# slow storage.
#
# data notes:
-# * NNTP article numbers are UIDs and message sequence numbers (MSNs)
-# * Message sequence numbers (MSNs) can be stable since we're read-only.
-# Most IMAP clients use UIDs (I hope). We may return a dummy message
-# in the future if a client requests a non-existent MSN, but that seems
-# unecessary with mutt.
+#
+# * NNTP article numbers are UIDs, mm->created_at is UIDVALIDITY
+#
+# * public-inboxes are sliced into mailboxes of 50K messages
+# to not overload MUAs: $NEWSGROUP_NAME.$SLICE_INDEX
+# Slices are similar in concept to v2 "epochs". Epochs
+# are for the limitations of git clients, while slices are
+# for the limitations of IMAP clients.
+#
+# * sequence numbers are estimated based on slice
package PublicInbox::IMAP;
use strict;
use base qw(PublicInbox::DS);
-use fields qw(imapd logged_in ibx long_cb -login_tag
- uid_min -idle_tag -idle_max);
+use fields qw(imapd ibx long_cb -login_tag
+ uid_base -idle_tag -idle_max);
use PublicInbox::Eml;
use PublicInbox::EmlContentFoo qw(parse_content_disposition);
use PublicInbox::DS qw(now);
use Errno qw(EAGAIN);
use Time::Local qw(timegm);
use POSIX qw(strftime);
+use Hash::Util qw(unlock_hash); # dependency of fields for perl 5.10+, anyways
+use PublicInbox::Search;
+*mdocid = \&PublicInbox::Search::mdocid;
my $Address;
for my $mod (qw(Email::Address::XS Mail::Address)) {
}
die "neither Email::Address::XS nor Mail::Address loaded: $@" if !$Address;
-sub LINE_MAX () { 512 } # does RFC 3501 have a limit like RFC 977?
+sub LINE_MAX () { 8000 } # RFC 2683 3.2.1.5
# changing this will cause grief for clients which cache
-sub UID_BLOCK () { 50_000 }
-
-my %FETCH_NEED_BLOB = ( # for future optimization
- 'BODY[HEADER]' => 1,
- 'BODY[TEXT]' => 1,
- 'BODY[]' => 1,
- 'RFC822.HEADER' => 1,
- 'RFC822.SIZE' => 1, # needs CRLF conversion :<
- 'RFC822.TEXT' => 1,
- BODY => 1,
- BODYSTRUCTURE => 1,
- ENVELOPE => 1,
- FLAGS => 0,
- INTERNALDATE => 0,
- RFC822 => 1,
- UID => 0,
+sub UID_SLICE () { 50_000 }
+
+# these values area also used for sorting
+sub NEED_SMSG () { 1 }
+sub NEED_BLOB () { NEED_SMSG|2 }
+sub CRLF_BREF () { 4 }
+sub EML_HDR () { 8 }
+sub CRLF_HDR () { 16 }
+sub EML_BDY () { 32 }
+sub CRLF_BDY () { 64 }
+my $OP_EML_NEW = [ EML_HDR - 1, \&op_eml_new ];
+my $OP_CRLF_BREF = [ CRLF_BREF, \&op_crlf_bref ];
+my $OP_CRLF_HDR = [ CRLF_HDR, \&op_crlf_hdr ];
+my $OP_CRLF_BDY = [ CRLF_BDY, \&op_crlf_bdy ];
+
+my %FETCH_NEED = (
+ 'BODY[HEADER]' => [ NEED_BLOB|EML_HDR|CRLF_HDR, \&emit_rfc822_header ],
+ 'BODY[TEXT]' => [ NEED_BLOB|EML_BDY|CRLF_BDY, \&emit_rfc822_text ],
+ 'BODY[]' => [ NEED_BLOB|CRLF_BREF, \&emit_rfc822 ],
+ 'RFC822.HEADER' => [ NEED_BLOB|EML_HDR|CRLF_HDR, \&emit_rfc822_header ],
+ 'RFC822.TEXT' => [ NEED_BLOB|EML_BDY|CRLF_BDY, \&emit_rfc822_text ],
+ 'RFC822.SIZE' => [ NEED_SMSG, \&emit_rfc822_size ],
+ RFC822 => [ NEED_BLOB|CRLF_BREF, \&emit_rfc822 ],
+ BODY => [ NEED_BLOB|EML_HDR|EML_BDY, \&emit_body ],
+ BODYSTRUCTURE => [ NEED_BLOB|EML_HDR|EML_BDY, \&emit_bodystructure ],
+ ENVELOPE => [ NEED_BLOB|EML_HDR, \&emit_envelope ],
+ FLAGS => [ 0, \&emit_flags ],
+ INTERNALDATE => [ NEED_SMSG, \&emit_internaldate ],
);
-my %FETCH_ATT = map { $_ => [ $_ ] } keys %FETCH_NEED_BLOB;
+my %FETCH_ATT = map { $_ => [ $_ ] } keys %FETCH_NEED;
# aliases (RFC 3501 section 6.4.5)
$FETCH_ATT{FAST} = [ qw(FLAGS INTERNALDATE RFC822.SIZE) ];
$FETCH_ATT{FULL} = [ @{$FETCH_ATT{ALL}}, 'BODY' ];
for my $att (keys %FETCH_ATT) {
- my %h = map { $_ => 1 } @{$FETCH_ATT{$att}};
+ my %h = map { $_ => $FETCH_NEED{$_} } @{$FETCH_ATT{$att}};
$FETCH_ATT{$att} = \%h;
}
+undef %FETCH_NEED;
my $valid_range = '[0-9]+|[0-9]+:[0-9]+|[0-9]+:\*';
$valid_range = qr/\A(?:$valid_range)(?:,(?:$valid_range))*\z/;
my %MoY;
@MoY{@MoY} = (0..11);
+# RFC 3501 5.4. Autologout Timer needs to be >= 30min
+$PublicInbox::DS::EXPTIME = 60 * 30;
+
sub greet ($) {
my ($self) = @_;
my $capa = capa($self);
sub new ($$$) {
my ($class, $sock, $imapd) = @_;
- my $self = fields::new($class);
+ my $self = fields::new('PublicInbox::IMAP_preauth');
+ unlock_hash(%$self);
my $ev = EPOLLIN;
my $wbuf;
if ($sock->can('accept_SSL') && !$sock->accept_SSL) {
$self;
}
+sub logged_in { 1 }
+
sub capa ($) {
my ($self) = @_;
# dovecot advertises IDLE pre-login; perhaps because some clients
# depend on it, so we'll do the same
my $capa = 'CAPABILITY IMAP4rev1 IDLE';
- if ($self->{logged_in}) {
+ if ($self->logged_in) {
$capa .= ' COMPRESS=DEFLATE';
} else {
if (!($self->{sock} // $self)->can('accept_SSL') &&
sub login_success ($$) {
my ($self, $tag) = @_;
- $self->{logged_in} = 1;
+ bless $self, 'PublicInbox::IMAP';
my $capa = capa($self);
"$tag OK [$capa] Logged in\r\n";
}
sub cmd_close ($$) {
my ($self, $tag) = @_;
- delete $self->{uid_min};
+ delete $self->{uid_base};
delete $self->{ibx} ? "$tag OK Close done\r\n"
: "$tag BAD No mailbox\r\n";
}
sub cmd_logout ($$) {
my ($self, $tag) = @_;
- delete $self->{logged_in};
+ delete $self->{-idle_tag};
$self->write(\"* BYE logging out\r\n$tag OK Logout done\r\n");
$self->shutdn; # PublicInbox::DS::shutdn
undef;
# called by PublicInbox::InboxIdle
sub on_inbox_unlock {
my ($self, $ibx) = @_;
- my $new = $ibx->mm->max;
+ my $new = $ibx->over->max;
+ my $uid_base = $self->{uid_base} // 0;
+ my $uid_end = $uid_base + UID_SLICE;
defined(my $old = $self->{-idle_max}) or die 'BUG: -idle_max unset';
+ $new = $uid_end if $new > $uid_end;
if ($new > $old) {
$self->{-idle_max} = $new;
+ $new -= $uid_base;
+ $old -= $uid_base;
$self->msg_more("* $_ EXISTS\r\n") for (($old + 1)..($new - 1));
$self->write(\"* $new EXISTS\r\n");
+ } elsif ($new == $uid_end) { # max exceeded $uid_end
+ # continue idling w/o inotify
+ delete $self->{-idle_max};
+ my $sock = $self->{sock} or return;
+ $ibx->unsubscribe_unlock(fileno($sock));
+ }
+}
+
+# called every X minute(s) or so by PublicInbox::DS::later
+my $IDLERS = {};
+my $idle_timer;
+sub idle_tick_all {
+ my $old = $IDLERS;
+ $IDLERS = {};
+ for my $i (values %$old) {
+ next if ($i->{wbuf} || !exists($i->{-idle_tag}));
+ $i->update_idle_time or next;
+ $IDLERS->{fileno($i->{sock})} = $i;
+ $i->write(\"* OK Still here\r\n");
}
+ $idle_timer = scalar keys %$IDLERS ?
+ PublicInbox::DS::later(\&idle_tick_all) : undef;
}
sub cmd_idle ($$) {
my ($self, $tag) = @_;
# IDLE seems allowed by dovecot w/o a mailbox selected *shrug*
my $ibx = $self->{ibx} or return "$tag BAD no mailbox selected\r\n";
- $ibx->subscribe_unlock(fileno($self->{sock}), $self);
- $self->{imapd}->idler_start;
$self->{-idle_tag} = $tag;
- $self->{-idle_max} = $ibx->mm->max // 0;
- "+ idling\r\n"
+ my $max = $ibx->over->max;
+ my $uid_end = $self->{uid_base} + UID_SLICE;
+ my $sock = $self->{sock} or return;
+ my $fd = fileno($sock);
+ # only do inotify on most recent slice
+ if ($max < $uid_end) {
+ $ibx->subscribe_unlock($fd, $self);
+ $self->{imapd}->idler_start;
+ $self->{-idle_max} = $max;
+ }
+ $idle_timer //= PublicInbox::DS::later(\&idle_tick_all);
+ $IDLERS->{$fd} = $self;
+ \"+ idling\r\n"
+}
+
+sub stop_idle ($$) {
+ my ($self, $ibx);
+ my $sock = $self->{sock} or return;
+ my $fd = fileno($sock);
+ delete $IDLERS->{$fd};
+ $ibx->unsubscribe_unlock($fd);
}
sub cmd_done ($$) {
warn "BUG: idle_tag set w/o inbox";
return "$tag BAD internal bug\r\n";
};
- $ibx->unsubscribe_unlock(fileno($self->{sock}));
+ stop_idle($self, $ibx);
"$idle_tag OK Idle done\r\n";
}
-sub ensure_ranges_exist ($$$) {
+sub ensure_slices_exist ($$$) {
my ($imapd, $ibx, $max) = @_;
+ defined(my $mb_top = $ibx->{newsgroup}) or return;
my $mailboxes = $imapd->{mailboxes};
- my $mb_top = $ibx->{newsgroup};
my @created;
- for (my $i = int($max/UID_BLOCK); $i >= 0; --$i) {
+ for (my $i = int($max/UID_SLICE); $i >= 0; --$i) {
my $sub_mailbox = "$mb_top.$i";
last if exists $mailboxes->{$sub_mailbox};
$mailboxes->{$sub_mailbox} = $ibx;
+ $sub_mailbox =~ s/\Ainbox\./INBOX./i; # more familiar to users
push @created, $sub_mailbox;
}
return unless @created;
sub inbox_lookup ($$) {
my ($self, $mailbox) = @_;
- my ($ibx, $exists, $uidnext);
+ my ($ibx, $exists, $uidnext, $uid_base);
if ($mailbox =~ /\A(.+)\.([0-9]+)\z/) {
- # old mail: inbox.comp.foo.$uid_block_idx
- my ($mb_top, $uid_min) = ($1, $2 * UID_BLOCK + 1);
-
+ # old mail: inbox.comp.foo.$SLICE_IDX
+ my $mb_top = $1;
+ $uid_base = $2 * UID_SLICE;
$ibx = $self->{imapd}->{mailboxes}->{lc $mailbox} or return;
- $exists = $ibx->mm->max // 0;
- $self->{uid_min} = $uid_min;
- ensure_ranges_exist($self->{imapd}, $ibx, $exists);
- my $uid_end = $uid_min + UID_BLOCK - 1;
- $exists = $uid_end if $exists > $uid_end;
- $uidnext = $exists + 1;
+ my $max;
+ ($exists, $uidnext, $max) = $ibx->over->imap_status($uid_base,
+ $uid_base + UID_SLICE);
+ ensure_slices_exist($self->{imapd}, $ibx, $max);
} else { # check for dummy inboxes
- $ibx = $self->{imapd}->{mailboxes}->{lc $mailbox} or return;
- delete $self->{uid_min};
- $exists = 0;
+ $mailbox = lc $mailbox;
+ $ibx = $self->{imapd}->{mailboxes}->{$mailbox} or return;
+
+ # if "INBOX.foo.bar" is selected and "INBOX.foo.bar.0",
+ # check for new UID ranges (e.g. "INBOX.foo.bar.1")
+ if (my $z = $self->{imapd}->{mailboxes}->{"$mailbox.0"}) {
+ ensure_slices_exist($self->{imapd}, $z, $z->over->max);
+ }
+
+ $uid_base = $exists = 0;
$uidnext = 1;
}
- ($ibx, $exists, $uidnext);
+ ($ibx, $exists, $uidnext, $uid_base);
}
sub cmd_examine ($$$) {
my ($self, $tag, $mailbox) = @_;
- my ($ibx, $exists, $uidnext) = inbox_lookup($self, $mailbox);
+ my ($ibx, $exists, $uidnext, $base) = inbox_lookup($self, $mailbox);
return "$tag NO Mailbox doesn't exist: $mailbox\r\n" if !$ibx;
+ $self->{uid_base} = $base;
# XXX: do we need this? RFC 5162/7162
my $ret = $self->{ibx} ? "* OK [CLOSED] previous closed\r\n" : '';
$self->requeue if $new_size == 1;
}
-sub uid_fetch_cb { # called by git->cat_async via git_async_cat
- my ($bref, $oid, $type, $size, $fetch_m_arg) = @_;
- my ($self, undef, $msgs, undef, $want) = @$fetch_m_arg;
+# my ($msn, $UID) = @_;
+sub fetch_msn_uid ($$) { '* '.(${$_[0]}++).' FETCH (UID '.$_[1] }
+
+sub fetch_run_ops {
+ my ($self, $msn, $smsg, $bref, $ops, $partial) = @_;
+ $self->msg_more(fetch_msn_uid($msn, $smsg->{num}));
+ my ($eml, $k);
+ for (my $i = 0; $i < @$ops;) {
+ $k = $ops->[$i++];
+ $ops->[$i++]->($self, $k, $smsg, $bref, $eml);
+ }
+ partial_emit($self, $partial, $eml) if $partial;
+ $self->msg_more(")\r\n");
+}
+
+sub fetch_blob_cb { # called by git->cat_async via git_async_cat
+ my ($bref, $oid, $type, $size, $fetch_arg) = @_;
+ my ($self, undef, $msgs, $range_info, $ops, $partial) = @$fetch_arg;
my $smsg = shift @$msgs or die 'BUG: no smsg';
if (!defined($oid)) {
# it's possible to have TOCTOU if an admin runs
} else {
$smsg->{blob} eq $oid or die "BUG: $smsg->{blob} != $oid";
}
+ fetch_run_ops($self, $range_info->[3], $smsg, $bref, $ops, $partial);
+ requeue_once($self);
+}
- $$bref =~ s/(?<!\r)\n/\r\n/sg; # make strict clients happy
+sub emit_rfc822 {
+ my ($self, $k, undef, $bref) = @_;
+ $self->msg_more(" $k {" . length($$bref)."}\r\n");
+ $self->msg_more($$bref);
+}
- # fixup old bug from import (pre-a0c07cba0e5d8b6a)
- $$bref =~ s/\A[\r\n]*From [^\r\n]*\r\n//s;
+# Mail::IMAPClient::message_string cares about this by default,
+# (->Ignoresizeerrors attribute). Admins are encouraged to
+# --reindex for IMAP support, anyways.
+sub emit_rfc822_size {
+ my ($self, $k, $smsg) = @_;
+ $self->msg_more(' RFC822.SIZE ' . $smsg->{bytes});
+}
- $self->msg_more("* $smsg->{num} FETCH (UID $smsg->{num}");
+sub emit_internaldate {
+ my ($self, undef, $smsg) = @_;
+ $self->msg_more(' INTERNALDATE "'.$smsg->internaldate.'"');
+}
- $want->{'RFC822.SIZE'} and
- $self->msg_more(' RFC822.SIZE '.length($$bref));
- $want->{INTERNALDATE} and
- $self->msg_more(' INTERNALDATE "'.$smsg->internaldate.'"');
- $want->{FLAGS} and $self->msg_more(' FLAGS ()');
- for ('RFC822', 'BODY[]') {
- $want->{$_} or next;
- $self->msg_more(" $_ {".length($$bref)."}\r\n");
- $self->msg_more($$bref);
- }
+sub emit_flags { $_[0]->msg_more(' FLAGS ()') }
- my $eml = PublicInbox::Eml->new($bref);
+sub emit_envelope {
+ my ($self, undef, undef, undef, $eml) = @_;
+ $self->msg_more(' ENVELOPE '.eml_envelope($eml));
+}
- $want->{ENVELOPE} and
- $self->msg_more(' ENVELOPE '.eml_envelope($eml));
+sub emit_rfc822_header {
+ my ($self, $k, undef, undef, $eml) = @_;
+ $self->msg_more(" $k {".length(${$eml->{hdr}})."}\r\n");
+ $self->msg_more(${$eml->{hdr}});
+}
- for ('RFC822.HEADER', 'BODY[HEADER]') {
- $want->{$_} or next;
- $self->msg_more(" $_ {".length(${$eml->{hdr}})."}\r\n");
- $self->msg_more(${$eml->{hdr}});
- }
- for ('RFC822.TEXT', 'BODY[TEXT]') {
- $want->{$_} or next;
- $self->msg_more(" $_ {".length($$bref)."}\r\n");
- $self->msg_more($$bref);
- }
- $want->{BODYSTRUCTURE} and
- $self->msg_more(' BODYSTRUCTURE '.fetch_body($eml, 1));
- $want->{BODY} and
- $self->msg_more(' BODY '.fetch_body($eml));
- if (my $partial = $want->{-partial}) {
- partial_emit($self, $partial, $eml);
- }
- $self->msg_more(")\r\n");
- requeue_once($self);
+# n.b. this is sorted to be after any emit_eml_new ops
+sub emit_rfc822_text {
+ my ($self, $k, undef, $bref) = @_;
+ $self->msg_more(" $k {".length($$bref)."}\r\n");
+ $self->msg_more($$bref);
+}
+
+sub emit_bodystructure {
+ my ($self, undef, undef, undef, $eml) = @_;
+ $self->msg_more(' BODYSTRUCTURE '.fetch_body($eml, 1));
+}
+
+sub emit_body {
+ my ($self, undef, undef, undef, $eml) = @_;
+ $self->msg_more(' BODY '.fetch_body($eml));
+}
+
+# set $eml once ($_[4] == $eml, $_[3] == $bref)
+sub op_eml_new { $_[4] = PublicInbox::Eml->new($_[3]) }
+
+# s/From / fixes old bug from import (pre-a0c07cba0e5d8b6a)
+sub to_crlf_full {
+ ${$_[0]} =~ s/(?<!\r)\n/\r\n/sg;
+ ${$_[0]} =~ s/\A[\r\n]*From [^\r\n]*\r\n//s;
}
+sub op_crlf_bref { to_crlf_full($_[3]) }
+
+sub op_crlf_hdr { to_crlf_full($_[4]->{hdr}) }
+
+sub op_crlf_bdy { ${$_[4]->{bdy}} =~ s/(?<!\r)\n/\r\n/sg if $_[4]->{bdy} }
+
sub uid_clamp ($$$) {
my ($self, $beg, $end) = @_;
- my $uid_min = $self->{uid_min} or return;
- my $uid_end = $uid_min + UID_BLOCK - 1;
+ my $uid_min = $self->{uid_base} + 1;
+ my $uid_end = $uid_min + UID_SLICE - 1;
$$beg = $uid_min if $$beg < $uid_min;
$$end = $uid_end if $$end > $uid_end;
}
$range = $$range_csv;
$$range_csv = undef;
}
+ my $uid_base = $self->{uid_base};
+ my $uid_end = $uid_base + UID_SLICE;
if ($range =~ /\A([0-9]+):([0-9]+)\z/) {
($beg, $end) = ($1 + 0, $2 + 0);
+ uid_clamp($self, \$beg, \$end);
} elsif ($range =~ /\A([0-9]+):\*\z/) {
$beg = $1 + 0;
- $end = $self->{ibx}->mm->max // 0;
- my $uid_end = ($self->{uid_min} // 1) - 1 + UID_BLOCK;
+ $end = $self->{ibx}->over->max;
$end = $uid_end if $end > $uid_end;
$beg = $end if $beg > $end;
+ uid_clamp($self, \$beg, \$end);
} elsif ($range =~ /\A[0-9]+\z/) {
$beg = $end = $range + 0;
- undef $range;
+ # just let the caller do an out-of-range query if a single
+ # UID is out-of-range
+ ++$beg if ($beg <= $uid_base || $end > $uid_end);
} else {
return 'BAD fetch range';
}
- uid_clamp($self, \$beg, \$end) if defined($range);
- [ $beg, $end, $$range_csv ];
+ my $msn = $beg - $uid_base;
+ [ $beg, $end, $$range_csv, \$msn ];
}
sub refill_range ($$$) {
undef; # keep looping
}
-sub uid_fetch_m { # long_response
- my ($self, $tag, $msgs, $range_info, $want) = @_;
+sub fetch_blob { # long_response
+ my ($self, $tag, $msgs, $range_info, $ops, $partial) = @_;
while (!@$msgs) { # rare
if (my $end = refill_range($self, $msgs, $range_info)) {
$self->write(\"$tag $end\r\n");
}
}
git_async_cat($self->{ibx}->git, $msgs->[0]->{blob},
- \&uid_fetch_cb, \@_);
+ \&fetch_blob_cb, \@_);
+}
+
+sub fetch_smsg { # long_response
+ my ($self, $tag, $msgs, $range_info, $ops) = @_;
+ while (!@$msgs) { # rare
+ if (my $end = refill_range($self, $msgs, $range_info)) {
+ $self->write(\"$tag $end\r\n");
+ return;
+ }
+ }
+ fetch_run_ops($self, $range_info->[3], $_, undef, $ops) for @$msgs;
+ @$msgs = ();
+ 1; # more
+}
+
+sub refill_uids ($$$;$) {
+ my ($self, $uids, $range_info, $sql) = @_;
+ my ($beg, $end, $range_csv) = @$range_info;
+ my $over = $self->{ibx}->over;
+ while (1) {
+ if (scalar(@$uids = @{$over->uid_range($beg, $end, $sql)})) {
+ $range_info->[0] = $uids->[-1] + 1; # update $beg
+ return;
+ } elsif (!$range_csv) {
+ return 0;
+ } else {
+ my $next_range = range_step($self, \$range_csv);
+ return $next_range if !ref($next_range); # error
+ ($beg, $end, $range_csv) = @$range_info = @$next_range;
+ # continue looping
+ }
+ }
+}
+
+sub fetch_uid { # long_response
+ my ($self, $tag, $uids, $range_info, $ops) = @_;
+ if (defined(my $err = refill_uids($self, $uids, $range_info))) {
+ $err ||= 'OK Fetch done';
+ $self->write("$tag $err\r\n");
+ return;
+ }
+ my ($i, $k);
+ my $msn = $range_info->[3];
+ for (@$uids) {
+ $self->msg_more(fetch_msn_uid($msn, $_));
+ for ($i = 0; $i < @$ops;) {
+ $k = $ops->[$i++];
+ $ops->[$i++]->($self, $k);
+ }
+ $self->msg_more(")\r\n");
+ }
+ @$uids = ();
+ 1; # more
}
sub cmd_status ($$$;@) {
# request for hierarchy delimiter
$l = [ qq[* LIST (\\Noselect) "." ""\r\n] ];
} elsif ($refname ne '' || $wildcard ne '*') {
- $wildcard = lc $wildcard;
- $wildcard =~ s!([^a-z0-9_])!$patmap{$1} // "\Q$1"!eg;
- $l = [ grep(/ \Q$refname\E$wildcard\r\n\z/s, @$l) ];
+ $wildcard =~ s!([^a-z0-9_])!$patmap{$1} // "\Q$1"!egi;
+ $l = [ grep(/ \Q$refname\E$wildcard\r\n\z/is, @$l) ];
}
\(join('', @$l, "$tag OK List done\r\n"));
}
}
my $str = $eml->header_obj->as_string;
$str =~ s/$hdrs_re//g;
+ $str =~ s/(?<!\r)\n/\r\n/sg;
$str .= "\r\n";
}
$eml = eml_body_idx($eml, $section_idx) or return;
}
my $str = $eml->header_obj->as_string;
- join('', ($str =~ m/($hdrs_re)/g), "\r\n");
+ $str = join('', ($str =~ m/($hdrs_re)/g));
+ $str =~ s/(?<!\r)\n/\r\n/sg;
+ $str .= "\r\n";
}
-sub partial_prepare ($$$) {
- my ($partial, $want, $att) = @_;
+sub partial_prepare ($$$$) {
+ my ($need, $partial, $want, $att) = @_;
# recombine [ "BODY[1.HEADER.FIELDS", "(foo", "bar)]" ]
# back to: "BODY[1.HEADER.FIELDS (foo bar)]"
(?:\.(HEADER|MIME|TEXT))? # 2 - section_name
\](?:<([0-9]+)(?:\.([0-9]+))?>)?\z/sx) { # 3, 4
$partial->{$att} = [ \&partial_body, $1, $2, $3, $4 ];
+ $$need |= CRLF_BREF|EML_HDR|EML_BDY;
} elsif ($att =~ /\ABODY\[(?:([0-9]+(?:\.[0-9]+)*)\.)? # 1 - section_idx
(?:HEADER\.FIELDS(\.NOT)?)\x20 # 2
\(([A-Z0-9\-\x20]+)\) # 3 - hdrs
: \&partial_hdr_get,
$1, undef, $4, $5 ];
$tmp->[2] = hdrs_regexp($3);
+
+ # don't emit CRLF_HDR instruction, here, partial_hdr_*
+ # will do CRLF conversion with only the extracted result
+ # and not waste time converting lines we don't care about.
+ $$need |= EML_HDR;
} else {
undef;
}
}
}
-sub fetch_common ($$$$) {
- my ($self, $tag, $range_csv, $want) = @_;
- my $ibx = $self->{ibx} or return "$tag BAD No mailbox selected\r\n";
+sub fetch_compile ($) {
+ my ($want) = @_;
if ($want->[0] =~ s/\A\(//s) {
- $want->[-1] =~ s/\)\z//s or return "$tag BAD no rparen\r\n";
+ $want->[-1] =~ s/\)\z//s or return 'BAD no rparen';
}
- my (%partial, %want);
+ my (%partial, %seen, @op);
+ my $need = 0;
while (defined(my $att = shift @$want)) {
$att = uc($att);
+ next if $att eq 'UID'; # always returned
$att =~ s/\ABODY\.PEEK\[/BODY\[/; # we're read-only
my $x = $FETCH_ATT{$att};
if ($x) {
- %want = (%want, %$x);
- } elsif (!partial_prepare(\%partial, $want, $att)) {
- return "$tag BAD param: $att\r\n";
+ while (my ($k, $fl_cb) = each %$x) {
+ next if $seen{$k}++;
+ $need |= $fl_cb->[0];
+ push @op, [ @$fl_cb, $k ];
+ }
+ } elsif (!partial_prepare(\$need, \%partial, $want, $att)) {
+ return "BAD param: $att";
}
}
+ my @r;
# stabilize partial order for consistency and ease-of-debugging:
if (scalar keys %partial) {
- $want{-partial} = [ map {;
- [ $_, @{$partial{$_}} ]
- } sort keys %partial ];
+ $need |= NEED_BLOB;
+ $r[2] = [ map { [ $_, @{$partial{$_}} ] } sort keys %partial ];
+ }
+
+ push @op, $OP_EML_NEW if ($need & (EML_HDR|EML_BDY));
+
+ # do we need CRLF conversion?
+ if ($need & CRLF_BREF) {
+ push @op, $OP_CRLF_BREF;
+ } elsif (my $crlf = ($need & (CRLF_HDR|CRLF_BDY))) {
+ if ($crlf == (CRLF_HDR|CRLF_BDY)) {
+ push @op, $OP_CRLF_BREF;
+ } elsif ($need & CRLF_HDR) {
+ push @op, $OP_CRLF_HDR;
+ } else {
+ push @op, $OP_CRLF_BDY;
+ }
}
+
+ $r[0] = $need & NEED_BLOB ? \&fetch_blob :
+ ($need & NEED_SMSG ? \&fetch_smsg : \&fetch_uid);
+
+ # r[1] = [ $key1, $cb1, $key2, $cb2, ... ]
+ use sort 'stable'; # makes output more consistent
+ $r[1] = [ map { ($_->[2], $_->[1]) } sort { $a->[0] <=> $b->[0] } @op ];
+ @r;
+}
+
+sub cmd_uid_fetch ($$$$;@) {
+ my ($self, $tag, $range_csv, @want) = @_;
+ my $ibx = $self->{ibx} or return "$tag BAD No mailbox selected\r\n";
+ my ($cb, $ops, $partial) = fetch_compile(\@want);
+ return "$tag $cb\r\n" unless $ops;
+
$range_csv = 'bad' if $range_csv !~ $valid_range;
my $range_info = range_step($self, \$range_csv);
return "$tag $range_info\r\n" if !ref($range_info);
- [ $tag, [], $range_info, \%want ];
+ long_response($self, $cb, $tag, [], $range_info, $ops, $partial);
+}
+
+sub msn_to_uid_range ($$) {
+ my $uid_base = $_[0]->{uid_base};
+ $_[1] =~ s/([0-9]+)/$uid_base + $1/sge;
}
-sub cmd_uid_fetch ($$$;@) {
+sub cmd_fetch ($$$$;@) {
my ($self, $tag, $range_csv, @want) = @_;
- my $args = fetch_common($self, $tag, $range_csv, \@want);
- ref($args) eq 'ARRAY' ?
- long_response($self, \&uid_fetch_m, @$args) :
- $args; # error
+ my $ibx = $self->{ibx} or return "$tag BAD No mailbox selected\r\n";
+ my ($cb, $ops, $partial) = fetch_compile(\@want);
+ return "$tag $cb\r\n" unless $ops;
+
+ # cb is one of fetch_blob, fetch_smsg, fetch_uid
+ $range_csv = 'bad' if $range_csv !~ $valid_range;
+ msn_to_uid_range($self, $range_csv);
+ my $range_info = range_step($self, \$range_csv);
+ return "$tag $range_info\r\n" if !ref($range_info);
+ long_response($self, $cb, $tag, [], $range_info, $ops, $partial);
}
sub parse_date ($) { # 02-Oct-1993
timegm(0, 0, 0, $dd, $mm, $yyyy);
}
-sub uid_search_uid_range { # long_response
- my ($self, $tag, $beg, $end, $sql) = @_;
- my $uids = $self->{ibx}->over->uid_range($$beg, $end, $sql);
- if (@$uids) {
- $$beg = $uids->[-1] + 1;
- $self->msg_more(join(' ', '', @$uids));
- } else {
- $self->write(\"\r\n$tag OK Search done\r\n");
- undef;
+sub search_uid_range { # long_response
+ my ($self, $tag, $sql, $range_info) = @_;
+ my $uids = [];
+ if (defined(my $err = refill_uids($self, $uids, $range_info, $sql))) {
+ $err ||= 'OK Search done';
+ $self->write("\r\n$tag $err\r\n");
+ return;
}
+ $self->msg_more(join(' ', '', @$uids));
+ 1; # more
}
sub date_search {
# KEYWORD # TODO ? dfpre,dfpost,...
);
+# IMAP allows searching arbitrary headers via "HEADER $HDR_NAME $HDR_VAL"
+# which gets silly expensive. We only allow the headers we already index.
+my %H2X = (%I2X, 'MESSAGE-ID' => 'm:', 'LIST-ID' => 'l:');
+
+sub xap_append ($$$$) {
+ my ($q, $rest, $k, $xk) = @_;
+ delete $q->{sql}; # can't use over.sqlite3
+ defined(my $arg = shift @$rest) or return "BAD $k no arg";
+
+ # AFAIK Xapian can't handle [*"] in probabilistic terms
+ $arg =~ tr/*"//d;
+ ${$q->{xap}} .= qq[ $xk"$arg"];
+ undef;
+}
+
sub parse_query {
my ($self, $rest) = @_;
if (uc($rest->[0]) eq 'CHARSET') {
}
my $sql = ''; # date conditions, {sql} deleted if Xapian is needed
- my $q = { xap => '', sql => \$sql };
+ my $xap = '';
+ my $q = { sql => \$sql, xap => \$xap };
while (@$rest) {
my $k = uc(shift @$rest);
# default criteria
next if $k =~ /\A(?:ALL|RECENT|UNSEEN|NEW)\z/;
next if $k eq 'AND'; # the default, until we support OR
- if ($k =~ $valid_range) { # sequence numbers == UIDs
+ if ($k =~ $valid_range) { # convert sequence numbers to UIDs
+ msn_to_uid_range($self, $k);
push @{$q->{uid}}, $k;
} elsif ($k eq 'UID') {
$k = shift(@$rest) // '';
delete $q->{sql}; # can't use over.sqlite3
my $bytes = shift(@$rest) // '';
$bytes =~ /\A[0-9]+\z/ or return "BAD $k not a number";
- $q->{xap} .= ' bytes:' . ($k eq 'SMALLER' ?
+ $xap .= ' bytes:' . ($k eq 'SMALLER' ?
'..'.(--$bytes) :
(++$bytes).'..');
+ } elsif ($k eq 'HEADER') {
+ $k = uc(shift(@$rest) // '');
+ my $xk = $H2X{$k} or
+ return "BAD HEADER $k not supported";
+ my $err = xap_append($q, $rest, $k, $xk);
+ return $err if $err;
} elsif (defined(my $xk = $I2X{$k})) {
- delete $q->{sql}; # can't use over.sqlite3
- my $arg = shift @$rest;
- defined($arg) or return "BAD $k no arg";
-
- # Xapian can't handle [*"] in probabilistic terms
- $arg =~ tr/*"//d;
- $q->{xap} .= qq[ $xk:"$arg"];
+ my $err = xap_append($q, $rest, $k, $xk);
+ return $err if $err;
} else {
# TODO: parentheses, OR, NOT ...
return "BAD $k not supported (yet?)";
} elsif (!$self->{ibx}->search) {
return 'BAD Xapian not configured for mailbox';
}
-
- if (my $uid = $q->{uid}) {
- ((@$uid > 1) || $uid->[0] =~ /,/) and
- return 'BAD multiple ranges not supported, yet';
- ($q->{sql} // $q->{xap}) and
- return 'BAD ranges and queries do not mix, yet';
- $q->{uid} = join(',', @$uid); # TODO: multiple ranges
+ my $max = $self->{ibx}->over->max;
+ if (my $uid = delete $q->{uid}) {
+ my $range_csv = join(',', @$uid);
+ do {
+ my $nxt = range_step($self, \$range_csv);
+ my ($beg, $end) = @$nxt;
+ if ($xap) {
+ $xap .= " uid:$beg..$end";
+ } elsif ($beg == $end) {
+ $sql .= " AND num = $beg";
+ } else {
+ $sql .= " AND num >= $beg AND num <= $end";
+ }
+ } while ($range_csv);
}
+ my $beg = 1;
+ uid_clamp($self, \$beg, \$max);
+ $q->{range_info} = [ $beg, $max ];
$q;
}
+sub refill_xap ($$$$) {
+ my ($self, $uids, $range_info, $q) = @_;
+ my ($beg, $end) = @$range_info;
+ my $srch = $self->{ibx}->search;
+ my $opt = { mset => 2, limit => 1000 };
+ my $nshard = $srch->{nshard} // 1;
+ while (1) {
+ my $mset = $srch->query("$$q uid:$beg..$end", $opt);
+ @$uids = map { mdocid($nshard, $_) } $mset->items;
+ if (@$uids) {
+ $range_info->[0] = $uids->[-1] + 1; # update $beg
+ return;
+ } else { # all done
+ return 0;
+ }
+ }
+}
+
+sub search_xap_range { # long_response
+ my ($self, $tag, $q, $range_info) = @_;
+ my $uids = [];
+ if (defined(my $err = refill_xap($self, $uids, $range_info, $q))) {
+ $err ||= 'OK Search done';
+ $self->write("\r\n$tag $err\r\n");
+ return;
+ }
+ $self->msg_more(join(' ', '', @$uids));
+ 1; # more
+}
+
sub cmd_uid_search ($$$;) {
my ($self, $tag) = splice(@_, 0, 2);
my $ibx = $self->{ibx} or return "$tag BAD No mailbox selected\r\n";
my $q = parse_query($self, \@_);
return "$tag $q\r\n" if !ref($q);
- my $sql = delete $q->{sql};
-
- if (!scalar(keys %$q)) {
+ my ($sql, $range_info) = delete @$q{qw(sql range_info)};
+ if (!scalar(keys %$q)) { # overview.sqlite3
$self->msg_more('* SEARCH');
- my $beg = $self->{uid_min} // 1;
- my $end = $ibx->mm->max;
- uid_clamp($self, \$beg, \$end);
- long_response($self, \&uid_search_uid_range,
- $tag, \$beg, $end, $sql);
- } elsif (my $uid = $q->{uid}) {
- if ($uid =~ /\A([0-9]+):([0-9]+|\*)\z/s) {
- my ($beg, $end) = ($1, $2);
- $end = $ibx->mm->max if $end eq '*';
- uid_clamp($self, \$beg, \$end);
- $self->msg_more('* SEARCH');
- long_response($self, \&uid_search_uid_range,
- $tag, \$beg, $end, $sql);
- } elsif ($uid =~ /\A[0-9]+\z/s) {
- $uid = $ibx->over->get_art($uid) ? " $uid" : '';
- "* SEARCH$uid\r\n$tag OK Search done\r\n";
- } else {
- "$tag BAD Error\r\n";
- }
+ long_response($self, \&search_uid_range,
+ $tag, $sql, $range_info);
+ } elsif ($q = $q->{xap}) {
+ $self->msg_more('* SEARCH');
+ long_response($self, \&search_xap_range,
+ $tag, $q, $range_info);
} else {
"$tag BAD Error\r\n";
}
}
+# note: MSN SEARCH is NOT supported. Do any widely-used MUAs
+# rely on MSNs from SEARCH results? Let us know at meta@public-inbox.org
+
sub args_ok ($$) { # duplicated from PublicInbox::NNTP
my ($cb, $argc) = @_;
my $tot = prototype $cb;
# returns 1 if we can continue, 0 if not due to buffered writes or disconnect
sub process_line ($$) {
my ($self, $l) = @_;
+
+ # TODO: IMAP allows literals for big requests to upload messages
+ # (which we don't support) but maybe some big search queries use it.
my ($tag, $req, @args) = parse_line('[ \t]+', 0, $l);
pop(@args) if (@args && !defined($args[-1]));
if (@args && uc($req) eq 'UID') {
my $rbuf = $self->{rbuf} // \(my $x = '');
my $line = index($$rbuf, "\n");
while ($line < 0) {
- return $self->close if length($$rbuf) >= LINE_MAX;
+ if (length($$rbuf) >= LINE_MAX) {
+ $self->write(\"\* BAD request too long\r\n");
+ return $self->close;
+ }
$self->do_read($rbuf, LINE_MAX, length($$rbuf)) or return;
$line = index($$rbuf, "\n");
}
sub close {
my ($self) = @_;
if (my $ibx = delete $self->{ibx}) {
- if (my $sock = $self->{sock}) {;
- $ibx->unsubscribe_unlock(fileno($sock));
- }
+ stop_idle($self, $ibx);
}
$self->SUPER::close; # PublicInbox::DS::close
}
# we're read-only, so SELECT and EXAMINE do the same thing
no warnings 'once';
*cmd_select = \&cmd_examine;
-*cmd_fetch = \&cmd_uid_fetch;
+
+package PublicInbox::IMAP_preauth;
+our @ISA = qw(PublicInbox::IMAP);
+
+sub logged_in { 0 }
1;