Re: Yet another washing function.

[Greg Stark <gsstark@MIT.EDU>]

Please try to avoid generating garbage whenever possible, the regular
expression functions generate a lot of wasted memory and the loops in elisp
are very inefficient compared to using primitives that do loops in C. Also,
replace-match probably moves the gap to that point which means you're doing
the equivalent of memoving the entire article a small bit at a time. 

I've noticed that article washing is a noticeable delay. It might be worth
looking through the existing washing functions for code that generates excess
garbage, do loops in elisp that can be done in C with primitives, or move the
gap unecessarily. Here's an implementation that does none of these:

(subst-char-in-region (point-min) (point-max) ?\221 ?`)
(subst-char-in-region (point-min) (point-max) ?\222 ?')
(subst-char-in-region (point-min) (point-max) ?\223 ?\")
(subst-char-in-region (point-min) (point-max) ?\224 ?\")

An even better alternative might be to use translate-region but it's a bit
awkward and likely to cause problems with XEmacs 20 and Emacs 20 where
characters and integers are distinct. It may not actually be faster than a few
loops through the text, though i'm not sure if the break-even point would be
more or less than four.

(let ((x (make-string 225 ?x))
      (i -1))
  (while (< (incf i) (length x))
    (aset x i i))
  (aset x ?\221 ?`)
  (aset x ?\222 ?')
  (aset x ?\223 ?\")
  (aset x ?\224 ?\")
  x
  (translate-region (point-min) (point-max) x))