Emacs: Typing dashes in html mode (update for Emacs 24) (Shallow Thoughts)

Emacs: Typing dashes in html mode (update for Emacs 24)

Back in 2006, I wrote an article on making a modified copy of sgml-mode.el to make it possible to use double-dashed clauses -- like this -- in HTML without messing up auto-fill mode.

That worked, but the problem is that if you use your own copy of sgml-mode.el, you miss out on any other improvements to HTML and SGML mode. There have been some good ones, like smarter rewrap of paragraphs. I had previously tried lots of ways of customizing sgml-mode without actually replacing it, but never found a way.

Now, in emacs 24.5.1, I've found a easier way that seems to work. The annoying mis-indentation comes from the function sgml-comment-indent-new-line, which sets variables comment-start, comment-start-skip and comment-end and then calls comment-indent-new-line.

All I had to do was redefine sgml-comment-indent-new-line to call comment-indent-new-line without first defining the comment characters:

(defun sgml-comment-indent-new-line (&optional soft)
  (comment-indent-new-line soft))

Finding emacs source

I wondered if it might be better to call whatever underlying indent-new-line function comment-indent-new-line calls, or maybe just to call (newline-and-indent). But how to find the code of comment-indent-new-line?

Happily, describe-function (on C-h f, or if like me you use C-h for backspace, try F-1 h) tells you exactly what file defines a function, and it even gives you a link to click on to view the source. Wonderful!

It turned out just calling (newline-and-indent) wasn't enough, because sgml-comment-indent-new-line typically calls comment-indent-new-line when you've typed a space on the end of a line, and that space gets wrapped and then messes up indentation. But you can fix that by copying just a couple of lines from the source of comment-indent-new-line:

(defun sgml-comment-indent-new-line (&optional soft)
  (save-excursion (forward-char -1) (delete-horizontal-space))
  (delete-horizontal-space)
  (newline-and-indent))

That's a little longer than the other definition, but it's cleaner since comment-indent-new-line is doing all sorts of extra work you don't need if you're not handling comments. I'm not sure that both of the delete-horizontal-space lines are needed: the documentation for delete-horizontal-space says it deletes both forward and backward. But I have to assume they had a good reason for having both: maybe the (forward-char -1) is to guard against spurious spaces already having been inserted in the next line. I'm keeping it, to be safe.