feat(typesetters): Support for speaker change introduced by em-dash

Some conventions denote speaker changes in dialogues by starting
a paragraph with an em-dash.
In these cases, the space after the em-dash shall be fixed by
default.

documentation/c07-settings.sil

@@ -189,6 +189,11 @@ Note that non-breaking spaces (U+00A0), following the guidelines of Unicode Anne
 
 If you want to disable this behavior, the \autodoc:setting{languages.fixedNbsp} setting may be set to \code{true} to enforce fixed-width non-breaking spaces.
 
+Some typography conventions use an em-dash at the start of a paragraph line to denote a speaker change in a dialogue.
+This is the case in particular in French and Turkish typography.
+By default, all spaces following an em-dash at the beginning of a paragraph in your input are replaced by a single \em{fixed} inter-word space, so that subsequent dialogue lines all start identically, while other inter-word spaces may still be variable for justification purposes.
+To cancel this behavior, the \autodoc:setting{typesetter.fixedSpacingAfterInitialEmdash} setting may be set to \code{false}.
+
 \subsection{Letter spacing settings}
 
 You can also put spaces in between \em{letters} with the \autodoc:setting{document.letterspaceglue} setting.

typesetters/base.lua

@@ -145,6 +145,13 @@ function typesetter.declareSettings(_)
     default = false,
     help = "When true, a warning is issued when a soft hyphen is encountered"
   })
+
+  SILE.settings:declare({
+    parameter = "typesetter.fixedSpacingAfterInitialEmdash",
+    type = "boolean",
+    default = true,
+    help = "When true, em-dash starting a paragraph is considered as a speaker change in a dialogue"
+  })
 end
 
 function typesetter:initState ()
@@ -330,6 +337,12 @@ function typesetter:endline ()
   SILE.documentState.documentClass.endPar(self)
 end
 
+-- Just compute once, to avoid unicode characters in source code.
+local speakerChangePattern = "^"
+   .. luautf8.char(0x2014) -- emdash
+   .. "[ " .. luautf8.char(0x00A0) .. luautf8.char(0x202F) -- regular space or NBSP or NNBSP
+   .. "]+"
+
 -- Takes string, writes onto self.state.nodes
 function typesetter:setpar (text)
   text = text:gsub("\r?\n", " "):gsub("\t", " ")
@@ -338,6 +351,21 @@ function typesetter:setpar (text)
       text = text:gsub("^%s+", "")
     end
     self:initline()
+
+    if SILE.settings:get("typesetter.fixedSpacingAfterInitialEmdash") and not SILE.settings:get("typesetter.obeyspaces") then
+      local speakerChange = false
+      local dialogue = luautf8.gsub(text, speakerChangePattern, function ()
+        speakerChange = true
+        return ""
+      end)
+      if speakerChange then
+        text = dialogue
+        local fontoptions = SILE.font.loadDefaults({})
+        local fiwsp = SILE.shaper:measureSpace(fontoptions).length -- fixed inter-word space
+        self:pushUnshaped({ text = luautf8.char(0x2014), options = fontoptions })
+        self:pushHorizontal(SILE.nodefactory.kern(fiwsp))
+      end
+    end
   end
   if #text >0 then
     self:pushUnshaped({ text = text, options= SILE.font.loadDefaults({})})