1 # Copyright (C) 2014-2021 all contributors <meta@public-inbox.org>
2 # License: AGPL-3.0+ <https://www.gnu.org/licenses/agpl-3.0.txt>
4 # Main web interface for mailing list archives
6 # We focus on the lowest common denominators here:
7 # - targeted at text-only console browsers (w3m, links, etc..)
8 # - Only basic HTML, CSS only for line-wrapping <pre> text content for GUIs
9 # and diff/syntax-highlighting (optional)
10 # - No JavaScript, graphics or icons allowed.
11 # - Must not rely on static content
12 # - UTF-8 is only for user-content, 7-bit US-ASCII for us
13 package PublicInbox::WWW;
17 use bytes (); # only for bytes::length
18 use PublicInbox::Config;
19 use PublicInbox::Hval;
20 use URI::Escape qw(uri_unescape);
21 use PublicInbox::MID qw(mid_escape);
22 use PublicInbox::GitHTTPBackend;
23 use PublicInbox::UserContent;
24 use PublicInbox::WwwStatic qw(r path_info_raw);
27 # TODO: consider a routing tree now that we have more endpoints:
28 our $INBOX_RE = qr!\A/([\w\-][\w\.\-]*)!;
29 our $MID_RE = qr!([^/]+)!;
30 our $END_RE = qr!(T/|t/|t\.mbox(?:\.gz)?|t\.atom|raw|)!;
31 our $ATTACH_RE = qr!([0-9][0-9\.]*)-($PublicInbox::Hval::FN)!;
32 our $OID_RE = qr![a-f0-9]{7,}!;
35 my ($class, $pi_cfg) = @_;
36 bless { pi_cfg => $pi_cfg // PublicInbox::Config->new }, $class;
39 # backwards compatibility, do not use
41 my ($req, $method) = @_;
42 PublicInbox::WWW->new->call($req->env);
46 my ($self, $env) = @_;
47 my $ctx = { env => $env, www => $self };
49 # we don't care about multi-value
53 my ($k, $v) = split(/=/, $_, 2);
54 # none of the keys we care about will need escaping
55 ($k // '', uri_unescape($v // ''))
56 } split(/[&;]+/, $env->{QUERY_STRING});
58 my $path_info = path_info_raw($env);
59 my $method = $env->{REQUEST_METHOD};
61 if ($method eq 'POST') {
62 if ($path_info =~ m!$INBOX_RE/(?:(?:git/)?([0-9]+)(?:\.git)?/)?
63 (git-upload-pack)\z!x) {
64 my ($epoch, $path) = ($2, $3);
65 return invalid_inbox($ctx, $1) ||
66 serve_git($ctx, $epoch, $path);
67 } elsif ($path_info =~ m!$INBOX_RE/(\w+)\.sql\.gz\z!o) {
68 return get_altid_dump($ctx, $1, $2);
69 } elsif ($path_info =~ m!$INBOX_RE/!o) {
70 return invalid_inbox($ctx, $1) || mbox_results($ctx);
73 elsif ($method !~ /\A(?:GET|HEAD)\z/) {
77 # top-level indices and feeds
78 if ($path_info eq '/') {
79 require PublicInbox::WwwListing;
80 PublicInbox::WwwListing->response($ctx);
81 } elsif ($path_info eq '/manifest.js.gz') {
82 require PublicInbox::ManifestJsGz;
83 PublicInbox::ManifestJsGz->response($ctx);
84 } elsif ($path_info =~ m!$INBOX_RE\z!o) {
85 invalid_inbox($ctx, $1) || r301($ctx, $1);
86 } elsif ($path_info =~ m!$INBOX_RE(?:/|/index\.html)?\z!o) {
87 invalid_inbox($ctx, $1) || get_index($ctx);
88 } elsif ($path_info =~ m!$INBOX_RE/(?:atom\.xml|new\.atom)\z!o) {
89 invalid_inbox($ctx, $1) || get_atom($ctx);
90 } elsif ($path_info =~ m!$INBOX_RE/new\.html\z!o) {
91 invalid_inbox($ctx, $1) || get_new($ctx);
92 } elsif ($path_info =~ m!$INBOX_RE/description\z!o) {
93 get_description($ctx, $1);
94 } elsif ($path_info =~ m!$INBOX_RE/(?:(?:git/)?([0-9]+)(?:\.git)?/)?
95 ($PublicInbox::GitHTTPBackend::ANY)\z!ox) {
96 my ($epoch, $path) = ($2, $3);
97 invalid_inbox($ctx, $1) || serve_git($ctx, $epoch, $path);
98 } elsif ($path_info =~ m!$INBOX_RE/([a-zA-Z0-9_\-]+).mbox\.gz\z!o) {
99 serve_mbox_range($ctx, $1, $2);
100 } elsif ($path_info =~ m!$INBOX_RE/$MID_RE/$END_RE\z!o) {
101 msg_page($ctx, $1, $2, $3);
103 } elsif ($path_info =~ m!$INBOX_RE/$MID_RE/$ATTACH_RE\z!o) {
104 my ($idx, $fn) = ($3, $4);
105 invalid_inbox_mid($ctx, $1, $2) || get_attach($ctx, $idx, $fn);
106 # in case people leave off the trailing slash:
107 } elsif ($path_info =~ m!$INBOX_RE/$MID_RE/(T|t)\z!o) {
108 my ($inbox, $mid_ue, $suffix) = ($1, $2, $3);
109 $suffix .= $suffix =~ /\A[tT]\z/ ? '/#u' : '/';
110 r301($ctx, $inbox, $mid_ue, $suffix);
112 } elsif ($path_info =~ m!$INBOX_RE/$MID_RE/R/?\z!o) {
113 my ($inbox, $mid_ue) = ($1, $2);
114 r301($ctx, $inbox, $mid_ue, '#R');
116 } elsif ($path_info =~ m!$INBOX_RE/$MID_RE/f/?\z!o) {
118 } elsif ($path_info =~ m!$INBOX_RE/_/text(?:/(.*))?\z!o) {
119 get_text($ctx, $1, $2);
120 } elsif ($path_info =~ m!$INBOX_RE/([a-zA-Z0-9_\-\.]+)\.css\z!o) {
121 get_css($ctx, $1, $2);
122 } elsif ($path_info =~ m!$INBOX_RE/manifest\.js\.gz\z!o) {
123 get_inbox_manifest($ctx, $1, $2);
124 } elsif ($path_info =~ m!$INBOX_RE/($OID_RE)/s/\z!o) {
125 get_vcs_object($ctx, $1, $2);
126 } elsif ($path_info =~ m!$INBOX_RE/($OID_RE)/s/
127 ($PublicInbox::Hval::FN)\z!ox) {
128 get_vcs_object($ctx, $1, $2, $3);
129 } elsif ($path_info =~ m!$INBOX_RE/($OID_RE)/s\z!o) {
130 r301($ctx, $1, $2, 's/');
131 } elsif ($path_info =~ m!$INBOX_RE/(\w+)\.sql\.gz\z!o) {
132 get_altid_dump($ctx, $1, $2);
133 # convenience redirects order matters
134 } elsif ($path_info =~ m!$INBOX_RE/([^/]{2,})\z!o) {
138 legacy_redirects($ctx, $path_info);
142 # for CoW-friendliness, MOOOOO! Even for single-process setups,
143 # we want to get all immortal allocations done early to avoid heap
144 # fragmentation since common allocators favor a large contiguous heap.
148 # populate caches used by Encode internally, since emails
149 # may show up with any encoding.
151 Encode::find_encoding($_) for Encode->encodings(':all');
153 require PublicInbox::ExtMsg;
154 require PublicInbox::Feed;
155 require PublicInbox::View;
156 require PublicInbox::SearchThread;
157 require PublicInbox::Eml;
158 require PublicInbox::Mbox;
159 require PublicInbox::ViewVCS;
160 require PublicInbox::WwwText;
161 require PublicInbox::WwwAttach;
163 require PublicInbox::Search;
164 PublicInbox::Search::load_xapian();
166 for (qw(SearchView MboxGz WwwAltId)) {
167 eval "require PublicInbox::$_;";
170 my $pi_cfg = $self->{pi_cfg};
171 if (defined($pi_cfg->{'publicinbox.cgitrc'})) {
172 $pi_cfg->limiter('-cgit');
175 $self->stylesheets_prepare($_) for ('', '../', '../../');
177 $pi_cfg->each_inbox(\&preload_inbox);
189 # private functions below
193 if ($ctx && $ctx->{mid}) {
194 require PublicInbox::ExtMsg;
195 return PublicInbox::ExtMsg::ext_msg($ctx);
200 sub news_cgit_fallback ($) {
202 my $www = $ctx->{www};
203 my $env = $ctx->{env};
204 my $res = $www->news_www->call($env);
205 $res->[0] == 404 ? $www->cgit->call($env) : $res;
208 # returns undef if valid, array ref response if invalid
209 sub invalid_inbox ($$) {
210 my ($ctx, $inbox) = @_;
211 my $ibx = $ctx->{www}->{pi_cfg}->lookup_name($inbox) //
212 $ctx->{www}->{pi_cfg}->lookup_ei($inbox);
218 # sometimes linkifiers (not ours!) screw up automatic link
219 # generation and link things intended for nntp:// to https?://,
220 # so try to infer links and redirect them to the appropriate
222 news_cgit_fallback($ctx);
225 # returns undef if valid, array ref response if invalid
226 sub invalid_inbox_mid {
227 my ($ctx, $inbox, $mid_ue) = @_;
228 my $ret = invalid_inbox($ctx, $inbox);
231 my $mid = $ctx->{mid} = uri_unescape($mid_ue);
232 my $ibx = $ctx->{ibx};
233 if ($mid =~ m!\A([a-f0-9]{2})([a-f0-9]{38})\z!) {
234 my ($x2, $x38) = ($1, $2);
235 # this is horrifically wasteful for legacy URLs:
236 my $str = $ctx->{ibx}->msg_by_path("$x2/$x38") or return;
237 my $s = PublicInbox::Eml->new($str);
238 $mid = PublicInbox::MID::mid_clean($s->header_raw('Message-ID'));
239 return r301($ctx, $inbox, mid_escape($mid));
244 # /$INBOX/new.atom -> Atom feed, includes replies
247 require PublicInbox::Feed;
248 PublicInbox::Feed::generate($ctx);
251 # /$INBOX/new.html -> HTML only
254 require PublicInbox::Feed;
255 PublicInbox::Feed::new_html($ctx);
258 # /$INBOX/?r=$GIT_COMMIT -> HTML only
261 require PublicInbox::Feed;
262 if ($ctx->{env}->{QUERY_STRING} =~ /(?:\A|[&;])q=/) {
263 require PublicInbox::SearchView;
264 PublicInbox::SearchView::sres_top_html($ctx);
266 PublicInbox::Feed::generate_html_index($ctx);
270 # /$INBOX/$MESSAGE_ID/raw -> raw mbox
273 require PublicInbox::Mbox;
274 PublicInbox::Mbox::emit_raw($ctx) || r(404);
277 # /$INBOX/$MESSAGE_ID/ -> HTML content (short quotes)
280 require PublicInbox::View;
281 PublicInbox::View::msg_page($ctx) || r404($ctx);
284 # /$INBOX/$MESSAGE_ID/t/
286 my ($ctx, $flat) = @_;
287 $ctx->{ibx}->over or return need($ctx, 'Overview');
288 $ctx->{flat} = $flat;
289 require PublicInbox::View;
290 PublicInbox::View::thread_html($ctx);
293 # /$INBOX/_/text/$KEY/
294 # /$INBOX/_/text/$KEY/raw
295 # KEY may contain slashes
297 my ($ctx, $inbox, $key) = @_;
298 my $r404 = invalid_inbox($ctx, $inbox);
299 return $r404 if $r404;
301 require PublicInbox::WwwText;
302 PublicInbox::WwwText::get_text($ctx, $key);
305 # show git objects (blobs and commits)
306 # /$INBOX/$GIT_OBJECT_ID/s/
307 # /$INBOX/$GIT_OBJECT_ID/s/$FILENAME
308 sub get_vcs_object ($$$;$) {
309 my ($ctx, $inbox, $oid, $filename) = @_;
310 my $r404 = invalid_inbox($ctx, $inbox);
311 return $r404 if $r404 || !$ctx->{www}->{pi_cfg}->repo_objs($ctx->{ibx});
312 require PublicInbox::ViewVCS;
313 PublicInbox::ViewVCS::show($ctx, $oid, $filename);
317 my ($ctx, $inbox, $altid_pfx) =@_;
318 my $r404 = invalid_inbox($ctx, $inbox);
319 return $r404 if $r404;
320 eval { require PublicInbox::WwwAltId } or return need($ctx, 'sqlite3');
321 PublicInbox::WwwAltId::sqldump($ctx, $altid_pfx);
325 my ($ctx, $extra) = @_;
326 require PublicInbox::WwwStream;
327 PublicInbox::WwwStream::html_oneshot($ctx, 501, \<<EOF);
328 <pre>$extra is not available for this public-inbox
329 <a\nhref="../">Return to index</a></pre>
333 # /$INBOX/$MESSAGE_ID/t.mbox -> thread as mbox
334 # /$INBOX/$MESSAGE_ID/t.mbox.gz -> thread as gzipped mbox
335 # note: I'm not a big fan of other compression formats since they're
336 # significantly more expensive on CPU than gzip and less-widely available,
337 # especially on older systems. Stick to zlib since that's what git uses.
338 sub get_thread_mbox {
339 my ($ctx, $sfx) = @_;
340 my $over = $ctx->{ibx}->over or return need($ctx, 'Overview');
341 require PublicInbox::Mbox;
342 PublicInbox::Mbox::thread_mbox($ctx, $over, $sfx);
346 # /$INBOX/$MESSAGE_ID/t.atom -> thread as Atom feed
347 sub get_thread_atom {
349 $ctx->{ibx}->over or return need($ctx, 'Overview');
350 require PublicInbox::Feed;
351 PublicInbox::Feed::generate_thread_atom($ctx);
354 sub legacy_redirects {
355 my ($ctx, $path_info) = @_;
357 # single-message pages
358 if ($path_info =~ m!$INBOX_RE/m/(\S+)/\z!o) {
360 } elsif ($path_info =~ m!$INBOX_RE/m/(\S+)/raw\z!o) {
361 r301($ctx, $1, $2, 'raw');
363 } elsif ($path_info =~ m!$INBOX_RE/f/(\S+)/\z!o) {
367 } elsif ($path_info =~ m!$INBOX_RE/t/(\S+)/\z!o) {
368 r301($ctx, $1, $2, 't/#u');
370 } elsif ($path_info =~ m!$INBOX_RE/t/(\S+)/mbox(\.gz)?\z!o) {
371 r301($ctx, $1, $2, "t.mbox$3");
373 # even older legacy redirects
374 } elsif ($path_info =~ m!$INBOX_RE/m/(\S+)\.html\z!o) {
377 } elsif ($path_info =~ m!$INBOX_RE/t/(\S+)\.html\z!o) {
378 r301($ctx, $1, $2, 't/#u');
380 } elsif ($path_info =~ m!$INBOX_RE/f/(\S+)\.html\z!o) {
383 } elsif ($path_info =~ m!$INBOX_RE/(?:m|f)/(\S+)\.txt\z!o) {
384 r301($ctx, $1, $2, 'raw');
386 } elsif ($path_info =~ m!$INBOX_RE/t/(\S+)(\.mbox(?:\.gz)?)\z!o) {
387 r301($ctx, $1, $2, "t$3");
389 # legacy convenience redirects, order still matters
390 } elsif ($path_info =~ m!$INBOX_RE/m/(\S+)\z!o) {
392 } elsif ($path_info =~ m!$INBOX_RE/t/(\S+)\z!o) {
393 r301($ctx, $1, $2, 't/#u');
394 } elsif ($path_info =~ m!$INBOX_RE/f/(\S+)\z!o) {
397 # some Message-IDs have slashes in them and the HTTP server
398 # may try to be clever and unescape them :<
399 } elsif ($path_info =~ m!$INBOX_RE/(\S+/\S+)/$END_RE\z!o) {
400 msg_page($ctx, $1, $2, $3);
402 # in case people leave off the trailing slash:
403 } elsif ($path_info =~ m!$INBOX_RE/(\S+/\S+)/(T|t)\z!o) {
404 r301($ctx, $1, $2, $3 eq 't' ? 't/#u' : $3);
405 } elsif ($path_info =~ m!$INBOX_RE/(\S+/\S+)/f\z!o) {
408 news_cgit_fallback($ctx);
413 my ($ctx, $inbox, $mid_ue, $suffix) = @_;
414 my $ibx = $ctx->{ibx};
416 my $r404 = invalid_inbox($ctx, $inbox);
417 return $r404 if $r404;
420 my $url = $ibx->base_url($ctx->{env});
421 my $qs = $ctx->{env}->{QUERY_STRING};
422 if (defined $mid_ue) {
423 # common, and much nicer as '@' than '%40':
424 $mid_ue =~ s/%40/@/g;
425 $url .= $mid_ue . '/';
427 $url .= $suffix if (defined $suffix);
428 $url .= "?$qs" if $qs ne '';
431 [ Location => $url, 'Content-Type' => 'text/plain' ],
432 [ "Redirecting to $url\n" ] ]
436 my ($ctx, $inbox, $mid_ue, $e) = @_;
438 $ret = invalid_inbox_mid($ctx, $inbox, $mid_ue) and return $ret;
439 '' eq $e and return get_mid_html($ctx);
440 'T/' eq $e and return get_thread($ctx, 1);
441 't/' eq $e and return get_thread($ctx);
442 't.atom' eq $e and return get_thread_atom($ctx);
443 't.mbox' eq $e and return get_thread_mbox($ctx);
444 't.mbox.gz' eq $e and return get_thread_mbox($ctx, '.gz');
445 'raw' eq $e and return get_mid_txt($ctx);
447 # legacy, but no redirect for compatibility:
448 'f/' eq $e and return get_mid_html($ctx);
453 my ($ctx, $epoch, $path) = @_;
454 my $env = $ctx->{env};
455 my $ibx = $ctx->{ibx};
456 my $git = defined $epoch ? $ibx->git_epoch($epoch) : $ibx->git;
457 $git ? PublicInbox::GitHTTPBackend::serve($env, $git, $path) : r404();
462 if ($ctx->{env}->{QUERY_STRING} =~ /(?:\A|[&;])q=/) {
463 $ctx->{ibx}->isrch or return need($ctx, 'search');
464 require PublicInbox::SearchView;
465 return PublicInbox::SearchView::mbox_results($ctx);
470 sub serve_mbox_range {
471 my ($ctx, $inbox, $range) = @_;
472 invalid_inbox($ctx, $inbox) || eval {
473 require PublicInbox::Mbox;
474 PublicInbox::Mbox::emit_range($ctx, $range);
480 $self->{news_www} ||= do {
481 require PublicInbox::NewsWWW;
482 PublicInbox::NewsWWW->new($self->{pi_cfg});
488 $self->{cgit} ||= do {
489 my $pi_cfg = $self->{pi_cfg};
491 if (defined($pi_cfg->{'publicinbox.cgitrc'})) {
492 require PublicInbox::Cgit;
493 PublicInbox::Cgit->new($pi_cfg);
496 Plack::Util::inline_object(call => sub { r404() });
501 # GET $INBOX/manifest.js.gz
502 sub get_inbox_manifest ($$$) {
503 my ($ctx, $inbox, $key) = @_;
504 my $r404 = invalid_inbox($ctx, $inbox);
505 return $r404 if $r404;
506 require PublicInbox::ManifestJsGz;
507 PublicInbox::ManifestJsGz::per_inbox($ctx);
511 my ($ctx, $idx, $fn) = @_;
512 require PublicInbox::WwwAttach;
513 PublicInbox::WwwAttach::get_attach($ctx, $idx, $fn);
516 # User-generated content (UGC) may have excessively long lines
517 # and screw up rendering on some browsers, so we use pre-wrap.
519 # We also force everything to the same scaled font-size because GUI
520 # browsers (tested both Firefox and surf (webkit)) uses a larger font
521 # for the Search <form> element than the rest of the page. Font size
522 # uniformity is important to people who rely on gigantic fonts.
523 # Finally, we use monospace to ensure the Search field and button
524 # has the same size and spacing as everything else which is
525 # <pre>-formatted anyways.
526 our $STYLE = 'pre{white-space:pre-wrap}*{font-size:100%;font-family:monospace}';
528 sub stylesheets_prepare ($$) {
529 my ($self, $upfx) = @_;
531 require CSS::Minifier;
532 sub { CSS::Minifier::minify(input => $_[0]) };
534 require CSS::Minifier::XS;
535 sub { CSS::Minifier::XS::minify($_[0]) };
539 my $stylesheets = $self->{pi_cfg}->{css} || [];
543 foreach my $s (@$stylesheets) {
546 foreach my $k (qw(media title href)) {
547 if (s/\s*$k='([^']+)'// || s/\s*$k=(\S+)//) {
552 if (defined $attr->{href}) {
556 my ($key) = (m!([^/]+?)(?:\.css)?\z!i);
557 if ($key !~ /\A[a-zA-Z0-9_\-\.]+\z/) {
558 warn "ignoring $fn, non-ASCII word character\n";
561 open(my $fh, '<', $fn) or do {
562 warn "failed to open $fn: $!\n";
566 my $local = do { local $/; <$fh> };
567 if ($local =~ /\S/) {
568 $ctime = sprintf('%x',(stat($fh))[10]);
569 $local = $mini->($local);
572 # do not let BOFHs override userContent.css:
573 if ($local =~ /!\s*important\b/i) {
574 warn "ignoring $fn since it uses `!important'\n";
578 $css_map->{$key} = $local;
579 $attr->{href} = "$upfx$key.css?$ctime";
580 if (defined($attr->{title})) {
582 } elsif (($attr->{media}||'screen') eq 'screen') {
583 $attr->{-inline} = $local;
589 my $buf = "<style>$STYLE";
591 my @ext; # for media=print and whatnot
592 foreach my $attr (@$links) {
593 if (defined(my $str = delete $attr->{-inline})) {
604 foreach my $attr (@$links) {
605 delete $attr->{-inline};
606 $buf .= "<link\ntype=text/css\nrel=stylesheet";
607 while (my ($k, $v) = each %$attr) {
608 $v = qq{"$v"} if $v =~ /[\s=]/;
613 $self->{"-style-$upfx"} = $buf;
615 $self->{-style_inline} = $buf;
617 $self->{-css_map} = $css_map;
620 # returns an HTML fragment with <style> or <link> tags in them
621 # Called by WwwStream by nearly every HTML page
623 my ($self, $upfx) = @_;
624 $self->{-style_inline} || $self->{"-style-$upfx"} || do {
625 stylesheets_prepare($self, $upfx);
626 $self->{-style_inline} || $self->{"-style-$upfx"}
630 # /$INBOX/$KEY.css endpoint
631 # CSS is configured globally for all inboxes, but we access them on
632 # a per-inbox basis. This allows administrators to setup per-inbox
633 # static routes to intercept the request before it hits PSGI
635 my ($ctx, $inbox, $key) = @_;
636 my $r404 = invalid_inbox($ctx, $inbox);
637 return $r404 if $r404;
638 my $self = $ctx->{www};
639 my $css_map = $self->{-css_map} || stylesheets_prepare($self, '');
640 my $css = $css_map->{$key};
641 if (!defined($css) && $key eq 'userContent') {
642 my $env = $ctx->{env};
643 $css = PublicInbox::UserContent::sample($ctx->{ibx}, $env);
645 defined $css or return r404();
646 my $h = [ 'Content-Length', bytes::length($css),
647 'Content-Type', 'text/css' ];
648 PublicInbox::GitHTTPBackend::cache_one_year($h);
649 [ 200, $h, [ $css ] ];
652 sub get_description {
653 my ($ctx, $inbox) = @_;
654 invalid_inbox($ctx, $inbox) || do {
655 my $d = $ctx->{ibx}->description . "\n";
656 [ 200, [ 'Content-Length', bytes::length($d),
657 'Content-Type', 'text/plain' ], [ $d ] ];