1 # Copyright (C) 2016-2019 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, public-inbox-learn
6 # and public-inbox-watch. Not the WWW or NNTP code which only
7 # requires read-only access.
8 package PublicInbox::Import;
11 use base qw(PublicInbox::Lock);
12 use PublicInbox::Spawn qw(spawn);
13 use PublicInbox::MID qw(mids mid2path);
14 use PublicInbox::Address;
15 use PublicInbox::MsgTime qw(msg_timestamp msg_datestamp);
16 use PublicInbox::ContentId qw(content_digest);
18 use POSIX qw(strftime);
21 # we can't change arg order, this is documented in POD
22 # and external projects may rely on it:
23 my ($class, $git, $name, $email, $ibx) = @_;
24 my $ref = 'refs/heads/master';
26 $ref = $ibx->{ref_head} || 'refs/heads/master';
27 $name ||= $ibx->{name};
28 $email ||= $ibx->{-primary_address};
33 ident => "$name <$email>",
37 path_type => '2/38', # or 'v2'
38 lock_path => "$git->{git_dir}/ssoma.lock", # v2 changes this
43 # idempotent start function
47 return ($self->{in}, $self->{out}) if $self->{pid};
49 my ($in_r, $in_w, $out_r, $out_w);
50 pipe($in_r, $in_w) or die "pipe failed: $!";
51 pipe($out_r, $out_w) or die "pipe failed: $!";
52 my $git = $self->{git};
57 my $ref = $self->{ref};
58 chomp($self->{tip} = $git->qx(qw(rev-parse --revs-only), $ref));
59 if ($self->{path_type} ne '2/38' && $self->{tip}) {
61 my @tree = $git->qx(qw(ls-tree -r -z --name-only), $ref);
63 $self->{-tree} = { map { $_ => 1 } @tree };
66 my $git_dir = $git->{git_dir};
67 my @cmd = ('git', "--git-dir=$git_dir", qw(fast-import
68 --quiet --done --date-format=raw));
69 my $rdr = { 0 => $out_r, 1 => $in_w };
70 my $pid = spawn(\@cmd, undef, $rdr);
73 $self->{out} = $out_w;
76 binmode $out_w, ':raw' or die "binmode :raw failed: $!";
77 binmode $in_r, ':raw' or die "binmode :raw failed: $!";
81 sub wfail () { die "write to fast-import failed: $!" }
83 sub now_raw () { time . ' +0000' }
87 my $b = $mime->body_raw;
88 $b =~ s/(\r?\n)+\z//s;
92 # only used for v1 (ssoma) inboxes
93 sub _check_path ($$$$) {
94 my ($r, $w, $tip, $path) = @_;
96 print $w "ls $tip $path\n" or wfail;
98 defined(my $info = <$r>) or die "EOF from fast-import: $!";
99 $info =~ /\Amissing / ? undef : $info;
102 sub _cat_blob ($$$) {
103 my ($r, $w, $oid) = @_;
104 print $w "cat-blob $oid\n" or wfail;
107 defined $info or die "EOF from fast-import / cat-blob: $!";
108 $info =~ /\A[a-f0-9]{40} blob ([0-9]+)\n\z/ or return;
114 $n = read($r, $buf, $left, $offset);
115 defined($n) or die "read cat-blob failed: $!";
116 $n == 0 and die 'fast-export (cat-blob) died';
120 $n = read($r, my $lf, 1);
121 defined($n) or die "read final byte of cat-blob failed: $!";
122 die "bad read on final byte: <$lf>" if $lf ne "\n";
124 # fixup some bugginess in old versions:
125 $buf =~ s/\A[\r\n]*From [^\r\n]*\r?\n//s;
130 my ($self, $oid) = @_;
131 my ($r, $w) = $self->gfi_start;
132 _cat_blob($r, $w, $oid);
135 sub check_remove_v1 {
136 my ($r, $w, $tip, $path, $mime) = @_;
138 my $info = _check_path($r, $w, $tip, $path) or return ('MISSING',undef);
139 $info =~ m!\A100644 blob ([a-f0-9]{40})\t!s or die "not blob: $info";
141 my $msg = _cat_blob($r, $w, $oid) or die "BUG: cat-blob $1 failed";
142 my $cur = PublicInbox::MIME->new($msg);
143 my $cur_s = $cur->header('Subject');
144 $cur_s = '' unless defined $cur_s;
145 my $cur_m = $mime->header('Subject');
146 $cur_m = '' unless defined $cur_m;
147 if ($cur_s ne $cur_m || norm_body($cur) ne norm_body($mime)) {
148 return ('MISMATCH', $cur);
155 return unless $self->{pid};
156 print { $self->{out} } "checkpoint\n" or wfail;
161 my ($self, $msg) = @_;
162 return unless $self->{pid};
163 print { $self->{out} } "progress $msg\n" or wfail;
164 $self->{in}->getline eq "progress $msg\n" or die
165 "progress $msg not received\n";
169 sub _update_git_info ($$) {
170 my ($self, $do_gc) = @_;
171 # for compatibility with existing ssoma installations
172 # we can probably remove this entirely by 2020
173 my $git_dir = $self->{git}->{git_dir};
174 my @cmd = ('git', "--git-dir=$git_dir");
175 my $index = "$git_dir/ssoma.index";
176 if (-e $index && !$ENV{FAST}) {
177 my $env = { GIT_INDEX_FILE => $index };
178 run_die([@cmd, qw(read-tree -m -v -i), $self->{ref}], $env);
180 run_die([@cmd, 'update-server-info']);
181 my $ibx = $self->{-inbox};
182 ($ibx && $self->{path_type} eq '2/38') and eval {
183 require PublicInbox::SearchIdx;
184 my $s = PublicInbox::SearchIdx->new($ibx);
185 $s->index_sync({ ref => $self->{ref} });
187 eval { run_die([@cmd, qw(gc --auto)]) } if $do_gc;
193 # For safety, we ensure git checkpoint is complete before because
194 # the data in git is still more important than what is in Xapian
195 # in v2. Performance may be gained by delaying the ->progress
196 # call but we lose safety
199 $self->progress('checkpoint');
200 _update_git_info($self, 0);
207 my ($self, $mark) = @_;
208 die "not active\n" unless $self->{pid};
209 my ($r, $w) = $self->gfi_start;
210 print $w "get-mark $mark\n" or wfail;
211 defined(my $oid = <$r>) or die "get-mark failed, need git 2.6.0+\n";
216 # returns undef on non-existent
217 # ('MISMATCH', Email::MIME) on mismatch
218 # (:MARK, Email::MIME) on success
220 # v2 callers should check with Xapian before calling this as
221 # it is not idempotent.
223 my ($self, $mime, $msg) = @_; # mime = Email::MIME
225 my $path_type = $self->{path_type};
226 my ($path, $err, $cur, $blob);
228 my ($r, $w) = $self->gfi_start;
229 my $tip = $self->{tip};
230 if ($path_type eq '2/38') {
231 $path = mid2path(v1_mid0($mime));
232 ($err, $cur) = check_remove_v1($r, $w, $tip, $path, $mime);
233 return ($err, $cur) if $err;
236 if (ref($mime) eq 'SCALAR') { # optimization used by V2Writable
238 } else { # XXX should not be necessary:
239 my $str = $mime->as_string;
242 my $len = length($$sref);
243 $blob = $self->{mark}++;
244 print $w "blob\nmark :$blob\ndata $len\n",
245 $$sref, "\n" or wfail;
248 my $ref = $self->{ref};
249 my $commit = $self->{mark}++;
250 my $parent = $tip =~ /\A:/ ? $tip : undef;
252 print $w "reset $ref\n" or wfail;
254 my $ident = $self->{ident};
257 my $len = length($msg) + 1;
258 print $w "commit $ref\nmark :$commit\n",
259 "author $ident $now\n",
260 "committer $ident $now\n",
261 "data $len\n$msg\n\n",
262 'from ', ($parent ? $parent : $tip), "\n" or wfail;
264 print $w "D $path\n\n" or wfail;
266 clean_tree_v2($self, $w, 'd');
267 print $w "M 100644 :$blob d\n\n" or wfail;
270 (($self->{tip} = ":$commit"), $cur);
274 my ($ts, $zone) = @_;
275 $ts = 0 if $ts < 0; # git uses unsigned times
279 sub extract_cmt_info ($) {
283 my $from = $mime->header('From');
285 my ($email) = PublicInbox::Address::emails($from);
286 my ($name) = PublicInbox::Address::names($from);
287 if (!defined($name) || !defined($email)) {
288 $sender = $mime->header('Sender');
290 if (!defined($name)) {
291 ($name) = PublicInbox::Address::names($sender);
293 if (!defined($email)) {
294 ($email) = PublicInbox::Address::emails($sender);
297 if (defined $email) {
298 # quiet down wide character warnings with utf8::encode
299 utf8::encode($email);
302 warn "no email in From: $from or Sender: $sender\n";
305 # git gets confused with:
306 # "'A U Thor <u@example.com>' via foo" <foo@example.com>
308 # <CAD0k6qSUYANxbjjbE4jTW4EeVwOYgBD=bXkSu=akiYC_CB7Ffw@mail.gmail.com>
314 warn "no name in From: $from or Sender: $sender\n";
317 my $hdr = $mime->header_obj;
319 my $subject = $hdr->header('Subject');
320 $subject = '(no subject)' unless defined $subject;
321 # Mime decoding can create nulls replace them with spaces to protect git
322 $subject =~ tr/\0/ /;
323 utf8::encode($subject);
324 my $at = git_timestamp(my @at = msg_datestamp($hdr));
325 my $ct = git_timestamp(my @ct = msg_timestamp($hdr));
326 ($name, $email, $at, $ct, $subject);
329 # kill potentially confusing/misleading headers
330 sub drop_unwanted_headers ($) {
333 $mime->header_set($_) for qw(bytes lines content-length status);
334 $mime->header_set($_) for @PublicInbox::MDA::BAD_HEADERS;
337 # used by V2Writable, too
338 sub append_mid ($$) {
339 my ($hdr, $mid0) = @_;
340 # @cur is likely empty if we need to call this sub, but it could
341 # have random unparseable crap which we'll preserve, too.
342 my @cur = $hdr->header_raw('Message-ID');
343 $hdr->header_set('Message-ID', @cur, "<$mid0>");
348 my $hdr = $mime->header_obj;
349 my $mids = mids($hdr);
351 if (!scalar(@$mids)) { # spam often has no Message-Id
352 my $mid0 = digest2mid(content_digest($mime), $hdr);
353 append_mid($hdr, $mid0);
358 sub clean_tree_v2 ($$$) {
359 my ($self, $w, $keep) = @_;
360 my $tree = $self->{-tree} or return; #v2 only
361 delete $tree->{$keep};
362 foreach (keys %$tree) {
363 print $w "D $_\n" or wfail;
365 %$tree = ($keep => 1);
368 # returns undef on duplicate
369 # returns the :MARK of the most recent commit
371 my ($self, $mime, $check_cb) = @_; # mime = Email::MIME
373 my ($name, $email, $at, $ct, $subject) = extract_cmt_info($mime);
374 my $path_type = $self->{path_type};
376 if ($path_type eq '2/38') {
377 $path = mid2path(v1_mid0($mime));
378 } else { # v2 layout, one file:
382 my ($r, $w) = $self->gfi_start;
383 my $tip = $self->{tip};
384 if ($path_type eq '2/38') {
385 _check_path($r, $w, $tip, $path) and return;
388 drop_unwanted_headers($mime);
392 $mime = $check_cb->($mime) or return;
395 my $blob = $self->{mark}++;
396 my $raw_email = $mime->{-public_inbox_raw} // $mime->as_string;
397 my $n = length($raw_email);
398 $self->{bytes_added} += $n;
399 print $w "blob\nmark :$blob\ndata ", $n, "\n" or wfail;
400 print $w $raw_email, "\n" or wfail;
402 # v2: we need this for Xapian
403 if ($self->{want_object_info}) {
404 my $oid = $self->get_mark(":$blob");
405 $self->{last_object} = [ $oid, $n, \$raw_email ];
407 my $ref = $self->{ref};
408 my $commit = $self->{mark}++;
409 my $parent = $tip =~ /\A:/ ? $tip : undef;
412 print $w "reset $ref\n" or wfail;
415 print $w "commit $ref\nmark :$commit\n",
416 "author $name <$email> $at\n",
417 "committer $self->{ident} $ct\n" or wfail;
418 print $w "data ", (length($subject) + 1), "\n",
419 $subject, "\n\n" or wfail;
421 print $w 'from ', ($parent ? $parent : $tip), "\n" or wfail;
423 clean_tree_v2($self, $w, $path);
424 print $w "M 100644 :$blob $path\n\n" or wfail;
426 $self->{tip} = ":$commit";
430 my ($cmd, $env, $rdr) = @_;
431 my $pid = spawn($cmd, $env, $rdr);
432 waitpid($pid, 0) == $pid or die join(' ', @$cmd) .' did not finish';
433 $? == 0 or die join(' ', @$cmd) . " failed: $?\n";
438 my @cmd = (qw(git init --bare -q), $dir);
440 # set a reasonable default:
441 @cmd = (qw/git config/, "--file=$dir/config",
442 'repack.writeBitmaps', 'true');
448 my $w = delete $self->{out} or return;
449 my $r = delete $self->{in} or die 'BUG: missing {in} when done';
450 print $w "done\n" or wfail;
451 my $pid = delete $self->{pid} or die 'BUG: missing {pid} when done';
452 waitpid($pid, 0) == $pid or die 'fast-import did not finish';
453 $? == 0 or die "fast-import failed: $?";
455 _update_git_info($self, 1) if delete $self->{nchg};
459 $self->{git}->cleanup;
464 foreach my $f (qw(in out)) {
465 next unless defined($self->{$f});
466 close $self->{$f} or die "failed to close import[$f]: $!\n";
470 sub digest2mid ($$) {
471 my ($dig, $hdr) = @_;
472 my $b64 = $dig->clone->b64digest;
473 # Make our own URLs nicer:
474 # See "Base 64 Encoding with URL and Filename Safe Alphabet" in RFC4648
477 # Add a date prefix to prevent a leading '-' in case that trips
478 # up some tools (e.g. if a Message-ID were a expected as a
480 my $dt = msg_datestamp($hdr);
481 $dt = POSIX::strftime('%Y%m%d%H%M%S', gmtime($dt));
485 sub rewrite_commit ($$$$) {
486 my ($self, $oids, $buf, $mime) = @_;
487 my ($name, $email, $at, $ct, $subject);
489 ($name, $email, $at, $ct, $subject) = extract_cmt_info($mime);
492 $subject = 'purged '.join(' ', @$oids);
496 foreach my $i (0..$#$buf) {
498 if ($l =~ /^author .* ([0-9]+ [\+-]?[0-9]+)$/) {
500 $buf->[$i] = "author $name <$email> $at\n";
501 } elsif ($l =~ /^committer .* ([0-9]+ [\+-]?[0-9]+)$/) {
503 $buf->[$i] = "committer $self->{ident} $ct\n";
504 } elsif ($l =~ /^data ([0-9]+)/) {
505 $buf->[$i++] = "data " . length($subject) . "\n";
506 $buf->[$i] = $subject;
512 # returns the new commit OID if a replacement was done
513 # returns undef if nothing was done
515 my ($self, $mime, $replace_map) = @_; # oid => raw string
516 my $tmp = "refs/heads/replace-".((keys %$replace_map)[0]);
517 my $old = $self->{'ref'};
518 my $git = $self->{git};
519 my @export = (qw(fast-export --no-data --use-done-feature), $old);
520 my $rd = $git->popen(@export);
521 my ($r, $w) = $self->gfi_start;
526 my $tree = $self->{-tree};
528 if (/^reset (?:.+)/) {
529 push @buf, "reset $tmp\n";
530 } elsif (/^commit (?:.+)/) {
532 $w->print(@buf) or wfail;
535 push @buf, "commit $tmp\n";
536 } elsif (/^data ([0-9]+)/) {
537 # only commit message, so $len is small:
538 my $len = $1; # + 1 for trailing "\n"
540 my $n = read($rd, my $buf, $len) or die "read: $!";
541 $len == $n or die "short read ($n < $len)";
543 } elsif (/^M 100644 ([a-f0-9]+) (\w+)/) {
544 my ($oid, $path) = ($1, $2);
546 my $sref = $replace_map->{$oid};
549 my $n = length($$sref);
550 push @buf, "M 100644 inline $path\ndata $n\n";
551 push @buf, $$sref; # hope CoW works...
556 } elsif (/^D (\w+)/) {
558 push @buf, $_ if $tree->{$path};
559 } elsif ($_ eq "\n") {
562 my $out = join('', @buf);
564 warn "purge rewriting\n", $out, "\n";
566 rewrite_commit($self, \@oids, \@buf, $mime);
569 $w->print(@buf, "\n") or wfail;
571 } elsif ($_ eq "done\n") {
573 } elsif (/^mark :([0-9]+)$/) {
580 close $rd or die "close fast-export failed: $?";
582 $w->print(@buf) or wfail;
584 die 'done\n not seen from fast-export' unless $done;
585 chomp(my $cmt = $self->get_mark(":$mark")) if $nreplace;
586 $self->{nchg} = 0; # prevent _update_git_info until update-ref:
588 my @git = ('git', "--git-dir=$git->{git_dir}");
590 run_die([@git, qw(update-ref), $old, $tmp]) if $nreplace;
592 run_die([@git, qw(update-ref -d), $tmp]);
594 return if $nreplace == 0;
596 run_die([@git, qw(-c gc.reflogExpire=now gc --prune=all --quiet)]);
598 # check that old OIDs are gone
600 foreach my $oid (keys %$replace_map) {
601 my @info = $git->check($oid);
603 warn "$oid not replaced\n";
607 _update_git_info($self, 0);
608 die "Failed to replace $err object(s)\n" if $err;
618 PublicInbox::Import - message importer for public-inbox v1 inboxes
627 use PublicInbox::Git;
628 use PublicInbox::Import;
630 chomp(my $git_dir = `git rev-parse --git-dir`);
631 $git_dir or die "GIT_DIR= must be specified\n";
632 my $git = PublicInbox::Git->new($git_dir);
633 my @committer = ('inbox', 'inbox@example.org');
634 my $im = PublicInbox::Import->new($git, @committer);
637 my $message = "From: <u\@example.org>\n".
638 "Subject: test message \n" .
639 "Date: Thu, 01 Jan 1970 00:00:00 +0000\n" .
640 "Message-ID: <m\@example.org>\n".
642 my $parsed = Email::MIME->new($message);
643 my $ret = $im->add($parsed);
646 $parsed->header_obj->header_raw('Message-ID'), "\n";
648 print "imported at mark $ret\n";
652 # to remove a message
653 my $junk = Email::MIME->new($message);
654 my ($mark, $orig) = $im->remove($junk);
655 if ($mark eq 'MISSING') {
657 } elsif ($mark eq 'MISMATCH') {
658 print "Message exists but does not match\n\n",
659 $orig->as_string, "\n",;
661 print "removed at mark $mark\n\n",
662 $orig->as_string, "\n";
668 An importer and remover for public-inboxes which takes L<Email::MIME>
669 messages as input and stores them in a git repository as
670 documented in L<https://public-inbox.org/public-inbox-v1-format.txt>,
671 except it does not allow duplicate Message-IDs.
673 It requires L<git(1)> and L<git-fast-import(1)> to be installed.
681 my $im = PublicInbox::Import->new($git, @committer);
683 Initialize a new PublicInbox::Import object.
687 my $parsed = Email::MIME->new($message);
690 Adds a message to to the git repository. This will acquire
691 C<$GIT_DIR/ssoma.lock> and start L<git-fast-import(1)> if necessary.
693 Messages added will not be visible to other processes until L</done>
694 is called, but L</remove> may be called on them.
698 my $junk = Email::MIME->new($message);
699 my ($code, $orig) = $im->remove($junk);
701 Removes a message from the repository. On success, it returns
702 a ':'-prefixed numeric code representing the git-fast-import
703 mark and the original messages as an Email::MIME object.
704 If the message could not be found, the code is "MISSING"
705 and the original message is undef. If there is a mismatch where
706 the "Message-ID" is matched but the subject and body do not match,
707 the returned code is "MISMATCH" and the conflicting message
712 Finalizes the L<git-fast-import(1)> and unlocks the repository.
713 Calling this is required to finalize changes to a repository.
721 All feedback welcome via plain-text mail to L<mailto:meta@public-inbox.org>
723 The mail archives are hosted at L<https://public-inbox.org/meta/>
727 Copyright (C) 2016 all contributors L<mailto:meta@public-inbox.org>
729 License: AGPL-3.0+ L<http://www.gnu.org/licenses/agpl-3.0.txt>