X-Git-Url: https://git.deb.at/?p=deb%2Fpackages.git;a=blobdiff_plain;f=bin%2Fparse-contents;h=12b7ed824a40dbb933d0bc7b7a10e675153b9081;hp=50f5ee2982d9700e5ae5fae262770025e5b8ea1b;hb=56d4b3ed57e20359e5ea46feabb850cf4a94caf2;hpb=9d463c289353df50118033b581bc60bd291c9d01 diff --git a/bin/parse-contents b/bin/parse-contents index 50f5ee2..12b7ed8 100755 --- a/bin/parse-contents +++ b/bin/parse-contents @@ -22,40 +22,62 @@ use strict; use lib './lib'; +$| = 1; + +# Important, we want sorting and such to happen like in the C locale: binary, +# without any fancy collation. FIXME: is this actually adequate? +$ENV{"LC_ALL"} = 'C'; + my $what = $ARGV[0] ? "head -10000|" : ""; +# More RAM vs more disk I/O tradeoff parameters, does not change +# functionality. True will always use more RAM at the benefit of less +# temporary files, and is adviced when possible +my $SORT_REVERSE_CONCURRENTLY = 1; + use DB_File; use Storable; +use File::Path; use Packages::Config qw( $TOPDIR $DBDIR @ARCHIVES @SUITES @ARCHITECTURES ); &Packages::Config::init( './' ); -my %filenames = (); my @archives =( 'us'); #@ARCHIVES # NOT-IMPLEMENTED-YET my @suites = @SUITES; my @archs = @ARCHITECTURES; -for my $archive (@archives) { for my $suite (@suites) { for my $arch (@archs) { +$DBDIR .= "/contents"; +-d $DBDIR || mkpath( $DBDIR ); + +for my $archive (@archives) { for my $suite (@suites) { + + for my $arch (@archs) { my $filename = "$TOPDIR/archive/$archive/$suite/Contents-$arch.gz"; - my $db = "$DBDIR/packages_contents_${suite}_${arch}.db"; + my $filelist_db = "$DBDIR/filelists_${suite}_${arch}.db"; next unless -f $filename; - my $ftime = (stat $filename)[9]; - my $dbtime = (stat $db)[9]; - next unless $ftime > $dbtime; + my $ftime = (stat $filename)[10]; # Note: ctime, because mtime is set back via rsync + my $dbtime = (stat $filelist_db)[9]; + next if defined $dbtime and $dbtime > $ftime; print "Reading $archive/$suite/$arch...\n"; my %packages_contents = (); my %packages_contents_nr = (); my %packages_contents_lastword = (); - my %contents_packages_reverse = (); - open CONT, "zcat $filename|$what"; + my $extra = ""; + $extra = "|sort" if $SORT_REVERSE_CONCURRENTLY; + + open REVERSED, "$extra>$DBDIR/reverse.tmp" + or die "Failed to open output reverse file: $!"; + + open CONT, "zcat $filename|$what" + or die $!; while () {last if /^FILE/mo;} while () { my $data = ""; my %data = (); chomp; - print "Doing line $.\n" if $. % 10000 == 0; + print "Doing line ".($./1000)."k (out of approx 1.5M)\n" if $. % 250000 == 0; /^(.+?)\s+(\S+)$/o; my ($file, $value) = ($1, $2); $value =~ s#[^,/]+/##og; @@ -72,23 +94,19 @@ for my $archive (@archives) { for my $suite (@suites) { for my $arch (@archs) { } # Searches are case-insensitive $file =~ tr [A-Z] [a-z]; - my $filename = $file; - $filename =~ s,.*/,,; - $filenames{$filename} = 1; - $contents_packages_reverse{reverse $file} = join "\0", @packages; + print REVERSED (reverse $file)."\0".(join ":$arch\0", @packages).":$arch\n"; } - my %contents_packages_reverse_db; - tie %contents_packages_reverse_db, "DB_File", "$DBDIR/contents_packages_reverse_${suite}_${arch}.db.new", - O_RDWR|O_CREAT, 0666, $DB_BTREE - or die "Error creating DB: $!"; - while (my ($x, $y) = each(%contents_packages_reverse)) { - $contents_packages_reverse_db{$x} = $y; - } - untie %contents_packages_reverse_db; + close CONT; + close REVERSED; + + print "Sorting reverse list if needed\n"; + system("cd $DBDIR && sort reverse.tmp > reverse.sorted && mv reverse.{sorted,tmp}") == 0 + or die "Failed to sort reverse" + unless $SORT_REVERSE_CONCURRENTLY; - my %packages_contents_db; - tie %packages_contents_db, "DB_File", "$DBDIR/packages_contents_${suite}_${arch}.db.new", + print "Writing filelist db\n"; + tie my %packages_contents_db, "DB_File", "$filelist_db.new", O_RDWR|O_CREAT, 0666, $DB_BTREE or die "Error creating DB: $!"; while (my ($k, $v) = each(%packages_contents)) { @@ -96,20 +114,66 @@ for my $archive (@archives) { for my $suite (@suites) { for my $arch (@archs) { . $v; } untie %packages_contents_db; -}}} - -print "Writing databases...\n"; - -# FIXME: missing filenames due to optimising above. Need to store filenames -# per-suite/arch, but merge them in the end for better cached searching -open FILENAMES, "> $DBDIR/filenames.txt.new"; -for (keys %filenames) { - print FILENAMES "$_\n"; -} -close FILENAMES; - -rename("$DBDIR/filenames.txt.new", "$DBDIR/filenames.txt"); -for my $archive (@archives) { for my $suite (@suites) { for my $arch (@archs) { - rename("$DBDIR/packages_contents_${suite}_${arch}.db.new", "$DBDIR/packages_contents_${suite}_${arch}.db"); - rename("$DBDIR/contents_packages_reverse_${suite}_${arch}.db.new", "$DBDIR/contents_packages_reverse_${suite}_${arch}.db"); -}}} + + rename("$DBDIR/reverse.tmp", "$DBDIR/reverse_${suite}_${arch}.txt"); + + rename("$filelist_db.new", $filelist_db); + system("ln -sf $filelist_db $DBDIR/filelists_${suite}_all.db") == 0 + or die "Oops"; + } + + + my $go = 0; + my $suite_mtime = (stat "$DBDIR/reverse_$suite.db")[9]; + for my $file (glob "$DBDIR/reverse_${suite}_*.txt") { + $go = 1 if not defined $suite_mtime + or $suite_mtime < (stat $file)[9]; + } + next unless $go; + + print "Merging reverse path lists for ${suite}...\n"; + + open MERGED, "sort -m $DBDIR/reverse_${suite}_*.txt |" + or die "Failed to open merged list"; + open FILENAMES, "> $DBDIR/filenames_$suite.txt.new" + or die "Failed to open filenames list"; + tie my %reverse_path_db, "DB_File", "$DBDIR/reverse_${suite}.db.new", + O_RDWR|O_CREAT, 0666, $DB_BTREE + or die "Error creating DB: $!"; + + my $lastpath = ""; + my $lastfile = ""; + my @matches = (); + while () { + print "Doing line ".($./1000000)."M (out of approx. 16M)\n" if $. % 1000000 == 0; + chomp; + my @line = split /\0/o, $_; + my $revpath = shift @line; + if ($revpath ne $lastpath) { + # Wrap: Do useful stuff with this ($lastpath, @matches) + $reverse_path_db{$lastpath} = join "\0", @matches if $lastpath ne ""; + $lastpath =~ s,/.*,,o; + if ($lastfile ne $lastpath) { + $lastfile = $lastpath; + print FILENAMES (reverse $lastfile)."\n"; + } + # + $lastpath = $revpath; + @matches = @line; + next; + } + push @matches, @line + } + # Note: do useful stuff here too, for out last entry. Maybe prevent this by + # adding a fake ultimate entry? + $reverse_path_db{$lastpath} = join "\0", @matches; + + untie %reverse_path_db; + close FILENAMES; + close MERGED; + + rename "$DBDIR/filenames_$suite.txt.new", "$DBDIR/filenames_$suite.txt"; + rename "$DBDIR/reverse_$suite.db.new", "$DBDIR/reverse_$suite.db"; +}} + +# vim: set ts=4