[ViewVC] Diff of: cvs/AnyEvent-MP/MP.pm

Comparing AnyEvent-MP/MP.pm (file contents):
Revision 1.1 by root, Thu Jul 30 08:38:50 2009 UTC vs.
Revision 1.108 by root, Wed Dec 30 13:37:53 2009 UTC

1	=head1 NAME	1	=head1 NAME
2		2
3	AnyEvent::MP - multi-processing/message-passing framework	3	AnyEvent::MP - erlang-style multi-processing/message-passing framework
4		4
5	=head1 SYNOPSIS	5	=head1 SYNOPSIS
6		6
7	use AnyEvent::MP;	7	use AnyEvent::MP;
8		8
		9	$NODE # contains this node's node ID
		10	NODE # returns this node's node ID
		11
		12	$SELF # receiving/own port id in rcv callbacks
		13
		14	# initialise the node so it can send/receive messages
		15	configure;
		16
		17	# ports are message destinations
		18
		19	# sending messages
		20	snd $port, type => data...;
		21	snd $port, @msg;
		22	snd @msg_with_first_element_being_a_port;
		23
		24	# creating/using ports, the simple way
		25	my $simple_port = port { my @msg = @_ };
		26
		27	# creating/using ports, tagged message matching
		28	my $port = port;
		29	rcv $port, ping => sub { snd $_[0], "pong" };
		30	rcv $port, pong => sub { warn "pong received\n" };
		31
		32	# create a port on another node
		33	my $port = spawn $node, $initfunc, @initdata;
		34
		35	# destroy a prot again
		36	kil $port; # "normal" kill
		37	kil $port, my_error => "everything is broken"; # error kill
		38
		39	# monitoring
		40	mon $localport, $cb->(@msg) # callback is invoked on death
		41	mon $localport, $otherport # kill otherport on abnormal death
		42	mon $localport, $otherport, @msg # send message on death
		43
		44	# temporarily execute code in port context
		45	peval $port, sub { die "kill the port!" };
		46
		47	# execute callbacks in $SELF port context
		48	my $timer = AE::timer 1, 0, psub {
		49	die "kill the port, delayed";
		50	};
		51
		52	=head1 CURRENT STATUS
		53
		54	bin/aemp - stable.
		55	AnyEvent::MP - stable API, should work.
		56	AnyEvent::MP::Intro - explains most concepts.
		57	AnyEvent::MP::Kernel - mostly stable API.
		58	AnyEvent::MP::Global - stable API.
		59
9	=head1 DESCRIPTION	60	=head1 DESCRIPTION
10		61
		62	This module (-family) implements a simple message passing framework.
		63
		64	Despite its simplicity, you can securely message other processes running
		65	on the same or other hosts, and you can supervise entities remotely.
		66
		67	For an introduction to this module family, see the L<AnyEvent::MP::Intro>
		68	manual page and the examples under F<eg/>.
		69
		70	=head1 CONCEPTS
		71
		72	=over 4
		73
		74	=item port
		75
		76	Not to be confused with a TCP port, a "port" is something you can send
		77	messages to (with the C<snd> function).
		78
		79	Ports allow you to register C<rcv> handlers that can match all or just
		80	some messages. Messages send to ports will not be queued, regardless of
		81	anything was listening for them or not.
		82
		83	=item port ID - C<nodeid#portname>
		84
		85	A port ID is the concatenation of a node ID, a hash-mark (C<#>) as
		86	separator, and a port name (a printable string of unspecified format).
		87
		88	=item node
		89
		90	A node is a single process containing at least one port - the node port,
		91	which enables nodes to manage each other remotely, and to create new
		92	ports.
		93
		94	Nodes are either public (have one or more listening ports) or private
		95	(no listening ports). Private nodes cannot talk to other private nodes
		96	currently.
		97
		98	=item node ID - C<[A-Z_][a-zA-Z0-9_\-.:]*>
		99
		100	A node ID is a string that uniquely identifies the node within a
		101	network. Depending on the configuration used, node IDs can look like a
		102	hostname, a hostname and a port, or a random string. AnyEvent::MP itself
		103	doesn't interpret node IDs in any way.
		104
		105	=item binds - C<ip:port>
		106
		107	Nodes can only talk to each other by creating some kind of connection to
		108	each other. To do this, nodes should listen on one or more local transport
		109	endpoints - binds. Currently, only standard C<ip:port> specifications can
		110	be used, which specify TCP ports to listen on.
		111
		112	=item seed nodes
		113
		114	When a node starts, it knows nothing about the network. To teach the node
		115	about the network it first has to contact some other node within the
		116	network. This node is called a seed.
		117
		118	Apart from the fact that other nodes know them as seed nodes and they have
		119	to have fixed listening addresses, seed nodes are perfectly normal nodes -
		120	any node can function as a seed node for others.
		121
		122	In addition to discovering the network, seed nodes are also used to
		123	maintain the network and to connect nodes that otherwise would have
		124	trouble connecting. They form the backbone of an AnyEvent::MP network.
		125
		126	Seed nodes are expected to be long-running, and at least one seed node
		127	should always be available. They should also be relatively responsive - a
		128	seed node that blocks for long periods will slow down everybody else.
		129
		130	=item seeds - C<host:port>
		131
		132	Seeds are transport endpoint(s) (usually a hostname/IP address and a
		133	TCP port) of nodes that should be used as seed nodes.
		134
		135	The nodes listening on those endpoints are expected to be long-running,
		136	and at least one of those should always be available. When nodes run out
		137	of connections (e.g. due to a network error), they try to re-establish
		138	connections to some seednodes again to join the network.
		139
		140	=back
		141
		142	=head1 VARIABLES/FUNCTIONS
		143
		144	=over 4
		145
11	=cut	146	=cut
12		147
13	package AnyEvent::MP;	148	package AnyEvent::MP;
14		149
		150	use AnyEvent::MP::Kernel;
		151
15	use common::sense;	152	use common::sense;
16		153
		154	use Carp ();
		155
17	use AE ();	156	use AE ();
18		157
		158	use base "Exporter";
		159
19	our $VERSION = '0.0';	160	our $VERSION = 1.24;
20		161
21	sub nonce($) {	162	our @EXPORT = qw(
22	my $nonce;	163	NODE $NODE *SELF node_of after
		164	configure
		165	snd rcv mon mon_guard kil psub peval spawn cal
		166	port
		167	);
23		168
24	if (open my $fh, "</dev/urandom") {	169	our $SELF;
25	sysread $fh, $nonce, $_[0];	170
26	} else {	171	sub _self_die() {
		172	my $msg = $@;
		173	$msg =~ s/\n+$// unless ref $msg;
		174	kil $SELF, die => $msg;
		175	}
		176
		177	=item $thisnode = NODE / $NODE
		178
		179	The C<NODE> function returns, and the C<$NODE> variable contains, the node
		180	ID of the node running in the current process. This value is initialised by
		181	a call to C<configure>.
		182
		183	=item $nodeid = node_of $port
		184
		185	Extracts and returns the node ID from a port ID or a node ID.
		186
		187	=item configure $profile, key => value...
		188
		189	=item configure key => value...
		190
		191	Before a node can talk to other nodes on the network (i.e. enter
		192	"distributed mode") it has to configure itself - the minimum a node needs
		193	to know is its own name, and optionally it should know the addresses of
		194	some other nodes in the network to discover other nodes.
		195
		196	The key/value pairs are basically the same ones as documented for the
		197	F<aemp> command line utility (sans the set/del prefix).
		198
		199	This function configures a node - it must be called exactly once (or
		200	never) before calling other AnyEvent::MP functions.
		201
		202	=over 4
		203
		204	=item step 1, gathering configuration from profiles
		205
		206	The function first looks up a profile in the aemp configuration (see the
		207	L<aemp> commandline utility). The profile name can be specified via the
		208	named C<profile> parameter or can simply be the first parameter). If it is
		209	missing, then the nodename (F<uname -n>) will be used as profile name.
		210
		211	The profile data is then gathered as follows:
		212
		213	First, all remaining key => value pairs (all of which are conveniently
		214	undocumented at the moment) will be interpreted as configuration
		215	data. Then they will be overwritten by any values specified in the global
		216	default configuration (see the F<aemp> utility), then the chain of
		217	profiles chosen by the profile name (and any C<parent> attributes).
		218
		219	That means that the values specified in the profile have highest priority
		220	and the values specified directly via C<configure> have lowest priority,
		221	and can only be used to specify defaults.
		222
		223	If the profile specifies a node ID, then this will become the node ID of
		224	this process. If not, then the profile name will be used as node ID. The
		225	special node ID of C<anon/> will be replaced by a random node ID.
		226
		227	=item step 2, bind listener sockets
		228
		229	The next step is to look up the binds in the profile, followed by binding
		230	aemp protocol listeners on all binds specified (it is possible and valid
		231	to have no binds, meaning that the node cannot be contacted form the
		232	outside. This means the node cannot talk to other nodes that also have no
		233	binds, but it can still talk to all "normal" nodes).
		234
		235	If the profile does not specify a binds list, then a default of C<*> is
		236	used, meaning the node will bind on a dynamically-assigned port on every
		237	local IP address it finds.
		238
		239	=item step 3, connect to seed nodes
		240
		241	As the last step, the seeds list from the profile is passed to the
		242	L<AnyEvent::MP::Global> module, which will then use it to keep
		243	connectivity with at least one node at any point in time.
		244
		245	=back
		246
		247	Example: become a distributed node using the local node name as profile.
		248	This should be the most common form of invocation for "daemon"-type nodes.
		249
		250	configure
		251
		252	Example: become an anonymous node. This form is often used for commandline
		253	clients.
		254
		255	configure nodeid => "anon/";
		256
		257	Example: configure a node using a profile called seed, which si suitable
		258	for a seed node as it binds on all local addresses on a fixed port (4040,
		259	customary for aemp).
		260
		261	# use the aemp commandline utility
		262	# aemp profile seed nodeid anon/ binds '*:4040'
		263
		264	# then use it
		265	configure profile => "seed";
		266
		267	# or simply use aemp from the shell again:
		268	# aemp run profile seed
		269
		270	# or provide a nicer-to-remember nodeid
		271	# aemp run profile seed nodeid "$(hostname)"
		272
		273	=item $SELF
		274
		275	Contains the current port id while executing C<rcv> callbacks or C<psub>
		276	blocks.
		277
		278	=item *SELF, SELF, %SELF, @SELF...
		279
		280	Due to some quirks in how perl exports variables, it is impossible to
		281	just export C<$SELF>, all the symbols named C<SELF> are exported by this
		282	module, but only C<$SELF> is currently used.
		283
		284	=item snd $port, type => @data
		285
		286	=item snd $port, @msg
		287
		288	Send the given message to the given port, which can identify either a
		289	local or a remote port, and must be a port ID.
		290
		291	While the message can be almost anything, it is highly recommended to
		292	use a string as first element (a port ID, or some word that indicates a
		293	request type etc.) and to consist if only simple perl values (scalars,
		294	arrays, hashes) - if you think you need to pass an object, think again.
		295
		296	The message data logically becomes read-only after a call to this
		297	function: modifying any argument (or values referenced by them) is
		298	forbidden, as there can be considerable time between the call to C<snd>
		299	and the time the message is actually being serialised - in fact, it might
		300	never be copied as within the same process it is simply handed to the
		301	receiving port.
		302
		303	The type of data you can transfer depends on the transport protocol: when
		304	JSON is used, then only strings, numbers and arrays and hashes consisting
		305	of those are allowed (no objects). When Storable is used, then anything
		306	that Storable can serialise and deserialise is allowed, and for the local
		307	node, anything can be passed. Best rely only on the common denominator of
		308	these.
		309
		310	=item $local_port = port
		311
		312	Create a new local port object and returns its port ID. Initially it has
		313	no callbacks set and will throw an error when it receives messages.
		314
		315	=item $local_port = port { my @msg = @_ }
		316
		317	Creates a new local port, and returns its ID. Semantically the same as
		318	creating a port and calling C<rcv $port, $callback> on it.
		319
		320	The block will be called for every message received on the port, with the
		321	global variable C<$SELF> set to the port ID. Runtime errors will cause the
		322	port to be C<kil>ed. The message will be passed as-is, no extra argument
		323	(i.e. no port ID) will be passed to the callback.
		324
		325	If you want to stop/destroy the port, simply C<kil> it:
		326
		327	my $port = port {
		328	my @msg = @_;
27	# shit...	329	...
28	our $nonce_init;	330	kil $SELF;
29	unless ($nonce_init++) {
30	srand time ^ $$ ^ unpack "%L*", qx"ps -edalf" . qx"ipconfig /all";
31	}
32
33	$nonce = join "", map +(chr rand 256), 1 .. $_[0]
34	}	331	};
35		332
36	$nonce	333	=cut
37	}
38		334
39	our $DEFAULT_SECRET;	335	sub rcv($@);
40		336
41	sub default_secret {	337	sub _kilme {
42	unless (defined $DEFAULT_SECRET) {	338	die "received message on port without callback";
43	if (open my $fh, "<$ENV{HOME}/.aemp-secret") {	339	}
44	sysread $fh, $DEFAULT_SECRET, -s $fh;	340
		341	sub port(;&) {
		342	my $id = "$UNIQ." . $ID++;
		343	my $port = "$NODE#$id";
		344
		345	rcv $port, shift \|\| \&_kilme;
		346
		347	$port
		348	}
		349
		350	=item rcv $local_port, $callback->(@msg)
		351
		352	Replaces the default callback on the specified port. There is no way to
		353	remove the default callback: use C<sub { }> to disable it, or better
		354	C<kil> the port when it is no longer needed.
		355
		356	The global C<$SELF> (exported by this module) contains C<$port> while
		357	executing the callback. Runtime errors during callback execution will
		358	result in the port being C<kil>ed.
		359
		360	The default callback received all messages not matched by a more specific
		361	C<tag> match.
		362
		363	=item rcv $local_port, tag => $callback->(@msg_without_tag), ...
		364
		365	Register (or replace) callbacks to be called on messages starting with the
		366	given tag on the given port (and return the port), or unregister it (when
		367	C<$callback> is C<$undef> or missing). There can only be one callback
		368	registered for each tag.
		369
		370	The original message will be passed to the callback, after the first
		371	element (the tag) has been removed. The callback will use the same
		372	environment as the default callback (see above).
		373
		374	Example: create a port and bind receivers on it in one go.
		375
		376	my $port = rcv port,
		377	msg1 => sub { ... },
		378	msg2 => sub { ... },
		379	;
		380
		381	Example: create a port, bind receivers and send it in a message elsewhere
		382	in one go:
		383
		384	snd $otherport, reply =>
		385	rcv port,
		386	msg1 => sub { ... },
		387	...
		388	;
		389
		390	Example: temporarily register a rcv callback for a tag matching some port
		391	(e.g. for an rpc reply) and unregister it after a message was received.
		392
		393	rcv $port, $otherport => sub {
		394	my @reply = @_;
		395
		396	rcv $SELF, $otherport;
		397	};
		398
		399	=cut
		400
		401	sub rcv($@) {
		402	my $port = shift;
		403	my ($nodeid, $portid) = split /#/, $port, 2;
		404
		405	$NODE{$nodeid} == $NODE{""}
		406	or Carp::croak "$port: rcv can only be called on local ports, caught";
		407
		408	while (@_) {
		409	if (ref $_[0]) {
		410	if (my $self = $PORT_DATA{$portid}) {
		411	"AnyEvent::MP::Port" eq ref $self
		412	or Carp::croak "$port: rcv can only be called on message matching ports, caught";
		413
		414	$self->[0] = shift;
45	} else {	415	} else {
46	$DEFAULT_SECRET = nonce 32;	416	my $cb = shift;
		417	$PORT{$portid} = sub {
		418	local $SELF = $port;
		419	eval { &$cb }; _self_die if $@;
		420	};
		421	}
		422	} elsif (defined $_[0]) {
		423	my $self = $PORT_DATA{$portid} \|\|= do {
		424	my $self = bless [$PORT{$portid} \|\| sub { }, { }, $port], "AnyEvent::MP::Port";
		425
		426	$PORT{$portid} = sub {
		427	local $SELF = $port;
		428
		429	if (my $cb = $self->[1]{$_[0]}) {
		430	shift;
		431	eval { &$cb }; _self_die if $@;
		432	} else {
		433	&{ $self->[0] };
		434	}
		435	};
		436
		437	$self
		438	};
		439
		440	"AnyEvent::MP::Port" eq ref $self
		441	or Carp::croak "$port: rcv can only be called on message matching ports, caught";
		442
		443	my ($tag, $cb) = splice @_, 0, 2;
		444
		445	if (defined $cb) {
		446	$self->[1]{$tag} = $cb;
		447	} else {
		448	delete $self->[1]{$tag};
		449	}
47	}	450	}
48	}	451	}
49		452
50	$DEFAULT_SECRET	453	$port
51	}	454	}
		455
		456	=item peval $port, $coderef[, @args]
		457
		458	Evaluates the given C<$codref> within the contetx of C<$port>, that is,
		459	when the code throews an exception the C<$port> will be killed.
		460
		461	Any remaining args will be passed to the callback. Any return values will
		462	be returned to the caller.
		463
		464	This is useful when you temporarily want to execute code in the context of
		465	a port.
		466
		467	Example: create a port and run some initialisation code in it's context.
		468
		469	my $port = port { ... };
		470
		471	peval $port, sub {
		472	init
		473	or die "unable to init";
		474	};
		475
		476	=cut
		477
		478	sub peval($$) {
		479	local $SELF = shift;
		480	my $cb = shift;
		481
		482	if (wantarray) {
		483	my @res = eval { &$cb };
		484	_self_die if $@;
		485	@res
		486	} else {
		487	my $res = eval { &$cb };
		488	_self_die if $@;
		489	$res
		490	}
		491	}
		492
		493	=item $closure = psub { BLOCK }
		494
		495	Remembers C<$SELF> and creates a closure out of the BLOCK. When the
		496	closure is executed, sets up the environment in the same way as in C<rcv>
		497	callbacks, i.e. runtime errors will cause the port to get C<kil>ed.
		498
		499	The effect is basically as if it returned C<< sub { peval $SELF, sub {
		500	BLOCK } } >>.
		501
		502	This is useful when you register callbacks from C<rcv> callbacks:
		503
		504	rcv delayed_reply => sub {
		505	my ($delay, @reply) = @_;
		506	my $timer = AE::timer $delay, 0, psub {
		507	snd @reply, $SELF;
		508	};
		509	};
		510
		511	=cut
		512
		513	sub psub(&) {
		514	my $cb = shift;
		515
		516	my $port = $SELF
		517	or Carp::croak "psub can only be called from within rcv or psub callbacks, not";
		518
		519	sub {
		520	local $SELF = $port;
		521
		522	if (wantarray) {
		523	my @res = eval { &$cb };
		524	_self_die if $@;
		525	@res
		526	} else {
		527	my $res = eval { &$cb };
		528	_self_die if $@;
		529	$res
		530	}
		531	}
		532	}
		533
		534	=item $guard = mon $port, $cb->(@reason) # call $cb when $port dies
		535
		536	=item $guard = mon $port, $rcvport # kill $rcvport when $port dies
		537
		538	=item $guard = mon $port # kill $SELF when $port dies
		539
		540	=item $guard = mon $port, $rcvport, @msg # send a message when $port dies
		541
		542	Monitor the given port and do something when the port is killed or
		543	messages to it were lost, and optionally return a guard that can be used
		544	to stop monitoring again.
		545
		546	In the first form (callback), the callback is simply called with any
		547	number of C<@reason> elements (no @reason means that the port was deleted
		548	"normally"). Note also that I<< the callback B<must> never die >>, so use
		549	C<eval> if unsure.
		550
		551	In the second form (another port given), the other port (C<$rcvport>)
		552	will be C<kil>'ed with C<@reason>, if a @reason was specified, i.e. on
		553	"normal" kils nothing happens, while under all other conditions, the other
		554	port is killed with the same reason.
		555
		556	The third form (kill self) is the same as the second form, except that
		557	C<$rvport> defaults to C<$SELF>.
		558
		559	In the last form (message), a message of the form C<@msg, @reason> will be
		560	C<snd>.
		561
		562	Monitoring-actions are one-shot: once messages are lost (and a monitoring
		563	alert was raised), they are removed and will not trigger again.
		564
		565	As a rule of thumb, monitoring requests should always monitor a port from
		566	a local port (or callback). The reason is that kill messages might get
		567	lost, just like any other message. Another less obvious reason is that
		568	even monitoring requests can get lost (for example, when the connection
		569	to the other node goes down permanently). When monitoring a port locally
		570	these problems do not exist.
		571
		572	C<mon> effectively guarantees that, in the absence of hardware failures,
		573	after starting the monitor, either all messages sent to the port will
		574	arrive, or the monitoring action will be invoked after possible message
		575	loss has been detected. No messages will be lost "in between" (after
		576	the first lost message no further messages will be received by the
		577	port). After the monitoring action was invoked, further messages might get
		578	delivered again.
		579
		580	Inter-host-connection timeouts and monitoring depend on the transport
		581	used. The only transport currently implemented is TCP, and AnyEvent::MP
		582	relies on TCP to detect node-downs (this can take 10-15 minutes on a
		583	non-idle connection, and usually around two hours for idle connections).
		584
		585	This means that monitoring is good for program errors and cleaning up
		586	stuff eventually, but they are no replacement for a timeout when you need
		587	to ensure some maximum latency.
		588
		589	Example: call a given callback when C<$port> is killed.
		590
		591	mon $port, sub { warn "port died because of <@_>\n" };
		592
		593	Example: kill ourselves when C<$port> is killed abnormally.
		594
		595	mon $port;
		596
		597	Example: send us a restart message when another C<$port> is killed.
		598
		599	mon $port, $self => "restart";
		600
		601	=cut
		602
		603	sub mon {
		604	my ($nodeid, $port) = split /#/, shift, 2;
		605
		606	my $node = $NODE{$nodeid} \|\| add_node $nodeid;
		607
		608	my $cb = @_ ? shift : $SELF \|\| Carp::croak 'mon: called with one argument only, but $SELF not set,';
		609
		610	unless (ref $cb) {
		611	if (@_) {
		612	# send a kill info message
		613	my (@msg) = ($cb, @_);
		614	$cb = sub { snd @msg, @_ };
		615	} else {
		616	# simply kill other port
		617	my $port = $cb;
		618	$cb = sub { kil $port, @_ if @_ };
		619	}
		620	}
		621
		622	$node->monitor ($port, $cb);
		623
		624	defined wantarray
		625	and ($cb += 0, AnyEvent::Util::guard { $node->unmonitor ($port, $cb) })
		626	}
		627
		628	=item $guard = mon_guard $port, $ref, $ref...
		629
		630	Monitors the given C<$port> and keeps the passed references. When the port
		631	is killed, the references will be freed.
		632
		633	Optionally returns a guard that will stop the monitoring.
		634
		635	This function is useful when you create e.g. timers or other watchers and
		636	want to free them when the port gets killed (note the use of C<psub>):
		637
		638	$port->rcv (start => sub {
		639	my $timer; $timer = mon_guard $port, AE::timer 1, 1, psub {
		640	undef $timer if 0.9 < rand;
		641	});
		642	});
		643
		644	=cut
		645
		646	sub mon_guard {
		647	my ($port, @refs) = @_;
		648
		649	#TODO: mon-less form?
		650
		651	mon $port, sub { 0 && @refs }
		652	}
		653
		654	=item kil $port[, @reason]
		655
		656	Kill the specified port with the given C<@reason>.
		657
		658	If no C<@reason> is specified, then the port is killed "normally" -
		659	monitor callback will be invoked, but the kil will not cause linked ports
		660	(C<mon $mport, $lport> form) to get killed.
		661
		662	If a C<@reason> is specified, then linked ports (C<mon $mport, $lport>
		663	form) get killed with the same reason.
		664
		665	Runtime errors while evaluating C<rcv> callbacks or inside C<psub> blocks
		666	will be reported as reason C<< die => $@ >>.
		667
		668	Transport/communication errors are reported as C<< transport_error =>
		669	$message >>.
		670
		671	=cut
		672
		673	=item $port = spawn $node, $initfunc[, @initdata]
		674
		675	Creates a port on the node C<$node> (which can also be a port ID, in which
		676	case it's the node where that port resides).
		677
		678	The port ID of the newly created port is returned immediately, and it is
		679	possible to immediately start sending messages or to monitor the port.
		680
		681	After the port has been created, the init function is called on the remote
		682	node, in the same context as a C<rcv> callback. This function must be a
		683	fully-qualified function name (e.g. C<MyApp::Chat::Server::init>). To
		684	specify a function in the main program, use C<::name>.
		685
		686	If the function doesn't exist, then the node tries to C<require>
		687	the package, then the package above the package and so on (e.g.
		688	C<MyApp::Chat::Server>, C<MyApp::Chat>, C<MyApp>) until the function
		689	exists or it runs out of package names.
		690
		691	The init function is then called with the newly-created port as context
		692	object (C<$SELF>) and the C<@initdata> values as arguments. It I<must>
		693	call one of the C<rcv> functions to set callbacks on C<$SELF>, otherwise
		694	the port might not get created.
		695
		696	A common idiom is to pass a local port, immediately monitor the spawned
		697	port, and in the remote init function, immediately monitor the passed
		698	local port. This two-way monitoring ensures that both ports get cleaned up
		699	when there is a problem.
		700
		701	C<spawn> guarantees that the C<$initfunc> has no visible effects on the
		702	caller before C<spawn> returns (by delaying invocation when spawn is
		703	called for the local node).
		704
		705	Example: spawn a chat server port on C<$othernode>.
		706
		707	# this node, executed from within a port context:
		708	my $server = spawn $othernode, "MyApp::Chat::Server::connect", $SELF;
		709	mon $server;
		710
		711	# init function on C<$othernode>
		712	sub connect {
		713	my ($srcport) = @_;
		714
		715	mon $srcport;
		716
		717	rcv $SELF, sub {
		718	...
		719	};
		720	}
		721
		722	=cut
		723
		724	sub _spawn {
		725	my $port = shift;
		726	my $init = shift;
		727
		728	# rcv will create the actual port
		729	local $SELF = "$NODE#$port";
		730	eval {
		731	&{ load_func $init }
		732	};
		733	_self_die if $@;
		734	}
		735
		736	sub spawn(@) {
		737	my ($nodeid, undef) = split /#/, shift, 2;
		738
		739	my $id = "$RUNIQ." . $ID++;
		740
		741	$_[0] =~ /::/
		742	or Carp::croak "spawn init function must be a fully-qualified name, caught";
		743
		744	snd_to_func $nodeid, "AnyEvent::MP::_spawn" => $id, @_;
		745
		746	"$nodeid#$id"
		747	}
		748
		749	=item after $timeout, @msg
		750
		751	=item after $timeout, $callback
		752
		753	Either sends the given message, or call the given callback, after the
		754	specified number of seconds.
		755
		756	This is simply a utility function that comes in handy at times - the
		757	AnyEvent::MP author is not convinced of the wisdom of having it, though,
		758	so it may go away in the future.
		759
		760	=cut
		761
		762	sub after($@) {
		763	my ($timeout, @action) = @_;
		764
		765	my $t; $t = AE::timer $timeout, 0, sub {
		766	undef $t;
		767	ref $action[0]
		768	? $action[0]()
		769	: snd @action;
		770	};
		771	}
		772
		773	=item cal $port, @msg, $callback[, $timeout]
		774
		775	A simple form of RPC - sends a message to the given C<$port> with the
		776	given contents (C<@msg>), but adds a reply port to the message.
		777
		778	The reply port is created temporarily just for the purpose of receiving
		779	the reply, and will be C<kil>ed when no longer needed.
		780
		781	A reply message sent to the port is passed to the C<$callback> as-is.
		782
		783	If an optional time-out (in seconds) is given and it is not C<undef>,
		784	then the callback will be called without any arguments after the time-out
		785	elapsed and the port is C<kil>ed.
		786
		787	If no time-out is given (or it is C<undef>), then the local port will
		788	monitor the remote port instead, so it eventually gets cleaned-up.
		789
		790	Currently this function returns the temporary port, but this "feature"
		791	might go in future versions unless you can make a convincing case that
		792	this is indeed useful for something.
		793
		794	=cut
		795
		796	sub cal(@) {
		797	my $timeout = ref $_[-1] ? undef : pop;
		798	my $cb = pop;
		799
		800	my $port = port {
		801	undef $timeout;
		802	kil $SELF;
		803	&$cb;
		804	};
		805
		806	if (defined $timeout) {
		807	$timeout = AE::timer $timeout, 0, sub {
		808	undef $timeout;
		809	kil $port;
		810	$cb->();
		811	};
		812	} else {
		813	mon $_[0], sub {
		814	kil $port;
		815	$cb->();
		816	};
		817	}
		818
		819	push @_, $port;
		820	&snd;
		821
		822	$port
		823	}
		824
		825	=back
		826
		827	=head1 AnyEvent::MP vs. Distributed Erlang
		828
		829	AnyEvent::MP got lots of its ideas from distributed Erlang (Erlang node
		830	== aemp node, Erlang process == aemp port), so many of the documents and
		831	programming techniques employed by Erlang apply to AnyEvent::MP. Here is a
		832	sample:
		833
		834	http://www.erlang.se/doc/programming_rules.shtml
		835	http://erlang.org/doc/getting_started/part_frame.html # chapters 3 and 4
		836	http://erlang.org/download/erlang-book-part1.pdf # chapters 5 and 6
		837	http://erlang.org/download/armstrong_thesis_2003.pdf # chapters 4 and 5
		838
		839	Despite the similarities, there are also some important differences:
		840
		841	=over 4
		842
		843	=item * Node IDs are arbitrary strings in AEMP.
		844
		845	Erlang relies on special naming and DNS to work everywhere in the same
		846	way. AEMP relies on each node somehow knowing its own address(es) (e.g. by
		847	configuration or DNS), and possibly the addresses of some seed nodes, but
		848	will otherwise discover other nodes (and their IDs) itself.
		849
		850	=item * Erlang has a "remote ports are like local ports" philosophy, AEMP
		851	uses "local ports are like remote ports".
		852
		853	The failure modes for local ports are quite different (runtime errors
		854	only) then for remote ports - when a local port dies, you I<know> it dies,
		855	when a connection to another node dies, you know nothing about the other
		856	port.
		857
		858	Erlang pretends remote ports are as reliable as local ports, even when
		859	they are not.
		860
		861	AEMP encourages a "treat remote ports differently" philosophy, with local
		862	ports being the special case/exception, where transport errors cannot
		863	occur.
		864
		865	=item * Erlang uses processes and a mailbox, AEMP does not queue.
		866
		867	Erlang uses processes that selectively receive messages, and therefore
		868	needs a queue. AEMP is event based, queuing messages would serve no
		869	useful purpose. For the same reason the pattern-matching abilities of
		870	AnyEvent::MP are more limited, as there is little need to be able to
		871	filter messages without dequeuing them.
		872
		873	(But see L<Coro::MP> for a more Erlang-like process model on top of AEMP).
		874
		875	=item * Erlang sends are synchronous, AEMP sends are asynchronous.
		876
		877	Sending messages in Erlang is synchronous and blocks the process (and
		878	so does not need a queue that can overflow). AEMP sends are immediate,
		879	connection establishment is handled in the background.
		880
		881	=item * Erlang suffers from silent message loss, AEMP does not.
		882
		883	Erlang implements few guarantees on messages delivery - messages can get
		884	lost without any of the processes realising it (i.e. you send messages a,
		885	b, and c, and the other side only receives messages a and c).
		886
		887	AEMP guarantees correct ordering, and the guarantee that after one message
		888	is lost, all following ones sent to the same port are lost as well, until
		889	monitoring raises an error, so there are no silent "holes" in the message
		890	sequence.
		891
		892	=item * Erlang can send messages to the wrong port, AEMP does not.
		893
		894	In Erlang it is quite likely that a node that restarts reuses a process ID
		895	known to other nodes for a completely different process, causing messages
		896	destined for that process to end up in an unrelated process.
		897
		898	AEMP never reuses port IDs, so old messages or old port IDs floating
		899	around in the network will not be sent to an unrelated port.
		900
		901	=item * Erlang uses unprotected connections, AEMP uses secure
		902	authentication and can use TLS.
		903
		904	AEMP can use a proven protocol - TLS - to protect connections and
		905	securely authenticate nodes.
		906
		907	=item * The AEMP protocol is optimised for both text-based and binary
		908	communications.
		909
		910	The AEMP protocol, unlike the Erlang protocol, supports both programming
		911	language independent text-only protocols (good for debugging) and binary,
		912	language-specific serialisers (e.g. Storable). By default, unless TLS is
		913	used, the protocol is actually completely text-based.
		914
		915	It has also been carefully designed to be implementable in other languages
		916	with a minimum of work while gracefully degrading functionality to make the
		917	protocol simple.
		918
		919	=item * AEMP has more flexible monitoring options than Erlang.
		920
		921	In Erlang, you can chose to receive I<all> exit signals as messages
		922	or I<none>, there is no in-between, so monitoring single processes is
		923	difficult to implement. Monitoring in AEMP is more flexible than in
		924	Erlang, as one can choose between automatic kill, exit message or callback
		925	on a per-process basis.
		926
		927	=item * Erlang tries to hide remote/local connections, AEMP does not.
		928
		929	Monitoring in Erlang is not an indicator of process death/crashes, in the
		930	same way as linking is (except linking is unreliable in Erlang).
		931
		932	In AEMP, you don't "look up" registered port names or send to named ports
		933	that might or might not be persistent. Instead, you normally spawn a port
		934	on the remote node. The init function monitors you, and you monitor the
		935	remote port. Since both monitors are local to the node, they are much more
		936	reliable (no need for C<spawn_link>).
		937
		938	This also saves round-trips and avoids sending messages to the wrong port
		939	(hard to do in Erlang).
		940
		941	=back
		942
		943	=head1 RATIONALE
		944
		945	=over 4
		946
		947	=item Why strings for port and node IDs, why not objects?
		948
		949	We considered "objects", but found that the actual number of methods
		950	that can be called are quite low. Since port and node IDs travel over
		951	the network frequently, the serialising/deserialising would add lots of
		952	overhead, as well as having to keep a proxy object everywhere.
		953
		954	Strings can easily be printed, easily serialised etc. and need no special
		955	procedures to be "valid".
		956
		957	And as a result, a miniport consists of a single closure stored in a
		958	global hash - it can't become much cheaper.
		959
		960	=item Why favour JSON, why not a real serialising format such as Storable?
		961
		962	In fact, any AnyEvent::MP node will happily accept Storable as framing
		963	format, but currently there is no way to make a node use Storable by
		964	default (although all nodes will accept it).
		965
		966	The default framing protocol is JSON because a) JSON::XS is many times
		967	faster for small messages and b) most importantly, after years of
		968	experience we found that object serialisation is causing more problems
		969	than it solves: Just like function calls, objects simply do not travel
		970	easily over the network, mostly because they will always be a copy, so you
		971	always have to re-think your design.
		972
		973	Keeping your messages simple, concentrating on data structures rather than
		974	objects, will keep your messages clean, tidy and efficient.
		975
		976	=back
52		977
53	=head1 SEE ALSO	978	=head1 SEE ALSO
		979
		980	L<AnyEvent::MP::Intro> - a gentle introduction.
		981
		982	L<AnyEvent::MP::Kernel> - more, lower-level, stuff.
		983
		984	L<AnyEvent::MP::Global> - network maintainance and port groups, to find
		985	your applications.
		986
		987	L<AnyEvent::MP::DataConn> - establish data connections between nodes.
		988
		989	L<AnyEvent::MP::LogCatcher> - simple service to display log messages from
		990	all nodes.
54		991
55	L<AnyEvent>.	992	L<AnyEvent>.
56		993
57	=head1 AUTHOR	994	=head1 AUTHOR
58		995

Diff Legend

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

Comparing AnyEvent-MP/MP.pm (file contents): Revision 1.1 by root, Thu Jul 30 08:38:50 2009 UTC vs. Revision 1.108 by root, Wed Dec 30 13:37:53 2009 UTC

Diff Legend

Comparing AnyEvent-MP/MP.pm (file contents):
Revision 1.1 by root, Thu Jul 30 08:38:50 2009 UTC vs.
Revision 1.108 by root, Wed Dec 30 13:37:53 2009 UTC