1 # Copyright (C) 2014-2019 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 require PublicInbox::Git;
23 use PublicInbox::GitHTTPBackend;
24 use PublicInbox::UserContent;
25 use PublicInbox::WwwStatic qw(r);
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 $OID_RE = qr![a-f0-9]{7,40}!;
35 my ($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);
46 # PATH_INFO is decoded, and we want the undecoded original
48 sub path_info_raw ($) {
50 my $sn = $env->{SCRIPT_NAME};
51 my $re = $path_re_cache{$sn} ||= do {
52 $sn = '/'.$sn unless index($sn, '/') == 0;
54 qr!\A(?:https?://[^/]+)?\Q$sn\E(/[^\?\#]+)!;
56 $env->{REQUEST_URI} =~ $re ? $1 : $env->{PATH_INFO};
60 my ($self, $env) = @_;
61 my $ctx = { env => $env, www => $self };
63 # we don't care about multi-value
67 my ($k, $v) = split('=', $_, 2);
68 $v = uri_unescape($v // '');
69 # none of the keys we care about will need escaping
71 } split(/[&;]+/, $env->{QUERY_STRING});
73 my $path_info = path_info_raw($env);
74 my $method = $env->{REQUEST_METHOD};
76 if ($method eq 'POST') {
77 if ($path_info =~ m!$INBOX_RE/(?:(?:git/)?([0-9]+)(?:\.git)?/)?
78 (git-upload-pack)\z!x) {
79 my ($epoch, $path) = ($2, $3);
80 return invalid_inbox($ctx, $1) ||
81 serve_git($ctx, $epoch, $path);
82 } elsif ($path_info =~ m!$INBOX_RE/!o) {
83 return invalid_inbox($ctx, $1) || mbox_results($ctx);
86 elsif ($method !~ /\AGET|HEAD\z/) {
90 # top-level indices and feeds
91 if ($path_info eq '/' || $path_info eq '/manifest.js.gz') {
92 www_listing($self)->call($env);
93 } elsif ($path_info =~ m!$INBOX_RE\z!o) {
94 invalid_inbox($ctx, $1) || r301($ctx, $1);
95 } elsif ($path_info =~ m!$INBOX_RE(?:/|/index\.html)?\z!o) {
96 invalid_inbox($ctx, $1) || get_index($ctx);
97 } elsif ($path_info =~ m!$INBOX_RE/(?:atom\.xml|new\.atom)\z!o) {
98 invalid_inbox($ctx, $1) || get_atom($ctx);
99 } elsif ($path_info =~ m!$INBOX_RE/new\.html\z!o) {
100 invalid_inbox($ctx, $1) || get_new($ctx);
101 } elsif ($path_info =~ m!$INBOX_RE/(?:(?:git/)?([0-9]+)(?:\.git)?/)?
102 ($PublicInbox::GitHTTPBackend::ANY)\z!ox) {
103 my ($epoch, $path) = ($2, $3);
104 invalid_inbox($ctx, $1) || serve_git($ctx, $epoch, $path);
105 } elsif ($path_info =~ m!$INBOX_RE/([a-zA-Z0-9_\-]+).mbox\.gz\z!o) {
106 serve_mbox_range($ctx, $1, $2);
107 } elsif ($path_info =~ m!$INBOX_RE/$MID_RE/$END_RE\z!o) {
108 msg_page($ctx, $1, $2, $3);
110 } elsif ($path_info =~ m!$INBOX_RE/$MID_RE/$ATTACH_RE\z!o) {
111 my ($idx, $fn) = ($3, $4);
112 invalid_inbox_mid($ctx, $1, $2) || get_attach($ctx, $idx, $fn);
113 # in case people leave off the trailing slash:
114 } elsif ($path_info =~ m!$INBOX_RE/$MID_RE/(T|t)\z!o) {
115 my ($inbox, $mid_ue, $suffix) = ($1, $2, $3);
116 $suffix .= $suffix =~ /\A[tT]\z/ ? '/#u' : '/';
117 r301($ctx, $inbox, $mid_ue, $suffix);
119 } elsif ($path_info =~ m!$INBOX_RE/$MID_RE/R/?\z!o) {
120 my ($inbox, $mid_ue) = ($1, $2);
121 r301($ctx, $inbox, $mid_ue, '#R');
123 } elsif ($path_info =~ m!$INBOX_RE/$MID_RE/f/?\z!o) {
125 } elsif ($path_info =~ m!$INBOX_RE/_/text(?:/(.*))?\z!o) {
126 get_text($ctx, $1, $2);
127 } elsif ($path_info =~ m!$INBOX_RE/([a-zA-Z0-9_\-\.]+)\.css\z!o) {
128 get_css($ctx, $1, $2);
129 } elsif ($path_info =~ m!$INBOX_RE/manifest\.js\.gz\z!o) {
130 get_inbox_manifest($ctx, $1, $2);
131 } elsif ($path_info =~ m!$INBOX_RE/($OID_RE)/s/\z!o) {
132 get_vcs_object($ctx, $1, $2);
133 } elsif ($path_info =~ m!$INBOX_RE/($OID_RE)/s/
134 ($PublicInbox::Hval::FN)\z!ox) {
135 get_vcs_object($ctx, $1, $2, $3);
136 } elsif ($path_info =~ m!$INBOX_RE/($OID_RE)/s\z!o) {
137 r301($ctx, $1, $2, 's/');
138 # convenience redirects order matters
139 } elsif ($path_info =~ m!$INBOX_RE/([^/]{2,})\z!o) {
143 legacy_redirects($ctx, $path_info);
147 # for CoW-friendliness, MOOOOO!
150 require PublicInbox::Feed;
151 require PublicInbox::View;
152 require PublicInbox::SearchThread;
153 require PublicInbox::MIME;
157 require PublicInbox::Search;
158 PublicInbox::Search::load_xapian();
160 foreach (qw(PublicInbox::SearchView
161 PublicInbox::Mbox IO::Compress::Gzip
162 PublicInbox::NewsWWW)) {
167 $self->stylesheets_prepare($_) for ('', '../', '../../');
172 # private functions below
176 if ($ctx && $ctx->{mid}) {
177 require PublicInbox::ExtMsg;
178 return PublicInbox::ExtMsg::ext_msg($ctx);
183 sub news_cgit_fallback ($) {
185 my $www = $ctx->{www};
186 my $env = $ctx->{env};
187 my $res = $www->news_www->call($env);
188 $res->[0] == 404 ? $www->cgit->call($env) : $res;
191 # returns undef if valid, array ref response if invalid
192 sub invalid_inbox ($$) {
193 my ($ctx, $inbox) = @_;
194 my $ibx = $ctx->{www}->{pi_config}->lookup_name($inbox);
196 $ctx->{-inbox} = $ibx;
200 # sometimes linkifiers (not ours!) screw up automatic link
201 # generation and link things intended for nntp:// to https?://,
202 # so try to infer links and redirect them to the appropriate
204 news_cgit_fallback($ctx);
207 # returns undef if valid, array ref response if invalid
208 sub invalid_inbox_mid {
209 my ($ctx, $inbox, $mid_ue) = @_;
210 my $ret = invalid_inbox($ctx, $inbox);
213 my $mid = $ctx->{mid} = uri_unescape($mid_ue);
214 my $ibx = $ctx->{-inbox};
215 if ($mid =~ m!\A([a-f0-9]{2})([a-f0-9]{38})\z!) {
216 my ($x2, $x38) = ($1, $2);
217 # this is horrifically wasteful for legacy URLs:
218 my $str = $ctx->{-inbox}->msg_by_path("$x2/$x38") or return;
219 require Email::Simple;
220 my $s = Email::Simple->new($str);
221 $mid = PublicInbox::MID::mid_clean($s->header('Message-ID'));
222 return r301($ctx, $inbox, mid_escape($mid));
227 # /$INBOX/new.atom -> Atom feed, includes replies
230 require PublicInbox::Feed;
231 PublicInbox::Feed::generate($ctx);
234 # /$INBOX/new.html -> HTML only
237 require PublicInbox::Feed;
238 PublicInbox::Feed::new_html($ctx);
241 # /$INBOX/?r=$GIT_COMMIT -> HTML only
244 require PublicInbox::Feed;
245 if ($ctx->{env}->{QUERY_STRING} =~ /(?:\A|[&;])q=/) {
246 require PublicInbox::SearchView;
247 PublicInbox::SearchView::sres_top_html($ctx);
249 PublicInbox::Feed::generate_html_index($ctx);
253 # /$INBOX/$MESSAGE_ID/raw -> raw mbox
256 require PublicInbox::Mbox;
257 PublicInbox::Mbox::emit_raw($ctx) || r404($ctx);
260 # /$INBOX/$MESSAGE_ID/ -> HTML content (short quotes)
263 require PublicInbox::View;
264 PublicInbox::View::msg_page($ctx) || r404($ctx);
267 # /$INBOX/$MESSAGE_ID/t/
269 my ($ctx, $flat) = @_;
270 $ctx->{-inbox}->over or return need($ctx, 'Overview');
271 $ctx->{flat} = $flat;
272 require PublicInbox::View;
273 PublicInbox::View::thread_html($ctx);
276 # /$INBOX/_/text/$KEY/
277 # /$INBOX/_/text/$KEY/raw
278 # KEY may contain slashes
280 my ($ctx, $inbox, $key) = @_;
281 my $r404 = invalid_inbox($ctx, $inbox);
282 return $r404 if $r404;
284 require PublicInbox::WwwText;
285 PublicInbox::WwwText::get_text($ctx, $key);
288 # show git objects (blobs and commits)
289 # /$INBOX/_/$OBJECT_ID/show
290 # /$INBOX/_/${OBJECT_ID}_${FILENAME}
291 # KEY may contain slashes
292 sub get_vcs_object ($$$;$) {
293 my ($ctx, $inbox, $oid, $filename) = @_;
294 my $r404 = invalid_inbox($ctx, $inbox);
295 return $r404 if $r404;
296 require PublicInbox::ViewVCS;
297 PublicInbox::ViewVCS::show($ctx, $oid, $filename);
301 my ($ctx, $extra) = @_;
303 <html><head><title>$extra not available for this
304 public-inbox</title><body><pre>$extra is not available for this public-inbox
305 <a href="../">Return to index</a></pre></body></html>
307 [ 501, [ 'Content-Type' => 'text/html; charset=UTF-8' ], [ $msg ] ];
310 # /$INBOX/$MESSAGE_ID/t.mbox -> thread as mbox
311 # /$INBOX/$MESSAGE_ID/t.mbox.gz -> thread as gzipped mbox
312 # note: I'm not a big fan of other compression formats since they're
313 # significantly more expensive on CPU than gzip and less-widely available,
314 # especially on older systems. Stick to zlib since that's what git uses.
315 sub get_thread_mbox {
316 my ($ctx, $sfx) = @_;
317 my $over = $ctx->{-inbox}->over or return need($ctx, 'Overview');
318 require PublicInbox::Mbox;
319 PublicInbox::Mbox::thread_mbox($ctx, $over, $sfx);
323 # /$INBOX/$MESSAGE_ID/t.atom -> thread as Atom feed
324 sub get_thread_atom {
326 $ctx->{-inbox}->over or return need($ctx, 'Overview');
327 require PublicInbox::Feed;
328 PublicInbox::Feed::generate_thread_atom($ctx);
331 sub legacy_redirects {
332 my ($ctx, $path_info) = @_;
334 # single-message pages
335 if ($path_info =~ m!$INBOX_RE/m/(\S+)/\z!o) {
337 } elsif ($path_info =~ m!$INBOX_RE/m/(\S+)/raw\z!o) {
338 r301($ctx, $1, $2, 'raw');
340 } elsif ($path_info =~ m!$INBOX_RE/f/(\S+)/\z!o) {
344 } elsif ($path_info =~ m!$INBOX_RE/t/(\S+)/\z!o) {
345 r301($ctx, $1, $2, 't/#u');
347 } elsif ($path_info =~ m!$INBOX_RE/t/(\S+)/mbox(\.gz)?\z!o) {
348 r301($ctx, $1, $2, "t.mbox$3");
350 # even older legacy redirects
351 } elsif ($path_info =~ m!$INBOX_RE/m/(\S+)\.html\z!o) {
354 } elsif ($path_info =~ m!$INBOX_RE/t/(\S+)\.html\z!o) {
355 r301($ctx, $1, $2, 't/#u');
357 } elsif ($path_info =~ m!$INBOX_RE/f/(\S+)\.html\z!o) {
360 } elsif ($path_info =~ m!$INBOX_RE/(?:m|f)/(\S+)\.txt\z!o) {
361 r301($ctx, $1, $2, 'raw');
363 } elsif ($path_info =~ m!$INBOX_RE/t/(\S+)(\.mbox(?:\.gz)?)\z!o) {
364 r301($ctx, $1, $2, "t$3");
366 # legacy convenience redirects, order still matters
367 } elsif ($path_info =~ m!$INBOX_RE/m/(\S+)\z!o) {
369 } elsif ($path_info =~ m!$INBOX_RE/t/(\S+)\z!o) {
370 r301($ctx, $1, $2, 't/#u');
371 } elsif ($path_info =~ m!$INBOX_RE/f/(\S+)\z!o) {
374 # some Message-IDs have slashes in them and the HTTP server
375 # may try to be clever and unescape them :<
376 } elsif ($path_info =~ m!$INBOX_RE/(\S+/\S+)/$END_RE\z!o) {
377 msg_page($ctx, $1, $2, $3);
379 # in case people leave off the trailing slash:
380 } elsif ($path_info =~ m!$INBOX_RE/(\S+/\S+)/(T|t)\z!o) {
381 r301($ctx, $1, $2, $3 eq 't' ? 't/#u' : $3);
382 } elsif ($path_info =~ m!$INBOX_RE/(\S+/\S+)/f\z!o) {
385 news_cgit_fallback($ctx);
390 my ($ctx, $inbox, $mid_ue, $suffix) = @_;
391 my $ibx = $ctx->{-inbox};
393 my $r404 = invalid_inbox($ctx, $inbox);
394 return $r404 if $r404;
395 $ibx = $ctx->{-inbox};
397 my $url = $ibx->base_url($ctx->{env});
398 my $qs = $ctx->{env}->{QUERY_STRING};
399 if (defined $mid_ue) {
400 # common, and much nicer as '@' than '%40':
401 $mid_ue =~ s/%40/@/g;
402 $url .= $mid_ue . '/';
404 $url .= $suffix if (defined $suffix);
405 $url .= "?$qs" if $qs ne '';
408 [ Location => $url, 'Content-Type' => 'text/plain' ],
409 [ "Redirecting to $url\n" ] ]
413 my ($ctx, $inbox, $mid_ue, $e) = @_;
415 $ret = invalid_inbox_mid($ctx, $inbox, $mid_ue) and return $ret;
416 '' eq $e and return get_mid_html($ctx);
417 'T/' eq $e and return get_thread($ctx, 1);
418 't/' eq $e and return get_thread($ctx);
419 't.atom' eq $e and return get_thread_atom($ctx);
420 't.mbox' eq $e and return get_thread_mbox($ctx);
421 't.mbox.gz' eq $e and return get_thread_mbox($ctx, '.gz');
422 'raw' eq $e and return get_mid_txt($ctx);
424 # legacy, but no redirect for compatibility:
425 'f/' eq $e and return get_mid_html($ctx);
430 my ($ctx, $epoch, $path) = @_;
431 my $env = $ctx->{env};
432 my $ibx = $ctx->{-inbox};
433 my $git = defined $epoch ? $ibx->git_epoch($epoch) : $ibx->git;
434 $git ? PublicInbox::GitHTTPBackend::serve($env, $git, $path) : r404();
439 if ($ctx->{env}->{QUERY_STRING} =~ /(?:\A|[&;])q=/) {
440 $ctx->{-inbox}->search or return need($ctx, 'search');
441 require PublicInbox::SearchView;
442 return PublicInbox::SearchView::mbox_results($ctx);
447 sub serve_mbox_range {
448 my ($ctx, $inbox, $range) = @_;
449 invalid_inbox($ctx, $inbox) || eval {
450 require PublicInbox::Mbox;
451 PublicInbox::Mbox::emit_range($ctx, $range);
457 $self->{news_www} ||= do {
458 require PublicInbox::NewsWWW;
459 PublicInbox::NewsWWW->new($self->{pi_config});
465 $self->{cgit} ||= do {
466 my $pi_config = $self->{pi_config};
468 if (defined($pi_config->{'publicinbox.cgitrc'})) {
469 require PublicInbox::Cgit;
470 PublicInbox::Cgit->new($pi_config);
473 Plack::Util::inline_object(call => sub { r404() });
480 $self->{www_listing} ||= do {
481 require PublicInbox::WwwListing;
482 PublicInbox::WwwListing->new($self);
486 # GET $INBOX/manifest.js.gz
487 sub get_inbox_manifest ($$$) {
488 my ($ctx, $inbox, $key) = @_;
489 my $r404 = invalid_inbox($ctx, $inbox);
490 return $r404 if $r404;
491 require PublicInbox::WwwListing;
492 PublicInbox::WwwListing::js($ctx->{env}, [$ctx->{-inbox}]);
496 my ($ctx, $idx, $fn) = @_;
497 require PublicInbox::WwwAttach;
498 PublicInbox::WwwAttach::get_attach($ctx, $idx, $fn);
501 # User-generated content (UGC) may have excessively long lines
502 # and screw up rendering on some browsers, so we use pre-wrap.
504 # We also force everything to the same scaled font-size because GUI
505 # browsers (tested both Firefox and surf (webkit)) uses a larger font
506 # for the Search <form> element than the rest of the page. Font size
507 # uniformity is important to people who rely on gigantic fonts.
508 # Finally, we use monospace to ensure the Search field and button
509 # has the same size and spacing as everything else which is
510 # <pre>-formatted anyways.
511 our $STYLE = 'pre{white-space:pre-wrap}*{font-size:100%;font-family:monospace}';
513 sub stylesheets_prepare ($$) {
514 my ($self, $upfx) = @_;
516 require CSS::Minifier;
517 sub { CSS::Minifier::minify(input => $_[0]) };
519 require CSS::Minifier::XS;
520 sub { CSS::Minifier::XS::minify($_[0]) };
524 my $stylesheets = $self->{pi_config}->{css} || [];
528 foreach my $s (@$stylesheets) {
531 foreach my $k (qw(media title href)) {
532 if (s/\s*$k='([^']+)'// || s/\s*$k=(\S+)//) {
537 if (defined $attr->{href}) {
541 my ($key) = (m!([^/]+?)(?:\.css)?\z!i);
542 if ($key !~ /\A[a-zA-Z0-9_\-\.]+\z/) {
543 warn "ignoring $fn, non-ASCII word character\n";
546 open(my $fh, '<', $fn) or do {
547 warn "failed to open $fn: $!\n";
551 my $local = do { local $/; <$fh> };
552 if ($local =~ /\S/) {
553 $ctime = sprintf('%x',(stat($fh))[10]);
554 $local = $mini->($local);
557 # do not let BOFHs override userContent.css:
558 if ($local =~ /!\s*important\b/i) {
559 warn "ignoring $fn since it uses `!important'\n";
563 $css_map->{$key} = $local;
564 $attr->{href} = "$upfx$key.css?$ctime";
565 if (defined($attr->{title})) {
567 } elsif (($attr->{media}||'screen') eq 'screen') {
568 $attr->{-inline} = $local;
574 my $buf = "<style>$STYLE";
576 my @ext; # for media=print and whatnot
577 foreach my $attr (@$links) {
578 if (defined(my $str = delete $attr->{-inline})) {
589 foreach my $attr (@$links) {
590 delete $attr->{-inline};
591 $buf .= "<link\ntype=text/css\nrel=stylesheet";
592 while (my ($k, $v) = each %$attr) {
593 $v = qq{"$v"} if $v =~ /[\s=]/;
598 $self->{"-style-$upfx"} = $buf;
600 $self->{-style_inline} = $buf;
602 $self->{-css_map} = $css_map;
605 # returns an HTML fragment with <style> or <link> tags in them
606 # Called by WwwStream by nearly every HTML page
608 my ($self, $upfx) = @_;
609 $self->{-style_inline} || $self->{"-style-$upfx"} || do {
610 stylesheets_prepare($self, $upfx);
611 $self->{-style_inline} || $self->{"-style-$upfx"}
615 # /$INBOX/$KEY.css endpoint
616 # CSS is configured globally for all inboxes, but we access them on
617 # a per-inbox basis. This allows administrators to setup per-inbox
618 # static routes to intercept the request before it hits PSGI
620 my ($ctx, $inbox, $key) = @_;
621 my $r404 = invalid_inbox($ctx, $inbox);
622 return $r404 if $r404;
623 my $self = $ctx->{www};
624 my $css_map = $self->{-css_map} || stylesheets_prepare($self, '');
625 my $css = $css_map->{$key};
626 if (!defined($css) && $key eq 'userContent') {
627 my $env = $ctx->{env};
628 $css = PublicInbox::UserContent::sample($ctx->{-inbox}, $env);
630 defined $css or return r404();
631 my $h = [ 'Content-Length', bytes::length($css),
632 'Content-Type', 'text/css' ];
633 PublicInbox::GitHTTPBackend::cache_one_year($h);
634 [ 200, $h, [ $css ] ];