Re: elisp: Replacing dots in strings (. -> \.)

[Jeronimo Pellegrini <j_p@aleph0.info>]

On Sun, Jul 08, 2007 at 03:27:14PM +0200, Nikolaj Schumacher wrote:
> Jeronimo Pellegrini <j_p@aleph0.info> wrote:
> 
> > ;; Adds one symbol to be highlighted as special-face.
> > ;;
> > (defun add-special-at-point ()
> >   "Adds the symbol at point to the list of symbols to be highlighted
> >    with special-face."
> >   (interactive)
> >   (let ((name (symbol-name (symbol-at-point))))
> >     (message name) 
> >     (let ((newname (replace-regexp-in-string "\\." "\\\\." name t t)))
> >       (font-lock-add-keywords nil `((,name . 'special-face))))))
> 
> Shouldn't you pass newname instead of name to `font-lock-add-keywords'?

Ouch! I read it several times and missed that.

> > This doesn't work really well, because when I write a symbol that starts
> > with a dot:
> >
> > .my_special_var
> 
> Beware that `symbol-at-point' will only include that period, if it has
> symbol syntax in the current buffer.

Yes, that's OK -- I'll use this with Lisp mode.

> > (replace-regexp-in-string "\\." "\\\." "aaa . bbb" t t)
> >       ==> "aaa \\. bbb"
> >
> > Now, that confuses me since I was explicitly asking the replacement to
> > be treated literally. What I am doing wrong?
> 
> You're asking the function to treat the input literally, yet the
> backslash has a special meaning while being _parsed_.  That is "\\" is a
> string with just one backslash.  Try `insert' on those results, and
> there will be fewer of them.
> The reason some of them are missing, is that Emacs apparently ignores
> the single backslash in "\.", which is actually undefined

Ah, now I see what happened.

> > Or, is there an easier way to replace the dots in a string so that
> > function will work properly?
> 
> Yes.  You're probably looking for `regexp-quote'.

And that definetely fixed it. Thanks!

> Also, you might want to check out this package of mine:
> http://nschum.de/src/emacs/highlight-symbol/

Thanks -- that's also very nice! I see it uses hi-lock, so I'll probably
keep my function until I find some time to switch to hi-lock (I use
font-lock currently).

J.