1 # Copyright (C) 2016-2020 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;
10 use parent qw(PublicInbox::Lock);
12 use PublicInbox::Spawn qw(spawn popen_rd);
13 use PublicInbox::MID qw(mids mid2path);
14 use PublicInbox::Address;
15 use PublicInbox::Smsg;
16 use PublicInbox::MsgTime qw(msg_datestamp);
17 use PublicInbox::ContentHash qw(content_digest);
20 use POSIX qw(strftime);
23 # we can't change arg order, this is documented in POD
24 # and external projects may rely on it:
25 my ($class, $git, $name, $email, $ibx) = @_;
26 my $ref = 'refs/heads/master';
28 $ref = $ibx->{ref_head} // 'refs/heads/master';
29 $name //= $ibx->{name};
30 $email //= $ibx->{-primary_address};
35 ident => "$name <$email>",
39 path_type => '2/38', # or 'v2'
40 lock_path => "$git->{git_dir}/ssoma.lock", # v2 changes this
45 # idempotent start function
49 return ($self->{in}, $self->{out}) if $self->{pid};
51 my ($in_r, $pid, $out_r, $out_w);
52 pipe($out_r, $out_w) or die "pipe failed: $!";
56 my ($git, $ref) = @$self{qw(git ref)};
58 chomp($self->{tip} = $git->qx(qw(rev-parse --revs-only), $ref));
59 die "fatal: rev-parse --revs-only $ref: \$?=$?" if $?;
60 if ($self->{path_type} ne '2/38' && $self->{tip}) {
62 my @t = $git->qx(qw(ls-tree -r -z --name-only), $ref);
63 die "fatal: ls-tree -r -z --name-only $ref: \$?=$?" if $?;
65 $self->{-tree} = { map { $_ => 1 } @t };
67 my @cmd = ('git', "--git-dir=$git->{git_dir}",
68 qw(fast-import --quiet --done --date-format=raw));
69 ($in_r, $pid) = popen_rd(\@cmd, undef, { 0 => $out_r });
72 $self->{out} = $out_w;
83 sub wfail () { die "write to fast-import failed: $!" }
85 sub now_raw () { time . ' +0000' }
89 my $b = $mime->body_raw;
90 $b =~ s/(\r?\n)+\z//s;
94 # only used for v1 (ssoma) inboxes
95 sub _check_path ($$$$) {
96 my ($r, $w, $tip, $path) = @_;
98 print $w "ls $tip $path\n" or wfail;
100 defined(my $info = <$r>) or die "EOF from fast-import: $!";
101 $info =~ /\Amissing / ? undef : $info;
104 sub _cat_blob ($$$) {
105 my ($r, $w, $oid) = @_;
106 print $w "cat-blob $oid\n" or wfail;
109 defined $info or die "EOF from fast-import / cat-blob: $!";
110 $info =~ /\A[a-f0-9]{40,} blob ([0-9]+)\n\z/ or return;
116 $n = read($r, $buf, $left, $offset);
117 defined($n) or die "read cat-blob failed: $!";
118 $n == 0 and die 'fast-export (cat-blob) died';
122 $n = read($r, my $lf, 1);
123 defined($n) or die "read final byte of cat-blob failed: $!";
124 die "bad read on final byte: <$lf>" if $lf ne "\n";
126 # fixup some bugginess in old versions:
127 $buf =~ s/\A[\r\n]*From [^\r\n]*\r?\n//s;
132 my ($self, $oid) = @_;
133 my ($r, $w) = $self->gfi_start;
134 _cat_blob($r, $w, $oid);
137 sub check_remove_v1 {
138 my ($r, $w, $tip, $path, $mime) = @_;
140 my $info = _check_path($r, $w, $tip, $path) or return ('MISSING',undef);
141 $info =~ m!\A100644 blob ([a-f0-9]{40,})\t!s or die "not blob: $info";
143 my $msg = _cat_blob($r, $w, $oid) or die "BUG: cat-blob $1 failed";
144 my $cur = PublicInbox::Eml->new($msg);
145 my $cur_s = $cur->header('Subject');
146 $cur_s = '' unless defined $cur_s;
147 my $cur_m = $mime->header('Subject');
148 $cur_m = '' unless defined $cur_m;
149 if ($cur_s ne $cur_m || norm_body($cur) ne norm_body($mime)) {
150 return ('MISMATCH', $cur);
157 return unless $self->{pid};
158 print { $self->{out} } "checkpoint\n" or wfail;
163 my ($self, $msg) = @_;
164 return unless $self->{pid};
165 print { $self->{out} } "progress $msg\n" or wfail;
166 readline($self->{in}) eq "progress $msg\n" or die
167 "progress $msg not received\n";
171 sub _update_git_info ($$) {
172 my ($self, $do_gc) = @_;
173 # for compatibility with existing ssoma installations
174 # we can probably remove this entirely by 2020
175 my $git_dir = $self->{git}->{git_dir};
176 my @cmd = ('git', "--git-dir=$git_dir");
177 my $index = "$git_dir/ssoma.index";
178 if (-e $index && !$ENV{FAST}) {
179 my $env = { GIT_INDEX_FILE => $index };
180 run_die([@cmd, qw(read-tree -m -v -i), $self->{ref}], $env);
182 eval { run_die([@cmd, 'update-server-info']) };
183 my $ibx = $self->{ibx};
184 if ($ibx && $ibx->version == 1 && -d "$ibx->{inboxdir}/public-inbox" &&
185 eval { require PublicInbox::SearchIdx }) {
187 my $s = PublicInbox::SearchIdx->new($ibx);
188 $s->index_sync({ ref => $self->{ref} });
190 warn "$ibx->{inboxdir} index failed: $@\n" if $@;
192 eval { run_die([@cmd, qw(gc --auto)]) } if $do_gc;
198 # For safety, we ensure git checkpoint is complete before because
199 # the data in git is still more important than what is in Xapian
200 # in v2. Performance may be gained by delaying the ->progress
201 # call but we lose safety
204 $self->progress('checkpoint');
205 _update_git_info($self, 0);
212 my ($self, $mark) = @_;
213 die "not active\n" unless $self->{pid};
214 my ($r, $w) = $self->gfi_start;
215 print $w "get-mark $mark\n" or wfail;
216 defined(my $oid = <$r>) or die "get-mark failed, need git 2.6.0+\n";
221 # returns undef on non-existent
222 # ('MISMATCH', PublicInbox::Eml) on mismatch
223 # (:MARK, PublicInbox::Eml) on success
225 # v2 callers should check with Xapian before calling this as
226 # it is not idempotent.
228 my ($self, $mime, $msg) = @_; # mime = PublicInbox::Eml or Email::MIME
230 my $path_type = $self->{path_type};
231 my ($path, $err, $cur, $blob);
233 my ($r, $w) = $self->gfi_start;
234 my $tip = $self->{tip};
235 if ($path_type eq '2/38') {
236 $path = mid2path(v1_mid0($mime));
237 ($err, $cur) = check_remove_v1($r, $w, $tip, $path, $mime);
238 return ($err, $cur) if $err;
241 if (ref($mime) eq 'SCALAR') { # optimization used by V2Writable
243 } else { # XXX should not be necessary:
244 my $str = $mime->as_string;
247 my $len = length($$sref);
248 $blob = $self->{mark}++;
249 print $w "blob\nmark :$blob\ndata $len\n",
250 $$sref, "\n" or wfail;
253 my $ref = $self->{ref};
254 my $commit = $self->{mark}++;
255 my $parent = $tip =~ /\A:/ ? $tip : undef;
257 print $w "reset $ref\n" or wfail;
259 my $ident = $self->{ident};
262 my $len = length($msg) + 1;
263 print $w "commit $ref\nmark :$commit\n",
264 "author $ident $now\n",
265 "committer $ident $now\n",
266 "data $len\n$msg\n\n",
267 'from ', ($parent ? $parent : $tip), "\n" or wfail;
269 print $w "D $path\n\n" or wfail;
271 clean_tree_v2($self, $w, 'd');
272 print $w "M 100644 :$blob d\n\n" or wfail;
275 (($self->{tip} = ":$commit"), $cur);
278 sub git_timestamp ($) {
279 my ($ts, $zone) = @{$_[0]};
280 $ts = 0 if $ts < 0; # git uses unsigned times
284 sub extract_cmt_info ($;$) {
285 my ($mime, $smsg) = @_;
286 # $mime is PublicInbox::Eml, but remains Email::MIME-compatible
287 $smsg //= bless {}, 'PublicInbox::Smsg';
289 $smsg->populate($mime);
292 my $from = delete($smsg->{From}) // '';
293 my ($email) = PublicInbox::Address::emails($from);
294 my ($name) = PublicInbox::Address::names($from);
295 if (!defined($name) || !defined($email)) {
296 $sender = $mime->header('Sender') // '';
297 $name //= (PublicInbox::Address::names($sender))[0];
298 $email //= (PublicInbox::Address::emails($sender))[0];
300 if (defined $email) {
301 # Email::Address::XS may leave quoted '<' in addresses,
302 # which git-fast-import doesn't like
305 # quiet down wide character warnings with utf8::encode
306 utf8::encode($email);
309 warn "no email in From: $from or Sender: $sender\n";
312 # git gets confused with:
313 # "'A U Thor <u@example.com>' via foo" <foo@example.com>
315 # <CAD0k6qSUYANxbjjbE4jTW4EeVwOYgBD=bXkSu=akiYC_CB7Ffw@mail.gmail.com>
321 warn "no name in From: $from or Sender: $sender\n";
324 my $subject = delete($smsg->{Subject}) // '(no subject)';
325 utf8::encode($subject);
326 my $at = git_timestamp(delete $smsg->{-ds});
327 my $ct = git_timestamp(delete $smsg->{-ts});
328 ("$name <$email>", $at, $ct, $subject);
331 # kill potentially confusing/misleading headers
332 our @UNWANTED_HEADERS = (qw(Bytes Lines Content-Length),
333 qw(Status X-Status));
334 sub drop_unwanted_headers ($) {
336 for (@UNWANTED_HEADERS, @PublicInbox::MDA::BAD_HEADERS) {
337 $eml->header_set($_);
341 # used by V2Writable, too
342 sub append_mid ($$) {
343 my ($hdr, $mid0) = @_;
344 # @cur is likely empty if we need to call this sub, but it could
345 # have random unparseable crap which we'll preserve, too.
346 my @cur = $hdr->header_raw('Message-ID');
347 $hdr->header_set('Message-ID', @cur, "<$mid0>");
352 my $mids = mids($eml);
354 if (!scalar(@$mids)) { # spam often has no Message-ID
355 my $mid0 = digest2mid(content_digest($eml), $eml);
356 append_mid($eml, $mid0);
361 sub clean_tree_v2 ($$$) {
362 my ($self, $w, $keep) = @_;
363 my $tree = $self->{-tree} or return; #v2 only
364 delete $tree->{$keep};
365 foreach (keys %$tree) {
366 print $w "D $_\n" or wfail;
368 %$tree = ($keep => 1);
371 # returns undef on duplicate
372 # returns the :MARK of the most recent commit
374 my ($self, $mime, $check_cb, $smsg) = @_;
376 my ($author, $at, $ct, $subject) = extract_cmt_info($mime, $smsg);
377 my $path_type = $self->{path_type};
379 if ($path_type eq '2/38') {
380 $path = mid2path(v1_mid0($mime));
381 } else { # v2 layout, one file:
385 my ($r, $w) = $self->gfi_start;
386 my $tip = $self->{tip};
387 if ($path_type eq '2/38') {
388 _check_path($r, $w, $tip, $path) and return;
391 drop_unwanted_headers($mime);
395 $mime = $check_cb->($mime, $self->{ibx}) or return;
398 my $blob = $self->{mark}++;
399 my $raw_email = $mime->{-public_inbox_raw} // $mime->as_string;
400 my $n = length($raw_email);
401 $self->{bytes_added} += $n;
402 print $w "blob\nmark :$blob\ndata ", $n, "\n" or wfail;
403 print $w $raw_email, "\n" or wfail;
405 # v2: we need this for Xapian
407 $smsg->{blob} = $self->get_mark(":$blob");
408 $smsg->{raw_bytes} = $n;
409 $smsg->{-raw_email} = \$raw_email;
411 my $ref = $self->{ref};
412 my $commit = $self->{mark}++;
413 my $parent = $tip =~ /\A:/ ? $tip : undef;
416 print $w "reset $ref\n" or wfail;
419 print $w "commit $ref\nmark :$commit\n",
420 "author $author $at\n",
421 "committer $self->{ident} $ct\n" or wfail;
422 print $w "data ", (length($subject) + 1), "\n",
423 $subject, "\n\n" or wfail;
425 print $w 'from ', ($parent ? $parent : $tip), "\n" or wfail;
427 clean_tree_v2($self, $w, $path);
428 print $w "M 100644 :$blob $path\n\n" or wfail;
430 $self->{tip} = ":$commit";
434 my ($cmd, $env, $rdr) = @_;
435 my $pid = spawn($cmd, $env, $rdr);
436 waitpid($pid, 0) == $pid or die join(' ', @$cmd) .' did not finish';
437 $? == 0 or die join(' ', @$cmd) . " failed: $?\n";
440 my @INIT_FILES = ('HEAD' => "ref: refs/heads/master\n",
441 'description' => <<EOD,
442 Unnamed repository; edit this file 'description' to name the repository.
446 repositoryFormatVersion = 0
454 my ($dir) = @_; # or self
455 $dir = $dir->{git}->{git_dir} if ref($dir);
457 File::Path::mkpath([ map { "$dir/$_" } qw(objects/info refs/heads) ]);
458 for (my $i = 0; $i < @INIT_FILES; $i++) {
459 my $f = $dir.'/'.$INIT_FILES[$i++];
461 open my $fh, '>', $f or die "open $f: $!";
462 print $fh $INIT_FILES[$i] or die "print $f: $!";
463 close $fh or die "close $f: $!";
467 # true if locked and active
468 sub active { !!$_[0]->{out} }
472 my $w = delete $self->{out} or return;
474 my $r = delete $self->{in} or die 'BUG: missing {in} when done';
475 print $w "done\n" or wfail;
476 my $pid = delete $self->{pid} or
477 die 'BUG: missing {pid} when done';
478 waitpid($pid, 0) == $pid or die 'fast-import did not finish';
479 $? == 0 or die "fast-import failed: $?";
482 my $nchg = delete $self->{nchg};
483 if ($nchg && !$wait_err) {
484 eval { _update_git_info($self, 1) };
485 warn "E: $self->{git}->{git_dir} update info: $@\n" if $@;
487 $self->lock_release(!!$nchg);
488 $self->{git}->cleanup;
489 die $wait_err if $wait_err;
494 foreach my $f (qw(in out)) {
495 next unless defined($self->{$f});
496 close $self->{$f} or die "failed to close import[$f]: $!\n";
500 sub digest2mid ($$) {
501 my ($dig, $hdr) = @_;
502 my $b64 = $dig->clone->b64digest;
503 # Make our own URLs nicer:
504 # See "Base 64 Encoding with URL and Filename Safe Alphabet" in RFC4648
507 # Add a date prefix to prevent a leading '-' in case that trips
508 # up some tools (e.g. if a Message-ID were a expected as a
510 my $dt = msg_datestamp($hdr);
511 $dt = POSIX::strftime('%Y%m%d%H%M%S', gmtime($dt));
515 sub rewrite_commit ($$$$) {
516 my ($self, $oids, $buf, $mime) = @_;
517 my ($author, $at, $ct, $subject);
519 ($author, $at, $ct, $subject) = extract_cmt_info($mime);
522 $subject = 'purged '.join(' ', @$oids);
526 foreach my $i (0..$#$buf) {
528 if ($l =~ /^author .* ([0-9]+ [\+-]?[0-9]+)$/) {
530 $buf->[$i] = "author $author $at\n";
531 } elsif ($l =~ /^committer .* ([0-9]+ [\+-]?[0-9]+)$/) {
533 $buf->[$i] = "committer $self->{ident} $ct\n";
534 } elsif ($l =~ /^data ([0-9]+)/) {
535 $buf->[$i++] = "data " . length($subject) . "\n";
536 $buf->[$i] = $subject;
542 # returns the new commit OID if a replacement was done
543 # returns undef if nothing was done
545 my ($self, $mime, $replace_map) = @_; # oid => raw string
546 my $tmp = "refs/heads/replace-".((keys %$replace_map)[0]);
547 my $old = $self->{'ref'};
548 my $git = $self->{git};
549 my @export = (qw(fast-export --no-data --use-done-feature), $old);
550 my $rd = $git->popen(@export);
551 my ($r, $w) = $self->gfi_start;
556 my $tree = $self->{-tree};
558 if (/^reset (?:.+)/) {
559 push @buf, "reset $tmp\n";
560 } elsif (/^commit (?:.+)/) {
562 print $w @buf or wfail;
565 push @buf, "commit $tmp\n";
566 } elsif (/^data ([0-9]+)/) {
567 # only commit message, so $len is small:
568 my $len = $1; # + 1 for trailing "\n"
570 my $n = read($rd, my $buf, $len) or die "read: $!";
571 $len == $n or die "short read ($n < $len)";
573 } elsif (/^M 100644 ([a-f0-9]+) (\w+)/) {
574 my ($oid, $path) = ($1, $2);
576 my $sref = $replace_map->{$oid};
579 my $n = length($$sref);
580 push @buf, "M 100644 inline $path\ndata $n\n";
581 push @buf, $$sref; # hope CoW works...
586 } elsif (/^D (\w+)/) {
588 push @buf, $_ if $tree->{$path};
589 } elsif ($_ eq "\n") {
592 my $out = join('', @buf);
594 warn "purge rewriting\n", $out, "\n";
596 rewrite_commit($self, \@oids, \@buf, $mime);
599 print $w @buf, "\n" or wfail;
601 } elsif ($_ eq "done\n") {
603 } elsif (/^mark :([0-9]+)$/) {
610 close $rd or die "close fast-export failed: $?";
612 print $w @buf or wfail;
614 die 'done\n not seen from fast-export' unless $done;
615 chomp(my $cmt = $self->get_mark(":$mark")) if $nreplace;
616 $self->{nchg} = 0; # prevent _update_git_info until update-ref:
618 my @git = ('git', "--git-dir=$git->{git_dir}");
620 run_die([@git, qw(update-ref), $old, $tmp]) if $nreplace;
622 run_die([@git, qw(update-ref -d), $tmp]);
624 return if $nreplace == 0;
626 run_die([@git, qw(-c gc.reflogExpire=now gc --prune=all --quiet)]);
628 # check that old OIDs are gone
630 foreach my $oid (keys %$replace_map) {
631 my @info = $git->check($oid);
633 warn "$oid not replaced\n";
637 _update_git_info($self, 0);
638 die "Failed to replace $err object(s)\n" if $err;
648 PublicInbox::Import - message importer for public-inbox v1 inboxes
656 use PublicInbox::Eml;
657 # PublicInbox::Eml exists as of public-inbox 1.5.0,
658 # Email::MIME was used in older versions
660 use PublicInbox::Git;
661 use PublicInbox::Import;
663 chomp(my $git_dir = `git rev-parse --git-dir`);
664 $git_dir or die "GIT_DIR= must be specified\n";
665 my $git = PublicInbox::Git->new($git_dir);
666 my @committer = ('inbox', 'inbox@example.org');
667 my $im = PublicInbox::Import->new($git, @committer);
670 my $message = "From: <u\@example.org>\n".
671 "Subject: test message \n" .
672 "Date: Thu, 01 Jan 1970 00:00:00 +0000\n" .
673 "Message-ID: <m\@example.org>\n".
675 my $parsed = PublicInbox::Eml->new($message);
676 my $ret = $im->add($parsed);
678 warn "duplicate: ", $parsed->header_raw('Message-ID'), "\n";
680 print "imported at mark $ret\n";
684 # to remove a message
685 my $junk = PublicInbox::Eml->new($message);
686 my ($mark, $orig) = $im->remove($junk);
687 if ($mark eq 'MISSING') {
689 } elsif ($mark eq 'MISMATCH') {
690 print "Message exists but does not match\n\n",
691 $orig->as_string, "\n",;
693 print "removed at mark $mark\n\n",
694 $orig->as_string, "\n";
700 An importer and remover for public-inboxes which takes C<PublicInbox::Eml>
701 or L<Email::MIME> messages as input and stores them in a git repository as
702 documented in L<https://public-inbox.org/public-inbox-v1-format.txt>,
703 except it does not allow duplicate Message-IDs.
705 It requires L<git(1)> and L<git-fast-import(1)> to be installed.
713 my $im = PublicInbox::Import->new($git, @committer);
715 Initialize a new PublicInbox::Import object.
719 my $parsed = PublicInbox::Eml->new($message);
722 Adds a message to to the git repository. This will acquire
723 C<$GIT_DIR/ssoma.lock> and start L<git-fast-import(1)> if necessary.
725 Messages added will not be visible to other processes until L</done>
726 is called, but L</remove> may be called on them.
730 my $junk = PublicInbox::Eml->new($message);
731 my ($code, $orig) = $im->remove($junk);
733 Removes a message from the repository. On success, it returns
734 a ':'-prefixed numeric code representing the git-fast-import
735 mark and the original messages as a PublicInbox::Eml
736 (or Email::MIME) object.
737 If the message could not be found, the code is "MISSING"
738 and the original message is undef. If there is a mismatch where
739 the "Message-ID" is matched but the subject and body do not match,
740 the returned code is "MISMATCH" and the conflicting message
745 Finalizes the L<git-fast-import(1)> and unlocks the repository.
746 Calling this is required to finalize changes to a repository.
754 All feedback welcome via plain-text mail to L<mailto:meta@public-inbox.org>
756 The mail archives are hosted at L<https://public-inbox.org/meta/>
760 Copyright (C) 2016-2020 all contributors L<mailto:meta@public-inbox.org>
762 License: AGPL-3.0+ L<http://www.gnu.org/licenses/agpl-3.0.txt>