nnmail-split-fancy usability problem

nnmail-split-fancy usability problem

[Florian Weimer]

Let's say we have list postings with the following header:

List-Id: An example list <list.example.com>

Say we want to use a generic rule which processes all lists at
example.com.  The following doesn't work because "<" is not a word
constituent:

("List-Id" "<\\([a-z0-9-]+\\)\\.example\\.com>" "lists.example.\\1")

This doesn't work either because ".*" is greedy.

("List-Id" ".*\\([a-z0-9-]+\\)\\.example\\.com" "lists.example.\\1")

After rewriting, this results in an illegal regular expression (".*"
at the beginning is dropped unconditionally).

("List-Id" ".*?\\([a-z0-9-]+\\)\\.example\\.com" "lists.example.\\1")

Fortunately, the rule below works because the initial ".*" suppresses
the word constituent check:

("List-Id" ".*<\\([a-z0-9-]+\\)\\.example\\.com>.*" "lists.example.\\1")

However, we are a bit lucky.  If there weren't the "<" character as a
marker, I think we would have a hard time working around the greedy
".*" in the header field matcher.

The following patch changes the implicit ".*" to ".*?".  Okay to
apply?

(However, IMHO, the whole behavior of nnmail-split-fancy is highly
non-intuitive.)

2003-06-29  Florian Weimer  <fw@deneb.enyo.de>

	* nnmail.el (nnmail-split-it): Make header field match
          non-greedy.

--- nnmail.el.~6.77.~	2003-06-29 12:50:50.000000000 +0200
+++ nnmail.el	2003-06-29 13:19:13.000000000 +0200
@@ -1306,7 +1306,7 @@
 	    (push split nnmail-split-trace))
 	  (let ((split-rest (cddr split))
 		(end (match-end 0))
-		;; The searched regexp is \(\(FIELD\).*\)\(VALUE\).
+		;; The searched regexp is \(\(FIELD\).*?\)\(VALUE\).
 		;; So, start-of-value is the point just before the
 		;; beginning of the value, whereas after-header-name
 		;; is the point just after the field name.
@@ -1366,7 +1366,7 @@
 			     (if (symbolp field)
 				 (cdr (assq field nnmail-split-abbrev-alist))
 			       field)
-			     "\\):.*\\)"
+			     "\\):.*?\\)"
 			     (or partial-front "\\<")
 			     "\\("
 			     value

Re: nnmail-split-fancy usability problem

[Michael Cook]

Florian Weimer <fw@deneb.enyo.de> writes:

> The following patch changes the implicit ".*" to ".*?".  Okay to
> apply?

i think that's an xemacs feature that emacs doesn't have yet.

m.

Re: nnmail-split-fancy usability problem

[Florian Weimer]

Michael Cook <michael@waxrat.com> writes:

>> The following patch changes the implicit ".*" to ".*?".  Okay to
>> apply?
>
> i think that's an xemacs feature that emacs doesn't have yet.

It's also available in GNU Emacs 21.

| ** Regular expressions now support intervals \{n,m\} as well as
| Perl's shy-groups \(?:...\) and non-greedy *? +? and ?? operators.

Re: nnmail-split-fancy usability problem

[Johan Bockgård]

Florian Weimer <fw@deneb.enyo.de> writes:

> The following patch changes the implicit ".*" to ".*?".  Okay to
> apply?

--8<---------------cut here---------------start------------->8---

From:  Lars Magne Ingebrigtsen <larsi@gnus.org>
Newsgroups: gnu.emacs.gnus
Subject: Re: [patch] Greedy pattern in nnmail-split-it
Date: Sat, 07 Jun 2003 19:51:41 +0200
Message-ID: <m3brx9n5ci.fsf@quimbies.gnus.org>

Dmitry Astapov <adept@umc.com.ua> writes:

> Here greedy match is used to glue together regexp for FIELD and
> regexp for VALUE, and this greedy match eats too much. Basically, it
> prevents user from using non-greedy matches at the beginning of
> VALUE in the nnmail-split-fancy at all.
>
> I propose to change ":.*" to ":.*?"

Well, this can't be changed in the stable Gnus series, so could you
suggest this again once the next development series starts?

--8<---------------cut here---------------end--------------->8---