ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/AnyEvent-MP/MP.pm
Revision: 1.133
Committed: Mon Mar 12 10:34:06 2012 UTC (12 years, 2 months ago) by root
Branch: MAIN
Changes since 1.132: +14 -3 lines
Log Message:
*** empty log message ***

File Contents

# Content
1 =head1 NAME
2
3 AnyEvent::MP - erlang-style multi-processing/message-passing framework
4
5 =head1 SYNOPSIS
6
7 use AnyEvent::MP;
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 port again
36 kil $port; # "normal" kill
37 kil $port, my_error => "everything is broken"; # error kill
38
39 # monitoring
40 mon $port, $cb->(@msg) # callback is invoked on death
41 mon $port, $localport # kill localport on abnormal death
42 mon $port, $localport, @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
60 =head1 DESCRIPTION
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 Ports are represented by (printable) strings called "port IDs".
84
85 =item port ID - C<nodeid#portname>
86
87 A port ID is the concatenation of a node ID, a hash-mark (C<#>)
88 as separator, and a port name (a printable string of unspecified
89 format created by AnyEvent::MP).
90
91 =item node
92
93 A node is a single process containing at least one port - the node port,
94 which enables nodes to manage each other remotely, and to create new
95 ports.
96
97 Nodes are either public (have one or more listening ports) or private
98 (no listening ports). Private nodes cannot talk to other private nodes
99 currently, but all nodes can talk to public nodes.
100
101 Nodes is represented by (printable) strings called "node IDs".
102
103 =item node ID - C<[A-Za-z0-9_\-.:]*>
104
105 A node ID is a string that uniquely identifies the node within a
106 network. Depending on the configuration used, node IDs can look like a
107 hostname, a hostname and a port, or a random string. AnyEvent::MP itself
108 doesn't interpret node IDs in any way except to uniquely identify a node.
109
110 =item binds - C<ip:port>
111
112 Nodes can only talk to each other by creating some kind of connection to
113 each other. To do this, nodes should listen on one or more local transport
114 endpoints - binds.
115
116 Currently, only standard C<ip:port> specifications can be used, which
117 specify TCP ports to listen on. So a bind is basically just a tcp socket
118 in listening mode thta accepts conenctions form other nodes.
119
120 =item seed nodes
121
122 When a node starts, it knows nothing about the network it is in - it
123 needs to connect to at least one other node that is already in the
124 network. These other nodes are called "seed nodes".
125
126 Seed nodes themselves are not special - they are seed nodes only because
127 some other node I<uses> them as such, but any node can be used as seed
128 node for other nodes, and eahc node cna use a different set of seed nodes.
129
130 In addition to discovering the network, seed nodes are also used to
131 maintain the network - all nodes using the same seed node form are part of
132 the same network. If a network is split into multiple subnets because e.g.
133 the network link between the parts goes down, then using the same seed
134 nodes for all nodes ensures that eventually the subnets get merged again.
135
136 Seed nodes are expected to be long-running, and at least one seed node
137 should always be available. They should also be relatively responsive - a
138 seed node that blocks for long periods will slow down everybody else.
139
140 For small networks, it's best if every node uses the same set of seed
141 nodes. For large networks, it can be useful to specify "regional" seed
142 nodes for most nodes in an area, and use all seed nodes as seed nodes for
143 each other. What's important is that all seed nodes connections form a
144 complete graph, so that the network cannot split into separate subnets
145 forever.
146
147 Seed nodes are represented by seed IDs.
148
149 =item seed IDs - C<host:port>
150
151 Seed IDs are transport endpoint(s) (usually a hostname/IP address and a
152 TCP port) of nodes that should be used as seed nodes.
153
154 =item global nodes
155
156 An AEMP network needs a discovery service - nodes need to know how to
157 connect to other nodes they only know by name. In addition, AEMP offers a
158 distributed "group database", which maps group names to a list of strings
159 - for example, to register worker ports.
160
161 A network needs at least one global node to work, and allows every node to
162 be a global node.
163
164 Any node that loads the L<AnyEvent::MP::Global> module becomes a global
165 node and tries to keep connections to all other nodes. So while it can
166 make sense to make every node "global" in small networks, it usually makes
167 sense to only make seed nodes into global nodes in large networks (nodes
168 keep connections to seed nodes and global nodes, so makign them the same
169 reduces overhead).
170
171 =back
172
173 =head1 VARIABLES/FUNCTIONS
174
175 =over 4
176
177 =cut
178
179 package AnyEvent::MP;
180
181 use AnyEvent::MP::Config ();
182 use AnyEvent::MP::Kernel;
183 use AnyEvent::MP::Kernel qw(%NODE %PORT %PORT_DATA $UNIQ $RUNIQ $ID);
184
185 use common::sense;
186
187 use Carp ();
188
189 use AE ();
190 use Guard ();
191
192 use base "Exporter";
193
194 our $VERSION = $AnyEvent::MP::Config::VERSION;
195
196 our @EXPORT = qw(
197 NODE $NODE *SELF node_of after
198 configure
199 snd rcv mon mon_guard kil psub peval spawn cal
200 port
201 db_set db_del db_reg
202 db_mon db_family db_keys db_values
203 );
204
205 our $SELF;
206
207 sub _self_die() {
208 my $msg = $@;
209 $msg =~ s/\n+$// unless ref $msg;
210 kil $SELF, die => $msg;
211 }
212
213 =item $thisnode = NODE / $NODE
214
215 The C<NODE> function returns, and the C<$NODE> variable contains, the node
216 ID of the node running in the current process. This value is initialised by
217 a call to C<configure>.
218
219 =item $nodeid = node_of $port
220
221 Extracts and returns the node ID from a port ID or a node ID.
222
223 =item configure $profile, key => value...
224
225 =item configure key => value...
226
227 Before a node can talk to other nodes on the network (i.e. enter
228 "distributed mode") it has to configure itself - the minimum a node needs
229 to know is its own name, and optionally it should know the addresses of
230 some other nodes in the network to discover other nodes.
231
232 This function configures a node - it must be called exactly once (or
233 never) before calling other AnyEvent::MP functions.
234
235 The key/value pairs are basically the same ones as documented for the
236 F<aemp> command line utility (sans the set/del prefix), with these additions:
237
238 =over 4
239
240 =item norc => $boolean (default false)
241
242 If true, then the rc file (e.g. F<~/.perl-anyevent-mp>) will I<not>
243 be consulted - all configuraiton options must be specified in the
244 C<configure> call.
245
246 =item force => $boolean (default false)
247
248 IF true, then the values specified in the C<configure> will take
249 precedence over any values configured via the rc file. The default is for
250 the rc file to override any options specified in the program.
251
252 =item secure => $pass->($nodeid)
253
254 In addition to specifying a boolean, you can specify a code reference that
255 is called for every remote execution attempt - the execution request is
256 granted iff the callback returns a true value.
257
258 See F<semp setsecure> for more info.
259
260 =back
261
262 =over 4
263
264 =item step 1, gathering configuration from profiles
265
266 The function first looks up a profile in the aemp configuration (see the
267 L<aemp> commandline utility). The profile name can be specified via the
268 named C<profile> parameter or can simply be the first parameter). If it is
269 missing, then the nodename (F<uname -n>) will be used as profile name.
270
271 The profile data is then gathered as follows:
272
273 First, all remaining key => value pairs (all of which are conveniently
274 undocumented at the moment) will be interpreted as configuration
275 data. Then they will be overwritten by any values specified in the global
276 default configuration (see the F<aemp> utility), then the chain of
277 profiles chosen by the profile name (and any C<parent> attributes).
278
279 That means that the values specified in the profile have highest priority
280 and the values specified directly via C<configure> have lowest priority,
281 and can only be used to specify defaults.
282
283 If the profile specifies a node ID, then this will become the node ID of
284 this process. If not, then the profile name will be used as node ID, with
285 a unique randoms tring (C</%u>) appended.
286
287 The node ID can contain some C<%> sequences that are expanded: C<%n>
288 is expanded to the local nodename, C<%u> is replaced by a random
289 strign to make the node unique. For example, the F<aemp> commandline
290 utility uses C<aemp/%n/%u> as nodename, which might expand to
291 C<aemp/cerebro/ZQDGSIkRhEZQDGSIkRhE>.
292
293 =item step 2, bind listener sockets
294
295 The next step is to look up the binds in the profile, followed by binding
296 aemp protocol listeners on all binds specified (it is possible and valid
297 to have no binds, meaning that the node cannot be contacted form the
298 outside. This means the node cannot talk to other nodes that also have no
299 binds, but it can still talk to all "normal" nodes).
300
301 If the profile does not specify a binds list, then a default of C<*> is
302 used, meaning the node will bind on a dynamically-assigned port on every
303 local IP address it finds.
304
305 =item step 3, connect to seed nodes
306
307 As the last step, the seed ID list from the profile is passed to the
308 L<AnyEvent::MP::Global> module, which will then use it to keep
309 connectivity with at least one node at any point in time.
310
311 =back
312
313 Example: become a distributed node using the local node name as profile.
314 This should be the most common form of invocation for "daemon"-type nodes.
315
316 configure
317
318 Example: become a semi-anonymous node. This form is often used for
319 commandline clients.
320
321 configure nodeid => "myscript/%n/%u";
322
323 Example: configure a node using a profile called seed, which is suitable
324 for a seed node as it binds on all local addresses on a fixed port (4040,
325 customary for aemp).
326
327 # use the aemp commandline utility
328 # aemp profile seed binds '*:4040'
329
330 # then use it
331 configure profile => "seed";
332
333 # or simply use aemp from the shell again:
334 # aemp run profile seed
335
336 # or provide a nicer-to-remember nodeid
337 # aemp run profile seed nodeid "$(hostname)"
338
339 =item $SELF
340
341 Contains the current port id while executing C<rcv> callbacks or C<psub>
342 blocks.
343
344 =item *SELF, SELF, %SELF, @SELF...
345
346 Due to some quirks in how perl exports variables, it is impossible to
347 just export C<$SELF>, all the symbols named C<SELF> are exported by this
348 module, but only C<$SELF> is currently used.
349
350 =item snd $port, type => @data
351
352 =item snd $port, @msg
353
354 Send the given message to the given port, which can identify either a
355 local or a remote port, and must be a port ID.
356
357 While the message can be almost anything, it is highly recommended to
358 use a string as first element (a port ID, or some word that indicates a
359 request type etc.) and to consist if only simple perl values (scalars,
360 arrays, hashes) - if you think you need to pass an object, think again.
361
362 The message data logically becomes read-only after a call to this
363 function: modifying any argument (or values referenced by them) is
364 forbidden, as there can be considerable time between the call to C<snd>
365 and the time the message is actually being serialised - in fact, it might
366 never be copied as within the same process it is simply handed to the
367 receiving port.
368
369 The type of data you can transfer depends on the transport protocol: when
370 JSON is used, then only strings, numbers and arrays and hashes consisting
371 of those are allowed (no objects). When Storable is used, then anything
372 that Storable can serialise and deserialise is allowed, and for the local
373 node, anything can be passed. Best rely only on the common denominator of
374 these.
375
376 =item $local_port = port
377
378 Create a new local port object and returns its port ID. Initially it has
379 no callbacks set and will throw an error when it receives messages.
380
381 =item $local_port = port { my @msg = @_ }
382
383 Creates a new local port, and returns its ID. Semantically the same as
384 creating a port and calling C<rcv $port, $callback> on it.
385
386 The block will be called for every message received on the port, with the
387 global variable C<$SELF> set to the port ID. Runtime errors will cause the
388 port to be C<kil>ed. The message will be passed as-is, no extra argument
389 (i.e. no port ID) will be passed to the callback.
390
391 If you want to stop/destroy the port, simply C<kil> it:
392
393 my $port = port {
394 my @msg = @_;
395 ...
396 kil $SELF;
397 };
398
399 =cut
400
401 sub rcv($@);
402
403 my $KILME = sub {
404 (my $tag = substr $_[0], 0, 30) =~ s/([\x20-\x7e])/./g;
405 kil $SELF, unhandled_message => "no callback set for message (first element $tag)";
406 };
407
408 sub port(;&) {
409 my $id = $UNIQ . ++$ID;
410 my $port = "$NODE#$id";
411
412 rcv $port, shift || $KILME;
413
414 $port
415 }
416
417 =item rcv $local_port, $callback->(@msg)
418
419 Replaces the default callback on the specified port. There is no way to
420 remove the default callback: use C<sub { }> to disable it, or better
421 C<kil> the port when it is no longer needed.
422
423 The global C<$SELF> (exported by this module) contains C<$port> while
424 executing the callback. Runtime errors during callback execution will
425 result in the port being C<kil>ed.
426
427 The default callback receives all messages not matched by a more specific
428 C<tag> match.
429
430 =item rcv $local_port, tag => $callback->(@msg_without_tag), ...
431
432 Register (or replace) callbacks to be called on messages starting with the
433 given tag on the given port (and return the port), or unregister it (when
434 C<$callback> is C<$undef> or missing). There can only be one callback
435 registered for each tag.
436
437 The original message will be passed to the callback, after the first
438 element (the tag) has been removed. The callback will use the same
439 environment as the default callback (see above).
440
441 Example: create a port and bind receivers on it in one go.
442
443 my $port = rcv port,
444 msg1 => sub { ... },
445 msg2 => sub { ... },
446 ;
447
448 Example: create a port, bind receivers and send it in a message elsewhere
449 in one go:
450
451 snd $otherport, reply =>
452 rcv port,
453 msg1 => sub { ... },
454 ...
455 ;
456
457 Example: temporarily register a rcv callback for a tag matching some port
458 (e.g. for an rpc reply) and unregister it after a message was received.
459
460 rcv $port, $otherport => sub {
461 my @reply = @_;
462
463 rcv $SELF, $otherport;
464 };
465
466 =cut
467
468 sub rcv($@) {
469 my $port = shift;
470 my ($nodeid, $portid) = split /#/, $port, 2;
471
472 $NODE{$nodeid} == $NODE{""}
473 or Carp::croak "$port: rcv can only be called on local ports, caught";
474
475 while (@_) {
476 if (ref $_[0]) {
477 if (my $self = $PORT_DATA{$portid}) {
478 "AnyEvent::MP::Port" eq ref $self
479 or Carp::croak "$port: rcv can only be called on message matching ports, caught";
480
481 $self->[0] = shift;
482 } else {
483 my $cb = shift;
484 $PORT{$portid} = sub {
485 local $SELF = $port;
486 eval { &$cb }; _self_die if $@;
487 };
488 }
489 } elsif (defined $_[0]) {
490 my $self = $PORT_DATA{$portid} ||= do {
491 my $self = bless [$PORT{$portid} || sub { }, { }, $port], "AnyEvent::MP::Port";
492
493 $PORT{$portid} = sub {
494 local $SELF = $port;
495
496 if (my $cb = $self->[1]{$_[0]}) {
497 shift;
498 eval { &$cb }; _self_die if $@;
499 } else {
500 &{ $self->[0] };
501 }
502 };
503
504 $self
505 };
506
507 "AnyEvent::MP::Port" eq ref $self
508 or Carp::croak "$port: rcv can only be called on message matching ports, caught";
509
510 my ($tag, $cb) = splice @_, 0, 2;
511
512 if (defined $cb) {
513 $self->[1]{$tag} = $cb;
514 } else {
515 delete $self->[1]{$tag};
516 }
517 }
518 }
519
520 $port
521 }
522
523 =item peval $port, $coderef[, @args]
524
525 Evaluates the given C<$codref> within the contetx of C<$port>, that is,
526 when the code throews an exception the C<$port> will be killed.
527
528 Any remaining args will be passed to the callback. Any return values will
529 be returned to the caller.
530
531 This is useful when you temporarily want to execute code in the context of
532 a port.
533
534 Example: create a port and run some initialisation code in it's context.
535
536 my $port = port { ... };
537
538 peval $port, sub {
539 init
540 or die "unable to init";
541 };
542
543 =cut
544
545 sub peval($$) {
546 local $SELF = shift;
547 my $cb = shift;
548
549 if (wantarray) {
550 my @res = eval { &$cb };
551 _self_die if $@;
552 @res
553 } else {
554 my $res = eval { &$cb };
555 _self_die if $@;
556 $res
557 }
558 }
559
560 =item $closure = psub { BLOCK }
561
562 Remembers C<$SELF> and creates a closure out of the BLOCK. When the
563 closure is executed, sets up the environment in the same way as in C<rcv>
564 callbacks, i.e. runtime errors will cause the port to get C<kil>ed.
565
566 The effect is basically as if it returned C<< sub { peval $SELF, sub {
567 BLOCK }, @_ } >>.
568
569 This is useful when you register callbacks from C<rcv> callbacks:
570
571 rcv delayed_reply => sub {
572 my ($delay, @reply) = @_;
573 my $timer = AE::timer $delay, 0, psub {
574 snd @reply, $SELF;
575 };
576 };
577
578 =cut
579
580 sub psub(&) {
581 my $cb = shift;
582
583 my $port = $SELF
584 or Carp::croak "psub can only be called from within rcv or psub callbacks, not";
585
586 sub {
587 local $SELF = $port;
588
589 if (wantarray) {
590 my @res = eval { &$cb };
591 _self_die if $@;
592 @res
593 } else {
594 my $res = eval { &$cb };
595 _self_die if $@;
596 $res
597 }
598 }
599 }
600
601 =item $guard = mon $port, $cb->(@reason) # call $cb when $port dies
602
603 =item $guard = mon $port, $rcvport # kill $rcvport when $port dies
604
605 =item $guard = mon $port # kill $SELF when $port dies
606
607 =item $guard = mon $port, $rcvport, @msg # send a message when $port dies
608
609 Monitor the given port and do something when the port is killed or
610 messages to it were lost, and optionally return a guard that can be used
611 to stop monitoring again.
612
613 In the first form (callback), the callback is simply called with any
614 number of C<@reason> elements (no @reason means that the port was deleted
615 "normally"). Note also that I<< the callback B<must> never die >>, so use
616 C<eval> if unsure.
617
618 In the second form (another port given), the other port (C<$rcvport>)
619 will be C<kil>'ed with C<@reason>, if a @reason was specified, i.e. on
620 "normal" kils nothing happens, while under all other conditions, the other
621 port is killed with the same reason.
622
623 The third form (kill self) is the same as the second form, except that
624 C<$rvport> defaults to C<$SELF>.
625
626 In the last form (message), a message of the form C<@msg, @reason> will be
627 C<snd>.
628
629 Monitoring-actions are one-shot: once messages are lost (and a monitoring
630 alert was raised), they are removed and will not trigger again.
631
632 As a rule of thumb, monitoring requests should always monitor a port from
633 a local port (or callback). The reason is that kill messages might get
634 lost, just like any other message. Another less obvious reason is that
635 even monitoring requests can get lost (for example, when the connection
636 to the other node goes down permanently). When monitoring a port locally
637 these problems do not exist.
638
639 C<mon> effectively guarantees that, in the absence of hardware failures,
640 after starting the monitor, either all messages sent to the port will
641 arrive, or the monitoring action will be invoked after possible message
642 loss has been detected. No messages will be lost "in between" (after
643 the first lost message no further messages will be received by the
644 port). After the monitoring action was invoked, further messages might get
645 delivered again.
646
647 Inter-host-connection timeouts and monitoring depend on the transport
648 used. The only transport currently implemented is TCP, and AnyEvent::MP
649 relies on TCP to detect node-downs (this can take 10-15 minutes on a
650 non-idle connection, and usually around two hours for idle connections).
651
652 This means that monitoring is good for program errors and cleaning up
653 stuff eventually, but they are no replacement for a timeout when you need
654 to ensure some maximum latency.
655
656 Example: call a given callback when C<$port> is killed.
657
658 mon $port, sub { warn "port died because of <@_>\n" };
659
660 Example: kill ourselves when C<$port> is killed abnormally.
661
662 mon $port;
663
664 Example: send us a restart message when another C<$port> is killed.
665
666 mon $port, $self => "restart";
667
668 =cut
669
670 sub mon {
671 my ($nodeid, $port) = split /#/, shift, 2;
672
673 my $node = $NODE{$nodeid} || add_node $nodeid;
674
675 my $cb = @_ ? shift : $SELF || Carp::croak 'mon: called with one argument only, but $SELF not set,';
676
677 unless (ref $cb) {
678 if (@_) {
679 # send a kill info message
680 my (@msg) = ($cb, @_);
681 $cb = sub { snd @msg, @_ };
682 } else {
683 # simply kill other port
684 my $port = $cb;
685 $cb = sub { kil $port, @_ if @_ };
686 }
687 }
688
689 $node->monitor ($port, $cb);
690
691 defined wantarray
692 and ($cb += 0, Guard::guard { $node->unmonitor ($port, $cb) })
693 }
694
695 =item $guard = mon_guard $port, $ref, $ref...
696
697 Monitors the given C<$port> and keeps the passed references. When the port
698 is killed, the references will be freed.
699
700 Optionally returns a guard that will stop the monitoring.
701
702 This function is useful when you create e.g. timers or other watchers and
703 want to free them when the port gets killed (note the use of C<psub>):
704
705 $port->rcv (start => sub {
706 my $timer; $timer = mon_guard $port, AE::timer 1, 1, psub {
707 undef $timer if 0.9 < rand;
708 });
709 });
710
711 =cut
712
713 sub mon_guard {
714 my ($port, @refs) = @_;
715
716 #TODO: mon-less form?
717
718 mon $port, sub { 0 && @refs }
719 }
720
721 =item kil $port[, @reason]
722
723 Kill the specified port with the given C<@reason>.
724
725 If no C<@reason> is specified, then the port is killed "normally" -
726 monitor callback will be invoked, but the kil will not cause linked ports
727 (C<mon $mport, $lport> form) to get killed.
728
729 If a C<@reason> is specified, then linked ports (C<mon $mport, $lport>
730 form) get killed with the same reason.
731
732 Runtime errors while evaluating C<rcv> callbacks or inside C<psub> blocks
733 will be reported as reason C<< die => $@ >>.
734
735 Transport/communication errors are reported as C<< transport_error =>
736 $message >>.
737
738 Common idioms:
739
740 # silently remove yourself, do not kill linked ports
741 kil $SELF;
742
743 # report a failure in some detail
744 kil $SELF, failure_mode_1 => "it failed with too high temperature";
745
746 # do not waste much time with killing, just die when something goes wrong
747 open my $fh, "<file"
748 or die "file: $!";
749
750 =item $port = spawn $node, $initfunc[, @initdata]
751
752 Creates a port on the node C<$node> (which can also be a port ID, in which
753 case it's the node where that port resides).
754
755 The port ID of the newly created port is returned immediately, and it is
756 possible to immediately start sending messages or to monitor the port.
757
758 After the port has been created, the init function is called on the remote
759 node, in the same context as a C<rcv> callback. This function must be a
760 fully-qualified function name (e.g. C<MyApp::Chat::Server::init>). To
761 specify a function in the main program, use C<::name>.
762
763 If the function doesn't exist, then the node tries to C<require>
764 the package, then the package above the package and so on (e.g.
765 C<MyApp::Chat::Server>, C<MyApp::Chat>, C<MyApp>) until the function
766 exists or it runs out of package names.
767
768 The init function is then called with the newly-created port as context
769 object (C<$SELF>) and the C<@initdata> values as arguments. It I<must>
770 call one of the C<rcv> functions to set callbacks on C<$SELF>, otherwise
771 the port might not get created.
772
773 A common idiom is to pass a local port, immediately monitor the spawned
774 port, and in the remote init function, immediately monitor the passed
775 local port. This two-way monitoring ensures that both ports get cleaned up
776 when there is a problem.
777
778 C<spawn> guarantees that the C<$initfunc> has no visible effects on the
779 caller before C<spawn> returns (by delaying invocation when spawn is
780 called for the local node).
781
782 Example: spawn a chat server port on C<$othernode>.
783
784 # this node, executed from within a port context:
785 my $server = spawn $othernode, "MyApp::Chat::Server::connect", $SELF;
786 mon $server;
787
788 # init function on C<$othernode>
789 sub connect {
790 my ($srcport) = @_;
791
792 mon $srcport;
793
794 rcv $SELF, sub {
795 ...
796 };
797 }
798
799 =cut
800
801 sub _spawn {
802 my $port = shift;
803 my $init = shift;
804
805 # rcv will create the actual port
806 local $SELF = "$NODE#$port";
807 eval {
808 &{ load_func $init }
809 };
810 _self_die if $@;
811 }
812
813 sub spawn(@) {
814 my ($nodeid, undef) = split /#/, shift, 2;
815
816 my $id = $RUNIQ . ++$ID;
817
818 $_[0] =~ /::/
819 or Carp::croak "spawn init function must be a fully-qualified name, caught";
820
821 snd_to_func $nodeid, "AnyEvent::MP::_spawn" => $id, @_;
822
823 "$nodeid#$id"
824 }
825
826
827 =item after $timeout, @msg
828
829 =item after $timeout, $callback
830
831 Either sends the given message, or call the given callback, after the
832 specified number of seconds.
833
834 This is simply a utility function that comes in handy at times - the
835 AnyEvent::MP author is not convinced of the wisdom of having it, though,
836 so it may go away in the future.
837
838 =cut
839
840 sub after($@) {
841 my ($timeout, @action) = @_;
842
843 my $t; $t = AE::timer $timeout, 0, sub {
844 undef $t;
845 ref $action[0]
846 ? $action[0]()
847 : snd @action;
848 };
849 }
850
851 #=item $cb2 = timeout $seconds, $cb[, @args]
852
853 =item cal $port, @msg, $callback[, $timeout]
854
855 A simple form of RPC - sends a message to the given C<$port> with the
856 given contents (C<@msg>), but adds a reply port to the message.
857
858 The reply port is created temporarily just for the purpose of receiving
859 the reply, and will be C<kil>ed when no longer needed.
860
861 A reply message sent to the port is passed to the C<$callback> as-is.
862
863 If an optional time-out (in seconds) is given and it is not C<undef>,
864 then the callback will be called without any arguments after the time-out
865 elapsed and the port is C<kil>ed.
866
867 If no time-out is given (or it is C<undef>), then the local port will
868 monitor the remote port instead, so it eventually gets cleaned-up.
869
870 Currently this function returns the temporary port, but this "feature"
871 might go in future versions unless you can make a convincing case that
872 this is indeed useful for something.
873
874 =cut
875
876 sub cal(@) {
877 my $timeout = ref $_[-1] ? undef : pop;
878 my $cb = pop;
879
880 my $port = port {
881 undef $timeout;
882 kil $SELF;
883 &$cb;
884 };
885
886 if (defined $timeout) {
887 $timeout = AE::timer $timeout, 0, sub {
888 undef $timeout;
889 kil $port;
890 $cb->();
891 };
892 } else {
893 mon $_[0], sub {
894 kil $port;
895 $cb->();
896 };
897 }
898
899 push @_, $port;
900 &snd;
901
902 $port
903 }
904
905 =back
906
907 =head1 DISTRIBUTED DATABASE
908
909 AnyEvent::MP comes with a simple distributed database. The database will
910 be mirrored asynchronously on all global nodes. Other nodes bind to one
911 of the global nodes for their needs. Every node has a "local database"
912 which contains all the values that are set locally. All local databases
913 are merged together to form the global database, which can be queried.
914
915 The database structure is that of a two-level hash - the database hash
916 contains hashes which contain values, similarly to a perl hash of hashes,
917 i.e.:
918
919 $DATABASE{$family}{$subkey} = $value
920
921 The top level hash key is called "family", and the second-level hash key
922 is called "subkey" or simply "key".
923
924 The family must be alphanumeric, i.e. start with a letter and consist
925 of letters, digits, underscores and colons (C<[A-Za-z][A-Za-z0-9_:]*>,
926 pretty much like Perl module names.
927
928 As the family namespace is global, it is recommended to prefix family names
929 with the name of the application or module using it.
930
931 The subkeys must be non-empty strings, with no further restrictions.
932
933 The values should preferably be strings, but other perl scalars should
934 work as well (such as C<undef>, arrays and hashes).
935
936 Every database entry is owned by one node - adding the same family/subkey
937 combination on multiple nodes will not cause discomfort for AnyEvent::MP,
938 but the result might be nondeterministic, i.e. the key might have
939 different values on different nodes.
940
941 Different subkeys in the same family can be owned by different nodes
942 without problems, and in fact, this is the common method to create worker
943 pools. For example, a worker port for image scaling might do this:
944
945 db_set my_image_scalers => $port;
946
947 And clients looking for an image scaler will want to get the
948 C<my_image_scalers> keys from time to time:
949
950 db_keys my_image_scalers => sub {
951 @ports = @{ $_[0] };
952 };
953
954 Or better yet, they want to monitor the database family, so they always
955 have a reasonable up-to-date copy:
956
957 db_mon my_image_scalers => sub {
958 @ports = keys %{ $_[0] };
959 };
960
961 In general, you can set or delete single subkeys, but query and monitor
962 whole families only.
963
964 If you feel the need to monitor or query a single subkey, try giving it
965 it's own family.
966
967 =over
968
969 =item db_set $family => $subkey [=> $value]
970
971 Sets (or replaces) a key to the database - if C<$value> is omitted,
972 C<undef> is used instead.
973
974 =item db_del $family => $subkey...
975
976 Deletes one or more subkeys from the database family.
977
978 =item $guard = db_reg $family => $subkey [=> $value]
979
980 Sets the key on the database and returns a guard. When the guard is
981 destroyed, the key is deleted from the database. If C<$value> is missing,
982 then C<undef> is used.
983
984 =item db_family $family => $cb->(\%familyhash)
985
986 Queries the named database C<$family> and call the callback with the
987 family represented as a hash. You can keep and freely modify the hash.
988
989 =item db_keys $family => $cb->(\@keys)
990
991 Same as C<db_family>, except it only queries the family I<subkeys> and passes
992 them as array reference to the callback.
993
994 =item db_values $family => $cb->(\@values)
995
996 Same as C<db_family>, except it only queries the family I<values> and passes them
997 as array reference to the callback.
998
999 =item $guard = db_mon $family => $cb->($familyhash, \@added, \@changed, \@deleted)
1000
1001 Creates a monitor on the given database family. Each time a key is set
1002 or or is deleted the callback is called with a hash containing the
1003 database family and three lists of added, changed and deleted subkeys,
1004 respectively. If no keys have changed then the array reference might be
1005 C<undef> or even missing.
1006
1007 If not called in void context, a guard object is returned that, when
1008 destroyed, stops the monitor.
1009
1010 The family hash reference and the key arrays belong to AnyEvent::MP and
1011 B<must not be modified or stored> by the callback. When in doubt, make a
1012 copy.
1013
1014 As soon as possible after the monitoring starts, the callback will be
1015 called with the intiial contents of the family, even if it is empty,
1016 i.e. there will always be a timely call to the callback with the current
1017 contents.
1018
1019 It is possible that the callback is called with a change event even though
1020 the subkey is already present and the value has not changed.
1021
1022 The monitoring stops when the guard object is destroyed.
1023
1024 Example: on every change to the family "mygroup", print out all keys.
1025
1026 my $guard = db_mon mygroup => sub {
1027 my ($family, $a, $c, $d) = @_;
1028 print "mygroup members: ", (join " ", keys %$family), "\n";
1029 };
1030
1031 Exmaple: wait until the family "My::Module::workers" is non-empty.
1032
1033 my $guard; $guard = db_mon My::Module::workers => sub {
1034 my ($family, $a, $c, $d) = @_;
1035 return unless %$family;
1036 undef $guard;
1037 print "My::Module::workers now nonempty\n";
1038 };
1039
1040 Example: print all changes to the family "AnyRvent::Fantasy::Module".
1041
1042 my $guard = db_mon AnyRvent::Fantasy::Module => sub {
1043 my ($family, $a, $c, $d) = @_;
1044
1045 print "+$_=$family->{$_}\n" for @$a;
1046 print "*$_=$family->{$_}\n" for @$c;
1047 print "-$_=$family->{$_}\n" for @$d;
1048 };
1049
1050 =cut
1051
1052 =back
1053
1054 =head1 AnyEvent::MP vs. Distributed Erlang
1055
1056 AnyEvent::MP got lots of its ideas from distributed Erlang (Erlang node
1057 == aemp node, Erlang process == aemp port), so many of the documents and
1058 programming techniques employed by Erlang apply to AnyEvent::MP. Here is a
1059 sample:
1060
1061 http://www.erlang.se/doc/programming_rules.shtml
1062 http://erlang.org/doc/getting_started/part_frame.html # chapters 3 and 4
1063 http://erlang.org/download/erlang-book-part1.pdf # chapters 5 and 6
1064 http://erlang.org/download/armstrong_thesis_2003.pdf # chapters 4 and 5
1065
1066 Despite the similarities, there are also some important differences:
1067
1068 =over 4
1069
1070 =item * Node IDs are arbitrary strings in AEMP.
1071
1072 Erlang relies on special naming and DNS to work everywhere in the same
1073 way. AEMP relies on each node somehow knowing its own address(es) (e.g. by
1074 configuration or DNS), and possibly the addresses of some seed nodes, but
1075 will otherwise discover other nodes (and their IDs) itself.
1076
1077 =item * Erlang has a "remote ports are like local ports" philosophy, AEMP
1078 uses "local ports are like remote ports".
1079
1080 The failure modes for local ports are quite different (runtime errors
1081 only) then for remote ports - when a local port dies, you I<know> it dies,
1082 when a connection to another node dies, you know nothing about the other
1083 port.
1084
1085 Erlang pretends remote ports are as reliable as local ports, even when
1086 they are not.
1087
1088 AEMP encourages a "treat remote ports differently" philosophy, with local
1089 ports being the special case/exception, where transport errors cannot
1090 occur.
1091
1092 =item * Erlang uses processes and a mailbox, AEMP does not queue.
1093
1094 Erlang uses processes that selectively receive messages out of order, and
1095 therefore needs a queue. AEMP is event based, queuing messages would serve
1096 no useful purpose. For the same reason the pattern-matching abilities
1097 of AnyEvent::MP are more limited, as there is little need to be able to
1098 filter messages without dequeuing them.
1099
1100 This is not a philosophical difference, but simply stems from AnyEvent::MP
1101 being event-based, while Erlang is process-based.
1102
1103 You cna have a look at L<Coro::MP> for a more Erlang-like process model on
1104 top of AEMP and Coro threads.
1105
1106 =item * Erlang sends are synchronous, AEMP sends are asynchronous.
1107
1108 Sending messages in Erlang is synchronous and blocks the process until
1109 a conenction has been established and the message sent (and so does not
1110 need a queue that can overflow). AEMP sends return immediately, connection
1111 establishment is handled in the background.
1112
1113 =item * Erlang suffers from silent message loss, AEMP does not.
1114
1115 Erlang implements few guarantees on messages delivery - messages can get
1116 lost without any of the processes realising it (i.e. you send messages a,
1117 b, and c, and the other side only receives messages a and c).
1118
1119 AEMP guarantees (modulo hardware errors) correct ordering, and the
1120 guarantee that after one message is lost, all following ones sent to the
1121 same port are lost as well, until monitoring raises an error, so there are
1122 no silent "holes" in the message sequence.
1123
1124 If you want your software to be very reliable, you have to cope with
1125 corrupted and even out-of-order messages in both Erlang and AEMP. AEMP
1126 simply tries to work better in common error cases, such as when a network
1127 link goes down.
1128
1129 =item * Erlang can send messages to the wrong port, AEMP does not.
1130
1131 In Erlang it is quite likely that a node that restarts reuses an Erlang
1132 process ID known to other nodes for a completely different process,
1133 causing messages destined for that process to end up in an unrelated
1134 process.
1135
1136 AEMP does not reuse port IDs, so old messages or old port IDs floating
1137 around in the network will not be sent to an unrelated port.
1138
1139 =item * Erlang uses unprotected connections, AEMP uses secure
1140 authentication and can use TLS.
1141
1142 AEMP can use a proven protocol - TLS - to protect connections and
1143 securely authenticate nodes.
1144
1145 =item * The AEMP protocol is optimised for both text-based and binary
1146 communications.
1147
1148 The AEMP protocol, unlike the Erlang protocol, supports both programming
1149 language independent text-only protocols (good for debugging), and binary,
1150 language-specific serialisers (e.g. Storable). By default, unless TLS is
1151 used, the protocol is actually completely text-based.
1152
1153 It has also been carefully designed to be implementable in other languages
1154 with a minimum of work while gracefully degrading functionality to make the
1155 protocol simple.
1156
1157 =item * AEMP has more flexible monitoring options than Erlang.
1158
1159 In Erlang, you can chose to receive I<all> exit signals as messages or
1160 I<none>, there is no in-between, so monitoring single Erlang processes is
1161 difficult to implement.
1162
1163 Monitoring in AEMP is more flexible than in Erlang, as one can choose
1164 between automatic kill, exit message or callback on a per-port basis.
1165
1166 =item * Erlang tries to hide remote/local connections, AEMP does not.
1167
1168 Monitoring in Erlang is not an indicator of process death/crashes, in the
1169 same way as linking is (except linking is unreliable in Erlang).
1170
1171 In AEMP, you don't "look up" registered port names or send to named ports
1172 that might or might not be persistent. Instead, you normally spawn a port
1173 on the remote node. The init function monitors you, and you monitor the
1174 remote port. Since both monitors are local to the node, they are much more
1175 reliable (no need for C<spawn_link>).
1176
1177 This also saves round-trips and avoids sending messages to the wrong port
1178 (hard to do in Erlang).
1179
1180 =back
1181
1182 =head1 RATIONALE
1183
1184 =over 4
1185
1186 =item Why strings for port and node IDs, why not objects?
1187
1188 We considered "objects", but found that the actual number of methods
1189 that can be called are quite low. Since port and node IDs travel over
1190 the network frequently, the serialising/deserialising would add lots of
1191 overhead, as well as having to keep a proxy object everywhere.
1192
1193 Strings can easily be printed, easily serialised etc. and need no special
1194 procedures to be "valid".
1195
1196 And as a result, a port with just a default receiver consists of a single
1197 code reference stored in a global hash - it can't become much cheaper.
1198
1199 =item Why favour JSON, why not a real serialising format such as Storable?
1200
1201 In fact, any AnyEvent::MP node will happily accept Storable as framing
1202 format, but currently there is no way to make a node use Storable by
1203 default (although all nodes will accept it).
1204
1205 The default framing protocol is JSON because a) JSON::XS is many times
1206 faster for small messages and b) most importantly, after years of
1207 experience we found that object serialisation is causing more problems
1208 than it solves: Just like function calls, objects simply do not travel
1209 easily over the network, mostly because they will always be a copy, so you
1210 always have to re-think your design.
1211
1212 Keeping your messages simple, concentrating on data structures rather than
1213 objects, will keep your messages clean, tidy and efficient.
1214
1215 =back
1216
1217 =head1 SEE ALSO
1218
1219 L<AnyEvent::MP::Intro> - a gentle introduction.
1220
1221 L<AnyEvent::MP::Kernel> - more, lower-level, stuff.
1222
1223 L<AnyEvent::MP::Global> - network maintenance and port groups, to find
1224 your applications.
1225
1226 L<AnyEvent::MP::DataConn> - establish data connections between nodes.
1227
1228 L<AnyEvent::MP::LogCatcher> - simple service to display log messages from
1229 all nodes.
1230
1231 L<AnyEvent>.
1232
1233 =head1 AUTHOR
1234
1235 Marc Lehmann <schmorp@schmorp.de>
1236 http://home.schmorp.de/
1237
1238 =cut
1239
1240 1
1241