sattr)
    if a then
      n.char = a
      node.unset_attribute(n, sattr)
    end
  end
end
local function myhyph(head, tail)
  long_to_s(head, tail)
  lang.hyphenate(head, tail)
  s_to_long(head, tail)
end
luatexbase.add_to_callback("hyphenate",myhyph,"hyphenate with modified s")
\end{luacode*}
\begin{document}
XXX Gesellschaft Gesellschaft Gesellschaft Gesellschaft Gesellschaft Gesellschaft Gesellschaft Gesellschaft Gesellschaft

XXX Geſellſchaft Geſellſchaft Geſellſchaft Geſellſchaft Geſellſchaft Geſellſchaft Geſellſchaft Geſellſchaft Geſellſchaft
\end{document}

LuaTeX also allows you to manipulate the hyphenation pattern during a run, so you can also use (this is an automated version of David Carlisles choice (b)):

\documentclass{article}
\usepackage[ngerman]{babel}
\usepackage{luacode}
\begin{luacode*}
  local l = lang.new(tex.language)
  l:patterns(l:patterns():gsub('s', 'ſ'))
\end{luacode*}
\begin{document}
XXX Gesellschaft Gesellschaft Gesellschaft Gesellschaft Gesellschaft Gesellschaft Gesellschaft Gesellschaft Gesellschaft

XXX Geſellſchaft Geſellſchaft Geſellſchaft Geſellſchaft Geſellſchaft Geſellſchaft Geſellſchaft Geſellſchaft Geſellſchaft
\end{document}

Answer 2

A simple way to do this is to choose a font that supports ſ as an open type character variant, e.g., EB Garamond. Then you can just select that variant when you need it.

(Re-reading the comments above, I see this corresponds to option (c) from David Carlisle, which you said wasn't suitable, but this MWE shows you can have both kinds of s with this method.)

Update showing iſt and ſelbes.

\documentclass{article}
\usepackage[ngerman]{babel}
\babelfont{rm}{EB Garamond}
\begin{document}
XXX Gesellschaft Gesellschaft Gesellschaft Gesellschaft Gesellschaft Gesellschaft Gesellschaft Gesellschaft Gesellschaft

\addfontfeature{CharacterVariant=1}

XXX Gesellschaft Gesellschaft Gesellschaft Gesellschaft Gesellschaft Gesellschaft Gesellschaft Gesellschaft Gesellschaft

ist selbes
\end{document}