* `optimise-store.pl' reduces disk space consumption by hard-linking

all identitical files in the Nix store to each other.  (Previously
  it only computed the size that would be saved by doing so.)
This commit is contained in:
Eelco Dolstra 2006-09-19 14:58:35 +00:00
parent 07cec27848
commit 6dbed1bf38

View file

@ -1,61 +1,85 @@
#! /usr/bin/perl -w #! /usr/bin/perl -w
use strict; use strict;
use File::Basename;
#{ my $ofh = select STDOUT;
# $| = 1;
# select $ofh;
#}
#my @paths = ("/nix/store/d49mc94xwwd7wf1xzfh4ch4cypn0ajjr-glibc-2.3.6", "/nix/store/1mgfgy3ga4m9z60747s0yzxl0g6w5kxz-glibc-2.3.6");
my @paths = ("/nix/store"); my @paths = ("/nix/store");
print "hashing...\n";
my $hashList = "/tmp/nix-optimise-hash-list"; my $hashList = "/tmp/nix-optimise-hash-list";
system("find @paths -type f -print0 | xargs -0 md5sum -- > $hashList") == 0 system("find @paths -type f -print0 | xargs -0 md5sum -- > $hashList") == 0
or die "cannot hash store files"; or die "cannot hash store files";
print "sorting by hash...\n";
system("sort $hashList > $hashList.sorted") == 0 system("sort $hashList > $hashList.sorted") == 0
or die "cannot sort list"; or die "cannot sort list";
sub atomicLink {
my $target = shift;
my $new = shift;
my $tmpNew = "${new}_optimise.$$";
# Make the directory writable temporarily.
my $dir = dirname $new;
my @st = stat $dir or die;
chmod ($st[2] | 0200, $dir) or die "cannot make `$dir' writable: $!";
link $target, $tmpNew or die "cannot create hard link `$tmpNew': $!";
rename $tmpNew, $new or die "cannot rename `$tmpNew' to `$new': $!";
chmod ($st[2], $dir) or die "cannot restore permission on `$dir': $!";
}
print "hard-linking...\n";
open LIST, "<$hashList.sorted" or die; open LIST, "<$hashList.sorted" or die;
my $prevFile; my $prevFile;
my $prevHash; my $prevHash;
my $prevInode;
my $totalSpace = 0; my $totalSpace = 0;
my $savedSpace = 0; my $savedSpace = 0;
my $files = 0;
while (<LIST>) { while (<LIST>) {
/^([0-9a-f]*)\s+(.*)$/ or die; /^([0-9a-f]*)\s+(.*)$/ or die;
my $curFile = $2; my $curFile = $2;
my $curHash = $1; my $curHash = $1;
my $fileSize = (stat $curFile)[7]; my @st = stat $curFile or die;
next if ($st[2] & 0222) != 0; # skip writable files
my $fileSize = $st[7];
$totalSpace += $fileSize; $totalSpace += $fileSize;
if (defined $prevHash && $curHash eq $prevHash) { if (defined $prevHash && $curHash eq $prevHash) {
print "$curFile = $prevFile\n"; if ($st[1] != $prevInode) {
print "$curFile = $prevFile\n";
$savedSpace += $fileSize; atomicLink $prevFile, $curFile;
$savedSpace += $fileSize;
}
} else { } else {
$prevFile = $curFile; $prevFile = $curFile;
$prevHash = $curHash; $prevHash = $curHash;
$prevInode = $st[1];
} }
# print "." if ($files++ % 100 == 0);
} }
#print "\n";
print "total space = $totalSpace\n"; print "total space = $totalSpace\n";
print "saved space = $savedSpace\n"; print "saved space = $savedSpace\n";
my $savings = ($savedSpace / $totalSpace) * 100.0; my $savings = ($savedSpace / $totalSpace) * 100.0;
print "savings = $savings %\n"; print "savings = $savings %\n";
close LIST; close LIST;