Merge pull request #615 from dbic/enh-anon-cmd

Add a bash anon-cmd to be used to incrementally anonymize sids

utils/anon-cmd

@@ -0,0 +1,54 @@
+#!/bin/bash
+# Generic anonymization script which would anonymize sid based on what it had
+# seen in the past or simply what the translation dict already has.
+
+set -eu
+
+debug() {
+    : echo "DEBUG: $*" >&2
+}
+
+# Translation file location
+# Store under .git by default to guarantee that it is not committed or locked by git-annex etc
+# But it might not fit some usecases where there is no .git
+anon_file_default=$(dirname "$0")/../.git/anon_sid_map.csv
+anon_file="${AC_ANON_FILE:-$anon_file_default}"
+anon_fmt="${AC_ANON_FMT:-%03d}"
+
+sid="$1"
+
+# harmonize since elderly  awk on rolando seems to have no clue about IGNORECASE
+sid=$(echo "$sid" | tr  '[:lower:]' '[:upper:]')
+
+debug "Using $anon_file to map $sid"
+
+if [ ! -e "$anon_file" ]; then
+    touch "$anon_file"  # initiate it
+fi
+
+# apparently heudiconv passes even those we provided in `-s` CLI option
+# to anonymization script.  So, we will have to match those by our format
+# and then give back if matches. That would forbid plain remapping though if
+# original ids are in the same format, so some folks might want to disable that!
+sid_input_fmted=$(echo "$sid" | sed -e 's,^0*,,g' | xargs printf "$anon_fmt" 2>&1 || :)
+if [ "$sid" = "$sid_input_fmted" ]; then
+	debug  already in the anonymized format
+	echo "$sid"
+	exit 0
+fi
+
+res=$(grep "^$sid," "$anon_file" | head -n 1)
+if [ -n "$res" ]; then
+    ann="${res##*,}"
+    debug "Found $ann in '$res'"
+else
+    echo "We have all sids mapped already! Will not create a new one for $sid" >&2; exit 1
+    # need to take the latest one
+    largest=$(sed -e 's/.*,//g' "$anon_file" | sort -n | tail -n1 | sed -e 's,^0*,,g')
+    next=$((largest+1))
+    # shellcheck disable=SC2059
+    ann=$(printf "$anon_fmt" $next)
+    debug "Found $largest and $next to get $ann, storing"
+    echo "$sid,$ann" >> "$anon_file"
+fi
+echo "$ann"