On Fri, Oct 01, 1999 at 09:38:17AM -0400 or thereabouts, Daniel Eisenbud wrote:
> On Fri, Oct 01, 1999 at 03:08:56PM +0300, Mikko Hänninen <[EMAIL PROTECTED]>
>wrote:
> > Telsa <[EMAIL PROTECTED]> wrote on Fri, 01 Oct 1999:
> > > I'm having some difficulties with the sorting by score ability of Mutt.
> >
> > Don't have any ideas on that, sorry... Unless Mutt does not support
> > every pattern match operator for scoring, only some. But that doesn't
> > sound likely or make any sense (what's different between looking at ~e
> > or ~h?).
>
> Mutt only supports some pattern operators for scoring. There's a
> reason, which I forget.
Someone else emailed me to say much the same thing, and said that
efficiency had had something to do with it, if they remembered right.
My immediate thought was "Well, which does it support?" and I spent
an entertaining evening resorting different folders according to
different criteria. I don't vouch for 100% accuracy for the whole
lot of these: but I can confidently assert that these worked for me:
~A (all), ~c (cc'd) ~C (to or cc'd), ~d (date) ~e (sender field), ~g
(PGP-signed) and ~N (new). That's as far through the alphabet as I
got before my brain melted.
A friend who understands C then told me that anything which has
M_FULL_MSG in the same line in the structure at the start of
pattern.c won't work. That happens to be the three I already
found definitely don't work: ~b, ~B and ~h. I also can't get
~l ("to a known mailing list") to work.
Perhaps it is worth adding a brief note to section 3.19 of the
manual to make it clear that those first three (or all four?)
aren't used for scoring?
I couldn't get ~k (contains PGP key material) to work on a message
containing my PGP key, although ~g worked to spot signed messages.
And ~I (message-ID, although why one would want to score by that
I'm not sure: I was just in a fit of trying things out at the time)
said, "I: invalid command".
> I had a discussion with Michael Elkins about
> this a long time ago. I think the conclusion was that scoring would
> have to be moved to a different pass than reading the messages for this
> to work. He didn't seem against the idea, IIRC, but I don't think
> anybody cared enough to actually do it.
I gather checking the entire message body would be a very big job
and result in slowing things down. I like mutt's speed, so I'm not
going to push this :)
Thanks for the enlightenment. At least I know it's not (all) due
to pattern-matching mistakes on my part now! I hope the partial
list of what does and doesn't work is (a) accurate, and (b) of
use to someone. I'm not desperate to sort on message bodies: or
at least not desperate enough to start reading the procmail pages
again, but I suppose that's one answer for anyone who is.
Telsa
