add shell script to convert result of zonewalk to input for gnunet-gns-benchmark

diff --git a/contrib/zonewalk-to-types.sh b/contrib/zonewalk-to-types.sh
new file mode 100755 (executable)
index 0000000..c5096e6
--- /dev/null
+++ b/contrib/zonewalk-to-types.sh
@@ -0,0 +1,35 @@
+#!/bin/sh
+# This script is in the public domain.
+# Converts the output of gnunet-zonewalk (DNS resolutions)
+# into a proper input for gnunet-gns-benchmark.
+
+NUM_CLIENTS=3
+# How many different groups of names should we
+# create?  1/N will be in the 'shared' group.
+
+# FILE ($1) contains results from DNS lookup; strip
+# everything but the hostnames, remove duplicates
+# and then randomize the order.
+cat $1 | awk '{print $1}' | sort | uniq | shuf > $1.tmp
+TOTAL=`cat $1.tmp | wc -l`
+GROUP_SIZE=`expr $TOTAL / \( $NUM_TYPES + 1 \)`
+
+# First group (0) is to be shared among all clients
+for i in `seq 1 $NUM_CLIENTS`
+do
+  cat $1.tmp | head -n $GROUP_SIZE | awk "{print 0 \" \" \$1}" > $1.$i.tmp
+done
+
+# Second group (1) is unique per client
+OFF=0
+for i in `seq 0 $NUM_CLIENTS`
+do
+  END=`expr $OFF + $GROUP_SIZE`
+  cat $1.tmp | head -n $END | tail -n $GROUP_SIZE | awk "{print 1 \" \" \$1}" >> $1.$i.tmp
+# Shuffle again, so we mix the different request categories in terms of
+# when we issue the queries.
+  cat $1.$i.tmp | shuf > $1.$i
+  OFF="$END"
+  rm $1.$i.tmp
+done
+rm $1.tmp