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 $self->{pid};
52 pipe($out_r, $out_w) or die "pipe failed: $!";
53 my $git = $self->{git};
58 my $ref = $self->{ref};
59 chomp($self->{tip} = $git->qx(qw(rev-parse --revs-only), $ref));
60 if ($self->{path_type} ne '2/38' && $self->{tip}) {
62 my @tree = $git->qx(qw(ls-tree -r -z --name-only), $ref);
64 $self->{-tree} = { map { $_ => 1 } @tree };
67 my $git_dir = $git->{git_dir};
68 my @cmd = ('git', "--git-dir=$git_dir", qw(fast-import
69 --quiet --done --date-format=raw));
70 my ($in_r, $pid) = popen_rd(\@cmd, undef, { 0 => $out_r });
73 $self->{out} = $out_w;
79 sub wfail () { die "write to fast-import failed: $!" }
81 sub now_raw () { time . ' +0000' }
85 my $b = $mime->body_raw;
86 $b =~ s/(\r?\n)+\z//s;
90 # only used for v1 (ssoma) inboxes
91 sub _check_path ($$$$) {
92 my ($r, $w, $tip, $path) = @_;
94 print $w "ls $tip $path\n" or wfail;
96 defined(my $info = <$r>) or die "EOF from fast-import: $!";
97 $info =~ /\Amissing / ? undef : $info;
100 sub _cat_blob ($$$) {
101 my ($r, $w, $oid) = @_;
102 print $w "cat-blob $oid\n" or wfail;
105 defined $info or die "EOF from fast-import / cat-blob: $!";
106 $info =~ /\A[a-f0-9]{40} blob ([0-9]+)\n\z/ or return;
112 $n = read($r, $buf, $left, $offset);
113 defined($n) or die "read cat-blob failed: $!";
114 $n == 0 and die 'fast-export (cat-blob) died';
118 $n = read($r, my $lf, 1);
119 defined($n) or die "read final byte of cat-blob failed: $!";
120 die "bad read on final byte: <$lf>" if $lf ne "\n";
122 # fixup some bugginess in old versions:
123 $buf =~ s/\A[\r\n]*From [^\r\n]*\r?\n//s;
128 my ($self, $oid) = @_;
129 my ($r, $w) = $self->gfi_start;
130 _cat_blob($r, $w, $oid);
133 sub check_remove_v1 {
134 my ($r, $w, $tip, $path, $mime) = @_;
136 my $info = _check_path($r, $w, $tip, $path) or return ('MISSING',undef);
137 $info =~ m!\A100644 blob ([a-f0-9]{40})\t!s or die "not blob: $info";
139 my $msg = _cat_blob($r, $w, $oid) or die "BUG: cat-blob $1 failed";
140 my $cur = PublicInbox::Eml->new($msg);
141 my $cur_s = $cur->header('Subject');
142 $cur_s = '' unless defined $cur_s;
143 my $cur_m = $mime->header('Subject');
144 $cur_m = '' unless defined $cur_m;
145 if ($cur_s ne $cur_m || norm_body($cur) ne norm_body($mime)) {
146 return ('MISMATCH', $cur);
153 return unless $self->{pid};
154 print { $self->{out} } "checkpoint\n" or wfail;
159 my ($self, $msg) = @_;
160 return unless $self->{pid};
161 print { $self->{out} } "progress $msg\n" or wfail;
162 readline($self->{in}) eq "progress $msg\n" or die
163 "progress $msg not received\n";
167 sub _update_git_info ($$) {
168 my ($self, $do_gc) = @_;
169 # for compatibility with existing ssoma installations
170 # we can probably remove this entirely by 2020
171 my $git_dir = $self->{git}->{git_dir};
172 my @cmd = ('git', "--git-dir=$git_dir");
173 my $index = "$git_dir/ssoma.index";
174 if (-e $index && !$ENV{FAST}) {
175 my $env = { GIT_INDEX_FILE => $index };
176 run_die([@cmd, qw(read-tree -m -v -i), $self->{ref}], $env);
178 run_die([@cmd, 'update-server-info']);
179 my $ibx = $self->{ibx};
180 ($ibx && $self->{path_type} eq '2/38') and eval {
181 require PublicInbox::SearchIdx;
182 my $s = PublicInbox::SearchIdx->new($ibx);
183 $s->index_sync({ ref => $self->{ref} });
185 eval { run_die([@cmd, qw(gc --auto)]) } if $do_gc;
191 # For safety, we ensure git checkpoint is complete before because
192 # the data in git is still more important than what is in Xapian
193 # in v2. Performance may be gained by delaying the ->progress
194 # call but we lose safety
197 $self->progress('checkpoint');
198 _update_git_info($self, 0);
205 my ($self, $mark) = @_;
206 die "not active\n" unless $self->{pid};
207 my ($r, $w) = $self->gfi_start;
208 print $w "get-mark $mark\n" or wfail;
209 defined(my $oid = <$r>) or die "get-mark failed, need git 2.6.0+\n";
214 # returns undef on non-existent
215 # ('MISMATCH', PublicInbox::Eml) on mismatch
216 # (:MARK, PublicInbox::Eml) on success
218 # v2 callers should check with Xapian before calling this as
219 # it is not idempotent.
221 my ($self, $mime, $msg) = @_; # mime = PublicInbox::Eml or Email::MIME
223 my $path_type = $self->{path_type};
224 my ($path, $err, $cur, $blob);
226 my ($r, $w) = $self->gfi_start;
227 my $tip = $self->{tip};
228 if ($path_type eq '2/38') {
229 $path = mid2path(v1_mid0($mime));
230 ($err, $cur) = check_remove_v1($r, $w, $tip, $path, $mime);
231 return ($err, $cur) if $err;
234 if (ref($mime) eq 'SCALAR') { # optimization used by V2Writable
236 } else { # XXX should not be necessary:
237 my $str = $mime->as_string;
240 my $len = length($$sref);
241 $blob = $self->{mark}++;
242 print $w "blob\nmark :$blob\ndata $len\n",
243 $$sref, "\n" or wfail;
246 my $ref = $self->{ref};
247 my $commit = $self->{mark}++;
248 my $parent = $tip =~ /\A:/ ? $tip : undef;
250 print $w "reset $ref\n" or wfail;
252 my $ident = $self->{ident};
255 my $len = length($msg) + 1;
256 print $w "commit $ref\nmark :$commit\n",
257 "author $ident $now\n",
258 "committer $ident $now\n",
259 "data $len\n$msg\n\n",
260 'from ', ($parent ? $parent : $tip), "\n" or wfail;
262 print $w "D $path\n\n" or wfail;
264 clean_tree_v2($self, $w, 'd');
265 print $w "M 100644 :$blob d\n\n" or wfail;
268 (($self->{tip} = ":$commit"), $cur);
271 sub git_timestamp ($) {
272 my ($ts, $zone) = @{$_[0]};
273 $ts = 0 if $ts < 0; # git uses unsigned times
277 sub extract_cmt_info ($;$) {
278 my ($mime, $smsg) = @_;
279 # $mime is PublicInbox::Eml, but remains Email::MIME-compatible
280 $smsg //= bless {}, 'PublicInbox::Smsg';
282 my $hdr = $mime->header_obj;
283 $smsg->populate($hdr);
286 my $from = delete($smsg->{From}) // '';
287 my ($email) = PublicInbox::Address::emails($from);
288 my ($name) = PublicInbox::Address::names($from);
289 if (!defined($name) || !defined($email)) {
290 $sender = $hdr->header('Sender') // '';
291 $name //= (PublicInbox::Address::names($sender))[0];
292 $email //= (PublicInbox::Address::emails($sender))[0];
294 if (defined $email) {
295 # Email::Address::XS may leave quoted '<' in addresses,
296 # which git-fast-import doesn't like
299 # quiet down wide character warnings with utf8::encode
300 utf8::encode($email);
303 warn "no email in From: $from or Sender: $sender\n";
306 # git gets confused with:
307 # "'A U Thor <u@example.com>' via foo" <foo@example.com>
309 # <CAD0k6qSUYANxbjjbE4jTW4EeVwOYgBD=bXkSu=akiYC_CB7Ffw@mail.gmail.com>
315 warn "no name in From: $from or Sender: $sender\n";
318 my $subject = delete($smsg->{Subject}) // '(no subject)';
319 utf8::encode($subject);
320 my $at = git_timestamp(delete $smsg->{-ds});
321 my $ct = git_timestamp(delete $smsg->{-ts});
322 ("$name <$email>", $at, $ct, $subject);
325 # kill potentially confusing/misleading headers
326 sub drop_unwanted_headers ($) {
329 $mime->header_set($_) for qw(Bytes Lines Content-Length Status);
330 $mime->header_set($_) for @PublicInbox::MDA::BAD_HEADERS;
333 # used by V2Writable, too
334 sub append_mid ($$) {
335 my ($hdr, $mid0) = @_;
336 # @cur is likely empty if we need to call this sub, but it could
337 # have random unparseable crap which we'll preserve, too.
338 my @cur = $hdr->header_raw('Message-ID');
339 $hdr->header_set('Message-ID', @cur, "<$mid0>");
344 my $hdr = $mime->header_obj;
345 my $mids = mids($hdr);
347 if (!scalar(@$mids)) { # spam often has no Message-ID
348 my $mid0 = digest2mid(content_digest($mime), $hdr);
349 append_mid($hdr, $mid0);
354 sub clean_tree_v2 ($$$) {
355 my ($self, $w, $keep) = @_;
356 my $tree = $self->{-tree} or return; #v2 only
357 delete $tree->{$keep};
358 foreach (keys %$tree) {
359 print $w "D $_\n" or wfail;
361 %$tree = ($keep => 1);
364 # returns undef on duplicate
365 # returns the :MARK of the most recent commit
367 my ($self, $mime, $check_cb, $smsg) = @_;
369 my ($author, $at, $ct, $subject) = extract_cmt_info($mime, $smsg);
370 my $path_type = $self->{path_type};
372 if ($path_type eq '2/38') {
373 $path = mid2path(v1_mid0($mime));
374 } else { # v2 layout, one file:
378 my ($r, $w) = $self->gfi_start;
379 my $tip = $self->{tip};
380 if ($path_type eq '2/38') {
381 _check_path($r, $w, $tip, $path) and return;
384 drop_unwanted_headers($mime);
388 $mime = $check_cb->($mime, $self->{ibx}) or return;
391 my $blob = $self->{mark}++;
392 my $raw_email = $mime->{-public_inbox_raw} // $mime->as_string;
393 my $n = length($raw_email);
394 $self->{bytes_added} += $n;
395 print $w "blob\nmark :$blob\ndata ", $n, "\n" or wfail;
396 print $w $raw_email, "\n" or wfail;
398 # v2: we need this for Xapian
400 $smsg->{blob} = $self->get_mark(":$blob");
401 $smsg->{raw_bytes} = $n;
402 $smsg->{-raw_email} = \$raw_email;
404 my $ref = $self->{ref};
405 my $commit = $self->{mark}++;
406 my $parent = $tip =~ /\A:/ ? $tip : undef;
409 print $w "reset $ref\n" or wfail;
412 print $w "commit $ref\nmark :$commit\n",
413 "author $author $at\n",
414 "committer $self->{ident} $ct\n" or wfail;
415 print $w "data ", (length($subject) + 1), "\n",
416 $subject, "\n\n" or wfail;
418 print $w 'from ', ($parent ? $parent : $tip), "\n" or wfail;
420 clean_tree_v2($self, $w, $path);
421 print $w "M 100644 :$blob $path\n\n" or wfail;
423 $self->{tip} = ":$commit";
427 my ($cmd, $env, $rdr) = @_;
428 my $pid = spawn($cmd, $env, $rdr);
429 waitpid($pid, 0) == $pid or die join(' ', @$cmd) .' did not finish';
430 $? == 0 or die join(' ', @$cmd) . " failed: $?\n";
433 my @INIT_FILES = ('HEAD' => "ref: refs/heads/master\n",
434 'description' => <<EOD,
435 Unnamed repository; edit this file 'description' to name the repository.
439 repositoryFormatVersion = 0
447 my ($dir) = @_; # or self
448 $dir = $dir->{git}->{git_dir} if ref($dir);
450 File::Path::mkpath([ map { "$dir/$_" } qw(objects/info refs/heads) ]);
451 for (my $i = 0; $i < @INIT_FILES; $i++) {
452 my $f = $dir.'/'.$INIT_FILES[$i++];
454 open my $fh, '>', $f or die "open $f: $!";
455 print $fh $INIT_FILES[$i] or die "print $f: $!";
456 close $fh or die "close $f: $!";
462 my $w = delete $self->{out} or return;
463 my $r = delete $self->{in} or die 'BUG: missing {in} when done';
464 print $w "done\n" or wfail;
465 my $pid = delete $self->{pid} or die 'BUG: missing {pid} when done';
466 waitpid($pid, 0) == $pid or die 'fast-import did not finish';
467 $? == 0 or die "fast-import failed: $?";
469 my $nchg = delete $self->{nchg};
470 _update_git_info($self, 1) if $nchg;
471 $self->lock_release(!!$nchg);
473 $self->{git}->cleanup;
478 foreach my $f (qw(in out)) {
479 next unless defined($self->{$f});
480 close $self->{$f} or die "failed to close import[$f]: $!\n";
484 sub digest2mid ($$) {
485 my ($dig, $hdr) = @_;
486 my $b64 = $dig->clone->b64digest;
487 # Make our own URLs nicer:
488 # See "Base 64 Encoding with URL and Filename Safe Alphabet" in RFC4648
491 # Add a date prefix to prevent a leading '-' in case that trips
492 # up some tools (e.g. if a Message-ID were a expected as a
494 my $dt = msg_datestamp($hdr);
495 $dt = POSIX::strftime('%Y%m%d%H%M%S', gmtime($dt));
499 sub rewrite_commit ($$$$) {
500 my ($self, $oids, $buf, $mime) = @_;
501 my ($author, $at, $ct, $subject);
503 ($author, $at, $ct, $subject) = extract_cmt_info($mime);
506 $subject = 'purged '.join(' ', @$oids);
510 foreach my $i (0..$#$buf) {
512 if ($l =~ /^author .* ([0-9]+ [\+-]?[0-9]+)$/) {
514 $buf->[$i] = "author $author $at\n";
515 } elsif ($l =~ /^committer .* ([0-9]+ [\+-]?[0-9]+)$/) {
517 $buf->[$i] = "committer $self->{ident} $ct\n";
518 } elsif ($l =~ /^data ([0-9]+)/) {
519 $buf->[$i++] = "data " . length($subject) . "\n";
520 $buf->[$i] = $subject;
526 # returns the new commit OID if a replacement was done
527 # returns undef if nothing was done
529 my ($self, $mime, $replace_map) = @_; # oid => raw string
530 my $tmp = "refs/heads/replace-".((keys %$replace_map)[0]);
531 my $old = $self->{'ref'};
532 my $git = $self->{git};
533 my @export = (qw(fast-export --no-data --use-done-feature), $old);
534 my $rd = $git->popen(@export);
535 my ($r, $w) = $self->gfi_start;
540 my $tree = $self->{-tree};
542 if (/^reset (?:.+)/) {
543 push @buf, "reset $tmp\n";
544 } elsif (/^commit (?:.+)/) {
546 print $w @buf or wfail;
549 push @buf, "commit $tmp\n";
550 } elsif (/^data ([0-9]+)/) {
551 # only commit message, so $len is small:
552 my $len = $1; # + 1 for trailing "\n"
554 my $n = read($rd, my $buf, $len) or die "read: $!";
555 $len == $n or die "short read ($n < $len)";
557 } elsif (/^M 100644 ([a-f0-9]+) (\w+)/) {
558 my ($oid, $path) = ($1, $2);
560 my $sref = $replace_map->{$oid};
563 my $n = length($$sref);
564 push @buf, "M 100644 inline $path\ndata $n\n";
565 push @buf, $$sref; # hope CoW works...
570 } elsif (/^D (\w+)/) {
572 push @buf, $_ if $tree->{$path};
573 } elsif ($_ eq "\n") {
576 my $out = join('', @buf);
578 warn "purge rewriting\n", $out, "\n";
580 rewrite_commit($self, \@oids, \@buf, $mime);
583 print $w @buf, "\n" or wfail;
585 } elsif ($_ eq "done\n") {
587 } elsif (/^mark :([0-9]+)$/) {
594 close $rd or die "close fast-export failed: $?";
596 print $w @buf or wfail;
598 die 'done\n not seen from fast-export' unless $done;
599 chomp(my $cmt = $self->get_mark(":$mark")) if $nreplace;
600 $self->{nchg} = 0; # prevent _update_git_info until update-ref:
602 my @git = ('git', "--git-dir=$git->{git_dir}");
604 run_die([@git, qw(update-ref), $old, $tmp]) if $nreplace;
606 run_die([@git, qw(update-ref -d), $tmp]);
608 return if $nreplace == 0;
610 run_die([@git, qw(-c gc.reflogExpire=now gc --prune=all --quiet)]);
612 # check that old OIDs are gone
614 foreach my $oid (keys %$replace_map) {
615 my @info = $git->check($oid);
617 warn "$oid not replaced\n";
621 _update_git_info($self, 0);
622 die "Failed to replace $err object(s)\n" if $err;
632 PublicInbox::Import - message importer for public-inbox v1 inboxes
640 use PublicInbox::Eml;
641 # PublicInbox::Eml exists as of public-inbox 1.5.0,
642 # Email::MIME was used in older versions
644 use PublicInbox::Git;
645 use PublicInbox::Import;
647 chomp(my $git_dir = `git rev-parse --git-dir`);
648 $git_dir or die "GIT_DIR= must be specified\n";
649 my $git = PublicInbox::Git->new($git_dir);
650 my @committer = ('inbox', 'inbox@example.org');
651 my $im = PublicInbox::Import->new($git, @committer);
654 my $message = "From: <u\@example.org>\n".
655 "Subject: test message \n" .
656 "Date: Thu, 01 Jan 1970 00:00:00 +0000\n" .
657 "Message-ID: <m\@example.org>\n".
659 my $parsed = PublicInbox::Eml->new($message);
660 my $ret = $im->add($parsed);
663 $parsed->header_obj->header_raw('Message-ID'), "\n";
665 print "imported at mark $ret\n";
669 # to remove a message
670 my $junk = PublicInbox::Eml->new($message);
671 my ($mark, $orig) = $im->remove($junk);
672 if ($mark eq 'MISSING') {
674 } elsif ($mark eq 'MISMATCH') {
675 print "Message exists but does not match\n\n",
676 $orig->as_string, "\n",;
678 print "removed at mark $mark\n\n",
679 $orig->as_string, "\n";
685 An importer and remover for public-inboxes which takes C<PublicInbox::Eml>
686 or L<Email::MIME> messages as input and stores them in a git repository as
687 documented in L<https://public-inbox.org/public-inbox-v1-format.txt>,
688 except it does not allow duplicate Message-IDs.
690 It requires L<git(1)> and L<git-fast-import(1)> to be installed.
698 my $im = PublicInbox::Import->new($git, @committer);
700 Initialize a new PublicInbox::Import object.
704 my $parsed = PublicInbox::Eml->new($message);
707 Adds a message to to the git repository. This will acquire
708 C<$GIT_DIR/ssoma.lock> and start L<git-fast-import(1)> if necessary.
710 Messages added will not be visible to other processes until L</done>
711 is called, but L</remove> may be called on them.
715 my $junk = PublicInbox::Eml->new($message);
716 my ($code, $orig) = $im->remove($junk);
718 Removes a message from the repository. On success, it returns
719 a ':'-prefixed numeric code representing the git-fast-import
720 mark and the original messages as a PublicInbox::Eml
721 (or Email::MIME) object.
722 If the message could not be found, the code is "MISSING"
723 and the original message is undef. If there is a mismatch where
724 the "Message-ID" is matched but the subject and body do not match,
725 the returned code is "MISMATCH" and the conflicting message
730 Finalizes the L<git-fast-import(1)> and unlocks the repository.
731 Calling this is required to finalize changes to a repository.
739 All feedback welcome via plain-text mail to L<mailto:meta@public-inbox.org>
741 The mail archives are hosted at L<https://public-inbox.org/meta/>
745 Copyright (C) 2016-2020 all contributors L<mailto:meta@public-inbox.org>
747 License: AGPL-3.0+ L<http://www.gnu.org/licenses/agpl-3.0.txt>