[benchmarks] word_freq example

Tickled the bash assoc array bug that Koichi worked around!  Not sure if
this is version-specific.

benchmarks/compute/word_freq.py

@@ -0,0 +1,36 @@
+#!/usr/bin/env python2
+"""
+word_freq.py
+"""
+from __future__ import print_function
+
+import sys
+
+
+def main(argv):
+  try:
+    iters = int(argv[1])
+  except IndexError:
+    iters = 1000
+
+  text = sys.stdin.read()
+
+  words = {}
+
+  for i in xrange(iters):
+    for word in text.split():
+      if word in words:
+        words[word] += 1
+      else:
+        words[word] = 1
+
+  for word in words:
+    print("%d %s" % (words[word], word))
+
+
+if __name__ == '__main__':
+  try:
+    main(sys.argv)
+  except RuntimeError as e:
+    print('FATAL: %s' % e, file=sys.stderr)
+    sys.exit(1)

benchmarks/compute/word_freq.sh

@@ -0,0 +1,35 @@
+#!/bin/bash
+
+main() {
+  iters=${1:-1000}
+
+  # read it once
+  read -d '' text
+
+  declare -A words
+
+  # do it a bunch of times
+  for (( i = 0; i < iters; ++i )); do
+    for word in $text; do  # relies on word splitting
+
+      # Hm this isn't correct in bash!
+      #(( words["$word"] += 1 ))
+
+      # This seems to work?  wtf?
+      # This causes a parse error in OSH though... Do we need two benchmarks?
+
+      # Or maybe we need something to turn of static parsing?
+      # similar to git
+
+      (( words[\$word] += 1 ))
+    done
+  done
+
+  # note: we can sort the output in the benchmark and assert that it's the same?
+
+  for word in "${!words[@]}"; do
+    echo "${words["$word"]} $word"
+  done
+}
+
+main "$@"