Visualize CI runs by test type and job category
This visualization is a stacked bar chart with each bar representing a
job category and each colored segment represesenting a test type.
Change-Id: I4b8c470cef40266bf3dabd1f35a135592619a12c
diff --git a/script/graphs/categorize-tests.bash b/script/graphs/categorize-tests.bash
new file mode 100644
index 0000000..92d286c
--- /dev/null
+++ b/script/graphs/categorize-tests.bash
@@ -0,0 +1,64 @@
+#
+# Copyright (c) 2021 Arm Limited. All rights reserved.
+#
+# SPDX-License-Identifier: BSD-3-Clause
+#
+#!/usr/bin/env bash
+set -euo pipefail
+
+# This script plots the categories of tests by group. It does this by combining
+# the awk script and gnuplot script of the same name. This script accepts an
+# argument, in the format of a grep expression, that allows the tests to be
+# filtered before categorization. This script produces the plot as a png on
+# stdout.
+
+# Variables
+# ^^^^^^^^^
+#
+# We are located in a specific location with this repo, so we can take
+# advantage of that to avoid any issues with running this from an unexpected
+# directory.
+rootdir=$(realpath $(dirname $(realpath $0))/../..)
+
+# I would like to use process-substitution for this, so that we can avoid
+# making a file on disk and keep everything in memory, removing the need to
+# clean anything up on exit and preventing any chance of polluting the user's
+# filesystem. However, when gnuplot is asked to plot from the same file more
+# than once, it will seek to the start of the file for every subsequent plot
+# after the first. Unix Pipes do not support this operation, and plotting fails
+# under these circumstances. Instead, we use an intermediate file, which is
+# removed on success.
+categories=$(mktemp "XXXXXXX-test-categories.dat")
+
+# We change a portion of the title for our graph based on the argument passed
+# to this script.
+title=$(if [[ $# -ge 1 ]] ; then echo $1 ; else echo "All Tests" ; fi)
+
+# Generate Data into the $categories file
+# ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+#
+# The following pipeline is the heart of the implementation, and has four
+# stages: find, ???, awk, and sort. The ??? stage of the pipeline is determined
+# by the bash if statement, which switches between a filter, when an argument
+# is passed, and a passthrough, implemented as `cat -`, when no filter
+# argument is passed.
+#
+# Note that the env -C before the find is to enforce that it produces
+# directories relative to the $rootdir, so that it does not trip up the awk
+# script.
+env -C $rootdir find group -type f |\
+ if [[ $# -ge 1 ]] ; then
+ grep -e "$1" -
+ else
+ cat -
+ fi | awk -f ${0%bash}awk | sort > $categories
+
+# Generate a Plot (on stdout)
+gnuplot -e "subtitle='$title'" -c ${0%bash}plot $categories
+
+# Dump data to stderr
+echo name build static component inegration 1>&2
+cat $categories 1>&2
+
+# Clean up temporary files
+rm $categories