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);
17 use POSIX qw(strftime);
20 my ($class, $git, $name, $email, $ibx) = @_;
21 my $ref = 'refs/heads/master';
23 $ref = $ibx->{ref_head} || 'refs/heads/master';
24 $name ||= $ibx->{name};
25 $email ||= $ibx->{-primary_address};
29 ident => "$name <$email>",
33 path_type => '2/38', # or 'v2'
34 lock_path => "$git->{git_dir}/ssoma.lock", # v2 changes this
39 # idempotent start function
43 return ($self->{in}, $self->{out}) if $self->{pid};
45 my ($in_r, $in_w, $out_r, $out_w);
46 pipe($in_r, $in_w) or die "pipe failed: $!";
47 pipe($out_r, $out_w) or die "pipe failed: $!";
48 my $git = $self->{git};
53 my $ref = $self->{ref};
54 chomp($self->{tip} = $git->qx(qw(rev-parse --revs-only), $ref));
55 if ($self->{path_type} ne '2/38' && $self->{tip}) {
57 my @tree = $git->qx(qw(ls-tree -r -z --name-only), $ref);
59 $self->{-tree} = { map { $_ => 1 } @tree };
62 my $git_dir = $git->{git_dir};
63 my @cmd = ('git', "--git-dir=$git_dir", qw(fast-import
64 --quiet --done --date-format=raw));
65 my $rdr = { 0 => fileno($out_r), 1 => fileno($in_w) };
66 my $pid = spawn(\@cmd, undef, $rdr);
67 die "spawn fast-import failed: $!" unless defined $pid;
70 $self->{out} = $out_w;
73 binmode $out_w, ':raw' or die "binmode :raw failed: $!";
74 binmode $in_r, ':raw' or die "binmode :raw failed: $!";
78 sub wfail () { die "write to fast-import failed: $!" }
80 sub now_raw () { time . ' +0000' }
84 my $b = $mime->body_raw;
85 $b =~ s/(\r?\n)+\z//s;
89 # only used for v1 (ssoma) inboxes
90 sub _check_path ($$$$) {
91 my ($r, $w, $tip, $path) = @_;
93 print $w "ls $tip $path\n" or wfail;
95 defined(my $info = <$r>) or die "EOF from fast-import: $!";
96 $info =~ /\Amissing / ? undef : $info;
100 my ($r, $w, $oid) = @_;
101 print $w "cat-blob $oid\n" or wfail;
104 defined $info or die "EOF from fast-import / cat-blob: $!";
105 $info =~ /\A[a-f0-9]{40} blob (\d+)\n\z/ or return;
111 $n = read($r, $buf, $left, $offset);
112 defined($n) or die "read cat-blob failed: $!";
113 $n == 0 and die 'fast-export (cat-blob) died';
117 $n = read($r, my $lf, 1);
118 defined($n) or die "read final byte of cat-blob failed: $!";
119 die "bad read on final byte: <$lf>" if $lf ne "\n";
124 my ($self, $oid) = @_;
125 my ($r, $w) = $self->gfi_start;
126 _cat_blob($r, $w, $oid);
129 sub check_remove_v1 {
130 my ($r, $w, $tip, $path, $mime) = @_;
132 my $info = _check_path($r, $w, $tip, $path) or return ('MISSING',undef);
133 $info =~ m!\A100644 blob ([a-f0-9]{40})\t!s or die "not blob: $info";
135 my $msg = _cat_blob($r, $w, $oid) or die "BUG: cat-blob $1 failed";
136 my $cur = PublicInbox::MIME->new($msg);
137 my $cur_s = $cur->header('Subject');
138 $cur_s = '' unless defined $cur_s;
139 my $cur_m = $mime->header('Subject');
140 $cur_m = '' unless defined $cur_m;
141 if ($cur_s ne $cur_m || norm_body($cur) ne norm_body($mime)) {
142 return ('MISMATCH', $cur);
149 return unless $self->{pid};
150 print { $self->{out} } "checkpoint\n" or wfail;
155 my ($self, $msg) = @_;
156 return unless $self->{pid};
157 print { $self->{out} } "progress $msg\n" or wfail;
158 $self->{in}->getline eq "progress $msg\n" or die
159 "progress $msg not received\n";
163 sub _update_git_info ($$) {
164 my ($self, $do_gc) = @_;
165 # for compatibility with existing ssoma installations
166 # we can probably remove this entirely by 2020
167 my $git_dir = $self->{git}->{git_dir};
168 my @cmd = ('git', "--git-dir=$git_dir");
169 my $index = "$git_dir/ssoma.index";
170 if (-e $index && !$ENV{FAST}) {
171 my $env = { GIT_INDEX_FILE => $index };
172 run_die([@cmd, qw(read-tree -m -v -i), $self->{ref}], $env);
174 run_die([@cmd, 'update-server-info'], undef);
175 ($self->{path_type} eq '2/38') and eval {
176 require PublicInbox::SearchIdx;
177 my $inbox = $self->{inbox} || $git_dir;
178 my $s = PublicInbox::SearchIdx->new($inbox);
179 $s->index_sync({ ref => $self->{ref} });
181 eval { run_die([@cmd, qw(gc --auto)], undef) } if $do_gc;
187 # For safety, we ensure git checkpoint is complete before because
188 # the data in git is still more important than what is in Xapian
189 # in v2. Performance may be gained by delaying the ->progress
190 # call but we lose safety
193 $self->progress('checkpoint');
194 _update_git_info($self, 0);
201 my ($self, $mark) = @_;
202 die "not active\n" unless $self->{pid};
203 my ($r, $w) = $self->gfi_start;
204 print $w "get-mark $mark\n" or wfail;
205 defined(my $oid = <$r>) or die "get-mark failed, need git 2.6.0+\n";
210 # returns undef on non-existent
211 # ('MISMATCH', Email::MIME) on mismatch
212 # (:MARK, Email::MIME) on success
214 # v2 callers should check with Xapian before calling this as
215 # it is not idempotent.
217 my ($self, $mime, $msg) = @_; # mime = Email::MIME
219 my $path_type = $self->{path_type};
220 my ($path, $err, $cur, $blob);
222 my ($r, $w) = $self->gfi_start;
223 my $tip = $self->{tip};
224 if ($path_type eq '2/38') {
225 $path = mid2path(v1_mid0($mime));
226 ($err, $cur) = check_remove_v1($r, $w, $tip, $path, $mime);
227 return ($err, $cur) if $err;
230 if (ref($mime) eq 'SCALAR') { # optimization used by V2Writable
232 } else { # XXX should not be necessary:
233 my $str = $mime->as_string;
236 my $len = length($$sref);
237 $blob = $self->{mark}++;
238 print $w "blob\nmark :$blob\ndata $len\n",
239 $$sref, "\n" or wfail;
242 my $ref = $self->{ref};
243 my $commit = $self->{mark}++;
244 my $parent = $tip =~ /\A:/ ? $tip : undef;
246 print $w "reset $ref\n" or wfail;
248 my $ident = $self->{ident};
251 my $len = length($msg) + 1;
252 print $w "commit $ref\nmark :$commit\n",
253 "author $ident $now\n",
254 "committer $ident $now\n",
255 "data $len\n$msg\n\n",
256 'from ', ($parent ? $parent : $tip), "\n" or wfail;
258 print $w "D $path\n\n" or wfail;
260 clean_tree_v2($self, $w, 'd');
261 print $w "M 100644 :$blob d\n\n" or wfail;
264 (($self->{tip} = ":$commit"), $cur);
268 my ($ts, $zone) = @_;
269 $ts = 0 if $ts < 0; # git uses unsigned times
273 sub extract_author_info ($) {
277 my $from = $mime->header('From');
278 my ($email) = PublicInbox::Address::emails($from);
279 my ($name) = PublicInbox::Address::names($from);
280 if (!defined($name) || !defined($email)) {
281 $sender = $mime->header('Sender');
282 if (!defined($name)) {
283 ($name) = PublicInbox::Address::names($sender);
285 if (!defined($email)) {
286 ($email) = PublicInbox::Address::emails($sender);
289 if (defined $email) {
290 # quiet down wide character warnings with utf8::encode
291 utf8::encode($email);
294 warn "no email in From: $from or Sender: $sender\n";
297 # git gets confused with:
298 # "'A U Thor <u@example.com>' via foo" <foo@example.com>
300 # <CAD0k6qSUYANxbjjbE4jTW4EeVwOYgBD=bXkSu=akiYC_CB7Ffw@mail.gmail.com>
306 warn "no name in From: $from or Sender: $sender\n";
311 # kill potentially confusing/misleading headers
312 sub drop_unwanted_headers ($) {
315 $mime->header_set($_) for qw(bytes lines content-length status);
316 $mime->header_set($_) for @PublicInbox::MDA::BAD_HEADERS;
319 # used by V2Writable, too
320 sub append_mid ($$) {
321 my ($hdr, $mid0) = @_;
322 # @cur is likely empty if we need to call this sub, but it could
323 # have random unparseable crap which we'll preserve, too.
324 my @cur = $hdr->header_raw('Message-ID');
325 $hdr->header_set('Message-ID', @cur, "<$mid0>");
330 my $hdr = $mime->header_obj;
331 my $mids = mids($hdr);
333 if (!scalar(@$mids)) { # spam often has no Message-Id
334 my $mid0 = digest2mid(content_digest($mime), $hdr);
335 append_mid($hdr, $mid0);
340 sub clean_tree_v2 ($$$) {
341 my ($self, $w, $keep) = @_;
342 my $tree = $self->{-tree} or return; #v2 only
343 delete $tree->{$keep};
344 foreach (keys %$tree) {
345 print $w "D $_\n" or wfail;
347 %$tree = ($keep => 1);
350 # returns undef on duplicate
351 # returns the :MARK of the most recent commit
353 my ($self, $mime, $check_cb) = @_; # mime = Email::MIME
355 my ($name, $email) = extract_author_info($mime);
356 my $hdr = $mime->header_obj;
357 my @at = msg_datestamp($hdr);
358 my @ct = msg_timestamp($hdr);
359 my $author_time_raw = git_timestamp(@at);
360 my $commit_time_raw = git_timestamp(@ct);
361 my $subject = $mime->header('Subject');
362 $subject = '(no subject)' unless defined $subject;
363 my $path_type = $self->{path_type};
366 if ($path_type eq '2/38') {
367 $path = mid2path(v1_mid0($mime));
368 } else { # v2 layout, one file:
372 my ($r, $w) = $self->gfi_start;
373 my $tip = $self->{tip};
374 if ($path_type eq '2/38') {
375 _check_path($r, $w, $tip, $path) and return;
378 drop_unwanted_headers($mime);
382 $mime = $check_cb->($mime) or return;
385 my $blob = $self->{mark}++;
386 my $str = $mime->as_string;
387 my $n = length($str);
388 $self->{bytes_added} += $n;
389 print $w "blob\nmark :$blob\ndata ", $n, "\n" or wfail;
390 print $w $str, "\n" or wfail;
392 # v2: we need this for Xapian
393 if ($self->{want_object_info}) {
394 my $oid = $self->get_mark(":$blob");
395 $self->{last_object} = [ $oid, $n, \$str ];
397 my $ref = $self->{ref};
398 my $commit = $self->{mark}++;
399 my $parent = $tip =~ /\A:/ ? $tip : undef;
402 print $w "reset $ref\n" or wfail;
405 utf8::encode($subject);
406 print $w "commit $ref\nmark :$commit\n",
407 "author $name <$email> $author_time_raw\n",
408 "committer $self->{ident} $commit_time_raw\n" or wfail;
409 print $w "data ", (length($subject) + 1), "\n",
410 $subject, "\n\n" or wfail;
412 print $w 'from ', ($parent ? $parent : $tip), "\n" or wfail;
414 clean_tree_v2($self, $w, $path);
415 print $w "M 100644 :$blob $path\n\n" or wfail;
417 $self->{tip} = ":$commit";
421 my ($cmd, $env, $rdr) = @_;
422 my $pid = spawn($cmd, $env, $rdr);
423 defined $pid or die "spawning ".join(' ', @$cmd)." failed: $!";
424 waitpid($pid, 0) == $pid or die join(' ', @$cmd) .' did not finish';
425 $? == 0 or die join(' ', @$cmd) . " failed: $?\n";
430 my $w = delete $self->{out} or return;
431 my $r = delete $self->{in} or die 'BUG: missing {in} when done';
432 print $w "done\n" or wfail;
433 my $pid = delete $self->{pid} or die 'BUG: missing {pid} when done';
434 waitpid($pid, 0) == $pid or die 'fast-import did not finish';
435 $? == 0 or die "fast-import failed: $?";
437 _update_git_info($self, 1) if delete $self->{nchg};
444 foreach my $f (qw(in out)) {
445 close $self->{$f} or die "failed to close import[$f]: $!\n";
449 sub digest2mid ($$) {
450 my ($dig, $hdr) = @_;
451 my $b64 = $dig->clone->b64digest;
452 # Make our own URLs nicer:
453 # See "Base 64 Encoding with URL and Filename Safe Alphabet" in RFC4648
456 # Add a date prefix to prevent a leading '-' in case that trips
457 # up some tools (e.g. if a Message-ID were a expected as a
459 my $dt = msg_datestamp($hdr);
460 $dt = POSIX::strftime('%Y%m%d%H%M%S', gmtime($dt));
464 sub clean_purge_buffer {
465 my ($oids, $buf) = @_;
466 my $cmt_msg = 'purged '.join(' ',@$oids)."\n";
469 foreach my $i (0..$#$buf) {
471 if ($l =~ /^author .* (\d+ [\+-]?\d+)$/) {
472 $buf->[$i] = "author <> $1\n";
473 } elsif ($l =~ /^data (\d+)/) {
474 $buf->[$i++] = "data " . length($cmt_msg) . "\n";
475 $buf->[$i] = $cmt_msg;
482 my ($self, $purge) = @_;
483 my $tmp = "refs/heads/purge-".((keys %$purge)[0]);
484 my $old = $self->{'ref'};
485 my $git = $self->{git};
486 my @export = (qw(fast-export --no-data --use-done-feature), $old);
487 my ($rd, $pid) = $git->popen(@export);
488 my ($r, $w) = $self->gfi_start;
493 my $tree = $self->{-tree};
495 if (/^reset (?:.+)/) {
496 push @buf, "reset $tmp\n";
497 } elsif (/^commit (?:.+)/) {
499 $w->print(@buf) or wfail;
502 push @buf, "commit $tmp\n";
503 } elsif (/^data (\d+)/) {
504 # only commit message, so $len is small:
505 my $len = $1; # + 1 for trailing "\n"
507 my $n = read($rd, my $buf, $len) or die "read: $!";
508 $len == $n or die "short read ($n < $len)";
510 } elsif (/^M 100644 ([a-f0-9]+) (\w+)/) {
511 my ($oid, $path) = ($1, $2);
512 if ($purge->{$oid}) {
514 delete $tree->{$path};
519 } elsif (/^D (\w+)/) {
521 push @buf, $_ if $tree->{$path};
522 } elsif ($_ eq "\n") {
524 my $out = join('', @buf);
526 warn "purge rewriting\n", $out, "\n";
527 clean_purge_buffer(\@oids, \@buf);
530 $w->print(@buf, "\n") or wfail;
532 } elsif ($_ eq "done\n") {
534 } elsif (/^mark :(\d+)$/) {
542 $w->print(@buf) or wfail;
544 die 'done\n not seen from fast-export' unless $done;
545 chomp(my $cmt = $self->get_mark(":$mark")) if $npurge;
546 $self->{nchg} = 0; # prevent _update_git_info until update-ref:
548 my @git = ('git', "--git-dir=$git->{git_dir}");
550 run_die([@git, qw(update-ref), $old, $tmp]) if $npurge;
552 run_die([@git, qw(update-ref -d), $tmp]);
554 return if $npurge == 0;
556 run_die([@git, qw(-c gc.reflogExpire=now gc --prune=all)]);
558 foreach my $oid (keys %$purge) {
559 my @info = $git->check($oid);
561 warn "$oid not purged\n";
565 _update_git_info($self, 0);
566 die "Failed to purge $err object(s)\n" if $err;
576 PublicInbox::Import - message importer for public-inbox
585 use PublicInbox::Git;
586 use PublicInbox::Import;
588 chomp(my $git_dir = `git rev-parse --git-dir`);
589 $git_dir or die "GIT_DIR= must be specified\n";
590 my $git = PublicInbox::Git->new($git_dir);
591 my @committer = ('inbox', 'inbox@example.org');
592 my $im = PublicInbox::Import->new($git, @committer);
595 my $message = "From: <u\@example.org>\n".
596 "Subject: test message \n" .
597 "Date: Thu, 01 Jan 1970 00:00:00 +0000\n" .
598 "Message-ID: <m\@example.org>\n".
600 my $parsed = Email::MIME->new($message);
601 my $ret = $im->add($parsed);
604 $parsed->header_obj->header_raw('Message-ID'), "\n";
606 print "imported at mark $ret\n";
610 # to remove a message
611 my $junk = Email::MIME->new($message);
612 my ($mark, $orig) = $im->remove($junk);
613 if ($mark eq 'MISSING') {
615 } elsif ($mark eq 'MISMATCH') {
616 print "Message exists but does not match\n\n",
617 $orig->as_string, "\n",;
619 print "removed at mark $mark\n\n",
620 $orig->as_string, "\n";
626 An importer and remover for public-inboxes which takes L<Email::MIME>
627 messages as input and stores them in a ssoma repository as
628 documented in L<https://ssoma.public-inbox.org/ssoma_repository.txt>,
629 except it does not allow duplicate Message-IDs.
631 It requires L<git(1)> and L<git-fast-import(1)> to be installed.
639 my $im = PublicInbox::Import->new($git, @committer);
641 Initialize a new PublicInbox::Import object.
645 my $parsed = Email::MIME->new($message);
648 Adds a message to to the git repository. This will acquire
649 C<$GIT_DIR/ssoma.lock> and start L<git-fast-import(1)> if necessary.
651 Messages added will not be visible to other processes until L</done>
652 is called, but L</remove> may be called on them.
656 my $junk = Email::MIME->new($message);
657 my ($code, $orig) = $im->remove($junk);
659 Removes a message from the repository. On success, it returns
660 a ':'-prefixed numeric code representing the git-fast-import
661 mark and the original messages as an Email::MIME object.
662 If the message could not be found, the code is "MISSING"
663 and the original message is undef. If there is a mismatch where
664 the "Message-ID" is matched but the subject and body do not match,
665 the returned code is "MISMATCH" and the conflicting message
670 Finalizes the L<git-fast-import(1)> and unlocks the repository.
671 Calling this is required to finalize changes to a repository.
679 All feedback welcome via plain-text mail to L<mailto:meta@public-inbox.org>
681 The mail archives are hosted at L<https://public-inbox.org/meta/>
685 Copyright (C) 2016 all contributors L<mailto:meta@public-inbox.org>
687 License: AGPL-3.0+ L<http://www.gnu.org/licenses/agpl-3.0.txt>