ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/AnyEvent-HTTP/HTTP.pm
(Generate patch)

Comparing AnyEvent-HTTP/HTTP.pm (file contents):
Revision 1.29 by root, Wed Oct 22 23:28:11 2008 UTC vs.
Revision 1.134 by root, Fri Sep 7 22:11:31 2018 UTC

15This module is an L<AnyEvent> user, you need to make sure that you use and 15This module is an L<AnyEvent> user, you need to make sure that you use and
16run a supported event loop. 16run a supported event loop.
17 17
18This module implements a simple, stateless and non-blocking HTTP 18This module implements a simple, stateless and non-blocking HTTP
19client. It supports GET, POST and other request methods, cookies and more, 19client. It supports GET, POST and other request methods, cookies and more,
20all on a very low level. It can follow redirects supports proxies and 20all on a very low level. It can follow redirects, supports proxies, and
21automatically limits the number of connections to the values specified in 21automatically limits the number of connections to the values specified in
22the RFC. 22the RFC.
23 23
24It should generally be a "good client" that is enough for most HTTP 24It should generally be a "good client" that is enough for most HTTP
25tasks. Simple tasks should be simple, but complex tasks should still be 25tasks. Simple tasks should be simple, but complex tasks should still be
36 36
37=cut 37=cut
38 38
39package AnyEvent::HTTP; 39package AnyEvent::HTTP;
40 40
41use strict; 41use common::sense;
42no warnings;
43 42
44use Carp; 43use Errno ();
45 44
46use AnyEvent (); 45use AnyEvent 5.0 ();
47use AnyEvent::Util (); 46use AnyEvent::Util ();
48use AnyEvent::Socket ();
49use AnyEvent::Handle (); 47use AnyEvent::Handle ();
50 48
51use base Exporter::; 49use base Exporter::;
52 50
53our $VERSION = '1.05'; 51our $VERSION = 2.24;
54 52
55our @EXPORT = qw(http_get http_post http_head http_request); 53our @EXPORT = qw(http_get http_post http_head http_request);
56 54
57our $USERAGENT = "Mozilla/5.0 (compatible; AnyEvent::HTTP/$VERSION; +http://software.schmorp.de/pkg/AnyEvent)"; 55our $USERAGENT = "Mozilla/5.0 (compatible; U; AnyEvent-HTTP/$VERSION; +http://software.schmorp.de/pkg/AnyEvent)";
58our $MAX_RECURSE = 10; 56our $MAX_RECURSE = 10;
59our $MAX_PERSISTENT = 8;
60our $PERSISTENT_TIMEOUT = 2; 57our $PERSISTENT_TIMEOUT = 3;
61our $TIMEOUT = 300; 58our $TIMEOUT = 300;
62 59our $MAX_PER_HOST = 4; # changing this is evil
63# changing these is evil
64our $MAX_PERSISTENT_PER_HOST = 2;
65our $MAX_PER_HOST = 4;
66 60
67our $PROXY; 61our $PROXY;
68our $ACTIVE = 0; 62our $ACTIVE = 0;
69 63
70my %KA_COUNT; # number of open keep-alive connections per host 64my %KA_CACHE; # indexed by uhost currently, points to [$handle...] array
71my %CO_SLOT; # number of open connections, and wait queue, per host 65my %CO_SLOT; # number of open connections, and wait queue, per host
72 66
73=item http_get $url, key => value..., $cb->($data, $headers) 67=item http_get $url, key => value..., $cb->($data, $headers)
74 68
75Executes an HTTP-GET request. See the http_request function for details on 69Executes an HTTP-GET request. See the http_request function for details on
92must be an absolute http or https URL. 86must be an absolute http or https URL.
93 87
94When called in void context, nothing is returned. In other contexts, 88When called in void context, nothing is returned. In other contexts,
95C<http_request> returns a "cancellation guard" - you have to keep the 89C<http_request> returns a "cancellation guard" - you have to keep the
96object at least alive until the callback get called. If the object gets 90object at least alive until the callback get called. If the object gets
97destroyed before the callbakc is called, the request will be cancelled. 91destroyed before the callback is called, the request will be cancelled.
98 92
99The callback will be called with the response data as first argument 93The callback will be called with the response body data as first argument
100(or C<undef> if it wasn't available due to errors), and a hash-ref with 94(or C<undef> if an error occurred), and a hash-ref with response headers
101response headers as second argument. 95(and trailers) as second argument.
102 96
103All the headers in that hash are lowercased. In addition to the response 97All the headers in that hash are lowercased. In addition to the response
104headers, the "pseudo-headers" C<HTTPVersion>, C<Status> and C<Reason> 98headers, the "pseudo-headers" (uppercase to avoid clashing with possible
99response headers) C<HTTPVersion>, C<Status> and C<Reason> contain the
105contain the three parts of the HTTP Status-Line of the same name. The 100three parts of the HTTP Status-Line of the same name. If an error occurs
101during the body phase of a request, then the original C<Status> and
102C<Reason> values from the header are available as C<OrigStatus> and
103C<OrigReason>.
104
106pseudo-header C<URL> contains the original URL (which can differ from the 105The pseudo-header C<URL> contains the actual URL (which can differ from
107requested URL when following redirects). 106the requested URL when following redirects - for example, you might get
107an error that your URL scheme is not supported even though your URL is a
108valid http URL because it redirected to an ftp URL, in which case you can
109look at the URL pseudo header).
108 110
111The pseudo-header C<Redirect> only exists when the request was a result
112of an internal redirect. In that case it is an array reference with
113the C<($data, $headers)> from the redirect response. Note that this
114response could in turn be the result of a redirect itself, and C<<
115$headers->{Redirect}[1]{Redirect} >> will then contain the original
116response, and so on.
117
109If the server sends a header multiple lines, then their contents will be 118If the server sends a header multiple times, then their contents will be
110joined together with C<\x00>. 119joined together with a comma (C<,>), as per the HTTP spec.
111 120
112If an internal error occurs, such as not being able to resolve a hostname, 121If an internal error occurs, such as not being able to resolve a hostname,
113then C<$data> will be C<undef>, C<< $headers->{Status} >> will be C<599> 122then C<$data> will be C<undef>, C<< $headers->{Status} >> will be
114and the C<Reason> pseudo-header will contain an error message. 123C<590>-C<599> and the C<Reason> pseudo-header will contain an error
124message. Currently the following status codes are used:
125
126=over 4
127
128=item 595 - errors during connection establishment, proxy handshake.
129
130=item 596 - errors during TLS negotiation, request sending and header processing.
131
132=item 597 - errors during body receiving or processing.
133
134=item 598 - user aborted request via C<on_header> or C<on_body>.
135
136=item 599 - other, usually nonretryable, errors (garbled URL etc.).
137
138=back
115 139
116A typical callback might look like this: 140A typical callback might look like this:
117 141
118 sub { 142 sub {
119 my ($body, $hdr) = @_; 143 my ($body, $hdr) = @_;
130 154
131=over 4 155=over 4
132 156
133=item recurse => $count (default: $MAX_RECURSE) 157=item recurse => $count (default: $MAX_RECURSE)
134 158
135Whether to recurse requests or not, e.g. on redirects, authentication 159Whether to recurse requests or not, e.g. on redirects, authentication and
136retries and so on, and how often to do so. 160other retries and so on, and how often to do so.
161
162Only redirects to http and https URLs are supported. While most common
163redirection forms are handled entirely within this module, some require
164the use of the optional L<URI> module. If it is required but missing, then
165the request will fail with an error.
137 166
138=item headers => hashref 167=item headers => hashref
139 168
140The request headers to use. Currently, C<http_request> may provide its 169The request headers to use. Currently, C<http_request> may provide its own
141own C<Host:>, C<Content-Length:>, C<Connection:> and C<Cookie:> headers 170C<Host:>, C<Content-Length:>, C<Connection:> and C<Cookie:> headers and
142and will provide defaults for C<User-Agent:> and C<Referer:>. 171will provide defaults at least for C<TE:>, C<Referer:> and C<User-Agent:>
172(this can be suppressed by using C<undef> for these headers in which case
173they won't be sent at all).
174
175You really should provide your own C<User-Agent:> header value that is
176appropriate for your program - I wouldn't be surprised if the default
177AnyEvent string gets blocked by webservers sooner or later.
178
179Also, make sure that your headers names and values do not contain any
180embedded newlines.
143 181
144=item timeout => $seconds 182=item timeout => $seconds
145 183
146The time-out to use for various stages - each connect attempt will reset 184The time-out to use for various stages - each connect attempt will reset
147the timeout, as will read or write activity. Default timeout is 5 minutes. 185the timeout, as will read or write activity, i.e. this is not an overall
186timeout.
187
188Default timeout is 5 minutes.
148 189
149=item proxy => [$host, $port[, $scheme]] or undef 190=item proxy => [$host, $port[, $scheme]] or undef
150 191
151Use the given http proxy for all requests. If not specified, then the 192Use the given http proxy for all requests, or no proxy if C<undef> is
152default proxy (as specified by C<$ENV{http_proxy}>) is used. 193used.
153 194
154C<$scheme> must be either missing or C<http> for HTTP, or C<https> for 195C<$scheme> must be either missing or must be C<http> for HTTP.
155HTTPS. 196
197If not specified, then the default proxy is used (see
198C<AnyEvent::HTTP::set_proxy>).
199
200Currently, if your proxy requires authorization, you have to specify an
201appropriate "Proxy-Authorization" header in every request.
156 202
157=item body => $string 203=item body => $string
158 204
159The request body, usually empty. Will be-sent as-is (future versions of 205The request body, usually empty. Will be sent as-is (future versions of
160this module might offer more options). 206this module might offer more options).
161 207
162=item cookie_jar => $hash_ref 208=item cookie_jar => $hash_ref
163 209
164Passing this parameter enables (simplified) cookie-processing, loosely 210Passing this parameter enables (simplified) cookie-processing, loosely
165based on the original netscape specification. 211based on the original netscape specification.
166 212
167The C<$hash_ref> must be an (initially empty) hash reference which will 213The C<$hash_ref> must be an (initially empty) hash reference which
168get updated automatically. It is possible to save the cookie_jar to 214will get updated automatically. It is possible to save the cookie jar
169persistent storage with something like JSON or Storable, but this is not 215to persistent storage with something like JSON or Storable - see the
170recommended, as expire times are currently being ignored. 216C<AnyEvent::HTTP::cookie_jar_expire> function if you wish to remove
217expired or session-only cookies, and also for documentation on the format
218of the cookie jar.
171 219
172Note that this cookie implementation is not of very high quality, nor 220Note that this cookie implementation is not meant to be complete. If
173meant to be complete. If you want complete cookie management you have to 221you want complete cookie management you have to do that on your
174do that on your own. C<cookie_jar> is meant as a quick fix to get some 222own. C<cookie_jar> is meant as a quick fix to get most cookie-using sites
175cookie-using sites working. Cookies are a privacy disaster, do not use 223working. Cookies are a privacy disaster, do not use them unless required
176them unless required to. 224to.
225
226When cookie processing is enabled, the C<Cookie:> and C<Set-Cookie:>
227headers will be set and handled by this module, otherwise they will be
228left untouched.
229
230=item tls_ctx => $scheme | $tls_ctx
231
232Specifies the AnyEvent::TLS context to be used for https connections. This
233parameter follows the same rules as the C<tls_ctx> parameter to
234L<AnyEvent::Handle>, but additionally, the two strings C<low> or
235C<high> can be specified, which give you a predefined low-security (no
236verification, highest compatibility) and high-security (CA and common-name
237verification) TLS context.
238
239The default for this option is C<low>, which could be interpreted as "give
240me the page, no matter what".
241
242See also the C<sessionid> parameter.
243
244=item session => $string
245
246The module might reuse connections to the same host internally. Sometimes
247(e.g. when using TLS), you do not want to reuse connections from other
248sessions. This can be achieved by setting this parameter to some unique
249ID (such as the address of an object storing your state data, or the TLS
250context) - only connections using the same unique ID will be reused.
251
252=item on_prepare => $callback->($fh)
253
254In rare cases you need to "tune" the socket before it is used to
255connect (for example, to bind it on a given IP address). This parameter
256overrides the prepare callback passed to C<AnyEvent::Socket::tcp_connect>
257and behaves exactly the same way (e.g. it has to provide a
258timeout). See the description for the C<$prepare_cb> argument of
259C<AnyEvent::Socket::tcp_connect> for details.
260
261=item tcp_connect => $callback->($host, $service, $connect_cb, $prepare_cb)
262
263In even rarer cases you want total control over how AnyEvent::HTTP
264establishes connections. Normally it uses L<AnyEvent::Socket::tcp_connect>
265to do this, but you can provide your own C<tcp_connect> function -
266obviously, it has to follow the same calling conventions, except that it
267may always return a connection guard object.
268
269The connections made by this hook will be treated as equivalent to
270connecitons made the built-in way, specifically, they will be put into
271and taken from the persistent conneciton cache. If your C<$tcp_connect>
272function is incompatible with this kind of re-use, consider switching off
273C<persistent> connections and/or providing a C<session> identifier.
274
275There are probably lots of weird uses for this function, starting from
276tracing the hosts C<http_request> actually tries to connect, to (inexact
277but fast) host => IP address caching or even socks protocol support.
278
279=item on_header => $callback->($headers)
280
281When specified, this callback will be called with the header hash as soon
282as headers have been successfully received from the remote server (not on
283locally-generated errors).
284
285It has to return either true (in which case AnyEvent::HTTP will continue),
286or false, in which case AnyEvent::HTTP will cancel the download (and call
287the finish callback with an error code of C<598>).
288
289This callback is useful, among other things, to quickly reject unwanted
290content, which, if it is supposed to be rare, can be faster than first
291doing a C<HEAD> request.
292
293The downside is that cancelling the request makes it impossible to re-use
294the connection. Also, the C<on_header> callback will not receive any
295trailer (headers sent after the response body).
296
297Example: cancel the request unless the content-type is "text/html".
298
299 on_header => sub {
300 $_[0]{"content-type"} =~ /^text\/html\s*(?:;|$)/
301 },
302
303=item on_body => $callback->($partial_body, $headers)
304
305When specified, all body data will be passed to this callback instead of
306to the completion callback. The completion callback will get the empty
307string instead of the body data.
308
309It has to return either true (in which case AnyEvent::HTTP will continue),
310or false, in which case AnyEvent::HTTP will cancel the download (and call
311the completion callback with an error code of C<598>).
312
313The downside to cancelling the request is that it makes it impossible to
314re-use the connection.
315
316This callback is useful when the data is too large to be held in memory
317(so the callback writes it to a file) or when only some information should
318be extracted, or when the body should be processed incrementally.
319
320It is usually preferred over doing your own body handling via
321C<want_body_handle>, but in case of streaming APIs, where HTTP is
322only used to create a connection, C<want_body_handle> is the better
323alternative, as it allows you to install your own event handler, reducing
324resource usage.
325
326=item want_body_handle => $enable
327
328When enabled (default is disabled), the behaviour of AnyEvent::HTTP
329changes considerably: after parsing the headers, and instead of
330downloading the body (if any), the completion callback will be
331called. Instead of the C<$body> argument containing the body data, the
332callback will receive the L<AnyEvent::Handle> object associated with the
333connection. In error cases, C<undef> will be passed. When there is no body
334(e.g. status C<304>), the empty string will be passed.
335
336The handle object might or might not be in TLS mode, might be connected
337to a proxy, be a persistent connection, use chunked transfer encoding
338etc., and configured in unspecified ways. The user is responsible for this
339handle (it will not be used by this module anymore).
340
341This is useful with some push-type services, where, after the initial
342headers, an interactive protocol is used (typical example would be the
343push-style twitter API which starts a JSON/XML stream).
344
345If you think you need this, first have a look at C<on_body>, to see if
346that doesn't solve your problem in a better way.
347
348=item persistent => $boolean
349
350Try to create/reuse a persistent connection. When this flag is set
351(default: true for idempotent requests, false for all others), then
352C<http_request> tries to re-use an existing (previously-created)
353persistent connection to same host (i.e. identical URL scheme, hostname,
354port and session) and, failing that, tries to create a new one.
355
356Requests failing in certain ways will be automatically retried once, which
357is dangerous for non-idempotent requests, which is why it defaults to off
358for them. The reason for this is because the bozos who designed HTTP/1.1
359made it impossible to distinguish between a fatal error and a normal
360connection timeout, so you never know whether there was a problem with
361your request or not.
362
363When reusing an existent connection, many parameters (such as TLS context)
364will be ignored. See the C<session> parameter for a workaround.
365
366=item keepalive => $boolean
367
368Only used when C<persistent> is also true. This parameter decides whether
369C<http_request> tries to handshake a HTTP/1.0-style keep-alive connection
370(as opposed to only a HTTP/1.1 persistent connection).
371
372The default is true, except when using a proxy, in which case it defaults
373to false, as HTTP/1.0 proxies cannot support this in a meaningful way.
374
375=item handle_params => { key => value ... }
376
377The key-value pairs in this hash will be passed to any L<AnyEvent::Handle>
378constructor that is called - not all requests will create a handle, and
379sometimes more than one is created, so this parameter is only good for
380setting hints.
381
382Example: set the maximum read size to 4096, to potentially conserve memory
383at the cost of speed.
384
385 handle_params => {
386 max_read_size => 4096,
387 },
177 388
178=back 389=back
179 390
180Example: make a simple HTTP GET request for http://www.nethype.de/ 391Example: do a simple HTTP GET request for http://www.nethype.de/ and print
392the response body.
181 393
182 http_request GET => "http://www.nethype.de/", sub { 394 http_request GET => "http://www.nethype.de/", sub {
183 my ($body, $hdr) = @_; 395 my ($body, $hdr) = @_;
184 print "$body\n"; 396 print "$body\n";
185 }; 397 };
186 398
187Example: make a HTTP HEAD request on https://www.google.com/, use a 399Example: do a HTTP HEAD request on https://www.google.com/, use a
188timeout of 30 seconds. 400timeout of 30 seconds.
189 401
190 http_request 402 http_request
191 GET => "https://www.google.com", 403 HEAD => "https://www.google.com",
404 headers => { "user-agent" => "MySearchClient 1.0" },
192 timeout => 30, 405 timeout => 30,
193 sub { 406 sub {
194 my ($body, $hdr) = @_; 407 my ($body, $hdr) = @_;
195 use Data::Dumper; 408 use Data::Dumper;
196 print Dumper $hdr; 409 print Dumper $hdr;
197 } 410 }
198 ; 411 ;
199 412
200Example: make another simple HTTP GET request, but immediately try to 413Example: do another simple HTTP GET request, but immediately try to
201cancel it. 414cancel it.
202 415
203 my $request = http_request GET => "http://www.nethype.de/", sub { 416 my $request = http_request GET => "http://www.nethype.de/", sub {
204 my ($body, $hdr) = @_; 417 my ($body, $hdr) = @_;
205 print "$body\n"; 418 print "$body\n";
206 }; 419 };
207 420
208 undef $request; 421 undef $request;
209 422
210=cut 423=cut
424
425#############################################################################
426# wait queue/slots
211 427
212sub _slot_schedule; 428sub _slot_schedule;
213sub _slot_schedule($) { 429sub _slot_schedule($) {
214 my $host = shift; 430 my $host = shift;
215 431
237 push @{ $CO_SLOT{$_[0]}[1] }, $_[1]; 453 push @{ $CO_SLOT{$_[0]}[1] }, $_[1];
238 454
239 _slot_schedule $_[0]; 455 _slot_schedule $_[0];
240} 456}
241 457
458#############################################################################
459# cookie handling
460
461# expire cookies
462sub cookie_jar_expire($;$) {
463 my ($jar, $session_end) = @_;
464
465 %$jar = () if $jar->{version} != 2;
466
467 my $anow = AE::now;
468
469 while (my ($chost, $paths) = each %$jar) {
470 next unless ref $paths;
471
472 while (my ($cpath, $cookies) = each %$paths) {
473 while (my ($cookie, $kv) = each %$cookies) {
474 if (exists $kv->{_expires}) {
475 delete $cookies->{$cookie}
476 if $anow > $kv->{_expires};
477 } elsif ($session_end) {
478 delete $cookies->{$cookie};
479 }
480 }
481
482 delete $paths->{$cpath}
483 unless %$cookies;
484 }
485
486 delete $jar->{$chost}
487 unless %$paths;
488 }
489}
490
491# extract cookies from jar
492sub cookie_jar_extract($$$$) {
493 my ($jar, $scheme, $host, $path) = @_;
494
495 %$jar = () if $jar->{version} != 2;
496
497 $host = AnyEvent::Util::idn_to_ascii $host
498 if $host =~ /[^\x00-\x7f]/;
499
500 my @cookies;
501
502 while (my ($chost, $paths) = each %$jar) {
503 next unless ref $paths;
504
505 # exact match or suffix including . match
506 $chost eq $host or ".$chost" eq substr $host, -1 - length $chost
507 or next;
508
509 while (my ($cpath, $cookies) = each %$paths) {
510 next unless $cpath eq substr $path, 0, length $cpath;
511
512 while (my ($cookie, $kv) = each %$cookies) {
513 next if $scheme ne "https" && exists $kv->{secure};
514
515 if (exists $kv->{_expires} and AE::now > $kv->{_expires}) {
516 delete $cookies->{$cookie};
517 next;
518 }
519
520 my $value = $kv->{value};
521
522 if ($value =~ /[=;,[:space:]]/) {
523 $value =~ s/([\\"])/\\$1/g;
524 $value = "\"$value\"";
525 }
526
527 push @cookies, "$cookie=$value";
528 }
529 }
530 }
531
532 \@cookies
533}
534
535# parse set_cookie header into jar
536sub cookie_jar_set_cookie($$$$) {
537 my ($jar, $set_cookie, $host, $date) = @_;
538
539 %$jar = () if $jar->{version} != 2;
540
541 my $anow = int AE::now;
542 my $snow; # server-now
543
544 for ($set_cookie) {
545 # parse NAME=VALUE
546 my @kv;
547
548 # expires is not http-compliant in the original cookie-spec,
549 # we support the official date format and some extensions
550 while (
551 m{
552 \G\s*
553 (?:
554 expires \s*=\s* ([A-Z][a-z][a-z]+,\ [^,;]+)
555 | ([^=;,[:space:]]+) (?: \s*=\s* (?: "((?:[^\\"]+|\\.)*)" | ([^;,[:space:]]*) ) )?
556 )
557 }gcxsi
558 ) {
559 my $name = $2;
560 my $value = $4;
561
562 if (defined $1) {
563 # expires
564 $name = "expires";
565 $value = $1;
566 } elsif (defined $3) {
567 # quoted
568 $value = $3;
569 $value =~ s/\\(.)/$1/gs;
570 }
571
572 push @kv, @kv ? lc $name : $name, $value;
573
574 last unless /\G\s*;/gc;
575 }
576
577 last unless @kv;
578
579 my $name = shift @kv;
580 my %kv = (value => shift @kv, @kv);
581
582 if (exists $kv{"max-age"}) {
583 $kv{_expires} = $anow + delete $kv{"max-age"};
584 } elsif (exists $kv{expires}) {
585 $snow ||= parse_date ($date) || $anow;
586 $kv{_expires} = $anow + (parse_date (delete $kv{expires}) - $snow);
587 } else {
588 delete $kv{_expires};
589 }
590
591 my $cdom;
592 my $cpath = (delete $kv{path}) || "/";
593
594 if (exists $kv{domain}) {
595 $cdom = $kv{domain};
596
597 $cdom =~ s/^\.?/./; # make sure it starts with a "."
598
599 next if $cdom =~ /\.$/;
600
601 # this is not rfc-like and not netscape-like. go figure.
602 my $ndots = $cdom =~ y/.//;
603 next if $ndots < ($cdom =~ /\.[^.][^.]\.[^.][^.]$/ ? 3 : 2);
604
605 $cdom = substr $cdom, 1; # remove initial .
606 } else {
607 $cdom = $host;
608 }
609
610 # store it
611 $jar->{version} = 2;
612 $jar->{lc $cdom}{$cpath}{$name} = \%kv;
613
614 redo if /\G\s*,/gc;
615 }
616}
617
618#############################################################################
619# keepalive/persistent connection cache
620
621# fetch a connection from the keepalive cache
622sub ka_fetch($) {
623 my $ka_key = shift;
624
625 my $hdl = pop @{ $KA_CACHE{$ka_key} }; # currently we reuse the MOST RECENTLY USED connection
626 delete $KA_CACHE{$ka_key}
627 unless @{ $KA_CACHE{$ka_key} };
628
629 $hdl
630}
631
632sub ka_store($$) {
633 my ($ka_key, $hdl) = @_;
634
635 my $kaa = $KA_CACHE{$ka_key} ||= [];
636
637 my $destroy = sub {
638 my @ka = grep $_ != $hdl, @{ $KA_CACHE{$ka_key} };
639
640 $hdl->destroy;
641
642 @ka
643 ? $KA_CACHE{$ka_key} = \@ka
644 : delete $KA_CACHE{$ka_key};
645 };
646
647 # on error etc., destroy
648 $hdl->on_error ($destroy);
649 $hdl->on_eof ($destroy);
650 $hdl->on_read ($destroy);
651 $hdl->timeout ($PERSISTENT_TIMEOUT);
652
653 push @$kaa, $hdl;
654 shift @$kaa while @$kaa > $MAX_PER_HOST;
655}
656
657#############################################################################
658# utilities
659
660# continue to parse $_ for headers and place them into the arg
661sub _parse_hdr() {
662 my %hdr;
663
664 # things seen, not parsed:
665 # p3pP="NON CUR OTPi OUR NOR UNI"
666
667 $hdr{lc $1} .= ",$2"
668 while /\G
669 ([^:\000-\037]*):
670 [\011\040]*
671 ((?: [^\012]+ | \012[\011\040] )*)
672 \012
673 /gxc;
674
675 /\G$/
676 or return;
677
678 # remove the "," prefix we added to all headers above
679 substr $_, 0, 1, ""
680 for values %hdr;
681
682 \%hdr
683}
684
685#############################################################################
686# http_get
687
688our $qr_nlnl = qr{(?<![^\012])\015?\012};
689
690our $TLS_CTX_LOW = { cache => 1, sslv2 => 1 };
691our $TLS_CTX_HIGH = { cache => 1, verify => 1, verify_peername => "https" };
692
693# maybe it should just become a normal object :/
694
695sub _destroy_state(\%) {
696 my ($state) = @_;
697
698 $state->{handle}->destroy if $state->{handle};
699 %$state = ();
700}
701
702sub _error(\%$$) {
703 my ($state, $cb, $hdr) = @_;
704
705 &_destroy_state ($state);
706
707 $cb->(undef, $hdr);
708 ()
709}
710
711our %IDEMPOTENT = (
712 DELETE => 1,
713 GET => 1,
714 HEAD => 1,
715 OPTIONS => 1,
716 PUT => 1,
717 TRACE => 1,
718
719 ACL => 1,
720 "BASELINE-CONTROL" => 1,
721 BIND => 1,
722 CHECKIN => 1,
723 CHECKOUT => 1,
724 COPY => 1,
725 LABEL => 1,
726 LINK => 1,
727 MERGE => 1,
728 MKACTIVITY => 1,
729 MKCALENDAR => 1,
730 MKCOL => 1,
731 MKREDIRECTREF => 1,
732 MKWORKSPACE => 1,
733 MOVE => 1,
734 ORDERPATCH => 1,
735 PROPFIND => 1,
736 PROPPATCH => 1,
737 REBIND => 1,
738 REPORT => 1,
739 SEARCH => 1,
740 UNBIND => 1,
741 UNCHECKOUT => 1,
742 UNLINK => 1,
743 UNLOCK => 1,
744 UPDATE => 1,
745 UPDATEREDIRECTREF => 1,
746 "VERSION-CONTROL" => 1,
747);
748
242sub http_request($$@) { 749sub http_request($$@) {
243 my $cb = pop; 750 my $cb = pop;
244 my ($method, $url, %arg) = @_; 751 my ($method, $url, %arg) = @_;
245 752
246 my %hdr; 753 my %hdr;
754
755 $arg{tls_ctx} = $TLS_CTX_LOW if $arg{tls_ctx} eq "low" || !exists $arg{tls_ctx};
756 $arg{tls_ctx} = $TLS_CTX_HIGH if $arg{tls_ctx} eq "high";
247 757
248 $method = uc $method; 758 $method = uc $method;
249 759
250 if (my $hdr = $arg{headers}) { 760 if (my $hdr = $arg{headers}) {
251 while (my ($k, $v) = each %$hdr) { 761 while (my ($k, $v) = each %$hdr) {
252 $hdr{lc $k} = $v; 762 $hdr{lc $k} = $v;
253 } 763 }
254 } 764 }
255 765
766 # pseudo headers for all subsequent responses
767 my @pseudo = (URL => $url);
768 push @pseudo, Redirect => delete $arg{Redirect} if exists $arg{Redirect};
769
256 my $recurse = exists $arg{recurse} ? delete $arg{recurse} : $MAX_RECURSE; 770 my $recurse = exists $arg{recurse} ? delete $arg{recurse} : $MAX_RECURSE;
257 771
258 return $cb->(undef, { Status => 599, Reason => "recursion limit reached", URL => $url }) 772 return $cb->(undef, { @pseudo, Status => 599, Reason => "Too many redirections" })
259 if $recurse < 0; 773 if $recurse < 0;
260 774
261 my $proxy = $arg{proxy} || $PROXY; 775 my $proxy = exists $arg{proxy} ? $arg{proxy} : $PROXY;
262 my $timeout = $arg{timeout} || $TIMEOUT; 776 my $timeout = $arg{timeout} || $TIMEOUT;
263 777
264 $hdr{"user-agent"} ||= $USERAGENT;
265
266 my ($scheme, $authority, $upath, $query, $fragment) = 778 my ($uscheme, $uauthority, $upath, $query, undef) = # ignore fragment
267 $url =~ m|(?:([^:/?#]+):)?(?://([^/?#]*))?([^?#]*)(?:\?([^#]*))?(?:#(.*))?|; 779 $url =~ m|^([^:]+):(?://([^/?#]*))?([^?#]*)(?:(\?[^#]*))?(?:#(.*))?$|;
268 780
269 $scheme = lc $scheme; 781 $uscheme = lc $uscheme;
270 782
271 my $uport = $scheme eq "http" ? 80 783 my $uport = $uscheme eq "http" ? 80
272 : $scheme eq "https" ? 443 784 : $uscheme eq "https" ? 443
273 : return $cb->(undef, { Status => 599, Reason => "only http and https URL schemes supported", URL => $url }); 785 : return $cb->(undef, { @pseudo, Status => 599, Reason => "Only http and https URL schemes supported" });
274 786
275 $hdr{referer} ||= "$scheme://$authority$upath"; # leave out fragment and query string, just a heuristic
276
277 $authority =~ /^(?: .*\@ )? ([^\@:]+) (?: : (\d+) )?$/x 787 $uauthority =~ /^(?: .*\@ )? ([^\@]+?) (?: : (\d+) )?$/x
278 or return $cb->(undef, { Status => 599, Reason => "unparsable URL", URL => $url }); 788 or return $cb->(undef, { @pseudo, Status => 599, Reason => "Unparsable URL" });
279 789
280 my $uhost = $1; 790 my $uhost = lc $1;
281 $uport = $2 if defined $2; 791 $uport = $2 if defined $2;
282 792
793 $hdr{host} = defined $2 ? "$uhost:$2" : "$uhost"
794 unless exists $hdr{host};
795
283 $uhost =~ s/^\[(.*)\]$/$1/; 796 $uhost =~ s/^\[(.*)\]$/$1/;
284 $upath .= "?$query" if length $query; 797 $upath .= $query if length $query;
285 798
286 $upath =~ s%^/?%/%; 799 $upath =~ s%^/?%/%;
287 800
288 # cookie processing 801 # cookie processing
289 if (my $jar = $arg{cookie_jar}) { 802 if (my $jar = $arg{cookie_jar}) {
290 %$jar = () if $jar->{version} < 1; 803 my $cookies = cookie_jar_extract $jar, $uscheme, $uhost, $upath;
291 804
805 $hdr{cookie} = join "; ", @$cookies
292 my @cookie; 806 if @$cookies;
293 807 }
294 while (my ($chost, $v) = each %$jar) { 808
295 next unless $chost eq substr $uhost, -length $chost; 809 my ($rhost, $rport, $rscheme, $rpath); # request host, port, path
296 next unless $chost =~ /^\./; 810
297 811 if ($proxy) {
298 while (my ($cpath, $v) = each %$v) { 812 ($rpath, $rhost, $rport, $rscheme) = ($url, @$proxy);
299 next unless $cpath eq substr $upath, 0, length $cpath; 813
300 814 $rscheme = "http" unless defined $rscheme;
301 while (my ($k, $v) = each %$v) { 815
302 next if $scheme ne "https" && exists $v->{secure}; 816 # don't support https requests over https-proxy transport,
303 push @cookie, "$k=$v->{value}"; 817 # can't be done with tls as spec'ed, unless you double-encrypt.
818 $rscheme = "http" if $uscheme eq "https" && $rscheme eq "https";
819
820 $rhost = lc $rhost;
821 $rscheme = lc $rscheme;
822 } else {
823 ($rhost, $rport, $rscheme, $rpath) = ($uhost, $uport, $uscheme, $upath);
824 }
825
826 # leave out fragment and query string, just a heuristic
827 $hdr{referer} = "$uscheme://$uauthority$upath" unless exists $hdr{referer};
828 $hdr{"user-agent"} = $USERAGENT unless exists $hdr{"user-agent"};
829
830 $hdr{"content-length"} = length $arg{body}
831 if length $arg{body} || $method ne "GET";
832
833 my $idempotent = $IDEMPOTENT{$method};
834
835 # default value for keepalive is true iff the request is for an idempotent method
836 my $persistent = exists $arg{persistent} ? !!$arg{persistent} : $idempotent;
837 my $keepalive = exists $arg{keepalive} ? !!$arg{keepalive} : !$proxy;
838 my $was_persistent; # true if this is actually a recycled connection
839
840 # the key to use in the keepalive cache
841 my $ka_key = "$uscheme\x00$uhost\x00$uport\x00$arg{sessionid}";
842
843 $hdr{connection} = ($persistent ? $keepalive ? "keep-alive, " : "" : "close, ") . "Te"; #1.1
844 $hdr{te} = "trailers" unless exists $hdr{te}; #1.1
845
846 my %state = (connect_guard => 1);
847
848 my $ae_error = 595; # connecting
849
850 # handle actual, non-tunneled, request
851 my $handle_actual_request = sub {
852 $ae_error = 596; # request phase
853
854 my $hdl = $state{handle};
855
856 $hdl->starttls ("connect") if $uscheme eq "https" && !exists $hdl->{tls};
857
858 # send request
859 $hdl->push_write (
860 "$method $rpath HTTP/1.1\015\012"
861 . (join "", map "\u$_: $hdr{$_}\015\012", grep defined $hdr{$_}, keys %hdr)
862 . "\015\012"
863 . $arg{body}
864 );
865
866 # return if error occurred during push_write()
867 return unless %state;
868
869 # reduce memory usage, save a kitten, also re-use it for the response headers.
870 %hdr = ();
871
872 # status line and headers
873 $state{read_response} = sub {
874 return unless %state;
875
876 for ("$_[1]") {
877 y/\015//d; # weed out any \015, as they show up in the weirdest of places.
878
879 /^HTTP\/0*([0-9\.]+) \s+ ([0-9]{3}) (?: \s+ ([^\012]*) )? \012/gxci
880 or return _error %state, $cb, { @pseudo, Status => 599, Reason => "Invalid server response" };
881
882 # 100 Continue handling
883 # should not happen as we don't send expect: 100-continue,
884 # but we handle it just in case.
885 # since we send the request body regardless, if we get an error
886 # we are out of-sync, which we currently do NOT handle correctly.
887 return $state{handle}->push_read (line => $qr_nlnl, $state{read_response})
888 if $2 eq 100;
889
890 push @pseudo,
891 HTTPVersion => $1,
892 Status => $2,
893 Reason => $3,
894 ;
895
896 my $hdr = _parse_hdr
897 or return _error %state, $cb, { @pseudo, Status => 599, Reason => "Garbled response headers" };
898
899 %hdr = (%$hdr, @pseudo);
900 }
901
902 # redirect handling
903 # relative uri handling forced by microsoft and other shitheads.
904 # we give our best and fall back to URI if available.
905 if (exists $hdr{location}) {
906 my $loc = $hdr{location};
907
908 if ($loc =~ m%^//%) { # //
909 $loc = "$uscheme:$loc";
910
911 } elsif ($loc eq "") {
912 $loc = $url;
913
914 } elsif ($loc !~ /^(?: $ | [^:\/?\#]+ : )/x) { # anything "simple"
915 $loc =~ s/^\.\/+//;
916
917 if ($loc !~ m%^[.?#]%) {
918 my $prefix = "$uscheme://$uauthority";
919
920 unless ($loc =~ s/^\///) {
921 $prefix .= $upath;
922 $prefix =~ s/\/[^\/]*$//;
923 }
924
925 $loc = "$prefix/$loc";
926
927 } elsif (eval { require URI }) { # uri
928 $loc = URI->new_abs ($loc, $url)->as_string;
929
930 } else {
931 return _error %state, $cb, { @pseudo, Status => 599, Reason => "Cannot parse Location (URI module missing)" };
932 #$hdr{Status} = 599;
933 #$hdr{Reason} = "Unparsable Redirect (URI module missing)";
934 #$recurse = 0;
935 }
936 }
937
938 $hdr{location} = $loc;
939 }
940
941 my $redirect;
942
943 if ($recurse) {
944 my $status = $hdr{Status};
945
946 # industry standard is to redirect POST as GET for
947 # 301, 302 and 303, in contrast to HTTP/1.0 and 1.1.
948 # also, the UA should ask the user for 301 and 307 and POST,
949 # industry standard seems to be to simply follow.
950 # we go with the industry standard. 308 is defined
951 # by rfc7538
952 if ($status == 301 or $status == 302 or $status == 303) {
953 $redirect = 1;
954 # HTTP/1.1 is unclear on how to mutate the method
955 unless ($method eq "HEAD") {
956 $method = "GET";
957 delete $arg{body};
958 }
959 } elsif ($status == 307 or $status == 308) {
960 $redirect = 1;
304 } 961 }
305 } 962 }
963
964 my $finish = sub { # ($data, $err_status, $err_reason[, $persistent])
965 if ($state{handle}) {
966 # handle keepalive
967 if (
968 $persistent
969 && $_[3]
970 && ($hdr{HTTPVersion} < 1.1
971 ? $hdr{connection} =~ /\bkeep-?alive\b/i
972 : $hdr{connection} !~ /\bclose\b/i)
973 ) {
974 ka_store $ka_key, delete $state{handle};
975 } else {
976 # no keepalive, destroy the handle
977 $state{handle}->destroy;
978 }
979 }
980
981 %state = ();
982
983 if (defined $_[1]) {
984 $hdr{OrigStatus} = $hdr{Status}; $hdr{Status} = $_[1];
985 $hdr{OrigReason} = $hdr{Reason}; $hdr{Reason} = $_[2];
986 }
987
988 # set-cookie processing
989 if ($arg{cookie_jar}) {
990 cookie_jar_set_cookie $arg{cookie_jar}, $hdr{"set-cookie"}, $uhost, $hdr{date};
991 }
992
993 if ($redirect && exists $hdr{location}) {
994 # we ignore any errors, as it is very common to receive
995 # Content-Length != 0 but no actual body
996 # we also access %hdr, as $_[1] might be an erro
997 $state{recurse} =
998 http_request (
999 $method => $hdr{location},
1000 %arg,
1001 recurse => $recurse - 1,
1002 Redirect => [$_[0], \%hdr],
1003 sub {
1004 %state = ();
1005 &$cb
1006 },
1007 );
1008 } else {
1009 $cb->($_[0], \%hdr);
1010 }
1011 };
1012
1013 $ae_error = 597; # body phase
1014
1015 my $chunked = $hdr{"transfer-encoding"} =~ /\bchunked\b/i; # not quite correct...
1016
1017 my $len = $chunked ? undef : $hdr{"content-length"};
1018
1019 # body handling, many different code paths
1020 # - no body expected
1021 # - want_body_handle
1022 # - te chunked
1023 # - 2x length known (with or without on_body)
1024 # - 2x length not known (with or without on_body)
1025 if (!$redirect && $arg{on_header} && !$arg{on_header}(\%hdr)) {
1026 $finish->(undef, 598 => "Request cancelled by on_header");
1027 } elsif (
1028 $hdr{Status} =~ /^(?:1..|204|205|304)$/
1029 or $method eq "HEAD"
1030 or (defined $len && $len == 0) # == 0, not !, because "0 " is true
1031 ) {
1032 # no body
1033 $finish->("", undef, undef, 1);
1034
1035 } elsif (!$redirect && $arg{want_body_handle}) {
1036 $_[0]->on_eof (undef);
1037 $_[0]->on_error (undef);
1038 $_[0]->on_read (undef);
1039
1040 $finish->(delete $state{handle});
1041
1042 } elsif ($chunked) {
1043 my $cl = 0;
1044 my $body = "";
1045 my $on_body = (!$redirect && $arg{on_body}) || sub { $body .= shift; 1 };
1046
1047 $state{read_chunk} = sub {
1048 $_[1] =~ /^([0-9a-fA-F]+)/
1049 or return $finish->(undef, $ae_error => "Garbled chunked transfer encoding");
1050
1051 my $len = hex $1;
1052
1053 if ($len) {
1054 $cl += $len;
1055
1056 $_[0]->push_read (chunk => $len, sub {
1057 $on_body->($_[1], \%hdr)
1058 or return $finish->(undef, 598 => "Request cancelled by on_body");
1059
1060 $_[0]->push_read (line => sub {
1061 length $_[1]
1062 and return $finish->(undef, $ae_error => "Garbled chunked transfer encoding");
1063 $_[0]->push_read (line => $state{read_chunk});
1064 });
1065 });
1066 } else {
1067 $hdr{"content-length"} ||= $cl;
1068
1069 $_[0]->push_read (line => $qr_nlnl, sub {
1070 if (length $_[1]) {
1071 for ("$_[1]") {
1072 y/\015//d; # weed out any \015, as they show up in the weirdest of places.
1073
1074 my $hdr = _parse_hdr
1075 or return $finish->(undef, $ae_error => "Garbled response trailers");
1076
1077 %hdr = (%hdr, %$hdr);
1078 }
1079 }
1080
1081 $finish->($body, undef, undef, 1);
1082 });
1083 }
1084 };
1085
1086 $_[0]->push_read (line => $state{read_chunk});
1087
1088 } elsif (!$redirect && $arg{on_body}) {
1089 if (defined $len) {
1090 $_[0]->on_read (sub {
1091 $len -= length $_[0]{rbuf};
1092
1093 $arg{on_body}(delete $_[0]{rbuf}, \%hdr)
1094 or return $finish->(undef, 598 => "Request cancelled by on_body");
1095
1096 $len > 0
1097 or $finish->("", undef, undef, 1);
1098 });
1099 } else {
1100 $_[0]->on_eof (sub {
1101 $finish->("");
1102 });
1103 $_[0]->on_read (sub {
1104 $arg{on_body}(delete $_[0]{rbuf}, \%hdr)
1105 or $finish->(undef, 598 => "Request cancelled by on_body");
1106 });
1107 }
1108 } else {
1109 $_[0]->on_eof (undef);
1110
1111 if (defined $len) {
1112 $_[0]->on_read (sub {
1113 $finish->((substr delete $_[0]{rbuf}, 0, $len, ""), undef, undef, 1)
1114 if $len <= length $_[0]{rbuf};
1115 });
1116 } else {
1117 $_[0]->on_error (sub {
1118 ($! == Errno::EPIPE || !$!)
1119 ? $finish->(delete $_[0]{rbuf})
1120 : $finish->(undef, $ae_error => $_[2]);
1121 });
1122 $_[0]->on_read (sub { });
1123 }
1124 }
1125 };
1126
1127 # if keepalive is enabled, then the server closing the connection
1128 # before a response can happen legally - we retry on idempotent methods.
1129 if ($was_persistent && $idempotent) {
1130 my $old_eof = $hdl->{on_eof};
1131 $hdl->{on_eof} = sub {
1132 _destroy_state %state;
1133
1134 %state = ();
1135 $state{recurse} =
1136 http_request (
1137 $method => $url,
1138 %arg,
1139 recurse => $recurse - 1,
1140 persistent => 0,
1141 sub {
1142 %state = ();
1143 &$cb
1144 }
1145 );
1146 };
1147 $hdl->on_read (sub {
1148 return unless %state;
1149
1150 # as soon as we receive something, a connection close
1151 # once more becomes a hard error
1152 $hdl->{on_eof} = $old_eof;
1153 $hdl->push_read (line => $qr_nlnl, $state{read_response});
1154 });
1155 } else {
1156 $hdl->push_read (line => $qr_nlnl, $state{read_response});
306 } 1157 }
307
308 $hdr{cookie} = join "; ", @cookie
309 if @cookie;
310 } 1158 };
311 1159
312 my ($rhost, $rport, $rpath); # request host, port, path 1160 my $prepare_handle = sub {
1161 my ($hdl) = $state{handle};
313 1162
314 if ($proxy) { 1163 $hdl->on_error (sub {
315 ($rhost, $rport, $scheme) = @$proxy; 1164 _error %state, $cb, { @pseudo, Status => $ae_error, Reason => $_[2] };
316 $rpath = $url; 1165 });
317 } else { 1166 $hdl->on_eof (sub {
318 ($rhost, $rport, $rpath) = ($uhost, $uport, $upath); 1167 _error %state, $cb, { @pseudo, Status => $ae_error, Reason => "Unexpected end-of-file" };
319 $hdr{host} = $uhost; 1168 });
1169 $hdl->timeout_reset;
1170 $hdl->timeout ($timeout);
320 } 1171 };
321 1172
322 $hdr{"content-length"} = length $arg{body}; 1173 # connected to proxy (or origin server)
1174 my $connect_cb = sub {
1175 my $fh = shift
1176 or return _error %state, $cb, { @pseudo, Status => $ae_error, Reason => "$!" };
323 1177
324 my %state = (connect_guard => 1); 1178 return unless delete $state{connect_guard};
1179
1180 # get handle
1181 $state{handle} = new AnyEvent::Handle
1182 %{ $arg{handle_params} },
1183 fh => $fh,
1184 peername => $uhost,
1185 tls_ctx => $arg{tls_ctx},
1186 ;
1187
1188 $prepare_handle->();
1189
1190 #$state{handle}->starttls ("connect") if $rscheme eq "https";
1191
1192 # now handle proxy-CONNECT method
1193 if ($proxy && $uscheme eq "https") {
1194 # oh dear, we have to wrap it into a connect request
1195
1196 my $auth = exists $hdr{"proxy-authorization"}
1197 ? "proxy-authorization: " . (delete $hdr{"proxy-authorization"}) . "\015\012"
1198 : "";
1199
1200 # maybe re-use $uauthority with patched port?
1201 $state{handle}->push_write ("CONNECT $uhost:$uport HTTP/1.0\015\012$auth\015\012");
1202 $state{handle}->push_read (line => $qr_nlnl, sub {
1203 $_[1] =~ /^HTTP\/([0-9\.]+) \s+ ([0-9]{3}) (?: \s+ ([^\015\012]*) )?/ix
1204 or return _error %state, $cb, { @pseudo, Status => 599, Reason => "Invalid proxy connect response ($_[1])" };
1205
1206 if ($2 == 200) {
1207 $rpath = $upath;
1208 $handle_actual_request->();
1209 } else {
1210 _error %state, $cb, { @pseudo, Status => $2, Reason => $3 };
1211 }
1212 });
1213 } else {
1214 delete $hdr{"proxy-authorization"} unless $proxy;
1215
1216 $handle_actual_request->();
1217 }
1218 };
325 1219
326 _get_slot $uhost, sub { 1220 _get_slot $uhost, sub {
327 $state{slot_guard} = shift; 1221 $state{slot_guard} = shift;
328 1222
329 return unless $state{connect_guard}; 1223 return unless $state{connect_guard};
330 1224
331 $state{connect_guard} = AnyEvent::Socket::tcp_connect $rhost, $rport, sub { 1225 # try to use an existing keepalive connection, but only if we, ourselves, plan
332 $state{fh} = shift 1226 # on a keepalive request (in theory, this should be a separate config option).
333 or return $cb->(undef, { Status => 599, Reason => "$!", URL => $url }); 1227 if ($persistent && $KA_CACHE{$ka_key}) {
1228 $was_persistent = 1;
334 1229
335 delete $state{connect_guard}; # reduce memory usage, save a tree 1230 $state{handle} = ka_fetch $ka_key;
1231# $state{handle}->destroyed
1232# and die "AnyEvent::HTTP: unexpectedly got a destructed handle (1), please report.";#d#
1233 $prepare_handle->();
1234# $state{handle}->destroyed
1235# and die "AnyEvent::HTTP: unexpectedly got a destructed handle (2), please report.";#d#
1236 $handle_actual_request->();
336 1237
337 # get handle
338 $state{handle} = new AnyEvent::Handle
339 fh => $state{fh},
340 ($scheme eq "https" ? (tls => "connect") : ());
341
342 # limit the number of persistent connections
343 if ($KA_COUNT{$_[1]} < $MAX_PERSISTENT_PER_HOST) {
344 ++$KA_COUNT{$_[1]};
345 $state{handle}{ka_count_guard} = AnyEvent::Util::guard { --$KA_COUNT{$_[1]} };
346 $hdr{connection} = "keep-alive";
347 delete $hdr{connection}; # keep-alive not yet supported
348 } else { 1238 } else {
349 delete $hdr{connection}; 1239 my $tcp_connect = $arg{tcp_connect}
350 } 1240 || do { require AnyEvent::Socket; \&AnyEvent::Socket::tcp_connect };
351 1241
352 # (re-)configure handle 1242 $state{connect_guard} = $tcp_connect->($rhost, $rport, $connect_cb, $arg{on_prepare} || sub { $timeout });
353 $state{handle}->timeout ($timeout);
354 $state{handle}->on_error (sub {
355 my $errno = "$!";
356 %state = ();
357 $cb->(undef, { Status => 599, Reason => $errno, URL => $url });
358 });
359 $state{handle}->on_eof (sub {
360 %state = ();
361 $cb->(undef, { Status => 599, Reason => "unexpected end-of-file", URL => $url });
362 });
363
364 # send request
365 $state{handle}->push_write (
366 "$method $rpath HTTP/1.0\015\012"
367 . (join "", map "$_: $hdr{$_}\015\012", keys %hdr)
368 . "\015\012"
369 . (delete $arg{body})
370 );
371
372 %hdr = (); # reduce memory usage, save a kitten
373
374 # status line
375 $state{handle}->push_read (line => qr/\015?\012/, sub {
376 $_[1] =~ /^HTTP\/([0-9\.]+) \s+ ([0-9]{3}) (?: \s+ ([^\015\012]*) )?/ix
377 or return (%state = (), $cb->(undef, { Status => 599, Reason => "invalid server response ($_[1])", URL => $url }));
378
379 my %hdr = ( # response headers
380 HTTPVersion => "\x00$1",
381 Status => "\x00$2",
382 Reason => "\x00$3",
383 URL => "\x00$url"
384 );
385
386 # headers, could be optimized a bit
387 $state{handle}->unshift_read (line => qr/\015?\012\015?\012/, sub {
388 for ("$_[1]\012") {
389 # we support spaces in field names, as lotus domino
390 # creates them.
391 $hdr{lc $1} .= "\x00$2"
392 while /\G
393 ([^:\000-\037]+):
394 [\011\040]*
395 ((?: [^\015\012]+ | \015?\012[\011\040] )*)
396 \015?\012
397 /gxc;
398
399 /\G$/
400 or return (%state = (), $cb->(undef, { Status => 599, Reason => "garbled response headers", URL => $url }));
401 }
402
403 substr $_, 0, 1, ""
404 for values %hdr;
405
406 my $finish = sub {
407 %state = ();
408
409 # set-cookie processing
410 if ($arg{cookie_jar} && exists $hdr{"set-cookie"}) {
411 for (split /\x00/, $hdr{"set-cookie"}) {
412 my ($cookie, @arg) = split /;\s*/;
413 my ($name, $value) = split /=/, $cookie, 2;
414 my %kv = (value => $value, map { split /=/, $_, 2 } @arg);
415
416 my $cdom = (delete $kv{domain}) || $uhost;
417 my $cpath = (delete $kv{path}) || "/";
418
419 $cdom =~ s/^\.?/./; # make sure it starts with a "."
420
421 next if $cdom =~ /\.$/;
422
423 # this is not rfc-like and not netscape-like. go figure.
424 my $ndots = $cdom =~ y/.//;
425 next if $ndots < ($cdom =~ /\.[^.][^.]\.[^.][^.]$/ ? 3 : 2);
426
427 # store it
428 $arg{cookie_jar}{version} = 1;
429 $arg{cookie_jar}{$cdom}{$cpath}{$name} = \%kv;
430 }
431 }
432
433 # microsoft and other shitheads don't give a shit for following standards,
434 # try to support some common forms of broken Location headers.
435 if ($_[1]{location} !~ /^(?: $ | [^:\/?\#]+ : )/x) {
436 $_[1]{location} =~ s/^\.\/+//;
437
438 my $url = "$scheme://$uhost:$uport";
439
440 unless ($_[1]{location} =~ s/^\///) {
441 $url .= $upath;
442 $url =~ s/\/[^\/]*$//;
443 }
444
445 $_[1]{location} = "$url/$_[1]{location}";
446 }
447
448 if ($_[1]{Status} =~ /^30[12]$/ && $recurse && $method ne "POST") {
449 # apparently, mozilla et al. just change POST to GET here
450 # more research is needed before we do the same
451 http_request ($method, $_[1]{location}, %arg, recurse => $recurse - 1, $cb);
452 } elsif ($_[1]{Status} == 303 && $recurse) {
453 # even http/1.1 is unlear on how to mutate the method
454 $method = "GET" unless $method eq "HEAD";
455 http_request ($method => $_[1]{location}, %arg, recurse => $recurse - 1, $cb);
456 } elsif ($_[1]{Status} == 307 && $recurse && $method =~ /^(?:GET|HEAD)$/) {
457 http_request ($method => $_[1]{location}, %arg, recurse => $recurse - 1, $cb);
458 } else {
459 $cb->($_[0], $_[1]);
460 }
461 };
462
463 if ($hdr{Status} =~ /^(?:1..|204|304)$/ or $method eq "HEAD") {
464 $finish->(undef, \%hdr);
465 } else {
466 if (exists $hdr{"content-length"}) {
467 $_[0]->unshift_read (chunk => $hdr{"content-length"}, sub {
468 # could cache persistent connection now
469 if ($hdr{connection} =~ /\bkeep-alive\b/i) {
470 # but we don't, due to misdesigns, this is annoyingly complex
471 };
472
473 $finish->($_[1], \%hdr);
474 });
475 } else {
476 # too bad, need to read until we get an error or EOF,
477 # no way to detect winged data.
478 $_[0]->on_error (sub {
479 $finish->($_[0]{rbuf}, \%hdr);
480 });
481 $_[0]->on_eof (undef);
482 $_[0]->on_read (sub { });
483 }
484 }
485 });
486 });
487 }, sub {
488 $timeout
489 }; 1243 }
490 }; 1244 };
491 1245
492 defined wantarray && AnyEvent::Util::guard { %state = () } 1246 defined wantarray && AnyEvent::Util::guard { _destroy_state %state }
493} 1247}
494 1248
495sub http_get($@) { 1249sub http_get($@) {
496 unshift @_, "GET"; 1250 unshift @_, "GET";
497 &http_request 1251 &http_request
508 &http_request 1262 &http_request
509} 1263}
510 1264
511=back 1265=back
512 1266
1267=head2 DNS CACHING
1268
1269AnyEvent::HTTP uses the AnyEvent::Socket::tcp_connect function for
1270the actual connection, which in turn uses AnyEvent::DNS to resolve
1271hostnames. The latter is a simple stub resolver and does no caching
1272on its own. If you want DNS caching, you currently have to provide
1273your own default resolver (by storing a suitable resolver object in
1274C<$AnyEvent::DNS::RESOLVER>) or your own C<tcp_connect> callback.
1275
513=head2 GLOBAL FUNCTIONS AND VARIABLES 1276=head2 GLOBAL FUNCTIONS AND VARIABLES
514 1277
515=over 4 1278=over 4
516 1279
517=item AnyEvent::HTTP::set_proxy "proxy-url" 1280=item AnyEvent::HTTP::set_proxy "proxy-url"
518 1281
519Sets the default proxy server to use. The proxy-url must begin with a 1282Sets the default proxy server to use. The proxy-url must begin with a
520string of the form C<http://host:port> (optionally C<https:...>). 1283string of the form C<http://host:port>, croaks otherwise.
1284
1285To clear an already-set proxy, use C<undef>.
1286
1287When AnyEvent::HTTP is loaded for the first time it will query the
1288default proxy from the operating system, currently by looking at
1289C<$ENV{http_proxy>}.
1290
1291=item AnyEvent::HTTP::cookie_jar_expire $jar[, $session_end]
1292
1293Remove all cookies from the cookie jar that have been expired. If
1294C<$session_end> is given and true, then additionally remove all session
1295cookies.
1296
1297You should call this function (with a true C<$session_end>) before you
1298save cookies to disk, and you should call this function after loading them
1299again. If you have a long-running program you can additionally call this
1300function from time to time.
1301
1302A cookie jar is initially an empty hash-reference that is managed by this
1303module. Its format is subject to change, but currently it is as follows:
1304
1305The key C<version> has to contain C<2>, otherwise the hash gets
1306cleared. All other keys are hostnames or IP addresses pointing to
1307hash-references. The key for these inner hash references is the
1308server path for which this cookie is meant, and the values are again
1309hash-references. Each key of those hash-references is a cookie name, and
1310the value, you guessed it, is another hash-reference, this time with the
1311key-value pairs from the cookie, except for C<expires> and C<max-age>,
1312which have been replaced by a C<_expires> key that contains the cookie
1313expiry timestamp. Session cookies are indicated by not having an
1314C<_expires> key.
1315
1316Here is an example of a cookie jar with a single cookie, so you have a
1317chance of understanding the above paragraph:
1318
1319 {
1320 version => 2,
1321 "10.0.0.1" => {
1322 "/" => {
1323 "mythweb_id" => {
1324 _expires => 1293917923,
1325 value => "ooRung9dThee3ooyXooM1Ohm",
1326 },
1327 },
1328 },
1329 }
1330
1331=item $date = AnyEvent::HTTP::format_date $timestamp
1332
1333Takes a POSIX timestamp (seconds since the epoch) and formats it as a HTTP
1334Date (RFC 2616).
1335
1336=item $timestamp = AnyEvent::HTTP::parse_date $date
1337
1338Takes a HTTP Date (RFC 2616) or a Cookie date (netscape cookie spec) or a
1339bunch of minor variations of those, and returns the corresponding POSIX
1340timestamp, or C<undef> if the date cannot be parsed.
521 1341
522=item $AnyEvent::HTTP::MAX_RECURSE 1342=item $AnyEvent::HTTP::MAX_RECURSE
523 1343
524The default value for the C<recurse> request parameter (default: C<10>). 1344The default value for the C<recurse> request parameter (default: C<10>).
525 1345
1346=item $AnyEvent::HTTP::TIMEOUT
1347
1348The default timeout for connection operations (default: C<300>).
1349
526=item $AnyEvent::HTTP::USERAGENT 1350=item $AnyEvent::HTTP::USERAGENT
527 1351
528The default value for the C<User-Agent> header (the default is 1352The default value for the C<User-Agent> header (the default is
529C<Mozilla/5.0 (compatible; AnyEvent::HTTP/$VERSION; +http://software.schmorp.de/pkg/AnyEvent)>). 1353C<Mozilla/5.0 (compatible; U; AnyEvent-HTTP/$VERSION; +http://software.schmorp.de/pkg/AnyEvent)>).
530 1354
531=item $AnyEvent::HTTP::MAX_PERSISTENT 1355=item $AnyEvent::HTTP::MAX_PER_HOST
532 1356
533The maximum number of persistent connections to keep open (default: 8). 1357The maximum number of concurrent connections to the same host (identified
1358by the hostname). If the limit is exceeded, then additional requests
1359are queued until previous connections are closed. Both persistent and
1360non-persistent connections are counted in this limit.
534 1361
535Not implemented currently. 1362The default value for this is C<4>, and it is highly advisable to not
1363increase it much.
1364
1365For comparison: the RFC's recommend 4 non-persistent or 2 persistent
1366connections, older browsers used 2, newer ones (such as firefox 3)
1367typically use 6, and Opera uses 8 because like, they have the fastest
1368browser and give a shit for everybody else on the planet.
536 1369
537=item $AnyEvent::HTTP::PERSISTENT_TIMEOUT 1370=item $AnyEvent::HTTP::PERSISTENT_TIMEOUT
538 1371
539The maximum time to cache a persistent connection, in seconds (default: 2). 1372The time after which idle persistent connections get closed by
540 1373AnyEvent::HTTP (default: C<3>).
541Not implemented currently.
542 1374
543=item $AnyEvent::HTTP::ACTIVE 1375=item $AnyEvent::HTTP::ACTIVE
544 1376
545The number of active connections. This is not the number of currently 1377The number of active connections. This is not the number of currently
546running requests, but the number of currently open and non-idle TCP 1378running requests, but the number of currently open and non-idle TCP
547connections. This number of can be useful for load-leveling. 1379connections. This number can be useful for load-leveling.
548 1380
549=back 1381=back
550 1382
551=cut 1383=cut
552 1384
1385our @month = qw(Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec);
1386our @weekday = qw(Sun Mon Tue Wed Thu Fri Sat);
1387
1388sub format_date($) {
1389 my ($time) = @_;
1390
1391 # RFC 822/1123 format
1392 my ($S, $M, $H, $mday, $mon, $year, $wday, $yday, undef) = gmtime $time;
1393
1394 sprintf "%s, %02d %s %04d %02d:%02d:%02d GMT",
1395 $weekday[$wday], $mday, $month[$mon], $year + 1900,
1396 $H, $M, $S;
1397}
1398
1399sub parse_date($) {
1400 my ($date) = @_;
1401
1402 my ($d, $m, $y, $H, $M, $S);
1403
1404 if ($date =~ /^[A-Z][a-z][a-z]+, ([0-9][0-9]?)[\- ]([A-Z][a-z][a-z])[\- ]([0-9][0-9][0-9][0-9]) ([0-9][0-9]?):([0-9][0-9]?):([0-9][0-9]?) GMT$/) {
1405 # RFC 822/1123, required by RFC 2616 (with " ")
1406 # cookie dates (with "-")
1407
1408 ($d, $m, $y, $H, $M, $S) = ($1, $2, $3, $4, $5, $6);
1409
1410 } elsif ($date =~ /^[A-Z][a-z][a-z]+, ([0-9][0-9]?)-([A-Z][a-z][a-z])-([0-9][0-9]) ([0-9][0-9]?):([0-9][0-9]?):([0-9][0-9]?) GMT$/) {
1411 # RFC 850
1412 ($d, $m, $y, $H, $M, $S) = ($1, $2, $3 < 69 ? $3 + 2000 : $3 + 1900, $4, $5, $6);
1413
1414 } elsif ($date =~ /^[A-Z][a-z][a-z]+ ([A-Z][a-z][a-z]) ([0-9 ]?[0-9]) ([0-9][0-9]?):([0-9][0-9]?):([0-9][0-9]?) ([0-9][0-9][0-9][0-9])$/) {
1415 # ISO C's asctime
1416 ($d, $m, $y, $H, $M, $S) = ($2, $1, $6, $3, $4, $5);
1417 }
1418 # other formats fail in the loop below
1419
1420 for (0..11) {
1421 if ($m eq $month[$_]) {
1422 require Time::Local;
1423 return eval { Time::Local::timegm ($S, $M, $H, $d, $_, $y) };
1424 }
1425 }
1426
1427 undef
1428}
1429
553sub set_proxy($) { 1430sub set_proxy($) {
554 $PROXY = [$2, $3 || 3128, $1] if $_[0] =~ m%^(https?):// ([^:/]+) (?: : (\d*) )?%ix; 1431 if (length $_[0]) {
1432 $_[0] =~ m%^(http):// ([^:/]+) (?: : (\d*) )?%ix
1433 or Carp::croak "$_[0]: invalid proxy URL";
1434 $PROXY = [$2, $3 || 3128, $1]
1435 } else {
1436 undef $PROXY;
1437 }
555} 1438}
556 1439
557# initialise proxy from environment 1440# initialise proxy from environment
1441eval {
558set_proxy $ENV{http_proxy}; 1442 set_proxy $ENV{http_proxy};
1443};
1444
1445=head2 SHOWCASE
1446
1447This section contains some more elaborate "real-world" examples or code
1448snippets.
1449
1450=head2 HTTP/1.1 FILE DOWNLOAD
1451
1452Downloading files with HTTP can be quite tricky, especially when something
1453goes wrong and you want to resume.
1454
1455Here is a function that initiates and resumes a download. It uses the
1456last modified time to check for file content changes, and works with many
1457HTTP/1.0 servers as well, and usually falls back to a complete re-download
1458on older servers.
1459
1460It calls the completion callback with either C<undef>, which means a
1461nonretryable error occurred, C<0> when the download was partial and should
1462be retried, and C<1> if it was successful.
1463
1464 use AnyEvent::HTTP;
1465
1466 sub download($$$) {
1467 my ($url, $file, $cb) = @_;
1468
1469 open my $fh, "+<", $file
1470 or die "$file: $!";
1471
1472 my %hdr;
1473 my $ofs = 0;
1474
1475 if (stat $fh and -s _) {
1476 $ofs = -s _;
1477 warn "-s is ", $ofs;
1478 $hdr{"if-unmodified-since"} = AnyEvent::HTTP::format_date +(stat _)[9];
1479 $hdr{"range"} = "bytes=$ofs-";
1480 }
1481
1482 http_get $url,
1483 headers => \%hdr,
1484 on_header => sub {
1485 my ($hdr) = @_;
1486
1487 if ($hdr->{Status} == 200 && $ofs) {
1488 # resume failed
1489 truncate $fh, $ofs = 0;
1490 }
1491
1492 sysseek $fh, $ofs, 0;
1493
1494 1
1495 },
1496 on_body => sub {
1497 my ($data, $hdr) = @_;
1498
1499 if ($hdr->{Status} =~ /^2/) {
1500 length $data == syswrite $fh, $data
1501 or return; # abort on write errors
1502 }
1503
1504 1
1505 },
1506 sub {
1507 my (undef, $hdr) = @_;
1508
1509 my $status = $hdr->{Status};
1510
1511 if (my $time = AnyEvent::HTTP::parse_date $hdr->{"last-modified"}) {
1512 utime $time, $time, $fh;
1513 }
1514
1515 if ($status == 200 || $status == 206 || $status == 416) {
1516 # download ok || resume ok || file already fully downloaded
1517 $cb->(1, $hdr);
1518
1519 } elsif ($status == 412) {
1520 # file has changed while resuming, delete and retry
1521 unlink $file;
1522 $cb->(0, $hdr);
1523
1524 } elsif ($status == 500 or $status == 503 or $status =~ /^59/) {
1525 # retry later
1526 $cb->(0, $hdr);
1527
1528 } else {
1529 $cb->(undef, $hdr);
1530 }
1531 }
1532 ;
1533 }
1534
1535 download "http://server/somelargefile", "/tmp/somelargefile", sub {
1536 if ($_[0]) {
1537 print "OK!\n";
1538 } elsif (defined $_[0]) {
1539 print "please retry later\n";
1540 } else {
1541 print "ERROR\n";
1542 }
1543 };
1544
1545=head3 SOCKS PROXIES
1546
1547Socks proxies are not directly supported by AnyEvent::HTTP. You can
1548compile your perl to support socks, or use an external program such as
1549F<socksify> (dante) or F<tsocks> to make your program use a socks proxy
1550transparently.
1551
1552Alternatively, for AnyEvent::HTTP only, you can use your own
1553C<tcp_connect> function that does the proxy handshake - here is an example
1554that works with socks4a proxies:
1555
1556 use Errno;
1557 use AnyEvent::Util;
1558 use AnyEvent::Socket;
1559 use AnyEvent::Handle;
1560
1561 # host, port and username of/for your socks4a proxy
1562 my $socks_host = "10.0.0.23";
1563 my $socks_port = 9050;
1564 my $socks_user = "";
1565
1566 sub socks4a_connect {
1567 my ($host, $port, $connect_cb, $prepare_cb) = @_;
1568
1569 my $hdl = new AnyEvent::Handle
1570 connect => [$socks_host, $socks_port],
1571 on_prepare => sub { $prepare_cb->($_[0]{fh}) },
1572 on_error => sub { $connect_cb->() },
1573 ;
1574
1575 $hdl->push_write (pack "CCnNZ*Z*", 4, 1, $port, 1, $socks_user, $host);
1576
1577 $hdl->push_read (chunk => 8, sub {
1578 my ($hdl, $chunk) = @_;
1579 my ($status, $port, $ipn) = unpack "xCna4", $chunk;
1580
1581 if ($status == 0x5a) {
1582 $connect_cb->($hdl->{fh}, (format_address $ipn) . ":$port");
1583 } else {
1584 $! = Errno::ENXIO; $connect_cb->();
1585 }
1586 });
1587
1588 $hdl
1589 }
1590
1591Use C<socks4a_connect> instead of C<tcp_connect> when doing C<http_request>s,
1592possibly after switching off other proxy types:
1593
1594 AnyEvent::HTTP::set_proxy undef; # usually you do not want other proxies
1595
1596 http_get 'http://www.google.com', tcp_connect => \&socks4a_connect, sub {
1597 my ($data, $headers) = @_;
1598 ...
1599 };
559 1600
560=head1 SEE ALSO 1601=head1 SEE ALSO
561 1602
562L<AnyEvent>. 1603L<AnyEvent>.
563 1604
564=head1 AUTHOR 1605=head1 AUTHOR
565 1606
566 Marc Lehmann <schmorp@schmorp.de> 1607 Marc Lehmann <schmorp@schmorp.de>
567 http://home.schmorp.de/ 1608 http://home.schmorp.de/
568 1609
1610With many thanks to Дмитрий Шалашов, who provided countless
1611testcases and bugreports.
1612
569=cut 1613=cut
570 1614
5711 16151
572 1616

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines