0

I want to write a mode for the Cadabra 2 computer algebra system, which has Python-like syntax, with an exception of having the hash character # as a wild card in expressions, in addition to being a comment starter like in regular Python. Consider the following code:

\dalembert{ A?? }::LaTeXForm(\Box A??).

{ \mu, \nu }::Indices(vector).
\dalembert{#}::Derivative.
\partial{#}::PartialDerivative.
{ A_{\mu} }::Depends(\partial{#}).

# QED Lagrangian
lagrangian:= (a/2) A_{\mu} \dalembert{ A_{\mu} }
    + (b/2) A_{\mu} \partial_{\nu}{\partial_{\mu}{ A_{\nu} }}
    + (1/2) m**2 A_{\mu} * A^{\mu};

Here, the QED Lagrangian line is a comment, but }::Derivative. is not. I want Emacs to interpret # as a comment delimiter only if it is not enclosed by curly or regular braces, and am struggling to implement that. The way I believe it should be done is with the syntax propertization function:

;;;###autoload
(define-derived-mode my-cadabra2-mode python-mode "cadabra2 mode"
  "Major mode for Cadabra 2 computer algebra system"
  (setq-local font-lock-defaults '((my-cadabra2-font-lock-keywords)))
  (setq-local syntax-propertize-function
              (syntax-propertize-rules ("\\(?<!{\s*\\)#\\(?!\s*}\\)" (1 "< ")))))

I believe my regular expression is correct, it should match # if not preceded by { and not followed by } with any whitespace in between. Emacs still uses any hash as a comment starter however.

How do I implement this properly?

3
  • 1
    Well, just trying to eval your function in my *scratch* buffer gives an (invalid-regexp) in your argument to syntax-propetize-rules.
    – nega
    May 14, 2022 at 18:42
  • 1
    Looks like you've got a negative-lookbehind and a negative-lookahead, both of which aren't supported. Also it looks like you're trying to use \s to match whitespace. You want \s- instead. See emacswiki.org/emacs/RegularExpression for a good primer on emacs regexes.
    – nega
    May 14, 2022 at 19:23
  • @nega I'm trying to make this work with the simplest example, I tried (rx (and "#" (not "}"))) which evaluates just fine and should catch the non-comments in the provided example, but still does not work when I load a file with the mode May 17, 2022 at 9:57

1 Answer 1

1

Based on your recent comment it is working, except for the fact that you're not overriding the comment character, you're adding properties to it. Effectively your (define-derived-mode ...) is saying "add this additional property to the comment character" and not "set the comment character to this property".

Additionally, you want to use 0 in your rule. The integer there is the matching sub-group of the regex in the rule, with 0 being "the whole match".

With your (slightly modified) sample text

# -*- mode: my-cadabra2 -*-

# bol comment
\dalembert{ A?? }::LaTeXForm(\Box A??). # inline comment

{ \mu, \nu }::Indices(vector).
\dalembert{#}::Derivative.
\partial{#}::PartialDerivative.
{ A_{\mu} }::Depends(\partial{#}).

# QED Lagrangian
lagrangian:= (a/2) A_{\mu} \dalembert{ A_{\mu} }
    + (b/2) A_{\mu} \partial_{\nu}{\partial_{\mu}{ A_{\nu} }}
    + (1/2) m**2 A_{\mu} * A^{\mu};

And this version of your defined of your define-derived-mode (note the modify-syntax-entry line)

;;;###autoload
(define-derived-mode my-cadabra2-mode python-mode "cadabra2 mode"
  "Major mode for Cadabra 2 computer algebra system"
  (setq-local font-lock-defaults '(python-font-lock-keywords))
  (modify-syntax-entry ?# ".") ;; set # to be punctuation
  (set (make-local-variable 'syntax-propertize-function)
              (syntax-propertize-rules ((rx (and "#" (not "}"))) (0 "<")))))

I get:

derived mode with working comment highlighting

1
  • Is there a function to force drawing the propertized syntax? It takes a while for Emacs to start displaying normal comments properly after startup now May 17, 2022 at 21:36

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

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