news/batch/parse.pl

#!/usr/bin/perl
# Copyright (c) 2006 Eclipse Foundation, made available under EPL v1.0
# Contributors Ward Cunningham, Bjorn Freeman-Benson
#
# usage:
# cat projects.txt | ./extract.pl > TEXT_FILE_OF_ARTICLES
#
# alternate:
# extract.pl --ignoretime

use strict;


# Project: eclipse
# Newsgroup: eclipse.platform
# From: Tom Roche <tlroche@us.ibm.com>
# Date: Fri, 18 Apr 2003 00:07:12 -0400
# Message-ID: <3E9F79F0.2040101@us.ibm.com>
# In-Reply-To: <b7mqdn$kue$1@rogue.oti.com>

my ($project, $newsgroup, $email, $date, $messageid, $replyto);

print "#	MESSAGEID	EMAIL	DATE	NEWSGROUP	REPLYTOID	PROJECT\n";

for (<STDIN>) {
	chomp;
	if( /^--------------/ ) {
		write_data();
		$project = undef;
		$newsgroup = undef;
		$email = undef;
		$date = undef;
		$messageid = undef;
		$replyto = undef;
	} else {
		$project = $1 if( /^Project: ([A-Za-z._-]+)/ );
		$newsgroup= $1 if( /^Newsgroup: ([A-Za-z._-]+)/ );
		$email = $1 if( /^From: [^<]*<([^>]+)>/ );
		$email = $1 if( /^From: (\S+) \(/ );
		$date = $1 if( /^Date: (.*)/ );
		$messageid = $1 if( /^Message-ID: <(.*)>/ );
		$replyto = $1 if( /^In-Reply-To: <(.*)>/ );
	}
}

sub write_data {
	print "$messageid	$email	$date	$newsgroup	$replyto	$project\n";
}