1 # Copyright (C) 2014-2020 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_config) = @_;
36 $pi_config ||= PublicInbox::Config->new;
37 bless { pi_config => $pi_config }, $class;
40 # backwards compatibility, do not use
42 my ($req, $method) = @_;
43 PublicInbox::WWW->new->call($req->env);
47 my ($self, $env) = @_;
48 my $ctx = { env => $env, www => $self };
50 # we don't care about multi-value
54 my ($k, $v) = split('=', $_, 2);
55 $v = uri_unescape($v // '');
56 # none of the keys we care about will need escaping
58 } split(/[&;]+/, $env->{QUERY_STRING});
60 my $path_info = path_info_raw($env);
61 my $method = $env->{REQUEST_METHOD};
63 if ($method eq 'POST') {
64 if ($path_info =~ m!$INBOX_RE/(?:(?:git/)?([0-9]+)(?:\.git)?/)?
65 (git-upload-pack)\z!x) {
66 my ($epoch, $path) = ($2, $3);
67 return invalid_inbox($ctx, $1) ||
68 serve_git($ctx, $epoch, $path);
69 } elsif ($path_info =~ m!$INBOX_RE/(\w+)\.sql\.gz\z!o) {
70 return get_altid_dump($ctx, $1, $2);
71 } elsif ($path_info =~ m!$INBOX_RE/!o) {
72 return invalid_inbox($ctx, $1) || mbox_results($ctx);
75 elsif ($method !~ /\A(?:GET|HEAD)\z/) {
79 # top-level indices and feeds
80 if ($path_info eq '/') {
81 require PublicInbox::WwwListing;
82 PublicInbox::WwwListing->response($ctx);
83 } elsif ($path_info eq '/manifest.js.gz') {
84 require PublicInbox::ManifestJsGz;
85 PublicInbox::ManifestJsGz->response($ctx);
86 } elsif ($path_info =~ m!$INBOX_RE\z!o) {
87 invalid_inbox($ctx, $1) || r301($ctx, $1);
88 } elsif ($path_info =~ m!$INBOX_RE(?:/|/index\.html)?\z!o) {
89 invalid_inbox($ctx, $1) || get_index($ctx);
90 } elsif ($path_info =~ m!$INBOX_RE/(?:atom\.xml|new\.atom)\z!o) {
91 invalid_inbox($ctx, $1) || get_atom($ctx);
92 } elsif ($path_info =~ m!$INBOX_RE/new\.html\z!o) {
93 invalid_inbox($ctx, $1) || get_new($ctx);
94 } elsif ($path_info =~ m!$INBOX_RE/description\z!o) {
95 get_description($ctx, $1);
96 } elsif ($path_info =~ m!$INBOX_RE/(?:(?:git/)?([0-9]+)(?:\.git)?/)?
97 ($PublicInbox::GitHTTPBackend::ANY)\z!ox) {
98 my ($epoch, $path) = ($2, $3);
99 invalid_inbox($ctx, $1) || serve_git($ctx, $epoch, $path);
100 } elsif ($path_info =~ m!$INBOX_RE/([a-zA-Z0-9_\-]+).mbox\.gz\z!o) {
101 serve_mbox_range($ctx, $1, $2);
102 } elsif ($path_info =~ m!$INBOX_RE/$MID_RE/$END_RE\z!o) {
103 msg_page($ctx, $1, $2, $3);
105 } elsif ($path_info =~ m!$INBOX_RE/$MID_RE/$ATTACH_RE\z!o) {
106 my ($idx, $fn) = ($3, $4);
107 invalid_inbox_mid($ctx, $1, $2) || get_attach($ctx, $idx, $fn);
108 # in case people leave off the trailing slash:
109 } elsif ($path_info =~ m!$INBOX_RE/$MID_RE/(T|t)\z!o) {
110 my ($inbox, $mid_ue, $suffix) = ($1, $2, $3);
111 $suffix .= $suffix =~ /\A[tT]\z/ ? '/#u' : '/';
112 r301($ctx, $inbox, $mid_ue, $suffix);
114 } elsif ($path_info =~ m!$INBOX_RE/$MID_RE/R/?\z!o) {
115 my ($inbox, $mid_ue) = ($1, $2);
116 r301($ctx, $inbox, $mid_ue, '#R');
118 } elsif ($path_info =~ m!$INBOX_RE/$MID_RE/f/?\z!o) {
120 } elsif ($path_info =~ m!$INBOX_RE/_/text(?:/(.*))?\z!o) {
121 get_text($ctx, $1, $2);
122 } elsif ($path_info =~ m!$INBOX_RE/([a-zA-Z0-9_\-\.]+)\.css\z!o) {
123 get_css($ctx, $1, $2);
124 } elsif ($path_info =~ m!$INBOX_RE/manifest\.js\.gz\z!o) {
125 get_inbox_manifest($ctx, $1, $2);
126 } elsif ($path_info =~ m!$INBOX_RE/($OID_RE)/s/\z!o) {
127 get_vcs_object($ctx, $1, $2);
128 } elsif ($path_info =~ m!$INBOX_RE/($OID_RE)/s/
129 ($PublicInbox::Hval::FN)\z!ox) {
130 get_vcs_object($ctx, $1, $2, $3);
131 } elsif ($path_info =~ m!$INBOX_RE/($OID_RE)/s\z!o) {
132 r301($ctx, $1, $2, 's/');
133 } elsif ($path_info =~ m!$INBOX_RE/(\w+)\.sql\.gz\z!o) {
134 get_altid_dump($ctx, $1, $2);
135 # convenience redirects order matters
136 } elsif ($path_info =~ m!$INBOX_RE/([^/]{2,})\z!o) {
140 legacy_redirects($ctx, $path_info);
144 # for CoW-friendliness, MOOOOO! Even for single-process setups,
145 # we want to get all immortal allocations done early to avoid heap
146 # fragmentation since common allocators favor a large contiguous heap.
150 # populate caches used by Encode internally, since emails
151 # may show up with any encoding.
153 Encode::find_encoding($_) for Encode->encodings(':all');
155 require PublicInbox::ExtMsg;
156 require PublicInbox::Feed;
157 require PublicInbox::View;
158 require PublicInbox::SearchThread;
159 require PublicInbox::Eml;
160 require PublicInbox::Mbox;
161 require PublicInbox::ViewVCS;
162 require PublicInbox::WwwText;
163 require PublicInbox::WwwAttach;
165 require PublicInbox::Search;
166 PublicInbox::Search::load_xapian();
168 for (qw(SearchView MboxGz WwwAltId)) {
169 eval "require PublicInbox::$_;";
172 my $pi_config = $self->{pi_config};
173 if (defined($pi_config->{'publicinbox.cgitrc'})) {
174 $pi_config->limiter('-cgit');
177 $self->stylesheets_prepare($_) for ('', '../', '../../');
179 $pi_config->each_inbox(\&preload_inbox);
191 # private functions below
195 if ($ctx && $ctx->{mid}) {
196 require PublicInbox::ExtMsg;
197 return PublicInbox::ExtMsg::ext_msg($ctx);
202 sub news_cgit_fallback ($) {
204 my $www = $ctx->{www};
205 my $env = $ctx->{env};
206 my $res = $www->news_www->call($env);
207 $res->[0] == 404 ? $www->cgit->call($env) : $res;
210 # returns undef if valid, array ref response if invalid
211 sub invalid_inbox ($$) {
212 my ($ctx, $inbox) = @_;
213 my $ibx = $ctx->{www}->{pi_config}->lookup_name($inbox) //
214 $ctx->{www}->{pi_config}->lookup_ei($inbox);
216 $ctx->{-inbox} = $ibx;
220 # sometimes linkifiers (not ours!) screw up automatic link
221 # generation and link things intended for nntp:// to https?://,
222 # so try to infer links and redirect them to the appropriate
224 news_cgit_fallback($ctx);
227 # returns undef if valid, array ref response if invalid
228 sub invalid_inbox_mid {
229 my ($ctx, $inbox, $mid_ue) = @_;
230 my $ret = invalid_inbox($ctx, $inbox);
233 my $mid = $ctx->{mid} = uri_unescape($mid_ue);
234 my $ibx = $ctx->{-inbox};
235 if ($mid =~ m!\A([a-f0-9]{2})([a-f0-9]{38})\z!) {
236 my ($x2, $x38) = ($1, $2);
237 # this is horrifically wasteful for legacy URLs:
238 my $str = $ctx->{-inbox}->msg_by_path("$x2/$x38") or return;
239 my $s = PublicInbox::Eml->new($str);
240 $mid = PublicInbox::MID::mid_clean($s->header_raw('Message-ID'));
241 return r301($ctx, $inbox, mid_escape($mid));
246 # /$INBOX/new.atom -> Atom feed, includes replies
249 require PublicInbox::Feed;
250 PublicInbox::Feed::generate($ctx);
253 # /$INBOX/new.html -> HTML only
256 require PublicInbox::Feed;
257 PublicInbox::Feed::new_html($ctx);
260 # /$INBOX/?r=$GIT_COMMIT -> HTML only
263 require PublicInbox::Feed;
264 if ($ctx->{env}->{QUERY_STRING} =~ /(?:\A|[&;])q=/) {
265 require PublicInbox::SearchView;
266 PublicInbox::SearchView::sres_top_html($ctx);
268 PublicInbox::Feed::generate_html_index($ctx);
272 # /$INBOX/$MESSAGE_ID/raw -> raw mbox
275 require PublicInbox::Mbox;
276 PublicInbox::Mbox::emit_raw($ctx) || r404($ctx);
279 # /$INBOX/$MESSAGE_ID/ -> HTML content (short quotes)
282 require PublicInbox::View;
283 PublicInbox::View::msg_page($ctx) || r404($ctx);
286 # /$INBOX/$MESSAGE_ID/t/
288 my ($ctx, $flat) = @_;
289 $ctx->{-inbox}->over or return need($ctx, 'Overview');
290 $ctx->{flat} = $flat;
291 require PublicInbox::View;
292 PublicInbox::View::thread_html($ctx);
295 # /$INBOX/_/text/$KEY/
296 # /$INBOX/_/text/$KEY/raw
297 # KEY may contain slashes
299 my ($ctx, $inbox, $key) = @_;
300 my $r404 = invalid_inbox($ctx, $inbox);
301 return $r404 if $r404;
303 require PublicInbox::WwwText;
304 PublicInbox::WwwText::get_text($ctx, $key);
307 # show git objects (blobs and commits)
308 # /$INBOX/$GIT_OBJECT_ID/s/
309 # /$INBOX/$GIT_OBJECT_ID/s/$FILENAME
310 sub get_vcs_object ($$$;$) {
311 my ($ctx, $inbox, $oid, $filename) = @_;
312 my $r404 = invalid_inbox($ctx, $inbox);
313 return $r404 if $r404;
314 require PublicInbox::ViewVCS;
315 PublicInbox::ViewVCS::show($ctx, $oid, $filename);
319 my ($ctx, $inbox, $altid_pfx) =@_;
320 my $r404 = invalid_inbox($ctx, $inbox);
321 return $r404 if $r404;
322 eval { require PublicInbox::WwwAltId } or return need($ctx, 'sqlite3');
323 PublicInbox::WwwAltId::sqldump($ctx, $altid_pfx);
327 my ($ctx, $extra) = @_;
328 require PublicInbox::WwwStream;
329 PublicInbox::WwwStream::html_oneshot($ctx, 501, \<<EOF);
330 <pre>$extra is not available for this public-inbox
331 <a\nhref="../">Return to index</a></pre>
335 # /$INBOX/$MESSAGE_ID/t.mbox -> thread as mbox
336 # /$INBOX/$MESSAGE_ID/t.mbox.gz -> thread as gzipped mbox
337 # note: I'm not a big fan of other compression formats since they're
338 # significantly more expensive on CPU than gzip and less-widely available,
339 # especially on older systems. Stick to zlib since that's what git uses.
340 sub get_thread_mbox {
341 my ($ctx, $sfx) = @_;
342 my $over = $ctx->{-inbox}->over or return need($ctx, 'Overview');
343 require PublicInbox::Mbox;
344 PublicInbox::Mbox::thread_mbox($ctx, $over, $sfx);
348 # /$INBOX/$MESSAGE_ID/t.atom -> thread as Atom feed
349 sub get_thread_atom {
351 $ctx->{-inbox}->over or return need($ctx, 'Overview');
352 require PublicInbox::Feed;
353 PublicInbox::Feed::generate_thread_atom($ctx);
356 sub legacy_redirects {
357 my ($ctx, $path_info) = @_;
359 # single-message pages
360 if ($path_info =~ m!$INBOX_RE/m/(\S+)/\z!o) {
362 } elsif ($path_info =~ m!$INBOX_RE/m/(\S+)/raw\z!o) {
363 r301($ctx, $1, $2, 'raw');
365 } elsif ($path_info =~ m!$INBOX_RE/f/(\S+)/\z!o) {
369 } elsif ($path_info =~ m!$INBOX_RE/t/(\S+)/\z!o) {
370 r301($ctx, $1, $2, 't/#u');
372 } elsif ($path_info =~ m!$INBOX_RE/t/(\S+)/mbox(\.gz)?\z!o) {
373 r301($ctx, $1, $2, "t.mbox$3");
375 # even older legacy redirects
376 } elsif ($path_info =~ m!$INBOX_RE/m/(\S+)\.html\z!o) {
379 } elsif ($path_info =~ m!$INBOX_RE/t/(\S+)\.html\z!o) {
380 r301($ctx, $1, $2, 't/#u');
382 } elsif ($path_info =~ m!$INBOX_RE/f/(\S+)\.html\z!o) {
385 } elsif ($path_info =~ m!$INBOX_RE/(?:m|f)/(\S+)\.txt\z!o) {
386 r301($ctx, $1, $2, 'raw');
388 } elsif ($path_info =~ m!$INBOX_RE/t/(\S+)(\.mbox(?:\.gz)?)\z!o) {
389 r301($ctx, $1, $2, "t$3");
391 # legacy convenience redirects, order still matters
392 } elsif ($path_info =~ m!$INBOX_RE/m/(\S+)\z!o) {
394 } elsif ($path_info =~ m!$INBOX_RE/t/(\S+)\z!o) {
395 r301($ctx, $1, $2, 't/#u');
396 } elsif ($path_info =~ m!$INBOX_RE/f/(\S+)\z!o) {
399 # some Message-IDs have slashes in them and the HTTP server
400 # may try to be clever and unescape them :<
401 } elsif ($path_info =~ m!$INBOX_RE/(\S+/\S+)/$END_RE\z!o) {
402 msg_page($ctx, $1, $2, $3);
404 # in case people leave off the trailing slash:
405 } elsif ($path_info =~ m!$INBOX_RE/(\S+/\S+)/(T|t)\z!o) {
406 r301($ctx, $1, $2, $3 eq 't' ? 't/#u' : $3);
407 } elsif ($path_info =~ m!$INBOX_RE/(\S+/\S+)/f\z!o) {
410 news_cgit_fallback($ctx);
415 my ($ctx, $inbox, $mid_ue, $suffix) = @_;
416 my $ibx = $ctx->{-inbox};
418 my $r404 = invalid_inbox($ctx, $inbox);
419 return $r404 if $r404;
420 $ibx = $ctx->{-inbox};
422 my $url = $ibx->base_url($ctx->{env});
423 my $qs = $ctx->{env}->{QUERY_STRING};
424 if (defined $mid_ue) {
425 # common, and much nicer as '@' than '%40':
426 $mid_ue =~ s/%40/@/g;
427 $url .= $mid_ue . '/';
429 $url .= $suffix if (defined $suffix);
430 $url .= "?$qs" if $qs ne '';
433 [ Location => $url, 'Content-Type' => 'text/plain' ],
434 [ "Redirecting to $url\n" ] ]
438 my ($ctx, $inbox, $mid_ue, $e) = @_;
440 $ret = invalid_inbox_mid($ctx, $inbox, $mid_ue) and return $ret;
441 '' eq $e and return get_mid_html($ctx);
442 'T/' eq $e and return get_thread($ctx, 1);
443 't/' eq $e and return get_thread($ctx);
444 't.atom' eq $e and return get_thread_atom($ctx);
445 't.mbox' eq $e and return get_thread_mbox($ctx);
446 't.mbox.gz' eq $e and return get_thread_mbox($ctx, '.gz');
447 'raw' eq $e and return get_mid_txt($ctx);
449 # legacy, but no redirect for compatibility:
450 'f/' eq $e and return get_mid_html($ctx);
455 my ($ctx, $epoch, $path) = @_;
456 my $env = $ctx->{env};
457 my $ibx = $ctx->{-inbox};
458 my $git = defined $epoch ? $ibx->git_epoch($epoch) : $ibx->git;
459 $git ? PublicInbox::GitHTTPBackend::serve($env, $git, $path) : r404();
464 if ($ctx->{env}->{QUERY_STRING} =~ /(?:\A|[&;])q=/) {
465 $ctx->{-inbox}->search or return need($ctx, 'search');
466 require PublicInbox::SearchView;
467 return PublicInbox::SearchView::mbox_results($ctx);
472 sub serve_mbox_range {
473 my ($ctx, $inbox, $range) = @_;
474 invalid_inbox($ctx, $inbox) || eval {
475 require PublicInbox::Mbox;
476 PublicInbox::Mbox::emit_range($ctx, $range);
482 $self->{news_www} ||= do {
483 require PublicInbox::NewsWWW;
484 PublicInbox::NewsWWW->new($self->{pi_config});
490 $self->{cgit} ||= do {
491 my $pi_config = $self->{pi_config};
493 if (defined($pi_config->{'publicinbox.cgitrc'})) {
494 require PublicInbox::Cgit;
495 PublicInbox::Cgit->new($pi_config);
498 Plack::Util::inline_object(call => sub { r404() });
503 # GET $INBOX/manifest.js.gz
504 sub get_inbox_manifest ($$$) {
505 my ($ctx, $inbox, $key) = @_;
506 my $r404 = invalid_inbox($ctx, $inbox);
507 return $r404 if $r404;
508 require PublicInbox::ManifestJsGz;
509 PublicInbox::ManifestJsGz->response($ctx);
513 my ($ctx, $idx, $fn) = @_;
514 require PublicInbox::WwwAttach;
515 PublicInbox::WwwAttach::get_attach($ctx, $idx, $fn);
518 # User-generated content (UGC) may have excessively long lines
519 # and screw up rendering on some browsers, so we use pre-wrap.
521 # We also force everything to the same scaled font-size because GUI
522 # browsers (tested both Firefox and surf (webkit)) uses a larger font
523 # for the Search <form> element than the rest of the page. Font size
524 # uniformity is important to people who rely on gigantic fonts.
525 # Finally, we use monospace to ensure the Search field and button
526 # has the same size and spacing as everything else which is
527 # <pre>-formatted anyways.
528 our $STYLE = 'pre{white-space:pre-wrap}*{font-size:100%;font-family:monospace}';
530 sub stylesheets_prepare ($$) {
531 my ($self, $upfx) = @_;
533 require CSS::Minifier;
534 sub { CSS::Minifier::minify(input => $_[0]) };
536 require CSS::Minifier::XS;
537 sub { CSS::Minifier::XS::minify($_[0]) };
541 my $stylesheets = $self->{pi_config}->{css} || [];
545 foreach my $s (@$stylesheets) {
548 foreach my $k (qw(media title href)) {
549 if (s/\s*$k='([^']+)'// || s/\s*$k=(\S+)//) {
554 if (defined $attr->{href}) {
558 my ($key) = (m!([^/]+?)(?:\.css)?\z!i);
559 if ($key !~ /\A[a-zA-Z0-9_\-\.]+\z/) {
560 warn "ignoring $fn, non-ASCII word character\n";
563 open(my $fh, '<', $fn) or do {
564 warn "failed to open $fn: $!\n";
568 my $local = do { local $/; <$fh> };
569 if ($local =~ /\S/) {
570 $ctime = sprintf('%x',(stat($fh))[10]);
571 $local = $mini->($local);
574 # do not let BOFHs override userContent.css:
575 if ($local =~ /!\s*important\b/i) {
576 warn "ignoring $fn since it uses `!important'\n";
580 $css_map->{$key} = $local;
581 $attr->{href} = "$upfx$key.css?$ctime";
582 if (defined($attr->{title})) {
584 } elsif (($attr->{media}||'screen') eq 'screen') {
585 $attr->{-inline} = $local;
591 my $buf = "<style>$STYLE";
593 my @ext; # for media=print and whatnot
594 foreach my $attr (@$links) {
595 if (defined(my $str = delete $attr->{-inline})) {
606 foreach my $attr (@$links) {
607 delete $attr->{-inline};
608 $buf .= "<link\ntype=text/css\nrel=stylesheet";
609 while (my ($k, $v) = each %$attr) {
610 $v = qq{"$v"} if $v =~ /[\s=]/;
615 $self->{"-style-$upfx"} = $buf;
617 $self->{-style_inline} = $buf;
619 $self->{-css_map} = $css_map;
622 # returns an HTML fragment with <style> or <link> tags in them
623 # Called by WwwStream by nearly every HTML page
625 my ($self, $upfx) = @_;
626 $self->{-style_inline} || $self->{"-style-$upfx"} || do {
627 stylesheets_prepare($self, $upfx);
628 $self->{-style_inline} || $self->{"-style-$upfx"}
632 # /$INBOX/$KEY.css endpoint
633 # CSS is configured globally for all inboxes, but we access them on
634 # a per-inbox basis. This allows administrators to setup per-inbox
635 # static routes to intercept the request before it hits PSGI
637 my ($ctx, $inbox, $key) = @_;
638 my $r404 = invalid_inbox($ctx, $inbox);
639 return $r404 if $r404;
640 my $self = $ctx->{www};
641 my $css_map = $self->{-css_map} || stylesheets_prepare($self, '');
642 my $css = $css_map->{$key};
643 if (!defined($css) && $key eq 'userContent') {
644 my $env = $ctx->{env};
645 $css = PublicInbox::UserContent::sample($ctx->{-inbox}, $env);
647 defined $css or return r404();
648 my $h = [ 'Content-Length', bytes::length($css),
649 'Content-Type', 'text/css' ];
650 PublicInbox::GitHTTPBackend::cache_one_year($h);
651 [ 200, $h, [ $css ] ];
654 sub get_description {
655 my ($ctx, $inbox) = @_;
656 invalid_inbox($ctx, $inbox) || do {
657 my $d = $ctx->{-inbox}->description . "\n";
658 [ 200, [ 'Content-Length', bytes::length($d),
659 'Content-Type', 'text/plain' ], [ $d ] ];