X-Git-Url: http://www.git.stargrave.org/?a=blobdiff_plain;f=lib%2FPublicInbox%2FSearch.pm;h=5e820594c3777b322b2a36997cb08d50c95e9403;hb=227a1d886672767e37cc86a3432952c14eb8a143;hp=bb4bd454dd9242d51dd886ad3ae57a1c320c3e8e;hpb=c23df198cb3b101c8f0ede73092c8f3dae5eda35;p=public-inbox.git diff --git a/lib/PublicInbox/Search.pm b/lib/PublicInbox/Search.pm index bb4bd454..5e820594 100644 --- a/lib/PublicInbox/Search.pm +++ b/lib/PublicInbox/Search.pm @@ -1,4 +1,4 @@ -# Copyright (C) 2015-2018 all contributors +# Copyright (C) 2015-2019 all contributors # License: AGPL-3.0+ # based on notmuch, but with no concept of folders, files or flags # @@ -12,15 +12,48 @@ use constant TS => 0; # Received: header in Unix time use constant YYYYMMDD => 1; # Date: header for searching in the WWW UI use constant DT => 2; # Date: YYYYMMDDHHMMSS -use Search::Xapian qw/:standard/; use PublicInbox::SearchMsg; -use PublicInbox::MIME; -use PublicInbox::MID qw/id_compress/; use PublicInbox::Over; +my $QP_FLAGS; +our %X = map { $_ => 0 } qw(BoolWeight Database Enquire + NumberValueRangeProcessor QueryParser Stem); +our $Xap; # 'Search::Xapian' or 'Xapian' +my $ENQ_ASCENDING; + +sub load_xapian () { + return 1 if defined $Xap; + for my $x (qw(Search::Xapian Xapian)) { + eval "require $x"; + next if $@; + + $x->import(qw(:standard)); + $Xap = $x; + $X{$_} = $Xap.'::'.$_ for (keys %X); + + # ENQ_ASCENDING doesn't seem exported by SWIG Xapian.pm, + # so lets hope this part of the ABI is stable because it's + # just an integer: + $ENQ_ASCENDING = $x eq 'Xapian' ? + 1 : Search::Xapian::ENQ_ASCENDING(); + + # for SearchMsg: + *PublicInbox::SearchMsg::sortable_unserialise = + $Xap.'::sortable_unserialise'; + # n.b. FLAG_PURE_NOT is expensive not suitable for a public + # website as it could become a denial-of-service vector + # FLAG_PHRASE also seems to cause performance problems chert + # (and probably earlier Xapian DBs). glass seems fine... + # TODO: make this an option, maybe? + # or make indexlevel=medium as default + $QP_FLAGS = FLAG_PHRASE() | FLAG_BOOLEAN() | FLAG_LOVEHATE() | + FLAG_WILDCARD(); + return 1; + } + undef; +} # This is English-only, everything else is non-standard and may be confused as # a prefix common in patch emails -our $REPLY_RE = qr/^re:\s+/i; our $LANG = 'english'; use constant { @@ -41,15 +74,10 @@ use constant { # 13 - fix threading for empty References/In-Reply-To # (commit 83425ef12e4b65cdcecd11ddcb38175d4a91d5a0) # 14 - fix ghost root vivification + # 15 - see public-inbox-v2-format(5) + # further bumps likely unnecessary, we'll suggest in-place + # "--reindex" use for further fixes and tweaks SCHEMA_VERSION => 15, - - # n.b. FLAG_PURE_NOT is expensive not suitable for a public website - # as it could become a denial-of-service vector - # - # FLAG_PHRASE also seems to cause performance problems sometimes. - # TODO: make this an option, maybe? - # or make indexlevel=medium as default - QP_FLAGS => FLAG_BOOLEAN|FLAG_LOVEHATE|FLAG_WILDCARD, }; my %bool_pfx_external = ( @@ -118,58 +146,70 @@ EOF ); chomp @HELP; -sub xdir { - my ($self) = @_; +sub xdir ($;$) { + my ($self, $rdonly) = @_; if ($self->{version} == 1) { - "$self->{mainrepo}/public-inbox/xapian" . SCHEMA_VERSION; + "$self->{inboxdir}/public-inbox/xapian" . SCHEMA_VERSION; } else { - my $dir = "$self->{mainrepo}/xap" . SCHEMA_VERSION; - my $part = $self->{partition}; - defined $part or die "partition not given"; - $dir .= "/$part"; + my $dir = "$self->{inboxdir}/xap" . SCHEMA_VERSION; + return $dir if $rdonly; + + my $shard = $self->{shard}; + defined $shard or die "shard not given"; + $dir .= "/$shard"; } } -sub new { - my ($class, $mainrepo, $altid) = @_; - my $version = 1; - my $ibx = $mainrepo; - if (ref $ibx) { - $version = $ibx->{version} || 1; - $mainrepo = $ibx->{mainrepo}; - } - my $self = bless { - mainrepo => $mainrepo, - altid => $altid, - version => $version, - }, $class; - my $dir; - if ($version >= 2) { - $dir = "$self->{mainrepo}/xap" . SCHEMA_VERSION; - my $xdb; - my $parts = 0; - foreach my $part (<$dir/*>) { - -d $part && $part =~ m!/\d+\z! or next; - $parts++; - my $sub = Search::Xapian::Database->new($part); +sub _xdb ($) { + my ($self) = @_; + my $dir = xdir($self, 1); + my ($xdb, $slow_phrase); + my $qpf = \($self->{qp_flags} ||= $QP_FLAGS); + if ($self->{version} >= 2) { + foreach my $shard (<$dir/*>) { + -d $shard && $shard =~ m!/[0-9]+\z! or next; + my $sub = $X{Database}->new($shard); if ($xdb) { $xdb->add_database($sub); } else { $xdb = $sub; } + $slow_phrase ||= -f "$shard/iamchert"; } - $self->{xdb} = $xdb; } else { - $dir = $self->xdir; - $self->{xdb} = Search::Xapian::Database->new($dir); + $slow_phrase = -f "$dir/iamchert"; + $xdb = $X{Database}->new($dir); } + $$qpf |= FLAG_PHRASE() unless $slow_phrase; + $xdb; +} + +sub xdb ($) { + my ($self) = @_; + $self->{xdb} ||= do { + load_xapian(); + _xdb($self); + }; +} + +sub new { + my ($class, $ibx) = @_; + ref $ibx or die "BUG: expected PublicInbox::Inbox object: $ibx"; + my $self = bless { + inboxdir => $ibx->{inboxdir}, + altid => $ibx->{altid}, + version => $ibx->version, + }, $class; + my $dir = xdir($self, 1); $self->{over_ro} = PublicInbox::Over->new("$dir/over.sqlite3"); $self; } sub reopen { my ($self) = @_; - $self->{xdb}->reopen; + if (my $xdb = $self->{xdb}) { + $xdb->reopen; + } $self; # make chaining easier } @@ -180,36 +220,35 @@ sub query { if ($query_string eq '' && !$opts->{mset}) { $self->{over_ro}->recent($opts); } else { - my $query = $self->qp->parse_query($query_string, QP_FLAGS); + my $qp = qp($self); + my $qp_flags = $self->{qp_flags}; + my $query = $qp->parse_query($query_string, $qp_flags); $opts->{relevance} = 1 unless exists $opts->{relevance}; _do_enquire($self, $query, $opts); } } -sub get_thread { - my ($self, $mid, $prev) = @_; - $self->{over_ro}->get_thread($mid, $prev); -} - sub retry_reopen { - my ($self, $cb) = @_; + my ($self, $cb, $arg) = @_; for my $i (1..10) { if (wantarray) { my @ret; - eval { @ret = $cb->() }; + eval { @ret = $cb->($arg) }; return @ret unless $@; } else { my $ret; - eval { $ret = $cb->() }; + eval { $ret = $cb->($arg) }; return $ret unless $@; } # Exception: The revision being read has been discarded - # you should call Xapian::Database::reopen() - if (ref($@) eq 'Search::Xapian::DatabaseModifiedError') { + if (ref($@) =~ /\bDatabaseModifiedError\b/) { warn "reopen try #$i on $@\n"; reopen($self); } else { - warn "ref: ", ref($@), "\n"; + # let caller decide how to spew, because ExtMsg queries + # get wonky and trigger: + # "something terrible happened at .../Xapian/Enquire.pm" die; } } @@ -218,18 +257,19 @@ sub retry_reopen { sub _do_enquire { my ($self, $query, $opts) = @_; - retry_reopen($self, sub { _enquire_once($self, $query, $opts) }); + retry_reopen($self, \&_enquire_once, [ $self, $query, $opts ]); } -sub _enquire_once { - my ($self, $query, $opts) = @_; - my $enquire = Search::Xapian::Enquire->new($self->{xdb}); +sub _enquire_once { # retry_reopen callback + my ($self, $query, $opts) = @{$_[0]}; + my $xdb = xdb($self); + my $enquire = $X{Enquire}->new($xdb); $enquire->set_query($query); $opts ||= {}; my $desc = !$opts->{asc}; if (($opts->{mset} || 0) == 2) { - $enquire->set_docid_order(Search::Xapian::ENQ_ASCENDING()); - $enquire->set_weighting_scheme(Search::Xapian::BoolWeight->new); + $enquire->set_docid_order($ENQ_ASCENDING); + $enquire->set_weighting_scheme($X{BoolWeight}->new); } elsif ($opts->{relevance}) { $enquire->set_sort_by_relevance_then_value(TS, $desc); } else { @@ -239,16 +279,14 @@ sub _enquire_once { my $limit = $opts->{limit} || 50; my $mset = $enquire->get_mset($offset, $limit); return $mset if $opts->{mset}; - my @msgs = map { - PublicInbox::SearchMsg->load_doc($_->get_document); - } $mset->items; + my @msgs = map { PublicInbox::SearchMsg::from_mitem($_) } $mset->items; return \@msgs unless wantarray; ($mset->get_matches_estimated, \@msgs) } # read-write -sub stemmer { Search::Xapian::Stem->new($LANG) } +sub stemmer { $X{Stem}->new($LANG) } # read-only sub qp { @@ -256,18 +294,17 @@ sub qp { my $qp = $self->{query_parser}; return $qp if $qp; - + my $xdb = xdb($self); # new parser - $qp = Search::Xapian::QueryParser->new; - $qp->set_default_op(OP_AND); - $qp->set_database($self->{xdb}); + $qp = $X{QueryParser}->new; + $qp->set_default_op(OP_AND()); + $qp->set_database($xdb); $qp->set_stemmer($self->stemmer); - $qp->set_stemming_strategy(STEM_SOME); + $qp->set_stemming_strategy(STEM_SOME()); $qp->set_max_wildcard_expansion(100); - $qp->add_valuerangeprocessor( - Search::Xapian::NumberValueRangeProcessor->new(YYYYMMDD, 'd:')); - $qp->add_valuerangeprocessor( - Search::Xapian::NumberValueRangeProcessor->new(DT, 'dt:')); + my $nvrp = $X{NumberValueRangeProcessor}; + $qp->add_valuerangeprocessor($nvrp->new(YYYYMMDD, 'd:')); + $qp->add_valuerangeprocessor($nvrp->new(DT, 'dt:')); while (my ($name, $prefix) = each %bool_pfx_external) { $qp->add_boolean_prefix($name, $_) foreach split(/ /, $prefix); @@ -297,46 +334,6 @@ EOF $self->{query_parser} = $qp; } -# only used for NNTP server -sub query_xover { - my ($self, $beg, $end, $offset) = @_; - $self->{over_ro}->query_xover($beg, $end, $offset); -} - -sub query_ts { - my ($self, $ts, $prev) = @_; - $self->{over_ro}->query_ts($ts, $prev); -} - -sub lookup_article { - my ($self, $num) = @_; - $self->{over_ro}->get_art($num); -} - -sub next_by_mid { - my ($self, $mid, $id, $prev) = @_; - $self->{over_ro}->next_by_mid($mid, $id, $prev); -} - -# normalize subjects so they are suitable as pathnames for URLs -# XXX: consider for removal -sub subject_path { - my $subj = pop; - $subj = subject_normalized($subj); - $subj =~ s![^a-zA-Z0-9_\.~/\-]+!_!g; - lc($subj); -} - -sub subject_normalized { - my $subj = pop; - $subj =~ s/\A\s+//s; # no leading space - $subj =~ s/\s+\z//s; # no trailing space - $subj =~ s/\s+/ /gs; # no redundant spaces - $subj =~ s/\.+\z//; # no trailing '.' - $subj =~ s/$REPLY_RE//igo; # remove reply prefix - $subj; -} - sub help { my ($self) = @_; $self->qp; # parse altids