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

Comparing AnyEvent-MP/MP.pm (file contents):
Revision 1.7 by root, Sat Aug 1 15:04:30 2009 UTC vs.
Revision 1.125 by root, Sat Mar 3 13:07:19 2012 UTC

1=head1 NAME 1=head1 NAME
2 2
3AnyEvent::MP - multi-processing/message-passing framework 3AnyEvent::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 # returns this node identifier
10 $NODE # contains this node identifier 9 $NODE # contains this node's node ID
10 NODE # returns this node's node ID
11 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
12 snd $port, type => data...; 20 snd $port, type => data...;
21 snd $port, @msg;
22 snd @msg_with_first_element_being_a_port;
13 23
14 rcv $port, smartmatch => $cb->($port, @msg); 24 # creating/using ports, the simple way
25 my $simple_port = port { my @msg = @_ };
15 26
16 # examples: 27 # creating/using ports, tagged message matching
28 my $port = port;
17 rcv $port2, ping => sub { snd $_[0], "pong"; 0 }; 29 rcv $port, ping => sub { snd $_[0], "pong" };
18 rcv $port1, pong => sub { warn "pong received\n" }; 30 rcv $port, pong => sub { warn "pong received\n" };
19 snd $port2, ping => $port1;
20 31
21 # more, smarter, matches (_any_ is exported by this module) 32 # create a port on another node
22 rcv $port, [child_died => $pid] => sub { ... 33 my $port = spawn $node, $initfunc, @initdata;
23 rcv $port, [_any_, _any_, 3] => sub { .. $_[2] is 3 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 $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.
24 59
25=head1 DESCRIPTION 60=head1 DESCRIPTION
26 61
27This module (-family) implements a simple message passing framework. 62This module (-family) implements a simple message passing framework.
28 63
29Despite its simplicity, you can securely message other processes running 64Despite its simplicity, you can securely message other processes running
30on the same or other hosts. 65on the same or other hosts, and you can supervise entities remotely.
31 66
32At the moment, this module family is severly brokena nd underdocumented, 67For an introduction to this module family, see the L<AnyEvent::MP::Intro>
33so do not use. This was uploaded mainly to resreve the CPAN namespace - 68manual page and the examples under F<eg/>.
34stay tuned!
35 69
36=head1 CONCEPTS 70=head1 CONCEPTS
37 71
38=over 4 72=over 4
39 73
40=item port 74=item port
41 75
42A port is something you can send messages to with the C<snd> function, and 76Not to be confused with a TCP port, a "port" is something you can send
43you can register C<rcv> handlers with. All C<rcv> handlers will receive 77messages to (with the C<snd> function).
44messages they match, messages will not be queued.
45 78
79Ports allow you to register C<rcv> handlers that can match all or just
80some messages. Messages send to ports will not be queued, regardless of
81anything was listening for them or not.
82
83Ports are represented by (printable) strings called "port IDs".
84
46=item port id - C<noderef#portname> 85=item port ID - C<nodeid#portname>
47 86
48A port id is always the noderef, a hash-mark (C<#>) as separator, followed 87A port ID is the concatenation of a node ID, a hash-mark (C<#>)
49by a port name (a printable string of unspecified format). 88as separator, and a port name (a printable string of unspecified
89format created by AnyEvent::MP).
50 90
51=item node 91=item node
52 92
53A node is a single process containing at least one port - the node 93A node is a single process containing at least one port - the node port,
54port. You can send messages to node ports to let them create new ports, 94which enables nodes to manage each other remotely, and to create new
55among other things. 95ports.
56 96
57Initially, nodes are either private (single-process only) or hidden 97Nodes are either public (have one or more listening ports) or private
58(connected to a master node only). Only when they epxlicitly "become 98(no listening ports). Private nodes cannot talk to other private nodes
59public" can you send them messages from unrelated other nodes. 99currently, but all nodes can talk to public nodes.
60 100
61=item noderef - C<host:port,host:port...>, C<id@noderef>, C<id> 101Nodes is represented by (printable) strings called "node IDs".
62 102
103=item node ID - C<[A-Za-z0-9_\-.:]*>
104
63A noderef is a string that either uniquely identifies a given node (for 105A node ID is a string that uniquely identifies the node within a
64private and hidden nodes), or contains a recipe on how to reach a given 106network. Depending on the configuration used, node IDs can look like a
65node (for public nodes). 107hostname, a hostname and a port, or a random string. AnyEvent::MP itself
108doesn't interpret node IDs in any way except to uniquely identify a node.
109
110=item binds - C<ip:port>
111
112Nodes can only talk to each other by creating some kind of connection to
113each other. To do this, nodes should listen on one or more local transport
114endpoints - binds.
115
116Currently, only standard C<ip:port> specifications can be used, which
117specify TCP ports to listen on. So a bind is basically just a tcp socket
118in listening mode thta accepts conenctions form other nodes.
119
120=item seed nodes
121
122When a node starts, it knows nothing about the network it is in - it
123needs to connect to at least one other node that is already in the
124network. These other nodes are called "seed nodes".
125
126Seed nodes themselves are not special - they are seed nodes only because
127some other node I<uses> them as such, but any node can be used as seed
128node for other nodes, and eahc node cna use a different set of seed nodes.
129
130In addition to discovering the network, seed nodes are also used to
131maintain the network - all nodes using the same seed node form are part of
132the same network. If a network is split into multiple subnets because e.g.
133the network link between the parts goes down, then using the same seed
134nodes for all nodes ensures that eventually the subnets get merged again.
135
136Seed nodes are expected to be long-running, and at least one seed node
137should always be available. They should also be relatively responsive - a
138seed node that blocks for long periods will slow down everybody else.
139
140For small networks, it's best if every node uses the same set of seed
141nodes. For large networks, it can be useful to specify "regional" seed
142nodes for most nodes in an area, and use all seed nodes as seed nodes for
143each other. What's important is that all seed nodes connections form a
144complete graph, so that the network cannot split into separate subnets
145forever.
146
147Seed nodes are represented by seed IDs.
148
149=item seed IDs - C<host:port>
150
151Seed IDs are transport endpoint(s) (usually a hostname/IP address and a
152TCP port) of nodes that should be used as seed nodes.
153
154=item global nodes
155
156An AEMP network needs a discovery service - nodes need to know how to
157connect to other nodes they only know by name. In addition, AEMP offers a
158distributed "group database", which maps group names to a list of strings
159- for example, to register worker ports.
160
161A network needs at least one global node to work, and allows every node to
162be a global node.
163
164Any node that loads the L<AnyEvent::MP::Global> module becomes a global
165node and tries to keep connections to all other nodes. So while it can
166make sense to make every node "global" in small networks, it usually makes
167sense to only make seed nodes into global nodes in large networks (nodes
168keep connections to seed nodes and global nodes, so makign them the same
169reduces overhead).
66 170
67=back 171=back
68 172
69=head1 VARIABLES/FUNCTIONS 173=head1 VARIABLES/FUNCTIONS
70 174
72 176
73=cut 177=cut
74 178
75package AnyEvent::MP; 179package AnyEvent::MP;
76 180
77use AnyEvent::MP::Util (); 181use AnyEvent::MP::Config ();
78use AnyEvent::MP::Node; 182use AnyEvent::MP::Kernel;
79use AnyEvent::MP::Transport; 183use AnyEvent::MP::Kernel qw(%NODE %PORT %PORT_DATA $UNIQ $RUNIQ $ID);
80 184
81use utf8;
82use common::sense; 185use common::sense;
83 186
84use Carp (); 187use Carp ();
85 188
86use AE (); 189use AE ();
190use Guard ();
87 191
88use base "Exporter"; 192use base "Exporter";
89 193
90our $VERSION = '0.01'; 194our $VERSION = $AnyEvent::MP::Config::VERSION;
91our @EXPORT = qw(NODE $NODE $PORT snd rcv _any_);
92 195
93our $DEFAULT_SECRET; 196our @EXPORT = qw(
94our $DEFAULT_PORT = "4040"; 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);
95 203
96our $CONNECT_INTERVAL = 5; # new connect every 5s, at least 204our $SELF;
97our $CONNECT_TIMEOUT = 30; # includes handshake
98 205
99sub default_secret { 206sub _self_die() {
100 unless (defined $DEFAULT_SECRET) { 207 my $msg = $@;
101 if (open my $fh, "<$ENV{HOME}/.aemp-secret") { 208 $msg =~ s/\n+$// unless ref $msg;
102 sysread $fh, $DEFAULT_SECRET, -s $fh; 209 kil $SELF, die => $msg;
103 } else {
104 $DEFAULT_SECRET = AnyEvent::MP::Util::nonce 32;
105 }
106 }
107
108 $DEFAULT_SECRET
109} 210}
110 211
111=item NODE / $NODE 212=item $thisnode = NODE / $NODE
112 213
113The C<NODE ()> function and the C<$NODE> variable contain the noderef of 214The C<NODE> function returns, and the C<$NODE> variable contains, the node
114the local node. The value is initialised by a call to C<become_public> or 215ID of the node running in the current process. This value is initialised by
115C<become_slave>, after which all local port identifiers become invalid. 216a call to C<configure>.
116 217
117=cut 218=item $nodeid = node_of $port
118 219
119our $UNIQ = sprintf "%x.%x", $$, time; # per-process/node unique cookie 220Extracts and returns the node ID from a port ID or a node ID.
120our $ID = "a0";
121our $PUBLIC = 0;
122our $NODE;
123our $PORT;
124 221
125our %NODE; # node id to transport mapping, or "undef", for local node 222=item configure $profile, key => value...
126our %PORT; # local ports
127our %LISTENER; # local transports
128 223
129sub NODE() { $NODE } 224=item configure key => value...
130 225
131{ 226Before a node can talk to other nodes on the network (i.e. enter
132 use POSIX (); 227"distributed mode") it has to configure itself - the minimum a node needs
133 my $nodename = (POSIX::uname)[1]; 228to know is its own name, and optionally it should know the addresses of
134 $NODE = "$$\@$nodename"; 229some other nodes in the network to discover other nodes.
135}
136 230
137sub _ANY_() { 1 } 231This function configures a node - it must be called exactly once (or
138sub _any_() { \&_ANY_ } 232never) before calling other AnyEvent::MP functions.
139 233
140sub add_node { 234The key/value pairs are basically the same ones as documented for the
141 my ($noderef) = @_; 235F<aemp> command line utility (sans the set/del prefix), with two additions:
142 236
143 return $NODE{$noderef} 237=over 4
144 if exists $NODE{$noderef};
145 238
146 for (split /,/, $noderef) { 239=item norc => $boolean (default false)
147 return $NODE{$noderef} = $NODE{$_}
148 if exists $NODE{$_};
149 }
150 240
151 # for indirect sends, use a different class 241If true, then the rc file (e.g. F<~/.perl-anyevent-mp>) will I<not>
152 my $node = new AnyEvent::MP::Node::Direct $noderef; 242be consulted - all configuraiton options must be specified in the
243C<configure> call.
153 244
154 $NODE{$_} = $node 245=item force => $boolean (default false)
155 for $noderef, split /,/, $noderef;
156 246
157 $node 247IF true, then the values specified in the C<configure> will take
158} 248precedence over any values configured via the rc file. The default is for
249the rc file to override any options specified in the program.
159 250
251=back
252
253=over 4
254
255=item step 1, gathering configuration from profiles
256
257The function first looks up a profile in the aemp configuration (see the
258L<aemp> commandline utility). The profile name can be specified via the
259named C<profile> parameter or can simply be the first parameter). If it is
260missing, then the nodename (F<uname -n>) will be used as profile name.
261
262The profile data is then gathered as follows:
263
264First, all remaining key => value pairs (all of which are conveniently
265undocumented at the moment) will be interpreted as configuration
266data. Then they will be overwritten by any values specified in the global
267default configuration (see the F<aemp> utility), then the chain of
268profiles chosen by the profile name (and any C<parent> attributes).
269
270That means that the values specified in the profile have highest priority
271and the values specified directly via C<configure> have lowest priority,
272and can only be used to specify defaults.
273
274If the profile specifies a node ID, then this will become the node ID of
275this process. If not, then the profile name will be used as node ID, with
276a slash (C</>) attached.
277
278If the node ID (or profile name) ends with a slash (C</>), then a random
279string is appended to make it unique.
280
281=item step 2, bind listener sockets
282
283The next step is to look up the binds in the profile, followed by binding
284aemp protocol listeners on all binds specified (it is possible and valid
285to have no binds, meaning that the node cannot be contacted form the
286outside. This means the node cannot talk to other nodes that also have no
287binds, but it can still talk to all "normal" nodes).
288
289If the profile does not specify a binds list, then a default of C<*> is
290used, meaning the node will bind on a dynamically-assigned port on every
291local IP address it finds.
292
293=item step 3, connect to seed nodes
294
295As the last step, the seed ID list from the profile is passed to the
296L<AnyEvent::MP::Global> module, which will then use it to keep
297connectivity with at least one node at any point in time.
298
299=back
300
301Example: become a distributed node using the local node name as profile.
302This should be the most common form of invocation for "daemon"-type nodes.
303
304 configure
305
306Example: become an anonymous node. This form is often used for commandline
307clients.
308
309 configure nodeid => "anon/";
310
311Example: configure a node using a profile called seed, which is suitable
312for a seed node as it binds on all local addresses on a fixed port (4040,
313customary for aemp).
314
315 # use the aemp commandline utility
316 # aemp profile seed binds '*:4040'
317
318 # then use it
319 configure profile => "seed";
320
321 # or simply use aemp from the shell again:
322 # aemp run profile seed
323
324 # or provide a nicer-to-remember nodeid
325 # aemp run profile seed nodeid "$(hostname)"
326
327=item $SELF
328
329Contains the current port id while executing C<rcv> callbacks or C<psub>
330blocks.
331
332=item *SELF, SELF, %SELF, @SELF...
333
334Due to some quirks in how perl exports variables, it is impossible to
335just export C<$SELF>, all the symbols named C<SELF> are exported by this
336module, but only C<$SELF> is currently used.
337
160=item snd $portid, type => @data 338=item snd $port, type => @data
161 339
162=item snd $portid, @msg 340=item snd $port, @msg
163 341
164Send the given message to the given port ID, which can identify either a 342Send the given message to the given port, which can identify either a
165local or a remote port. 343local or a remote port, and must be a port ID.
166 344
167While the message can be about anything, it is highly recommended to use 345While the message can be almost anything, it is highly recommended to
168a constant string as first element. 346use a string as first element (a port ID, or some word that indicates a
347request type etc.) and to consist if only simple perl values (scalars,
348arrays, hashes) - if you think you need to pass an object, think again.
169 349
170The message data effectively becomes read-only after a call to this 350The message data logically becomes read-only after a call to this
171function: modifying any argument is not allowed and can cause many 351function: modifying any argument (or values referenced by them) is
172problems. 352forbidden, as there can be considerable time between the call to C<snd>
353and the time the message is actually being serialised - in fact, it might
354never be copied as within the same process it is simply handed to the
355receiving port.
173 356
174The type of data you can transfer depends on the transport protocol: when 357The type of data you can transfer depends on the transport protocol: when
175JSON is used, then only strings, numbers and arrays and hashes consisting 358JSON is used, then only strings, numbers and arrays and hashes consisting
176of those are allowed (no objects). When Storable is used, then anything 359of those are allowed (no objects). When Storable is used, then anything
177that Storable can serialise and deserialise is allowed, and for the local 360that Storable can serialise and deserialise is allowed, and for the local
178node, anything can be passed. 361node, anything can be passed. Best rely only on the common denominator of
362these.
179 363
180=cut 364=item $local_port = port
181 365
182sub snd(@) { 366Create a new local port object and returns its port ID. Initially it has
183 my ($noderef, $port) = split /#/, shift, 2; 367no callbacks set and will throw an error when it receives messages.
184 368
185 add_node $noderef 369=item $local_port = port { my @msg = @_ }
186 unless exists $NODE{$noderef};
187 370
188 $NODE{$noderef}->send (["$port", [@_]]); 371Creates a new local port, and returns its ID. Semantically the same as
372creating a port and calling C<rcv $port, $callback> on it.
373
374The block will be called for every message received on the port, with the
375global variable C<$SELF> set to the port ID. Runtime errors will cause the
376port to be C<kil>ed. The message will be passed as-is, no extra argument
377(i.e. no port ID) will be passed to the callback.
378
379If you want to stop/destroy the port, simply C<kil> it:
380
381 my $port = port {
382 my @msg = @_;
383 ...
384 kil $SELF;
385 };
386
387=cut
388
389sub rcv($@);
390
391sub _kilme {
392 die "received message on port without callback";
189} 393}
190 394
395sub port(;&) {
396 my $id = $UNIQ . ++$ID;
397 my $port = "$NODE#$id";
398
399 rcv $port, shift || \&_kilme;
400
401 $port
402}
403
191=item rcv $portid, type => $callback->(@msg) 404=item rcv $local_port, $callback->(@msg)
192 405
193=item rcv $portid, $smartmatch => $callback->(@msg) 406Replaces the default callback on the specified port. There is no way to
407remove the default callback: use C<sub { }> to disable it, or better
408C<kil> the port when it is no longer needed.
194 409
195=item rcv $portid, [$smartmatch...] => $callback->(@msg) 410The global C<$SELF> (exported by this module) contains C<$port> while
411executing the callback. Runtime errors during callback execution will
412result in the port being C<kil>ed.
196 413
197Register a callback on the port identified by C<$portid>, which I<must> be 414The default callback received all messages not matched by a more specific
198a local port. 415C<tag> match.
199 416
200The callback has to return a true value when its work is done, after 417=item rcv $local_port, tag => $callback->(@msg_without_tag), ...
201which is will be removed, or a false value in which case it will stay
202registered.
203 418
204If the match is an array reference, then it will be matched against the 419Register (or replace) callbacks to be called on messages starting with the
205first elements of the message, otherwise only the first element is being 420given tag on the given port (and return the port), or unregister it (when
206matched. 421C<$callback> is C<$undef> or missing). There can only be one callback
422registered for each tag.
207 423
208Any element in the match that is specified as C<_any_> (a function 424The original message will be passed to the callback, after the first
209exported by this module) matches any single element of the message. 425element (the tag) has been removed. The callback will use the same
426environment as the default callback (see above).
210 427
211While not required, it is highly recommended that the first matching 428Example: create a port and bind receivers on it in one go.
212element is a string identifying the message. The one-string-only match is 429
213also the most efficient match (by far). 430 my $port = rcv port,
431 msg1 => sub { ... },
432 msg2 => sub { ... },
433 ;
434
435Example: create a port, bind receivers and send it in a message elsewhere
436in one go:
437
438 snd $otherport, reply =>
439 rcv port,
440 msg1 => sub { ... },
441 ...
442 ;
443
444Example: temporarily register a rcv callback for a tag matching some port
445(e.g. for an rpc reply) and unregister it after a message was received.
446
447 rcv $port, $otherport => sub {
448 my @reply = @_;
449
450 rcv $SELF, $otherport;
451 };
214 452
215=cut 453=cut
216 454
217sub rcv($@) { 455sub rcv($@) {
218 my ($port, $match, $cb) = @_; 456 my $port = shift;
219
220 my $port = $PORT{$port}
221 or do {
222 my ($noderef, $lport) = split /#/, $port; 457 my ($nodeid, $portid) = split /#/, $port, 2;
223 "AnyEvent::MP::Node::Self" eq ref $NODE{$noderef} 458
459 $NODE{$nodeid} == $NODE{""}
224 or Carp::croak "$port: can only rcv on local ports"; 460 or Carp::croak "$port: rcv can only be called on local ports, caught";
225 461
226 $PORT{$lport} 462 while (@_) {
227 or Carp::croak "$port: port does not exist"; 463 if (ref $_[0]) {
228 464 if (my $self = $PORT_DATA{$portid}) {
229 $PORT{$port} = $PORT{$lport} # also return 465 "AnyEvent::MP::Port" eq ref $self
230 }; 466 or Carp::croak "$port: rcv can only be called on message matching ports, caught";
231 467
232 if (!ref $match) { 468 $self->[0] = shift;
233 push @{ $port->{rc0}{$match} }, [$cb];
234 } elsif (("ARRAY" eq ref $match && !ref $match->[0])) {
235 my ($type, @match) = @$match;
236 @match
237 ? push @{ $port->{rcv}{$match->[0]} }, [$cb, \@match]
238 : push @{ $port->{rc0}{$match->[0]} }, [$cb];
239 } else { 469 } else {
240 push @{ $port->{any} }, [$cb, $match]; 470 my $cb = shift;
241 } 471 $PORT{$portid} = sub {
242} 472 local $SELF = $port;
243 473 eval { &$cb }; _self_die if $@;
244sub _inject {
245 my ($port, $msg) = @{+shift};
246
247 $port = $PORT{$port}
248 or return;
249
250 @_ = @$msg;
251
252 for (@{ $port->{rc0}{$msg->[0]} }) {
253 $_ && &{$_->[0]}
254 && undef $_;
255 }
256
257 for (@{ $port->{rcv}{$msg->[0]} }) {
258 $_ && [@_[1..$#{$_->[1]}]] ~~ $_->[1]
259 && &{$_->[0]}
260 && undef $_;
261 }
262
263 for (@{ $port->{any} }) {
264 $_ && [@_[0..$#{$_->[1]}]] ~~ $_->[1]
265 && &{$_->[0]}
266 && undef $_;
267 }
268}
269
270sub normalise_noderef($) {
271 my ($noderef) = @_;
272
273 my $cv = AE::cv;
274 my @res;
275
276 $cv->begin (sub {
277 my %seen;
278 my @refs;
279 for (sort { $a->[0] <=> $b->[0] } @res) {
280 push @refs, $_->[1] unless $seen{$_->[1]}++
281 }
282 shift->send (join ",", @refs);
283 });
284
285 $noderef = $DEFAULT_PORT unless length $noderef;
286
287 my $idx;
288 for my $t (split /,/, $noderef) {
289 my $pri = ++$idx;
290
291 #TODO: this should be outside normalise_noderef and in become_public
292 if ($t =~ /^\d*$/) {
293 my $nodename = (POSIX::uname)[1];
294
295 $cv->begin;
296 AnyEvent::Socket::resolve_sockaddr $nodename, $t || "aemp=$DEFAULT_PORT", "tcp", 0, undef, sub {
297 for (@_) {
298 my ($service, $host) = AnyEvent::Socket::unpack_sockaddr $_->[3];
299 push @res, [
300 $pri += 1e-5,
301 AnyEvent::Socket::format_hostport AnyEvent::Socket::format_address $host, $service
302 ];
303 } 474 };
304 $cv->end; 475 }
476 } elsif (defined $_[0]) {
477 my $self = $PORT_DATA{$portid} ||= do {
478 my $self = bless [$PORT{$portid} || sub { }, { }, $port], "AnyEvent::MP::Port";
479
480 $PORT{$portid} = sub {
481 local $SELF = $port;
482
483 if (my $cb = $self->[1]{$_[0]}) {
484 shift;
485 eval { &$cb }; _self_die if $@;
486 } else {
487 &{ $self->[0] };
488 }
489 };
490
491 $self
305 }; 492 };
306 493
307# my (undef, undef, undef, undef, @ipv4) = gethostbyname $nodename; 494 "AnyEvent::MP::Port" eq ref $self
308# 495 or Carp::croak "$port: rcv can only be called on message matching ports, caught";
309# for (@ipv4) { 496
310# push @res, [ 497 my ($tag, $cb) = splice @_, 0, 2;
311# $pri, 498
312# AnyEvent::Socket::format_hostport AnyEvent::Socket::format_address $_, $t || $DEFAULT_PORT, 499 if (defined $cb) {
313# ]; 500 $self->[1]{$tag} = $cb;
314# }
315 } else { 501 } else {
316 my ($host, $port) = AnyEvent::Socket::parse_hostport $t, "aemp=$DEFAULT_PORT" 502 delete $self->[1]{$tag};
317 or Carp::croak "$t: unparsable transport descriptor";
318
319 $cv->begin;
320 AnyEvent::Socket::resolve_sockaddr $host, $port, "tcp", 0, undef, sub {
321 for (@_) {
322 my ($service, $host) = AnyEvent::Socket::unpack_sockaddr $_->[3];
323 push @res, [
324 $pri += 1e-5,
325 AnyEvent::Socket::format_hostport AnyEvent::Socket::format_address $host, $service
326 ];
327 }
328 $cv->end;
329 } 503 }
330 } 504 }
331 } 505 }
332 506
333 $cv->end; 507 $port
334
335 $cv
336} 508}
337 509
338sub become_public { 510=item peval $port, $coderef[, @args]
339 return if $PUBLIC;
340 511
341 my $noderef = join ",", ref $_[0] ? @{+shift} : shift; 512Evaluates the given C<$codref> within the contetx of C<$port>, that is,
342 my @args = @_; 513when the code throews an exception the C<$port> will be killed.
343 514
344 $NODE = (normalise_noderef $noderef)->recv; 515Any remaining args will be passed to the callback. Any return values will
516be returned to the caller.
345 517
346 for my $t (split /,/, $NODE) { 518This is useful when you temporarily want to execute code in the context of
347 $NODE{$t} = $NODE{""}; 519a port.
348 520
349 my ($host, $port) = AnyEvent::Socket::parse_hostport $t; 521Example: create a port and run some initialisation code in it's context.
350 522
351 $LISTENER{$t} = AnyEvent::MP::Transport::mp_server $host, $port, 523 my $port = port { ... };
352 @args,
353 on_error => sub {
354 die "on_error<@_>\n";#d#
355 },
356 on_connect => sub {
357 my ($tp) = @_;
358 524
359 $NODE{$tp->{remote_id}} = $_[0]; 525 peval $port, sub {
360 }, 526 init
361 sub { 527 or die "unable to init";
362 my ($tp) = @_; 528 };
363 529
364 $NODE{"$tp->{peerhost}:$tp->{peerport}"} = $tp; 530=cut
365 }, 531
366 ; 532sub peval($$) {
533 local $SELF = shift;
534 my $cb = shift;
535
536 if (wantarray) {
537 my @res = eval { &$cb };
538 _self_die if $@;
539 @res
540 } else {
541 my $res = eval { &$cb };
542 _self_die if $@;
543 $res
367 } 544 }
368
369 $PUBLIC = 1;
370} 545}
371 546
372=back 547=item $closure = psub { BLOCK }
373 548
374=head1 NODE MESSAGES 549Remembers C<$SELF> and creates a closure out of the BLOCK. When the
550closure is executed, sets up the environment in the same way as in C<rcv>
551callbacks, i.e. runtime errors will cause the port to get C<kil>ed.
375 552
376Nodes understand the following messages sent to them. Many of them take 553The effect is basically as if it returned C<< sub { peval $SELF, sub {
377arguments called C<@reply>, which will simply be used to compose a reply 554BLOCK }, @_ } >>.
378message - C<$reply[0]> is the port to reply to, C<$reply[1]> the type and
379the remaining arguments are simply the message data.
380 555
381=over 4 556This is useful when you register callbacks from C<rcv> callbacks:
382 557
383=cut 558 rcv delayed_reply => sub {
559 my ($delay, @reply) = @_;
560 my $timer = AE::timer $delay, 0, psub {
561 snd @reply, $SELF;
562 };
563 };
384 564
385############################################################################# 565=cut
386# self node code
387 566
388sub _new_port($) { 567sub psub(&) {
389 my ($name) = @_; 568 my $cb = shift;
390 569
570 my $port = $SELF
571 or Carp::croak "psub can only be called from within rcv or psub callbacks, not";
572
573 sub {
574 local $SELF = $port;
575
576 if (wantarray) {
577 my @res = eval { &$cb };
578 _self_die if $@;
579 @res
580 } else {
581 my $res = eval { &$cb };
582 _self_die if $@;
583 $res
584 }
585 }
586}
587
588=item $guard = mon $port, $cb->(@reason) # call $cb when $port dies
589
590=item $guard = mon $port, $rcvport # kill $rcvport when $port dies
591
592=item $guard = mon $port # kill $SELF when $port dies
593
594=item $guard = mon $port, $rcvport, @msg # send a message when $port dies
595
596Monitor the given port and do something when the port is killed or
597messages to it were lost, and optionally return a guard that can be used
598to stop monitoring again.
599
600In the first form (callback), the callback is simply called with any
601number of C<@reason> elements (no @reason means that the port was deleted
602"normally"). Note also that I<< the callback B<must> never die >>, so use
603C<eval> if unsure.
604
605In the second form (another port given), the other port (C<$rcvport>)
606will be C<kil>'ed with C<@reason>, if a @reason was specified, i.e. on
607"normal" kils nothing happens, while under all other conditions, the other
608port is killed with the same reason.
609
610The third form (kill self) is the same as the second form, except that
611C<$rvport> defaults to C<$SELF>.
612
613In the last form (message), a message of the form C<@msg, @reason> will be
614C<snd>.
615
616Monitoring-actions are one-shot: once messages are lost (and a monitoring
617alert was raised), they are removed and will not trigger again.
618
619As a rule of thumb, monitoring requests should always monitor a port from
620a local port (or callback). The reason is that kill messages might get
621lost, just like any other message. Another less obvious reason is that
622even monitoring requests can get lost (for example, when the connection
623to the other node goes down permanently). When monitoring a port locally
624these problems do not exist.
625
626C<mon> effectively guarantees that, in the absence of hardware failures,
627after starting the monitor, either all messages sent to the port will
628arrive, or the monitoring action will be invoked after possible message
629loss has been detected. No messages will be lost "in between" (after
630the first lost message no further messages will be received by the
631port). After the monitoring action was invoked, further messages might get
632delivered again.
633
634Inter-host-connection timeouts and monitoring depend on the transport
635used. The only transport currently implemented is TCP, and AnyEvent::MP
636relies on TCP to detect node-downs (this can take 10-15 minutes on a
637non-idle connection, and usually around two hours for idle connections).
638
639This means that monitoring is good for program errors and cleaning up
640stuff eventually, but they are no replacement for a timeout when you need
641to ensure some maximum latency.
642
643Example: call a given callback when C<$port> is killed.
644
645 mon $port, sub { warn "port died because of <@_>\n" };
646
647Example: kill ourselves when C<$port> is killed abnormally.
648
649 mon $port;
650
651Example: send us a restart message when another C<$port> is killed.
652
653 mon $port, $self => "restart";
654
655=cut
656
657sub mon {
391 my ($noderef, $portname) = split /#/, $name; 658 my ($nodeid, $port) = split /#/, shift, 2;
392 659
393 $PORT{$name} = 660 my $node = $NODE{$nodeid} || add_node $nodeid;
394 $PORT{$portname} = { 661
395 names => [$name, $portname], 662 my $cb = @_ ? shift : $SELF || Carp::croak 'mon: called with one argument only, but $SELF not set,';
663
664 unless (ref $cb) {
665 if (@_) {
666 # send a kill info message
667 my (@msg) = ($cb, @_);
668 $cb = sub { snd @msg, @_ };
669 } else {
670 # simply kill other port
671 my $port = $cb;
672 $cb = sub { kil $port, @_ if @_ };
673 }
674 }
675
676 $node->monitor ($port, $cb);
677
678 defined wantarray
679 and ($cb += 0, Guard::guard { $node->unmonitor ($port, $cb) })
680}
681
682=item $guard = mon_guard $port, $ref, $ref...
683
684Monitors the given C<$port> and keeps the passed references. When the port
685is killed, the references will be freed.
686
687Optionally returns a guard that will stop the monitoring.
688
689This function is useful when you create e.g. timers or other watchers and
690want to free them when the port gets killed (note the use of C<psub>):
691
692 $port->rcv (start => sub {
693 my $timer; $timer = mon_guard $port, AE::timer 1, 1, psub {
694 undef $timer if 0.9 < rand;
695 });
696 });
697
698=cut
699
700sub mon_guard {
701 my ($port, @refs) = @_;
702
703 #TODO: mon-less form?
704
705 mon $port, sub { 0 && @refs }
706}
707
708=item kil $port[, @reason]
709
710Kill the specified port with the given C<@reason>.
711
712If no C<@reason> is specified, then the port is killed "normally" -
713monitor callback will be invoked, but the kil will not cause linked ports
714(C<mon $mport, $lport> form) to get killed.
715
716If a C<@reason> is specified, then linked ports (C<mon $mport, $lport>
717form) get killed with the same reason.
718
719Runtime errors while evaluating C<rcv> callbacks or inside C<psub> blocks
720will be reported as reason C<< die => $@ >>.
721
722Transport/communication errors are reported as C<< transport_error =>
723$message >>.
724
725=cut
726
727=item $port = spawn $node, $initfunc[, @initdata]
728
729Creates a port on the node C<$node> (which can also be a port ID, in which
730case it's the node where that port resides).
731
732The port ID of the newly created port is returned immediately, and it is
733possible to immediately start sending messages or to monitor the port.
734
735After the port has been created, the init function is called on the remote
736node, in the same context as a C<rcv> callback. This function must be a
737fully-qualified function name (e.g. C<MyApp::Chat::Server::init>). To
738specify a function in the main program, use C<::name>.
739
740If the function doesn't exist, then the node tries to C<require>
741the package, then the package above the package and so on (e.g.
742C<MyApp::Chat::Server>, C<MyApp::Chat>, C<MyApp>) until the function
743exists or it runs out of package names.
744
745The init function is then called with the newly-created port as context
746object (C<$SELF>) and the C<@initdata> values as arguments. It I<must>
747call one of the C<rcv> functions to set callbacks on C<$SELF>, otherwise
748the port might not get created.
749
750A common idiom is to pass a local port, immediately monitor the spawned
751port, and in the remote init function, immediately monitor the passed
752local port. This two-way monitoring ensures that both ports get cleaned up
753when there is a problem.
754
755C<spawn> guarantees that the C<$initfunc> has no visible effects on the
756caller before C<spawn> returns (by delaying invocation when spawn is
757called for the local node).
758
759Example: spawn a chat server port on C<$othernode>.
760
761 # this node, executed from within a port context:
762 my $server = spawn $othernode, "MyApp::Chat::Server::connect", $SELF;
763 mon $server;
764
765 # init function on C<$othernode>
766 sub connect {
767 my ($srcport) = @_;
768
769 mon $srcport;
770
771 rcv $SELF, sub {
772 ...
773 };
774 }
775
776=cut
777
778sub _spawn {
779 my $port = shift;
780 my $init = shift;
781
782 # rcv will create the actual port
783 local $SELF = "$NODE#$port";
784 eval {
785 &{ load_func $init }
786 };
787 _self_die if $@;
788}
789
790sub spawn(@) {
791 my ($nodeid, undef) = split /#/, shift, 2;
792
793 my $id = $RUNIQ . ++$ID;
794
795 $_[0] =~ /::/
796 or Carp::croak "spawn init function must be a fully-qualified name, caught";
797
798 snd_to_func $nodeid, "AnyEvent::MP::_spawn" => $id, @_;
799
800 "$nodeid#$id"
801}
802
803
804=item after $timeout, @msg
805
806=item after $timeout, $callback
807
808Either sends the given message, or call the given callback, after the
809specified number of seconds.
810
811This is simply a utility function that comes in handy at times - the
812AnyEvent::MP author is not convinced of the wisdom of having it, though,
813so it may go away in the future.
814
815=cut
816
817sub after($@) {
818 my ($timeout, @action) = @_;
819
820 my $t; $t = AE::timer $timeout, 0, sub {
821 undef $t;
822 ref $action[0]
823 ? $action[0]()
824 : snd @action;
396 }; 825 };
397} 826}
398 827
399$NODE{""} = new AnyEvent::MP::Node::Self noderef => $NODE; 828=item cal $port, @msg, $callback[, $timeout]
400_new_port "";
401 829
402=item devnull => ... 830A simple form of RPC - sends a message to the given C<$port> with the
831given contents (C<@msg>), but adds a reply port to the message.
403 832
404Generic data sink/CPU heat conversion. 833The reply port is created temporarily just for the purpose of receiving
834the reply, and will be C<kil>ed when no longer needed.
405 835
406=cut 836A reply message sent to the port is passed to the C<$callback> as-is.
407 837
408rcv "", devnull => sub { () }; 838If an optional time-out (in seconds) is given and it is not C<undef>,
839then the callback will be called without any arguments after the time-out
840elapsed and the port is C<kil>ed.
409 841
410=item relay => $port, @msg 842If no time-out is given (or it is C<undef>), then the local port will
843monitor the remote port instead, so it eventually gets cleaned-up.
411 844
412Simply forwards the message to the given port. 845Currently this function returns the temporary port, but this "feature"
846might go in future versions unless you can make a convincing case that
847this is indeed useful for something.
413 848
414=cut 849=cut
415 850
416rcv "", relay => sub { \&snd; () }; 851sub cal(@) {
852 my $timeout = ref $_[-1] ? undef : pop;
853 my $cb = pop;
417 854
418=item eval => $string[ @reply] 855 my $port = port {
856 undef $timeout;
857 kil $SELF;
858 &$cb;
859 };
419 860
420Evaluates the given string. If C<@reply> is given, then a message of the 861 if (defined $timeout) {
421form C<@reply, $@, @evalres> is sent. 862 $timeout = AE::timer $timeout, 0, sub {
863 undef $timeout;
864 kil $port;
865 $cb->();
866 };
867 } else {
868 mon $_[0], sub {
869 kil $port;
870 $cb->();
871 };
872 }
422 873
423Example: crash another node. 874 push @_, $port;
875 &snd;
424 876
425 snd $othernode, eval => "exit"; 877 $port
426 878}
427=cut
428
429rcv "", eval => sub {
430 my (undef, $string, @reply) = @_;
431 my @res = eval $string;
432 snd @reply, "$@", @res if @reply;
433 ()
434};
435
436=item time => @reply
437
438Replies the the current node time to C<@reply>.
439
440Example: tell the current node to send the current time to C<$myport> in a
441C<timereply> message.
442
443 snd $NODE, time => $myport, timereply => 1, 2;
444 # => snd $myport, timereply => 1, 2, <time>
445
446=cut
447
448rcv "", time => sub { shift; snd @_, AE::time; () };
449 879
450=back 880=back
451 881
882=head1 DISTRIBUTED DATABASE
883
884AnyEvent::MP comes with a simple distributed database. The database will
885be mirrored asynchronously at all global nodes. Other nodes bind to one of
886the global nodes for their needs.
887
888The database consists of a two-level hash - a hash contains a hash which
889contains values.
890
891The top level hash key is called "family", and the second-level hash key
892is simply called "key".
893
894The family must be alphanumeric, i.e. start with a letter and consist
895of letters, digits, underscores and colons (C<[A-Za-z][A-Za-z0-9_:]*>,
896pretty much like Perl module names.
897
898As the family namespace is global, it is recommended to prefix family names
899with the name of the application or module using it.
900
901The keys must be strings, with no other limitations.
902
903The values should preferably be strings, but other perl scalars should
904work as well (such as undef, arrays and hashes).
905
906Every database entry is owned by one node - adding the same family/key
907combination on multiple nodes will not cause discomfort for AnyEvent::MP,
908but the result might be nondeterministic, i.e. the key might have
909different values on different nodes.
910
911=item db_set $family => $key => $value
912
913Sets (or replaces) a key to the database.
914
915=item db_del $family => $key
916
917Deletes a key from the database.
918
919=item $guard = db_reg $family => $key [=> $value]
920
921Sets the key on the database and returns a guard. When the guard is
922destroyed, the key is deleted from the database. If C<$value> is missing,
923then C<undef> is used.
924
925=cut
926
927=back
928
929=head1 AnyEvent::MP vs. Distributed Erlang
930
931AnyEvent::MP got lots of its ideas from distributed Erlang (Erlang node
932== aemp node, Erlang process == aemp port), so many of the documents and
933programming techniques employed by Erlang apply to AnyEvent::MP. Here is a
934sample:
935
936 http://www.erlang.se/doc/programming_rules.shtml
937 http://erlang.org/doc/getting_started/part_frame.html # chapters 3 and 4
938 http://erlang.org/download/erlang-book-part1.pdf # chapters 5 and 6
939 http://erlang.org/download/armstrong_thesis_2003.pdf # chapters 4 and 5
940
941Despite the similarities, there are also some important differences:
942
943=over 4
944
945=item * Node IDs are arbitrary strings in AEMP.
946
947Erlang relies on special naming and DNS to work everywhere in the same
948way. AEMP relies on each node somehow knowing its own address(es) (e.g. by
949configuration or DNS), and possibly the addresses of some seed nodes, but
950will otherwise discover other nodes (and their IDs) itself.
951
952=item * Erlang has a "remote ports are like local ports" philosophy, AEMP
953uses "local ports are like remote ports".
954
955The failure modes for local ports are quite different (runtime errors
956only) then for remote ports - when a local port dies, you I<know> it dies,
957when a connection to another node dies, you know nothing about the other
958port.
959
960Erlang pretends remote ports are as reliable as local ports, even when
961they are not.
962
963AEMP encourages a "treat remote ports differently" philosophy, with local
964ports being the special case/exception, where transport errors cannot
965occur.
966
967=item * Erlang uses processes and a mailbox, AEMP does not queue.
968
969Erlang uses processes that selectively receive messages out of order, and
970therefore needs a queue. AEMP is event based, queuing messages would serve
971no useful purpose. For the same reason the pattern-matching abilities
972of AnyEvent::MP are more limited, as there is little need to be able to
973filter messages without dequeuing them.
974
975This is not a philosophical difference, but simply stems from AnyEvent::MP
976being event-based, while Erlang is process-based.
977
978You cna have a look at L<Coro::MP> for a more Erlang-like process model on
979top of AEMP and Coro threads.
980
981=item * Erlang sends are synchronous, AEMP sends are asynchronous.
982
983Sending messages in Erlang is synchronous and blocks the process until
984a conenction has been established and the message sent (and so does not
985need a queue that can overflow). AEMP sends return immediately, connection
986establishment is handled in the background.
987
988=item * Erlang suffers from silent message loss, AEMP does not.
989
990Erlang implements few guarantees on messages delivery - messages can get
991lost without any of the processes realising it (i.e. you send messages a,
992b, and c, and the other side only receives messages a and c).
993
994AEMP guarantees (modulo hardware errors) correct ordering, and the
995guarantee that after one message is lost, all following ones sent to the
996same port are lost as well, until monitoring raises an error, so there are
997no silent "holes" in the message sequence.
998
999If you want your software to be very reliable, you have to cope with
1000corrupted and even out-of-order messages in both Erlang and AEMP. AEMP
1001simply tries to work better in common error cases, such as when a network
1002link goes down.
1003
1004=item * Erlang can send messages to the wrong port, AEMP does not.
1005
1006In Erlang it is quite likely that a node that restarts reuses an Erlang
1007process ID known to other nodes for a completely different process,
1008causing messages destined for that process to end up in an unrelated
1009process.
1010
1011AEMP does not reuse port IDs, so old messages or old port IDs floating
1012around in the network will not be sent to an unrelated port.
1013
1014=item * Erlang uses unprotected connections, AEMP uses secure
1015authentication and can use TLS.
1016
1017AEMP can use a proven protocol - TLS - to protect connections and
1018securely authenticate nodes.
1019
1020=item * The AEMP protocol is optimised for both text-based and binary
1021communications.
1022
1023The AEMP protocol, unlike the Erlang protocol, supports both programming
1024language independent text-only protocols (good for debugging), and binary,
1025language-specific serialisers (e.g. Storable). By default, unless TLS is
1026used, the protocol is actually completely text-based.
1027
1028It has also been carefully designed to be implementable in other languages
1029with a minimum of work while gracefully degrading functionality to make the
1030protocol simple.
1031
1032=item * AEMP has more flexible monitoring options than Erlang.
1033
1034In Erlang, you can chose to receive I<all> exit signals as messages or
1035I<none>, there is no in-between, so monitoring single Erlang processes is
1036difficult to implement.
1037
1038Monitoring in AEMP is more flexible than in Erlang, as one can choose
1039between automatic kill, exit message or callback on a per-port basis.
1040
1041=item * Erlang tries to hide remote/local connections, AEMP does not.
1042
1043Monitoring in Erlang is not an indicator of process death/crashes, in the
1044same way as linking is (except linking is unreliable in Erlang).
1045
1046In AEMP, you don't "look up" registered port names or send to named ports
1047that might or might not be persistent. Instead, you normally spawn a port
1048on the remote node. The init function monitors you, and you monitor the
1049remote port. Since both monitors are local to the node, they are much more
1050reliable (no need for C<spawn_link>).
1051
1052This also saves round-trips and avoids sending messages to the wrong port
1053(hard to do in Erlang).
1054
1055=back
1056
1057=head1 RATIONALE
1058
1059=over 4
1060
1061=item Why strings for port and node IDs, why not objects?
1062
1063We considered "objects", but found that the actual number of methods
1064that can be called are quite low. Since port and node IDs travel over
1065the network frequently, the serialising/deserialising would add lots of
1066overhead, as well as having to keep a proxy object everywhere.
1067
1068Strings can easily be printed, easily serialised etc. and need no special
1069procedures to be "valid".
1070
1071And as a result, a port with just a default receiver consists of a single
1072code reference stored in a global hash - it can't become much cheaper.
1073
1074=item Why favour JSON, why not a real serialising format such as Storable?
1075
1076In fact, any AnyEvent::MP node will happily accept Storable as framing
1077format, but currently there is no way to make a node use Storable by
1078default (although all nodes will accept it).
1079
1080The default framing protocol is JSON because a) JSON::XS is many times
1081faster for small messages and b) most importantly, after years of
1082experience we found that object serialisation is causing more problems
1083than it solves: Just like function calls, objects simply do not travel
1084easily over the network, mostly because they will always be a copy, so you
1085always have to re-think your design.
1086
1087Keeping your messages simple, concentrating on data structures rather than
1088objects, will keep your messages clean, tidy and efficient.
1089
1090=back
1091
452=head1 SEE ALSO 1092=head1 SEE ALSO
1093
1094L<AnyEvent::MP::Intro> - a gentle introduction.
1095
1096L<AnyEvent::MP::Kernel> - more, lower-level, stuff.
1097
1098L<AnyEvent::MP::Global> - network maintenance and port groups, to find
1099your applications.
1100
1101L<AnyEvent::MP::DataConn> - establish data connections between nodes.
1102
1103L<AnyEvent::MP::LogCatcher> - simple service to display log messages from
1104all nodes.
453 1105
454L<AnyEvent>. 1106L<AnyEvent>.
455 1107
456=head1 AUTHOR 1108=head1 AUTHOR
457 1109

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines