1 # This library is free software; you can redistribute it and/or modify
2 # it under the same terms as Perl itself.
4 # This license differs from the rest of public-inbox
6 # Our own jwz-style threading class based on Mail::Thread from CPAN.
7 # Mail::Thread is unmaintained and unavailable on some distros.
8 # We also do not want pruning or subject grouping, since we want
9 # to encourage strict threading and hopefully encourage people
10 # to use proper In-Reply-To/References.
12 # This includes fixes from several open bugs for Mail::Thread
14 # Avoid circular references
15 # - https://rt.cpan.org/Public/Bug/Display.html?id=22817
17 # And avoid recursion in recurse_down:
18 # - https://rt.cpan.org/Ticket/Display.html?id=116727
19 # - http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=833479
20 package PublicInbox::SearchThread;
23 use PublicInbox::MID qw($MID_EXTRACT);
26 my ($msgs, $ordersub, $ctx) = @_;
28 # A. put all current $msgs (non-ghosts) into %id_table
30 # this delete saves around 4K across 1K messages
31 # TODO: move this to a more appropriate place, breaks tests
32 # if we do it during psgi_cull
35 $_->{mid} => PublicInbox::SearchThread::Msg::cast($_);
38 # Sadly, we sort here anyways since the fill-in-the-blanks References:
39 # can be shakier if somebody used In-Reply-To with multiple, disparate
40 # messages. So, take the client Date: into account since we can't
41 # always determine ordering when somebody uses multiple In-Reply-To.
42 # We'll trust the client Date: header here instead of the Received:
43 # time since this is for display (and not retrieval)
44 _set_parent(\%id_table, $_) for sort { $a->{ds} <=> $b->{ds} } @$msgs;
45 my $ibx = $ctx->{ibx};
46 my $rootset = [ grep {
47 !delete($_->{parent}) && $_->visible($ibx)
49 $rootset = $ordersub->($rootset);
50 $_->order_children($ordersub, $ctx) for @$rootset;
54 sub _set_parent ($$) {
55 my ($id_table, $this) = @_;
57 # B. For each element in the message's References field:
58 defined(my $refs = $this->{references}) or return;
60 # This loop exists to help fill in gaps left from missing
61 # messages. It is not needed in a perfect world where
62 # everything is perfectly referenced, only the last ref
65 foreach my $ref ($refs =~ m/$MID_EXTRACT/go) {
66 # Find a Container object for the given Message-ID
67 my $cont = $id_table->{$ref} //=
68 PublicInbox::SearchThread::Msg::ghost($ref);
70 # Link the References field's Containers together in
71 # the order implied by the References header
73 # * If they are already linked don't change the
75 # * Do not add a link if adding that link would
78 !$cont->{parent} && # already linked
79 !$cont->has_descendent($prev) # would loop
81 $prev->add_child($cont);
86 # C. Set the parent of this message to be the last element in
88 if (defined $prev && !$this->has_descendent($prev)) { # would loop
89 $prev->add_child($this);
93 package PublicInbox::SearchThread::Msg;
94 use base qw(PublicInbox::Smsg);
99 # declare a ghost smsg (determined by absence of {blob})
103 children => {}, # becomes an array when sorted by ->order(...)
107 # give a existing smsg the methods of this class
110 $smsg->{children} = {};
111 bless $smsg, __PACKAGE__;
117 while (my $cont = shift @q) {
118 return $cont if $cont->{blob};
119 push @q, values %{$cont->{children}};
125 my ($self, $child) = @_;
126 croak "Cowardly refusing to become my own parent: $self"
129 my $cid = $child->{mid};
132 if (defined(my $parent = $child->{parent})) {
133 delete $parent->{children}->{$cid};
136 $self->{children}->{$cid} = $child;
137 $child->{parent} = $self;
141 my ($self, $child) = @_;
142 my %seen; # loop prevention
144 return 1 if $self == $child || $seen{$child}++;
145 $child = $child->{parent};
150 # Do not show/keep ghosts iff they have no children. Sometimes
151 # a ghost Message-ID is the result of a long header line
152 # being folded/mangled by a MUA, and not a missing message.
154 my ($self, $ibx) = @_;
155 return 1 if $self->{blob};
156 if (my $by_mid = $ibx->smsg_by_mid($self->{mid})) {
157 %$self = (%$self, %$by_mid);
160 (scalar values %{$self->{children}});
165 my ($cur, $ordersub, $ctx) = @_;
167 my %seen = ($cur => 1); # self-referential loop prevention
169 my $ibx = $ctx->{ibx};
170 while (defined($cur = shift @q)) {
171 my $c = $cur->{children}; # The hashref here...
173 $c = [ grep { !$seen{$_}++ && visible($_, $ibx) } values %$c ];
174 $c = $ordersub->($c) if scalar @$c > 1;
175 $cur->{children} = $c; # ...becomes an arrayref