1 |
=head1 NAME |
2 |
|
3 |
AnyEvent - provide framework for multiple event loops |
4 |
|
5 |
EV, Event, Glib, Tk, Perl, Event::Lib, Qt, POE - various supported event loops |
6 |
|
7 |
=head1 SYNOPSIS |
8 |
|
9 |
use AnyEvent; |
10 |
|
11 |
my $w = AnyEvent->io (fh => $fh, poll => "r|w", cb => sub { |
12 |
... |
13 |
}); |
14 |
|
15 |
my $w = AnyEvent->timer (after => $seconds, cb => sub { |
16 |
... |
17 |
}); |
18 |
|
19 |
my $w = AnyEvent->condvar; # stores whether a condition was flagged |
20 |
$w->wait; # enters "main loop" till $condvar gets ->send |
21 |
$w->send; # wake up current and all future wait's |
22 |
|
23 |
=head1 WHY YOU SHOULD USE THIS MODULE (OR NOT) |
24 |
|
25 |
Glib, POE, IO::Async, Event... CPAN offers event models by the dozen |
26 |
nowadays. So what is different about AnyEvent? |
27 |
|
28 |
Executive Summary: AnyEvent is I<compatible>, AnyEvent is I<free of |
29 |
policy> and AnyEvent is I<small and efficient>. |
30 |
|
31 |
First and foremost, I<AnyEvent is not an event model> itself, it only |
32 |
interfaces to whatever event model the main program happens to use in a |
33 |
pragmatic way. For event models and certain classes of immortals alike, |
34 |
the statement "there can only be one" is a bitter reality: In general, |
35 |
only one event loop can be active at the same time in a process. AnyEvent |
36 |
helps hiding the differences between those event loops. |
37 |
|
38 |
The goal of AnyEvent is to offer module authors the ability to do event |
39 |
programming (waiting for I/O or timer events) without subscribing to a |
40 |
religion, a way of living, and most importantly: without forcing your |
41 |
module users into the same thing by forcing them to use the same event |
42 |
model you use. |
43 |
|
44 |
For modules like POE or IO::Async (which is a total misnomer as it is |
45 |
actually doing all I/O I<synchronously>...), using them in your module is |
46 |
like joining a cult: After you joined, you are dependent on them and you |
47 |
cannot use anything else, as it is simply incompatible to everything that |
48 |
isn't itself. What's worse, all the potential users of your module are |
49 |
I<also> forced to use the same event loop you use. |
50 |
|
51 |
AnyEvent is different: AnyEvent + POE works fine. AnyEvent + Glib works |
52 |
fine. AnyEvent + Tk works fine etc. etc. but none of these work together |
53 |
with the rest: POE + IO::Async? no go. Tk + Event? no go. Again: if |
54 |
your module uses one of those, every user of your module has to use it, |
55 |
too. But if your module uses AnyEvent, it works transparently with all |
56 |
event models it supports (including stuff like POE and IO::Async, as long |
57 |
as those use one of the supported event loops. It is trivial to add new |
58 |
event loops to AnyEvent, too, so it is future-proof). |
59 |
|
60 |
In addition to being free of having to use I<the one and only true event |
61 |
model>, AnyEvent also is free of bloat and policy: with POE or similar |
62 |
modules, you get an enourmous amount of code and strict rules you have to |
63 |
follow. AnyEvent, on the other hand, is lean and up to the point, by only |
64 |
offering the functionality that is necessary, in as thin as a wrapper as |
65 |
technically possible. |
66 |
|
67 |
Of course, if you want lots of policy (this can arguably be somewhat |
68 |
useful) and you want to force your users to use the one and only event |
69 |
model, you should I<not> use this module. |
70 |
|
71 |
=head1 DESCRIPTION |
72 |
|
73 |
L<AnyEvent> provides an identical interface to multiple event loops. This |
74 |
allows module authors to utilise an event loop without forcing module |
75 |
users to use the same event loop (as only a single event loop can coexist |
76 |
peacefully at any one time). |
77 |
|
78 |
The interface itself is vaguely similar, but not identical to the L<Event> |
79 |
module. |
80 |
|
81 |
During the first call of any watcher-creation method, the module tries |
82 |
to detect the currently loaded event loop by probing whether one of the |
83 |
following modules is already loaded: L<EV>, |
84 |
L<Event>, L<Glib>, L<AnyEvent::Impl::Perl>, L<Tk>, L<Event::Lib>, L<Qt>, |
85 |
L<POE>. The first one found is used. If none are found, the module tries |
86 |
to load these modules (excluding Tk, Event::Lib, Qt and POE as the pure perl |
87 |
adaptor should always succeed) in the order given. The first one that can |
88 |
be successfully loaded will be used. If, after this, still none could be |
89 |
found, AnyEvent will fall back to a pure-perl event loop, which is not |
90 |
very efficient, but should work everywhere. |
91 |
|
92 |
Because AnyEvent first checks for modules that are already loaded, loading |
93 |
an event model explicitly before first using AnyEvent will likely make |
94 |
that model the default. For example: |
95 |
|
96 |
use Tk; |
97 |
use AnyEvent; |
98 |
|
99 |
# .. AnyEvent will likely default to Tk |
100 |
|
101 |
The I<likely> means that, if any module loads another event model and |
102 |
starts using it, all bets are off. Maybe you should tell their authors to |
103 |
use AnyEvent so their modules work together with others seamlessly... |
104 |
|
105 |
The pure-perl implementation of AnyEvent is called |
106 |
C<AnyEvent::Impl::Perl>. Like other event modules you can load it |
107 |
explicitly. |
108 |
|
109 |
=head1 WATCHERS |
110 |
|
111 |
AnyEvent has the central concept of a I<watcher>, which is an object that |
112 |
stores relevant data for each kind of event you are waiting for, such as |
113 |
the callback to call, the filehandle to watch, etc. |
114 |
|
115 |
These watchers are normal Perl objects with normal Perl lifetime. After |
116 |
creating a watcher it will immediately "watch" for events and invoke the |
117 |
callback when the event occurs (of course, only when the event model |
118 |
is in control). |
119 |
|
120 |
To disable the watcher you have to destroy it (e.g. by setting the |
121 |
variable you store it in to C<undef> or otherwise deleting all references |
122 |
to it). |
123 |
|
124 |
All watchers are created by calling a method on the C<AnyEvent> class. |
125 |
|
126 |
Many watchers either are used with "recursion" (repeating timers for |
127 |
example), or need to refer to their watcher object in other ways. |
128 |
|
129 |
An any way to achieve that is this pattern: |
130 |
|
131 |
my $w; $w = AnyEvent->type (arg => value ..., cb => sub { |
132 |
# you can use $w here, for example to undef it |
133 |
undef $w; |
134 |
}); |
135 |
|
136 |
Note that C<my $w; $w => combination. This is necessary because in Perl, |
137 |
my variables are only visible after the statement in which they are |
138 |
declared. |
139 |
|
140 |
=head2 I/O WATCHERS |
141 |
|
142 |
You can create an I/O watcher by calling the C<< AnyEvent->io >> method |
143 |
with the following mandatory key-value pairs as arguments: |
144 |
|
145 |
C<fh> the Perl I<file handle> (I<not> file descriptor) to watch |
146 |
for events. C<poll> must be a string that is either C<r> or C<w>, |
147 |
which creates a watcher waiting for "r"eadable or "w"ritable events, |
148 |
respectively. C<cb> is the callback to invoke each time the file handle |
149 |
becomes ready. |
150 |
|
151 |
Although the callback might get passed parameters, their value and |
152 |
presence is undefined and you cannot rely on them. Portable AnyEvent |
153 |
callbacks cannot use arguments passed to I/O watcher callbacks. |
154 |
|
155 |
The I/O watcher might use the underlying file descriptor or a copy of it. |
156 |
You must not close a file handle as long as any watcher is active on the |
157 |
underlying file descriptor. |
158 |
|
159 |
Some event loops issue spurious readyness notifications, so you should |
160 |
always use non-blocking calls when reading/writing from/to your file |
161 |
handles. |
162 |
|
163 |
Example: |
164 |
|
165 |
# wait for readability of STDIN, then read a line and disable the watcher |
166 |
my $w; $w = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub { |
167 |
chomp (my $input = <STDIN>); |
168 |
warn "read: $input\n"; |
169 |
undef $w; |
170 |
}); |
171 |
|
172 |
=head2 TIME WATCHERS |
173 |
|
174 |
You can create a time watcher by calling the C<< AnyEvent->timer >> |
175 |
method with the following mandatory arguments: |
176 |
|
177 |
C<after> specifies after how many seconds (fractional values are |
178 |
supported) the callback should be invoked. C<cb> is the callback to invoke |
179 |
in that case. |
180 |
|
181 |
Although the callback might get passed parameters, their value and |
182 |
presence is undefined and you cannot rely on them. Portable AnyEvent |
183 |
callbacks cannot use arguments passed to time watcher callbacks. |
184 |
|
185 |
The timer callback will be invoked at most once: if you want a repeating |
186 |
timer you have to create a new watcher (this is a limitation by both Tk |
187 |
and Glib). |
188 |
|
189 |
Example: |
190 |
|
191 |
# fire an event after 7.7 seconds |
192 |
my $w = AnyEvent->timer (after => 7.7, cb => sub { |
193 |
warn "timeout\n"; |
194 |
}); |
195 |
|
196 |
# to cancel the timer: |
197 |
undef $w; |
198 |
|
199 |
Example 2: |
200 |
|
201 |
# fire an event after 0.5 seconds, then roughly every second |
202 |
my $w; |
203 |
|
204 |
my $cb = sub { |
205 |
# cancel the old timer while creating a new one |
206 |
$w = AnyEvent->timer (after => 1, cb => $cb); |
207 |
}; |
208 |
|
209 |
# start the "loop" by creating the first watcher |
210 |
$w = AnyEvent->timer (after => 0.5, cb => $cb); |
211 |
|
212 |
=head3 TIMING ISSUES |
213 |
|
214 |
There are two ways to handle timers: based on real time (relative, "fire |
215 |
in 10 seconds") and based on wallclock time (absolute, "fire at 12 |
216 |
o'clock"). |
217 |
|
218 |
While most event loops expect timers to specified in a relative way, they |
219 |
use absolute time internally. This makes a difference when your clock |
220 |
"jumps", for example, when ntp decides to set your clock backwards from |
221 |
the wrong date of 2014-01-01 to 2008-01-01, a watcher that is supposed to |
222 |
fire "after" a second might actually take six years to finally fire. |
223 |
|
224 |
AnyEvent cannot compensate for this. The only event loop that is conscious |
225 |
about these issues is L<EV>, which offers both relative (ev_timer, based |
226 |
on true relative time) and absolute (ev_periodic, based on wallclock time) |
227 |
timers. |
228 |
|
229 |
AnyEvent always prefers relative timers, if available, matching the |
230 |
AnyEvent API. |
231 |
|
232 |
=head2 SIGNAL WATCHERS |
233 |
|
234 |
You can watch for signals using a signal watcher, C<signal> is the signal |
235 |
I<name> without any C<SIG> prefix, C<cb> is the Perl callback to |
236 |
be invoked whenever a signal occurs. |
237 |
|
238 |
Although the callback might get passed parameters, their value and |
239 |
presence is undefined and you cannot rely on them. Portable AnyEvent |
240 |
callbacks cannot use arguments passed to signal watcher callbacks. |
241 |
|
242 |
Multiple signal occurances can be clumped together into one callback |
243 |
invocation, and callback invocation will be synchronous. synchronous means |
244 |
that it might take a while until the signal gets handled by the process, |
245 |
but it is guarenteed not to interrupt any other callbacks. |
246 |
|
247 |
The main advantage of using these watchers is that you can share a signal |
248 |
between multiple watchers. |
249 |
|
250 |
This watcher might use C<%SIG>, so programs overwriting those signals |
251 |
directly will likely not work correctly. |
252 |
|
253 |
Example: exit on SIGINT |
254 |
|
255 |
my $w = AnyEvent->signal (signal => "INT", cb => sub { exit 1 }); |
256 |
|
257 |
=head2 CHILD PROCESS WATCHERS |
258 |
|
259 |
You can also watch on a child process exit and catch its exit status. |
260 |
|
261 |
The child process is specified by the C<pid> argument (if set to C<0>, it |
262 |
watches for any child process exit). The watcher will trigger as often |
263 |
as status change for the child are received. This works by installing a |
264 |
signal handler for C<SIGCHLD>. The callback will be called with the pid |
265 |
and exit status (as returned by waitpid), so unlike other watcher types, |
266 |
you I<can> rely on child watcher callback arguments. |
267 |
|
268 |
There is a slight catch to child watchers, however: you usually start them |
269 |
I<after> the child process was created, and this means the process could |
270 |
have exited already (and no SIGCHLD will be sent anymore). |
271 |
|
272 |
Not all event models handle this correctly (POE doesn't), but even for |
273 |
event models that I<do> handle this correctly, they usually need to be |
274 |
loaded before the process exits (i.e. before you fork in the first place). |
275 |
|
276 |
This means you cannot create a child watcher as the very first thing in an |
277 |
AnyEvent program, you I<have> to create at least one watcher before you |
278 |
C<fork> the child (alternatively, you can call C<AnyEvent::detect>). |
279 |
|
280 |
Example: fork a process and wait for it |
281 |
|
282 |
my $done = AnyEvent->condvar; |
283 |
|
284 |
AnyEvent::detect; # force event module to be initialised |
285 |
|
286 |
my $pid = fork or exit 5; |
287 |
|
288 |
my $w = AnyEvent->child ( |
289 |
pid => $pid, |
290 |
cb => sub { |
291 |
my ($pid, $status) = @_; |
292 |
warn "pid $pid exited with status $status"; |
293 |
$done->send; |
294 |
}, |
295 |
); |
296 |
|
297 |
# do something else, then wait for process exit |
298 |
$done->wait; |
299 |
|
300 |
=head2 CONDITION VARIABLES |
301 |
|
302 |
If you are familiar with some event loops you will know that all of them |
303 |
require you to run some blocking "loop", "run" or similar function that |
304 |
will actively watch for new events and call your callbacks. |
305 |
|
306 |
AnyEvent is different, it expects somebody else to run the event loop and |
307 |
will only block when necessary (usually when told by the user). |
308 |
|
309 |
The instrument to do that is called a "condition variable", so called |
310 |
because they represent a condition that must become true. |
311 |
|
312 |
Condition variables can be created by calling the C<< AnyEvent->condvar |
313 |
>> method, usually without arguments. The only argument pair allowed is |
314 |
C<cb>, which specifies a callback to be called when the condition variable |
315 |
becomes true. |
316 |
|
317 |
After creation, the conditon variable is "false" until it becomes "true" |
318 |
by calling the C<send> method. |
319 |
|
320 |
Condition variables are similar to callbacks, except that you can |
321 |
optionally wait for them. They can also be called merge points - points |
322 |
in time where multiple outstandign events have been processed. And yet |
323 |
another way to call them is transations - each condition variable can be |
324 |
used to represent a transaction, which finishes at some point and delivers |
325 |
a result. |
326 |
|
327 |
Condition variables are very useful to signal that something has finished, |
328 |
for example, if you write a module that does asynchronous http requests, |
329 |
then a condition variable would be the ideal candidate to signal the |
330 |
availability of results. The user can either act when the callback is |
331 |
called or can synchronously C<< ->wait >> for the results. |
332 |
|
333 |
You can also use them to simulate traditional event loops - for example, |
334 |
you can block your main program until an event occurs - for example, you |
335 |
could C<< ->wait >> in your main program until the user clicks the Quit |
336 |
button of your app, which would C<< ->send >> the "quit" event. |
337 |
|
338 |
Note that condition variables recurse into the event loop - if you have |
339 |
two pieces of code that call C<< ->wait >> in a round-robbin fashion, you |
340 |
lose. Therefore, condition variables are good to export to your caller, but |
341 |
you should avoid making a blocking wait yourself, at least in callbacks, |
342 |
as this asks for trouble. |
343 |
|
344 |
Condition variables are represented by hash refs in perl, and the keys |
345 |
used by AnyEvent itself are all named C<_ae_XXX> to make subclassing |
346 |
easy (it is often useful to build your own transaction class on top of |
347 |
AnyEvent). To subclass, use C<AnyEvent::CondVar> as base class and call |
348 |
it's C<new> method in your own C<new> method. |
349 |
|
350 |
There are two "sides" to a condition variable - the "producer side" which |
351 |
eventually calls C<< -> send >>, and the "consumer side", which waits |
352 |
for the send to occur. |
353 |
|
354 |
Example: |
355 |
|
356 |
# wait till the result is ready |
357 |
my $result_ready = AnyEvent->condvar; |
358 |
|
359 |
# do something such as adding a timer |
360 |
# or socket watcher the calls $result_ready->send |
361 |
# when the "result" is ready. |
362 |
# in this case, we simply use a timer: |
363 |
my $w = AnyEvent->timer ( |
364 |
after => 1, |
365 |
cb => sub { $result_ready->send }, |
366 |
); |
367 |
|
368 |
# this "blocks" (while handling events) till the callback |
369 |
# calls send |
370 |
$result_ready->wait; |
371 |
|
372 |
=head3 METHODS FOR PRODUCERS |
373 |
|
374 |
These methods should only be used by the producing side, i.e. the |
375 |
code/module that eventually sends the signal. Note that it is also |
376 |
the producer side which creates the condvar in most cases, but it isn't |
377 |
uncommon for the consumer to create it as well. |
378 |
|
379 |
=over 4 |
380 |
|
381 |
=item $cv->send (...) |
382 |
|
383 |
Flag the condition as ready - a running C<< ->wait >> and all further |
384 |
calls to C<wait> will (eventually) return after this method has been |
385 |
called. If nobody is waiting the send will be remembered. |
386 |
|
387 |
If a callback has been set on the condition variable, it is called |
388 |
immediately from within send. |
389 |
|
390 |
Any arguments passed to the C<send> call will be returned by all |
391 |
future C<< ->wait >> calls. |
392 |
|
393 |
=item $cv->croak ($error) |
394 |
|
395 |
Similar to send, but causes all call's wait C<< ->wait >> to invoke |
396 |
C<Carp::croak> with the given error message/object/scalar. |
397 |
|
398 |
This can be used to signal any errors to the condition variable |
399 |
user/consumer. |
400 |
|
401 |
=item $cv->begin ([group callback]) |
402 |
|
403 |
=item $cv->end |
404 |
|
405 |
These two methods can be used to combine many transactions/events into |
406 |
one. For example, a function that pings many hosts in parallel might want |
407 |
to use a condition variable for the whole process. |
408 |
|
409 |
Every call to C<< ->begin >> will increment a counter, and every call to |
410 |
C<< ->end >> will decrement it. If the counter reaches C<0> in C<< ->end |
411 |
>>, the (last) callback passed to C<begin> will be executed. That callback |
412 |
is I<supposed> to call C<< ->send >>, but that is not required. If no |
413 |
callback was set, C<send> will be called without any arguments. |
414 |
|
415 |
Let's clarify this with the ping example: |
416 |
|
417 |
my $cv = AnyEvent->condvar; |
418 |
|
419 |
my %result; |
420 |
$cv->begin (sub { $cv->send (\%result) }); |
421 |
|
422 |
for my $host (@list_of_hosts) { |
423 |
$cv->begin; |
424 |
ping_host_then_call_callback $host, sub { |
425 |
$result{$host} = ...; |
426 |
$cv->end; |
427 |
}; |
428 |
} |
429 |
|
430 |
$cv->end; |
431 |
|
432 |
This code fragment supposedly pings a number of hosts and calls |
433 |
C<send> after results for all then have have been gathered - in any |
434 |
order. To achieve this, the code issues a call to C<begin> when it starts |
435 |
each ping request and calls C<end> when it has received some result for |
436 |
it. Since C<begin> and C<end> only maintain a counter, the order in which |
437 |
results arrive is not relevant. |
438 |
|
439 |
There is an additional bracketing call to C<begin> and C<end> outside the |
440 |
loop, which serves two important purposes: first, it sets the callback |
441 |
to be called once the counter reaches C<0>, and second, it ensures that |
442 |
C<send> is called even when C<no> hosts are being pinged (the loop |
443 |
doesn't execute once). |
444 |
|
445 |
This is the general pattern when you "fan out" into multiple subrequests: |
446 |
use an outer C<begin>/C<end> pair to set the callback and ensure C<end> |
447 |
is called at least once, and then, for each subrequest you start, call |
448 |
C<begin> and for eahc subrequest you finish, call C<end>. |
449 |
|
450 |
=back |
451 |
|
452 |
=head3 METHODS FOR CONSUMERS |
453 |
|
454 |
These methods should only be used by the consuming side, i.e. the |
455 |
code awaits the condition. |
456 |
|
457 |
=over 4 |
458 |
|
459 |
=item $cv->wait |
460 |
|
461 |
Wait (blocking if necessary) until the C<< ->send >> or C<< ->croak |
462 |
>> methods have been called on c<$cv>, while servicing other watchers |
463 |
normally. |
464 |
|
465 |
You can only wait once on a condition - additional calls are valid but |
466 |
will return immediately. |
467 |
|
468 |
If an error condition has been set by calling C<< ->croak >>, then this |
469 |
function will call C<croak>. |
470 |
|
471 |
In list context, all parameters passed to C<send> will be returned, |
472 |
in scalar context only the first one will be returned. |
473 |
|
474 |
Not all event models support a blocking wait - some die in that case |
475 |
(programs might want to do that to stay interactive), so I<if you are |
476 |
using this from a module, never require a blocking wait>, but let the |
477 |
caller decide whether the call will block or not (for example, by coupling |
478 |
condition variables with some kind of request results and supporting |
479 |
callbacks so the caller knows that getting the result will not block, |
480 |
while still suppporting blocking waits if the caller so desires). |
481 |
|
482 |
Another reason I<never> to C<< ->wait >> in a module is that you cannot |
483 |
sensibly have two C<< ->wait >>'s in parallel, as that would require |
484 |
multiple interpreters or coroutines/threads, none of which C<AnyEvent> |
485 |
can supply. |
486 |
|
487 |
The L<Coro> module, however, I<can> and I<does> supply coroutines and, in |
488 |
fact, L<Coro::AnyEvent> replaces AnyEvent's condvars by coroutine-safe |
489 |
versions and also integrates coroutines into AnyEvent, making blocking |
490 |
C<< ->wait >> calls perfectly safe as long as they are done from another |
491 |
coroutine (one that doesn't run the event loop). |
492 |
|
493 |
You can ensure that C<< -wait >> never blocks by setting a callback and |
494 |
only calling C<< ->wait >> from within that callback (or at a later |
495 |
time). This will work even when the event loop does not support blocking |
496 |
waits otherwise. |
497 |
|
498 |
=item $bool = $cv->ready |
499 |
|
500 |
Returns true when the condition is "true", i.e. whether C<send> or |
501 |
C<croak> have been called. |
502 |
|
503 |
=item $cb = $cv->cb ([new callback]) |
504 |
|
505 |
This is a mutator function that returns the callback set and optionally |
506 |
replaces it before doing so. |
507 |
|
508 |
The callback will be called when the condition becomes "true", i.e. when |
509 |
C<send> or C<croak> are called. Calling C<wait> inside the callback |
510 |
or at any later time is guaranteed not to block. |
511 |
|
512 |
=back |
513 |
|
514 |
=head1 GLOBAL VARIABLES AND FUNCTIONS |
515 |
|
516 |
=over 4 |
517 |
|
518 |
=item $AnyEvent::MODEL |
519 |
|
520 |
Contains C<undef> until the first watcher is being created. Then it |
521 |
contains the event model that is being used, which is the name of the |
522 |
Perl class implementing the model. This class is usually one of the |
523 |
C<AnyEvent::Impl:xxx> modules, but can be any other class in the case |
524 |
AnyEvent has been extended at runtime (e.g. in I<rxvt-unicode>). |
525 |
|
526 |
The known classes so far are: |
527 |
|
528 |
AnyEvent::Impl::EV based on EV (an interface to libev, best choice). |
529 |
AnyEvent::Impl::Event based on Event, second best choice. |
530 |
AnyEvent::Impl::Perl pure-perl implementation, fast and portable. |
531 |
AnyEvent::Impl::Glib based on Glib, third-best choice. |
532 |
AnyEvent::Impl::Tk based on Tk, very bad choice. |
533 |
AnyEvent::Impl::Qt based on Qt, cannot be autoprobed (see its docs). |
534 |
AnyEvent::Impl::EventLib based on Event::Lib, leaks memory and worse. |
535 |
AnyEvent::Impl::POE based on POE, not generic enough for full support. |
536 |
|
537 |
There is no support for WxWidgets, as WxWidgets has no support for |
538 |
watching file handles. However, you can use WxWidgets through the |
539 |
POE Adaptor, as POE has a Wx backend that simply polls 20 times per |
540 |
second, which was considered to be too horrible to even consider for |
541 |
AnyEvent. Likewise, other POE backends can be used by AnyEvent by using |
542 |
it's adaptor. |
543 |
|
544 |
AnyEvent knows about L<Prima> and L<Wx> and will try to use L<POE> when |
545 |
autodetecting them. |
546 |
|
547 |
=item AnyEvent::detect |
548 |
|
549 |
Returns C<$AnyEvent::MODEL>, forcing autodetection of the event model |
550 |
if necessary. You should only call this function right before you would |
551 |
have created an AnyEvent watcher anyway, that is, as late as possible at |
552 |
runtime. |
553 |
|
554 |
=item @AnyEvent::detect |
555 |
|
556 |
If there are any code references in this array (you can C<push> to it |
557 |
before or after loading AnyEvent), then they will called directly after |
558 |
the event loop has been chosen. |
559 |
|
560 |
You should check C<$AnyEvent::MODEL> before adding to this array, though: |
561 |
if it contains a true value then the event loop has already been detected, |
562 |
and the array will be ignored. |
563 |
|
564 |
=back |
565 |
|
566 |
=head1 WHAT TO DO IN A MODULE |
567 |
|
568 |
As a module author, you should C<use AnyEvent> and call AnyEvent methods |
569 |
freely, but you should not load a specific event module or rely on it. |
570 |
|
571 |
Be careful when you create watchers in the module body - AnyEvent will |
572 |
decide which event module to use as soon as the first method is called, so |
573 |
by calling AnyEvent in your module body you force the user of your module |
574 |
to load the event module first. |
575 |
|
576 |
Never call C<< ->wait >> on a condition variable unless you I<know> that |
577 |
the C<< ->send >> method has been called on it already. This is |
578 |
because it will stall the whole program, and the whole point of using |
579 |
events is to stay interactive. |
580 |
|
581 |
It is fine, however, to call C<< ->wait >> when the user of your module |
582 |
requests it (i.e. if you create a http request object ad have a method |
583 |
called C<results> that returns the results, it should call C<< ->wait >> |
584 |
freely, as the user of your module knows what she is doing. always). |
585 |
|
586 |
=head1 WHAT TO DO IN THE MAIN PROGRAM |
587 |
|
588 |
There will always be a single main program - the only place that should |
589 |
dictate which event model to use. |
590 |
|
591 |
If it doesn't care, it can just "use AnyEvent" and use it itself, or not |
592 |
do anything special (it does not need to be event-based) and let AnyEvent |
593 |
decide which implementation to chose if some module relies on it. |
594 |
|
595 |
If the main program relies on a specific event model. For example, in |
596 |
Gtk2 programs you have to rely on the Glib module. You should load the |
597 |
event module before loading AnyEvent or any module that uses it: generally |
598 |
speaking, you should load it as early as possible. The reason is that |
599 |
modules might create watchers when they are loaded, and AnyEvent will |
600 |
decide on the event model to use as soon as it creates watchers, and it |
601 |
might chose the wrong one unless you load the correct one yourself. |
602 |
|
603 |
You can chose to use a rather inefficient pure-perl implementation by |
604 |
loading the C<AnyEvent::Impl::Perl> module, which gives you similar |
605 |
behaviour everywhere, but letting AnyEvent chose is generally better. |
606 |
|
607 |
=head1 OTHER MODULES |
608 |
|
609 |
The following is a non-exhaustive list of additional modules that use |
610 |
AnyEvent and can therefore be mixed easily with other AnyEvent modules |
611 |
in the same program. Some of the modules come with AnyEvent, some are |
612 |
available via CPAN. |
613 |
|
614 |
=over 4 |
615 |
|
616 |
=item L<AnyEvent::Util> |
617 |
|
618 |
Contains various utility functions that replace often-used but blocking |
619 |
functions such as C<inet_aton> by event-/callback-based versions. |
620 |
|
621 |
=item L<AnyEvent::Handle> |
622 |
|
623 |
Provide read and write buffers and manages watchers for reads and writes. |
624 |
|
625 |
=item L<AnyEvent::Socket> |
626 |
|
627 |
Provides a means to do non-blocking connects, accepts etc. |
628 |
|
629 |
=item L<AnyEvent::HTTPD> |
630 |
|
631 |
Provides a simple web application server framework. |
632 |
|
633 |
=item L<AnyEvent::DNS> |
634 |
|
635 |
Provides asynchronous DNS resolver capabilities, beyond what |
636 |
L<AnyEvent::Util> offers. |
637 |
|
638 |
=item L<AnyEvent::FastPing> |
639 |
|
640 |
The fastest ping in the west. |
641 |
|
642 |
=item L<Net::IRC3> |
643 |
|
644 |
AnyEvent based IRC client module family. |
645 |
|
646 |
=item L<Net::XMPP2> |
647 |
|
648 |
AnyEvent based XMPP (Jabber protocol) module family. |
649 |
|
650 |
=item L<Net::FCP> |
651 |
|
652 |
AnyEvent-based implementation of the Freenet Client Protocol, birthplace |
653 |
of AnyEvent. |
654 |
|
655 |
=item L<Event::ExecFlow> |
656 |
|
657 |
High level API for event-based execution flow control. |
658 |
|
659 |
=item L<Coro> |
660 |
|
661 |
Has special support for AnyEvent via L<Coro::AnyEvent>. |
662 |
|
663 |
=item L<IO::Lambda> |
664 |
|
665 |
The lambda approach to I/O - don't ask, look there. Can use AnyEvent. |
666 |
|
667 |
=item L<IO::AIO> |
668 |
|
669 |
Truly asynchronous I/O, should be in the toolbox of every event |
670 |
programmer. Can be trivially made to use AnyEvent. |
671 |
|
672 |
=item L<BDB> |
673 |
|
674 |
Truly asynchronous Berkeley DB access. Can be trivially made to use |
675 |
AnyEvent. |
676 |
|
677 |
=back |
678 |
|
679 |
=cut |
680 |
|
681 |
package AnyEvent; |
682 |
|
683 |
no warnings; |
684 |
use strict; |
685 |
|
686 |
use Carp; |
687 |
|
688 |
our $VERSION = '3.4'; |
689 |
our $MODEL; |
690 |
|
691 |
our $AUTOLOAD; |
692 |
our @ISA; |
693 |
|
694 |
our $verbose = $ENV{PERL_ANYEVENT_VERBOSE}*1; |
695 |
|
696 |
our @REGISTRY; |
697 |
|
698 |
my @models = ( |
699 |
[EV:: => AnyEvent::Impl::EV::], |
700 |
[Event:: => AnyEvent::Impl::Event::], |
701 |
[Tk:: => AnyEvent::Impl::Tk::], |
702 |
[Wx:: => AnyEvent::Impl::POE::], |
703 |
[Prima:: => AnyEvent::Impl::POE::], |
704 |
[AnyEvent::Impl::Perl:: => AnyEvent::Impl::Perl::], |
705 |
# everything below here will not be autoprobed as the pureperl backend should work everywhere |
706 |
[Glib:: => AnyEvent::Impl::Glib::], |
707 |
[Event::Lib:: => AnyEvent::Impl::EventLib::], # too buggy |
708 |
[Qt:: => AnyEvent::Impl::Qt::], # requires special main program |
709 |
[POE::Kernel:: => AnyEvent::Impl::POE::], # lasciate ogni speranza |
710 |
); |
711 |
|
712 |
our %method = map +($_ => 1), qw(io timer signal child condvar one_event DESTROY); |
713 |
|
714 |
our @detect; |
715 |
|
716 |
sub detect() { |
717 |
unless ($MODEL) { |
718 |
no strict 'refs'; |
719 |
|
720 |
if ($ENV{PERL_ANYEVENT_MODEL} =~ /^([a-zA-Z]+)$/) { |
721 |
my $model = "AnyEvent::Impl::$1"; |
722 |
if (eval "require $model") { |
723 |
$MODEL = $model; |
724 |
warn "AnyEvent: loaded model '$model' (forced by \$PERL_ANYEVENT_MODEL), using it.\n" if $verbose > 1; |
725 |
} else { |
726 |
warn "AnyEvent: unable to load model '$model' (from \$PERL_ANYEVENT_MODEL):\n$@" if $verbose; |
727 |
} |
728 |
} |
729 |
|
730 |
# check for already loaded models |
731 |
unless ($MODEL) { |
732 |
for (@REGISTRY, @models) { |
733 |
my ($package, $model) = @$_; |
734 |
if (${"$package\::VERSION"} > 0) { |
735 |
if (eval "require $model") { |
736 |
$MODEL = $model; |
737 |
warn "AnyEvent: autodetected model '$model', using it.\n" if $verbose > 1; |
738 |
last; |
739 |
} |
740 |
} |
741 |
} |
742 |
|
743 |
unless ($MODEL) { |
744 |
# try to load a model |
745 |
|
746 |
for (@REGISTRY, @models) { |
747 |
my ($package, $model) = @$_; |
748 |
if (eval "require $package" |
749 |
and ${"$package\::VERSION"} > 0 |
750 |
and eval "require $model") { |
751 |
$MODEL = $model; |
752 |
warn "AnyEvent: autoprobed model '$model', using it.\n" if $verbose > 1; |
753 |
last; |
754 |
} |
755 |
} |
756 |
|
757 |
$MODEL |
758 |
or die "No event module selected for AnyEvent and autodetect failed. Install any one of these modules: EV, Event or Glib."; |
759 |
} |
760 |
} |
761 |
|
762 |
unshift @ISA, $MODEL; |
763 |
push @{"$MODEL\::ISA"}, "AnyEvent::Base"; |
764 |
|
765 |
(shift @detect)->() while @detect; |
766 |
} |
767 |
|
768 |
$MODEL |
769 |
} |
770 |
|
771 |
sub AUTOLOAD { |
772 |
(my $func = $AUTOLOAD) =~ s/.*://; |
773 |
|
774 |
$method{$func} |
775 |
or croak "$func: not a valid method for AnyEvent objects"; |
776 |
|
777 |
detect unless $MODEL; |
778 |
|
779 |
my $class = shift; |
780 |
$class->$func (@_); |
781 |
} |
782 |
|
783 |
package AnyEvent::Base; |
784 |
|
785 |
# default implementation for ->condvar, ->wait, ->broadcast |
786 |
|
787 |
sub condvar { |
788 |
bless \my $flag, "AnyEvent::Base::CondVar" |
789 |
} |
790 |
|
791 |
sub AnyEvent::Base::CondVar::broadcast { |
792 |
${$_[0]}++; |
793 |
} |
794 |
|
795 |
sub AnyEvent::Base::CondVar::wait { |
796 |
AnyEvent->one_event while !${$_[0]}; |
797 |
} |
798 |
|
799 |
# default implementation for ->signal |
800 |
|
801 |
our %SIG_CB; |
802 |
|
803 |
sub signal { |
804 |
my (undef, %arg) = @_; |
805 |
|
806 |
my $signal = uc $arg{signal} |
807 |
or Carp::croak "required option 'signal' is missing"; |
808 |
|
809 |
$SIG_CB{$signal}{$arg{cb}} = $arg{cb}; |
810 |
$SIG{$signal} ||= sub { |
811 |
$_->() for values %{ $SIG_CB{$signal} || {} }; |
812 |
}; |
813 |
|
814 |
bless [$signal, $arg{cb}], "AnyEvent::Base::Signal" |
815 |
} |
816 |
|
817 |
sub AnyEvent::Base::Signal::DESTROY { |
818 |
my ($signal, $cb) = @{$_[0]}; |
819 |
|
820 |
delete $SIG_CB{$signal}{$cb}; |
821 |
|
822 |
$SIG{$signal} = 'DEFAULT' unless keys %{ $SIG_CB{$signal} }; |
823 |
} |
824 |
|
825 |
# default implementation for ->child |
826 |
|
827 |
our %PID_CB; |
828 |
our $CHLD_W; |
829 |
our $CHLD_DELAY_W; |
830 |
our $PID_IDLE; |
831 |
our $WNOHANG; |
832 |
|
833 |
sub _child_wait { |
834 |
while (0 < (my $pid = waitpid -1, $WNOHANG)) { |
835 |
$_->($pid, $?) for (values %{ $PID_CB{$pid} || {} }), |
836 |
(values %{ $PID_CB{0} || {} }); |
837 |
} |
838 |
|
839 |
undef $PID_IDLE; |
840 |
} |
841 |
|
842 |
sub _sigchld { |
843 |
# make sure we deliver these changes "synchronous" with the event loop. |
844 |
$CHLD_DELAY_W ||= AnyEvent->timer (after => 0, cb => sub { |
845 |
undef $CHLD_DELAY_W; |
846 |
&_child_wait; |
847 |
}); |
848 |
} |
849 |
|
850 |
sub child { |
851 |
my (undef, %arg) = @_; |
852 |
|
853 |
defined (my $pid = $arg{pid} + 0) |
854 |
or Carp::croak "required option 'pid' is missing"; |
855 |
|
856 |
$PID_CB{$pid}{$arg{cb}} = $arg{cb}; |
857 |
|
858 |
unless ($WNOHANG) { |
859 |
$WNOHANG = eval { require POSIX; &POSIX::WNOHANG } || 1; |
860 |
} |
861 |
|
862 |
unless ($CHLD_W) { |
863 |
$CHLD_W = AnyEvent->signal (signal => 'CHLD', cb => \&_sigchld); |
864 |
# child could be a zombie already, so make at least one round |
865 |
&_sigchld; |
866 |
} |
867 |
|
868 |
bless [$pid, $arg{cb}], "AnyEvent::Base::Child" |
869 |
} |
870 |
|
871 |
sub AnyEvent::Base::Child::DESTROY { |
872 |
my ($pid, $cb) = @{$_[0]}; |
873 |
|
874 |
delete $PID_CB{$pid}{$cb}; |
875 |
delete $PID_CB{$pid} unless keys %{ $PID_CB{$pid} }; |
876 |
|
877 |
undef $CHLD_W unless keys %PID_CB; |
878 |
} |
879 |
|
880 |
=head1 SUPPLYING YOUR OWN EVENT MODEL INTERFACE |
881 |
|
882 |
This is an advanced topic that you do not normally need to use AnyEvent in |
883 |
a module. This section is only of use to event loop authors who want to |
884 |
provide AnyEvent compatibility. |
885 |
|
886 |
If you need to support another event library which isn't directly |
887 |
supported by AnyEvent, you can supply your own interface to it by |
888 |
pushing, before the first watcher gets created, the package name of |
889 |
the event module and the package name of the interface to use onto |
890 |
C<@AnyEvent::REGISTRY>. You can do that before and even without loading |
891 |
AnyEvent, so it is reasonably cheap. |
892 |
|
893 |
Example: |
894 |
|
895 |
push @AnyEvent::REGISTRY, [urxvt => urxvt::anyevent::]; |
896 |
|
897 |
This tells AnyEvent to (literally) use the C<urxvt::anyevent::> |
898 |
package/class when it finds the C<urxvt> package/module is already loaded. |
899 |
|
900 |
When AnyEvent is loaded and asked to find a suitable event model, it |
901 |
will first check for the presence of urxvt by trying to C<use> the |
902 |
C<urxvt::anyevent> module. |
903 |
|
904 |
The class should provide implementations for all watcher types. See |
905 |
L<AnyEvent::Impl::EV> (source code), L<AnyEvent::Impl::Glib> (Source code) |
906 |
and so on for actual examples. Use C<perldoc -m AnyEvent::Impl::Glib> to |
907 |
see the sources. |
908 |
|
909 |
If you don't provide C<signal> and C<child> watchers than AnyEvent will |
910 |
provide suitable (hopefully) replacements. |
911 |
|
912 |
The above example isn't fictitious, the I<rxvt-unicode> (a.k.a. urxvt) |
913 |
terminal emulator uses the above line as-is. An interface isn't included |
914 |
in AnyEvent because it doesn't make sense outside the embedded interpreter |
915 |
inside I<rxvt-unicode>, and it is updated and maintained as part of the |
916 |
I<rxvt-unicode> distribution. |
917 |
|
918 |
I<rxvt-unicode> also cheats a bit by not providing blocking access to |
919 |
condition variables: code blocking while waiting for a condition will |
920 |
C<die>. This still works with most modules/usages, and blocking calls must |
921 |
not be done in an interactive application, so it makes sense. |
922 |
|
923 |
=head1 ENVIRONMENT VARIABLES |
924 |
|
925 |
The following environment variables are used by this module: |
926 |
|
927 |
=over 4 |
928 |
|
929 |
=item C<PERL_ANYEVENT_VERBOSE> |
930 |
|
931 |
By default, AnyEvent will be completely silent except in fatal |
932 |
conditions. You can set this environment variable to make AnyEvent more |
933 |
talkative. |
934 |
|
935 |
When set to C<1> or higher, causes AnyEvent to warn about unexpected |
936 |
conditions, such as not being able to load the event model specified by |
937 |
C<PERL_ANYEVENT_MODEL>. |
938 |
|
939 |
When set to C<2> or higher, cause AnyEvent to report to STDERR which event |
940 |
model it chooses. |
941 |
|
942 |
=item C<PERL_ANYEVENT_MODEL> |
943 |
|
944 |
This can be used to specify the event model to be used by AnyEvent, before |
945 |
autodetection and -probing kicks in. It must be a string consisting |
946 |
entirely of ASCII letters. The string C<AnyEvent::Impl::> gets prepended |
947 |
and the resulting module name is loaded and if the load was successful, |
948 |
used as event model. If it fails to load AnyEvent will proceed with |
949 |
autodetection and -probing. |
950 |
|
951 |
This functionality might change in future versions. |
952 |
|
953 |
For example, to force the pure perl model (L<AnyEvent::Impl::Perl>) you |
954 |
could start your program like this: |
955 |
|
956 |
PERL_ANYEVENT_MODEL=Perl perl ... |
957 |
|
958 |
=back |
959 |
|
960 |
=head1 EXAMPLE PROGRAM |
961 |
|
962 |
The following program uses an I/O watcher to read data from STDIN, a timer |
963 |
to display a message once per second, and a condition variable to quit the |
964 |
program when the user enters quit: |
965 |
|
966 |
use AnyEvent; |
967 |
|
968 |
my $cv = AnyEvent->condvar; |
969 |
|
970 |
my $io_watcher = AnyEvent->io ( |
971 |
fh => \*STDIN, |
972 |
poll => 'r', |
973 |
cb => sub { |
974 |
warn "io event <$_[0]>\n"; # will always output <r> |
975 |
chomp (my $input = <STDIN>); # read a line |
976 |
warn "read: $input\n"; # output what has been read |
977 |
$cv->broadcast if $input =~ /^q/i; # quit program if /^q/i |
978 |
}, |
979 |
); |
980 |
|
981 |
my $time_watcher; # can only be used once |
982 |
|
983 |
sub new_timer { |
984 |
$timer = AnyEvent->timer (after => 1, cb => sub { |
985 |
warn "timeout\n"; # print 'timeout' about every second |
986 |
&new_timer; # and restart the time |
987 |
}); |
988 |
} |
989 |
|
990 |
new_timer; # create first timer |
991 |
|
992 |
$cv->wait; # wait until user enters /^q/i |
993 |
|
994 |
=head1 REAL-WORLD EXAMPLE |
995 |
|
996 |
Consider the L<Net::FCP> module. It features (among others) the following |
997 |
API calls, which are to freenet what HTTP GET requests are to http: |
998 |
|
999 |
my $data = $fcp->client_get ($url); # blocks |
1000 |
|
1001 |
my $transaction = $fcp->txn_client_get ($url); # does not block |
1002 |
$transaction->cb ( sub { ... } ); # set optional result callback |
1003 |
my $data = $transaction->result; # possibly blocks |
1004 |
|
1005 |
The C<client_get> method works like C<LWP::Simple::get>: it requests the |
1006 |
given URL and waits till the data has arrived. It is defined to be: |
1007 |
|
1008 |
sub client_get { $_[0]->txn_client_get ($_[1])->result } |
1009 |
|
1010 |
And in fact is automatically generated. This is the blocking API of |
1011 |
L<Net::FCP>, and it works as simple as in any other, similar, module. |
1012 |
|
1013 |
More complicated is C<txn_client_get>: It only creates a transaction |
1014 |
(completion, result, ...) object and initiates the transaction. |
1015 |
|
1016 |
my $txn = bless { }, Net::FCP::Txn::; |
1017 |
|
1018 |
It also creates a condition variable that is used to signal the completion |
1019 |
of the request: |
1020 |
|
1021 |
$txn->{finished} = AnyAvent->condvar; |
1022 |
|
1023 |
It then creates a socket in non-blocking mode. |
1024 |
|
1025 |
socket $txn->{fh}, ...; |
1026 |
fcntl $txn->{fh}, F_SETFL, O_NONBLOCK; |
1027 |
connect $txn->{fh}, ... |
1028 |
and !$!{EWOULDBLOCK} |
1029 |
and !$!{EINPROGRESS} |
1030 |
and Carp::croak "unable to connect: $!\n"; |
1031 |
|
1032 |
Then it creates a write-watcher which gets called whenever an error occurs |
1033 |
or the connection succeeds: |
1034 |
|
1035 |
$txn->{w} = AnyEvent->io (fh => $txn->{fh}, poll => 'w', cb => sub { $txn->fh_ready_w }); |
1036 |
|
1037 |
And returns this transaction object. The C<fh_ready_w> callback gets |
1038 |
called as soon as the event loop detects that the socket is ready for |
1039 |
writing. |
1040 |
|
1041 |
The C<fh_ready_w> method makes the socket blocking again, writes the |
1042 |
request data and replaces the watcher by a read watcher (waiting for reply |
1043 |
data). The actual code is more complicated, but that doesn't matter for |
1044 |
this example: |
1045 |
|
1046 |
fcntl $txn->{fh}, F_SETFL, 0; |
1047 |
syswrite $txn->{fh}, $txn->{request} |
1048 |
or die "connection or write error"; |
1049 |
$txn->{w} = AnyEvent->io (fh => $txn->{fh}, poll => 'r', cb => sub { $txn->fh_ready_r }); |
1050 |
|
1051 |
Again, C<fh_ready_r> waits till all data has arrived, and then stores the |
1052 |
result and signals any possible waiters that the request ahs finished: |
1053 |
|
1054 |
sysread $txn->{fh}, $txn->{buf}, length $txn->{$buf}; |
1055 |
|
1056 |
if (end-of-file or data complete) { |
1057 |
$txn->{result} = $txn->{buf}; |
1058 |
$txn->{finished}->broadcast; |
1059 |
$txb->{cb}->($txn) of $txn->{cb}; # also call callback |
1060 |
} |
1061 |
|
1062 |
The C<result> method, finally, just waits for the finished signal (if the |
1063 |
request was already finished, it doesn't wait, of course, and returns the |
1064 |
data: |
1065 |
|
1066 |
$txn->{finished}->wait; |
1067 |
return $txn->{result}; |
1068 |
|
1069 |
The actual code goes further and collects all errors (C<die>s, exceptions) |
1070 |
that occured during request processing. The C<result> method detects |
1071 |
whether an exception as thrown (it is stored inside the $txn object) |
1072 |
and just throws the exception, which means connection errors and other |
1073 |
problems get reported tot he code that tries to use the result, not in a |
1074 |
random callback. |
1075 |
|
1076 |
All of this enables the following usage styles: |
1077 |
|
1078 |
1. Blocking: |
1079 |
|
1080 |
my $data = $fcp->client_get ($url); |
1081 |
|
1082 |
2. Blocking, but running in parallel: |
1083 |
|
1084 |
my @datas = map $_->result, |
1085 |
map $fcp->txn_client_get ($_), |
1086 |
@urls; |
1087 |
|
1088 |
Both blocking examples work without the module user having to know |
1089 |
anything about events. |
1090 |
|
1091 |
3a. Event-based in a main program, using any supported event module: |
1092 |
|
1093 |
use EV; |
1094 |
|
1095 |
$fcp->txn_client_get ($url)->cb (sub { |
1096 |
my $txn = shift; |
1097 |
my $data = $txn->result; |
1098 |
... |
1099 |
}); |
1100 |
|
1101 |
EV::loop; |
1102 |
|
1103 |
3b. The module user could use AnyEvent, too: |
1104 |
|
1105 |
use AnyEvent; |
1106 |
|
1107 |
my $quit = AnyEvent->condvar; |
1108 |
|
1109 |
$fcp->txn_client_get ($url)->cb (sub { |
1110 |
... |
1111 |
$quit->broadcast; |
1112 |
}); |
1113 |
|
1114 |
$quit->wait; |
1115 |
|
1116 |
|
1117 |
=head1 BENCHMARKS |
1118 |
|
1119 |
To give you an idea of the performance and overheads that AnyEvent adds |
1120 |
over the event loops themselves and to give you an impression of the speed |
1121 |
of various event loops I prepared some benchmarks. |
1122 |
|
1123 |
=head2 BENCHMARKING ANYEVENT OVERHEAD |
1124 |
|
1125 |
Here is a benchmark of various supported event models used natively and |
1126 |
through anyevent. The benchmark creates a lot of timers (with a zero |
1127 |
timeout) and I/O watchers (watching STDOUT, a pty, to become writable, |
1128 |
which it is), lets them fire exactly once and destroys them again. |
1129 |
|
1130 |
Source code for this benchmark is found as F<eg/bench> in the AnyEvent |
1131 |
distribution. |
1132 |
|
1133 |
=head3 Explanation of the columns |
1134 |
|
1135 |
I<watcher> is the number of event watchers created/destroyed. Since |
1136 |
different event models feature vastly different performances, each event |
1137 |
loop was given a number of watchers so that overall runtime is acceptable |
1138 |
and similar between tested event loop (and keep them from crashing): Glib |
1139 |
would probably take thousands of years if asked to process the same number |
1140 |
of watchers as EV in this benchmark. |
1141 |
|
1142 |
I<bytes> is the number of bytes (as measured by the resident set size, |
1143 |
RSS) consumed by each watcher. This method of measuring captures both C |
1144 |
and Perl-based overheads. |
1145 |
|
1146 |
I<create> is the time, in microseconds (millionths of seconds), that it |
1147 |
takes to create a single watcher. The callback is a closure shared between |
1148 |
all watchers, to avoid adding memory overhead. That means closure creation |
1149 |
and memory usage is not included in the figures. |
1150 |
|
1151 |
I<invoke> is the time, in microseconds, used to invoke a simple |
1152 |
callback. The callback simply counts down a Perl variable and after it was |
1153 |
invoked "watcher" times, it would C<< ->broadcast >> a condvar once to |
1154 |
signal the end of this phase. |
1155 |
|
1156 |
I<destroy> is the time, in microseconds, that it takes to destroy a single |
1157 |
watcher. |
1158 |
|
1159 |
=head3 Results |
1160 |
|
1161 |
name watchers bytes create invoke destroy comment |
1162 |
EV/EV 400000 244 0.56 0.46 0.31 EV native interface |
1163 |
EV/Any 100000 244 2.50 0.46 0.29 EV + AnyEvent watchers |
1164 |
CoroEV/Any 100000 244 2.49 0.44 0.29 coroutines + Coro::Signal |
1165 |
Perl/Any 100000 513 4.92 0.87 1.12 pure perl implementation |
1166 |
Event/Event 16000 516 31.88 31.30 0.85 Event native interface |
1167 |
Event/Any 16000 590 35.75 31.42 1.08 Event + AnyEvent watchers |
1168 |
Glib/Any 16000 1357 98.22 12.41 54.00 quadratic behaviour |
1169 |
Tk/Any 2000 1860 26.97 67.98 14.00 SEGV with >> 2000 watchers |
1170 |
POE/Event 2000 6644 108.64 736.02 14.73 via POE::Loop::Event |
1171 |
POE/Select 2000 6343 94.13 809.12 565.96 via POE::Loop::Select |
1172 |
|
1173 |
=head3 Discussion |
1174 |
|
1175 |
The benchmark does I<not> measure scalability of the event loop very |
1176 |
well. For example, a select-based event loop (such as the pure perl one) |
1177 |
can never compete with an event loop that uses epoll when the number of |
1178 |
file descriptors grows high. In this benchmark, all events become ready at |
1179 |
the same time, so select/poll-based implementations get an unnatural speed |
1180 |
boost. |
1181 |
|
1182 |
Also, note that the number of watchers usually has a nonlinear effect on |
1183 |
overall speed, that is, creating twice as many watchers doesn't take twice |
1184 |
the time - usually it takes longer. This puts event loops tested with a |
1185 |
higher number of watchers at a disadvantage. |
1186 |
|
1187 |
To put the range of results into perspective, consider that on the |
1188 |
benchmark machine, handling an event takes roughly 1600 CPU cycles with |
1189 |
EV, 3100 CPU cycles with AnyEvent's pure perl loop and almost 3000000 CPU |
1190 |
cycles with POE. |
1191 |
|
1192 |
C<EV> is the sole leader regarding speed and memory use, which are both |
1193 |
maximal/minimal, respectively. Even when going through AnyEvent, it uses |
1194 |
far less memory than any other event loop and is still faster than Event |
1195 |
natively. |
1196 |
|
1197 |
The pure perl implementation is hit in a few sweet spots (both the |
1198 |
constant timeout and the use of a single fd hit optimisations in the perl |
1199 |
interpreter and the backend itself). Nevertheless this shows that it |
1200 |
adds very little overhead in itself. Like any select-based backend its |
1201 |
performance becomes really bad with lots of file descriptors (and few of |
1202 |
them active), of course, but this was not subject of this benchmark. |
1203 |
|
1204 |
The C<Event> module has a relatively high setup and callback invocation |
1205 |
cost, but overall scores in on the third place. |
1206 |
|
1207 |
C<Glib>'s memory usage is quite a bit higher, but it features a |
1208 |
faster callback invocation and overall ends up in the same class as |
1209 |
C<Event>. However, Glib scales extremely badly, doubling the number of |
1210 |
watchers increases the processing time by more than a factor of four, |
1211 |
making it completely unusable when using larger numbers of watchers |
1212 |
(note that only a single file descriptor was used in the benchmark, so |
1213 |
inefficiencies of C<poll> do not account for this). |
1214 |
|
1215 |
The C<Tk> adaptor works relatively well. The fact that it crashes with |
1216 |
more than 2000 watchers is a big setback, however, as correctness takes |
1217 |
precedence over speed. Nevertheless, its performance is surprising, as the |
1218 |
file descriptor is dup()ed for each watcher. This shows that the dup() |
1219 |
employed by some adaptors is not a big performance issue (it does incur a |
1220 |
hidden memory cost inside the kernel which is not reflected in the figures |
1221 |
above). |
1222 |
|
1223 |
C<POE>, regardless of underlying event loop (whether using its pure perl |
1224 |
select-based backend or the Event module, the POE-EV backend couldn't |
1225 |
be tested because it wasn't working) shows abysmal performance and |
1226 |
memory usage with AnyEvent: Watchers use almost 30 times as much memory |
1227 |
as EV watchers, and 10 times as much memory as Event (the high memory |
1228 |
requirements are caused by requiring a session for each watcher). Watcher |
1229 |
invocation speed is almost 900 times slower than with AnyEvent's pure perl |
1230 |
implementation. |
1231 |
|
1232 |
The design of the POE adaptor class in AnyEvent can not really account |
1233 |
for the performance issues, though, as session creation overhead is |
1234 |
small compared to execution of the state machine, which is coded pretty |
1235 |
optimally within L<AnyEvent::Impl::POE> (and while everybody agrees that |
1236 |
using multiple sessions is not a good approach, especially regarding |
1237 |
memory usage, even the author of POE could not come up with a faster |
1238 |
design). |
1239 |
|
1240 |
=head3 Summary |
1241 |
|
1242 |
=over 4 |
1243 |
|
1244 |
=item * Using EV through AnyEvent is faster than any other event loop |
1245 |
(even when used without AnyEvent), but most event loops have acceptable |
1246 |
performance with or without AnyEvent. |
1247 |
|
1248 |
=item * The overhead AnyEvent adds is usually much smaller than the overhead of |
1249 |
the actual event loop, only with extremely fast event loops such as EV |
1250 |
adds AnyEvent significant overhead. |
1251 |
|
1252 |
=item * You should avoid POE like the plague if you want performance or |
1253 |
reasonable memory usage. |
1254 |
|
1255 |
=back |
1256 |
|
1257 |
=head2 BENCHMARKING THE LARGE SERVER CASE |
1258 |
|
1259 |
This benchmark atcually benchmarks the event loop itself. It works by |
1260 |
creating a number of "servers": each server consists of a socketpair, a |
1261 |
timeout watcher that gets reset on activity (but never fires), and an I/O |
1262 |
watcher waiting for input on one side of the socket. Each time the socket |
1263 |
watcher reads a byte it will write that byte to a random other "server". |
1264 |
|
1265 |
The effect is that there will be a lot of I/O watchers, only part of which |
1266 |
are active at any one point (so there is a constant number of active |
1267 |
fds for each loop iterstaion, but which fds these are is random). The |
1268 |
timeout is reset each time something is read because that reflects how |
1269 |
most timeouts work (and puts extra pressure on the event loops). |
1270 |
|
1271 |
In this benchmark, we use 10000 socketpairs (20000 sockets), of which 100 |
1272 |
(1%) are active. This mirrors the activity of large servers with many |
1273 |
connections, most of which are idle at any one point in time. |
1274 |
|
1275 |
Source code for this benchmark is found as F<eg/bench2> in the AnyEvent |
1276 |
distribution. |
1277 |
|
1278 |
=head3 Explanation of the columns |
1279 |
|
1280 |
I<sockets> is the number of sockets, and twice the number of "servers" (as |
1281 |
each server has a read and write socket end). |
1282 |
|
1283 |
I<create> is the time it takes to create a socketpair (which is |
1284 |
nontrivial) and two watchers: an I/O watcher and a timeout watcher. |
1285 |
|
1286 |
I<request>, the most important value, is the time it takes to handle a |
1287 |
single "request", that is, reading the token from the pipe and forwarding |
1288 |
it to another server. This includes deleting the old timeout and creating |
1289 |
a new one that moves the timeout into the future. |
1290 |
|
1291 |
=head3 Results |
1292 |
|
1293 |
name sockets create request |
1294 |
EV 20000 69.01 11.16 |
1295 |
Perl 20000 73.32 35.87 |
1296 |
Event 20000 212.62 257.32 |
1297 |
Glib 20000 651.16 1896.30 |
1298 |
POE 20000 349.67 12317.24 uses POE::Loop::Event |
1299 |
|
1300 |
=head3 Discussion |
1301 |
|
1302 |
This benchmark I<does> measure scalability and overall performance of the |
1303 |
particular event loop. |
1304 |
|
1305 |
EV is again fastest. Since it is using epoll on my system, the setup time |
1306 |
is relatively high, though. |
1307 |
|
1308 |
Perl surprisingly comes second. It is much faster than the C-based event |
1309 |
loops Event and Glib. |
1310 |
|
1311 |
Event suffers from high setup time as well (look at its code and you will |
1312 |
understand why). Callback invocation also has a high overhead compared to |
1313 |
the C<< $_->() for .. >>-style loop that the Perl event loop uses. Event |
1314 |
uses select or poll in basically all documented configurations. |
1315 |
|
1316 |
Glib is hit hard by its quadratic behaviour w.r.t. many watchers. It |
1317 |
clearly fails to perform with many filehandles or in busy servers. |
1318 |
|
1319 |
POE is still completely out of the picture, taking over 1000 times as long |
1320 |
as EV, and over 100 times as long as the Perl implementation, even though |
1321 |
it uses a C-based event loop in this case. |
1322 |
|
1323 |
=head3 Summary |
1324 |
|
1325 |
=over 4 |
1326 |
|
1327 |
=item * The pure perl implementation performs extremely well. |
1328 |
|
1329 |
=item * Avoid Glib or POE in large projects where performance matters. |
1330 |
|
1331 |
=back |
1332 |
|
1333 |
=head2 BENCHMARKING SMALL SERVERS |
1334 |
|
1335 |
While event loops should scale (and select-based ones do not...) even to |
1336 |
large servers, most programs we (or I :) actually write have only a few |
1337 |
I/O watchers. |
1338 |
|
1339 |
In this benchmark, I use the same benchmark program as in the large server |
1340 |
case, but it uses only eight "servers", of which three are active at any |
1341 |
one time. This should reflect performance for a small server relatively |
1342 |
well. |
1343 |
|
1344 |
The columns are identical to the previous table. |
1345 |
|
1346 |
=head3 Results |
1347 |
|
1348 |
name sockets create request |
1349 |
EV 16 20.00 6.54 |
1350 |
Perl 16 25.75 12.62 |
1351 |
Event 16 81.27 35.86 |
1352 |
Glib 16 32.63 15.48 |
1353 |
POE 16 261.87 276.28 uses POE::Loop::Event |
1354 |
|
1355 |
=head3 Discussion |
1356 |
|
1357 |
The benchmark tries to test the performance of a typical small |
1358 |
server. While knowing how various event loops perform is interesting, keep |
1359 |
in mind that their overhead in this case is usually not as important, due |
1360 |
to the small absolute number of watchers (that is, you need efficiency and |
1361 |
speed most when you have lots of watchers, not when you only have a few of |
1362 |
them). |
1363 |
|
1364 |
EV is again fastest. |
1365 |
|
1366 |
Perl again comes second. It is noticably faster than the C-based event |
1367 |
loops Event and Glib, although the difference is too small to really |
1368 |
matter. |
1369 |
|
1370 |
POE also performs much better in this case, but is is still far behind the |
1371 |
others. |
1372 |
|
1373 |
=head3 Summary |
1374 |
|
1375 |
=over 4 |
1376 |
|
1377 |
=item * C-based event loops perform very well with small number of |
1378 |
watchers, as the management overhead dominates. |
1379 |
|
1380 |
=back |
1381 |
|
1382 |
|
1383 |
=head1 FORK |
1384 |
|
1385 |
Most event libraries are not fork-safe. The ones who are usually are |
1386 |
because they rely on inefficient but fork-safe C<select> or C<poll> |
1387 |
calls. Only L<EV> is fully fork-aware. |
1388 |
|
1389 |
If you have to fork, you must either do so I<before> creating your first |
1390 |
watcher OR you must not use AnyEvent at all in the child. |
1391 |
|
1392 |
|
1393 |
=head1 SECURITY CONSIDERATIONS |
1394 |
|
1395 |
AnyEvent can be forced to load any event model via |
1396 |
$ENV{PERL_ANYEVENT_MODEL}. While this cannot (to my knowledge) be used to |
1397 |
execute arbitrary code or directly gain access, it can easily be used to |
1398 |
make the program hang or malfunction in subtle ways, as AnyEvent watchers |
1399 |
will not be active when the program uses a different event model than |
1400 |
specified in the variable. |
1401 |
|
1402 |
You can make AnyEvent completely ignore this variable by deleting it |
1403 |
before the first watcher gets created, e.g. with a C<BEGIN> block: |
1404 |
|
1405 |
BEGIN { delete $ENV{PERL_ANYEVENT_MODEL} } |
1406 |
|
1407 |
use AnyEvent; |
1408 |
|
1409 |
Similar considerations apply to $ENV{PERL_ANYEVENT_VERBOSE}, as that can |
1410 |
be used to probe what backend is used and gain other information (which is |
1411 |
probably even less useful to an attacker than PERL_ANYEVENT_MODEL). |
1412 |
|
1413 |
|
1414 |
=head1 SEE ALSO |
1415 |
|
1416 |
Event modules: L<EV>, L<EV::Glib>, L<Glib::EV>, L<Event>, L<Glib::Event>, |
1417 |
L<Glib>, L<Tk>, L<Event::Lib>, L<Qt>, L<POE>. |
1418 |
|
1419 |
Implementations: L<AnyEvent::Impl::EV>, L<AnyEvent::Impl::Event>, |
1420 |
L<AnyEvent::Impl::Glib>, L<AnyEvent::Impl::Tk>, L<AnyEvent::Impl::Perl>, |
1421 |
L<AnyEvent::Impl::EventLib>, L<AnyEvent::Impl::Qt>, |
1422 |
L<AnyEvent::Impl::POE>. |
1423 |
|
1424 |
Coroutine support: L<Coro>, L<Coro::AnyEvent>, L<Coro::EV>, L<Coro::Event>, |
1425 |
|
1426 |
Nontrivial usage examples: L<Net::FCP>, L<Net::XMPP2>. |
1427 |
|
1428 |
|
1429 |
=head1 AUTHOR |
1430 |
|
1431 |
Marc Lehmann <schmorp@schmorp.de> |
1432 |
http://home.schmorp.de/ |
1433 |
|
1434 |
=cut |
1435 |
|
1436 |
1 |
1437 |
|