Fix gfm_auto_identifiers behavior with emojis.

Closes #5813.

Note that we also now use emoji names for emojis
when `ascii_identifiers` is enabled.

MANUAL.txt

@@ -2575,6 +2575,7 @@ Changes the algorithm used by `auto_identifiers` to conform to
 GitHub's method.  Spaces are converted to dashes (`-`),
 uppercase characters to lowercase characters, and punctuation
 characters other than `-` and `_` are removed.
+Emojis are replaced by their names.
 
 ## Math Input
 

src/Text/Pandoc/Shared.hs

@@ -463,8 +463,15 @@ isPara _        = False
 -- letters, digits, and the characters _-.
 inlineListToIdentifier :: Extensions -> [Inline] -> String
 inlineListToIdentifier exts =
-  dropNonLetter . filterAscii . toIdent . stringify
+  dropNonLetter . filterAscii . toIdent . stringify . walk unEmojify
   where
+    unEmojify :: [Inline] -> [Inline]
+    unEmojify
+      | extensionEnabled Ext_gfm_auto_identifiers exts ||
+        extensionEnabled Ext_ascii_identifiers exts = walk unEmoji
+      | otherwise = id
+    unEmoji (Span ("",["emoji"],[("data-emoji",ename)]) _) = Str ename
+    unEmoji x = x
     dropNonLetter
       | extensionEnabled Ext_gfm_auto_identifiers exts = id
       | otherwise = dropWhile (not . isAlpha)

test/command/5813.md

@@ -0,0 +1,6 @@
+```
+% pandoc -f gfm
+### Jekyll Plugins & Gems :gem:
+^D
+<h3 id="jekyll-plugins--gems-gem">Jekyll Plugins &amp; Gems <span class="emoji" data-emoji="gem">💎</span></h3>
+```