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
19 use PublicInbox::Config;
20 use PublicInbox::Hval;
21 use URI::Escape qw(uri_unescape);
22 use PublicInbox::MID qw(mid_escape);
23 require PublicInbox::Git;
24 use PublicInbox::GitHTTPBackend;
25 use PublicInbox::UserContent;
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);
49 my $sn = $_[0]->{SCRIPT_NAME};
50 $path_re_cache{$sn} ||= do {
51 $sn = '/'.$sn unless index($sn, '/') == 0;
53 qr!\A(?:https?://[^/]+)?\Q$sn\E(/[^\?\#]+)!;
58 my ($self, $env) = @_;
59 my $ctx = { env => $env, www => $self };
61 # we don't care about multi-value
65 my ($k, $v) = split('=', $_, 2);
66 $v = uri_unescape($v // '');
67 # none of the keys we care about will need escaping
69 } split(/[&;]+/, $env->{QUERY_STRING});
71 # avoiding $env->{PATH_INFO} here since that's already decoded
72 my ($path_info) = ($env->{REQUEST_URI} =~ path_re($env));
73 $path_info //= $env->{PATH_INFO};
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 ($method !~ /\AGET|HEAD\z/) {
87 return r(405, 'Method Not Allowed');
90 # top-level indices and feeds
91 if ($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 # simple response for errors
184 sub r { [ $_[0], ['Content-Type' => 'text/plain'], [ join(' ', @_, "\n") ] ] }
186 sub news_cgit_fallback ($) {
188 my $www = $ctx->{www};
189 my $env = $ctx->{env};
190 my $res = $www->news_www->call($env);
191 $res->[0] == 404 ? $www->cgit->call($env) : $res;
194 # returns undef if valid, array ref response if invalid
195 sub invalid_inbox ($$) {
196 my ($ctx, $inbox) = @_;
197 my $ibx = $ctx->{www}->{pi_config}->lookup_name($inbox);
199 $ctx->{-inbox} = $ibx;
203 # sometimes linkifiers (not ours!) screw up automatic link
204 # generation and link things intended for nntp:// to https?://,
205 # so try to infer links and redirect them to the appropriate
207 news_cgit_fallback($ctx);
210 # returns undef if valid, array ref response if invalid
211 sub invalid_inbox_mid {
212 my ($ctx, $inbox, $mid_ue) = @_;
213 my $ret = invalid_inbox($ctx, $inbox);
216 my $mid = $ctx->{mid} = uri_unescape($mid_ue);
217 my $ibx = $ctx->{-inbox};
218 if ($mid =~ m!\A([a-f0-9]{2})([a-f0-9]{38})\z!) {
219 my ($x2, $x38) = ($1, $2);
220 # this is horrifically wasteful for legacy URLs:
221 my $str = $ctx->{-inbox}->msg_by_path("$x2/$x38") or return;
222 require Email::Simple;
223 my $s = Email::Simple->new($str);
224 $mid = PublicInbox::MID::mid_clean($s->header('Message-ID'));
225 return r301($ctx, $inbox, mid_escape($mid));
230 # /$INBOX/new.atom -> Atom feed, includes replies
233 require PublicInbox::Feed;
234 PublicInbox::Feed::generate($ctx);
237 # /$INBOX/new.html -> HTML only
240 require PublicInbox::Feed;
241 PublicInbox::Feed::new_html($ctx);
244 # /$INBOX/?r=$GIT_COMMIT -> HTML only
247 require PublicInbox::Feed;
248 if ($ctx->{env}->{QUERY_STRING} =~ /(?:\A|[&;])q=/) {
249 require PublicInbox::SearchView;
250 PublicInbox::SearchView::sres_top_html($ctx);
252 PublicInbox::Feed::generate_html_index($ctx);
256 # /$INBOX/$MESSAGE_ID/raw -> raw mbox
259 require PublicInbox::Mbox;
260 PublicInbox::Mbox::emit_raw($ctx) || r404($ctx);
263 # /$INBOX/$MESSAGE_ID/ -> HTML content (short quotes)
266 require PublicInbox::View;
267 PublicInbox::View::msg_page($ctx) || r404($ctx);
270 # /$INBOX/$MESSAGE_ID/t/
272 my ($ctx, $flat) = @_;
273 $ctx->{-inbox}->over or return need($ctx, 'Overview');
274 $ctx->{flat} = $flat;
275 require PublicInbox::View;
276 PublicInbox::View::thread_html($ctx);
279 # /$INBOX/_/text/$KEY/
280 # /$INBOX/_/text/$KEY/raw
281 # KEY may contain slashes
283 my ($ctx, $inbox, $key) = @_;
284 my $r404 = invalid_inbox($ctx, $inbox);
285 return $r404 if $r404;
287 require PublicInbox::WwwText;
288 PublicInbox::WwwText::get_text($ctx, $key);
291 # show git objects (blobs and commits)
292 # /$INBOX/_/$OBJECT_ID/show
293 # /$INBOX/_/${OBJECT_ID}_${FILENAME}
294 # KEY may contain slashes
295 sub get_vcs_object ($$$;$) {
296 my ($ctx, $inbox, $oid, $filename) = @_;
297 my $r404 = invalid_inbox($ctx, $inbox);
298 return $r404 if $r404;
299 require PublicInbox::ViewVCS;
300 PublicInbox::ViewVCS::show($ctx, $oid, $filename);
304 my ($ctx, $key) = @_;
305 my $val = $ctx->{$key};
306 (defined $val && $val ne '') or die "BUG: bad ctx, $key unusable";
311 my ($ctx, $extra) = @_;
313 <html><head><title>$extra not available for this
314 public-inbox</title><body><pre>$extra is not available for this public-inbox
315 <a href="../">Return to index</a></pre></body></html>
317 [ 501, [ 'Content-Type' => 'text/html; charset=UTF-8' ], [ $msg ] ];
320 # /$INBOX/$MESSAGE_ID/t.mbox -> thread as mbox
321 # /$INBOX/$MESSAGE_ID/t.mbox.gz -> thread as gzipped mbox
322 # note: I'm not a big fan of other compression formats since they're
323 # significantly more expensive on CPU than gzip and less-widely available,
324 # especially on older systems. Stick to zlib since that's what git uses.
325 sub get_thread_mbox {
326 my ($ctx, $sfx) = @_;
327 my $over = $ctx->{-inbox}->over or return need($ctx, 'Overview');
328 require PublicInbox::Mbox;
329 PublicInbox::Mbox::thread_mbox($ctx, $over, $sfx);
333 # /$INBOX/$MESSAGE_ID/t.atom -> thread as Atom feed
334 sub get_thread_atom {
336 $ctx->{-inbox}->over or return need($ctx, 'Overview');
337 require PublicInbox::Feed;
338 PublicInbox::Feed::generate_thread_atom($ctx);
341 sub legacy_redirects {
342 my ($ctx, $path_info) = @_;
344 # single-message pages
345 if ($path_info =~ m!$INBOX_RE/m/(\S+)/\z!o) {
347 } elsif ($path_info =~ m!$INBOX_RE/m/(\S+)/raw\z!o) {
348 r301($ctx, $1, $2, 'raw');
350 } elsif ($path_info =~ m!$INBOX_RE/f/(\S+)/\z!o) {
354 } elsif ($path_info =~ m!$INBOX_RE/t/(\S+)/\z!o) {
355 r301($ctx, $1, $2, 't/#u');
357 } elsif ($path_info =~ m!$INBOX_RE/t/(\S+)/mbox(\.gz)?\z!o) {
358 r301($ctx, $1, $2, "t.mbox$3");
360 # even older legacy redirects
361 } elsif ($path_info =~ m!$INBOX_RE/m/(\S+)\.html\z!o) {
364 } elsif ($path_info =~ m!$INBOX_RE/t/(\S+)\.html\z!o) {
365 r301($ctx, $1, $2, 't/#u');
367 } elsif ($path_info =~ m!$INBOX_RE/f/(\S+)\.html\z!o) {
370 } elsif ($path_info =~ m!$INBOX_RE/(?:m|f)/(\S+)\.txt\z!o) {
371 r301($ctx, $1, $2, 'raw');
373 } elsif ($path_info =~ m!$INBOX_RE/t/(\S+)(\.mbox(?:\.gz)?)\z!o) {
374 r301($ctx, $1, $2, "t$3");
376 # legacy convenience redirects, order still matters
377 } elsif ($path_info =~ m!$INBOX_RE/m/(\S+)\z!o) {
379 } elsif ($path_info =~ m!$INBOX_RE/t/(\S+)\z!o) {
380 r301($ctx, $1, $2, 't/#u');
381 } elsif ($path_info =~ m!$INBOX_RE/f/(\S+)\z!o) {
384 # some Message-IDs have slashes in them and the HTTP server
385 # may try to be clever and unescape them :<
386 } elsif ($path_info =~ m!$INBOX_RE/(\S+/\S+)/$END_RE\z!o) {
387 msg_page($ctx, $1, $2, $3);
389 # in case people leave off the trailing slash:
390 } elsif ($path_info =~ m!$INBOX_RE/(\S+/\S+)/(T|t)\z!o) {
391 r301($ctx, $1, $2, $3 eq 't' ? 't/#u' : $3);
392 } elsif ($path_info =~ m!$INBOX_RE/(\S+/\S+)/f\z!o) {
395 news_cgit_fallback($ctx);
400 my ($ctx, $inbox, $mid_ue, $suffix) = @_;
401 my $ibx = $ctx->{-inbox};
403 my $r404 = invalid_inbox($ctx, $inbox);
404 return $r404 if $r404;
405 $ibx = $ctx->{-inbox};
407 my $url = $ibx->base_url($ctx->{env});
408 my $qs = $ctx->{env}->{QUERY_STRING};
409 if (defined $mid_ue) {
410 # common, and much nicer as '@' than '%40':
411 $mid_ue =~ s/%40/@/g;
412 $url .= $mid_ue . '/';
414 $url .= $suffix if (defined $suffix);
415 $url .= "?$qs" if $qs ne '';
418 [ Location => $url, 'Content-Type' => 'text/plain' ],
419 [ "Redirecting to $url\n" ] ]
423 my ($ctx, $inbox, $mid_ue, $e) = @_;
425 $ret = invalid_inbox_mid($ctx, $inbox, $mid_ue) and return $ret;
426 '' eq $e and return get_mid_html($ctx);
427 'T/' eq $e and return get_thread($ctx, 1);
428 't/' eq $e and return get_thread($ctx);
429 't.atom' eq $e and return get_thread_atom($ctx);
430 't.mbox' eq $e and return get_thread_mbox($ctx);
431 't.mbox.gz' eq $e and return get_thread_mbox($ctx, '.gz');
432 'raw' eq $e and return get_mid_txt($ctx);
434 # legacy, but no redirect for compatibility:
435 'f/' eq $e and return get_mid_html($ctx);
440 my ($ctx, $epoch, $path) = @_;
441 my $env = $ctx->{env};
442 my $ibx = $ctx->{-inbox};
443 my $git = defined $epoch ? $ibx->git_epoch($epoch) : $ibx->git;
444 $git ? PublicInbox::GitHTTPBackend::serve($env, $git, $path) : r404();
449 if ($ctx->{env}->{QUERY_STRING} =~ /(?:\A|[&;])q=/) {
450 $ctx->{-inbox}->search or return need($ctx, 'search');
451 require PublicInbox::SearchView;
452 return PublicInbox::SearchView::mbox_results($ctx);
457 sub serve_mbox_range {
458 my ($ctx, $inbox, $range) = @_;
459 invalid_inbox($ctx, $inbox) || eval {
460 require PublicInbox::Mbox;
461 PublicInbox::Mbox::emit_range($ctx, $range);
467 $self->{news_www} ||= do {
468 require PublicInbox::NewsWWW;
469 PublicInbox::NewsWWW->new($self->{pi_config});
475 $self->{cgit} ||= do {
476 my $pi_config = $self->{pi_config};
478 if (defined($pi_config->{'publicinbox.cgitrc'})) {
479 require PublicInbox::Cgit;
480 PublicInbox::Cgit->new($pi_config);
482 Plack::Util::inline_object(call => sub { r404() });
489 $self->{www_listing} ||= do {
490 require PublicInbox::WwwListing;
491 PublicInbox::WwwListing->new($self);
495 # GET $INBOX/manifest.js.gz
496 sub get_inbox_manifest ($$$) {
497 my ($ctx, $inbox, $key) = @_;
498 my $r404 = invalid_inbox($ctx, $inbox);
499 return $r404 if $r404;
500 require PublicInbox::WwwListing;
501 PublicInbox::WwwListing::js($ctx->{env}, [$ctx->{-inbox}]);
505 my ($ctx, $idx, $fn) = @_;
506 require PublicInbox::WwwAttach;
507 PublicInbox::WwwAttach::get_attach($ctx, $idx, $fn);
510 # User-generated content (UGC) may have excessively long lines
511 # and screw up rendering on some browsers, so we use pre-wrap.
513 # We also force everything to the same scaled font-size because GUI
514 # browsers (tested both Firefox and surf (webkit)) uses a larger font
515 # for the Search <form> element than the rest of the page. Font size
516 # uniformity is important to people who rely on gigantic fonts.
517 # Finally, we use monospace to ensure the Search field and button
518 # has the same size and spacing as everything else which is
519 # <pre>-formatted anyways.
520 our $STYLE = 'pre{white-space:pre-wrap}*{font-size:100%;font-family:monospace}';
522 sub stylesheets_prepare ($$) {
523 my ($self, $upfx) = @_;
525 require CSS::Minifier;
526 sub { CSS::Minifier::minify(input => $_[0]) };
528 require CSS::Minifier::XS;
529 sub { CSS::Minifier::XS::minify($_[0]) };
533 my $stylesheets = $self->{pi_config}->{css} || [];
537 foreach my $s (@$stylesheets) {
540 foreach my $k (qw(media title href)) {
541 if (s/\s*$k='([^']+)'// || s/\s*$k=(\S+)//) {
546 if (defined $attr->{href}) {
550 my ($key) = (m!([^/]+?)(?:\.css)?\z!i);
551 if ($key !~ /\A[a-zA-Z0-9_\-\.]+\z/) {
552 warn "ignoring $fn, non-ASCII word character\n";
555 open(my $fh, '<', $fn) or do {
556 warn "failed to open $fn: $!\n";
560 my $local = do { local $/; <$fh> };
561 if ($local =~ /\S/) {
562 $ctime = sprintf('%x',(stat($fh))[10]);
563 $local = $mini->($local);
566 # do not let BOFHs override userContent.css:
567 if ($local =~ /!\s*important\b/i) {
568 warn "ignoring $fn since it uses `!important'\n";
572 $css_map->{$key} = $local;
573 $attr->{href} = "$upfx$key.css?$ctime";
574 if (defined($attr->{title})) {
576 } elsif (($attr->{media}||'screen') eq 'screen') {
577 $attr->{-inline} = $local;
583 my $buf = "<style>$STYLE";
585 my @ext; # for media=print and whatnot
586 foreach my $attr (@$links) {
587 if (defined(my $str = delete $attr->{-inline})) {
598 foreach my $attr (@$links) {
599 delete $attr->{-inline};
600 $buf .= "<link\ntype=text/css\nrel=stylesheet";
601 while (my ($k, $v) = each %$attr) {
602 $v = qq{"$v"} if $v =~ /[\s=]/;
607 $self->{"-style-$upfx"} = $buf;
609 $self->{-style_inline} = $buf;
611 $self->{-css_map} = $css_map;
614 # returns an HTML fragment with <style> or <link> tags in them
615 # Called by WwwStream by nearly every HTML page
617 my ($self, $upfx) = @_;
618 $self->{-style_inline} || $self->{"-style-$upfx"} || do {
619 stylesheets_prepare($self, $upfx);
620 $self->{-style_inline} || $self->{"-style-$upfx"}
624 # /$INBOX/$KEY.css endpoint
625 # CSS is configured globally for all inboxes, but we access them on
626 # a per-inbox basis. This allows administrators to setup per-inbox
627 # static routes to intercept the request before it hits PSGI
629 my ($ctx, $inbox, $key) = @_;
630 my $r404 = invalid_inbox($ctx, $inbox);
631 return $r404 if $r404;
632 my $self = $ctx->{www};
633 my $css_map = $self->{-css_map} || stylesheets_prepare($self, '');
634 my $css = $css_map->{$key};
635 if (!defined($css) && $key eq 'userContent') {
636 my $env = $ctx->{env};
637 $css = PublicInbox::UserContent::sample($ctx->{-inbox}, $env);
639 defined $css or return r404();
640 my $h = [ 'Content-Length', bytes::length($css),
641 'Content-Type', 'text/css' ];
642 PublicInbox::GitHTTPBackend::cache_one_year($h);
643 [ 200, $h, [ $css ] ];