[soil] Improve wedge size report

Also make a demo of 'tail -f' and 'multitail' tool, which seems
promising for us to use.

build/deps.sh

@@ -443,6 +443,17 @@ container-wedges() {
 
 }
 
+commas() {
+  # Wow I didn't know this :a trick
+  #
+  # OK this is a label and a loop, which makes sense.  You can't do it with
+  # pure regex.
+  #
+  # https://shallowsky.com/blog/linux/cmdline/sed-improve-comma-insertion.html
+  # https://shallowsky.com/blog/linux/cmdline/sed-improve-comma-insertion.html
+  sed ':a;s/\b\([0-9]\+\)\([0-9]\{3\}\)\b/\1,\2/;ta'   
+}
+
 show-wedge-tree() {
   # 4 levels deep shows the package
   if command -v tree > /dev/null; then
@@ -452,6 +463,51 @@ show-wedge-tree() {
 
   # Sizes
   du --si -s /wedge/*/*/* ~/wedge/*/*/*
+  echo
+
+  local tmp=_tmp/wedge-tree.txt
+
+  # Show the biggest files
+  find /wedge ~/wedge -type f -a -printf '%10s %P\n' > $tmp
+  sort -n $tmp | tail -n 20 | commas
+  echo
+
+  # Show the most common file extensions
+  #
+  # I feel like we should be able to get rid of .a files?  That's 92 MB, second
+  # most common
+
+  cat $tmp | python3 -c '
+import os, sys, collections
+
+bytes = collections.Counter()
+files = collections.Counter()
+
+for line in sys.stdin:
+  size, path = line.split(None, 1)
+  path = path.strip()  # remove newline
+  _, ext = os.path.splitext(path)
+  size = int(size)
+
+  bytes[ext] += size
+  files[ext] += 1
+
+#print(bytes)
+#print(files)
+
+n = 20
+
+print("Most common file types")
+for ext, count in files.most_common()[:n]:
+  print("%10d %s" % (count, ext))
+
+print()
+
+print("Total bytes by file type")
+for ext, total_bytes in bytes.most_common()[:n]:
+  print("%10d %s" % (total_bytes, ext))
+' | commas
+
 }
 
 run-task "$@"

demo/tail-multiplex.sh

@@ -0,0 +1,44 @@
+#!/usr/bin/env bash
+
+set -o nounset
+set -o pipefail
+set -o errexit
+
+print-output() {
+  local begin=${1:-1}
+  local end=${2:-10}
+
+  for i in $(seq $begin $end); do
+    echo $i
+    sleep 0.1
+  done
+}
+
+parallel() {
+  print-output 1 10 &
+  print-output 11 20 &
+  wait
+  wait
+  echo done
+}
+
+parallel2() {
+  mkdir -p _tmp
+  print-output 1 10 >_tmp/d1 &
+  print-output 11 20 >_tmp/d2 &
+
+  # Hm the output is not good because it prints too much
+  # also --pid would be nice for stopping
+  #tail -q -f _tmp/d1 _tmp/d2
+
+  multitail _tmp/d1 _tmp/d2
+
+  wait
+  wait
+  echo done
+}
+
+# TODO: try this
+# https://www.vanheusden.com/multitail/
+
+"$@"