fancy split regexp?

fancy split regexp?

[Josh Huber]

I must be being pretty dense here -- perhaps someone could help me
out.  Here's the rule:

	("Resent-To"
	 "\\(.+\\)@packages\\.qa\\.debian\\.org"
	 "mail.debian.pts.\\1")

This does not work.

Variations on this, such as .*, [a-z-_]+ (or *), \\w+, etc do not
work!  The versions with * always seem to match nothing, and the
message would be split into mail.debian.pts., the versions with +
always seem to match e: mail.debian.pts.e

The message I'm using for testing has this header:

Resent-To: developers-reference@packages.qa.debian.org

This works fine:

(let ((str "developers-reference@packages.qa.debian.org"))
  (string-match "\\<\\(.*\\)@packages\\.qa\\.debian\\.org\\>" str)
  (match-string 1 str))

=> "developers-reference"

What could possibly be going on here?

-- 
Josh Huber

Re: fancy split regexp?

[Bryan]

My guess is that this has to do with the fact that nnmail-split-it
uses re-search-backwards to do it's pattern matching. 

BrYan

Josh Huber <huber@alum.wpi.edu> writes:
> I must be being pretty dense here -- perhaps someone could help me
> out.  Here's the rule:
>
> 	("Resent-To"
> 	 "\\(.+\\)@packages\\.qa\\.debian\\.org"
> 	 "mail.debian.pts.\\1")
>
> This does not work.
>
> Variations on this, such as .*, [a-z-_]+ (or *), \\w+, etc do not
> work!  The versions with * always seem to match nothing, and the
> message would be split into mail.debian.pts., the versions with +
> always seem to match e: mail.debian.pts.e
>
> The message I'm using for testing has this header:
>
> Resent-To: developers-reference@packages.qa.debian.org
>
> This works fine:
>
> (let ((str "developers-reference@packages.qa.debian.org"))
>   (string-match "\\<\\(.*\\)@packages\\.qa\\.debian\\.org\\>" str)
>   (match-string 1 str))
>
> => "developers-reference"
>
> What could possibly be going on here?

Re: fancy split regexp?

[Josh Huber]

Bryan <bilko@onebabyzebra.com> writes:

> My guess is that this has to do with the fact that nnmail-split-it
> uses re-search-backwards to do it's pattern matching. 

Hmm, okay.

That really sucks.  I'm trying to fiddle with re-search-backward, but
I can't seem to get it to go past the word boundry (-) in
developers-reference.  I always get reference out.

I'll look into it further ;)

thanks,
-- 
Josh Huber

Re: fancy split regexp?

[Paul Jarc]

Josh Huber <huber@alum.wpi.edu> wrote:
> I'm trying to fiddle with re-search-backward, but I can't seem to
> get it to go past the word boundry (-) in developers-reference.  I
> always get reference out.

Try this:
("Resent-To"
"\\`\\(.*[^!#-'*+/-9=?A-Z^-~-]\\)?\\([!#-'*+/-9=?A-Z^-~-]*\\)@packages\\.qa\\.debian\\.org"
 "mail.debian.pts.\\2")


paul

Re: fancy split regexp?

[Josh Huber]

prj@po.cwru.edu (Paul Jarc) writes:

> Try this:
> ("Resent-To"
> "\\`\\(.*[^!#-'*+/-9=?A-Z^-~-]\\)?\\([!#-'*+/-9=?A-Z^-~-]*\\)@packages\\.qa\\.debian\\.org"
>  "mail.debian.pts.\\2")

Oh dear god...

Can we somehow use re-search-forward for splitting?  I tried this (the
above regexp), and it works in a scratch buffer, but not in the
fancy-split-rules.  In fact, any time I try and use \\` it seems to
never match.

So, I gave up... (oh, and I found another header to match on ;):

(defun jmh::pts-group (base)
  (goto-char (point-min))
  (when (re-search-forward "^X-PTS-Package: \\(.*\\)$" nil t)
    (concat base "." (match-string 1))))

and use it like so:

	(: jmh::pts-group "mail.debian.pts")

probably slower, but I fell a lot better about it :)

Thanks for ideas,

-- 
Josh Huber

Re: fancy split regexp?

[Kai Großjohann]

Josh Huber <huber@alum.wpi.edu> writes:

> Can we somehow use re-search-forward for splitting?  I tried this (the
> above regexp), and it works in a scratch buffer, but not in the
> fancy-split-rules.  In fact, any time I try and use \\` it seems to
> never match.

Sure, \\` only matches at beginning of buffer.  But most mail
addresses are not at the beginning of the buffer.  For example, the
header name comes before the address.

kai
-- 
Silence is foo!

Re: fancy split regexp?

[Josh Huber]

Kai.Grossjohann@CS.Uni-Dortmund.DE (Kai Großjohann) writes:

> Sure, \\` only matches at beginning of buffer.  But most mail
> addresses are not at the beginning of the buffer.  For example, the
> header name comes before the address.

Ah, of course.

I guess my point still stands -- since we use re-search-backward it
makes regexps for matching addresses pretty disgusting...IMHO.  Or,
near impossible.

Taking a peek at nnmail-split-it, I'm not ready to start screwing with
that code for fear in breaking people's split rules. =)

perhaps another time.

-- 
Josh Huber