#!/usr/bin/bash
# Usage: reExtract.sh 20..-.. < files...
# Input is list of paths to relative path of warc files
#  under /beegfs/common_crawl/CC-MAIN-$1

function edex () {
    echo "$1" | \
	cut -f 5 -d - | cut -f 1 -d .|sed 's/^00//' | \
	echo $(($(fgrep -n -f - ~/by11n.txt | cut -f 1 -d :) - 1))
    }

h=/beegfs/common_crawl/CC-MAIN-$1

mkdir -p /dev/shm/rex
cd /dev/shm/rex

sed 's/\// /;s/-00/ /;s/.warc.gz//' | \
 while read s p i
 do
    printf "%s\t%s\t%s\t%s\n" $(edex $i) $s $p $i
 done > in.txt
for s in $(cut -f 2 in.txt| sort -u)
do
    mkdir -p $s
    cd $s
    for e in $(egrep "\b$p\b" in.txt |cut -f 1 | sort -u)
    do
	tar -xf $h/$s/extract_$e.tar