1 |
=head1 NAME |
2 |
|
3 |
AnyEvent::MP - multi-processing/message-passing framework |
4 |
|
5 |
=head1 SYNOPSIS |
6 |
|
7 |
use AnyEvent::MP; |
8 |
|
9 |
$NODE # contains this node's noderef |
10 |
NODE # returns this node's noderef |
11 |
NODE $port # returns the noderef of the port |
12 |
|
13 |
$SELF # receiving/own port id in rcv callbacks |
14 |
|
15 |
# initialise the node so it can send/receive messages |
16 |
initialise_node; # -OR- |
17 |
initialise_node "localhost:4040"; # -OR- |
18 |
initialise_node "slave/", "localhost:4040" |
19 |
|
20 |
# ports are message endpoints |
21 |
|
22 |
# sending messages |
23 |
snd $port, type => data...; |
24 |
snd $port, @msg; |
25 |
snd @msg_with_first_element_being_a_port; |
26 |
|
27 |
# creating/using miniports |
28 |
my $miniport = port { my @msg = @_; 0 }; |
29 |
|
30 |
# creating/using full ports |
31 |
my $port = port; |
32 |
rcv $port, smartmatch => $cb->(@msg); |
33 |
rcv $port, ping => sub { snd $_[0], "pong"; 0 }; |
34 |
rcv $port, pong => sub { warn "pong received\n"; 0 }; |
35 |
|
36 |
# more, smarter, matches (_any_ is exported by this module) |
37 |
rcv $port, [child_died => $pid] => sub { ... |
38 |
rcv $port, [_any_, _any_, 3] => sub { .. $_[2] is 3 |
39 |
|
40 |
# create a port on another node |
41 |
my $port = spawn $node, $initfunc, @initdata; |
42 |
|
43 |
# monitoring |
44 |
mon $port, $cb->(@msg) # callback is invoked on death |
45 |
mon $port, $otherport # kill otherport on abnormal death |
46 |
mon $port, $otherport, @msg # send message on death |
47 |
|
48 |
=head1 CURRENT STATUS |
49 |
|
50 |
AnyEvent::MP - stable API, should work |
51 |
AnyEvent::MP::Intro - outdated |
52 |
AnyEvent::MP::Kernel - WIP |
53 |
AnyEvent::MP::Transport - mostly stable |
54 |
|
55 |
stay tuned. |
56 |
|
57 |
=head1 DESCRIPTION |
58 |
|
59 |
This module (-family) implements a simple message passing framework. |
60 |
|
61 |
Despite its simplicity, you can securely message other processes running |
62 |
on the same or other hosts. |
63 |
|
64 |
For an introduction to this module family, see the L<AnyEvent::MP::Intro> |
65 |
manual page. |
66 |
|
67 |
At the moment, this module family is severly broken and underdocumented, |
68 |
so do not use. This was uploaded mainly to reserve the CPAN namespace - |
69 |
stay tuned! |
70 |
|
71 |
=head1 CONCEPTS |
72 |
|
73 |
=over 4 |
74 |
|
75 |
=item port |
76 |
|
77 |
A port is something you can send messages to (with the C<snd> function). |
78 |
|
79 |
Some ports allow you to register C<rcv> handlers that can match specific |
80 |
messages. All C<rcv> handlers will receive messages they match, messages |
81 |
will not be queued. |
82 |
|
83 |
=item port id - C<noderef#portname> |
84 |
|
85 |
A port id is normaly the concatenation of a noderef, a hash-mark (C<#>) as |
86 |
separator, and a port name (a printable string of unspecified format). An |
87 |
exception is the the node port, whose ID is identical to its node |
88 |
reference. |
89 |
|
90 |
=item node |
91 |
|
92 |
A node is a single process containing at least one port - the node |
93 |
port. You can send messages to node ports to find existing ports or to |
94 |
create new ports, among other things. |
95 |
|
96 |
Nodes are either private (single-process only), slaves (connected to a |
97 |
master node only) or public nodes (connectable from unrelated nodes). |
98 |
|
99 |
=item noderef - C<host:port,host:port...>, C<id@noderef>, C<id> |
100 |
|
101 |
A node reference is a string that either simply identifies the node (for |
102 |
private and slave nodes), or contains a recipe on how to reach a given |
103 |
node (for public nodes). |
104 |
|
105 |
This recipe is simply a comma-separated list of C<address:port> pairs (for |
106 |
TCP/IP, other protocols might look different). |
107 |
|
108 |
Node references come in two flavours: resolved (containing only numerical |
109 |
addresses) or unresolved (where hostnames are used instead of addresses). |
110 |
|
111 |
Before using an unresolved node reference in a message you first have to |
112 |
resolve it. |
113 |
|
114 |
=back |
115 |
|
116 |
=head1 VARIABLES/FUNCTIONS |
117 |
|
118 |
=over 4 |
119 |
|
120 |
=cut |
121 |
|
122 |
package AnyEvent::MP; |
123 |
|
124 |
use AnyEvent::MP::Kernel; |
125 |
|
126 |
use common::sense; |
127 |
|
128 |
use Carp (); |
129 |
|
130 |
use AE (); |
131 |
|
132 |
use base "Exporter"; |
133 |
|
134 |
our $VERSION = $AnyEvent::MP::Kernel::VERSION; |
135 |
|
136 |
our @EXPORT = qw( |
137 |
NODE $NODE *SELF node_of _any_ |
138 |
resolve_node initialise_node |
139 |
snd rcv mon kil reg psub spawn |
140 |
port |
141 |
); |
142 |
|
143 |
our $SELF; |
144 |
|
145 |
sub _self_die() { |
146 |
my $msg = $@; |
147 |
$msg =~ s/\n+$// unless ref $msg; |
148 |
kil $SELF, die => $msg; |
149 |
} |
150 |
|
151 |
=item $thisnode = NODE / $NODE |
152 |
|
153 |
The C<NODE> function returns, and the C<$NODE> variable contains |
154 |
the noderef of the local node. The value is initialised by a call |
155 |
to C<become_public> or C<become_slave>, after which all local port |
156 |
identifiers become invalid. |
157 |
|
158 |
=item $noderef = node_of $port |
159 |
|
160 |
Extracts and returns the noderef from a portid or a noderef. |
161 |
|
162 |
=item initialise_node $noderef, $seednode, $seednode... |
163 |
|
164 |
=item initialise_node "slave/", $master, $master... |
165 |
|
166 |
Before a node can talk to other nodes on the network it has to initialise |
167 |
itself - the minimum a node needs to know is it's own name, and optionally |
168 |
it should know the noderefs of some other nodes in the network. |
169 |
|
170 |
This function initialises a node - it must be called exactly once (or |
171 |
never) before calling other AnyEvent::MP functions. |
172 |
|
173 |
All arguments (optionally except for the first) are noderefs, which can be |
174 |
either resolved or unresolved. |
175 |
|
176 |
The first argument will be looked up in the configuration database first |
177 |
(if it is C<undef> then the current nodename will be used instead) to find |
178 |
the relevant configuration profile (see L<aemp>). If none is found then |
179 |
the default configuration is used. The configuration supplies additional |
180 |
seed/master nodes and can override the actual noderef. |
181 |
|
182 |
There are two types of networked nodes, public nodes and slave nodes: |
183 |
|
184 |
=over 4 |
185 |
|
186 |
=item public nodes |
187 |
|
188 |
For public nodes, C<$noderef> (supplied either directly to |
189 |
C<initialise_node> or indirectly via a profile or the nodename) must be a |
190 |
noderef (possibly unresolved, in which case it will be resolved). |
191 |
|
192 |
After resolving, the node will bind itself on all endpoints and try to |
193 |
connect to all additional C<$seednodes> that are specified. Seednodes are |
194 |
optional and can be used to quickly bootstrap the node into an existing |
195 |
network. |
196 |
|
197 |
=item slave nodes |
198 |
|
199 |
When the C<$noderef> (either as given or overriden by the config file) |
200 |
is the special string C<slave/>, then the node will become a slave |
201 |
node. Slave nodes cannot be contacted from outside and will route most of |
202 |
their traffic to the master node that they attach to. |
203 |
|
204 |
At least one additional noderef is required (either by specifying it |
205 |
directly or because it is part of the configuration profile): The node |
206 |
will try to connect to all of them and will become a slave attached to the |
207 |
first node it can successfully connect to. |
208 |
|
209 |
=back |
210 |
|
211 |
This function will block until all nodes have been resolved and, for slave |
212 |
nodes, until it has successfully established a connection to a master |
213 |
server. |
214 |
|
215 |
Example: become a public node listening on the guessed noderef, or the one |
216 |
specified via C<aemp> for the current node. This should be the most common |
217 |
form of invocation for "daemon"-type nodes. |
218 |
|
219 |
initialise_node; |
220 |
|
221 |
Example: become a slave node to any of the the seednodes specified via |
222 |
C<aemp>. This form is often used for commandline clients. |
223 |
|
224 |
initialise_node "slave/"; |
225 |
|
226 |
Example: become a slave node to any of the specified master servers. This |
227 |
form is also often used for commandline clients. |
228 |
|
229 |
initialise_node "slave/", "master1", "192.168.13.17", "mp.example.net"; |
230 |
|
231 |
Example: become a public node, and try to contact some well-known master |
232 |
servers to become part of the network. |
233 |
|
234 |
initialise_node undef, "master1", "master2"; |
235 |
|
236 |
Example: become a public node listening on port C<4041>. |
237 |
|
238 |
initialise_node 4041; |
239 |
|
240 |
Example: become a public node, only visible on localhost port 4044. |
241 |
|
242 |
initialise_node "localhost:4044"; |
243 |
|
244 |
=item $cv = resolve_node $noderef |
245 |
|
246 |
Takes an unresolved node reference that may contain hostnames and |
247 |
abbreviated IDs, resolves all of them and returns a resolved node |
248 |
reference. |
249 |
|
250 |
In addition to C<address:port> pairs allowed in resolved noderefs, the |
251 |
following forms are supported: |
252 |
|
253 |
=over 4 |
254 |
|
255 |
=item the empty string |
256 |
|
257 |
An empty-string component gets resolved as if the default port (4040) was |
258 |
specified. |
259 |
|
260 |
=item naked port numbers (e.g. C<1234>) |
261 |
|
262 |
These are resolved by prepending the local nodename and a colon, to be |
263 |
further resolved. |
264 |
|
265 |
=item hostnames (e.g. C<localhost:1234>, C<localhost>) |
266 |
|
267 |
These are resolved by using AnyEvent::DNS to resolve them, optionally |
268 |
looking up SRV records for the C<aemp=4040> port, if no port was |
269 |
specified. |
270 |
|
271 |
=back |
272 |
|
273 |
=item $SELF |
274 |
|
275 |
Contains the current port id while executing C<rcv> callbacks or C<psub> |
276 |
blocks. |
277 |
|
278 |
=item SELF, %SELF, @SELF... |
279 |
|
280 |
Due to some quirks in how perl exports variables, it is impossible to |
281 |
just export C<$SELF>, all the symbols called C<SELF> are exported by this |
282 |
module, but only C<$SELF> is currently used. |
283 |
|
284 |
=item snd $port, type => @data |
285 |
|
286 |
=item snd $port, @msg |
287 |
|
288 |
Send the given message to the given port ID, which can identify either |
289 |
a local or a remote port, and can be either a string or soemthignt hat |
290 |
stringifies a sa port ID (such as a port object :). |
291 |
|
292 |
While the message can be about anything, it is highly recommended to use a |
293 |
string as first element (a portid, or some word that indicates a request |
294 |
type etc.). |
295 |
|
296 |
The message data effectively becomes read-only after a call to this |
297 |
function: modifying any argument is not allowed and can cause many |
298 |
problems. |
299 |
|
300 |
The type of data you can transfer depends on the transport protocol: when |
301 |
JSON is used, then only strings, numbers and arrays and hashes consisting |
302 |
of those are allowed (no objects). When Storable is used, then anything |
303 |
that Storable can serialise and deserialise is allowed, and for the local |
304 |
node, anything can be passed. |
305 |
|
306 |
=item $local_port = port |
307 |
|
308 |
Create a new local port object that can be used either as a pattern |
309 |
matching port ("full port") or a single-callback port ("miniport"), |
310 |
depending on how C<rcv> callbacks are bound to the object. |
311 |
|
312 |
=item $port = port { my @msg = @_; $finished } |
313 |
|
314 |
Creates a "miniport", that is, a very lightweight port without any pattern |
315 |
matching behind it, and returns its ID. Semantically the same as creating |
316 |
a port and calling C<rcv $port, $callback> on it. |
317 |
|
318 |
The block will be called for every message received on the port. When the |
319 |
callback returns a true value its job is considered "done" and the port |
320 |
will be destroyed. Otherwise it will stay alive. |
321 |
|
322 |
The message will be passed as-is, no extra argument (i.e. no port id) will |
323 |
be passed to the callback. |
324 |
|
325 |
If you need the local port id in the callback, this works nicely: |
326 |
|
327 |
my $port; $port = port { |
328 |
snd $otherport, reply => $port; |
329 |
}; |
330 |
|
331 |
=cut |
332 |
|
333 |
sub rcv($@); |
334 |
|
335 |
sub port(;&) { |
336 |
my $id = "$UNIQ." . $ID++; |
337 |
my $port = "$NODE#$id"; |
338 |
|
339 |
if (@_) { |
340 |
rcv $port, shift; |
341 |
} else { |
342 |
$PORT{$id} = sub { }; # nop |
343 |
} |
344 |
|
345 |
$port |
346 |
} |
347 |
|
348 |
=item reg $port, $name |
349 |
|
350 |
=item reg $name |
351 |
|
352 |
Registers the given port (or C<$SELF><<< if missing) under the name |
353 |
C<$name>. If the name already exists it is replaced. |
354 |
|
355 |
A port can only be registered under one well known name. |
356 |
|
357 |
A port automatically becomes unregistered when it is killed. |
358 |
|
359 |
=cut |
360 |
|
361 |
sub reg(@) { |
362 |
my $port = @_ > 1 ? shift : $SELF || Carp::croak 'reg: called with one argument only, but $SELF not set,'; |
363 |
|
364 |
$REG{$_[0]} = $port; |
365 |
} |
366 |
|
367 |
=item rcv $port, $callback->(@msg) |
368 |
|
369 |
Replaces the callback on the specified miniport (after converting it to |
370 |
one if required). |
371 |
|
372 |
=item rcv $port, tagstring => $callback->(@msg), ... |
373 |
|
374 |
=item rcv $port, $smartmatch => $callback->(@msg), ... |
375 |
|
376 |
=item rcv $port, [$smartmatch...] => $callback->(@msg), ... |
377 |
|
378 |
Register callbacks to be called on matching messages on the given full |
379 |
port (after converting it to one if required) and return the port. |
380 |
|
381 |
The callback has to return a true value when its work is done, after |
382 |
which is will be removed, or a false value in which case it will stay |
383 |
registered. |
384 |
|
385 |
The global C<$SELF> (exported by this module) contains C<$port> while |
386 |
executing the callback. |
387 |
|
388 |
Runtime errors during callback execution will result in the port being |
389 |
C<kil>ed. |
390 |
|
391 |
If the match is an array reference, then it will be matched against the |
392 |
first elements of the message, otherwise only the first element is being |
393 |
matched. |
394 |
|
395 |
Any element in the match that is specified as C<_any_> (a function |
396 |
exported by this module) matches any single element of the message. |
397 |
|
398 |
While not required, it is highly recommended that the first matching |
399 |
element is a string identifying the message. The one-string-only match is |
400 |
also the most efficient match (by far). |
401 |
|
402 |
Example: create a port and bind receivers on it in one go. |
403 |
|
404 |
my $port = rcv port, |
405 |
msg1 => sub { ...; 0 }, |
406 |
msg2 => sub { ...; 0 }, |
407 |
; |
408 |
|
409 |
Example: create a port, bind receivers and send it in a message elsewhere |
410 |
in one go: |
411 |
|
412 |
snd $otherport, reply => |
413 |
rcv port, |
414 |
msg1 => sub { ...; 0 }, |
415 |
... |
416 |
; |
417 |
|
418 |
=cut |
419 |
|
420 |
sub rcv($@) { |
421 |
my $port = shift; |
422 |
my ($noderef, $portid) = split /#/, $port, 2; |
423 |
|
424 |
($NODE{$noderef} || add_node $noderef) == $NODE{""} |
425 |
or Carp::croak "$port: rcv can only be called on local ports, caught"; |
426 |
|
427 |
if (@_ == 1) { |
428 |
my $cb = shift; |
429 |
delete $PORT_DATA{$portid}; |
430 |
$PORT{$portid} = sub { |
431 |
local $SELF = $port; |
432 |
eval { |
433 |
&$cb |
434 |
and kil $port; |
435 |
}; |
436 |
_self_die if $@; |
437 |
}; |
438 |
} else { |
439 |
my $self = $PORT_DATA{$portid} ||= do { |
440 |
my $self = bless { |
441 |
id => $port, |
442 |
}, "AnyEvent::MP::Port"; |
443 |
|
444 |
$PORT{$portid} = sub { |
445 |
local $SELF = $port; |
446 |
|
447 |
eval { |
448 |
for (@{ $self->{rc0}{$_[0]} }) { |
449 |
$_ && &{$_->[0]} |
450 |
&& undef $_; |
451 |
} |
452 |
|
453 |
for (@{ $self->{rcv}{$_[0]} }) { |
454 |
$_ && [@_[1 .. @{$_->[1]}]] ~~ $_->[1] |
455 |
&& &{$_->[0]} |
456 |
&& undef $_; |
457 |
} |
458 |
|
459 |
for (@{ $self->{any} }) { |
460 |
$_ && [@_[0 .. $#{$_->[1]}]] ~~ $_->[1] |
461 |
&& &{$_->[0]} |
462 |
&& undef $_; |
463 |
} |
464 |
}; |
465 |
_self_die if $@; |
466 |
}; |
467 |
|
468 |
$self |
469 |
}; |
470 |
|
471 |
"AnyEvent::MP::Port" eq ref $self |
472 |
or Carp::croak "$port: rcv can only be called on message matching ports, caught"; |
473 |
|
474 |
while (@_) { |
475 |
my ($match, $cb) = splice @_, 0, 2; |
476 |
|
477 |
if (!ref $match) { |
478 |
push @{ $self->{rc0}{$match} }, [$cb]; |
479 |
} elsif (("ARRAY" eq ref $match && !ref $match->[0])) { |
480 |
my ($type, @match) = @$match; |
481 |
@match |
482 |
? push @{ $self->{rcv}{$match->[0]} }, [$cb, \@match] |
483 |
: push @{ $self->{rc0}{$match->[0]} }, [$cb]; |
484 |
} else { |
485 |
push @{ $self->{any} }, [$cb, $match]; |
486 |
} |
487 |
} |
488 |
} |
489 |
|
490 |
$port |
491 |
} |
492 |
|
493 |
=item $closure = psub { BLOCK } |
494 |
|
495 |
Remembers C<$SELF> and creates a closure out of the BLOCK. When the |
496 |
closure is executed, sets up the environment in the same way as in C<rcv> |
497 |
callbacks, i.e. runtime errors will cause the port to get C<kil>ed. |
498 |
|
499 |
This is useful when you register callbacks from C<rcv> callbacks: |
500 |
|
501 |
rcv delayed_reply => sub { |
502 |
my ($delay, @reply) = @_; |
503 |
my $timer = AE::timer $delay, 0, psub { |
504 |
snd @reply, $SELF; |
505 |
}; |
506 |
}; |
507 |
|
508 |
=cut |
509 |
|
510 |
sub psub(&) { |
511 |
my $cb = shift; |
512 |
|
513 |
my $port = $SELF |
514 |
or Carp::croak "psub can only be called from within rcv or psub callbacks, not"; |
515 |
|
516 |
sub { |
517 |
local $SELF = $port; |
518 |
|
519 |
if (wantarray) { |
520 |
my @res = eval { &$cb }; |
521 |
_self_die if $@; |
522 |
@res |
523 |
} else { |
524 |
my $res = eval { &$cb }; |
525 |
_self_die if $@; |
526 |
$res |
527 |
} |
528 |
} |
529 |
} |
530 |
|
531 |
=item $guard = mon $port, $cb->(@reason) |
532 |
|
533 |
=item $guard = mon $port, $rcvport |
534 |
|
535 |
=item $guard = mon $port |
536 |
|
537 |
=item $guard = mon $port, $rcvport, @msg |
538 |
|
539 |
Monitor the given port and do something when the port is killed or |
540 |
messages to it were lost, and optionally return a guard that can be used |
541 |
to stop monitoring again. |
542 |
|
543 |
C<mon> effectively guarantees that, in the absence of hardware failures, |
544 |
that after starting the monitor, either all messages sent to the port |
545 |
will arrive, or the monitoring action will be invoked after possible |
546 |
message loss has been detected. No messages will be lost "in between" |
547 |
(after the first lost message no further messages will be received by the |
548 |
port). After the monitoring action was invoked, further messages might get |
549 |
delivered again. |
550 |
|
551 |
In the first form (callback), the callback is simply called with any |
552 |
number of C<@reason> elements (no @reason means that the port was deleted |
553 |
"normally"). Note also that I<< the callback B<must> never die >>, so use |
554 |
C<eval> if unsure. |
555 |
|
556 |
In the second form (another port given), the other port (C<$rcvport>) |
557 |
will be C<kil>'ed with C<@reason>, iff a @reason was specified, i.e. on |
558 |
"normal" kils nothing happens, while under all other conditions, the other |
559 |
port is killed with the same reason. |
560 |
|
561 |
The third form (kill self) is the same as the second form, except that |
562 |
C<$rvport> defaults to C<$SELF>. |
563 |
|
564 |
In the last form (message), a message of the form C<@msg, @reason> will be |
565 |
C<snd>. |
566 |
|
567 |
As a rule of thumb, monitoring requests should always monitor a port from |
568 |
a local port (or callback). The reason is that kill messages might get |
569 |
lost, just like any other message. Another less obvious reason is that |
570 |
even monitoring requests can get lost (for exmaple, when the connection |
571 |
to the other node goes down permanently). When monitoring a port locally |
572 |
these problems do not exist. |
573 |
|
574 |
Example: call a given callback when C<$port> is killed. |
575 |
|
576 |
mon $port, sub { warn "port died because of <@_>\n" }; |
577 |
|
578 |
Example: kill ourselves when C<$port> is killed abnormally. |
579 |
|
580 |
mon $port; |
581 |
|
582 |
Example: send us a restart message when another C<$port> is killed. |
583 |
|
584 |
mon $port, $self => "restart"; |
585 |
|
586 |
=cut |
587 |
|
588 |
sub mon { |
589 |
my ($noderef, $port) = split /#/, shift, 2; |
590 |
|
591 |
my $node = $NODE{$noderef} || add_node $noderef; |
592 |
|
593 |
my $cb = @_ ? shift : $SELF || Carp::croak 'mon: called with one argument only, but $SELF not set,'; |
594 |
|
595 |
unless (ref $cb) { |
596 |
if (@_) { |
597 |
# send a kill info message |
598 |
my (@msg) = ($cb, @_); |
599 |
$cb = sub { snd @msg, @_ }; |
600 |
} else { |
601 |
# simply kill other port |
602 |
my $port = $cb; |
603 |
$cb = sub { kil $port, @_ if @_ }; |
604 |
} |
605 |
} |
606 |
|
607 |
$node->monitor ($port, $cb); |
608 |
|
609 |
defined wantarray |
610 |
and AnyEvent::Util::guard { $node->unmonitor ($port, $cb) } |
611 |
} |
612 |
|
613 |
=item $guard = mon_guard $port, $ref, $ref... |
614 |
|
615 |
Monitors the given C<$port> and keeps the passed references. When the port |
616 |
is killed, the references will be freed. |
617 |
|
618 |
Optionally returns a guard that will stop the monitoring. |
619 |
|
620 |
This function is useful when you create e.g. timers or other watchers and |
621 |
want to free them when the port gets killed: |
622 |
|
623 |
$port->rcv (start => sub { |
624 |
my $timer; $timer = mon_guard $port, AE::timer 1, 1, sub { |
625 |
undef $timer if 0.9 < rand; |
626 |
}); |
627 |
}); |
628 |
|
629 |
=cut |
630 |
|
631 |
sub mon_guard { |
632 |
my ($port, @refs) = @_; |
633 |
|
634 |
#TODO: mon-less form? |
635 |
|
636 |
mon $port, sub { 0 && @refs } |
637 |
} |
638 |
|
639 |
=item kil $port[, @reason] |
640 |
|
641 |
Kill the specified port with the given C<@reason>. |
642 |
|
643 |
If no C<@reason> is specified, then the port is killed "normally" (linked |
644 |
ports will not be kileld, or even notified). |
645 |
|
646 |
Otherwise, linked ports get killed with the same reason (second form of |
647 |
C<mon>, see below). |
648 |
|
649 |
Runtime errors while evaluating C<rcv> callbacks or inside C<psub> blocks |
650 |
will be reported as reason C<< die => $@ >>. |
651 |
|
652 |
Transport/communication errors are reported as C<< transport_error => |
653 |
$message >>. |
654 |
|
655 |
=cut |
656 |
|
657 |
=item $port = spawn $node, $initfunc[, @initdata] |
658 |
|
659 |
Creates a port on the node C<$node> (which can also be a port ID, in which |
660 |
case it's the node where that port resides). |
661 |
|
662 |
The port ID of the newly created port is return immediately, and it is |
663 |
permissible to immediately start sending messages or monitor the port. |
664 |
|
665 |
After the port has been created, the init function is |
666 |
called. This function must be a fully-qualified function name |
667 |
(e.g. C<MyApp::Chat::Server::init>). To specify a function in the main |
668 |
program, use C<::name>. |
669 |
|
670 |
If the function doesn't exist, then the node tries to C<require> |
671 |
the package, then the package above the package and so on (e.g. |
672 |
C<MyApp::Chat::Server>, C<MyApp::Chat>, C<MyApp>) until the function |
673 |
exists or it runs out of package names. |
674 |
|
675 |
The init function is then called with the newly-created port as context |
676 |
object (C<$SELF>) and the C<@initdata> values as arguments. |
677 |
|
678 |
A common idiom is to pass your own port, monitor the spawned port, and |
679 |
in the init function, monitor the original port. This two-way monitoring |
680 |
ensures that both ports get cleaned up when there is a problem. |
681 |
|
682 |
Example: spawn a chat server port on C<$othernode>. |
683 |
|
684 |
# this node, executed from within a port context: |
685 |
my $server = spawn $othernode, "MyApp::Chat::Server::connect", $SELF; |
686 |
mon $server; |
687 |
|
688 |
# init function on C<$othernode> |
689 |
sub connect { |
690 |
my ($srcport) = @_; |
691 |
|
692 |
mon $srcport; |
693 |
|
694 |
rcv $SELF, sub { |
695 |
... |
696 |
}; |
697 |
} |
698 |
|
699 |
=cut |
700 |
|
701 |
sub _spawn { |
702 |
my $port = shift; |
703 |
my $init = shift; |
704 |
|
705 |
local $SELF = "$NODE#$port"; |
706 |
eval { |
707 |
&{ load_func $init } |
708 |
}; |
709 |
_self_die if $@; |
710 |
} |
711 |
|
712 |
sub spawn(@) { |
713 |
my ($noderef, undef) = split /#/, shift, 2; |
714 |
|
715 |
my $id = "$RUNIQ." . $ID++; |
716 |
|
717 |
$_[0] =~ /::/ |
718 |
or Carp::croak "spawn init function must be a fully-qualified name, caught"; |
719 |
|
720 |
($NODE{$noderef} || add_node $noderef) |
721 |
->send (["", "AnyEvent::MP::_spawn" => $id, @_]); |
722 |
|
723 |
"$noderef#$id" |
724 |
} |
725 |
|
726 |
=back |
727 |
|
728 |
=head1 NODE MESSAGES |
729 |
|
730 |
Nodes understand the following messages sent to them. Many of them take |
731 |
arguments called C<@reply>, which will simply be used to compose a reply |
732 |
message - C<$reply[0]> is the port to reply to, C<$reply[1]> the type and |
733 |
the remaining arguments are simply the message data. |
734 |
|
735 |
While other messages exist, they are not public and subject to change. |
736 |
|
737 |
=over 4 |
738 |
|
739 |
=cut |
740 |
|
741 |
=item lookup => $name, @reply |
742 |
|
743 |
Replies with the port ID of the specified well-known port, or C<undef>. |
744 |
|
745 |
=item devnull => ... |
746 |
|
747 |
Generic data sink/CPU heat conversion. |
748 |
|
749 |
=item relay => $port, @msg |
750 |
|
751 |
Simply forwards the message to the given port. |
752 |
|
753 |
=item eval => $string[ @reply] |
754 |
|
755 |
Evaluates the given string. If C<@reply> is given, then a message of the |
756 |
form C<@reply, $@, @evalres> is sent. |
757 |
|
758 |
Example: crash another node. |
759 |
|
760 |
snd $othernode, eval => "exit"; |
761 |
|
762 |
=item time => @reply |
763 |
|
764 |
Replies the the current node time to C<@reply>. |
765 |
|
766 |
Example: tell the current node to send the current time to C<$myport> in a |
767 |
C<timereply> message. |
768 |
|
769 |
snd $NODE, time => $myport, timereply => 1, 2; |
770 |
# => snd $myport, timereply => 1, 2, <time> |
771 |
|
772 |
=back |
773 |
|
774 |
=head1 AnyEvent::MP vs. Distributed Erlang |
775 |
|
776 |
AnyEvent::MP got lots of its ideas from distributed Erlang (Erlang node |
777 |
== aemp node, Erlang process == aemp port), so many of the documents and |
778 |
programming techniques employed by Erlang apply to AnyEvent::MP. Here is a |
779 |
sample: |
780 |
|
781 |
http://www.Erlang.se/doc/programming_rules.shtml |
782 |
http://Erlang.org/doc/getting_started/part_frame.html # chapters 3 and 4 |
783 |
http://Erlang.org/download/Erlang-book-part1.pdf # chapters 5 and 6 |
784 |
http://Erlang.org/download/armstrong_thesis_2003.pdf # chapters 4 and 5 |
785 |
|
786 |
Despite the similarities, there are also some important differences: |
787 |
|
788 |
=over 4 |
789 |
|
790 |
=item * Node references contain the recipe on how to contact them. |
791 |
|
792 |
Erlang relies on special naming and DNS to work everywhere in the |
793 |
same way. AEMP relies on each node knowing it's own address(es), with |
794 |
convenience functionality. |
795 |
|
796 |
This means that AEMP requires a less tightly controlled environment at the |
797 |
cost of longer node references and a slightly higher management overhead. |
798 |
|
799 |
=item * Erlang uses processes and a mailbox, AEMP does not queue. |
800 |
|
801 |
Erlang uses processes that selctively receive messages, and therefore |
802 |
needs a queue. AEMP is event based, queuing messages would serve no useful |
803 |
purpose. |
804 |
|
805 |
(But see L<Coro::MP> for a more Erlang-like process model on top of AEMP). |
806 |
|
807 |
=item * Erlang sends are synchronous, AEMP sends are asynchronous. |
808 |
|
809 |
Sending messages in Erlang is synchronous and blocks the process. AEMP |
810 |
sends are immediate, connection establishment is handled in the |
811 |
background. |
812 |
|
813 |
=item * Erlang can silently lose messages, AEMP cannot. |
814 |
|
815 |
Erlang makes few guarantees on messages delivery - messages can get lost |
816 |
without any of the processes realising it (i.e. you send messages a, b, |
817 |
and c, and the other side only receives messages a and c). |
818 |
|
819 |
AEMP guarantees correct ordering, and the guarantee that there are no |
820 |
holes in the message sequence. |
821 |
|
822 |
=item * In Erlang, processes can be declared dead and later be found to be |
823 |
alive. |
824 |
|
825 |
In Erlang it can happen that a monitored process is declared dead and |
826 |
linked processes get killed, but later it turns out that the process is |
827 |
still alive - and can receive messages. |
828 |
|
829 |
In AEMP, when port monitoring detects a port as dead, then that port will |
830 |
eventually be killed - it cannot happen that a node detects a port as dead |
831 |
and then later sends messages to it, finding it is still alive. |
832 |
|
833 |
=item * Erlang can send messages to the wrong port, AEMP does not. |
834 |
|
835 |
In Erlang it is quite possible that a node that restarts reuses a process |
836 |
ID known to other nodes for a completely different process, causing |
837 |
messages destined for that process to end up in an unrelated process. |
838 |
|
839 |
AEMP never reuses port IDs, so old messages or old port IDs floating |
840 |
around in the network will not be sent to an unrelated port. |
841 |
|
842 |
=item * Erlang uses unprotected connections, AEMP uses secure |
843 |
authentication and can use TLS. |
844 |
|
845 |
AEMP can use a proven protocol - SSL/TLS - to protect connections and |
846 |
securely authenticate nodes. |
847 |
|
848 |
=item * The AEMP protocol is optimised for both text-based and binary |
849 |
communications. |
850 |
|
851 |
The AEMP protocol, unlike the Erlang protocol, supports both |
852 |
language-independent text-only protocols (good for debugging) and binary, |
853 |
language-specific serialisers (e.g. Storable). |
854 |
|
855 |
It has also been carefully designed to be implementable in other languages |
856 |
with a minimum of work while gracefully degrading fucntionality to make the |
857 |
protocol simple. |
858 |
|
859 |
=item * AEMP has more flexible monitoring options than Erlang. |
860 |
|
861 |
In Erlang, you can chose to receive I<all> exit signals as messages |
862 |
or I<none>, there is no in-between, so monitoring single processes is |
863 |
difficult to implement. Monitoring in AEMP is more flexible than in |
864 |
Erlang, as one can choose between automatic kill, exit message or callback |
865 |
on a per-process basis. |
866 |
|
867 |
=item * Erlang tries to hide remote/local connections, AEMP does not. |
868 |
|
869 |
Monitoring in Erlang is not an indicator of process death/crashes, |
870 |
as linking is (except linking is unreliable in Erlang). |
871 |
|
872 |
In AEMP, you don't "look up" registered port names or send to named ports |
873 |
that might or might not be persistent. Instead, you normally spawn a port |
874 |
on the remote node. The init function monitors the you, and you monitor |
875 |
the remote port. Since both monitors are local to the node, they are much |
876 |
more reliable. |
877 |
|
878 |
This also saves round-trips and avoids sending messages to the wrong port |
879 |
(hard to do in Erlang). |
880 |
|
881 |
=back |
882 |
|
883 |
=head1 RATIONALE |
884 |
|
885 |
=over 4 |
886 |
|
887 |
=item Why strings for ports and noderefs, why not objects? |
888 |
|
889 |
We considered "objects", but found that the actual number of methods |
890 |
thatc an be called are very low. Since port IDs and noderefs travel over |
891 |
the network frequently, the serialising/deserialising would add lots of |
892 |
overhead, as well as having to keep a proxy object. |
893 |
|
894 |
Strings can easily be printed, easily serialised etc. and need no special |
895 |
procedures to be "valid". |
896 |
|
897 |
And a a miniport consists of a single closure stored in a global hash - it |
898 |
can't become much cheaper. |
899 |
|
900 |
=item Why favour JSON, why not real serialising format such as Storable? |
901 |
|
902 |
In fact, any AnyEvent::MP node will happily accept Storable as framing |
903 |
format, but currently there is no way to make a node use Storable by |
904 |
default. |
905 |
|
906 |
The default framing protocol is JSON because a) JSON::XS is many times |
907 |
faster for small messages and b) most importantly, after years of |
908 |
experience we found that object serialisation is causing more problems |
909 |
than it gains: Just like function calls, objects simply do not travel |
910 |
easily over the network, mostly because they will always be a copy, so you |
911 |
always have to re-think your design. |
912 |
|
913 |
Keeping your messages simple, concentrating on data structures rather than |
914 |
objects, will keep your messages clean, tidy and efficient. |
915 |
|
916 |
=back |
917 |
|
918 |
=head1 SEE ALSO |
919 |
|
920 |
L<AnyEvent>. |
921 |
|
922 |
=head1 AUTHOR |
923 |
|
924 |
Marc Lehmann <schmorp@schmorp.de> |
925 |
http://home.schmorp.de/ |
926 |
|
927 |
=cut |
928 |
|
929 |
1 |
930 |
|