From 6c0c347db28783a3412c72bf9b7ef18b9926e8e1 Mon Sep 17 00:00:00 2001
From: Tom Ryder <tom@sanctum.geek.nz>
Date: Fri, 9 Jun 2017 23:32:36 +1200
Subject: Make dub(1df) smarter, add a caveat

---
 bin/dub.sh       | 26 +++++++++++---------------
 man/man1/dub.1df |  4 ++++
 2 files changed, 15 insertions(+), 15 deletions(-)

diff --git a/bin/dub.sh b/bin/dub.sh
index f42c5ac9..41ef88d3 100644
--- a/bin/dub.sh
+++ b/bin/dub.sh
@@ -1,4 +1,5 @@
 # List the biggest files in a directory
+self=dub
 
 # First optional argument is the directory, defaulting to the
 # current dir; second optional argument is the number of files to
@@ -8,21 +9,16 @@ dir=${1:-.} lines=${2:-10}
 # Enter the target dir or bail
 cd -- "$dir" || exit
 
-# Add files matching glob, shift them off if unexpanded (first and
-# only entry doesn't exist)
-set -- *
-[ -e "$1" ] || shift
-
-# Add dot files, shift off the "." and ".." entries (sh(1)
-# implementations seem to vary on whether they include these)
-set -- .* "$@"
-[ -e "$1" ] || shift
-[ "$1" = . ] && shift
-[ "$1" = .. ] && shift
-
-# Run du(1) with POSIX compatible flags -k for kilobyte unit and
-# -s for total over the arguments
-du -ks -- "$@" |
+# Some find(1) devilry to deal with newlines as safely as possible. The idea is
+# not even to touch them, and warn about their presence; better the results are
+# wrong than malformed
+nl=$(printf '\n/')
+find . ! -name . -prune \( \
+    -name '*'"${nl%/}"'*'  \
+        -exec sh -c '
+            printf >&2 '\''%s: warning: skipped newline filename\n'\'' "$1"
+        ' _ "$self" \; \
+    -o -exec du -ksx -- {} + \) |
 
 # Sort the first field (the sizes) numerically, in reverse
 sort -k1,1nr |
diff --git a/man/man1/dub.1df b/man/man1/dub.1df
index 52900a36..69a4c8e0 100644
--- a/man/man1/dub.1df
+++ b/man/man1/dub.1df
@@ -15,6 +15,10 @@
 lists the biggest entries in a given directory, defaulting to the current
 directory. It defaults to printing 10 entries unless a second argument is
 given.
+.SH CAVEATS
+Skips filenames with newlines in them with an explicit warning to stderr, for
+the least dangerous POSIX-compatible approach. Even so, you probably shouldn't
+use this in critical scripts.
 .SH SEE ALSO
 du(1)
 .SH AUTHOR
-- 
cgit v1.2.3