X-Git-Url: https://git.deb.at/?a=blobdiff_plain;f=bin%2Fparse-contents;h=341ecd5783394cfcef859fae632d5896432cdc6a;hb=9c10f84340c63159bd3d242bab1fb3d9615eae3f;hp=ceb0cb0e5bce0eea3b958b93d6e5dfc4bc02a2d4;hpb=eca3f1d1d3ddba9041b4cae9cc7c54bcb95cc3ef;p=deb%2Fpackages.git diff --git a/bin/parse-contents b/bin/parse-contents index ceb0cb0..341ecd5 100755 --- a/bin/parse-contents +++ b/bin/parse-contents @@ -22,33 +22,60 @@ use strict; use lib './lib'; -my $what = $ARGV[0] ? "head -100000|" : ""; +# Important, we want sorting and such to happen like in the C locale: binary, +# without any fancy collation. FIXME: is this actually adequate? +$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 Packages::Config qw( $TOPDIR $DBDIR @ARCHIVES @SUITES ); +use Packages::Config qw( $TOPDIR $DBDIR @ARCHIVES @SUITES @ARCHITECTURES ); &Packages::Config::init( './' ); -my %packages_contents = (); -my %packages_contents_nr = (); -my %packages_contents_lastword = (); -my %file_reverse = (); -my @archives =( 'us'); #@ARCHIVES -my @suites = ('stable');#@SUITES +my @archives =( 'us'); #@ARCHIVES # NOT-IMPLEMENTED-YET +my @suites = @SUITES; +my @archs = @ARCHITECTURES; + +$DBDIR .= "/contents"; for my $archive (@archives) { for my $suite (@suites) { - print "Reading $archive/$suite/i386...\n"; - open CONT, "zcat /org/ftp.debian.org/ftp/dists/stable/Contents-i386.gz|$what"; - while (1) {$_ = ;last if /^FILE/mo;} + for my $arch (@archs) { + + my $filename = "$TOPDIR/archive/$archive/$suite/Contents-$arch.gz"; + my $filelist_db = "$DBDIR/filelists_${suite}_${arch}.db"; + next unless -f $filename; + 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 $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"; + while () {last if /^FILE/mo;} while () { my $data = ""; my %data = (); chomp; - print "Doing line $.\n" if $. % 10000 == 0; - /^(\S+)\s+(\S+)/; + print "Doing line ".($./1000)."k (out of approx 1.5M)\n" if $. % 250000 == 0; + /^(.+?)\s+(\S+)$/o; my ($file, $value) = ($1, $2); - $value =~ s#[^,/]+/##g; + $value =~ s#[^,/]+/##og; my @packages = split /,/, $value; for (@packages) { $packages_contents_nr{$_}++; @@ -63,36 +90,88 @@ for my $archive (@archives) { for my $suite (@suites) { # Searches are case-insensitive $file =~ tr [A-Z] [a-z]; - $file_reverse{reverse $file} = join "\0", @packages; + print REVERSED (reverse $file)."\0".(join ":$arch\0", @packages).":$arch\n"; + } + 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 $SORT_REVERSE_CONCURRENTLY; + + print "Writing filelist db\n"; + tie my %packages_contents_db, "DB_File", "$filelist_db.new", + O_RDWR|O_CREAT, 0666, $DB_BTREE + or die "Error creating DB: $!"; + while (my ($k, $v) = each(%packages_contents)) { + $packages_contents_db{$k} = (pack "L", $packages_contents_nr{$k}) + . $v; } + untie %packages_contents_db; + + rename("$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"; + } + + 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"; }} -print "Writing databases...\n"; -my %packages_contents_db; -tie %packages_contents_db, "DB_File", "$DBDIR/packages_contents.db.new", - O_RDWR|O_CREAT, 0666, $DB_BTREE - or die "Error creating DB: $!"; -while (my ($k, $v) = each(%packages_contents)) { - $packages_contents_db{$k} = (pack "L", $packages_contents_nr{$k}) - . $v; -} -untie %packages_contents_db; - -my %file_reverse_db; -tie %file_reverse_db, "DB_File", "$DBDIR/file_reverse.db.new", - O_RDWR|O_CREAT, 0666, $DB_BTREE - or die "Error creating DB: $!"; -while (my ($x, $y) = each(%file_reverse)) { -# $v =~ s/.$//s; -# my $nr = $v; -# $nr =~ s/[^\000]//g; -# $nr = length($nr) + 1; # < number of hits -# if ($nr > $MAX_file_reverse) { -# $v = "\001" . $nr; -# } - $file_reverse_db{$x} = $y; -} -untie %file_reverse_db; - -rename("$DBDIR/packages_contents.db.new", "$DBDIR/packages_contents.db"); -rename("$DBDIR/file_reverse.db.new", "$DBDIR/file_reverse.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"); + +# vim: set ts=4