Emacs syntax highlighting of optional parts

Question

I'm trying to write a derived mode from fundamental-mode. Assuming that I have this regexp : A ((foo)bar)? B, how can I tell emacs to use the following faces ?

• font-lock-keyword-face on A
• font-lock-warning-face on foo (but not bar)
• font-lock-constant-face on B

I have tried to use this following code :

(defvar myregexp
  "\\(A\\) \\(?:\\(foo\\)bar \\)?\\(B\\)")

(setq mylang-font-lock-keywords `(
  (, myregex 1 font-lock-keyword-face)
  (, myregex 2 font-lock-warning-face)
  (, myregex 3 font-lock-constant-face)
))

But it does not work with the string A B (emacs report a missing capture).

Answer 1

Use subexp highlighters with laxmatch enabled to have font-lock-mode ignore missing groups in matches:

(setq mylang-font-lock-keywords
      `((,myregexp (1 font-lock-keyword-face)
              (2 font-lock-warning-face nil t)
              (3 font-lock-constant-face))))

The forth element of each subexp highlighter is the laxmatch argument. If t font-lock-mode ignores this highlighter if the corresponding group in the first element is not found in the match result of myregexp.

See Search-based Fontification in the Emacs Lisp manual for more information.

Answer 2

The way your regexp is specified, you'd need two spaces between A and B. Either pull the second space inside the optional foobar part or use * or ? after it.