3

I am writing a major mode derived from f90-mode in order to fontify the syntax of a commonly used Fortran preprocessor, fypp (similar to jinja2 for HTML templating). Here's a contrived example of what that looks like, with plain f90-mode highlighting on the left and my fypp-mode on the right.

f90 vs fypp

The #: directive and the ${...}$ evaluator are new syntax atop Fortran, which is why f90-mode does not color them properly. I want my major mode to inherit the fontifications defined in f90-mode.el and add to them. Here's a minimal but representative implementation

;;; fypp-mode.el

(setq fypp-highlights
      '(("#:set" . font-lock-preprocessor-face)
        ("\\${\\|}\\$" . font-lock-preprocessor-face)))

(define-derived-mode fypp-mode f90-mode "fypp"
  "Major mode for editing Fortran code with fypp extensions"
  (font-lock-add-keywords nil fypp-highlights))

I have two complaints with this, seen in the right-side image above:

  1. The ${...}$ syntax in the subroutine name declaration breaks the fontification of the subroutine name.

  2. The ${...}$ inside a string literal does not get fontified, but I would like it to.

Issue #1 I am able to solve by modifying the syntax table entries for { and } to make them word class instead of parentheses. I do wonder if this is likely to bite me later, though.

Issue #2 is the trickier matter for me. Normally, you want strings fontified as strings, even if they happen to contain keywords. Here, though, the string "Hello, ${whom}$!" will get expanded by the preprocessor to "Hello, world!", so I want the ${...}$ to be fontified. I still want string literals to be fontified, I just want my preprocessing tokens to "take precedence". But I don't know how to do this cleanly, if at all.

2

For issue #1, setting { and } to word syntax is wrong. You can give them "symbol constituent" syntax, but not "word" (otherwise you break the expected behavior of M-f and friends). This said, it may indeed bite you later, depending on what the rest of f90.el does.

For Issue #2, you simply want to use the OVERRIDE field in your font-lock rule. See C-h o font-lock-keywords for details.

  • I see, good call on putting curly braces as symbols rather than words (they are not part of Fortran's syntax, so I expect not to step on any landmines.) Also, thanks for pointing out the OVERRIDE field, that's exactly the kind of thing I was searching for! – Endulum Jul 8 '20 at 6:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.