Using `links' to process HTML

[Hrvoje Niksic <hniksic@arsdigita.com>]

[ Note: I'm not on the list.  If you want me to see your replies, add Cc. ]

I've been using this for some time now and want to share it with
others.  It hooks into the processing of "text/html" and renders it by
noninteractively calling the textual browser `links'.  Although this
doesn't support images, colors, forms and whatnot, those are the
things I turn off anyway.  It does support tables, however.

Overall it does a much better job of rendering HTML than w3 does.  My
spams finally come out right before I ignore them.

Here is what I put in `.gnus'.  It relies on the program `links' being
somewhere on the PATH.  If you try it, please let me know if it works
for you.

(eval-after-load "mm-decode"
  '(push '("text/html"
	   my-html-handler
	   (lambda (handle) t))
	 mm-inline-media-tests))

;; If you're not running under XEmacs...
(unless (fboundp 'temp-directory)
  (defun temp-directory ()
    (or (getenv "TMPDIR") "/tmp")))

(defun my-html-handler (handle)
  (let ((source (mm-get-part handle))
	(tmpfile (make-temp-name (expand-file-name
				  "www" (temp-directory))))
	text)
    (with-temp-buffer
      (insert source)
      (write-region (point-min) (point-max) tmpfile nil 'silent)
      (delete-region (point-min) (point-max))
      (unwind-protect
	  ;; It would be nice if we could use `call-process-region' to
	  ;; feed the HTML to links's stdin thus avoiding the tmpfile.
	  ;; But `links -dump /dev/stdin' doesn't work when stdin is a
	  ;; pipe.
	  (call-process "links" nil t nil "-dump" tmpfile)
	(delete-file tmpfile))

      ;; Delete the annoying three spaces preceding each line of links
      ;; output.
      (goto-char (point-min))
      (while (re-search-forward "^   " nil t)
	(delete-region (match-beginning 0) (match-end 0)))

      (setq text (buffer-string)))

    (mm-insert-inline handle text)))