1 files changed, 53 insertions, 3 deletions

diff --git a/lib/PublicInbox/MID.pm b/lib/PublicInbox/MID.pm
index 2c9822f4..c82e8401 100644
--- a/lib/PublicInbox/MID.pm
+++ b/lib/PublicInbox/MID.pm
@@ -6,10 +6,14 @@ package PublicInbox::MID;
 use strict;
 use warnings;
 use base qw/Exporter/;
-our @EXPORT_OK = qw/mid_clean id_compress mid2path mid_mime mid_escape MID_ESC/;
+our @EXPORT_OK = qw/mid_clean id_compress mid2path mid_mime mid_escape MID_ESC
+        mids references/;
 use URI::Escape qw(uri_escape_utf8);
 use Digest::SHA qw/sha1_hex/;
-use constant MID_MAX => 40; # SHA-1 hex length
+use constant {
+        MID_MAX => 40, # SHA-1 hex length # TODO: get rid of this
+        MAX_MID_SIZE => 244, # max term size (Xapian limitation) - length('Q')
+};
 
 sub mid_clean {
         my ($mid) = @_;
@@ -46,7 +50,53 @@ sub mid2path {
         "$x2/$x38";
 }
 
-sub mid_mime ($) { $_[0]->header_obj->header_raw('Message-ID') }
+# Only for v1 code paths:
+sub mid_mime ($) { mids($_[0]->header_obj)->[0] }
+
+sub mids ($) {
+        my ($hdr) = @_;
+        my @mids;
+        my @v = $hdr->header_raw('Message-Id');
+        foreach my $v (@v) {
+                my @cur = ($v =~ /<([^>]+)>/sg);
+                if (@cur) {
+                        push(@mids, @cur);
+                } else {
+                        push(@mids, $v);
+                }
+        }
+        uniq_mids(\@mids);
+}
+
+# last References should be IRT, but some mail clients do things
+# out of order, so trust IRT over References iff IRT exists
+sub references ($) {
+        my ($hdr) = @_;
+        my @mids;
+        foreach my $f (qw(References In-Reply-To)) {
+                my @v = $hdr->header_raw($f);
+                foreach my $v (@v) {
+                        push(@mids, ($v =~ /<([^>]+)>/sg));
+                }
+        }
+        uniq_mids(\@mids);
+}
+
+sub uniq_mids ($) {
+        my ($mids) = @_;
+        my @ret;
+        my %seen;
+        foreach my $mid (@$mids) {
+                if (length($mid) > MAX_MID_SIZE) {
+                        warn "Message-ID: <$mid> too long, truncating\n";
+                        $mid = substr($mid, 0, MAX_MID_SIZE);
+                }
+                next if $seen{$mid};
+                push @ret, $mid;
+                $seen{$mid} = 1;
+        }
+        \@ret;
+}
 
 # RFC3986, section 3.3:
 sub MID_ESC () { '^A-Za-z0-9\-\._~!\$\&\';\(\)\*\+,;=:@' }