1 # Copyright (C) 2016-2018 all contributors <meta@public-inbox.org>
2 # License: AGPL-3.0+ <https://www.gnu.org/licenses/agpl-3.0.txt>
4 # git fast-import-based ssoma-mda MDA replacement
5 # This is only ever run by public-inbox-mda and public-inbox-learn,
6 # not the WWW or NNTP code which only requires read-only access.
7 package PublicInbox::Import;
10 use base qw(PublicInbox::Lock);
11 use PublicInbox::Spawn qw(spawn);
12 use PublicInbox::MID qw(mids mid_mime mid2path);
13 use PublicInbox::Address;
14 use PublicInbox::MsgTime qw(msg_timestamp msg_datestamp);
15 use PublicInbox::ContentId qw(content_digest);
19 my ($class, $git, $name, $email, $ibx) = @_;
20 my $ref = 'refs/heads/master';
22 $ref = $ibx->{ref_head} || 'refs/heads/master';
23 $name ||= $ibx->{name};
24 $email ||= $ibx->{-primary_address};
28 ident => "$name <$email>",
32 path_type => '2/38', # or 'v2'
33 lock_path => "$git->{git_dir}/ssoma.lock", # v2 changes this
38 # idempotent start function
42 return ($self->{in}, $self->{out}) if $self->{pid};
44 my ($in_r, $in_w, $out_r, $out_w);
45 pipe($in_r, $in_w) or die "pipe failed: $!";
46 pipe($out_r, $out_w) or die "pipe failed: $!";
47 my $git = $self->{git};
52 my $ref = $self->{ref};
53 chomp($self->{tip} = $git->qx(qw(rev-parse --revs-only), $ref));
54 if ($self->{path_type} ne '2/38' && $self->{tip}) {
56 my @tree = $git->qx(qw(ls-tree -r -z --name-only), $ref);
58 $self->{-tree} = { map { $_ => 1 } @tree };
61 my $git_dir = $git->{git_dir};
62 my @cmd = ('git', "--git-dir=$git_dir", qw(fast-import
63 --quiet --done --date-format=raw));
64 my $rdr = { 0 => fileno($out_r), 1 => fileno($in_w) };
65 my $pid = spawn(\@cmd, undef, $rdr);
66 die "spawn fast-import failed: $!" unless defined $pid;
69 $self->{out} = $out_w;
72 binmode $out_w, ':raw' or die "binmode :raw failed: $!";
73 binmode $in_r, ':raw' or die "binmode :raw failed: $!";
77 sub wfail () { die "write to fast-import failed: $!" }
79 sub now_raw () { time . ' +0000' }
83 my $b = $mime->body_raw;
84 $b =~ s/(\r?\n)+\z//s;
88 # only used for v1 (ssoma) inboxes
89 sub _check_path ($$$$) {
90 my ($r, $w, $tip, $path) = @_;
92 print $w "ls $tip $path\n" or wfail;
94 defined(my $info = <$r>) or die "EOF from fast-import: $!";
95 $info =~ /\Amissing / ? undef : $info;
99 my ($r, $w, $oid) = @_;
100 print $w "cat-blob $oid\n" or wfail;
103 defined $info or die "EOF from fast-import / cat-blob: $!";
104 $info =~ /\A[a-f0-9]{40} blob (\d+)\n\z/ or return;
110 $n = read($r, $buf, $left, $offset);
111 defined($n) or die "read cat-blob failed: $!";
112 $n == 0 and die 'fast-export (cat-blob) died';
116 $n = read($r, my $lf, 1);
117 defined($n) or die "read final byte of cat-blob failed: $!";
118 die "bad read on final byte: <$lf>" if $lf ne "\n";
123 my ($self, $oid) = @_;
124 my ($r, $w) = $self->gfi_start;
125 _cat_blob($r, $w, $oid);
128 sub check_remove_v1 {
129 my ($r, $w, $tip, $path, $mime) = @_;
131 my $info = _check_path($r, $w, $tip, $path) or return ('MISSING',undef);
132 $info =~ m!\A100644 blob ([a-f0-9]{40})\t!s or die "not blob: $info";
134 my $msg = _cat_blob($r, $w, $oid) or die "BUG: cat-blob $1 failed";
135 my $cur = PublicInbox::MIME->new($msg);
136 my $cur_s = $cur->header('Subject');
137 $cur_s = '' unless defined $cur_s;
138 my $cur_m = $mime->header('Subject');
139 $cur_m = '' unless defined $cur_m;
140 if ($cur_s ne $cur_m || norm_body($cur) ne norm_body($mime)) {
141 return ('MISMATCH', $cur);
148 return unless $self->{pid};
149 print { $self->{out} } "checkpoint\n" or wfail;
154 my ($self, $msg) = @_;
155 return unless $self->{pid};
156 print { $self->{out} } "progress $msg\n" or wfail;
157 $self->{in}->getline eq "progress $msg\n" or die
158 "progress $msg not received\n";
162 sub _update_git_info ($$) {
163 my ($self, $do_gc) = @_;
164 # for compatibility with existing ssoma installations
165 # we can probably remove this entirely by 2020
166 my $git_dir = $self->{git}->{git_dir};
167 my @cmd = ('git', "--git-dir=$git_dir");
168 my $index = "$git_dir/ssoma.index";
169 if (-e $index && !$ENV{FAST}) {
170 my $env = { GIT_INDEX_FILE => $index };
171 run_die([@cmd, qw(read-tree -m -v -i), $self->{ref}], $env);
173 run_die([@cmd, 'update-server-info'], undef);
174 ($self->{path_type} eq '2/38') and eval {
175 require PublicInbox::SearchIdx;
176 my $inbox = $self->{inbox} || $git_dir;
177 my $s = PublicInbox::SearchIdx->new($inbox);
178 $s->index_sync({ ref => $self->{ref} });
180 eval { run_die([@cmd, qw(gc --auto)], undef) } if $do_gc;
186 # For safety, we ensure git checkpoint is complete before because
187 # the data in git is still more important than what is in Xapian
188 # in v2. Performance may be gained by delaying the ->progress
189 # call but we lose safety
192 $self->progress('checkpoint');
193 _update_git_info($self, 0);
200 my ($self, $mark) = @_;
201 die "not active\n" unless $self->{pid};
202 my ($r, $w) = $self->gfi_start;
203 print $w "get-mark $mark\n" or wfail;
204 defined(my $oid = <$r>) or die "get-mark failed, need git 2.6.0+\n";
209 # returns undef on non-existent
210 # ('MISMATCH', Email::MIME) on mismatch
211 # (:MARK, Email::MIME) on success
213 # v2 callers should check with Xapian before calling this as
214 # it is not idempotent.
216 my ($self, $mime, $msg) = @_; # mime = Email::MIME
218 my $path_type = $self->{path_type};
219 my ($path, $err, $cur, $blob);
221 my ($r, $w) = $self->gfi_start;
222 my $tip = $self->{tip};
223 if ($path_type eq '2/38') {
224 $path = mid2path(v1_mid0($mime));
225 ($err, $cur) = check_remove_v1($r, $w, $tip, $path, $mime);
226 return ($err, $cur) if $err;
229 if (ref($mime) eq 'SCALAR') { # optimization used by V2Writable
231 } else { # XXX should not be necessary:
232 my $str = $mime->as_string;
235 my $len = length($$sref);
236 $blob = $self->{mark}++;
237 print $w "blob\nmark :$blob\ndata $len\n",
238 $$sref, "\n" or wfail;
241 my $ref = $self->{ref};
242 my $commit = $self->{mark}++;
243 my $parent = $tip =~ /\A:/ ? $tip : undef;
245 print $w "reset $ref\n" or wfail;
247 my $ident = $self->{ident};
250 my $len = length($msg) + 1;
251 print $w "commit $ref\nmark :$commit\n",
252 "author $ident $now\n",
253 "committer $ident $now\n",
254 "data $len\n$msg\n\n",
255 'from ', ($parent ? $parent : $tip), "\n" or wfail;
257 print $w "D $path\n\n" or wfail;
259 clean_tree_v2($self, $w, 'd');
260 print $w "M 100644 :$blob d\n\n" or wfail;
263 (($self->{tip} = ":$commit"), $cur);
267 my ($ts, $zone) = @_;
268 $ts = 0 if $ts < 0; # git uses unsigned times
272 sub extract_author_info ($) {
276 my $from = $mime->header('From');
277 my ($email) = PublicInbox::Address::emails($from);
278 my ($name) = PublicInbox::Address::names($from);
279 if (!defined($name) || !defined($email)) {
280 $sender = $mime->header('Sender');
281 if (!defined($name)) {
282 ($name) = PublicInbox::Address::names($sender);
284 if (!defined($email)) {
285 ($email) = PublicInbox::Address::emails($sender);
288 if (defined $email) {
289 # quiet down wide character warnings with utf8::encode
290 utf8::encode($email);
293 warn "no email in From: $from or Sender: $sender\n";
296 # git gets confused with:
297 # "'A U Thor <u@example.com>' via foo" <foo@example.com>
299 # <CAD0k6qSUYANxbjjbE4jTW4EeVwOYgBD=bXkSu=akiYC_CB7Ffw@mail.gmail.com>
305 warn "no name in From: $from or Sender: $sender\n";
310 # kill potentially confusing/misleading headers
311 sub drop_unwanted_headers ($) {
314 $mime->header_set($_) for qw(bytes lines content-length status);
315 $mime->header_set($_) for @PublicInbox::MDA::BAD_HEADERS;
318 # used by V2Writable, too
319 sub append_mid ($$) {
320 my ($hdr, $mid0) = @_;
321 # @cur is likely empty if we need to call this sub, but it could
322 # have random unparseable crap which we'll preserve, too.
323 my @cur = $hdr->header_raw('Message-ID');
324 $hdr->header_set('Message-ID', @cur, "<$mid0>");
329 my $hdr = $mime->header_obj;
330 my $mids = mids($hdr);
332 if (!scalar(@$mids)) { # spam often has no Message-Id
333 my $mid0 = digest2mid(content_digest($mime));
334 append_mid($hdr, $mid0);
339 sub clean_tree_v2 ($$$) {
340 my ($self, $w, $keep) = @_;
341 my $tree = $self->{-tree} or return; #v2 only
342 delete $tree->{$keep};
343 foreach (keys %$tree) {
344 print $w "D $_\n" or wfail;
346 %$tree = ($keep => 1);
349 # returns undef on duplicate
350 # returns the :MARK of the most recent commit
352 my ($self, $mime, $check_cb) = @_; # mime = Email::MIME
354 my ($name, $email) = extract_author_info($mime);
355 my $hdr = $mime->header_obj;
356 my @at = msg_datestamp($hdr);
357 my @ct = msg_timestamp($hdr);
358 my $author_time_raw = git_timestamp(@at);
359 my $commit_time_raw = git_timestamp(@ct);
360 my $subject = $mime->header('Subject');
361 $subject = '(no subject)' unless defined $subject;
362 my $path_type = $self->{path_type};
365 if ($path_type eq '2/38') {
366 $path = mid2path(v1_mid0($mime));
367 } else { # v2 layout, one file:
371 my ($r, $w) = $self->gfi_start;
372 my $tip = $self->{tip};
373 if ($path_type eq '2/38') {
374 _check_path($r, $w, $tip, $path) and return;
377 drop_unwanted_headers($mime);
381 $mime = $check_cb->($mime) or return;
384 my $blob = $self->{mark}++;
385 my $str = $mime->as_string;
386 my $n = length($str);
387 $self->{bytes_added} += $n;
388 print $w "blob\nmark :$blob\ndata ", $n, "\n" or wfail;
389 print $w $str, "\n" or wfail;
391 # v2: we need this for Xapian
392 if ($self->{want_object_info}) {
393 my $oid = $self->get_mark(":$blob");
394 $self->{last_object} = [ $oid, $n, \$str ];
396 my $ref = $self->{ref};
397 my $commit = $self->{mark}++;
398 my $parent = $tip =~ /\A:/ ? $tip : undef;
401 print $w "reset $ref\n" or wfail;
404 utf8::encode($subject);
405 print $w "commit $ref\nmark :$commit\n",
406 "author $name <$email> $author_time_raw\n",
407 "committer $self->{ident} $commit_time_raw\n" or wfail;
408 print $w "data ", (length($subject) + 1), "\n",
409 $subject, "\n\n" or wfail;
411 print $w 'from ', ($parent ? $parent : $tip), "\n" or wfail;
413 clean_tree_v2($self, $w, $path);
414 print $w "M 100644 :$blob $path\n\n" or wfail;
416 $self->{tip} = ":$commit";
420 my ($cmd, $env, $rdr) = @_;
421 my $pid = spawn($cmd, $env, $rdr);
422 defined $pid or die "spawning ".join(' ', @$cmd)." failed: $!";
423 waitpid($pid, 0) == $pid or die join(' ', @$cmd) .' did not finish';
424 $? == 0 or die join(' ', @$cmd) . " failed: $?\n";
429 my $w = delete $self->{out} or return;
430 my $r = delete $self->{in} or die 'BUG: missing {in} when done';
431 print $w "done\n" or wfail;
432 my $pid = delete $self->{pid} or die 'BUG: missing {pid} when done';
433 waitpid($pid, 0) == $pid or die 'fast-import did not finish';
434 $? == 0 or die "fast-import failed: $?";
436 _update_git_info($self, 1) if delete $self->{nchg};
443 foreach my $f (qw(in out)) {
444 close $self->{$f} or die "failed to close import[$f]: $!\n";
450 my $b64 = $dig->clone->b64digest;
451 # Make our own URLs nicer:
452 # See "Base 64 Encoding with URL and Filename Safe Alphabet" in RFC4648
455 # We can make this more meaningful with a date prefix or other things,
456 # but this is only needed for crap that fails to generate a Message-ID
457 # or reuses one. In other words, it's usually spammers who hit this
458 # so they don't deserve nice Message-IDs :P
462 sub clean_purge_buffer {
463 my ($oids, $buf) = @_;
464 my $cmt_msg = 'purged '.join(' ',@$oids)."\n";
467 foreach my $i (0..$#$buf) {
469 if ($l =~ /^author .* (\d+ [\+-]?\d+)$/) {
470 $buf->[$i] = "author <> $1\n";
471 } elsif ($l =~ /^data (\d+)/) {
472 $buf->[$i++] = "data " . length($cmt_msg) . "\n";
473 $buf->[$i] = $cmt_msg;
480 my ($self, $purge) = @_;
481 my $tmp = "refs/heads/purge-".((keys %$purge)[0]);
482 my $old = $self->{'ref'};
483 my $git = $self->{git};
484 my @export = (qw(fast-export --no-data --use-done-feature), $old);
485 my ($rd, $pid) = $git->popen(@export);
486 my ($r, $w) = $self->gfi_start;
491 my $tree = $self->{-tree};
493 if (/^reset (?:.+)/) {
494 push @buf, "reset $tmp\n";
495 } elsif (/^commit (?:.+)/) {
497 $w->print(@buf) or wfail;
500 push @buf, "commit $tmp\n";
501 } elsif (/^data (\d+)/) {
502 # only commit message, so $len is small:
503 my $len = $1; # + 1 for trailing "\n"
505 my $n = read($rd, my $buf, $len) or die "read: $!";
506 $len == $n or die "short read ($n < $len)";
508 } elsif (/^M 100644 ([a-f0-9]+) (\w+)/) {
509 my ($oid, $path) = ($1, $2);
510 if ($purge->{$oid}) {
512 delete $tree->{$path};
517 } elsif (/^D (\w+)/) {
519 push @buf, $_ if $tree->{$path};
520 } elsif ($_ eq "\n") {
522 my $out = join('', @buf);
524 warn "purge rewriting\n", $out, "\n";
525 clean_purge_buffer(\@oids, \@buf);
528 $w->print(@buf, "\n") or wfail;
530 } elsif ($_ eq "done\n") {
532 } elsif (/^mark :(\d+)$/) {
540 $w->print(@buf) or wfail;
542 die 'done\n not seen from fast-export' unless $done;
543 chomp(my $cmt = $self->get_mark(":$mark")) if $npurge;
544 $self->{nchg} = 0; # prevent _update_git_info until update-ref:
546 my @git = ('git', "--git-dir=$git->{git_dir}");
548 run_die([@git, qw(update-ref), $old, $tmp]) if $npurge;
550 run_die([@git, qw(update-ref -d), $tmp]);
552 return if $npurge == 0;
554 run_die([@git, qw(-c gc.reflogExpire=now gc --prune=all)]);
556 foreach my $oid (keys %$purge) {
557 my @info = $git->check($oid);
559 warn "$oid not purged\n";
563 _update_git_info($self, 0);
564 die "Failed to purge $err object(s)\n" if $err;
574 PublicInbox::Import - message importer for public-inbox
583 use PublicInbox::Git;
584 use PublicInbox::Import;
586 chomp(my $git_dir = `git rev-parse --git-dir`);
587 $git_dir or die "GIT_DIR= must be specified\n";
588 my $git = PublicInbox::Git->new($git_dir);
589 my @committer = ('inbox', 'inbox@example.org');
590 my $im = PublicInbox::Import->new($git, @committer);
593 my $message = "From: <u\@example.org>\n".
594 "Subject: test message \n" .
595 "Date: Thu, 01 Jan 1970 00:00:00 +0000\n" .
596 "Message-ID: <m\@example.org>\n".
598 my $parsed = Email::MIME->new($message);
599 my $ret = $im->add($parsed);
602 $parsed->header_obj->header_raw('Message-ID'), "\n";
604 print "imported at mark $ret\n";
608 # to remove a message
609 my $junk = Email::MIME->new($message);
610 my ($mark, $orig) = $im->remove($junk);
611 if ($mark eq 'MISSING') {
613 } elsif ($mark eq 'MISMATCH') {
614 print "Message exists but does not match\n\n",
615 $orig->as_string, "\n",;
617 print "removed at mark $mark\n\n",
618 $orig->as_string, "\n";
624 An importer and remover for public-inboxes which takes L<Email::MIME>
625 messages as input and stores them in a ssoma repository as
626 documented in L<https://ssoma.public-inbox.org/ssoma_repository.txt>,
627 except it does not allow duplicate Message-IDs.
629 It requires L<git(1)> and L<git-fast-import(1)> to be installed.
637 my $im = PublicInbox::Import->new($git, @committer);
639 Initialize a new PublicInbox::Import object.
643 my $parsed = Email::MIME->new($message);
646 Adds a message to to the git repository. This will acquire
647 C<$GIT_DIR/ssoma.lock> and start L<git-fast-import(1)> if necessary.
649 Messages added will not be visible to other processes until L</done>
650 is called, but L</remove> may be called on them.
654 my $junk = Email::MIME->new($message);
655 my ($code, $orig) = $im->remove($junk);
657 Removes a message from the repository. On success, it returns
658 a ':'-prefixed numeric code representing the git-fast-import
659 mark and the original messages as an Email::MIME object.
660 If the message could not be found, the code is "MISSING"
661 and the original message is undef. If there is a mismatch where
662 the "Message-ID" is matched but the subject and body do not match,
663 the returned code is "MISMATCH" and the conflicting message
668 Finalizes the L<git-fast-import(1)> and unlocks the repository.
669 Calling this is required to finalize changes to a repository.
677 All feedback welcome via plain-text mail to L<mailto:meta@public-inbox.org>
679 The mail archives are hosted at L<https://public-inbox.org/meta/>
683 Copyright (C) 2016 all contributors L<mailto:meta@public-inbox.org>
685 License: AGPL-3.0+ L<http://www.gnu.org/licenses/agpl-3.0.txt>