Re: Inefficiency in mm-uu.el

[Shenghuo ZHU <zsh@cs.rochester.edu>]

>>>>> "HN" == Hrvoje Niksic <hniksic@srce.hr> writes:

HN> The uuencode interface to mm seems very inefficient, because of
HN> two separate reasons.  When I `g' a message containing a, say, an
HN> uuencoded image, it insists on decoding the uuencoded region (I
HN> can tell by the time `g' takes and by the conducted profiles.)

HN> When I click on the image button to show it, it decodes it again.
HN> Why this double encoding?

The first decoding is due to mm-inlinable-p. With the attached patch,
each type is tested at most once in a session.

HN> The other question is about the strange default -- the code
HN> defaults to using the internal, elisp-based decoding, even when a
HN> perfectly valid `uudecode' is found on the system.  This makes
HN> decoding of large stuff (and uuencoded stuff is large, more often
HN> than not) abysmally slow, even on very very fast machines.  I can
HN> imagine that on slower machines it's closer to "unusable".

Do you mean to automatically setup mm-uu-decode-function by searching
uudecode on the system at loading time of mm-uu.el?  If so, please
feel free to patch it.

-- 
Shenghuo

:- cut -----
--- ChangeLog	1998/11/29 08:49:39	1.1
+++ ChangeLog	1998/11/29 08:51:02
@@ -1,3 +1,7 @@
+Sun Nov 29 03:50:20 1998  Shenghuo ZHU  <zsh@cs.rochester.edu>
+
+	* mm-decode.el (mm-inlinable-p): Save the result of test.
+
 Fri Nov 27 12:26:10 1998  Lars Magne Ingebrigtsen  <larsi@menja.ifi.uio.no>
 
 	* gnus.el: Pterodactyl Gnus v0.55 is released.

--- mm-decode.el	1998/11/29 08:44:18	1.1
+++ mm-decode.el	1998/11/29 08:47:55
@@ -357,9 +357,11 @@
 	test)
     (while alist
       (when (equal type (caar alist))
-	(setq test (caddar alist)
-	      alist nil)
-	(setq test (eval test)))
+	(setq test (caddar alist))
+	(when (consp test)
+	  (setq test (eval test))
+	  (setf (caddar alist) test))
+	(setq alist nil))
       (pop alist))
     test))