To: -email is unavailable-
cc: Paul Fox <pgf@foxharp.boston.ma.us>,
    nmh workers <nmh-workers@nongnu.org>
Subject: Re: [Nmh-workers] Additional subject prefixes to ignore
In-reply-to: Message from -email is unavailable- on
             Wed, 24 May 2006 21:47:36 -0400.<200605250147.k4P1lbB4006725@turing-police.cc.vt.edu>
References: <20694.1148518425@foxharp.boston.ma.us> <13316.1148519658@mittens.honig.net> <200605250147.k4P1lbB4006725@turing-police.cc.vt.edu>
From: jch
X-Mailer: MH-E 8.0.3+cvs; nmh 1.2; GNU Emacs 22.1.1
Fcc: mail/mh
--------
OK, only 29 months to reply to this one.

This is an interesting idea.  Since a resent post reminded me that
implementing something along these lines was warranted, I started
looking at it.  First I did a quick version fetching a ``Subject-Re-Regex''
from the profile, i.e.:

Subject-Re-Regex: ^([Rr][Ee]|[Ff][Ww][Dd]?):

And implemented it in repl and sortm (see diff attached).

Now I'm looking at your suggestion and it sounds interesting.  The issue
is that mh-format functions can take only one argument.  The first
solution I can think of is a function named ``resubject'' which would
apply the specified regular expression to the subject field.  How do we
handle RFC2047 subjects?  Have two functions resubject and deresubject?

Also, as you alluded to in your example, it would be nice not to assume
that the Re: field is a prefix and allow a regular expression
substitution, not just a match.  That would require a method to specify
both the match string and the replacement string, something like perl
uses (the first character is the separator). 

This would also leave sortm with a different mechanism.


Another alternative would be to specify the regular expression in the
profile (Regex-Pattern: %match%sub%) and then provide the string as the
argument %(regsub{subject}).  This would allow sortm to silently use the
same mechanism.


Anyone have any ideas to contribute?

Thanks (谢谢).

Jeff



(file #16725)

To: Jon Steinhart <jon@fourwinds.com>
cc: nmh workers <nmh-workers@nongnu.org>
Subject: Re: [Nmh-workers] Additional subject prefixes to ignore
From:  "Jeffrey C Honig" <jch@honig.net>
Date: Wed, 24 May 2006 22:59:56 -0400
Sender: -email is unavailable-

Jon Steinhart <jon@fourwinds.com> wrote:

> > I'm hesitant to try to use a regex because the only regex code in nmh is
> > strictly for pick and generalizing it or making it use POSIX regular
> > expressions is more than I want to take on at the moment.
> >
> > Yeah, I'm chicken, I admit it.
>
> The posix regular expression library calls are simple to use.
> It's not like you have to write your own from scratch anymore.

Yeah, that's easy.  The hard parts are:

1) If we start requiring regex, we need to fix configure.  Will this
   limit systems nmh can run on?  Do we support systems that do not have
   POSIX regex?

2) If we add some code that uses POSIX regex, we should update pick to
   be consistent.  Does the code in picksbr use the same syntax as the
   POSIX code?

3) This is going from a two hour enhancement to a multi-day enhancement
   and cleanup.

Thanks.

Jeff

To: Jeffrey C Honig <jch@honig.net>
Cc: Paul Fox <pgf@foxharp.boston.ma.us>, nmh workers <nmh-workers@nongnu.org>
Subject: Re: [Nmh-workers] Additional subject prefixes to ignore
From: -email is unavailable-
Date: Wed, 24 May 2006 21:47:36 -0400

On Wed, 24 May 2006 21:14:18 EDT, Jeffrey C Honig said:

> That's a good idea.  But I think that can be accomplished w/o a regex by
> listing `[listname]' as a prefix to list for each list you are on.
>
> I'm hesitant to try to use a regex because the only regex code in nmh is
> strictly for pick and generalizing it or making it use POSIX regular
> expressions is more than I want to take on at the moment.

On the other hand, generalizing it would be pretty nice, too.  Not only
would it be useful for 'sort', but for 'scan' as well, and even things
like 'repl' - using the regexp to trim off so we can re-sane the messy
'Re: FW: [list] re:' noise we see now.  Fortunately for scan and repl,
we can do all the heavy lifting in one place, in a filter %regexp of some
sort.

%regexp{pcre,string} - so we'd have:

Subject: Re: %regexp{"([Rr][Ee]:|[Ff][Ww][Dd]*:|\[[a-zA-Z-]\])*\(\1\)",%{subject}}

Or something like that....

To: "nmh workers" <nmh-workers@nongnu.org>
Subject: Re: [Nmh-workers] Additional subject prefixes to ignore
Date: Wed, 24 May 2006 20:53:45 -0400
From: Paul Fox <pgf@foxharp.boston.ma.us>
Sender: nmh-workers-bounces+jch=honig.net@nongnu.org

jeffrey c honig wrote:
 > I have to deal with mail from Outbreak, I mean Outlook, on a daily
 > basis.  At least I get paid for the privilege...
 >
 > Much of this mail comes with a subject prefix of FW: instead of Re:.
 > This messes up sorting by subject.
 >
 > I would like to train nmh to string FW: in addition to Re:.  But that
 > seems like too specific a solution to the problem.  A more general
 > solution would be to add flags to the relevant commands to add prefixes
 > to the list.

i think i understand what you're getting at, but i'd like to see it
be a regular expression that gets trimmed before the sort (i assume
that's what you're referring to).  in particular, i get mail from lists
which prepend the listname.  that coupled with mailers that do and
don't strip existing "Re:" strings gives subjects that begin with
any of:
    [listname] foo
    Re: [listname] foo
    [listname] Re: foo
    Re: [listname] Re: foo

(i realize that an expression that catches these risks corrupting
a valid subject.  so far, i'm willing to risk that.)

paul
=---------------------
 paul fox, -email is unavailable- (arlington, ma, where it's 50.7 degrees)