ViewVC Help

View File | Revision Log | Show Annotations | Download File

/cvs/AnyEvent-HTTP/HTTP.pm

Comparing AnyEvent-HTTP/HTTP.pm (file contents):
Revision 1.65 by root, Fri Dec 31 03:47:32 2010 UTC vs.
Revision 1.100 by root, Fri Jan 28 01:28:03 2011 UTC

…		…
15	This module is an L<AnyEvent> user, you need to make sure that you use and	15	This module is an L<AnyEvent> user, you need to make sure that you use and
16	run a supported event loop.	16	run a supported event loop.
17		17
18	This module implements a simple, stateless and non-blocking HTTP	18	This module implements a simple, stateless and non-blocking HTTP
19	client. It supports GET, POST and other request methods, cookies and more,	19	client. It supports GET, POST and other request methods, cookies and more,
20	all on a very low level. It can follow redirects supports proxies and	20	all on a very low level. It can follow redirects, supports proxies, and
21	automatically limits the number of connections to the values specified in	21	automatically limits the number of connections to the values specified in
22	the RFC.	22	the RFC.
23		23
24	It should generally be a "good client" that is enough for most HTTP	24	It should generally be a "good client" that is enough for most HTTP
25	tasks. Simple tasks should be simple, but complex tasks should still be	25	tasks. Simple tasks should be simple, but complex tasks should still be
…		…
36		36
37	=cut	37	=cut
38		38
39	package AnyEvent::HTTP;	39	package AnyEvent::HTTP;
40		40
41	use strict;	41	use common::sense;
42	no warnings;
43		42
44	use Errno ();	43	use Errno ();
45		44
46	use AnyEvent 5.0 ();	45	use AnyEvent 5.0 ();
47	use AnyEvent::Util ();	46	use AnyEvent::Util ();
48	use AnyEvent::Handle ();	47	use AnyEvent::Handle ();
49		48
50	use base Exporter::;	49	use base Exporter::;
51		50
52	our $VERSION = '1.5';	51	our $VERSION = '2.03';
53		52
54	our @EXPORT = qw(http_get http_post http_head http_request);	53	our @EXPORT = qw(http_get http_post http_head http_request);
55		54
56	our $USERAGENT = "Mozilla/5.0 (compatible; U; AnyEvent-HTTP/$VERSION; +http://software.schmorp.de/pkg/AnyEvent)";	55	our $USERAGENT = "Mozilla/5.0 (compatible; U; AnyEvent-HTTP/$VERSION; +http://software.schmorp.de/pkg/AnyEvent)";
57	our $MAX_RECURSE = 10;	56	our $MAX_RECURSE = 10;
58	our $MAX_PERSISTENT = 8;
59	our $PERSISTENT_TIMEOUT = 2;	57	our $PERSISTENT_TIMEOUT = 3;
60	our $TIMEOUT = 300;	58	our $TIMEOUT = 300;
61		59	our $MAX_PER_HOST = 4; # changing this is evil
62	# changing these is evil
63	our $MAX_PERSISTENT_PER_HOST = 0;
64	our $MAX_PER_HOST = 4;
65		60
66	our $PROXY;	61	our $PROXY;
67	our $ACTIVE = 0;	62	our $ACTIVE = 0;
68		63
69	my %KA_COUNT; # number of open keep-alive connections per host	64	my %KA_CACHE; # indexed by uhost currently, points to [$handle...] array
70	my %CO_SLOT; # number of open connections, and wait queue, per host	65	my %CO_SLOT; # number of open connections, and wait queue, per host
71		66
72	=item http_get $url, key => value..., $cb->($data, $headers)	67	=item http_get $url, key => value..., $cb->($data, $headers)
73		68
74	Executes an HTTP-GET request. See the http_request function for details on	69	Executes an HTTP-GET request. See the http_request function for details on
…		…
94	C<http_request> returns a "cancellation guard" - you have to keep the	89	C<http_request> returns a "cancellation guard" - you have to keep the
95	object at least alive until the callback get called. If the object gets	90	object at least alive until the callback get called. If the object gets
96	destroyed before the callback is called, the request will be cancelled.	91	destroyed before the callback is called, the request will be cancelled.
97		92
98	The callback will be called with the response body data as first argument	93	The callback will be called with the response body data as first argument
99	(or C<undef> if an error occured), and a hash-ref with response headers as	94	(or C<undef> if an error occured), and a hash-ref with response headers
100	second argument.	95	(and trailers) as second argument.
101		96
102	All the headers in that hash are lowercased. In addition to the response	97	All the headers in that hash are lowercased. In addition to the response
103	headers, the "pseudo-headers" (uppercase to avoid clashing with possible	98	headers, the "pseudo-headers" (uppercase to avoid clashing with possible
104	response headers) C<HTTPVersion>, C<Status> and C<Reason> contain the	99	response headers) C<HTTPVersion>, C<Status> and C<Reason> contain the
105	three parts of the HTTP Status-Line of the same name. If an error occurs	100	three parts of the HTTP Status-Line of the same name. If an error occurs
…		…
122		117
123	If the server sends a header multiple times, then their contents will be	118	If the server sends a header multiple times, then their contents will be
124	joined together with a comma (C<,>), as per the HTTP spec.	119	joined together with a comma (C<,>), as per the HTTP spec.
125		120
126	If an internal error occurs, such as not being able to resolve a hostname,	121	If an internal error occurs, such as not being able to resolve a hostname,
127	then C<$data> will be C<undef>, C<< $headers->{Status} >> will be C<59x>	122	then C<$data> will be C<undef>, C<< $headers->{Status} >> will be
128	(usually C<599>) and the C<Reason> pseudo-header will contain an error	123	C<590>-C<599> and the C<Reason> pseudo-header will contain an error
129	message.	124	message. Currently the following status codes are used:
		125
		126	=over 4
		127
		128	=item 595 - errors during connection etsbalishment, proxy handshake.
		129
		130	=item 596 - errors during TLS negotiation, request sending and header processing.
		131
		132	=item 597 - errors during body receiving or processing.
		133
		134	=item 598 - user aborted request via C<on_header> or C<on_body>.
		135
		136	=item 599 - other, usually nonretryable, errors (garbled URL etc.).
		137
		138	=back
130		139
131	A typical callback might look like this:	140	A typical callback might look like this:
132		141
133	sub {	142	sub {
134	my ($body, $hdr) = @_;	143	my ($body, $hdr) = @_;
…		…
150	Whether to recurse requests or not, e.g. on redirects, authentication	159	Whether to recurse requests or not, e.g. on redirects, authentication
151	retries and so on, and how often to do so.	160	retries and so on, and how often to do so.
152		161
153	=item headers => hashref	162	=item headers => hashref
154		163
155	The request headers to use. Currently, C<http_request> may provide its	164	The request headers to use. Currently, C<http_request> may provide its own
156	own C<Host:>, C<Content-Length:>, C<Connection:> and C<Cookie:> headers	165	C<Host:>, C<Content-Length:>, C<Connection:> and C<Cookie:> headers and
157	and will provide defaults for C<User-Agent:> and C<Referer:> (this can be	166	will provide defaults at least for C<TE:>, C<Referer:> and C<User-Agent:>
158	suppressed by using C<undef> for these headers in which case they won't be	167	(this can be suppressed by using C<undef> for these headers in which case
159	sent at all).	168	they won't be sent at all).
		169
		170	You really should provide your own C<User-Agent:> header value that is
		171	appropriate for your program - I wouldn't be surprised if the default
		172	AnyEvent string gets blocked by webservers sooner or later.
160		173
161	=item timeout => $seconds	174	=item timeout => $seconds
162		175
163	The time-out to use for various stages - each connect attempt will reset	176	The time-out to use for various stages - each connect attempt will reset
164	the timeout, as will read or write activity, i.e. this is not an overall	177	the timeout, as will read or write activity, i.e. this is not an overall
…		…
169	=item proxy => [$host, $port[, $scheme]] or undef	182	=item proxy => [$host, $port[, $scheme]] or undef
170		183
171	Use the given http proxy for all requests. If not specified, then the	184	Use the given http proxy for all requests. If not specified, then the
172	default proxy (as specified by C<$ENV{http_proxy}>) is used.	185	default proxy (as specified by C<$ENV{http_proxy}>) is used.
173		186
174	C<$scheme> must be either missing, C<http> for HTTP or C<https> for	187	C<$scheme> must be either missing or must be C<http> for HTTP.
175	HTTPS.
176		188
177	=item body => $string	189	=item body => $string
178		190
179	The request body, usually empty. Will be-sent as-is (future versions of	191	The request body, usually empty. Will be sent as-is (future versions of
180	this module might offer more options).	192	this module might offer more options).
181		193
182	=item cookie_jar => $hash_ref	194	=item cookie_jar => $hash_ref
183		195
184	Passing this parameter enables (simplified) cookie-processing, loosely	196	Passing this parameter enables (simplified) cookie-processing, loosely
185	based on the original netscape specification.	197	based on the original netscape specification.
186		198
187	The C<$hash_ref> must be an (initially empty) hash reference which will	199	The C<$hash_ref> must be an (initially empty) hash reference which
188	get updated automatically. It is possible to save the cookie_jar to	200	will get updated automatically. It is possible to save the cookie jar
189	persistent storage with something like JSON or Storable, but this is not	201	to persistent storage with something like JSON or Storable - see the
190	recommended, as expiry times are currently being ignored.	202	C<AnyEvent::HTTP::cookie_jar_expire> function if you wish to remove
		203	expired or session-only cookies, and also for documentation on the format
		204	of the cookie jar.
191		205
192	Note that this cookie implementation is not of very high quality, nor	206	Note that this cookie implementation is not meant to be complete. If
193	meant to be complete. If you want complete cookie management you have to	207	you want complete cookie management you have to do that on your
194	do that on your own. C<cookie_jar> is meant as a quick fix to get some	208	own. C<cookie_jar> is meant as a quick fix to get most cookie-using sites
195	cookie-using sites working. Cookies are a privacy disaster, do not use	209	working. Cookies are a privacy disaster, do not use them unless required
196	them unless required to.	210	to.
		211
		212	When cookie processing is enabled, the C<Cookie:> and C<Set-Cookie:>
		213	headers will be set and handled by this module, otherwise they will be
		214	left untouched.
197		215
198	=item tls_ctx => $scheme | $tls_ctx	216	=item tls_ctx => $scheme | $tls_ctx
199		217
200	Specifies the AnyEvent::TLS context to be used for https connections. This	218	Specifies the AnyEvent::TLS context to be used for https connections. This
201	parameter follows the same rules as the C<tls_ctx> parameter to	219	parameter follows the same rules as the C<tls_ctx> parameter to
…		…
205	verification) TLS context.	223	verification) TLS context.
206		224
207	The default for this option is C<low>, which could be interpreted as "give	225	The default for this option is C<low>, which could be interpreted as "give
208	me the page, no matter what".	226	me the page, no matter what".
209		227
		228	See also the C<sessionid> parameter.
		229
		230	=item session => $string
		231
		232	The module might reuse connections to the same host internally. Sometimes
		233	(e.g. when using TLS), you do not want to reuse connections from other
		234	sessions. This can be achieved by setting this parameter to some unique
		235	ID (such as the address of an object storing your state data, or the TLS
		236	context) - only connections using the same unique ID will be reused.
		237
210	=item on_prepare => $callback->($fh)	238	=item on_prepare => $callback->($fh)
211		239
212	In rare cases you need to "tune" the socket before it is used to	240	In rare cases you need to "tune" the socket before it is used to
213	connect (for exmaple, to bind it on a given IP address). This parameter	241	connect (for exmaple, to bind it on a given IP address). This parameter
214	overrides the prepare callback passed to C<AnyEvent::Socket::tcp_connect>	242	overrides the prepare callback passed to C<AnyEvent::Socket::tcp_connect>
…		…
240		268
241	This callback is useful, among other things, to quickly reject unwanted	269	This callback is useful, among other things, to quickly reject unwanted
242	content, which, if it is supposed to be rare, can be faster than first	270	content, which, if it is supposed to be rare, can be faster than first
243	doing a C<HEAD> request.	271	doing a C<HEAD> request.
244		272
		273	The downside is that cancelling the request makes it impossible to re-use
		274	the connection. Also, the C<on_header> callback will not receive any
		275	trailer (headers sent after the response body).
		276
245	Example: cancel the request unless the content-type is "text/html".	277	Example: cancel the request unless the content-type is "text/html".
246		278
247	on_header => sub {	279	on_header => sub {
248	$_[0]{"content-type"} =~ /^text\/html\s*(?:;|$)/	280	$_[0]{"content-type"} =~ /^text\/html\s*(?:;|$)/
249	},	281	},
…		…
255	string instead of the body data.	287	string instead of the body data.
256		288
257	It has to return either true (in which case AnyEvent::HTTP will continue),	289	It has to return either true (in which case AnyEvent::HTTP will continue),
258	or false, in which case AnyEvent::HTTP will cancel the download (and call	290	or false, in which case AnyEvent::HTTP will cancel the download (and call
259	the completion callback with an error code of C<598>).	291	the completion callback with an error code of C<598>).
		292
		293	The downside to cancelling the request is that it makes it impossible to
		294	re-use the connection.
260		295
261	This callback is useful when the data is too large to be held in memory	296	This callback is useful when the data is too large to be held in memory
262	(so the callback writes it to a file) or when only some information should	297	(so the callback writes it to a file) or when only some information should
263	be extracted, or when the body should be processed incrementally.	298	be extracted, or when the body should be processed incrementally.
264		299
…		…
276	called. Instead of the C<$body> argument containing the body data, the	311	called. Instead of the C<$body> argument containing the body data, the
277	callback will receive the L<AnyEvent::Handle> object associated with the	312	callback will receive the L<AnyEvent::Handle> object associated with the
278	connection. In error cases, C<undef> will be passed. When there is no body	313	connection. In error cases, C<undef> will be passed. When there is no body
279	(e.g. status C<304>), the empty string will be passed.	314	(e.g. status C<304>), the empty string will be passed.
280		315
281	The handle object might or might not be in TLS mode, might be connected to	316	The handle object might or might not be in TLS mode, might be connected
282	a proxy, be a persistent connection etc., and configured in unspecified	317	to a proxy, be a persistent connection, use chunked transfer encoding
283	ways. The user is responsible for this handle (it will not be used by this	318	etc., and configured in unspecified ways. The user is responsible for this
284	module anymore).	319	handle (it will not be used by this module anymore).
285		320
286	This is useful with some push-type services, where, after the initial	321	This is useful with some push-type services, where, after the initial
287	headers, an interactive protocol is used (typical example would be the	322	headers, an interactive protocol is used (typical example would be the
288	push-style twitter API which starts a JSON/XML stream).	323	push-style twitter API which starts a JSON/XML stream).
289		324
290	If you think you need this, first have a look at C<on_body>, to see if	325	If you think you need this, first have a look at C<on_body>, to see if
291	that doesn't solve your problem in a better way.	326	that doesn't solve your problem in a better way.
292		327
		328	=item persistent => $boolean
		329
		330	Try to create/reuse a persistent connection. When this flag is set
		331	(default: true for idempotent requests, false for all others), then
		332	C<http_request> tries to re-use an existing (previously-created)
		333	persistent connection to the host and, failing that, tries to create a new
		334	one.
		335
		336	Requests failing in certain ways will be automatically retried once, which
		337	is dangerous for non-idempotent requests, which is why it defaults to off
		338	for them. The reason for this is because the bozos who designed HTTP/1.1
		339	made it impossible to distinguish between a fatal error and a normal
		340	connection timeout, so you never know whether there was a problem with
		341	your request or not.
		342
		343	When reusing an existent connection, many parameters (such as TLS context)
		344	will be ignored. See the C<session> parameter for a workaround.
		345
		346	=item keepalive => $boolean
		347
		348	Only used when C<persistent> is also true. This parameter decides whether
		349	C<http_request> tries to handshake a HTTP/1.0-style keep-alive connection
		350	(as opposed to only a HTTP/1.1 persistent connection).
		351
		352	The default is true, except when using a proxy, in which case it defaults
		353	to false, as HTTP/1.0 proxies cannot support this in a meaningful way.
		354
		355	=item handle_params => { key => value ... }
		356
		357	The key-value pairs in this hash will be passed to any L<AnyEvent::Handle>
		358	constructor that is called - not all requests will create a handle, and
		359	sometimes more than one is created, so this parameter is only good for
		360	setting hints.
		361
		362	Example: set the maximum read size to 4096, to potentially conserve memory
		363	at the cost of speed.
		364
		365	handle_params => {
		366	max_read_size => 4096,
		367	},
		368
293	=back	369	=back
294		370
295	Example: make a simple HTTP GET request for http://www.nethype.de/	371	Example: do a simple HTTP GET request for http://www.nethype.de/ and print
		372	the response body.
296		373
297	http_request GET => "http://www.nethype.de/", sub {	374	http_request GET => "http://www.nethype.de/", sub {
298	my ($body, $hdr) = @_;	375	my ($body, $hdr) = @_;
299	print "$body\n";	376	print "$body\n";
300	};	377	};
301		378
302	Example: make a HTTP HEAD request on https://www.google.com/, use a	379	Example: do a HTTP HEAD request on https://www.google.com/, use a
303	timeout of 30 seconds.	380	timeout of 30 seconds.
304		381
305	http_request	382	http_request
306	GET => "https://www.google.com",	383	GET => "https://www.google.com",
		384	headers => { "user-agent" => "MySearchClient 1.0" },
307	timeout => 30,	385	timeout => 30,
308	sub {	386	sub {
309	my ($body, $hdr) = @_;	387	my ($body, $hdr) = @_;
310	use Data::Dumper;	388	use Data::Dumper;
311	print Dumper $hdr;	389	print Dumper $hdr;
312	}	390	}
313	;	391	;
314		392
315	Example: make another simple HTTP GET request, but immediately try to	393	Example: do another simple HTTP GET request, but immediately try to
316	cancel it.	394	cancel it.
317		395
318	my $request = http_request GET => "http://www.nethype.de/", sub {	396	my $request = http_request GET => "http://www.nethype.de/", sub {
319	my ($body, $hdr) = @_;	397	my ($body, $hdr) = @_;
320	print "$body\n";	398	print "$body\n";
321	};	399	};
322		400
323	undef $request;	401	undef $request;
324		402
325	=cut	403	=cut
		404
		405	#############################################################################
		406	# wait queue/slots
326		407
327	sub _slot_schedule;	408	sub _slot_schedule;
328	sub _slot_schedule($) {	409	sub _slot_schedule($) {
329	my $host = shift;	410	my $host = shift;
330		411
…		…
352	push @{ $CO_SLOT{$_[0]}[1] }, $_[1];	433	push @{ $CO_SLOT{$_[0]}[1] }, $_[1];
353		434
354	_slot_schedule $_[0];	435	_slot_schedule $_[0];
355	}	436	}
356		437
		438	#############################################################################
		439	# cookie handling
		440
		441	# expire cookies
		442	sub cookie_jar_expire($;$) {
		443	my ($jar, $session_end) = @_;
		444
		445	%$jar = () if $jar->{version} != 1;
		446
		447	my $anow = AE::now;
		448
		449	while (my ($chost, $paths) = each %$jar) {
		450	next unless ref $paths;
		451
		452	while (my ($cpath, $cookies) = each %$paths) {
		453	while (my ($cookie, $kv) = each %$cookies) {
		454	if (exists $kv->{_expires}) {
		455	delete $cookies->{$cookie}
		456	if $anow > $kv->{_expires};
		457	} elsif ($session_end) {
		458	delete $cookies->{$cookie};
		459	}
		460	}
		461
		462	delete $paths->{$cpath}
		463	unless %$cookies;
		464	}
		465
		466	delete $jar->{$chost}
		467	unless %$paths;
		468	}
		469	}
		470
		471	# extract cookies from jar
		472	sub cookie_jar_extract($$$$) {
		473	my ($jar, $scheme, $host, $path) = @_;
		474
		475	%$jar = () if $jar->{version} != 1;
		476
		477	my @cookies;
		478
		479	while (my ($chost, $paths) = each %$jar) {
		480	next unless ref $paths;
		481
		482	if ($chost =~ /^\./) {
		483	next unless $chost eq substr $host, -length $chost;
		484	} elsif ($chost =~ /\./) {
		485	next unless $chost eq $host;
		486	} else {
		487	next;
		488	}
		489
		490	while (my ($cpath, $cookies) = each %$paths) {
		491	next unless $cpath eq substr $path, 0, length $cpath;
		492
		493	while (my ($cookie, $kv) = each %$cookies) {
		494	next if $scheme ne "https" && exists $kv->{secure};
		495
		496	if (exists $kv->{_expires} and AE::now > $kv->{_expires}) {
		497	delete $cookies->{$cookie};
		498	next;
		499	}
		500
		501	my $value = $kv->{value};
		502
		503	if ($value =~ /[=;,[:space:]]/) {
		504	$value =~ s/([\\"])/\\$1/g;
		505	$value = "\"$value\"";
		506	}
		507
		508	push @cookies, "$cookie=$value";
		509	}
		510	}
		511	}
		512
		513	\@cookies
		514	}
		515
		516	# parse set_cookie header into jar
		517	sub cookie_jar_set_cookie($$$$) {
		518	my ($jar, $set_cookie, $host, $date) = @_;
		519
		520	my $anow = int AE::now;
		521	my $snow; # server-now
		522
		523	for ($set_cookie) {
		524	# parse NAME=VALUE
		525	my @kv;
		526
		527	# expires is not http-compliant in the original cookie-spec,
		528	# we support the official date format and some extensions
		529	while (
		530	m{
		531	\G\s*
		532	(?:
		533	expires \s*=\s* ([A-Z][a-z][a-z]+,\ [^,;]+)
		534	| ([^=;,[:space:]]+) (?: \s*=\s* (?: "((?:[^\\"]+|\\.)*)" | ([^=;,[:space:]]*) ) )?
		535	)
		536	}gcxsi
		537	) {
		538	my $name = $2;
		539	my $value = $4;
		540
		541	if (defined $1) {
		542	# expires
		543	$name = "expires";
		544	$value = $1;
		545	} elsif (defined $3) {
		546	# quoted
		547	$value = $3;
		548	$value =~ s/\\(.)/$1/gs;
		549	}
		550
		551	push @kv, @kv ? lc $name : $name, $value;
		552
		553	last unless /\G\s*;/gc;
		554	}
		555
		556	last unless @kv;
		557
		558	my $name = shift @kv;
		559	my %kv = (value => shift @kv, @kv);
		560
		561	if (exists $kv{"max-age"}) {
		562	$kv{_expires} = $anow + delete $kv{"max-age"};
		563	} elsif (exists $kv{expires}) {
		564	$snow ||= parse_date ($date) || $anow;
		565	$kv{_expires} = $anow + (parse_date (delete $kv{expires}) - $snow);
		566	} else {
		567	delete $kv{_expires};
		568	}
		569
		570	my $cdom;
		571	my $cpath = (delete $kv{path}) || "/";
		572
		573	if (exists $kv{domain}) {
		574	$cdom = delete $kv{domain};
		575
		576	$cdom =~ s/^\.?/./; # make sure it starts with a "."
		577
		578	next if $cdom =~ /\.$/;
		579
		580	# this is not rfc-like and not netscape-like. go figure.
		581	my $ndots = $cdom =~ y/.//;
		582	next if $ndots < ($cdom =~ /\.[^.][^.]\.[^.][^.]$/ ? 3 : 2);
		583	} else {
		584	$cdom = $host;
		585	}
		586
		587	# store it
		588	$jar->{version} = 1;
		589	$jar->{lc $cdom}{$cpath}{$name} = \%kv;
		590
		591	redo if /\G\s*,/gc;
		592	}
		593	}
		594
		595	#############################################################################
		596	# keepalive/persistent connection cache
		597
		598	# fetch a connection from the keepalive cache
		599	sub ka_fetch($) {
		600	my $ka_key = shift;
		601
		602	my $hdl = pop @{ $KA_CACHE{$ka_key} }; # currently we reuse the MOST RECENTLY USED connection
		603	delete $KA_CACHE{$ka_key}
		604	unless @{ $KA_CACHE{$ka_key} };
		605
		606	$hdl
		607	}
		608
		609	sub ka_store($$) {
		610	my ($ka_key, $hdl) = @_;
		611
		612	my $kaa = $KA_CACHE{$ka_key} ||= [];
		613
		614	my $destroy = sub {
		615	my @ka = grep $_ != $hdl, @{ $KA_CACHE{$ka_key} };
		616
		617	$hdl->destroy;
		618
		619	@ka
		620	? $KA_CACHE{$ka_key} = \@ka
		621	: delete $KA_CACHE{$ka_key};
		622	};
		623
		624	# on error etc., destroy
		625	$hdl->on_error ($destroy);
		626	$hdl->on_eof ($destroy);
		627	$hdl->on_read ($destroy);
		628	$hdl->timeout ($PERSISTENT_TIMEOUT);
		629
		630	push @$kaa, $hdl;
		631	shift @$kaa while @$kaa > $MAX_PER_HOST;
		632	}
		633
		634	#############################################################################
		635	# utilities
		636
		637	# continue to parse $_ for headers and place them into the arg
		638	sub _parse_hdr() {
		639	my %hdr;
		640
		641	# things seen, not parsed:
		642	# p3pP="NON CUR OTPi OUR NOR UNI"
		643
		644	$hdr{lc $1} .= ",$2"
		645	while /\G
		646	([^:\000-\037]*):
		647	[\011\040]*
		648	((?: [^\012]+ | \012[\011\040] )*)
		649	\012
		650	/gxc;
		651
		652	/\G$/
		653	or return;
		654
		655	# remove the "," prefix we added to all headers above
		656	substr $_, 0, 1, ""
		657	for values %hdr;
		658
		659	\%hdr
		660	}
		661
		662	#############################################################################
		663	# http_get
		664
357	our $qr_nlnl = qr{(?<![^\012])\015?\012};	665	our $qr_nlnl = qr{(?<![^\012])\015?\012};
358		666
359	our $TLS_CTX_LOW = { cache => 1, sslv2 => 1 };	667	our $TLS_CTX_LOW = { cache => 1, sslv2 => 1 };
360	our $TLS_CTX_HIGH = { cache => 1, verify => 1, verify_peername => "https" };	668	our $TLS_CTX_HIGH = { cache => 1, verify => 1, verify_peername => "https" };
		669
		670	# maybe it should just become a normal object :/
		671
		672	sub _destroy_state(\%) {
		673	my ($state) = @_;
		674
		675	$state->{handle}->destroy if $state->{handle};
		676	%$state = ();
		677	}
		678
		679	sub _error(\%$$) {
		680	my ($state, $cb, $hdr) = @_;
		681
		682	&_destroy_state ($state);
		683
		684	$cb->(undef, $hdr);
		685	()
		686	}
361		687
362	sub http_request($$@) {	688	sub http_request($$@) {
363	my $cb = pop;	689	my $cb = pop;
364	my ($method, $url, %arg) = @_;	690	my ($method, $url, %arg) = @_;
365		691
…		…
386	if $recurse < 0;	712	if $recurse < 0;
387		713
388	my $proxy = $arg{proxy} || $PROXY;	714	my $proxy = $arg{proxy} || $PROXY;
389	my $timeout = $arg{timeout} || $TIMEOUT;	715	my $timeout = $arg{timeout} || $TIMEOUT;
390		716
391	my ($uscheme, $uauthority, $upath, $query, $fragment) =	717	my ($uscheme, $uauthority, $upath, $query, undef) = # ignore fragment
392	$url =~ m|(?:([^:/?#]+):)?(?://([^/?#]*))?([^?#]*)(?:(\?[^#]*))?(?:#(.*))?|;	718	$url =~ m|^([^:]+):(?://([^/?#]*))?([^?#]*)(?:(\?[^#]*))?(?:#(.*))?$|;
393		719
394	$uscheme = lc $uscheme;	720	$uscheme = lc $uscheme;
395		721
396	my $uport = $uscheme eq "http" ? 80	722	my $uport = $uscheme eq "http" ? 80
397	: $uscheme eq "https" ? 443	723	: $uscheme eq "https" ? 443
398	: return $cb->(undef, { @pseudo, Status => 599, Reason => "Only http and https URL schemes supported" });	724	: return $cb->(undef, { @pseudo, Status => 599, Reason => "Only http and https URL schemes supported" });
399		725
400	$uauthority =~ /^(?: .*\@ )? ([^\@:]+) (?: : (\d+) )?$/x	726	$uauthority =~ /^(?: .*\@ )? ([^\@:]+) (?: : (\d+) )?$/x
401	or return $cb->(undef, { @pseudo, Status => 599, Reason => "Unparsable URL" });	727	or return $cb->(undef, { @pseudo, Status => 599, Reason => "Unparsable URL" });
402		728
403	my $uhost = $1;	729	my $uhost = lc $1;
404	$uport = $2 if defined $2;	730	$uport = $2 if defined $2;
405		731
406	$hdr{host} = defined $2 ? "$uhost:$2" : "$uhost"	732	$hdr{host} = defined $2 ? "$uhost:$2" : "$uhost"
407	unless exists $hdr{host};	733	unless exists $hdr{host};
408		734
…		…
411		737
412	$upath =~ s%^/?%/%;	738	$upath =~ s%^/?%/%;
413		739
414	# cookie processing	740	# cookie processing
415	if (my $jar = $arg{cookie_jar}) {	741	if (my $jar = $arg{cookie_jar}) {
416	%$jar = () if $jar->{version} != 1;	742	my $cookies = cookie_jar_extract $jar, $uscheme, $uhost, $upath;
417		743
418	my @cookie;
419
420	while (my ($chost, $v) = each %$jar) {
421	if ($chost =~ /^\./) {
422	next unless $chost eq substr $uhost, -length $chost;
423	} elsif ($chost =~ /\./) {
424	next unless $chost eq $uhost;
425	} else {
426	next;
427	}
428
429	while (my ($cpath, $v) = each %$v) {
430	next unless $cpath eq substr $upath, 0, length $cpath;
431
432	while (my ($k, $v) = each %$v) {
433	next if $uscheme ne "https" && exists $v->{secure};
434	my $value = $v->{value};
435	$value =~ s/([\\"])/\\$1/g;
436	push @cookie, "$k=\"$value\"";
437	}
438	}
439	}
440
441	$hdr{cookie} = join "; ", @cookie	744	$hdr{cookie} = join "; ", @$cookies
442	if @cookie;	745	if @$cookies;
443	}	746	}
444		747
445	my ($rhost, $rport, $rscheme, $rpath); # request host, port, path	748	my ($rhost, $rport, $rscheme, $rpath); # request host, port, path
446		749
447	if ($proxy) {	750	if ($proxy) {
…		…
450	$rscheme = "http" unless defined $rscheme;	753	$rscheme = "http" unless defined $rscheme;
451		754
452	# don't support https requests over https-proxy transport,	755	# don't support https requests over https-proxy transport,
453	# can't be done with tls as spec'ed, unless you double-encrypt.	756	# can't be done with tls as spec'ed, unless you double-encrypt.
454	$rscheme = "http" if $uscheme eq "https" && $rscheme eq "https";	757	$rscheme = "http" if $uscheme eq "https" && $rscheme eq "https";
		758
		759	$rhost = lc $rhost;
		760	$rscheme = lc $rscheme;
455	} else {	761	} else {
456	($rhost, $rport, $rscheme, $rpath) = ($uhost, $uport, $uscheme, $upath);	762	($rhost, $rport, $rscheme, $rpath) = ($uhost, $uport, $uscheme, $upath);
457	}	763	}
458		764
459	# leave out fragment and query string, just a heuristic	765	# leave out fragment and query string, just a heuristic
460	$hdr{referer} ||= "$uscheme://$uauthority$upath" unless exists $hdr{referer};	766	$hdr{referer} = "$uscheme://$uauthority$upath" unless exists $hdr{referer};
461	$hdr{"user-agent"} ||= $USERAGENT unless exists $hdr{"user-agent"};	767	$hdr{"user-agent"} = $USERAGENT unless exists $hdr{"user-agent"};
462		768
463	$hdr{"content-length"} = length $arg{body}	769	$hdr{"content-length"} = length $arg{body}
464	if length $arg{body} || $method ne "GET";	770	if length $arg{body} || $method ne "GET";
465		771
		772	my $idempotent = $method =~ /^(?:GET|HEAD|PUT|DELETE|OPTIONS|TRACE)$/;
		773
		774	# default value for keepalive is true iff the request is for an idempotent method
		775	my $keepalive = exists $arg{keepalive} ? !!$arg{keepalive} : $idempotent;
		776	my $keepalive10 = exists $arg{keepalive10} ? $arg{keepalive10} : !$proxy;
		777	my $keptalive; # true if this is actually a recycled connection
		778
		779	# the key to use in the keepalive cache
		780	my $ka_key = "$uhost\x00$arg{sessionid}";
		781
		782	$hdr{connection} = ($keepalive ? $keepalive10 ? "keep-alive " : "" : "close ") . "Te"; #1.1
		783	$hdr{te} = "trailers" unless exists $hdr{te}; #1.1
		784
466	my %state = (connect_guard => 1);	785	my %state = (connect_guard => 1);
		786
		787	my $ae_error = 595; # connecting
		788
		789	# handle actual, non-tunneled, request
		790	my $handle_actual_request = sub {
		791	$ae_error = 596; # request phase
		792
		793	my $hdl = $state{handle};
		794
		795	$hdl->starttls ("connect") if $uscheme eq "https" && !exists $hdl->{tls};
		796
		797	# send request
		798	$hdl->push_write (
		799	"$method $rpath HTTP/1.1\015\012"
		800	. (join "", map "\u$_: $hdr{$_}\015\012", grep defined $hdr{$_}, keys %hdr)
		801	. "\015\012"
		802	. (delete $arg{body})
		803	);
		804
		805	# return if error occured during push_write()
		806	return unless %state;
		807
		808	# reduce memory usage, save a kitten, also re-use it for the response headers.
		809	%hdr = ();
		810
		811	# status line and headers
		812	$state{read_response} = sub {
		813	return unless %state;
		814
		815	for ("$_[1]") {
		816	y/\015//d; # weed out any \015, as they show up in the weirdest of places.
		817
		818	/^HTTP\/0*([0-9\.]+) \s+ ([0-9]{3}) (?: \s+ ([^\012]*) )? \012/gxci
		819	or return _error %state, $cb, { @pseudo, Status => 599, Reason => "Invalid server response" };
		820
		821	# 100 Continue handling
		822	# should not happen as we don't send expect: 100-continue,
		823	# but we handle it just in case.
		824	# since we send the request body regardless, if we get an error
		825	# we are out of-sync, which we currently do NOT handle correctly.
		826	return $state{handle}->push_read (line => $qr_nlnl, $state{read_response})
		827	if $2 eq 100;
		828
		829	push @pseudo,
		830	HTTPVersion => $1,
		831	Status => $2,
		832	Reason => $3,
		833	;
		834
		835	my $hdr = _parse_hdr
		836	or return _error %state, $cb, { @pseudo, Status => 599, Reason => "Garbled response headers" };
		837
		838	%hdr = (%$hdr, @pseudo);
		839	}
		840
		841	# redirect handling
		842	# microsoft and other shitheads don't give a shit for following standards,
		843	# try to support some common forms of broken Location headers.
		844	if ($hdr{location} !~ /^(?: $ | [^:\/?\#]+ : )/x) {
		845	$hdr{location} =~ s/^\.\/+//;
		846
		847	my $url = "$rscheme://$uhost:$uport";
		848
		849	unless ($hdr{location} =~ s/^\///) {
		850	$url .= $upath;
		851	$url =~ s/\/[^\/]*$//;
		852	}
		853
		854	$hdr{location} = "$url/$hdr{location}";
		855	}
		856
		857	my $redirect;
		858
		859	if ($recurse) {
		860	my $status = $hdr{Status};
		861
		862	# industry standard is to redirect POST as GET for
		863	# 301, 302 and 303, in contrast to HTTP/1.0 and 1.1.
		864	# also, the UA should ask the user for 301 and 307 and POST,
		865	# industry standard seems to be to simply follow.
		866	# we go with the industry standard.
		867	if ($status == 301 or $status == 302 or $status == 303) {
		868	# HTTP/1.1 is unclear on how to mutate the method
		869	$method = "GET" unless $method eq "HEAD";
		870	$redirect = 1;
		871	} elsif ($status == 307) {
		872	$redirect = 1;
		873	}
		874	}
		875
		876	my $finish = sub { # ($data, $err_status, $err_reason[, $keepalive])
		877	if ($state{handle}) {
		878	# handle keepalive
		879	if (
		880	$keepalive
		881	&& $_[3]
		882	&& ($hdr{HTTPVersion} < 1.1
		883	? $hdr{connection} =~ /\bkeep-?alive\b/i
		884	: $hdr{connection} !~ /\bclose\b/i)
		885	) {
		886	ka_store $ka_key, delete $state{handle};
		887	} else {
		888	# no keepalive, destroy the handle
		889	$state{handle}->destroy;
		890	}
		891	}
		892
		893	%state = ();
		894
		895	if (defined $_[1]) {
		896	$hdr{OrigStatus} = $hdr{Status}; $hdr{Status} = $_[1];
		897	$hdr{OrigReason} = $hdr{Reason}; $hdr{Reason} = $_[2];
		898	}
		899
		900	# set-cookie processing
		901	if ($arg{cookie_jar}) {
		902	cookie_jar_set_cookie $arg{cookie_jar}, $hdr{"set-cookie"}, $uhost, $hdr{date};
		903	}
		904
		905	if ($redirect && exists $hdr{location}) {
		906	# we ignore any errors, as it is very common to receive
		907	# Content-Length != 0 but no actual body
		908	# we also access %hdr, as $_[1] might be an erro
		909	http_request (
		910	$method => $hdr{location},
		911	%arg,
		912	recurse => $recurse - 1,
		913	Redirect => [$_[0], \%hdr],
		914	$cb
		915	);
		916	} else {
		917	$cb->($_[0], \%hdr);
		918	}
		919	};
		920
		921	$ae_error = 597; # body phase
		922
		923	my $chunked = $hdr{"transfer-encoding"} =~ /\bchunked\b/i; # not quite correct...
		924
		925	my $len = $chunked ? undef : $hdr{"content-length"};
		926
		927	# body handling, many different code paths
		928	# - no body expected
		929	# - want_body_handle
		930	# - te chunked
		931	# - 2x length known (with or without on_body)
		932	# - 2x length not known (with or without on_body)
		933	if (!$redirect && $arg{on_header} && !$arg{on_header}(\%hdr)) {
		934	$finish->(undef, 598 => "Request cancelled by on_header");
		935	} elsif (
		936	$hdr{Status} =~ /^(?:1..|204|205|304)$/
		937	or $method eq "HEAD"
		938	or (defined $len && $len == 0) # == 0, not !, because "0 " is true
		939	) {
		940	# no body
		941	$finish->("", undef, undef, 1);
		942
		943	} elsif (!$redirect && $arg{want_body_handle}) {
		944	$_[0]->on_eof (undef);
		945	$_[0]->on_error (undef);
		946	$_[0]->on_read (undef);
		947
		948	$finish->(delete $state{handle});
		949
		950	} elsif ($chunked) {
		951	my $cl = 0;
		952	my $body = "";
		953	my $on_body = $arg{on_body} || sub { $body .= shift; 1 };
		954
		955	$state{read_chunk} = sub {
		956	$_[1] =~ /^([0-9a-fA-F]+)/
		957	or $finish->(undef, $ae_error => "Garbled chunked transfer encoding");
		958
		959	my $len = hex $1;
		960
		961	if ($len) {
		962	$cl += $len;
		963
		964	$_[0]->push_read (chunk => $len, sub {
		965	$on_body->($_[1], \%hdr)
		966	or return $finish->(undef, 598 => "Request cancelled by on_body");
		967
		968	$_[0]->push_read (line => sub {
		969	length $_[1]
		970	and return $finish->(undef, $ae_error => "Garbled chunked transfer encoding");
		971	$_[0]->push_read (line => $state{read_chunk});
		972	});
		973	});
		974	} else {
		975	$hdr{"content-length"} ||= $cl;
		976
		977	$_[0]->push_read (line => $qr_nlnl, sub {
		978	if (length $_[1]) {
		979	for ("$_[1]") {
		980	y/\015//d; # weed out any \015, as they show up in the weirdest of places.
		981
		982	my $hdr = _parse_hdr
		983	or return $finish->(undef, $ae_error => "Garbled response trailers");
		984
		985	%hdr = (%hdr, %$hdr);
		986	}
		987	}
		988
		989	$finish->($body, undef, undef, 1);
		990	});
		991	}
		992	};
		993
		994	$_[0]->push_read (line => $state{read_chunk});
		995
		996	} elsif ($arg{on_body}) {
		997	if (defined $len) {
		998	$_[0]->on_read (sub {
		999	$len -= length $_[0]{rbuf};
		1000
		1001	$arg{on_body}(delete $_[0]{rbuf}, \%hdr)
		1002	or return $finish->(undef, 598 => "Request cancelled by on_body");
		1003
		1004	$len > 0
		1005	or $finish->("", undef, undef, 1);
		1006	});
		1007	} else {
		1008	$_[0]->on_eof (sub {
		1009	$finish->("");
		1010	});
		1011	$_[0]->on_read (sub {
		1012	$arg{on_body}(delete $_[0]{rbuf}, \%hdr)
		1013	or $finish->(undef, 598 => "Request cancelled by on_body");
		1014	});
		1015	}
		1016	} else {
		1017	$_[0]->on_eof (undef);
		1018
		1019	if (defined $len) {
		1020	$_[0]->on_read (sub {
		1021	$finish->((substr delete $_[0]{rbuf}, 0, $len, ""), undef, undef, 1)
		1022	if $len <= length $_[0]{rbuf};
		1023	});
		1024	} else {
		1025	$_[0]->on_error (sub {
		1026	($! == Errno::EPIPE || !$!)
		1027	? $finish->(delete $_[0]{rbuf})
		1028	: $finish->(undef, $ae_error => $_[2]);
		1029	});
		1030	$_[0]->on_read (sub { });
		1031	}
		1032	}
		1033	};
		1034
		1035	# if keepalive is enabled, then the server closing the connection
		1036	# before a response can happen legally - we retry on idempotent methods.
		1037	if ($keptalive && $idempotent) {
		1038	my $old_eof = $hdl->{on_eof};
		1039	$hdl->{on_eof} = sub {
		1040	_destroy_state %state;
		1041
		1042	http_request (
		1043	$method => $url,
		1044	%arg,
		1045	keepalive => 0,
		1046	$cb
		1047	);
		1048	};
		1049	$hdl->on_read (sub {
		1050	return unless %state;
		1051
		1052	# as soon as we receive something, a connection close
		1053	# once more becomes a hard error
		1054	$hdl->{on_eof} = $old_eof;
		1055	$hdl->push_read (line => $qr_nlnl, $state{read_response});
		1056	});
		1057	} else {
		1058	$hdl->push_read (line => $qr_nlnl, $state{read_response});
		1059	}
		1060	};
		1061
		1062	my $prepare_handle = sub {
		1063	my ($hdl) = $state{handle};
		1064
		1065	$hdl->timeout ($timeout);
		1066	$hdl->on_error (sub {
		1067	_error %state, $cb, { @pseudo, Status => $ae_error, Reason => $_[2] };
		1068	});
		1069	$hdl->on_eof (sub {
		1070	_error %state, $cb, { @pseudo, Status => $ae_error, Reason => "Unexpected end-of-file" };
		1071	});
		1072	};
		1073
		1074	# connected to proxy (or origin server)
		1075	my $connect_cb = sub {
		1076	my $fh = shift
		1077	or return _error %state, $cb, { @pseudo, Status => $ae_error, Reason => "$!" };
		1078
		1079	return unless delete $state{connect_guard};
		1080
		1081	# get handle
		1082	$state{handle} = new AnyEvent::Handle
		1083	%{ $arg{handle_params} },
		1084	fh => $fh,
		1085	peername => $uhost,
		1086	tls_ctx => $arg{tls_ctx},
		1087	;
		1088
		1089	$prepare_handle->();
		1090
		1091	#$state{handle}->starttls ("connect") if $rscheme eq "https";
		1092
		1093	# now handle proxy-CONNECT method
		1094	if ($proxy && $uscheme eq "https") {
		1095	# oh dear, we have to wrap it into a connect request
		1096
		1097	# maybe re-use $uauthority with patched port?
		1098	$state{handle}->push_write ("CONNECT $uhost:$uport HTTP/1.0\015\012\015\012");
		1099	$state{handle}->push_read (line => $qr_nlnl, sub {
		1100	$_[1] =~ /^HTTP\/([0-9\.]+) \s+ ([0-9]{3}) (?: \s+ ([^\015\012]*) )?/ix
		1101	or return _error %state, $cb, { @pseudo, Status => 599, Reason => "Invalid proxy connect response ($_[1])" };
		1102
		1103	if ($2 == 200) {
		1104	$rpath = $upath;
		1105	$handle_actual_request->();
		1106	} else {
		1107	_error %state, $cb, { @pseudo, Status => $2, Reason => $3 };
		1108	}
		1109	});
		1110	} else {
		1111	$handle_actual_request->();
		1112	}
		1113	};
467		1114
468	_get_slot $uhost, sub {	1115	_get_slot $uhost, sub {
469	$state{slot_guard} = shift;	1116	$state{slot_guard} = shift;
470		1117
471	return unless $state{connect_guard};	1118	return unless $state{connect_guard};
472		1119
473	my $connect_cb = sub {	1120	# try to use an existing keepalive connection, but only if we, ourselves, plan
474	$state{fh} = shift	1121	# on a keepalive request (in theory, this should be a separate config option).
475	or do {	1122	if ($keepalive && $KA_CACHE{$ka_key}) {
476	my $err = "$!";	1123	$keptalive = 1;
477	%state = ();	1124	$state{handle} = ka_fetch $ka_key;
478	return $cb->(undef, { @pseudo, Status => 599, Reason => $err });	1125	$prepare_handle->();
479	};
480
481	pop; # free memory, save a tree
482
483	return unless delete $state{connect_guard};
484
485	# get handle
486	$state{handle} = new AnyEvent::Handle
487	fh => $state{fh},
488	peername => $rhost,
489	tls_ctx => $arg{tls_ctx},
490	# these need to be reconfigured on keepalive handles
491	timeout => $timeout,
492	on_error => sub {
493	%state = ();
494	$cb->(undef, { @pseudo, Status => 599, Reason => $_[2] });
495	},
496	on_eof => sub {
497	%state = ();
498	$cb->(undef, { @pseudo, Status => 599, Reason => "Unexpected end-of-file" });
499	},
500	;
501
502	# limit the number of persistent connections
503	# keepalive not yet supported
504	# if ($KA_COUNT{$_[1]} < $MAX_PERSISTENT_PER_HOST) {
505	# ++$KA_COUNT{$_[1]};
506	# $state{handle}{ka_count_guard} = AnyEvent::Util::guard {
507	# --$KA_COUNT{$_[1]}
508	# };
509	# $hdr{connection} = "keep-alive";
510	# } else {
511	delete $hdr{connection};
512	# }
513
514	$state{handle}->starttls ("connect") if $rscheme eq "https";
515
516	# handle actual, non-tunneled, request
517	my $handle_actual_request = sub {
518	$state{handle}->starttls ("connect") if $uscheme eq "https" && !exists $state{handle}{tls};
519
520	# send request
521	$state{handle}->push_write (
522	"$method $rpath HTTP/1.0\015\012"
523	. (join "", map "\u$_: $hdr{$_}\015\012", grep defined $hdr{$_}, keys %hdr)
524	. "\015\012"
525	. (delete $arg{body})
526	);
527
528	# return if error occured during push_write()
529	return unless %state;
530
531	%hdr = (); # reduce memory usage, save a kitten, also make it possible to re-use
532
533	# status line and headers
534	$state{handle}->push_read (line => $qr_nlnl, sub {
535	my $keepalive = pop;
536
537	for ("$_[1]") {
538	y/\015//d; # weed out any \015, as they show up in the weirdest of places.
539
540	/^HTTP\/([0-9\.]+) \s+ ([0-9]{3}) (?: \s+ ([^\015\012]*) )? \015?\012/igxc
541	or return (%state = (), $cb->(undef, { @pseudo, Status => 599, Reason => "Invalid server response" }));
542
543	push @pseudo,
544	HTTPVersion => $1,
545	Status => $2,
546	Reason => $3,
547	;
548
549	# things seen, not parsed:
550	# p3pP="NON CUR OTPi OUR NOR UNI"
551
552	$hdr{lc $1} .= ",$2"
553	while /\G
554	([^:\000-\037]*):
555	[\011\040]*
556	((?: [^\012]+ | \012[\011\040] )*)
557	\012
558	/gxc;
559
560	/\G$/
561	or return (%state = (), $cb->(undef, { @pseudo, Status => 599, Reason => "Garbled response headers" }));
562	}
563
564	# remove the "," prefix we added to all headers above
565	substr $_, 0, 1, ""
566	for values %hdr;
567
568	# patch in all pseudo headers
569	%hdr = (%hdr, @pseudo);
570
571	# redirect handling
572	# microsoft and other shitheads don't give a shit for following standards,
573	# try to support some common forms of broken Location headers.
574	if ($hdr{location} !~ /^(?: $ | [^:\/?\#]+ : )/x) {
575	$hdr{location} =~ s/^\.\/+//;
576
577	my $url = "$rscheme://$uhost:$uport";
578
579	unless ($hdr{location} =~ s/^\///) {
580	$url .= $upath;
581	$url =~ s/\/[^\/]*$//;
582	}
583
584	$hdr{location} = "$url/$hdr{location}";
585	}
586
587	my $redirect;
588
589	if ($recurse) {
590	my $status = $hdr{Status};
591
592	# industry standard is to redirect POST as GET for
593	# 301, 302 and 303, in contrast to http/1.0 and 1.1.
594	# also, the UA should ask the user for 301 and 307 and POST,
595	# industry standard seems to be to simply follow.
596	# we go with the industry standard.
597	if ($status == 301 or $status == 302 or $status == 303) {
598	# HTTP/1.1 is unclear on how to mutate the method
599	$method = "GET" unless $method eq "HEAD";
600	$redirect = 1;
601	} elsif ($status == 307) {
602	$redirect = 1;
603	}
604	}
605
606	my $finish = sub { # ($data, $err_status, $err_reason[, $keepalive])
607	$state{handle}->destroy if $state{handle};
608	%state = ();
609
610	if (defined $_[1]) {
611	$hdr{OrigStatus} = $hdr{Status}; $hdr{Status} = $_[1];
612	$hdr{OrigReason} = $hdr{Reason}; $hdr{Reason} = $_[2];
613	}
614
615	# set-cookie processing
616	if ($arg{cookie_jar}) {
617	for ($hdr{"set-cookie"}) {
618	# parse NAME=VALUE
619	my @kv;
620
621	while (/\G\s* ([^=;,[:space:]]+) \s*=\s* (?: "((?:[^\\"]+|\\.)*)" | ([^=;,[:space:]]*) )/gcxs) {
622	my $name = $1;
623	my $value = $3;
624
625	unless ($value) {
626	$value = $2;
627	$value =~ s/\\(.)/$1/gs;
628	}
629
630	push @kv, $name => $value;
631
632	last unless /\G\s*;/gc;
633	}
634
635	last unless @kv;
636
637	my $name = shift @kv;
638	my %kv = (value => shift @kv, @kv);
639
640	my $cdom;
641	my $cpath = (delete $kv{path}) || "/";
642
643	if (exists $kv{domain}) {
644	$cdom = delete $kv{domain};
645
646	$cdom =~ s/^\.?/./; # make sure it starts with a "."
647
648	next if $cdom =~ /\.$/;
649
650	# this is not rfc-like and not netscape-like. go figure.
651	my $ndots = $cdom =~ y/.//;
652	next if $ndots < ($cdom =~ /\.[^.][^.]\.[^.][^.]$/ ? 3 : 2);
653	} else {
654	$cdom = $uhost;
655	}
656
657	# store it
658	$arg{cookie_jar}{version} = 1;
659	$arg{cookie_jar}{$cdom}{$cpath}{$name} = \%kv;
660
661	redo if /\G\s*,/gc;
662	}
663	}
664
665	if ($redirect && exists $hdr{location}) {
666	# we ignore any errors, as it is very common to receive
667	# Content-Length != 0 but no actual body
668	# we also access %hdr, as $_[1] might be an erro
669	http_request (
670	$method => $hdr{location},
671	%arg,
672	recurse => $recurse - 1,
673	Redirect => [$_[0], \%hdr],
674	$cb);
675	} else {
676	$cb->($_[0], \%hdr);
677	}
678	};
679
680	my $len = $hdr{"content-length"};
681
682	if (!$redirect && $arg{on_header} && !$arg{on_header}(\%hdr)) {
683	$finish->(undef, 598 => "Request cancelled by on_header");
684	} elsif (
685	$hdr{Status} =~ /^(?:1..|204|205|304)$/
686	or $method eq "HEAD"
687	or (defined $len && !$len)
688	) {
689	# no body
690	$finish->("", undef, undef, 1);
691	} else {
692	# body handling, four different code paths
693	# for want_body_handle, on_body (2x), normal (2x)
694	# we might read too much here, but it does not matter yet (no pipelining)
695	if (!$redirect && $arg{want_body_handle}) {
696	$_[0]->on_eof (undef);
697	$_[0]->on_error (undef);
698	$_[0]->on_read (undef);
699
700	$finish->(delete $state{handle});
701
702	} elsif ($arg{on_body}) {
703	$_[0]->on_error (sub { $finish->(undef, 599 => $_[2]) });
704	if ($len) {
705	$_[0]->on_read (sub {
706	$len -= length $_[0]{rbuf};
707
708	$arg{on_body}(delete $_[0]{rbuf}, \%hdr)
709	or $finish->(undef, 598 => "Request cancelled by on_body");
710
711	$len > 0
712	or $finish->("", undef, undef, 1);
713	});
714	} else {
715	$_[0]->on_eof (sub {
716	$finish->("");
717	});
718	$_[0]->on_read (sub {
719	$arg{on_body}(delete $_[0]{rbuf}, \%hdr)
720	or $finish->(undef, 598 => "Request cancelled by on_body");
721	});
722	}
723	} else {
724	$_[0]->on_eof (undef);
725
726	if ($len) {
727	$_[0]->on_error (sub { $finish->(undef, 599 => $_[2]) });
728	$_[0]->on_read (sub {
729	$finish->((substr delete $_[0]{rbuf}, 0, $len, ""), undef, undef, 1)
730	if $len <= length $_[0]{rbuf};
731	});
732	} else {
733	$_[0]->on_error (sub {
734	($! == Errno::EPIPE || !$!)
735	? $finish->(delete $_[0]{rbuf})
736	: $finish->(undef, 599 => $_[2]);
737	});
738	$_[0]->on_read (sub { });
739	}
740	}
741	}
742	});
743	};
744
745	# now handle proxy-CONNECT method
746	if ($proxy && $uscheme eq "https") {
747	# oh dear, we have to wrap it into a connect request
748
749	# maybe re-use $uauthority with patched port?
750	$state{handle}->push_write ("CONNECT $uhost:$uport HTTP/1.0\015\012Host: $uhost\015\012\015\012");
751	$state{handle}->push_read (line => $qr_nlnl, sub {
752	$_[1] =~ /^HTTP\/([0-9\.]+) \s+ ([0-9]{3}) (?: \s+ ([^\015\012]*) )?/ix
753	or return (%state = (), $cb->(undef, { @pseudo, Status => 599, Reason => "Invalid proxy connect response ($_[1])" }));
754
755	if ($2 == 200) {
756	$rpath = $upath;
757	&$handle_actual_request;	1126	$handle_actual_request->();
758	} else {	1127
759	%state = ();
760	$cb->(undef, { @pseudo, Status => $2, Reason => $3 });
761	}
762	});
763	} else {	1128	} else {
764	&$handle_actual_request;
765	}
766	};
767
768	my $tcp_connect = $arg{tcp_connect}	1129	my $tcp_connect = $arg{tcp_connect}
769	|| do { require AnyEvent::Socket; \&AnyEvent::Socket::tcp_connect };	1130	|| do { require AnyEvent::Socket; \&AnyEvent::Socket::tcp_connect };
770		1131
771	$state{connect_guard} = $tcp_connect->($rhost, $rport, $connect_cb, $arg{on_prepare} || sub { $timeout });	1132	$state{connect_guard} = $tcp_connect->($rhost, $rport, $connect_cb, $arg{on_prepare} || sub { $timeout });
772		1133	}
773	};	1134	};
774		1135
775	defined wantarray && AnyEvent::Util::guard { %state = () }	1136	defined wantarray && AnyEvent::Util::guard { _destroy_state %state }
776	}	1137	}
777		1138
778	sub http_get($@) {	1139	sub http_get($@) {
779	unshift @_, "GET";	1140	unshift @_, "GET";
780	&http_request	1141	&http_request
…		…
798	AnyEvent::HTTP uses the AnyEvent::Socket::tcp_connect function for	1159	AnyEvent::HTTP uses the AnyEvent::Socket::tcp_connect function for
799	the actual connection, which in turn uses AnyEvent::DNS to resolve	1160	the actual connection, which in turn uses AnyEvent::DNS to resolve
800	hostnames. The latter is a simple stub resolver and does no caching	1161	hostnames. The latter is a simple stub resolver and does no caching
801	on its own. If you want DNS caching, you currently have to provide	1162	on its own. If you want DNS caching, you currently have to provide
802	your own default resolver (by storing a suitable resolver object in	1163	your own default resolver (by storing a suitable resolver object in
803	C<$AnyEvent::DNS::RESOLVER>).	1164	C<$AnyEvent::DNS::RESOLVER>) or your own C<tcp_connect> callback.
804		1165
805	=head2 GLOBAL FUNCTIONS AND VARIABLES	1166	=head2 GLOBAL FUNCTIONS AND VARIABLES
806		1167
807	=over 4	1168	=over 4
808		1169
809	=item AnyEvent::HTTP::set_proxy "proxy-url"	1170	=item AnyEvent::HTTP::set_proxy "proxy-url"
810		1171
811	Sets the default proxy server to use. The proxy-url must begin with a	1172	Sets the default proxy server to use. The proxy-url must begin with a
812	string of the form C<http://host:port> (optionally C<https:...>), croaks	1173	string of the form C<http://host:port>, croaks otherwise.
813	otherwise.
814		1174
815	To clear an already-set proxy, use C<undef>.	1175	To clear an already-set proxy, use C<undef>.
		1176
		1177	=item AnyEvent::HTTP::cookie_jar_expire $jar[, $session_end]
		1178
		1179	Remove all cookies from the cookie jar that have been expired. If
		1180	C<$session_end> is given and true, then additionally remove all session
		1181	cookies.
		1182
		1183	You should call this function (with a true C<$session_end>) before you
		1184	save cookies to disk, and you should call this function after loading them
		1185	again. If you have a long-running program you can additonally call this
		1186	function from time to time.
		1187
		1188	A cookie jar is initially an empty hash-reference that is managed by this
		1189	module. It's format is subject to change, but currently it is like this:
		1190
		1191	The key C<version> has to contain C<1>, otherwise the hash gets
		1192	emptied. All other keys are hostnames or IP addresses pointing to
		1193	hash-references. The key for these inner hash references is the
		1194	server path for which this cookie is meant, and the values are again
		1195	hash-references. The keys of those hash-references is the cookie name, and
		1196	the value, you guessed it, is another hash-reference, this time with the
		1197	key-value pairs from the cookie, except for C<expires> and C<max-age>,
		1198	which have been replaced by a C<_expires> key that contains the cookie
		1199	expiry timestamp.
		1200
		1201	Here is an example of a cookie jar with a single cookie, so you have a
		1202	chance of understanding the above paragraph:
		1203
		1204	{
		1205	version => 1,
		1206	"10.0.0.1" => {
		1207	"/" => {
		1208	"mythweb_id" => {
		1209	_expires => 1293917923,
		1210	value => "ooRung9dThee3ooyXooM1Ohm",
		1211	},
		1212	},
		1213	},
		1214	}
816		1215
817	=item $date = AnyEvent::HTTP::format_date $timestamp	1216	=item $date = AnyEvent::HTTP::format_date $timestamp
818		1217
819	Takes a POSIX timestamp (seconds since the epoch) and formats it as a HTTP	1218	Takes a POSIX timestamp (seconds since the epoch) and formats it as a HTTP
820	Date (RFC 2616).	1219	Date (RFC 2616).
821		1220
822	=item $timestamp = AnyEvent::HTTP::parse_date $date	1221	=item $timestamp = AnyEvent::HTTP::parse_date $date
823		1222
824	Takes a HTTP Date (RFC 2616) and returns the corresponding POSIX	1223	Takes a HTTP Date (RFC 2616) or a Cookie date (netscape cookie spec) or a
		1224	bunch of minor variations of those, and returns the corresponding POSIX
825	timestamp, or C<undef> if the date cannot be parsed.	1225	timestamp, or C<undef> if the date cannot be parsed.
826		1226
827	=item $AnyEvent::HTTP::MAX_RECURSE	1227	=item $AnyEvent::HTTP::MAX_RECURSE
828		1228
829	The default value for the C<recurse> request parameter (default: C<10>).	1229	The default value for the C<recurse> request parameter (default: C<10>).
		1230
		1231	=item $AnyEvent::HTTP::TIMEOUT
		1232
		1233	The default timeout for conenction operations (default: C<300>).
830		1234
831	=item $AnyEvent::HTTP::USERAGENT	1235	=item $AnyEvent::HTTP::USERAGENT
832		1236
833	The default value for the C<User-Agent> header (the default is	1237	The default value for the C<User-Agent> header (the default is
834	C<Mozilla/5.0 (compatible; U; AnyEvent-HTTP/$VERSION; +http://software.schmorp.de/pkg/AnyEvent)>).	1238	C<Mozilla/5.0 (compatible; U; AnyEvent-HTTP/$VERSION; +http://software.schmorp.de/pkg/AnyEvent)>).
835		1239
836	=item $AnyEvent::HTTP::MAX_PER_HOST	1240	=item $AnyEvent::HTTP::MAX_PER_HOST
837		1241
838	The maximum number of concurrent connections to the same host (identified	1242	The maximum number of concurrent connections to the same host (identified
839	by the hostname). If the limit is exceeded, then the additional requests	1243	by the hostname). If the limit is exceeded, then the additional requests
840	are queued until previous connections are closed.	1244	are queued until previous connections are closed. Both persistent and
		1245	non-persistent connections are counted in this limit.
841		1246
842	The default value for this is C<4>, and it is highly advisable to not	1247	The default value for this is C<4>, and it is highly advisable to not
843	increase it.	1248	increase it much.
		1249
		1250	For comparison: the RFC's recommend 4 non-persistent or 2 persistent
		1251	connections, older browsers used 2, newers (such as firefox 3) typically
		1252	use 6, and Opera uses 8 because like, they have the fastest browser and
		1253	give a shit for everybody else on the planet.
		1254
		1255	=item $AnyEvent::HTTP::PERSISTENT_TIMEOUT
		1256
		1257	The time after which idle persistent conenctions get closed by
		1258	AnyEvent::HTTP (default: C<3>).
844		1259
845	=item $AnyEvent::HTTP::ACTIVE	1260	=item $AnyEvent::HTTP::ACTIVE
846		1261
847	The number of active connections. This is not the number of currently	1262	The number of active connections. This is not the number of currently
848	running requests, but the number of currently open and non-idle TCP	1263	running requests, but the number of currently open and non-idle TCP
849	connections. This number of can be useful for load-leveling.	1264	connections. This number can be useful for load-leveling.
850		1265
851	=back	1266	=back
852		1267
853	=cut	1268	=cut
854		1269
…		…
869	sub parse_date($) {	1284	sub parse_date($) {
870	my ($date) = @_;	1285	my ($date) = @_;
871		1286
872	my ($d, $m, $y, $H, $M, $S);	1287	my ($d, $m, $y, $H, $M, $S);
873		1288
874	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$/) {	1289	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$/) {
875	# RFC 822/1123, required by RFC 2616	1290	# RFC 822/1123, required by RFC 2616 (with " ")
		1291	# cookie dates (with "-")
		1292
876	($d, $m, $y, $H, $M, $S) = ($1, $2, $3, $4, $5, $6);	1293	($d, $m, $y, $H, $M, $S) = ($1, $2, $3, $4, $5, $6);
877		1294
878	} elsif ($date =~ /^[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$/) {	1295	} 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$/) {
879	# RFC 850	1296	# RFC 850
880	($d, $m, $y, $H, $M, $S) = ($1, $2, $3 < 69 ? $3 + 2000 : $3 + 1900, $4, $5, $6);	1297	($d, $m, $y, $H, $M, $S) = ($1, $2, $3 < 69 ? $3 + 2000 : $3 + 1900, $4, $5, $6);
881		1298
882	} 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])$/) {	1299	} 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])$/) {
883	# ISO C's asctime	1300	# ISO C's asctime
884	($d, $m, $y, $H, $M, $S) = ($2, $1, $6, $3, $4, $5);	1301	($d, $m, $y, $H, $M, $S) = ($2, $1, $6, $3, $4, $5);
885	}	1302	}
886	# other formats fail in the loop below	1303	# other formats fail in the loop below
887		1304
…		…
895	undef	1312	undef
896	}	1313	}
897		1314
898	sub set_proxy($) {	1315	sub set_proxy($) {
899	if (length $_[0]) {	1316	if (length $_[0]) {
900	$_[0] =~ m%^(https?):// ([^:/]+) (?: : (\d*) )?%ix	1317	$_[0] =~ m%^(http):// ([^:/]+) (?: : (\d*) )?%ix
901	or Carp::croak "$_[0]: invalid proxy URL";	1318	or Carp::croak "$_[0]: invalid proxy URL";
902	$PROXY = [$2, $3 || 3128, $1]	1319	$PROXY = [$2, $3 || 3128, $1]
903	} else {	1320	} else {
904	undef $PROXY;	1321	undef $PROXY;
905	}	1322	}
…		…
908	# initialise proxy from environment	1325	# initialise proxy from environment
909	eval {	1326	eval {
910	set_proxy $ENV{http_proxy};	1327	set_proxy $ENV{http_proxy};
911	};	1328	};
912		1329
		1330	=head2 SHOWCASE
		1331
		1332	This section contaisn some more elaborate "real-world" examples or code
		1333	snippets.
		1334
		1335	=head2 HTTP/1.1 FILE DOWNLOAD
		1336
		1337	Downloading files with HTTP can be quite tricky, especially when something
		1338	goes wrong and you want to resume.
		1339
		1340	Here is a function that initiates and resumes a download. It uses the
		1341	last modified time to check for file content changes, and works with many
		1342	HTTP/1.0 servers as well, and usually falls back to a complete re-download
		1343	on older servers.
		1344
		1345	It calls the completion callback with either C<undef>, which means a
		1346	nonretryable error occured, C<0> when the download was partial and should
		1347	be retried, and C<1> if it was successful.
		1348
		1349	use AnyEvent::HTTP;
		1350
		1351	sub download($$$) {
		1352	my ($url, $file, $cb) = @_;
		1353
		1354	open my $fh, "+<", $file
		1355	or die "$file: $!";
		1356
		1357	my %hdr;
		1358	my $ofs = 0;
		1359
		1360	warn stat $fh;
		1361	warn -s _;
		1362	if (stat $fh and -s _) {
		1363	$ofs = -s _;
		1364	warn "-s is ", $ofs;#d#
		1365	$hdr{"if-unmodified-since"} = AnyEvent::HTTP::format_date +(stat _)[9];
		1366	$hdr{"range"} = "bytes=$ofs-";
		1367	}
		1368
		1369	http_get $url,
		1370	headers => \%hdr,
		1371	on_header => sub {
		1372	my ($hdr) = @_;
		1373
		1374	if ($hdr->{Status} == 200 && $ofs) {
		1375	# resume failed
		1376	truncate $fh, $ofs = 0;
		1377	}
		1378
		1379	sysseek $fh, $ofs, 0;
		1380
		1381	1
		1382	},
		1383	on_body => sub {
		1384	my ($data, $hdr) = @_;
		1385
		1386	if ($hdr->{Status} =~ /^2/) {
		1387	length $data == syswrite $fh, $data
		1388	or return; # abort on write errors
		1389	}
		1390
		1391	1
		1392	},
		1393	sub {
		1394	my (undef, $hdr) = @_;
		1395
		1396	my $status = $hdr->{Status};
		1397
		1398	if (my $time = AnyEvent::HTTP::parse_date $hdr->{"last-modified"}) {
		1399	utime $fh, $time, $time;
		1400	}
		1401
		1402	if ($status == 200 || $status == 206 || $status == 416) {
		1403	# download ok || resume ok || file already fully downloaded
		1404	$cb->(1, $hdr);
		1405
		1406	} elsif ($status == 412) {
		1407	# file has changed while resuming, delete and retry
		1408	unlink $file;
		1409	$cb->(0, $hdr);
		1410
		1411	} elsif ($status == 500 or $status == 503 or $status =~ /^59/) {
		1412	# retry later
		1413	$cb->(0, $hdr);
		1414
		1415	} else {
		1416	$cb->(undef, $hdr);
		1417	}
		1418	}
		1419	;
		1420	}
		1421
		1422	download "http://server/somelargefile", "/tmp/somelargefile", sub {
		1423	if ($_[0]) {
		1424	print "OK!\n";
		1425	} elsif (defined $_[0]) {
		1426	print "please retry later\n";
		1427	} else {
		1428	print "ERROR\n";
		1429	}
		1430	};
		1431
913	=head2 SOCKS PROXIES	1432	=head3 SOCKS PROXIES
914		1433
915	Socks proxies are not directly supported by AnyEvent::HTTP. You can	1434	Socks proxies are not directly supported by AnyEvent::HTTP. You can
916	compile your perl to support socks, or use an external program such as	1435	compile your perl to support socks, or use an external program such as
917	F<socksify> (dante) or F<tsocks> to make your program use a socks proxy	1436	F<socksify> (dante) or F<tsocks> to make your program use a socks proxy
918	transparently.	1437	transparently.

Diff Legend

–	Removed lines
+	Added lines
<	Changed lines
>	Changed lines