Giampaolo Tomassoni wrote:
> > From: LuKreme [mailto:krem...@kreme.com]
> >
> > On 25-Mar-2009, at 11:24, Giampaolo Tomassoni wrote:
> > > rawbody LARGETABLE
> > > m'<tr\W(?:[^<]|<(?!t[dr]\W))*(?:<td\W(?:[^<]|<(?!t[rd]\W))*){30,}</
> > > tr'is
> >
> >
> > Just to be sure my parsing is working correctly, that is flagging if
> > there are 30 or more TDs in a single TR?
>
> Right.
>
>
> > If so, couldn't that be
> > written a lot more compactly?
>
> Probably yes. The problem is that a simple way like
> '<tr\W.*(?:<td\W.*){30,}</tr' would easily fail because the '*'
> operator would work "greedly" here, consuming <td>s and <tr>s which
> should instead be counted.
Then why not use the non-greedy version?
<tr\W.*?(?:<td\W.*?){30,}</tr
On the other hand, '.*' of any kind is usually a bad idea in a
SpamAssassin rule. It should always be limited to avoid excessive
backtracking.
<tr\W.{,20}?(?:<td\W.{,20}?){30,}</tr
I pulled the 20 character limit out of thin air. Change it to whatever
makes sense for this rule.
--
Bowie
