doc: Add support for MyST roles

Officially, only the manpage role is supported at the moment.

Unlike in rST, the syntax uses braces instead of colons:

    {manpage}`nix.conf(5)`

doc/Makefile

@@ -6,8 +6,12 @@ pandoc_media_dir = media
 # NOTE: Keep in sync with NixOS manual (/nixos/doc/manual/md-to-db.sh).
 # TODO: Remove raw-attribute when we can get rid of DocBook altogether.
 pandoc_commonmark_enabled_extensions = +attributes+fenced_divs+footnotes+bracketed_spans+definition_lists+pipe_tables+raw_attribute
+# Not needed:
+# - docbook-reader/citerefentry-to-rst-role.lua (only relevant for DocBook → MarkDown/rST/MyST)
 pandoc_flags = --extract-media=$(pandoc_media_dir) \
 	--lua-filter=$(PANDOC_LUA_FILTERS_DIR)/diagram-generator.lua \
+	--lua-filter=build-aux/pandoc-filters/myst-reader/roles.lua \
+	--lua-filter=build-aux/pandoc-filters/docbook-writer/rst-roles.lua \
 	--lua-filter=build-aux/pandoc-filters/docbook-writer/labelless-link-is-xref.lua \
 	-f commonmark$(pandoc_commonmark_enabled_extensions)+smart
 

doc/build-aux/pandoc-filters/docbook-reader/citerefentry-to-rst-role.lua

@@ -0,0 +1,23 @@
+--[[
+Converts Code AST nodes produced by pandoc’s DocBook reader
+from citerefentry elements into AST for corresponding role
+for reStructuredText.
+
+We use subset of MyST syntax (CommonMark with features from rST)
+so let’s use the rST AST for rST features.
+
+Reference: https://www.sphinx-doc.org/en/master/usage/restructuredtext/roles.html#role-manpage
+]]
+
+function Code(elem)
+  elem.classes = elem.classes:map(function (x)
+    if x == 'citerefentry' then
+      elem.attributes['role'] = 'manpage'
+      return 'interpreted-text'
+    else
+      return x
+    end
+  end)
+
+  return elem
+end

doc/build-aux/pandoc-filters/docbook-writer/rst-roles.lua

@@ -0,0 +1,36 @@
+--[[
+Converts AST for reStructuredText roles into corresponding
+DocBook elements.
+
+Currently, only a subset of roles is supported.
+
+Reference:
+  List of roles:
+    https://www.sphinx-doc.org/en/master/usage/restructuredtext/roles.html
+  manpage:
+    https://tdg.docbook.org/tdg/5.1/citerefentry.html
+  file:
+    https://tdg.docbook.org/tdg/5.1/filename.html
+]]
+
+function Code(elem)
+  if elem.classes:includes('interpreted-text') then
+    local tag = nil
+    local content = elem.text
+    if elem.attributes['role'] == 'manpage' then
+      tag = 'citerefentry'
+      local title, volnum = content:match('^(.+)%((%w+)%)$')
+      if title == nil then
+        -- No volnum in parentheses.
+        title = content
+      end
+      content = '<refentrytitle>' .. title .. '</refentrytitle>' .. (volnum ~= nil and ('<manvolnum>' .. volnum .. '</manvolnum>') or '')
+    elseif elem.attributes['role'] == 'file' then
+      tag = 'filename'
+    end
+
+    if tag ~= nil then
+      return pandoc.RawInline('docbook', '<' .. tag .. '>' .. content .. '</' .. tag .. '>')
+    end
+  end
+end

doc/build-aux/pandoc-filters/myst-reader/roles.lua

@@ -0,0 +1,29 @@
+--[[
+Replaces Str AST nodes containing {role}, followed by a Code node
+by a Code node with attrs that would be produced by rST reader
+from the role syntax.
+
+This is to emulate MyST syntax in Pandoc.
+(MyST is a CommonMark flavour with rST features mixed in.)
+
+Reference: https://myst-parser.readthedocs.io/en/latest/syntax/syntax.html#roles-an-in-line-extension-point
+]]
+
+function Inlines(inlines)
+  for i = #inlines-1,1,-1 do
+    local first = inlines[i]
+    local second = inlines[i+1]
+    local correct_tags = first.tag == 'Str' and second.tag == 'Code'
+    if correct_tags then
+      -- docutils supports alphanumeric strings separated by [-._:]
+      -- We are slightly more liberal for simplicity.
+      local role = first.text:match('^{([-._+:%w]+)}$')
+      if role ~= nil then
+        inlines:remove(i)
+        second.attributes['role'] = role
+        second.classes:insert('interpreted-text')
+      end
+    end
+  end
+  return inlines
+end

doc/build-aux/pandoc-filters/myst-writer/roles.lua

@@ -0,0 +1,25 @@
+--[[
+Replaces Code nodes with attrs that would be produced by rST reader
+from the role syntax by a Str AST node containing {role}, followed by a Code node.
+
+This is to emulate MyST syntax in Pandoc.
+(MyST is a CommonMark flavour with rST features mixed in.)
+
+Reference: https://myst-parser.readthedocs.io/en/latest/syntax/syntax.html#roles-an-in-line-extension-point
+]]
+
+function Code(elem)
+  local role = elem.attributes['role']
+
+  if elem.classes:includes('interpreted-text') and role ~= nil then
+    elem.classes = elem.classes:filter(function (c)
+      return c ~= 'interpreted-text'
+    end)
+    elem.attributes['role'] = nil
+
+    return {
+      pandoc.Str('{' .. role .. '}'),
+      elem,
+    }
+  end
+end

doc/contributing/contributing-to-documentation.chapter.md

@@ -52,6 +52,11 @@ Additionally, the following syntax extensions are currently used:
 
   This syntax is taken from [MyST](https://myst-parser.readthedocs.io/en/latest/using/syntax.html#targets-and-cross-referencing).
 
+- []{#ssec-contributing-markup-inline-roles}
+  If you want to link to a man page, you can use ``{manpage}`nix.conf(5)```, which will turn into {manpage}`nix.conf(5)`.
+
+  This syntax is taken from [MyST](https://myst-parser.readthedocs.io/en/latest/syntax/syntax.html#roles-an-in-line-extension-point). Though, the feature originates from [reStructuredText](https://www.sphinx-doc.org/en/master/usage/restructuredtext/roles.html#role-manpage) with slightly different syntax.
+
 - []{#ssec-contributing-markup-admonitions}
   **Admonitions**, set off from the text to bring attention to something.
 

nixos/doc/manual/md-to-db.sh

@@ -12,7 +12,12 @@ pushd $DIR
 # TODO: Remove raw-attribute when we can get rid of DocBook altogether.
 pandoc_commonmark_enabled_extensions=+attributes+fenced_divs+footnotes+bracketed_spans+definition_lists+pipe_tables+raw_attribute
 pandoc_flags=(
-  # media extraction and diagram-generator.lua not needed
+  # Not needed:
+  # - diagram-generator.lua (we do not support that in NixOS manual to limit dependencies)
+  # - media extraction (was only required for diagram generator)
+  # - docbook-reader/citerefentry-to-rst-role.lua (only relevant for DocBook → MarkDown/rST/MyST)
+  "--lua-filter=$DIR/../../../doc/build-aux/pandoc-filters/myst-reader/roles.lua"
+  "--lua-filter=$DIR/../../../doc/build-aux/pandoc-filters/docbook-writer/rst-roles.lua"
   "--lua-filter=$DIR/../../../doc/build-aux/pandoc-filters/docbook-writer/labelless-link-is-xref.lua"
   -f "commonmark${pandoc_commonmark_enabled_extensions}+smart"
   -t docbook