How can I fontify a symbol by regexp match group (as in a naming convention)?

Question

I want to add better support for expl3 to AUCTeX. I have regular expressions that properly group according to the naming conventions for:

• functions (\\(?:__)?[a-z]+_)([a-z_]+)(:[nNpTFwcVvxof]*),
• variables (\\[cgl]__?)([a-z][a-z_]+)(_[A-Za-z]+), and
• key-value segments \s*([a-z\-]+)\s+(\.[a-z_]+):([nNpTFwcVvxof]*)\s+=

How can I apply faces to both

• the match as a whole and
• the individual groups (overriding the whole match where appropriate)?

Just a proof-of-concept using the function-regexp would be very helpful :)

rx Forms

(setq LaTeX-expl3-function-regexp
      (rx symbol-start
          (group "\\" (optional "__") (+ alpha) "_")
          (group (+ (any alpha "_")))
          (group ":" (* (any "nNpTFwcVvxofD")))
          symbol-end)

      LaTeX-expl3-variable-regexp
      (rx symbol-start
          (group "\\" (any "cgl") "_" (optional "_"))
          (group (or (seq (any alpha) "_") "@@_")) 
          (group (any alpha) (+ (any alpha "_")))
          (group "_" (+ alpha))
          symbol-end)

      LaTeX-expl3-keyval-regexp
      (rx (* whitespace)
          (group (any alpha "-"))
          (+ whitespace)
          (group "." (+ (any alpha)))
          (group ":" (* (any "nNpTFwcVvxofD")))
          (+ whitespace) "="))

Answer 1

You can use font-lock-add-keywords in a major mode hook to add new rules. The format of the rules is described in the help text to the variable font-lock-keywords. In short, your code would look something like:

(defun my-LaTeX-expl3-hook ()
  (font-lock-add-keywords nil
                          `((,LaTeX-expl3-function-regexp
                             (0 face-for-whole-match prepend)
                             (1 face-for-group-one prepend)
                             (2 face-for-group-two prepend))
                            (,LaTeX-expl3-variable-regexp
                             (0 font-lock-variable-name-face prepend))
                            (,LaTeX-expl3-keyval-regexp
                             (0 font-lock-keyword-face prepend)))
                          t)) ;; Append
(add-hook 'latex-mode-hook 'my-LaTeX-expl3-hook)

Here, I've used 0 to highlight the entire match. If you would like to highlight only parts, supply a integer 1, 2, 3 etc. that corresponds the the parenthesis you would like to highlight. You can supply more than one such rule to each regexp.

The prepend flag indicates that this takes precedence over already present fontification. (However, unlike passing t, it leaves, for example, the background intact even if it overrides the foreground.)

Finally, if you find that you keywords doesn't work as intended, I would recommend using font-lock-studio, an interactive debugger that lets you step the font-lock machinery one step at a time, it highlights matches in the source buffer and in the regexps using a palette of background colors etc.