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);
22 sub default_branch () {
23 state $default_branch = do {
24 my $r = popen_rd([qw(git config --global init.defaultBranch)]);
25 chomp(my $h = <$r> // '');
26 $h eq '' ? 'refs/heads/master' : $h;
31 # we can't change arg order, this is documented in POD
32 # and external projects may rely on it:
33 my ($class, $git, $name, $email, $ibx) = @_;
36 $ref = $ibx->{ref_head};
37 $name //= $ibx->{name};
38 $email //= $ibx->{-primary_address};
43 ident => "$name <$email>",
45 ref => $ref // default_branch,
47 path_type => '2/38', # or 'v2'
48 lock_path => "$git->{git_dir}/ssoma.lock", # v2 changes this
53 # idempotent start function
57 return ($self->{in}, $self->{out}) if $self->{pid};
59 my ($in_r, $pid, $out_r, $out_w);
60 pipe($out_r, $out_w) or die "pipe failed: $!";
64 my ($git, $ref) = @$self{qw(git ref)};
66 chomp($self->{tip} = $git->qx(qw(rev-parse --revs-only), $ref));
67 die "fatal: rev-parse --revs-only $ref: \$?=$?" if $?;
68 if ($self->{path_type} ne '2/38' && $self->{tip}) {
70 my @t = $git->qx(qw(ls-tree -r -z --name-only), $ref);
71 die "fatal: ls-tree -r -z --name-only $ref: \$?=$?" if $?;
73 $self->{-tree} = { map { $_ => 1 } @t };
75 my @cmd = ('git', "--git-dir=$git->{git_dir}",
76 qw(fast-import --quiet --done --date-format=raw));
77 ($in_r, $pid) = popen_rd(\@cmd, undef, { 0 => $out_r });
80 $self->{out} = $out_w;
91 sub wfail () { die "write to fast-import failed: $!" }
93 sub now_raw () { time . ' +0000' }
97 my $b = $mime->body_raw;
98 $b =~ s/(\r?\n)+\z//s;
102 # only used for v1 (ssoma) inboxes
103 sub _check_path ($$$$) {
104 my ($r, $w, $tip, $path) = @_;
105 return if $tip eq '';
106 print $w "ls $tip $path\n" or wfail;
108 defined(my $info = <$r>) or die "EOF from fast-import: $!";
109 $info =~ /\Amissing / ? undef : $info;
112 sub _cat_blob ($$$) {
113 my ($r, $w, $oid) = @_;
114 print $w "cat-blob $oid\n" or wfail;
117 defined $info or die "EOF from fast-import / cat-blob: $!";
118 $info =~ /\A[a-f0-9]{40,} blob ([0-9]+)\n\z/ or return;
124 $n = read($r, $buf, $left, $offset);
125 defined($n) or die "read cat-blob failed: $!";
126 $n == 0 and die 'fast-export (cat-blob) died';
130 $n = read($r, my $lf, 1);
131 defined($n) or die "read final byte of cat-blob failed: $!";
132 die "bad read on final byte: <$lf>" if $lf ne "\n";
134 # fixup some bugginess in old versions:
135 $buf =~ s/\A[\r\n]*From [^\r\n]*\r?\n//s;
140 my ($self, $oid) = @_;
141 my ($r, $w) = $self->gfi_start;
142 _cat_blob($r, $w, $oid);
145 sub check_remove_v1 {
146 my ($r, $w, $tip, $path, $mime) = @_;
148 my $info = _check_path($r, $w, $tip, $path) or return ('MISSING',undef);
149 $info =~ m!\A100644 blob ([a-f0-9]{40,})\t!s or die "not blob: $info";
151 my $msg = _cat_blob($r, $w, $oid) or die "BUG: cat-blob $1 failed";
152 my $cur = PublicInbox::Eml->new($msg);
153 my $cur_s = $cur->header('Subject');
154 $cur_s = '' unless defined $cur_s;
155 my $cur_m = $mime->header('Subject');
156 $cur_m = '' unless defined $cur_m;
157 if ($cur_s ne $cur_m || norm_body($cur) ne norm_body($mime)) {
158 return ('MISMATCH', $cur);
165 return unless $self->{pid};
166 print { $self->{out} } "checkpoint\n" or wfail;
171 my ($self, $msg) = @_;
172 return unless $self->{pid};
173 print { $self->{out} } "progress $msg\n" or wfail;
174 readline($self->{in}) eq "progress $msg\n" or die
175 "progress $msg not received\n";
179 sub _update_git_info ($$) {
180 my ($self, $do_gc) = @_;
181 # for compatibility with existing ssoma installations
182 # we can probably remove this entirely by 2020
183 my $git_dir = $self->{git}->{git_dir};
184 my @cmd = ('git', "--git-dir=$git_dir");
185 my $index = "$git_dir/ssoma.index";
186 if (-e $index && !$ENV{FAST}) {
187 my $env = { GIT_INDEX_FILE => $index };
188 run_die([@cmd, qw(read-tree -m -v -i), $self->{ref}], $env);
190 eval { run_die([@cmd, 'update-server-info']) };
191 my $ibx = $self->{ibx};
192 if ($ibx && $ibx->version == 1 && -d "$ibx->{inboxdir}/public-inbox" &&
193 eval { require PublicInbox::SearchIdx }) {
195 my $s = PublicInbox::SearchIdx->new($ibx);
196 $s->index_sync({ ref => $self->{ref} });
198 warn "$ibx->{inboxdir} index failed: $@\n" if $@;
200 eval { run_die([@cmd, qw(gc --auto)]) } if $do_gc;
206 # For safety, we ensure git checkpoint is complete before because
207 # the data in git is still more important than what is in Xapian
208 # in v2. Performance may be gained by delaying the ->progress
209 # call but we lose safety
212 $self->progress('checkpoint');
213 _update_git_info($self, 0);
220 my ($self, $mark) = @_;
221 die "not active\n" unless $self->{pid};
222 my ($r, $w) = $self->gfi_start;
223 print $w "get-mark $mark\n" or wfail;
224 defined(my $oid = <$r>) or die "get-mark failed, need git 2.6.0+\n";
229 # returns undef on non-existent
230 # ('MISMATCH', PublicInbox::Eml) on mismatch
231 # (:MARK, PublicInbox::Eml) on success
233 # v2 callers should check with Xapian before calling this as
234 # it is not idempotent.
236 my ($self, $mime, $msg) = @_; # mime = PublicInbox::Eml or Email::MIME
238 my $path_type = $self->{path_type};
239 my ($path, $err, $cur, $blob);
241 my ($r, $w) = $self->gfi_start;
242 my $tip = $self->{tip};
243 if ($path_type eq '2/38') {
244 $path = mid2path(v1_mid0($mime));
245 ($err, $cur) = check_remove_v1($r, $w, $tip, $path, $mime);
246 return ($err, $cur) if $err;
249 if (ref($mime) eq 'SCALAR') { # optimization used by V2Writable
251 } else { # XXX should not be necessary:
252 my $str = $mime->as_string;
255 my $len = length($$sref);
256 $blob = $self->{mark}++;
257 print $w "blob\nmark :$blob\ndata $len\n",
258 $$sref, "\n" or wfail;
261 my $ref = $self->{ref};
262 my $commit = $self->{mark}++;
263 my $parent = $tip =~ /\A:/ ? $tip : undef;
265 print $w "reset $ref\n" or wfail;
267 my $ident = $self->{ident};
270 my $len = length($msg) + 1;
271 print $w "commit $ref\nmark :$commit\n",
272 "author $ident $now\n",
273 "committer $ident $now\n",
274 "data $len\n$msg\n\n",
275 'from ', ($parent ? $parent : $tip), "\n" or wfail;
277 print $w "D $path\n\n" or wfail;
279 clean_tree_v2($self, $w, 'd');
280 print $w "M 100644 :$blob d\n\n" or wfail;
283 (($self->{tip} = ":$commit"), $cur);
286 sub git_timestamp ($) {
287 my ($ts, $zone) = @{$_[0]};
288 $ts = 0 if $ts < 0; # git uses unsigned times
292 sub extract_cmt_info ($;$) {
293 my ($mime, $smsg) = @_;
294 # $mime is PublicInbox::Eml, but remains Email::MIME-compatible
295 $smsg //= bless {}, 'PublicInbox::Smsg';
297 $smsg->populate($mime);
300 my $from = delete($smsg->{From}) // '';
301 my ($email) = PublicInbox::Address::emails($from);
302 my ($name) = PublicInbox::Address::names($from);
303 if (!defined($name) || !defined($email)) {
304 $sender = $mime->header('Sender') // '';
305 $name //= (PublicInbox::Address::names($sender))[0];
306 $email //= (PublicInbox::Address::emails($sender))[0];
308 if (defined $email) {
309 # Email::Address::XS may leave quoted '<' in addresses,
310 # which git-fast-import doesn't like
313 # quiet down wide character warnings with utf8::encode
314 utf8::encode($email);
317 warn "no email in From: $from or Sender: $sender\n";
320 # git gets confused with:
321 # "'A U Thor <u@example.com>' via foo" <foo@example.com>
323 # <CAD0k6qSUYANxbjjbE4jTW4EeVwOYgBD=bXkSu=akiYC_CB7Ffw@mail.gmail.com>
329 warn "no name in From: $from or Sender: $sender\n";
332 my $subject = delete($smsg->{Subject}) // '(no subject)';
333 utf8::encode($subject);
334 my $at = git_timestamp(delete $smsg->{-ds});
335 my $ct = git_timestamp(delete $smsg->{-ts});
336 ("$name <$email>", $at, $ct, $subject);
339 # kill potentially confusing/misleading headers
340 our @UNWANTED_HEADERS = (qw(Bytes Lines Content-Length),
341 qw(Status X-Status));
342 sub drop_unwanted_headers ($) {
344 for (@UNWANTED_HEADERS, @PublicInbox::MDA::BAD_HEADERS) {
345 $eml->header_set($_);
349 # used by V2Writable, too
350 sub append_mid ($$) {
351 my ($hdr, $mid0) = @_;
352 # @cur is likely empty if we need to call this sub, but it could
353 # have random unparseable crap which we'll preserve, too.
354 my @cur = $hdr->header_raw('Message-ID');
355 $hdr->header_set('Message-ID', @cur, "<$mid0>");
360 my $mids = mids($eml);
362 if (!scalar(@$mids)) { # spam often has no Message-ID
363 my $mid0 = digest2mid(content_digest($eml), $eml);
364 append_mid($eml, $mid0);
369 sub clean_tree_v2 ($$$) {
370 my ($self, $w, $keep) = @_;
371 my $tree = $self->{-tree} or return; #v2 only
372 delete $tree->{$keep};
373 foreach (keys %$tree) {
374 print $w "D $_\n" or wfail;
376 %$tree = ($keep => 1);
379 # returns undef on duplicate
380 # returns the :MARK of the most recent commit
382 my ($self, $mime, $check_cb, $smsg) = @_;
384 my ($author, $at, $ct, $subject) = extract_cmt_info($mime, $smsg);
385 my $path_type = $self->{path_type};
387 if ($path_type eq '2/38') {
388 $path = mid2path(v1_mid0($mime));
389 } else { # v2 layout, one file:
393 my ($r, $w) = $self->gfi_start;
394 my $tip = $self->{tip};
395 if ($path_type eq '2/38') {
396 _check_path($r, $w, $tip, $path) and return;
399 drop_unwanted_headers($mime);
403 $mime = $check_cb->($mime, $self->{ibx}) or return;
406 my $blob = $self->{mark}++;
407 my $raw_email = $mime->{-public_inbox_raw} // $mime->as_string;
408 my $n = length($raw_email);
409 $self->{bytes_added} += $n;
410 print $w "blob\nmark :$blob\ndata ", $n, "\n" or wfail;
411 print $w $raw_email, "\n" or wfail;
413 # v2: we need this for Xapian
415 $smsg->{blob} = $self->get_mark(":$blob");
416 $smsg->{raw_bytes} = $n;
417 if (my $oidx = delete $smsg->{-oidx}) { # used by LeiStore
418 return if $oidx->blob_exists($smsg->{blob});
420 # XXX do we need this? it's in git at this point
421 $smsg->{-raw_email} = \$raw_email;
423 my $ref = $self->{ref};
424 my $commit = $self->{mark}++;
425 my $parent = $tip =~ /\A:/ ? $tip : undef;
428 print $w "reset $ref\n" or wfail;
431 print $w "commit $ref\nmark :$commit\n",
432 "author $author $at\n",
433 "committer $self->{ident} $ct\n" or wfail;
434 print $w "data ", (length($subject) + 1), "\n",
435 $subject, "\n\n" or wfail;
437 print $w 'from ', ($parent ? $parent : $tip), "\n" or wfail;
439 clean_tree_v2($self, $w, $path);
440 print $w "M 100644 :$blob $path\n\n" or wfail;
442 $self->{tip} = ":$commit";
446 my ($cmd, $env, $rdr) = @_;
447 my $pid = spawn($cmd, $env, $rdr);
448 waitpid($pid, 0) == $pid or die join(' ', @$cmd) .' did not finish';
449 $? == 0 or die join(' ', @$cmd) . " failed: $?\n";
452 my @INIT_FILES = ('HEAD' => undef, # filled in at runtime
453 'description' => <<EOD,
454 Unnamed repository; edit this file 'description' to name the repository.
458 repositoryFormatVersion = 0
466 my ($dir, $head) = @_; # or self
467 $dir = $dir->{git}->{git_dir} if ref($dir);
469 File::Path::mkpath([ map { "$dir/$_" } qw(objects/info refs/heads) ]);
470 $INIT_FILES[1] //= 'ref: '.default_branch."\n";
471 my @fn_contents = @INIT_FILES;
472 $fn_contents[1] = "ref: refs/heads/$head\n" if defined $head;
473 while (my ($fn, $contents) = splice(@fn_contents, 0, 2)) {
474 my $f = $dir.'/'.$fn;
476 open my $fh, '>', $f or die "open $f: $!";
477 print $fh $contents or die "print $f: $!";
478 close $fh or die "close $f: $!";
482 # true if locked and active
483 sub active { !!$_[0]->{out} }
487 my $w = delete $self->{out} or return;
489 my $r = delete $self->{in} or die 'BUG: missing {in} when done';
490 print $w "done\n" or wfail;
491 my $pid = delete $self->{pid} or
492 die 'BUG: missing {pid} when done';
493 waitpid($pid, 0) == $pid or die 'fast-import did not finish';
494 $? == 0 or die "fast-import failed: $?";
497 my $nchg = delete $self->{nchg};
498 if ($nchg && !$wait_err) {
499 eval { _update_git_info($self, 1) };
500 warn "E: $self->{git}->{git_dir} update info: $@\n" if $@;
502 $self->lock_release(!!$nchg);
503 $self->{git}->cleanup;
504 die $wait_err if $wait_err;
509 foreach my $f (qw(in out)) {
510 next unless defined($self->{$f});
511 close $self->{$f} or die "failed to close import[$f]: $!\n";
515 sub digest2mid ($$) {
516 my ($dig, $hdr) = @_;
517 my $b64 = $dig->clone->b64digest;
518 # Make our own URLs nicer:
519 # See "Base 64 Encoding with URL and Filename Safe Alphabet" in RFC4648
522 # Add a date prefix to prevent a leading '-' in case that trips
523 # up some tools (e.g. if a Message-ID were a expected as a
525 my $dt = msg_datestamp($hdr);
526 $dt = POSIX::strftime('%Y%m%d%H%M%S', gmtime($dt));
530 sub rewrite_commit ($$$$) {
531 my ($self, $oids, $buf, $mime) = @_;
532 my ($author, $at, $ct, $subject);
534 ($author, $at, $ct, $subject) = extract_cmt_info($mime);
537 $subject = 'purged '.join(' ', @$oids);
541 foreach my $i (0..$#$buf) {
543 if ($l =~ /^author .* ([0-9]+ [\+-]?[0-9]+)$/) {
545 $buf->[$i] = "author $author $at\n";
546 } elsif ($l =~ /^committer .* ([0-9]+ [\+-]?[0-9]+)$/) {
548 $buf->[$i] = "committer $self->{ident} $ct\n";
549 } elsif ($l =~ /^data ([0-9]+)/) {
550 $buf->[$i++] = "data " . length($subject) . "\n";
551 $buf->[$i] = $subject;
557 # returns the new commit OID if a replacement was done
558 # returns undef if nothing was done
560 my ($self, $mime, $replace_map) = @_; # oid => raw string
561 my $tmp = "refs/heads/replace-".((keys %$replace_map)[0]);
562 my $old = $self->{'ref'};
563 my $git = $self->{git};
564 my @export = (qw(fast-export --no-data --use-done-feature), $old);
565 my $rd = $git->popen(@export);
566 my ($r, $w) = $self->gfi_start;
571 my $tree = $self->{-tree};
573 if (/^reset (?:.+)/) {
574 push @buf, "reset $tmp\n";
575 } elsif (/^commit (?:.+)/) {
577 print $w @buf or wfail;
580 push @buf, "commit $tmp\n";
581 } elsif (/^data ([0-9]+)/) {
582 # only commit message, so $len is small:
583 my $len = $1; # + 1 for trailing "\n"
585 my $n = read($rd, my $buf, $len) or die "read: $!";
586 $len == $n or die "short read ($n < $len)";
588 } elsif (/^M 100644 ([a-f0-9]+) (\w+)/) {
589 my ($oid, $path) = ($1, $2);
591 my $sref = $replace_map->{$oid};
594 my $n = length($$sref);
595 push @buf, "M 100644 inline $path\ndata $n\n";
596 push @buf, $$sref; # hope CoW works...
601 } elsif (/^D (\w+)/) {
603 push @buf, $_ if $tree->{$path};
604 } elsif ($_ eq "\n") {
607 my $out = join('', @buf);
609 warn "purge rewriting\n", $out, "\n";
611 rewrite_commit($self, \@oids, \@buf, $mime);
614 print $w @buf, "\n" or wfail;
616 } elsif ($_ eq "done\n") {
618 } elsif (/^mark :([0-9]+)$/) {
625 close $rd or die "close fast-export failed: $?";
627 print $w @buf or wfail;
629 die 'done\n not seen from fast-export' unless $done;
630 chomp(my $cmt = $self->get_mark(":$mark")) if $nreplace;
631 $self->{nchg} = 0; # prevent _update_git_info until update-ref:
633 my @git = ('git', "--git-dir=$git->{git_dir}");
635 run_die([@git, qw(update-ref), $old, $tmp]) if $nreplace;
637 run_die([@git, qw(update-ref -d), $tmp]);
639 return if $nreplace == 0;
641 run_die([@git, qw(-c gc.reflogExpire=now gc --prune=all --quiet)]);
643 # check that old OIDs are gone
645 foreach my $oid (keys %$replace_map) {
646 my @info = $git->check($oid);
648 warn "$oid not replaced\n";
652 _update_git_info($self, 0);
653 die "Failed to replace $err object(s)\n" if $err;
663 PublicInbox::Import - message importer for public-inbox v1 inboxes
671 use PublicInbox::Eml;
672 # PublicInbox::Eml exists as of public-inbox 1.5.0,
673 # Email::MIME was used in older versions
675 use PublicInbox::Git;
676 use PublicInbox::Import;
678 chomp(my $git_dir = `git rev-parse --git-dir`);
679 $git_dir or die "GIT_DIR= must be specified\n";
680 my $git = PublicInbox::Git->new($git_dir);
681 my @committer = ('inbox', 'inbox@example.org');
682 my $im = PublicInbox::Import->new($git, @committer);
685 my $message = "From: <u\@example.org>\n".
686 "Subject: test message \n" .
687 "Date: Thu, 01 Jan 1970 00:00:00 +0000\n" .
688 "Message-ID: <m\@example.org>\n".
690 my $parsed = PublicInbox::Eml->new($message);
691 my $ret = $im->add($parsed);
693 warn "duplicate: ", $parsed->header_raw('Message-ID'), "\n";
695 print "imported at mark $ret\n";
699 # to remove a message
700 my $junk = PublicInbox::Eml->new($message);
701 my ($mark, $orig) = $im->remove($junk);
702 if ($mark eq 'MISSING') {
704 } elsif ($mark eq 'MISMATCH') {
705 print "Message exists but does not match\n\n",
706 $orig->as_string, "\n",;
708 print "removed at mark $mark\n\n",
709 $orig->as_string, "\n";
715 An importer and remover for public-inboxes which takes C<PublicInbox::Eml>
716 or L<Email::MIME> messages as input and stores them in a git repository as
717 documented in L<https://public-inbox.org/public-inbox-v1-format.txt>,
718 except it does not allow duplicate Message-IDs.
720 It requires L<git(1)> and L<git-fast-import(1)> to be installed.
728 my $im = PublicInbox::Import->new($git, @committer);
730 Initialize a new PublicInbox::Import object.
734 my $parsed = PublicInbox::Eml->new($message);
737 Adds a message to to the git repository. This will acquire
738 C<$GIT_DIR/ssoma.lock> and start L<git-fast-import(1)> if necessary.
740 Messages added will not be visible to other processes until L</done>
741 is called, but L</remove> may be called on them.
745 my $junk = PublicInbox::Eml->new($message);
746 my ($code, $orig) = $im->remove($junk);
748 Removes a message from the repository. On success, it returns
749 a ':'-prefixed numeric code representing the git-fast-import
750 mark and the original messages as a PublicInbox::Eml
751 (or Email::MIME) object.
752 If the message could not be found, the code is "MISSING"
753 and the original message is undef. If there is a mismatch where
754 the "Message-ID" is matched but the subject and body do not match,
755 the returned code is "MISMATCH" and the conflicting message
760 Finalizes the L<git-fast-import(1)> and unlocks the repository.
761 Calling this is required to finalize changes to a repository.
769 All feedback welcome via plain-text mail to L<mailto:meta@public-inbox.org>
771 The mail archives are hosted at L<https://public-inbox.org/meta/>
775 Copyright (C) 2016-2020 all contributors L<mailto:meta@public-inbox.org>
777 License: AGPL-3.0+ L<http://www.gnu.org/licenses/agpl-3.0.txt>