tangetools/rrm/rclean

#!/usr/bin/perl

use Digest::MD5::File qw(dir_md5_hex file_md5_hex url_md5_hex);
use Digest::SHA;

my $dir = shift || ".";

chdir $dir;


# Table of which files have a given size
open(IN,"-|",'find "`pwd`" -type f -printf "%s\t%p\0"') || die;
$/="\0";
my %size;
while(<IN>) {
    chop; # Remove \0
    my($s,$f) = split /\t/,$_;
    push @{$size{$s}}, $f;
}
close IN;

# Read hash-value of removed files of a given size
my %rrm;
my $rrmfile = find_rrm_file(".") || ".rrm";
if(not -r $rrmfile) {
    print STDERR "Cannot read $rrmfile\n";
    exit(1);
}
    
open(RRM,"<",$rrmfile) || die;
while(<RRM>) {
    my($size,$hashval,$file) = split /\t/,$_;
    $rrm{0+$size}{$hashval}++;
}
close RRM;

# Which existing files are the same size as some of the removed files?
for my $size (keys %rrm) {
    for my $file (@{$size{$size}}) {
	if(-e $file) {
	    my $md5 = Digest::MD5->new;	
	    $md5->addpath($file);
	    my $sha = Digest::SHA->new(256);
	    $sha->addfile($file);
	    # Do they have the same hash-value?
	    if($rrm{$size}{$md5->hexdigest} or
	       $rrm{$size}{"SHA256:".$sha->b64digest}) {
		# remove this
		print "$file\n";
	    }
	}
    }
}

sub find_rrm_file {
    my $dir = shift;
    if(-r "$dir/.rrm") {
	return "$dir/.rrm";
    }
    if(join(" ",stat $dir) eq join(" ",stat "$dir/..")) {
	# root
	return undef;
    } else {
	return find_rrm_file("$dir/..");
    } 
}