File Coverage

blib/lib/File/FindSimilars.pm

Criterion	Covered	Total	%
statement	13	15	86.6
branch			n/a
condition			n/a
subroutine	5	5	100.0
pod			n/a
total	18	20	90.0

package File::FindSimilars;

# @Author: Tong SUN, (c)2001-2008, all right reserved
# @Version: $Date: 2009/01/02 21:10:20 $ $Revision: 2.6 $
# @HomeURL: http://xpt.sourceforge.net/

# {{{ LICENSE:

#
# Permission to use, copy, modify, and distribute this software and its
# documentation for any purpose and without fee is hereby granted, provided
# that the above copyright notices appear in all copies and that both those
# copyright notices and this permission notice appear in supporting
# documentation, and that the names of author not be used in advertising or
# publicity pertaining to distribution of the software without specific,
# written prior permission.  Tong Sun makes no representations about the
# suitability of this software for any purpose.  It is provided "as is"
# without express or implied warranty.
#
# TONG SUN DISCLAIM ALL WARRANTIES WITH REGARD TO THIS SOFTWARE, INCLUDING ALL
# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS, IN NO EVENT SHALL ADOBE
# SYSTEMS INCORPORATED AND DIGITAL EQUIPMENT CORPORATION BE LIABLE FOR ANY
# SPECIAL, INDIRECT OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES WHATSOEVER
# RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION OF
# CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN
# CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
#

# }}}

# {{{ POD, Intro:

=head1 NAME

File::FindSimilars - Fast similar-files finder

=head1 SYNOPSIS

  use File::FindSimilars;

  my $similars_finder =
      File::FindSimilars->new( { fc_level => $fc_level, } );
  $similars_finder->find_for(\@ARGV);
  $similars_finder->similarity_check();

=head1 DESCRIPTION

Extremely fast file similarity checker. Similar-sized and similar-named
files are picked out as suspicious candidates of duplicated files.

It uses advanced soundex vector algorithm to determine the similarity
between files. Generally it means that if there are n files, each having
approximately m words in the file name, the degree of calculation is merely

  O(n^2 * m)

which is over thousands times faster than any existing file fingerprinting
technology.

=head2 ALGORITHM EXPLANATION

The self-test output will help you understand what the module do and what
would you expect from the outcome.

  $ make test
  PERL_DL_NONLAZY=1 /usr/bin/perl "-Iblib/lib" "-Iblib/arch" test.pl
  1..5 todo 2;
  # Running under perl version 5.010000 for linux
  # Current time local: Wed Nov  5 17:45:19 2008
  # Current time GMT:   Wed Nov  5 22:45:19 2008
  # Using Test.pm version 1.25
  # Testing File::FindSimilars version 2.04

  . . . .

  == Testing 2, files under test/ subdir:

    9 test/(eBook) GNU - Python Standard Library 2001.pdf
    3 test/Audio Book - The Grey Coloured Bunnie.mp3
    5 test/ColoredGrayBunny.ogg
    5 test/GNU - 2001 - Python Standard Library.pdf
    4 test/GNU - Python Standard Library (2001).rar
    9 test/LayoutTest.java
    3 test/PopupTest.java
    2 test/Python Standard Library.zip
  ok 2 # (test.pl at line 83 TODO?!)

  Note:

  - The findsimilars script will pick out similar files from them in next test.
  - Let's assume that the number represent the file size in KB.

  == Testing 3 result should be:

  ## =========
             3 'Audio Book - The Grey Coloured Bunnie.mp3' 'test/'
             5 'ColoredGrayBunny.ogg'                      'test/'

  ## =========
             4 'GNU - Python Standard Library (2001).rar' 'test/'
             5 'GNU - 2001 - Python Standard Library.pdf' 'test/'
  ok 3

  Note:

  - There are 2 groups of similar files picked out by the script.
  - The similar files are picked because their file names look similar.
    Note that the first group looks different and spells differently too,
    which means that the script is versatile enough to handle file names that
    don't have space in it, and robust enough to deal with spelling mistakes.
  - Apart from the file name, the file size plays an important role as well.
  - There are 2 files in the second similar files group, the book files group.
  - The file 'Python Standard Library.zip' is not considered to be similar to
    the group because its size is not similar to the group.

  == Testing 4, if Python.zip is bigger, result should be:

  ## =========
             3 'Audio Book - The Grey Coloured Bunnie.mp3' 'test/'
             5 'ColoredGrayBunny.ogg'                      'test/'

  ## =========
             4 'Python Standard Library.zip' 'test/'
             4 'GNU - Python Standard Library (2001).rar' 'test/'
             5 'GNU - 2001 - Python Standard Library.pdf' 'test/'
  ok 4

  Note:

  - There are now 3 files in the book files group.
  - The file 'Python Standard Library.zip' is included in the
    group because its size is now similar to the group.

  == Testing 5, if Python.zip is even bigger, result should be:

  ## =========
             3 'Audio Book - The Grey Coloured Bunnie.mp3' 'test/'
             5 'ColoredGrayBunny.ogg'                      'test/'

  ## =========
             4 'GNU - Python Standard Library (2001).rar' 'test/'
             5 'GNU - 2001 - Python Standard Library.pdf' 'test/'
             6 'Python Standard Library.zip' 'test/'
             9 '(eBook) GNU - Python Standard Library 2001.pdf' 'test/'
  ok 5

  Note:

  - There are 4 files in the book files group now.
  - The file 'Python Standard Library.zip' is still in the group.
  - But this time, because it is also considered to be similar to the .pdf
    file (since their size are now similar, 6 vs 9), a 4th file the .pdf one
    is now included in the book group.
  - If the size of file 'Python Standard Library.zip' is 12(KB), then the
    book files group will be split into two. Do you know why and
    which files each group will contain?

The File::FindSimilars package comes with a fully functional demo
script findsimilars. Please refer to its help file for further
explanations.

This package is highly customizable. Refer to the class method C for
details.

=head1 DEPENDS

This module depends on L, but not L.

=cut

# }}}

# {{{ Global Declaration:

# ============================================================== &us ===
# ............................................................. Uses ...

# -- global modules
use strict;			# !

use Carp;
use Getopt::Long;
use File::Basename;
use Text::Soundex;

use base qw(Class::Accessor::Fast);

# -- local modules

sub dbg_show {};
#use MyDbg; $MyDbg::debugging=010;

# ============================================================== &gv ===
# .................................................. Global Varibles ...
#

our @EXPORT = (  ); # may even omit this line

use vars qw($progname $VERSION $debugging);
use vars qw(%config @filequeue @fileInfo %sdxCnt %wrdLst);

# @fileInfo: List of the following list:
my (
    $N_dName,			# dir name
    $N_fName,			# file name
    $N_fSize,			# file size
    $N_fSdxl,			# file soundex list, reference
    ) = (0..9);

# ============================================================== &cs ===
# ................................................. Constant setting ...
#
$VERSION = sprintf("%d.%02d", q$Revision: 2.6 $ =~ /(\d+)\.(\d+)/);

# }}}

# ############################################################## &ss ###
# ................................................ Subroutions start ...

=head1 METHODS

=head2 File::FindSimilars->new(\%config_param)

Initialize the object.

  my $similars_finder = File::FindSimilars->new();

or,

  my $similars_finder = File::FindSimilars->new( {} );

which are the same as:

  my $similars_finder = File::FindSimilars->new( {
     soundex_weight => 50,	# percentage of weight that soundex takes,
     				# the rest is for file size
     fc_threshold => 75,	# over which files are considered similar
     delimiter => "\n## =========\n",	# delimiter between files output
     format => "%12d '%s' %s'%s'", # file info print format
     fc_level => 0, 		# file comparison level
     verbose => 0,

  } );

What shown above are default settings. Any of the C<%config_param> attribute can be omitted when calling the new method.

The C is the only class method. All the rest methods are object methods.

=head2 Object attribute: soundex_weight([set_val])

Percentage of weight that soundex takes, the rest of percentage is for file size.

Provide the C to change the attribute, omitting it to retrieve the attribute value.

=head2 Object attribute: fc_threshold([set_val])

The threshold over which files are considered similar.

Provide the C to change the attribute, omitting it to retrieve the attribute value.

=head2 Object attribute: delimiter([set_val])

Delimiter printed between file info outputs.

Provide the C to change the attribute, omitting it to retrieve the attribute value.

=head2 Object attribute: format([set_val])

Format used to print file info.

Provide the C to change the attribute, omitting it to retrieve the attribute value.

=head2 Object attribute: fc_level([set_val])

File comparison level. Whether to check similar files within the same folder: 0, no; 1, yes.

Provide the C to change the attribute, omitting it to retrieve the attribute value.

=head2 Object attribute: verbose([set_val])

Verbose level. Whether to output progress info: 0, no; 1, yes.

Provide the C to change the attribute, omitting it to retrieve the attribute value.

=cut

File::FindSimilars
    ->mk_accessors(qw(soundex_weight fc_threshold
	delimiter format fc_level verbose));

%config =
    (

     soundex_weight => 50,	# percentage of weight that soundex takes,
     				# the rest is for file size
     fc_threshold => 75,	# over which files are considered similar
     delimiter => "\n## =========\n",	# delimiter between files output
     format => "%12d '%s' %s'%s'", # file info print format

     fc_level => 0, 		# file comparison level
     verbose => 0,
 );


# =========================================================== &s-sub ===

sub new {
    ref(my $class = shift)
	and croak "new is a class method. class name needed.";
    my ($arg_ref) = @_;
    my $self = $class->SUPER::new({%config, %$arg_ref});
    $config{soundex_weight} = $self->soundex_weight;
    $config{fc_threshold} = $self->fc_threshold;
    $config{delimiter} = $self->delimiter;
    $config{format} = $self->format;
    $config{fc_level} = $self->fc_level;
    $config{verbose} = $self->verbose;
    #$config{} = $self->;
    return $self;
}

# =========================================================== &s-sub ===

=head2 Object method: find_for($array_ref)

Set directory queue for similarity checking. Each entry in C<$array_ref>
is a directory to check into. E.g.,

  $similars_finder->find_for(\@ARGV);

=cut

sub find_for {
    my ($self, $init_dirs) = @_;

    # threshold $config{fc_threshold}
    print STDERR "Searching in directory(ies): @$init_dirs with level $config{fc_level}...\n\n"
	if $config{verbose};

    @filequeue = @fileInfo = ();
    @filequeue = (@filequeue, map { [$_, ''] } @$init_dirs);
    process_entries();

    dbg_show(100,"\@fileInfo", @fileInfo);
    dbg_show(100,"%sdxCnt", %sdxCnt);
    dbg_show(100,"%wrdLst", %wrdLst);
}

# =========================================================== &s-sub ===
# I -  Input: global array @filequeue
#      Input parameters: None
#
sub process_entries {
    my($dir, $qf) = ();
    #warn "] inside process_entries...\n";

    while ($qf = shift @filequeue) {
	($dir, $_) = ($qf->[0], $qf->[1]);
	#warn "] inside process_entries loop, $dir, $_, ...\n";
        next if /^..?$/;
        my $name = "$dir/$_";
	#warn "] processing file '$name'.\n";
	if ($name eq '-/') {
	    # get info from stdin
	    process_stdin();
	}
	elsif (-d $name) {
	    # a directory, process it recursively.
	    process_dir($name);
	}
	else {
	    process_file($dir, $_);
	}
    }
}

# =========================================================== &s-sub ===
# D -  Process info given from stdin, which should of form same as
#       find -printf "%p\t%s\n"
#
sub process_stdin {

  while (<>){
    croak "Wrong input format: '$_'" unless m{(.*)/(.+?)\t(\d+)$};
    my ($dn, $fn, $size) = ( $1, $2, $3 );
    my $fSdxl = [ get_soundex($fn) ]; # file soundex list
    push @fileInfo, [ $dn, $fn, $size, $fSdxl, ];

    dbg_show(100,"fileInfo",@fileInfo);
    map { $sdxCnt{$_}++ } @$fSdxl;
  }
}

# =========================================================== &s-sub ===
# D -  Process given dir recursively
# N -  BFS is more memory friendly than DFS
#
# T -  $dir="/home/tong/tmp"
sub process_dir {
    my($dir) = @_;
    #warn "] processing dir '$dir'...\n";

    opendir(DIR,$dir) || die "File::FindSimilars error: Can't open $dir";
    my @filenames = readdir(DIR);
    closedir(DIR);

    # record the dirname/fname pair to queue
    @filequeue = (@filequeue, map { [$dir, $_] } @filenames);
    dbg_show(100,"filequeue", @filequeue)
}

# =========================================================== &s-sub ===
# S -  process_file($dirname, $fname), process file $fname under $dirname
# D -  Process one file and update global vars
# U -
#
# I -  Input parameters:
#	$dirname: dir name string
#	$fname:	 file name string
# O -  Global vars get updated
#      fileInfo [ $dirname, $fname, $fsize, [ file_soundex ] ]
# T -

sub process_file {
    my ($dn, $fn) = @_;
    my ($dev,$ino,$mode,$nlink,$uid,$gid,$rdev,$size,@rest) =
	stat("$dn/$fn");
    my $fSdxl = [ get_soundex($fn) ]; # file soundex list
    push @fileInfo, [ $dn, $fn, $size, $fSdxl, ];

    dbg_show(100,"fileInfo",@fileInfo);
    map { $sdxCnt{$_}++ } @$fSdxl;
}

# =========================================================== &s-sub ===
# S -  get_soundex($fname), get soundex for file $fname
# D -  Return a list of soundex of each individual word in file name
# U -  $aref = [ get_soundex($fname) ];
#
# I -  Input parameters:
#	$fname:	 file name string
# O -  sorted anonymous soundex array w/ duplications removed
# T -  @out = get_soundex 'Java_RMI - _Remote_Method_Invocation_ch03.tgz';
#      @out = get_soundex 'ASuchKindOfFile.tgz';

sub get_soundex {
    my ($fn) = @_;
    # split to individual words
    my @fn_wlist = split /[-_[:cntrl:][:blank:][:punct:][:digit:]]/i, $fn;
    # discards file extension, if any
    pop @fn_wlist if @fn_wlist >= 1;
    # if it is single word, try further decompose SuchKindOfWord
    @fn_wlist = $fn_wlist[0] =~ /[A-Z][^A-Z]*/g
	if @fn_wlist == 1 && $fn_wlist[0] =~ /^[A-Z]/;
    # wash short
    dbg_show(100,"wlist 0",@fn_wlist);
    @fn_wlist = arrwash_short(\@fn_wlist);
    dbg_show(100,"wlist 1",@fn_wlist);

    # language specific handling
    @fn_wlist = arrwash_lang(\@fn_wlist);
    dbg_show(100,"wlist 2",@fn_wlist);

    # change word to soundex, record soundex/word in global hash
    map {
	if (/[[:alpha:]]/) {
	    my $sdx = soundex($_);
	    $wrdLst{$sdx}{$_}++;
	    s/^.*$/$sdx/;
	    }
	} @fn_wlist;
    dbg_show(1,"wrdLst",%wrdLst);

    # wash empty/duplicates
    @fn_wlist = grep(!/^$/, @fn_wlist);
    @fn_wlist = arrwash_dup(\@fn_wlist);

    return sort @fn_wlist;
}

# =========================================================== &s-sub ===
# S -  arrwash_short($arr_ref), wash short from array $arr_ref
# D -  weed out empty lines and less-than-3-letter words (e.g. ch12)
# U -  @fn_wlist = arrwash_short(\@fn_wlist);
#

sub arrwash_short($) {
    my ($arr_ref) = @_;
    return @$arr_ref unless @$arr_ref >= 1;
    my @r= grep tr/a-zA-Z// >=3, @$arr_ref;
    return @r if @r;
    return @$arr_ref		# for upper ASCII
	if grep(/[\200-\377]/, @$arr_ref);
    return @r;
}

# =========================================================== &s-sub ===
# S -  arrwash_dup($arr_ref), wash duplicates from array $arr_ref
# D -  weed out duplicates
# U -  @fn_wlist = arrwash_dup(\@fn_wlist);
#

sub arrwash_dup($) {
    my ($arr_ref) = @_;
    my %saw;
    return grep !$saw{$_}++, @$arr_ref;
}

# =========================================================== &s-sub ===
# S -  arrwash_lang($arr_ref), language specific washing from array $arr_ref
# U -  @fn_wlist = arrwash_lang(\@fn_wlist);
#

sub arrwash_lang($) {
    my ($arr_ref) = @_;

    # split Chinese into individual chars
    my @r;
    map {
	if (/[\200-\377]{2}/) {
	    @r = (@r, /[\200-\377]{2}/g);
	}
	else {
	    @r = (@r, $_);
	}
    } @$arr_ref;

    return @r;
}

=head2 Object method: similarity_check()

Do similarity check on the queued directories.  Print similar files info on
stdout according to the configured format and delimiters. E.g.,

  $similars_finder->similarity_check();

=cut

# =========================================================== &s-sub ===
# S -  similarity_check: similarity check on glabal array @fileInfo
# U -  similarity_check();
#
# I -  Input parameters: None
# O -  similar files printed on stdout

sub similarity_check {

    # get a ordered (by soundex count and file name) of file Info array
    # (Use short file names to compare to long file names)
    # use Schwartzian Transform to sort on 2 fields for efficiency
    my @fileInfos = map { $_->[0] }
	sort { $a->[1] cmp $b->[1] }
	    map  { [ $_,
		     sprintf "%3d%6s", $#{$_->[$N_fSdxl]}, $_->[$N_fSdxl][0]
		] } @fileInfo;
    dbg_show(100,"\@fileInfos", @fileInfos);

    my @saw = (0) x ($#fileInfos+1);
    foreach my $ii (0..$#fileInfos) {
	#warn "] ii=$ii\n";
	my @similar = ();
	my $fnl;

	dbg_show(100,"\@fileInfos", $fileInfos[$ii]);
	push @similar, [$ii, $ii, $fileInfos[$ii]->[$N_fSize] ];
	foreach my $jj (($ii+1) ..$#fileInfos) {
	    $fnl=0;		# 0 is good enough since file at [ii] is
				# shorter in name than  the one at [jj]
	    # don't care about same dir files?
	    next
		if (!$config{fc_level} && ($fileInfos[$ii]->[$N_dName]
		    eq $fileInfos[$jj]->[$N_dName])) ;
	    if (file_diff(\@fileInfos, $ii, $jj) >= $config{fc_threshold}) {
		push @similar, [$ii, $jj, $fileInfos[$jj]->[$N_fSize] ];
		$fnl= length($fileInfos[$jj]->[$N_fName]) if
		    $fnl < length($fileInfos[$jj]->[$N_fName]);
	    }
	}
	dbg_show(100,"\@similar", @similar);
	# output unvisited potential similars by each row, order by fSize
	@similar = grep {!$saw[$_->[1]]}
	  sort { $a->[2] <=> $b->[2] } @similar;
	next unless @similar>1;
	print $config{delimiter};
	foreach my $similar (@similar) {
	    print file_info(\@fileInfos, $similar->[1], $fnl). "\n";
	    $saw[$similar->[1]]++;
	}
    }
}

# =========================================================== &s-sub ===
sub file_info ($$$) {
    my ($fileInfos, $ndx, $fnl) = @_;
    return sprintf($config{format}, $fileInfos->[$ndx]->[$N_fSize],
		   $fileInfos->[$ndx]->[$N_fName],
		   ' ' x ($fnl - length($fileInfos->[$ndx]->[$N_fName])),
		   "$fileInfos->[$ndx]->[$N_dName]");
}

# =========================================================== &s-sub ===
# S -  file_diff: determind how difference two files are by name & size
# U -  file_diff($fileInfos, $ndx1, $ndx2);
#
# I -  $fileInfos:	reference to @fileInfos
#	$ndx1, $ndx2:	index to the two file in @fileInfos
# O -  100%: files are identical
#	0%: no similarity at all
sub file_diff ($$$) {
    my ($fileInfos, $ndx1, $ndx2) = @_;

    return 0 unless @{$fileInfos->[$ndx1]->[$N_fSdxl]};

    # find intersection in two soudex array
    my %count = ();
    foreach my $element
	(@{$fileInfos->[$ndx1]->[$N_fSdxl]},
	 @{$fileInfos->[$ndx2]->[$N_fSdxl]}) { $count{$element}++ }
    # since there is no duplication in each of file soudex
    my $intersection =
	grep $count{$_} > 1, keys %count;
    # return p * normal(\common soudex) + (1-p) * ( 1 - normal(\delta fSize))
    # so the bigger the return value is, the similar the two files are
    $intersection *= $config{soundex_weight} /
	(@{$fileInfos->[$ndx1]->[$N_fSdxl]});
    dbg_show(100,"intersection", $intersection, $ndx1, $ndx2);
    my $WeightfSzie = 100 - $config{soundex_weight};
    my $dfSize = abs($fileInfos->[$ndx1]->[$N_fSize] -
		     $fileInfos->[$ndx2]->[$N_fSize]) * $WeightfSzie /
		($fileInfos->[$ndx1]->[$N_fSize] + 1);
    $dfSize = $dfSize > $WeightfSzie ? $WeightfSzie : $dfSize;
    my $file_diff = $intersection + ($WeightfSzie - $dfSize);
    if ($file_diff >= $config{fc_threshold}) {
	dbg_show(010,"file_diff",
		 @{$fileInfos->[$ndx1]},
		 @{$fileInfos->[$ndx2]},
		 $intersection, $dfSize, $file_diff
		 );
    }
    return $file_diff;
}


1;
__END__