Pull a random sample of lines from the files in a directory (http://www.progski.net/blog/2011/random_lines_in_bash.html)

sample-lines.sh

#! /bin/bash
#
# Pull a random sample of size 1000 from a directory of files and
# write to standard out.
#
# Usage:
#       ./sample-lines /path/to/dir > mysample.log

if [ $# -lt 1 ]
then
    echo "Usage: $0 PATH" >&2
fi

TMP=$(mktemp -t sample_lines.XXXX)
trap "rm -f $TMP" EXIT

DIR=$1
LC=$(wc -l $DIR/* | grep total | grep -oE [[:digit:]]+)

echo pulling 1000 samples from $LC lines in $DIR >&2

rand()
{
    od -N4 -An -tu4 /dev/urandom \
        | sed -n 1p \
        | awk -v lim="$LC" '{printf "%i\n", 0 + (lim * ($1 / 4294967296))}' 
}

for i in {1..1000}
do
    echo $(rand)p >> $TMP
done

sed -n -f "$TMP" $DIR/*