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 1.36 by root, Thu Aug 6 10:46:48 2009 UTC vs.
Revision 1.68 by root, Fri Aug 28 23:06:33 2009 UTC

8 8
9 $NODE # contains this node's noderef 9 $NODE # contains this node's noderef
10 NODE # returns this node's noderef 10 NODE # returns this node's noderef
11 NODE $port # returns the noderef of the port 11 NODE $port # returns the noderef of the port
12 12
13 $SELF # receiving/own port id in rcv callbacks
14
15 # initialise the node so it can send/receive messages
16 initialise_node;
17
18 # ports are message endpoints
19
20 # sending messages
13 snd $port, type => data...; 21 snd $port, type => data...;
22 snd $port, @msg;
23 snd @msg_with_first_element_being_a_port;
14 24
15 $SELF # receiving/own port id in rcv callbacks 25 # creating/using ports, the simple way
26 my $simple_port = port { my @msg = @_; 0 };
16 27
17 rcv $port, smartmatch => $cb->($port, @msg); 28 # creating/using ports, tagged message matching
18 29 my $port = port;
19 # examples:
20 rcv $port2, ping => sub { snd $_[0], "pong"; 0 }; 30 rcv $port, ping => sub { snd $_[0], "pong"; 0 };
21 rcv $port1, pong => sub { warn "pong received\n" }; 31 rcv $port, pong => sub { warn "pong received\n"; 0 };
22 snd $port2, ping => $port1;
23 32
24 # more, smarter, matches (_any_ is exported by this module) 33 # create a port on another node
25 rcv $port, [child_died => $pid] => sub { ... 34 my $port = spawn $node, $initfunc, @initdata;
26 rcv $port, [_any_, _any_, 3] => sub { .. $_[2] is 3
27
28 # linking two ports, so they both crash together
29 lnk $port1, $port2;
30 35
31 # monitoring 36 # monitoring
32 mon $port, $cb->(@msg) # callback is invoked on death 37 mon $port, $cb->(@msg) # callback is invoked on death
33 mon $port, $otherport # kill otherport on abnormal death 38 mon $port, $otherport # kill otherport on abnormal death
34 mon $port, $otherport, @msg # send message on death 39 mon $port, $otherport, @msg # send message on death
35 40
41=head1 CURRENT STATUS
42
43 AnyEvent::MP - stable API, should work
44 AnyEvent::MP::Intro - outdated
45 AnyEvent::MP::Kernel - mostly stable
46 AnyEvent::MP::Global - mostly stable
47 AnyEvent::MP::Node - mostly stable, but internal anyways
48 AnyEvent::MP::Transport - mostly stable, but internal anyways
49
50 stay tuned.
51
36=head1 DESCRIPTION 52=head1 DESCRIPTION
37 53
38This module (-family) implements a simple message passing framework. 54This module (-family) implements a simple message passing framework.
39 55
40Despite its simplicity, you can securely message other processes running 56Despite its simplicity, you can securely message other processes running
41on the same or other hosts. 57on the same or other hosts, and you can supervise entities remotely.
42 58
43For an introduction to this module family, see the L<AnyEvent::MP::Intro> 59For an introduction to this module family, see the L<AnyEvent::MP::Intro>
44manual page. 60manual page and the examples under F<eg/>.
45 61
46At the moment, this module family is severly broken and underdocumented, 62At the moment, this module family is a bit underdocumented.
47so do not use. This was uploaded mainly to reserve the CPAN namespace -
48stay tuned! The basic API should be finished, however.
49 63
50=head1 CONCEPTS 64=head1 CONCEPTS
51 65
52=over 4 66=over 4
53 67
54=item port 68=item port
55 69
56A port is something you can send messages to (with the C<snd> function). 70A port is something you can send messages to (with the C<snd> function).
57 71
58Some ports allow you to register C<rcv> handlers that can match specific 72Ports allow you to register C<rcv> handlers that can match all or just
59messages. All C<rcv> handlers will receive messages they match, messages 73some messages. Messages send to ports will not be queued, regardless of
60will not be queued. 74anything was listening for them or not.
61 75
62=item port id - C<noderef#portname> 76=item port ID - C<nodeid#portname>
63 77
64A port id is normaly the concatenation of a noderef, a hash-mark (C<#>) as 78A port ID is the concatenation of a node ID, a hash-mark (C<#>) as
65separator, and a port name (a printable string of unspecified format). An 79separator, and a port name (a printable string of unspecified format).
66exception is the the node port, whose ID is identical to its node
67reference.
68 80
69=item node 81=item node
70 82
71A node is a single process containing at least one port - the node 83A node is a single process containing at least one port - the node port,
72port. You can send messages to node ports to find existing ports or to 84which enables nodes to manage each other remotely, and to create new
73create new ports, among other things. 85ports.
74 86
75Nodes are either private (single-process only), slaves (connected to a 87Nodes are either public (have one or more listening ports) or private
76master node only) or public nodes (connectable from unrelated nodes). 88(no listening ports). Private nodes cannot talk to other private nodes
89currently.
77 90
78=item noderef - C<host:port,host:port...>, C<id@noderef>, C<id> 91=item node ID - C<[a-za-Z0-9_\-.:]+>
79 92
80A node reference is a string that either simply identifies the node (for 93A node ID is a string that uniquely identifies the node within a
81private and slave nodes), or contains a recipe on how to reach a given 94network. Depending on the configuration used, node IDs can look like a
82node (for public nodes). 95hostname, a hostname and a port, or a random string. AnyEvent::MP itself
96doesn't interpret node IDs in any way.
83 97
84This recipe is simply a comma-separated list of C<address:port> pairs (for 98=item binds - C<ip:port>
85TCP/IP, other protocols might look different).
86 99
87Node references come in two flavours: resolved (containing only numerical 100Nodes can only talk to each other by creating some kind of connection to
88addresses) or unresolved (where hostnames are used instead of addresses). 101each other. To do this, nodes should listen on one or more local transport
102endpoints - binds. Currently, only standard C<ip:port> specifications can
103be used, which specify TCP ports to listen on.
89 104
90Before using an unresolved node reference in a message you first have to 105=item seeds - C<host:port>
91resolve it. 106
107When a node starts, it knows nothing about the network. To teach the node
108about the network it first has to contact some other node within the
109network. This node is called a seed.
110
111Seeds are transport endpoint(s) of as many nodes as one wants. Those nodes
112are expected to be long-running, and at least one of those should always
113be available. When nodes run out of connections (e.g. due to a network
114error), they try to re-establish connections to some seednodes again to
115join the network.
116
117Apart from being sued for seeding, seednodes are not special in any way -
118every public node can be a seednode.
92 119
93=back 120=back
94 121
95=head1 VARIABLES/FUNCTIONS 122=head1 VARIABLES/FUNCTIONS
96 123
98 125
99=cut 126=cut
100 127
101package AnyEvent::MP; 128package AnyEvent::MP;
102 129
103use AnyEvent::MP::Base; 130use AnyEvent::MP::Kernel;
104 131
105use common::sense; 132use common::sense;
106 133
107use Carp (); 134use Carp ();
108 135
109use AE (); 136use AE ();
110 137
111use base "Exporter"; 138use base "Exporter";
112 139
113our $VERSION = '0.1'; 140our $VERSION = $AnyEvent::MP::Kernel::VERSION;
141
114our @EXPORT = qw( 142our @EXPORT = qw(
115 NODE $NODE *SELF node_of _any_ 143 NODE $NODE *SELF node_of after
116 resolve_node initialise_node 144 initialise_node
117 snd rcv mon kil reg psub 145 snd rcv mon mon_guard kil reg psub spawn
118 port 146 port
119); 147);
120 148
121our $SELF; 149our $SELF;
122 150
126 kil $SELF, die => $msg; 154 kil $SELF, die => $msg;
127} 155}
128 156
129=item $thisnode = NODE / $NODE 157=item $thisnode = NODE / $NODE
130 158
131The C<NODE> function returns, and the C<$NODE> variable contains 159The C<NODE> function returns, and the C<$NODE> variable contains, the node
132the noderef of the local node. The value is initialised by a call 160ID of the node running in the current process. This value is initialised by
133to C<become_public> or C<become_slave>, after which all local port 161a call to C<initialise_node>.
134identifiers become invalid.
135 162
136=item $noderef = node_of $port 163=item $nodeid = node_of $port
137 164
138Extracts and returns the noderef from a portid or a noderef. 165Extracts and returns the node ID from a port ID or a node ID.
139 166
140=item initialise_node $noderef, $seednode, $seednode... 167=item initialise_node $profile_name
141 168
142=item initialise_node "slave/", $master, $master...
143
144Before a node can talk to other nodes on the network it has to initialise 169Before a node can talk to other nodes on the network (i.e. enter
145itself - the minimum a node needs to know is it's own name, and optionally 170"distributed mode") it has to initialise itself - the minimum a node needs
146it should know the noderefs of some other nodes in the network. 171to know is its own name, and optionally it should know the addresses of
172some other nodes in the network to discover other nodes.
147 173
148This function initialises a node - it must be called exactly once (or 174This function initialises a node - it must be called exactly once (or
149never) before calling other AnyEvent::MP functions. 175never) before calling other AnyEvent::MP functions.
150 176
151All arguments are noderefs, which can be either resolved or unresolved. 177The first argument is a profile name. If it is C<undef> or missing, then
178the current nodename will be used instead (i.e. F<uname -n>).
152 179
153There are two types of networked nodes, public nodes and slave nodes: 180The function then looks up the profile in the aemp configuration (see the
181L<aemp> commandline utility).
154 182
155=over 4 183If the profile specifies a node ID, then this will become the node ID of
184this process. If not, then the profile name will be used as node ID. The
185special node ID of C<anon/> will be replaced by a random node ID.
156 186
157=item public nodes 187The next step is to look up the binds in the profile, followed by binding
188aemp protocol listeners on all binds specified (it is possible and valid
189to have no binds, meaning that the node cannot be contacted form the
190outside. This means the node cannot talk to other nodes that also have no
191binds, but it can still talk to all "normal" nodes).
158 192
159For public nodes, C<$noderef> must either be a (possibly unresolved) 193If the profile does not specify a binds list, then the node ID will be
160noderef, in which case it will be resolved, or C<undef> (or missing), in 194treated as if it were of the form C<host:port>, which will be resolved and
161which case the noderef will be guessed. 195used as binds list.
162 196
163Afterwards, the node will bind itself on all endpoints and try to connect 197Lastly, the seeds list from the profile is passed to the
164to all additional C<$seednodes> that are specified. Seednodes are optional 198L<AnyEvent::MP::Global> module, which will then use it to keep
165and can be used to quickly bootstrap the node into an existing network. 199connectivity with at least on of those seed nodes at any point in time.
166 200
167=item slave nodes 201Example: become a distributed node listening on the guessed noderef, or
168 202the one specified via C<aemp> for the current node. This should be the
169When the C<$noderef> is the special string C<slave/>, then the node will 203most common form of invocation for "daemon"-type nodes.
170become a slave node. Slave nodes cannot be contacted from outside and will
171route most of their traffic to the master node that they attach to.
172
173At least one additional noderef is required: The node will try to connect
174to all of them and will become a slave attached to the first node it can
175successfully connect to.
176
177=back
178
179This function will block until all nodes have been resolved and, for slave
180nodes, until it has successfully established a connection to a master
181server.
182
183Example: become a public node listening on the default node.
184 204
185 initialise_node; 205 initialise_node;
186 206
187Example: become a public node, and try to contact some well-known master 207Example: become an anonymous node. This form is often used for commandline
188servers to become part of the network. 208clients.
189 209
190 initialise_node undef, "master1", "master2";
191
192Example: become a public node listening on port C<4041>.
193
194 initialise_node 4041; 210 initialise_node "anon/";
195 211
196Example: become a public node, only visible on localhost port 4044. 212Example: become a distributed node. If there is no profile of the given
213name, or no binds list was specified, resolve C<localhost:4044> and bind
214on the resulting addresses.
197 215
198 initialise_node "locahost:4044"; 216 initialise_node "localhost:4044";
199
200Example: become a slave node to any of the specified master servers.
201
202 initialise_node "slave/", "master1", "192.168.13.17", "mp.example.net";
203
204=item $cv = resolve_node $noderef
205
206Takes an unresolved node reference that may contain hostnames and
207abbreviated IDs, resolves all of them and returns a resolved node
208reference.
209
210In addition to C<address:port> pairs allowed in resolved noderefs, the
211following forms are supported:
212
213=over 4
214
215=item the empty string
216
217An empty-string component gets resolved as if the default port (4040) was
218specified.
219
220=item naked port numbers (e.g. C<1234>)
221
222These are resolved by prepending the local nodename and a colon, to be
223further resolved.
224
225=item hostnames (e.g. C<localhost:1234>, C<localhost>)
226
227These are resolved by using AnyEvent::DNS to resolve them, optionally
228looking up SRV records for the C<aemp=4040> port, if no port was
229specified.
230
231=back
232 217
233=item $SELF 218=item $SELF
234 219
235Contains the current port id while executing C<rcv> callbacks or C<psub> 220Contains the current port id while executing C<rcv> callbacks or C<psub>
236blocks. 221blocks.
237 222
238=item SELF, %SELF, @SELF... 223=item *SELF, SELF, %SELF, @SELF...
239 224
240Due to some quirks in how perl exports variables, it is impossible to 225Due to some quirks in how perl exports variables, it is impossible to
241just export C<$SELF>, all the symbols called C<SELF> are exported by this 226just export C<$SELF>, all the symbols named C<SELF> are exported by this
242module, but only C<$SELF> is currently used. 227module, but only C<$SELF> is currently used.
243 228
244=item snd $port, type => @data 229=item snd $port, type => @data
245 230
246=item snd $port, @msg 231=item snd $port, @msg
247 232
248Send the given message to the given port ID, which can identify either 233Send the given message to the given port, which can identify either a
249a local or a remote port, and can be either a string or soemthignt hat 234local or a remote port, and must be a port ID.
250stringifies a sa port ID (such as a port object :).
251 235
252While the message can be about anything, it is highly recommended to use a 236While the message can be almost anything, it is highly recommended to
253string as first element (a portid, or some word that indicates a request 237use a string as first element (a port ID, or some word that indicates a
254type etc.). 238request type etc.) and to consist if only simple perl values (scalars,
239arrays, hashes) - if you think you need to pass an object, think again.
255 240
256The message data effectively becomes read-only after a call to this 241The message data logically becomes read-only after a call to this
257function: modifying any argument is not allowed and can cause many 242function: modifying any argument (or values referenced by them) is
258problems. 243forbidden, as there can be considerable time between the call to C<snd>
244and the time the message is actually being serialised - in fact, it might
245never be copied as within the same process it is simply handed to the
246receiving port.
259 247
260The type of data you can transfer depends on the transport protocol: when 248The type of data you can transfer depends on the transport protocol: when
261JSON is used, then only strings, numbers and arrays and hashes consisting 249JSON is used, then only strings, numbers and arrays and hashes consisting
262of those are allowed (no objects). When Storable is used, then anything 250of those are allowed (no objects). When Storable is used, then anything
263that Storable can serialise and deserialise is allowed, and for the local 251that Storable can serialise and deserialise is allowed, and for the local
264node, anything can be passed. 252node, anything can be passed. Best rely only on the common denominator of
253these.
265 254
266=item $local_port = port 255=item $local_port = port
267 256
268Create a new local port object that can be used either as a pattern 257Create a new local port object and returns its port ID. Initially it has
269matching port ("full port") or a single-callback port ("miniport"), 258no callbacks set and will throw an error when it receives messages.
270depending on how C<rcv> callbacks are bound to the object.
271 259
272=item $port = port { my @msg = @_; $finished } 260=item $local_port = port { my @msg = @_ }
273 261
274Creates a "miniport", that is, a very lightweight port without any pattern 262Creates a new local port, and returns its ID. Semantically the same as
275matching behind it, and returns its ID. Semantically the same as creating
276a port and calling C<rcv $port, $callback> on it. 263creating a port and calling C<rcv $port, $callback> on it.
277 264
278The block will be called for every message received on the port. When the 265The block will be called for every message received on the port, with the
279callback returns a true value its job is considered "done" and the port 266global variable C<$SELF> set to the port ID. Runtime errors will cause the
280will be destroyed. Otherwise it will stay alive. 267port to be C<kil>ed. The message will be passed as-is, no extra argument
268(i.e. no port ID) will be passed to the callback.
281 269
282The message will be passed as-is, no extra argument (i.e. no port id) will 270If you want to stop/destroy the port, simply C<kil> it:
283be passed to the callback.
284 271
285If you need the local port id in the callback, this works nicely: 272 my $port = port {
286 273 my @msg = @_;
287 my $port; $port = port { 274 ...
288 snd $otherport, reply => $port; 275 kil $SELF;
289 }; 276 };
290 277
291=cut 278=cut
292 279
293sub rcv($@); 280sub rcv($@);
281
282sub _kilme {
283 die "received message on port without callback";
284}
294 285
295sub port(;&) { 286sub port(;&) {
296 my $id = "$UNIQ." . $ID++; 287 my $id = "$UNIQ." . $ID++;
297 my $port = "$NODE#$id"; 288 my $port = "$NODE#$id";
298 289
299 if (@_) { 290 rcv $port, shift || \&_kilme;
300 rcv $port, shift;
301 } else {
302 $PORT{$id} = sub { }; # nop
303 }
304 291
305 $port 292 $port
306} 293}
307 294
308=item reg $port, $name
309
310=item reg $name
311
312Registers the given port (or C<$SELF><<< if missing) under the name
313C<$name>. If the name already exists it is replaced.
314
315A port can only be registered under one well known name.
316
317A port automatically becomes unregistered when it is killed.
318
319=cut
320
321sub reg(@) {
322 my $port = @_ > 1 ? shift : $SELF || Carp::croak 'reg: called with one argument only, but $SELF not set,';
323
324 $REG{$_[0]} = $port;
325}
326
327=item rcv $port, $callback->(@msg) 295=item rcv $local_port, $callback->(@msg)
328 296
329Replaces the callback on the specified miniport (after converting it to 297Replaces the default callback on the specified port. There is no way to
330one if required). 298remove the default callback: use C<sub { }> to disable it, or better
331 299C<kil> the port when it is no longer needed.
332=item rcv $port, tagstring => $callback->(@msg), ...
333
334=item rcv $port, $smartmatch => $callback->(@msg), ...
335
336=item rcv $port, [$smartmatch...] => $callback->(@msg), ...
337
338Register callbacks to be called on matching messages on the given full
339port (after converting it to one if required) and return the port.
340
341The callback has to return a true value when its work is done, after
342which is will be removed, or a false value in which case it will stay
343registered.
344 300
345The global C<$SELF> (exported by this module) contains C<$port> while 301The global C<$SELF> (exported by this module) contains C<$port> while
346executing the callback. 302executing the callback. Runtime errors during callback execution will
303result in the port being C<kil>ed.
347 304
348Runtime errors wdurign callback execution will result in the port being 305The default callback received all messages not matched by a more specific
349C<kil>ed. 306C<tag> match.
350 307
351If the match is an array reference, then it will be matched against the 308=item rcv $local_port, tag => $callback->(@msg_without_tag), ...
352first elements of the message, otherwise only the first element is being
353matched.
354 309
355Any element in the match that is specified as C<_any_> (a function 310Register (or replace) callbacks to be called on messages starting with the
356exported by this module) matches any single element of the message. 311given tag on the given port (and return the port), or unregister it (when
312C<$callback> is C<$undef> or missing). There can only be one callback
313registered for each tag.
357 314
358While not required, it is highly recommended that the first matching 315The original message will be passed to the callback, after the first
359element is a string identifying the message. The one-string-only match is 316element (the tag) has been removed. The callback will use the same
360also the most efficient match (by far). 317environment as the default callback (see above).
361 318
362Example: create a port and bind receivers on it in one go. 319Example: create a port and bind receivers on it in one go.
363 320
364 my $port = rcv port, 321 my $port = rcv port,
365 msg1 => sub { ...; 0 }, 322 msg1 => sub { ... },
366 msg2 => sub { ...; 0 }, 323 msg2 => sub { ... },
367 ; 324 ;
368 325
369Example: create a port, bind receivers and send it in a message elsewhere 326Example: create a port, bind receivers and send it in a message elsewhere
370in one go: 327in one go:
371 328
372 snd $otherport, reply => 329 snd $otherport, reply =>
373 rcv port, 330 rcv port,
374 msg1 => sub { ...; 0 }, 331 msg1 => sub { ... },
375 ... 332 ...
376 ; 333 ;
334
335Example: temporarily register a rcv callback for a tag matching some port
336(e.g. for a rpc reply) and unregister it after a message was received.
337
338 rcv $port, $otherport => sub {
339 my @reply = @_;
340
341 rcv $SELF, $otherport;
342 };
377 343
378=cut 344=cut
379 345
380sub rcv($@) { 346sub rcv($@) {
381 my $port = shift; 347 my $port = shift;
382 my ($noderef, $portid) = split /#/, $port, 2; 348 my ($noderef, $portid) = split /#/, $port, 2;
383 349
384 ($NODE{$noderef} || add_node $noderef) == $NODE{""} 350 $NODE{$noderef} == $NODE{""}
385 or Carp::croak "$port: rcv can only be called on local ports, caught"; 351 or Carp::croak "$port: rcv can only be called on local ports, caught";
386 352
387 if (@_ == 1) { 353 while (@_) {
354 if (ref $_[0]) {
355 if (my $self = $PORT_DATA{$portid}) {
356 "AnyEvent::MP::Port" eq ref $self
357 or Carp::croak "$port: rcv can only be called on message matching ports, caught";
358
359 $self->[2] = shift;
360 } else {
388 my $cb = shift; 361 my $cb = shift;
389 delete $PORT_DATA{$portid};
390 $PORT{$portid} = sub { 362 $PORT{$portid} = sub {
391 local $SELF = $port; 363 local $SELF = $port;
392 eval { 364 eval { &$cb }; _self_die if $@;
393 &$cb 365 };
394 and kil $port;
395 }; 366 }
396 _self_die if $@; 367 } elsif (defined $_[0]) {
397 };
398 } else {
399 my $self = $PORT_DATA{$portid} ||= do { 368 my $self = $PORT_DATA{$portid} ||= do {
400 my $self = bless { 369 my $self = bless [$PORT{$port} || sub { }, { }, $port], "AnyEvent::MP::Port";
401 id => $port,
402 }, "AnyEvent::MP::Port";
403 370
404 $PORT{$portid} = sub { 371 $PORT{$portid} = sub {
405 local $SELF = $port; 372 local $SELF = $port;
406 373
407 eval {
408 for (@{ $self->{rc0}{$_[0]} }) { 374 if (my $cb = $self->[1]{$_[0]}) {
409 $_ && &{$_->[0]} 375 shift;
410 && undef $_; 376 eval { &$cb }; _self_die if $@;
411 } 377 } else {
412
413 for (@{ $self->{rcv}{$_[0]} }) {
414 $_ && [@_[1 .. @{$_->[1]}]] ~~ $_->[1]
415 && &{$_->[0]} 378 &{ $self->[0] };
416 && undef $_;
417 }
418
419 for (@{ $self->{any} }) {
420 $_ && [@_[0 .. $#{$_->[1]}]] ~~ $_->[1]
421 && &{$_->[0]}
422 && undef $_;
423 } 379 }
424 }; 380 };
425 _self_die if $@; 381
382 $self
426 }; 383 };
427 384
428 $self
429 };
430
431 "AnyEvent::MP::Port" eq ref $self 385 "AnyEvent::MP::Port" eq ref $self
432 or Carp::croak "$port: rcv can only be called on message matching ports, caught"; 386 or Carp::croak "$port: rcv can only be called on message matching ports, caught";
433 387
434 while (@_) {
435 my ($match, $cb) = splice @_, 0, 2; 388 my ($tag, $cb) = splice @_, 0, 2;
436 389
437 if (!ref $match) { 390 if (defined $cb) {
438 push @{ $self->{rc0}{$match} }, [$cb]; 391 $self->[1]{$tag} = $cb;
439 } elsif (("ARRAY" eq ref $match && !ref $match->[0])) {
440 my ($type, @match) = @$match;
441 @match
442 ? push @{ $self->{rcv}{$match->[0]} }, [$cb, \@match]
443 : push @{ $self->{rc0}{$match->[0]} }, [$cb];
444 } else { 392 } else {
445 push @{ $self->{any} }, [$cb, $match]; 393 delete $self->[1]{$tag};
446 } 394 }
447 } 395 }
448 } 396 }
449 397
450 $port 398 $port
486 $res 434 $res
487 } 435 }
488 } 436 }
489} 437}
490 438
491=item $guard = mon $port, $cb->(@reason) 439=item $guard = mon $port, $cb->(@reason) # call $cb when $port dies
492 440
493=item $guard = mon $port, $rcvport 441=item $guard = mon $port, $rcvport # kill $rcvport when $port dies
494 442
495=item $guard = mon $port 443=item $guard = mon $port # kill $SELF when $port dies
496 444
497=item $guard = mon $port, $rcvport, @msg 445=item $guard = mon $port, $rcvport, @msg # send a message when $port dies
498 446
499Monitor the given port and do something when the port is killed, and 447Monitor the given port and do something when the port is killed or
500optionally return a guard that can be used to stop monitoring again. 448messages to it were lost, and optionally return a guard that can be used
449to stop monitoring again.
450
451C<mon> effectively guarantees that, in the absence of hardware failures,
452after starting the monitor, either all messages sent to the port will
453arrive, or the monitoring action will be invoked after possible message
454loss has been detected. No messages will be lost "in between" (after
455the first lost message no further messages will be received by the
456port). After the monitoring action was invoked, further messages might get
457delivered again.
458
459Note that monitoring-actions are one-shot: once messages are lost (and a
460monitoring alert was raised), they are removed and will not trigger again.
501 461
502In the first form (callback), the callback is simply called with any 462In the first form (callback), the callback is simply called with any
503number of C<@reason> elements (no @reason means that the port was deleted 463number of C<@reason> elements (no @reason means that the port was deleted
504"normally"). Note also that I<< the callback B<must> never die >>, so use 464"normally"). Note also that I<< the callback B<must> never die >>, so use
505C<eval> if unsure. 465C<eval> if unsure.
506 466
507In the second form (another port given), the other port (C<$rcvport) 467In the second form (another port given), the other port (C<$rcvport>)
508will be C<kil>'ed with C<@reason>, iff a @reason was specified, i.e. on 468will be C<kil>'ed with C<@reason>, iff a @reason was specified, i.e. on
509"normal" kils nothing happens, while under all other conditions, the other 469"normal" kils nothing happens, while under all other conditions, the other
510port is killed with the same reason. 470port is killed with the same reason.
511 471
512The third form (kill self) is the same as the second form, except that 472The third form (kill self) is the same as the second form, except that
513C<$rvport> defaults to C<$SELF>. 473C<$rvport> defaults to C<$SELF>.
514 474
515In the last form (message), a message of the form C<@msg, @reason> will be 475In the last form (message), a message of the form C<@msg, @reason> will be
516C<snd>. 476C<snd>.
517 477
478As a rule of thumb, monitoring requests should always monitor a port from
479a local port (or callback). The reason is that kill messages might get
480lost, just like any other message. Another less obvious reason is that
481even monitoring requests can get lost (for exmaple, when the connection
482to the other node goes down permanently). When monitoring a port locally
483these problems do not exist.
484
518Example: call a given callback when C<$port> is killed. 485Example: call a given callback when C<$port> is killed.
519 486
520 mon $port, sub { warn "port died because of <@_>\n" }; 487 mon $port, sub { warn "port died because of <@_>\n" };
521 488
522Example: kill ourselves when C<$port> is killed abnormally. 489Example: kill ourselves when C<$port> is killed abnormally.
532sub mon { 499sub mon {
533 my ($noderef, $port) = split /#/, shift, 2; 500 my ($noderef, $port) = split /#/, shift, 2;
534 501
535 my $node = $NODE{$noderef} || add_node $noderef; 502 my $node = $NODE{$noderef} || add_node $noderef;
536 503
537 my $cb = @_ ? $_[0] : $SELF || Carp::croak 'mon: called with one argument only, but $SELF not set,'; 504 my $cb = @_ ? shift : $SELF || Carp::croak 'mon: called with one argument only, but $SELF not set,';
538 505
539 unless (ref $cb) { 506 unless (ref $cb) {
540 if (@_) { 507 if (@_) {
541 # send a kill info message 508 # send a kill info message
542 my (@msg) = @_; 509 my (@msg) = ($cb, @_);
543 $cb = sub { snd @msg, @_ }; 510 $cb = sub { snd @msg, @_ };
544 } else { 511 } else {
545 # simply kill other port 512 # simply kill other port
546 my $port = $cb; 513 my $port = $cb;
547 $cb = sub { kil $port, @_ if @_ }; 514 $cb = sub { kil $port, @_ if @_ };
560is killed, the references will be freed. 527is killed, the references will be freed.
561 528
562Optionally returns a guard that will stop the monitoring. 529Optionally returns a guard that will stop the monitoring.
563 530
564This function is useful when you create e.g. timers or other watchers and 531This function is useful when you create e.g. timers or other watchers and
565want to free them when the port gets killed: 532want to free them when the port gets killed (note the use of C<psub>):
566 533
567 $port->rcv (start => sub { 534 $port->rcv (start => sub {
568 my $timer; $timer = mon_guard $port, AE::timer 1, 1, sub { 535 my $timer; $timer = mon_guard $port, AE::timer 1, 1, psub {
569 undef $timer if 0.9 < rand; 536 undef $timer if 0.9 < rand;
570 }); 537 });
571 }); 538 });
572 539
573=cut 540=cut
578 #TODO: mon-less form? 545 #TODO: mon-less form?
579 546
580 mon $port, sub { 0 && @refs } 547 mon $port, sub { 0 && @refs }
581} 548}
582 549
583=item lnk $port1, $port2
584
585=item lnk $otherport
586
587Link two ports. This is simply a shorthand for:
588
589 mon $port1, $port2;
590 mon $port2, $port1;
591
592It means that if either one is killed abnormally, the other one gets
593killed as well.
594
595The one-argument form assumes that one port is C<$SELF>.
596
597=cut
598
599sub lnk {
600 my $port1 = shift;
601 my $port2 = @_ ? shift : $SELF || Carp::croak 'lnk: called with one argument only, but $SELF not set,';
602
603 mon $port1, $port2;
604 mon $port2, $port1;
605}
606
607=item kil $port[, @reason] 550=item kil $port[, @reason]
608 551
609Kill the specified port with the given C<@reason>. 552Kill the specified port with the given C<@reason>.
610 553
611If no C<@reason> is specified, then the port is killed "normally" (linked 554If no C<@reason> is specified, then the port is killed "normally" (ports
612ports will not be kileld, or even notified). 555monitoring other ports will not necessarily die because a port dies
556"normally").
613 557
614Otherwise, linked ports get killed with the same reason (second form of 558Otherwise, linked ports get killed with the same reason (second form of
615C<mon>, see below). 559C<mon>, see above).
616 560
617Runtime errors while evaluating C<rcv> callbacks or inside C<psub> blocks 561Runtime errors while evaluating C<rcv> callbacks or inside C<psub> blocks
618will be reported as reason C<< die => $@ >>. 562will be reported as reason C<< die => $@ >>.
619 563
620Transport/communication errors are reported as C<< transport_error => 564Transport/communication errors are reported as C<< transport_error =>
621$message >>. 565$message >>.
622 566
623=back
624
625=head1 NODE MESSAGES
626
627Nodes understand the following messages sent to them. Many of them take
628arguments called C<@reply>, which will simply be used to compose a reply
629message - C<$reply[0]> is the port to reply to, C<$reply[1]> the type and
630the remaining arguments are simply the message data.
631
632While other messages exist, they are not public and subject to change.
633
634=over 4
635
636=cut 567=cut
637 568
638=item lookup => $name, @reply 569=item $port = spawn $node, $initfunc[, @initdata]
639 570
640Replies with the port ID of the specified well-known port, or C<undef>. 571Creates a port on the node C<$node> (which can also be a port ID, in which
572case it's the node where that port resides).
641 573
642=item devnull => ... 574The port ID of the newly created port is returned immediately, and it is
575possible to immediately start sending messages or to monitor the port.
643 576
644Generic data sink/CPU heat conversion. 577After the port has been created, the init function is called on the remote
578node, in the same context as a C<rcv> callback. This function must be a
579fully-qualified function name (e.g. C<MyApp::Chat::Server::init>). To
580specify a function in the main program, use C<::name>.
645 581
646=item relay => $port, @msg 582If the function doesn't exist, then the node tries to C<require>
583the package, then the package above the package and so on (e.g.
584C<MyApp::Chat::Server>, C<MyApp::Chat>, C<MyApp>) until the function
585exists or it runs out of package names.
647 586
648Simply forwards the message to the given port. 587The init function is then called with the newly-created port as context
588object (C<$SELF>) and the C<@initdata> values as arguments.
649 589
650=item eval => $string[ @reply] 590A common idiom is to pass a local port, immediately monitor the spawned
591port, and in the remote init function, immediately monitor the passed
592local port. This two-way monitoring ensures that both ports get cleaned up
593when there is a problem.
651 594
652Evaluates the given string. If C<@reply> is given, then a message of the 595Example: spawn a chat server port on C<$othernode>.
653form C<@reply, $@, @evalres> is sent.
654 596
655Example: crash another node. 597 # this node, executed from within a port context:
598 my $server = spawn $othernode, "MyApp::Chat::Server::connect", $SELF;
599 mon $server;
656 600
657 snd $othernode, eval => "exit"; 601 # init function on C<$othernode>
602 sub connect {
603 my ($srcport) = @_;
658 604
659=item time => @reply 605 mon $srcport;
660 606
661Replies the the current node time to C<@reply>. 607 rcv $SELF, sub {
608 ...
609 };
610 }
662 611
663Example: tell the current node to send the current time to C<$myport> in a 612=cut
664C<timereply> message.
665 613
666 snd $NODE, time => $myport, timereply => 1, 2; 614sub _spawn {
667 # => snd $myport, timereply => 1, 2, <time> 615 my $port = shift;
616 my $init = shift;
617
618 local $SELF = "$NODE#$port";
619 eval {
620 &{ load_func $init }
621 };
622 _self_die if $@;
623}
624
625sub spawn(@) {
626 my ($noderef, undef) = split /#/, shift, 2;
627
628 my $id = "$RUNIQ." . $ID++;
629
630 $_[0] =~ /::/
631 or Carp::croak "spawn init function must be a fully-qualified name, caught";
632
633 snd_to_func $noderef, "AnyEvent::MP::_spawn" => $id, @_;
634
635 "$noderef#$id"
636}
637
638=item after $timeout, @msg
639
640=item after $timeout, $callback
641
642Either sends the given message, or call the given callback, after the
643specified number of seconds.
644
645This is simply a utility function that comes in handy at times - the
646AnyEvent::MP author is not convinced of the wisdom of having it, though,
647so it may go away in the future.
648
649=cut
650
651sub after($@) {
652 my ($timeout, @action) = @_;
653
654 my $t; $t = AE::timer $timeout, 0, sub {
655 undef $t;
656 ref $action[0]
657 ? $action[0]()
658 : snd @action;
659 };
660}
668 661
669=back 662=back
670 663
671=head1 AnyEvent::MP vs. Distributed Erlang 664=head1 AnyEvent::MP vs. Distributed Erlang
672 665
682 675
683Despite the similarities, there are also some important differences: 676Despite the similarities, there are also some important differences:
684 677
685=over 4 678=over 4
686 679
687=item * Node references contain the recipe on how to contact them. 680=item * Node IDs are arbitrary strings in AEMP.
688 681
689Erlang relies on special naming and DNS to work everywhere in the 682Erlang relies on special naming and DNS to work everywhere in the same
690same way. AEMP relies on each node knowing it's own address(es), with 683way. AEMP relies on each node somehow knowing its own address(es) (e.g. by
691convenience functionality. 684configuraiton or DNS), but will otherwise discover other odes itself.
692 685
693This means that AEMP requires a less tightly controlled environment at the 686=item * Erlang has a "remote ports are like local ports" philosophy, AEMP
694cost of longer node references and a slightly higher management overhead. 687uses "local ports are like remote ports".
688
689The failure modes for local ports are quite different (runtime errors
690only) then for remote ports - when a local port dies, you I<know> it dies,
691when a connection to another node dies, you know nothing about the other
692port.
693
694Erlang pretends remote ports are as reliable as local ports, even when
695they are not.
696
697AEMP encourages a "treat remote ports differently" philosophy, with local
698ports being the special case/exception, where transport errors cannot
699occur.
695 700
696=item * Erlang uses processes and a mailbox, AEMP does not queue. 701=item * Erlang uses processes and a mailbox, AEMP does not queue.
697 702
698Erlang uses processes that selctively receive messages, and therefore 703Erlang uses processes that selectively receive messages, and therefore
699needs a queue. AEMP is event based, queuing messages would serve no useful 704needs a queue. AEMP is event based, queuing messages would serve no
700purpose. 705useful purpose. For the same reason the pattern-matching abilities of
706AnyEvent::MP are more limited, as there is little need to be able to
707filter messages without dequeing them.
701 708
702(But see L<Coro::MP> for a more Erlang-like process model on top of AEMP). 709(But see L<Coro::MP> for a more Erlang-like process model on top of AEMP).
703 710
704=item * Erlang sends are synchronous, AEMP sends are asynchronous. 711=item * Erlang sends are synchronous, AEMP sends are asynchronous.
705 712
706Sending messages in Erlang is synchronous and blocks the process. AEMP 713Sending messages in Erlang is synchronous and blocks the process (and
707sends are immediate, connection establishment is handled in the 714so does not need a queue that can overflow). AEMP sends are immediate,
708background. 715connection establishment is handled in the background.
709 716
710=item * Erlang can silently lose messages, AEMP cannot. 717=item * Erlang suffers from silent message loss, AEMP does not.
711 718
712Erlang makes few guarantees on messages delivery - messages can get lost 719Erlang makes few guarantees on messages delivery - messages can get lost
713without any of the processes realising it (i.e. you send messages a, b, 720without any of the processes realising it (i.e. you send messages a, b,
714and c, and the other side only receives messages a and c). 721and c, and the other side only receives messages a and c).
715 722
716AEMP guarantees correct ordering, and the guarantee that there are no 723AEMP guarantees correct ordering, and the guarantee that after one message
717holes in the message sequence. 724is lost, all following ones sent to the same port are lost as well, until
718 725monitoring raises an error, so there are no silent "holes" in the message
719=item * In Erlang, processes can be declared dead and later be found to be 726sequence.
720alive.
721
722In Erlang it can happen that a monitored process is declared dead and
723linked processes get killed, but later it turns out that the process is
724still alive - and can receive messages.
725
726In AEMP, when port monitoring detects a port as dead, then that port will
727eventually be killed - it cannot happen that a node detects a port as dead
728and then later sends messages to it, finding it is still alive.
729 727
730=item * Erlang can send messages to the wrong port, AEMP does not. 728=item * Erlang can send messages to the wrong port, AEMP does not.
731 729
732In Erlang it is quite possible that a node that restarts reuses a process 730In Erlang it is quite likely that a node that restarts reuses a process ID
733ID known to other nodes for a completely different process, causing 731known to other nodes for a completely different process, causing messages
734messages destined for that process to end up in an unrelated process. 732destined for that process to end up in an unrelated process.
735 733
736AEMP never reuses port IDs, so old messages or old port IDs floating 734AEMP never reuses port IDs, so old messages or old port IDs floating
737around in the network will not be sent to an unrelated port. 735around in the network will not be sent to an unrelated port.
738 736
739=item * Erlang uses unprotected connections, AEMP uses secure 737=item * Erlang uses unprotected connections, AEMP uses secure
740authentication and can use TLS. 738authentication and can use TLS.
741 739
742AEMP can use a proven protocol - SSL/TLS - to protect connections and 740AEMP can use a proven protocol - TLS - to protect connections and
743securely authenticate nodes. 741securely authenticate nodes.
744 742
745=item * The AEMP protocol is optimised for both text-based and binary 743=item * The AEMP protocol is optimised for both text-based and binary
746communications. 744communications.
747 745
748The AEMP protocol, unlike the Erlang protocol, supports both 746The AEMP protocol, unlike the Erlang protocol, supports both programming
749language-independent text-only protocols (good for debugging) and binary, 747language independent text-only protocols (good for debugging) and binary,
750language-specific serialisers (e.g. Storable). 748language-specific serialisers (e.g. Storable). By default, unless TLS is
749used, the protocol is actually completely text-based.
751 750
752It has also been carefully designed to be implementable in other languages 751It has also been carefully designed to be implementable in other languages
753with a minimum of work while gracefully degrading fucntionality to make the 752with a minimum of work while gracefully degrading functionality to make the
754protocol simple. 753protocol simple.
755 754
756=item * AEMP has more flexible monitoring options than Erlang. 755=item * AEMP has more flexible monitoring options than Erlang.
757 756
758In Erlang, you can chose to receive I<all> exit signals as messages 757In Erlang, you can chose to receive I<all> exit signals as messages
759or I<none>, there is no in-between, so monitoring single processes is 758or I<none>, there is no in-between, so monitoring single processes is
760difficult to implement. Monitoring in AEMP is more flexible than in 759difficult to implement. Monitoring in AEMP is more flexible than in
761Erlang, as one can choose between automatic kill, exit message or callback 760Erlang, as one can choose between automatic kill, exit message or callback
762on a per-process basis. 761on a per-process basis.
763 762
764=item * Erlang has different semantics for monitoring and linking, AEMP has the same. 763=item * Erlang tries to hide remote/local connections, AEMP does not.
765 764
766Monitoring in Erlang is not an indicator of process death/crashes, 765Monitoring in Erlang is not an indicator of process death/crashes, in the
767as linking is (except linking is unreliable in Erlang). In AEMP, the 766same way as linking is (except linking is unreliable in Erlang).
768semantics of monitoring and linking are identical, linking is simply 767
769two-way monitoring with automatic kill. 768In AEMP, you don't "look up" registered port names or send to named ports
769that might or might not be persistent. Instead, you normally spawn a port
770on the remote node. The init function monitors you, and you monitor the
771remote port. Since both monitors are local to the node, they are much more
772reliable (no need for C<spawn_link>).
773
774This also saves round-trips and avoids sending messages to the wrong port
775(hard to do in Erlang).
770 776
771=back 777=back
772 778
779=head1 RATIONALE
780
781=over 4
782
783=item Why strings for port and node IDs, why not objects?
784
785We considered "objects", but found that the actual number of methods
786that can be called are quite low. Since port and node IDs travel over
787the network frequently, the serialising/deserialising would add lots of
788overhead, as well as having to keep a proxy object everywhere.
789
790Strings can easily be printed, easily serialised etc. and need no special
791procedures to be "valid".
792
793And as a result, a miniport consists of a single closure stored in a
794global hash - it can't become much cheaper.
795
796=item Why favour JSON, why not a real serialising format such as Storable?
797
798In fact, any AnyEvent::MP node will happily accept Storable as framing
799format, but currently there is no way to make a node use Storable by
800default (although all nodes will accept it).
801
802The default framing protocol is JSON because a) JSON::XS is many times
803faster for small messages and b) most importantly, after years of
804experience we found that object serialisation is causing more problems
805than it solves: Just like function calls, objects simply do not travel
806easily over the network, mostly because they will always be a copy, so you
807always have to re-think your design.
808
809Keeping your messages simple, concentrating on data structures rather than
810objects, will keep your messages clean, tidy and efficient.
811
812=back
813
773=head1 SEE ALSO 814=head1 SEE ALSO
815
816L<AnyEvent::MP::Intro> - a gentle introduction.
817
818L<AnyEvent::MP::Kernel> - more, lower-level, stuff.
819
820L<AnyEvent::MP::Global> - network maintainance and port groups, to find
821your applications.
774 822
775L<AnyEvent>. 823L<AnyEvent>.
776 824
777=head1 AUTHOR 825=head1 AUTHOR
778 826

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines