February 2001

The Shell Corner: cmptree

Ed Schaefer

Last month, Gerard van Wageningen submitted convert, a Korn script to display numbers in binary, decimal, hexadecimal, and base 36. After my deadline, Gerard submitted a change, adding base 64. If anyone desires this change, please email me and I'll send it to you.

This month, Ives Aerts submits cmptree, a bash script that compares the contents of two directories and reports if like-named files are different. It also reports a missing file or directory as not existing. Simply execute:

cmptree directory1 directory2</Pre>


Comparing directories is not new. SCO Unix has provided shell script <code>dircmp</code> since before System V, and Solaris also provides a version of <code>dircmp</code>. For Unix variants that dont have a directory compare, such as RedHat Linux 6.1, <code>cmptree</code> is a worthy addition.</P>


While <code>dircmp</code> executes a <code>find</code> on each directory comparing contents, <code>cmptree</code> elegantly compares the two directories recursively. The simplicity of <code>cmptree</code> eliminates some of the functionality of <code>dircmp</code>.</P>
<B>
<h3><code>cmptree</code> Limitations</h3>
</B>


<code>cmptree</code> reports only if files are different -- not what the differences are.</P>


<code>cmptree</code> only compares normal files. Special files, such as block, character, named pipe, semaphore, and shared memory are marked as "unknown."</P>


Also, <code>cmptree</code> uses <code>test -e</code> to check whether an object exists; thus, no read permissions causes an "object does not exist" error.</P>


Soft links report only if they are different. For example, two directories exist: <code>/tmp/dir1</code> and <code>/tmp/dir2</code>. In the present working directory, create two soft links:</P>
<Pre>ln -s /tmp/dir1 linkdir1
ln -s /tmp/dir2 linkdir2

cmptree /tmp/dir1 /tmp/dir2

cmptree linkdir1 linkdir2

  cmp -s $1 $2
  if [ $? -gt 0 ]; then
    echo "$1 different from $2"
  fi

#!/bin/bash
#
# cmptree: compare directory trees recursively and report the differences.
# Author: Ives Aerts

function gettype () {
  if [ -L $1 ]; then
    echo "softlink"
  elif [ -f $1 ]; then
    echo "file"
  elif [ -d $1 ]; then
    echo "directory"
  else
    echo "unknown"
  fi
}

function exists () {
  if [ -e $1 -o -L $1 ]; then
    return 0;
  else
    echo "$1 does not exist."
    return 1;
  fi
}

function comparefile () {
  cmp -s $1 $2
  if [ $? -gt 0 ]; then
    echo "$1 different from $2"
#  else
#    echo "$1 same as $2"
  fi
  return
}

function comparedirectory () {
  local result=0
  for i in `(ls -A $1 && ls -A $2) | sort | uniq`; do
    compare $1/$i $2/$i || result=1
  done
  return $result
}

function comparesoftlink () {
  local dest1=`ls -l $1 | awk '{ print $11 }'`
  local dest2=`ls -l $2 | awk '{ print $11 }'`

  if [ $dest1 = $dest2 ]; then
    return 0
  else
    echo "different link targets $1 -> $dest1, $2 -> $dest2"
    return 1
  fi
}

# compare a file, directory, or softlink
function compare () {
  (exists $1 && exists $2) || return 1;

  local type1=$(gettype $1)
  local type2=$(gettype $2)

  if [ $type1 = $type2 ]; then
    case $type1 in
      file)
        comparefile $1 $2
        ;;
      directory)
        comparedirectory $1 $2
        ;;
      softlink)
        comparesoftlink $1 $2
        ;;
      *)
        echo "$1 of unknown type"
        false
        ;;
    esac
  else
    echo "type mismatch: $type1 ($1) and $type2 ($2)."
    false
  fi

  return
}

if [ 2 -ne $# ]; then
cat << EOU
Usage: $0 dir1 dir2
Compare directory trees:
  files are binary compared (cmp)
  directories are checked for identical content
  soft links are checked for identical targets
EOU
  exit 10
fi

compare $1 $2
exit $?

Check out past winners of shell corner!!!

By day, Ed Schaefer is a mild-mannered senior programmer-analyst for Intel's Factory Integrated Information Systems (FIIS). The standard employer-employee disclaimer is in effect: In this forum, Ed doesn't speak for Intel and his views on Unix and all other topics are his own.