about summary refs log tree commit homepage
path: root/lib
diff options
context:
space:
mode:
Diffstat (limited to 'lib')
-rw-r--r--lib/PublicInbox/HTTPD.pm1
-rw-r--r--lib/PublicInbox/Inbox.pm5
-rw-r--r--lib/PublicInbox/Over.pm19
-rw-r--r--lib/PublicInbox/SearchIdx.pm6
-rw-r--r--lib/PublicInbox/SearchMsg.pm49
-rw-r--r--lib/PublicInbox/SearchThread.pm5
-rw-r--r--lib/PublicInbox/SearchView.pm1
-rw-r--r--lib/PublicInbox/View.pm10
8 files changed, 54 insertions, 42 deletions
diff --git a/lib/PublicInbox/HTTPD.pm b/lib/PublicInbox/HTTPD.pm
index 8cf365d0..38517710 100644
--- a/lib/PublicInbox/HTTPD.pm
+++ b/lib/PublicInbox/HTTPD.pm
@@ -29,7 +29,6 @@ sub new {
                 'psgi.run_once'         => Plack::Util::FALSE,
                 'psgi.multithread' => Plack::Util::FALSE,
                 'psgi.multiprocess' => Plack::Util::TRUE,
-                'psgix.harakiri'=> Plack::Util::FALSE,
                 'psgix.input.buffered' => Plack::Util::TRUE,
 
                 # XXX unstable API!
diff --git a/lib/PublicInbox/Inbox.pm b/lib/PublicInbox/Inbox.pm
index 73f5761a..d57e46d2 100644
--- a/lib/PublicInbox/Inbox.pm
+++ b/lib/PublicInbox/Inbox.pm
@@ -302,8 +302,9 @@ sub smsg_by_mid ($$) {
         my ($self, $mid) = @_;
         my $srch = search($self) or return;
         # favor the Message-ID we used for the NNTP article number:
-        my $num = mid2num($self, $mid);
-        defined $num ? $srch->lookup_article($num) : undef;
+        defined(my $num = mid2num($self, $mid)) or return;
+        my $smsg = $srch->lookup_article($num) or return;
+        PublicInbox::SearchMsg::psgi_cull($smsg);
 }
 
 sub msg_by_mid ($$;$) {
diff --git a/lib/PublicInbox/Over.pm b/lib/PublicInbox/Over.pm
index dda1e6d0..598c9fcb 100644
--- a/lib/PublicInbox/Over.pm
+++ b/lib/PublicInbox/Over.pm
@@ -40,13 +40,16 @@ sub disconnect { $_[0]->{dbh} = undef }
 
 sub connect { $_[0]->{dbh} ||= $_[0]->dbh_new }
 
-sub load_from_row {
-        my ($smsg) = @_;
+sub load_from_row ($;$) {
+        my ($smsg, $cull) = @_;
         bless $smsg, 'PublicInbox::SearchMsg';
         if (defined(my $data = delete $smsg->{ddd})) {
                 $data = uncompress($data);
                 utf8::decode($data);
-                $smsg->load_from_data($data);
+                PublicInbox::SearchMsg::load_from_data($smsg, $data);
+
+                # saves over 600K for 1000+ message threads
+                PublicInbox::SearchMsg::psgi_cull($smsg) if $cull;
         }
         $smsg
 }
@@ -57,7 +60,8 @@ sub do_get {
         my $lim = (($opts->{limit} || 0) + 0) || DEFAULT_LIMIT;
         $sql .= "LIMIT $lim";
         my $msgs = $dbh->selectall_arrayref($sql, { Slice => {} }, @args);
-        load_from_row($_) for @$msgs;
+        my $cull = $opts->{cull};
+        load_from_row($_, $cull) for @$msgs;
         $msgs
 }
 
@@ -82,6 +86,7 @@ sub nothing () { wantarray ? (0, []) : [] };
 sub get_thread {
         my ($self, $mid, $prev) = @_;
         my $dbh = $self->connect;
+        my $opts = { cull => 1 };
 
         my $id = $dbh->selectrow_array(<<'', undef, $mid);
 SELECT id FROM msgid WHERE mid = ? LIMIT 1
@@ -109,7 +114,7 @@ SELECT tid,sid FROM over WHERE num = ? LIMIT 1
 
         my $cols = 'num,ts,ds,ddd';
         unless (wantarray) {
-                return do_get($self, <<"", {}, $tid, $sid, $num);
+                return do_get($self, <<"", $opts, $tid, $sid, $num);
 SELECT $cols FROM over WHERE $cond_all
 ORDER BY $sort_col ASC
 
@@ -123,14 +128,14 @@ SELECT COUNT(num) FROM over WHERE $cond_all
 
         # giant thread, prioritize strict (tid) matches and throw
         # in the loose (sid) matches at the end
-        my $msgs = do_get($self, <<"", {}, $tid, $num);
+        my $msgs = do_get($self, <<"", $opts, $tid, $num);
 SELECT $cols FROM over WHERE tid = ? AND num > ?
 ORDER BY $sort_col ASC
 
         # do we have room for loose matches? get the most recent ones, first:
         my $lim = DEFAULT_LIMIT - scalar(@$msgs);
         if ($lim > 0) {
-                my $opts = { limit => $lim };
+                $opts->{limit} = $lim;
                 my $loose = do_get($self, <<"", $opts, $tid, $sid, $num);
 SELECT $cols FROM over WHERE tid != ? AND sid = ? AND num > ?
 ORDER BY $sort_col DESC
diff --git a/lib/PublicInbox/SearchIdx.pm b/lib/PublicInbox/SearchIdx.pm
index cc1ac56b..8810fe76 100644
--- a/lib/PublicInbox/SearchIdx.pm
+++ b/lib/PublicInbox/SearchIdx.pm
@@ -282,7 +282,7 @@ sub index_body ($$$) {
 sub add_xapian ($$$$$) {
         my ($self, $mime, $num, $oid, $mids, $mid0) = @_;
         my $smsg = PublicInbox::SearchMsg->new($mime);
-        my $doc = $smsg->{doc};
+        my $doc = Search::Xapian::Document->new;
         my $subj = $smsg->subject;
         add_val($doc, PublicInbox::Search::TS(), $smsg->ts);
         my @ds = gmtime($smsg->ds);
@@ -439,8 +439,8 @@ sub remove_by_oid {
         for (; $head != $tail; $head->inc) {
                 my $docid = $head->get_docid;
                 my $doc = $db->get_document($docid);
-                my $smsg = PublicInbox::SearchMsg->wrap($doc, $mid);
-                $smsg->load_expand;
+                my $smsg = PublicInbox::SearchMsg->wrap($mid);
+                $smsg->load_expand($doc);
                 if ($smsg->{blob} eq $oid) {
                         push(@delete, $docid);
                 }
diff --git a/lib/PublicInbox/SearchMsg.pm b/lib/PublicInbox/SearchMsg.pm
index 32cceb25..ceb6edad 100644
--- a/lib/PublicInbox/SearchMsg.pm
+++ b/lib/PublicInbox/SearchMsg.pm
@@ -15,20 +15,18 @@ use Time::Local qw(timegm);
 
 sub new {
         my ($class, $mime) = @_;
-        my $doc = Search::Xapian::Document->new;
-        bless { doc => $doc, mime => $mime }, $class;
+        bless { mime => $mime }, $class;
 }
 
 sub wrap {
-        my ($class, $doc, $mid) = @_;
-        bless { doc => $doc, mime => undef, mid => $mid }, $class;
+        my ($class, $mid) = @_;
+        bless { mid => $mid }, $class;
 }
 
 sub get {
         my ($class, $head, $db, $mid) = @_;
         my $doc_id = $head->get_docid;
-        my $doc = $db->get_document($doc_id);
-        load_expand(wrap($class, $doc, $mid))
+        load_expand(wrap($class, $mid), $db->get_document($doc_id));
 }
 
 sub get_val ($$) {
@@ -61,19 +59,21 @@ sub load_from_data ($$) {
 
                 # To: and Cc: are stored to optimize HDR/XHDR in NNTP since
                 # some NNTP clients will use that for message displays.
+                # NNTP only, and only stored in Over(view), not Xapian
                 $self->{to},
                 $self->{cc},
 
                 $self->{blob},
                 $self->{mid},
+
+                # NNTP only
                 $self->{bytes},
                 $self->{lines}
         ) = split(/\n/, $_[1]);
 }
 
 sub load_expand {
-        my ($self) = @_;
-        my $doc = $self->{doc};
+        my ($self, $doc) = @_;
         my $data = $doc->get_data or return;
         $self->{ts} = get_val($doc, PublicInbox::Search::TS());
         my $dt = get_val($doc, PublicInbox::Search::DT());
@@ -84,10 +84,21 @@ sub load_expand {
         $self;
 }
 
+sub psgi_cull ($) {
+        my ($self) = @_;
+        from_name($self); # fill in {from_name} so we can delete {from}
+
+        # drop NNTP-only fields which aren't relevant to PSGI results:
+        # saves ~80K on a 200 item search result:
+        delete @$self{qw(from ts to cc bytes lines)};
+        $self;
+}
+
+# Only called by PSGI interface, not NNTP
 sub load_doc {
         my ($class, $doc) = @_;
-        my $self = bless { doc => $doc }, $class;
-        $self->load_expand;
+        my $self = bless {}, $class;
+        psgi_cull(load_expand($self, $doc));
 }
 
 # :bytes and :lines metadata in RFC 3977
@@ -159,29 +170,15 @@ sub references {
         defined $x ? $x : '';
 }
 
-sub _get_term_val ($$$) {
-        my ($self, $pfx, $re) = @_;
-        my $doc = $self->{doc};
-        my $end = $doc->termlist_end;
-        my $i = $doc->termlist_begin;
-        $i->skip_to($pfx);
-        if ($i != $end) {
-                my $val = $i->get_termname;
-                $val =~ s/$re// and return $val;
-        }
-        undef;
-}
-
 sub mid ($;$) {
         my ($self, $mid) = @_;
 
         if (defined $mid) {
                 $self->{mid} = $mid;
-        } elsif (my $rv = $self->{mid}) {
+        } elsif (defined(my $rv = $self->{mid})) {
                 $rv;
-        } elsif ($self->{doc}) {
-                $self->{mid} = _get_term_val($self, 'Q', qr/\AQ/);
         } else {
+                die "NO {mime} for mid\n" unless $self->{mime};
                 $self->_extract_mid; # v1 w/o Xapian
         }
 }
diff --git a/lib/PublicInbox/SearchThread.pm b/lib/PublicInbox/SearchThread.pm
index be290980..931bd579 100644
--- a/lib/PublicInbox/SearchThread.pm
+++ b/lib/PublicInbox/SearchThread.pm
@@ -53,6 +53,11 @@ sub _add_message ($$) {
         my $this = _get_cont_for_id($id_table, $smsg->{mid});
         $this->{smsg} = $smsg;
 
+        # saves around 4K across 1K messages
+        # TODO: move this to a more appropriate place, breaks tests
+        # if we do it during psgi_cull
+        delete $smsg->{num};
+
         # B. For each element in the message's References field:
         defined(my $refs = $smsg->{references}) or return;
 
diff --git a/lib/PublicInbox/SearchView.pm b/lib/PublicInbox/SearchView.pm
index e47dcfd0..c33caec9 100644
--- a/lib/PublicInbox/SearchView.pm
+++ b/lib/PublicInbox/SearchView.pm
@@ -240,7 +240,6 @@ sub mset_thread {
         $ctx->{pct} = \%pct;
         $ctx->{prev_attr} = '';
         $ctx->{prev_level} = 0;
-        $ctx->{seen} = {};
         $ctx->{s_nr} = scalar(@$msgs).'+ results';
 
         # reduce hash lookups in skel_dump
diff --git a/lib/PublicInbox/View.pm b/lib/PublicInbox/View.pm
index 5ddb8425..cd125e00 100644
--- a/lib/PublicInbox/View.pm
+++ b/lib/PublicInbox/View.pm
@@ -219,7 +219,10 @@ sub index_entry {
         $rv .= _th_index_lite($mid_raw, \$irt, $id, $ctx);
         my @tocc;
         my $ds = $smsg->ds; # for v1 non-Xapian/SQLite users
-        my $mime = delete $smsg->{mime}; # critical to memory use
+        # deleting {mime} is critical to memory use,
+        # the rest of the fields saves about 400K as we iterate across 1K msgs
+        my ($mime) = delete @$smsg{qw(mime ds ts blob subject)};
+
         my $hdr = $mime->header_obj;
         my $from = _hdr_names_html($hdr, 'From');
         obfuscate_addrs($obfs_ibx, $from) if $obfs_ibx;
@@ -311,7 +314,10 @@ sub _th_index_lite {
         my $nr_s = 0;
         my $siblings;
         if (my $smsg = $node->{smsg}) {
-                ($$irt) = (($smsg->{references} || '') =~ m/<([^>]+)>\z/);
+                # delete saves about 200KB on a 1K message thread
+                if (my $refs = delete $smsg->{references}) {
+                        ($$irt) = ($refs =~ m/<([^>]+)>\z/);
+                }
         }
         my $irt_map = $mapping->{$$irt} if defined $$irt;
         if (defined $irt_map) {