ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/AnyEvent/README
(Generate patch)

Comparing AnyEvent/README (file contents):
Revision 1.15 by root, Wed Apr 16 15:10:10 2008 UTC vs.
Revision 1.22 by root, Sat May 24 17:58:33 2008 UTC

1NAME 1=> NAME
2 AnyEvent - provide framework for multiple event loops 2 AnyEvent - provide framework for multiple event loops
3 3
4 EV, Event, Coro::EV, Coro::Event, Glib, Tk, Perl - various supported 4 EV, Event, Glib, Tk, Perl, Event::Lib, Qt, POE - various supported event
5 event loops 5 loops
6 6
7SYNOPSIS 7SYNOPSIS
8 use AnyEvent; 8 use AnyEvent;
9 9
10 my $w = AnyEvent->io (fh => $fh, poll => "r|w", cb => sub { 10 my $w = AnyEvent->io (fh => $fh, poll => "r|w", cb => sub {
13 13
14 my $w = AnyEvent->timer (after => $seconds, cb => sub { 14 my $w = AnyEvent->timer (after => $seconds, cb => sub {
15 ... 15 ...
16 }); 16 });
17 17
18 my $w = AnyEvent->condvar; # stores wether a condition was flagged 18 my $w = AnyEvent->condvar; # stores whether a condition was flagged
19 $w->send; # wake up current and all future recv's
19 $w->wait; # enters "main loop" till $condvar gets ->broadcast 20 $w->recv; # enters "main loop" till $condvar gets ->send
20 $w->broadcast; # wake up current and all future wait's
21 21
22WHY YOU SHOULD USE THIS MODULE (OR NOT) 22WHY YOU SHOULD USE THIS MODULE (OR NOT)
23 Glib, POE, IO::Async, Event... CPAN offers event models by the dozen 23 Glib, POE, IO::Async, Event... CPAN offers event models by the dozen
24 nowadays. So what is different about AnyEvent? 24 nowadays. So what is different about AnyEvent?
25 25
27 policy* and AnyEvent is *small and efficient*. 27 policy* and AnyEvent is *small and efficient*.
28 28
29 First and foremost, *AnyEvent is not an event model* itself, it only 29 First and foremost, *AnyEvent is not an event model* itself, it only
30 interfaces to whatever event model the main program happens to use in a 30 interfaces to whatever event model the main program happens to use in a
31 pragmatic way. For event models and certain classes of immortals alike, 31 pragmatic way. For event models and certain classes of immortals alike,
32 the statement "there can only be one" is a bitter reality, and AnyEvent 32 the statement "there can only be one" is a bitter reality: In general,
33 helps hiding the differences. 33 only one event loop can be active at the same time in a process.
34 AnyEvent helps hiding the differences between those event loops.
34 35
35 The goal of AnyEvent is to offer module authors the ability to do event 36 The goal of AnyEvent is to offer module authors the ability to do event
36 programming (waiting for I/O or timer events) without subscribing to a 37 programming (waiting for I/O or timer events) without subscribing to a
37 religion, a way of living, and most importantly: without forcing your 38 religion, a way of living, and most importantly: without forcing your
38 module users into the same thing by forcing them to use the same event 39 module users into the same thing by forcing them to use the same event
39 model you use. 40 model you use.
40 41
41 For modules like POE or IO::Async (which is actually doing all I/O 42 For modules like POE or IO::Async (which is a total misnomer as it is
42 *synchronously*...), using them in your module is like joining a cult: 43 actually doing all I/O *synchronously*...), using them in your module is
43 After you joined, you are dependent on them and you cannot use anything 44 like joining a cult: After you joined, you are dependent on them and you
44 else, as it is simply incompatible to everything that isn't itself. 45 cannot use anything else, as it is simply incompatible to everything
46 that isn't itself. What's worse, all the potential users of your module
47 are *also* forced to use the same event loop you use.
45 48
46 AnyEvent + POE works fine. AnyEvent + Glib works fine. AnyEvent + Tk 49 AnyEvent is different: AnyEvent + POE works fine. AnyEvent + Glib works
47 works fine etc. etc. but none of these work together with the rest: POE 50 fine. AnyEvent + Tk works fine etc. etc. but none of these work together
48 + IO::Async? no go. Tk + Event? no go. If your module uses one of those, 51 with the rest: POE + IO::Async? no go. Tk + Event? no go. Again: if your
49 every user of your module has to use it, too. If your module uses 52 module uses one of those, every user of your module has to use it, too.
50 AnyEvent, it works transparently with all event models it supports 53 But if your module uses AnyEvent, it works transparently with all event
51 (including stuff like POE and IO::Async). 54 models it supports (including stuff like POE and IO::Async, as long as
55 those use one of the supported event loops. It is trivial to add new
56 event loops to AnyEvent, too, so it is future-proof).
52 57
53 In addition of being free of having to use *the one and only true event 58 In addition to being free of having to use *the one and only true event
54 model*, AnyEvent also is free of bloat and policy: with POE or similar 59 model*, AnyEvent also is free of bloat and policy: with POE or similar
55 modules, you get an enourmous amount of code and strict rules you have 60 modules, you get an enormous amount of code and strict rules you have to
56 to follow. AnyEvent, on the other hand, is lean and to the point by only 61 follow. AnyEvent, on the other hand, is lean and up to the point, by
57 offering the functionality that is useful, in as thin as a wrapper as 62 only offering the functionality that is necessary, in as thin as a
58 technically possible. 63 wrapper as technically possible.
59 64
60 Of course, if you want lots of policy (this can arguably be somewhat 65 Of course, if you want lots of policy (this can arguably be somewhat
61 useful) and you want to force your users to use the one and only event 66 useful) and you want to force your users to use the one and only event
62 model, you should *not* use this module. 67 model, you should *not* use this module.
63 68
65 AnyEvent provides an identical interface to multiple event loops. This 70 AnyEvent provides an identical interface to multiple event loops. This
66 allows module authors to utilise an event loop without forcing module 71 allows module authors to utilise an event loop without forcing module
67 users to use the same event loop (as only a single event loop can 72 users to use the same event loop (as only a single event loop can
68 coexist peacefully at any one time). 73 coexist peacefully at any one time).
69 74
70 The interface itself is vaguely similar but not identical to the Event 75 The interface itself is vaguely similar, but not identical to the Event
71 module. 76 module.
72 77
73 On the first call of any method, the module tries to detect the 78 During the first call of any watcher-creation method, the module tries
74 currently loaded event loop by probing wether any of the following 79 to detect the currently loaded event loop by probing whether one of the
75 modules is loaded: Coro::EV, Coro::Event, EV, Event, Glib, Tk. The first 80 following modules is already loaded: EV, Event, Glib,
81 AnyEvent::Impl::Perl, Tk, Event::Lib, Qt, POE. The first one found is
76 one found is used. If none are found, the module tries to load these 82 used. If none are found, the module tries to load these modules
77 modules in the order given. The first one that could be successfully 83 (excluding Tk, Event::Lib, Qt and POE as the pure perl adaptor should
78 loaded will be used. If still none could be found, AnyEvent will fall 84 always succeed) in the order given. The first one that can be
79 back to a pure-perl event loop, which is also not very efficient. 85 successfully loaded will be used. If, after this, still none could be
86 found, AnyEvent will fall back to a pure-perl event loop, which is not
87 very efficient, but should work everywhere.
80 88
81 Because AnyEvent first checks for modules that are already loaded, 89 Because AnyEvent first checks for modules that are already loaded,
82 loading an Event model explicitly before first using AnyEvent will 90 loading an event model explicitly before first using AnyEvent will
83 likely make that model the default. For example: 91 likely make that model the default. For example:
84 92
85 use Tk; 93 use Tk;
86 use AnyEvent; 94 use AnyEvent;
87 95
88 # .. AnyEvent will likely default to Tk 96 # .. AnyEvent will likely default to Tk
97
98 The *likely* means that, if any module loads another event model and
99 starts using it, all bets are off. Maybe you should tell their authors
100 to use AnyEvent so their modules work together with others seamlessly...
89 101
90 The pure-perl implementation of AnyEvent is called 102 The pure-perl implementation of AnyEvent is called
91 "AnyEvent::Impl::Perl". Like other event modules you can load it 103 "AnyEvent::Impl::Perl". Like other event modules you can load it
92 explicitly. 104 explicitly.
93 105
94WATCHERS 106WATCHERS
95 AnyEvent has the central concept of a *watcher*, which is an object that 107 AnyEvent has the central concept of a *watcher*, which is an object that
96 stores relevant data for each kind of event you are waiting for, such as 108 stores relevant data for each kind of event you are waiting for, such as
97 the callback to call, the filehandle to watch, etc. 109 the callback to call, the file handle to watch, etc.
98 110
99 These watchers are normal Perl objects with normal Perl lifetime. After 111 These watchers are normal Perl objects with normal Perl lifetime. After
100 creating a watcher it will immediately "watch" for events and invoke the 112 creating a watcher it will immediately "watch" for events and invoke the
113 callback when the event occurs (of course, only when the event model is
114 in control).
115
101 callback. To disable the watcher you have to destroy it (e.g. by setting 116 To disable the watcher you have to destroy it (e.g. by setting the
102 the variable that stores it to "undef" or otherwise deleting all 117 variable you store it in to "undef" or otherwise deleting all references
103 references to it). 118 to it).
104 119
105 All watchers are created by calling a method on the "AnyEvent" class. 120 All watchers are created by calling a method on the "AnyEvent" class.
106 121
122 Many watchers either are used with "recursion" (repeating timers for
123 example), or need to refer to their watcher object in other ways.
124
125 An any way to achieve that is this pattern:
126
127 my $w; $w = AnyEvent->type (arg => value ..., cb => sub {
128 # you can use $w here, for example to undef it
129 undef $w;
130 });
131
132 Note that "my $w; $w =" combination. This is necessary because in Perl,
133 my variables are only visible after the statement in which they are
134 declared.
135
107 IO WATCHERS 136 I/O WATCHERS
108 You can create I/O watcher by calling the "AnyEvent->io" method with the 137 You can create an I/O watcher by calling the "AnyEvent->io" method with
109 following mandatory arguments: 138 the following mandatory key-value pairs as arguments:
110 139
111 "fh" the Perl *filehandle* (not filedescriptor) to watch for events. 140 "fh" the Perl *file handle* (*not* file descriptor) to watch for events.
112 "poll" must be a string that is either "r" or "w", that creates a 141 "poll" must be a string that is either "r" or "w", which creates a
113 watcher waiting for "r"eadable or "w"ritable events. "cb" the callback 142 watcher waiting for "r"eadable or "w"ritable events, respectively. "cb"
114 to invoke everytime the filehandle becomes ready. 143 is the callback to invoke each time the file handle becomes ready.
115 144
116 Filehandles will be kept alive, so as long as the watcher exists, the 145 Although the callback might get passed parameters, their value and
117 filehandle exists, too. 146 presence is undefined and you cannot rely on them. Portable AnyEvent
147 callbacks cannot use arguments passed to I/O watcher callbacks.
148
149 The I/O watcher might use the underlying file descriptor or a copy of
150 it. You must not close a file handle as long as any watcher is active on
151 the underlying file descriptor.
152
153 Some event loops issue spurious readyness notifications, so you should
154 always use non-blocking calls when reading/writing from/to your file
155 handles.
118 156
119 Example: 157 Example:
120 158
121 # wait for readability of STDIN, then read a line and disable the watcher 159 # wait for readability of STDIN, then read a line and disable the watcher
122 my $w; $w = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub { 160 my $w; $w = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub {
127 165
128 TIME WATCHERS 166 TIME WATCHERS
129 You can create a time watcher by calling the "AnyEvent->timer" method 167 You can create a time watcher by calling the "AnyEvent->timer" method
130 with the following mandatory arguments: 168 with the following mandatory arguments:
131 169
132 "after" after how many seconds (fractions are supported) should the 170 "after" specifies after how many seconds (fractional values are
133 timer activate. "cb" the callback to invoke. 171 supported) the callback should be invoked. "cb" is the callback to
172 invoke in that case.
173
174 Although the callback might get passed parameters, their value and
175 presence is undefined and you cannot rely on them. Portable AnyEvent
176 callbacks cannot use arguments passed to time watcher callbacks.
134 177
135 The timer callback will be invoked at most once: if you want a repeating 178 The timer callback will be invoked at most once: if you want a repeating
136 timer you have to create a new watcher (this is a limitation by both Tk 179 timer you have to create a new watcher (this is a limitation by both Tk
137 and Glib). 180 and Glib).
138 181
144 }); 187 });
145 188
146 # to cancel the timer: 189 # to cancel the timer:
147 undef $w; 190 undef $w;
148 191
149 CONDITION WATCHERS 192 Example 2:
193
194 # fire an event after 0.5 seconds, then roughly every second
195 my $w;
196
197 my $cb = sub {
198 # cancel the old timer while creating a new one
199 $w = AnyEvent->timer (after => 1, cb => $cb);
200 };
201
202 # start the "loop" by creating the first watcher
203 $w = AnyEvent->timer (after => 0.5, cb => $cb);
204
205 TIMING ISSUES
206 There are two ways to handle timers: based on real time (relative, "fire
207 in 10 seconds") and based on wallclock time (absolute, "fire at 12
208 o'clock").
209
210 While most event loops expect timers to specified in a relative way,
211 they use absolute time internally. This makes a difference when your
212 clock "jumps", for example, when ntp decides to set your clock backwards
213 from the wrong date of 2014-01-01 to 2008-01-01, a watcher that is
214 supposed to fire "after" a second might actually take six years to
215 finally fire.
216
217 AnyEvent cannot compensate for this. The only event loop that is
218 conscious about these issues is EV, which offers both relative
219 (ev_timer, based on true relative time) and absolute (ev_periodic, based
220 on wallclock time) timers.
221
222 AnyEvent always prefers relative timers, if available, matching the
223 AnyEvent API.
224
225 SIGNAL WATCHERS
226 You can watch for signals using a signal watcher, "signal" is the signal
227 *name* without any "SIG" prefix, "cb" is the Perl callback to be invoked
228 whenever a signal occurs.
229
230 Although the callback might get passed parameters, their value and
231 presence is undefined and you cannot rely on them. Portable AnyEvent
232 callbacks cannot use arguments passed to signal watcher callbacks.
233
234 Multiple signal occurrences can be clumped together into one callback
235 invocation, and callback invocation will be synchronous. Synchronous
236 means that it might take a while until the signal gets handled by the
237 process, but it is guaranteed not to interrupt any other callbacks.
238
239 The main advantage of using these watchers is that you can share a
240 signal between multiple watchers.
241
242 This watcher might use %SIG, so programs overwriting those signals
243 directly will likely not work correctly.
244
245 Example: exit on SIGINT
246
247 my $w = AnyEvent->signal (signal => "INT", cb => sub { exit 1 });
248
249 CHILD PROCESS WATCHERS
250 You can also watch on a child process exit and catch its exit status.
251
252 The child process is specified by the "pid" argument (if set to 0, it
253 watches for any child process exit). The watcher will trigger as often
254 as status change for the child are received. This works by installing a
255 signal handler for "SIGCHLD". The callback will be called with the pid
256 and exit status (as returned by waitpid), so unlike other watcher types,
257 you *can* rely on child watcher callback arguments.
258
259 There is a slight catch to child watchers, however: you usually start
260 them *after* the child process was created, and this means the process
261 could have exited already (and no SIGCHLD will be sent anymore).
262
263 Not all event models handle this correctly (POE doesn't), but even for
264 event models that *do* handle this correctly, they usually need to be
265 loaded before the process exits (i.e. before you fork in the first
266 place).
267
268 This means you cannot create a child watcher as the very first thing in
269 an AnyEvent program, you *have* to create at least one watcher before
270 you "fork" the child (alternatively, you can call "AnyEvent::detect").
271
272 Example: fork a process and wait for it
273
274 my $done = AnyEvent->condvar;
275
276 my $pid = fork or exit 5;
277
278 my $w = AnyEvent->child (
279 pid => $pid,
280 cb => sub {
281 my ($pid, $status) = @_;
282 warn "pid $pid exited with status $status";
283 $done->send;
284 },
285 );
286
287 # do something else, then wait for process exit
288 $done->recv;
289
290 CONDITION VARIABLES
291 If you are familiar with some event loops you will know that all of them
292 require you to run some blocking "loop", "run" or similar function that
293 will actively watch for new events and call your callbacks.
294
295 AnyEvent is different, it expects somebody else to run the event loop
296 and will only block when necessary (usually when told by the user).
297
298 The instrument to do that is called a "condition variable", so called
299 because they represent a condition that must become true.
300
150 Condition watchers can be created by calling the "AnyEvent->condvar" 301 Condition variables can be created by calling the "AnyEvent->condvar"
151 method without any arguments. 302 method, usually without arguments. The only argument pair allowed is
303 "cb", which specifies a callback to be called when the condition
304 variable becomes true.
152 305
153 A condition watcher watches for a condition - precisely that the 306 After creation, the condition variable is "false" until it becomes
154 "->broadcast" method has been called. 307 "true" by calling the "send" method (or calling the condition variable
308 as if it were a callback).
155 309
310 Condition variables are similar to callbacks, except that you can
311 optionally wait for them. They can also be called merge points - points
312 in time where multiple outstanding events have been processed. And yet
313 another way to call them is transactions - each condition variable can
314 be used to represent a transaction, which finishes at some point and
315 delivers a result.
316
317 Condition variables are very useful to signal that something has
318 finished, for example, if you write a module that does asynchronous http
319 requests, then a condition variable would be the ideal candidate to
320 signal the availability of results. The user can either act when the
321 callback is called or can synchronously "->recv" for the results.
322
323 You can also use them to simulate traditional event loops - for example,
324 you can block your main program until an event occurs - for example, you
325 could "->recv" in your main program until the user clicks the Quit
326 button of your app, which would "->send" the "quit" event.
327
156 Note that condition watchers recurse into the event loop - if you have 328 Note that condition variables recurse into the event loop - if you have
157 two watchers that call "->wait" in a round-robbin fashion, you lose. 329 two pieces of code that call "->recv" in a round-robin fashion, you
158 Therefore, condition watchers are good to export to your caller, but you 330 lose. Therefore, condition variables are good to export to your caller,
159 should avoid making a blocking wait, at least in callbacks, as this 331 but you should avoid making a blocking wait yourself, at least in
160 usually asks for trouble. 332 callbacks, as this asks for trouble.
161 333
162 The watcher has only two methods: 334 Condition variables are represented by hash refs in perl, and the keys
335 used by AnyEvent itself are all named "_ae_XXX" to make subclassing easy
336 (it is often useful to build your own transaction class on top of
337 AnyEvent). To subclass, use "AnyEvent::CondVar" as base class and call
338 it's "new" method in your own "new" method.
163 339
164 $cv->wait 340 There are two "sides" to a condition variable - the "producer side"
341 which eventually calls "-> send", and the "consumer side", which waits
342 for the send to occur.
343
344 Example: wait for a timer.
345
346 # wait till the result is ready
347 my $result_ready = AnyEvent->condvar;
348
349 # do something such as adding a timer
350 # or socket watcher the calls $result_ready->send
351 # when the "result" is ready.
352 # in this case, we simply use a timer:
353 my $w = AnyEvent->timer (
354 after => 1,
355 cb => sub { $result_ready->send },
356 );
357
358 # this "blocks" (while handling events) till the callback
359 # calls send
360 $result_ready->recv;
361
362 Example: wait for a timer, but take advantage of the fact that condition
363 variables are also code references.
364
365 my $done = AnyEvent->condvar;
366 my $delay = AnyEvent->timer (after => 5, cb => $done);
367 $done->recv;
368
369 METHODS FOR PRODUCERS
370 These methods should only be used by the producing side, i.e. the
371 code/module that eventually sends the signal. Note that it is also the
372 producer side which creates the condvar in most cases, but it isn't
373 uncommon for the consumer to create it as well.
374
375 $cv->send (...)
376 Flag the condition as ready - a running "->recv" and all further
377 calls to "recv" will (eventually) return after this method has been
378 called. If nobody is waiting the send will be remembered.
379
380 If a callback has been set on the condition variable, it is called
381 immediately from within send.
382
383 Any arguments passed to the "send" call will be returned by all
384 future "->recv" calls.
385
386 Condition variables are overloaded so one can call them directly (as
387 a code reference). Calling them directly is the same as calling
388 "send".
389
390 $cv->croak ($error)
391 Similar to send, but causes all call's to "->recv" to invoke
392 "Carp::croak" with the given error message/object/scalar.
393
394 This can be used to signal any errors to the condition variable
395 user/consumer.
396
397 $cv->begin ([group callback])
398 $cv->end
399 These two methods are EXPERIMENTAL and MIGHT CHANGE.
400
401 These two methods can be used to combine many transactions/events
402 into one. For example, a function that pings many hosts in parallel
403 might want to use a condition variable for the whole process.
404
405 Every call to "->begin" will increment a counter, and every call to
406 "->end" will decrement it. If the counter reaches 0 in "->end", the
407 (last) callback passed to "begin" will be executed. That callback is
408 *supposed* to call "->send", but that is not required. If no
409 callback was set, "send" will be called without any arguments.
410
411 Let's clarify this with the ping example:
412
413 my $cv = AnyEvent->condvar;
414
415 my %result;
416 $cv->begin (sub { $cv->send (\%result) });
417
418 for my $host (@list_of_hosts) {
419 $cv->begin;
420 ping_host_then_call_callback $host, sub {
421 $result{$host} = ...;
422 $cv->end;
423 };
424 }
425
426 $cv->end;
427
428 This code fragment supposedly pings a number of hosts and calls
429 "send" after results for all then have have been gathered - in any
430 order. To achieve this, the code issues a call to "begin" when it
431 starts each ping request and calls "end" when it has received some
432 result for it. Since "begin" and "end" only maintain a counter, the
433 order in which results arrive is not relevant.
434
435 There is an additional bracketing call to "begin" and "end" outside
436 the loop, which serves two important purposes: first, it sets the
437 callback to be called once the counter reaches 0, and second, it
438 ensures that "send" is called even when "no" hosts are being pinged
439 (the loop doesn't execute once).
440
441 This is the general pattern when you "fan out" into multiple
442 subrequests: use an outer "begin"/"end" pair to set the callback and
443 ensure "end" is called at least once, and then, for each subrequest
444 you start, call "begin" and for each subrequest you finish, call
445 "end".
446
447 METHODS FOR CONSUMERS
448 These methods should only be used by the consuming side, i.e. the code
449 awaits the condition.
450
451 $cv->recv
165 Wait (blocking if necessary) until the "->broadcast" method has been 452 Wait (blocking if necessary) until the "->send" or "->croak" methods
166 called on c<$cv>, while servicing other watchers normally. 453 have been called on c<$cv>, while servicing other watchers normally.
167 454
168 You can only wait once on a condition - additional calls will return 455 You can only wait once on a condition - additional calls are valid
169 immediately. 456 but will return immediately.
457
458 If an error condition has been set by calling "->croak", then this
459 function will call "croak".
460
461 In list context, all parameters passed to "send" will be returned,
462 in scalar context only the first one will be returned.
170 463
171 Not all event models support a blocking wait - some die in that case 464 Not all event models support a blocking wait - some die in that case
172 (programs might want to do that so they stay interactive), so *if 465 (programs might want to do that to stay interactive), so *if you are
173 you are using this from a module, never require a blocking wait*, 466 using this from a module, never require a blocking wait*, but let
174 but let the caller decide wether the call will block or not (for 467 the caller decide whether the call will block or not (for example,
175 example, by coupling condition variables with some kind of request 468 by coupling condition variables with some kind of request results
176 results and supporting callbacks so the caller knows that getting 469 and supporting callbacks so the caller knows that getting the result
177 the result will not block, while still suppporting blocking waits if 470 will not block, while still supporting blocking waits if the caller
178 the caller so desires). 471 so desires).
179 472
180 Another reason *never* to "->wait" in a module is that you cannot 473 Another reason *never* to "->recv" in a module is that you cannot
181 sensibly have two "->wait"'s in parallel, as that would require 474 sensibly have two "->recv"'s in parallel, as that would require
182 multiple interpreters or coroutines/threads, none of which 475 multiple interpreters or coroutines/threads, none of which
183 "AnyEvent" can supply (the coroutine-aware backends "Coro::EV" and 476 "AnyEvent" can supply.
184 "Coro::Event" explicitly support concurrent "->wait"'s from
185 different coroutines, however).
186 477
187 $cv->broadcast 478 The Coro module, however, *can* and *does* supply coroutines and, in
188 Flag the condition as ready - a running "->wait" and all further 479 fact, Coro::AnyEvent replaces AnyEvent's condvars by coroutine-safe
189 calls to "wait" will return after this method has been called. If 480 versions and also integrates coroutines into AnyEvent, making
190 nobody is waiting the broadcast will be remembered.. 481 blocking "->recv" calls perfectly safe as long as they are done from
482 another coroutine (one that doesn't run the event loop).
191 483
192 Example: 484 You can ensure that "-recv" never blocks by setting a callback and
485 only calling "->recv" from within that callback (or at a later
486 time). This will work even when the event loop does not support
487 blocking waits otherwise.
193 488
194 # wait till the result is ready 489 $bool = $cv->ready
195 my $result_ready = AnyEvent->condvar; 490 Returns true when the condition is "true", i.e. whether "send" or
491 "croak" have been called.
196 492
197 # do something such as adding a timer 493 $cb = $cv->cb ([new callback])
198 # or socket watcher the calls $result_ready->broadcast 494 This is a mutator function that returns the callback set and
199 # when the "result" is ready. 495 optionally replaces it before doing so.
200 496
201 $result_ready->wait; 497 The callback will be called when the condition becomes "true", i.e.
498 when "send" or "croak" are called. Calling "recv" inside the
499 callback or at any later time is guaranteed not to block.
202 500
203 SIGNAL WATCHERS 501GLOBAL VARIABLES AND FUNCTIONS
204 You can listen for signals using a signal watcher, "signal" is the
205 signal *name* without any "SIG" prefix. Multiple signals events can be
206 clumped together into one callback invocation, and callback invocation
207 might or might not be asynchronous.
208
209 These watchers might use %SIG, so programs overwriting those signals
210 directly will likely not work correctly.
211
212 Example: exit on SIGINT
213
214 my $w = AnyEvent->signal (signal => "INT", cb => sub { exit 1 });
215
216 CHILD PROCESS WATCHERS
217 You can also listen for the status of a child process specified by the
218 "pid" argument (or any child if the pid argument is 0). The watcher will
219 trigger as often as status change for the child are received. This works
220 by installing a signal handler for "SIGCHLD". The callback will be
221 called with the pid and exit status (as returned by waitpid).
222
223 Example: wait for pid 1333
224
225 my $w = AnyEvent->child (pid => 1333, cb => sub { warn "exit status $?" });
226
227GLOBALS
228 $AnyEvent::MODEL 502 $AnyEvent::MODEL
229 Contains "undef" until the first watcher is being created. Then it 503 Contains "undef" until the first watcher is being created. Then it
230 contains the event model that is being used, which is the name of 504 contains the event model that is being used, which is the name of
231 the Perl class implementing the model. This class is usually one of 505 the Perl class implementing the model. This class is usually one of
232 the "AnyEvent::Impl:xxx" modules, but can be any other class in the 506 the "AnyEvent::Impl:xxx" modules, but can be any other class in the
233 case AnyEvent has been extended at runtime (e.g. in *rxvt-unicode*). 507 case AnyEvent has been extended at runtime (e.g. in *rxvt-unicode*).
234 508
235 The known classes so far are: 509 The known classes so far are:
236 510
237 AnyEvent::Impl::CoroEV based on Coro::EV, best choice.
238 AnyEvent::Impl::CoroEvent based on Coro::Event, second best choice.
239 AnyEvent::Impl::EV based on EV (an interface to libev, also best choice). 511 AnyEvent::Impl::EV based on EV (an interface to libev, best choice).
240 AnyEvent::Impl::Event based on Event, also second best choice :) 512 AnyEvent::Impl::Event based on Event, second best choice.
513 AnyEvent::Impl::Perl pure-perl implementation, fast and portable.
241 AnyEvent::Impl::Glib based on Glib, third-best choice. 514 AnyEvent::Impl::Glib based on Glib, third-best choice.
242 AnyEvent::Impl::Tk based on Tk, very bad choice. 515 AnyEvent::Impl::Tk based on Tk, very bad choice.
243 AnyEvent::Impl::Perl pure-perl implementation, inefficient but portable. 516 AnyEvent::Impl::Qt based on Qt, cannot be autoprobed (see its docs).
517 AnyEvent::Impl::EventLib based on Event::Lib, leaks memory and worse.
518 AnyEvent::Impl::POE based on POE, not generic enough for full support.
519
520 There is no support for WxWidgets, as WxWidgets has no support for
521 watching file handles. However, you can use WxWidgets through the
522 POE Adaptor, as POE has a Wx backend that simply polls 20 times per
523 second, which was considered to be too horrible to even consider for
524 AnyEvent. Likewise, other POE backends can be used by AnyEvent by
525 using it's adaptor.
526
527 AnyEvent knows about Prima and Wx and will try to use POE when
528 autodetecting them.
244 529
245 AnyEvent::detect 530 AnyEvent::detect
246 Returns $AnyEvent::MODEL, forcing autodetection of the event model 531 Returns $AnyEvent::MODEL, forcing autodetection of the event model
247 if necessary. You should only call this function right before you 532 if necessary. You should only call this function right before you
248 would have created an AnyEvent watcher anyway, that is, very late at 533 would have created an AnyEvent watcher anyway, that is, as late as
249 runtime. 534 possible at runtime.
535
536 $guard = AnyEvent::post_detect { BLOCK }
537 Arranges for the code block to be executed as soon as the event
538 model is autodetected (or immediately if this has already happened).
539
540 If called in scalar or list context, then it creates and returns an
541 object that automatically removes the callback again when it is
542 destroyed. See Coro::BDB for a case where this is useful.
543
544 @AnyEvent::post_detect
545 If there are any code references in this array (you can "push" to it
546 before or after loading AnyEvent), then they will called directly
547 after the event loop has been chosen.
548
549 You should check $AnyEvent::MODEL before adding to this array,
550 though: if it contains a true value then the event loop has already
551 been detected, and the array will be ignored.
552
553 Best use "AnyEvent::post_detect { BLOCK }" instead.
250 554
251WHAT TO DO IN A MODULE 555WHAT TO DO IN A MODULE
252 As a module author, you should "use AnyEvent" and call AnyEvent methods 556 As a module author, you should "use AnyEvent" and call AnyEvent methods
253 freely, but you should not load a specific event module or rely on it. 557 freely, but you should not load a specific event module or rely on it.
254 558
255 Be careful when you create watchers in the module body - Anyevent will 559 Be careful when you create watchers in the module body - AnyEvent will
256 decide which event module to use as soon as the first method is called, 560 decide which event module to use as soon as the first method is called,
257 so by calling AnyEvent in your module body you force the user of your 561 so by calling AnyEvent in your module body you force the user of your
258 module to load the event module first. 562 module to load the event module first.
259 563
564 Never call "->recv" on a condition variable unless you *know* that the
565 "->send" method has been called on it already. This is because it will
566 stall the whole program, and the whole point of using events is to stay
567 interactive.
568
569 It is fine, however, to call "->recv" when the user of your module
570 requests it (i.e. if you create a http request object ad have a method
571 called "results" that returns the results, it should call "->recv"
572 freely, as the user of your module knows what she is doing. always).
573
260WHAT TO DO IN THE MAIN PROGRAM 574WHAT TO DO IN THE MAIN PROGRAM
261 There will always be a single main program - the only place that should 575 There will always be a single main program - the only place that should
262 dictate which event model to use. 576 dictate which event model to use.
263 577
264 If it doesn't care, it can just "use AnyEvent" and use it itself, or not 578 If it doesn't care, it can just "use AnyEvent" and use it itself, or not
265 do anything special and let AnyEvent decide which implementation to 579 do anything special (it does not need to be event-based) and let
266 chose. 580 AnyEvent decide which implementation to chose if some module relies on
581 it.
267 582
268 If the main program relies on a specific event model (for example, in 583 If the main program relies on a specific event model. For example, in
269 Gtk2 programs you have to rely on either Glib or Glib::Event), you 584 Gtk2 programs you have to rely on the Glib module. You should load the
270 should load it before loading AnyEvent or any module that uses it, 585 event module before loading AnyEvent or any module that uses it:
271 generally, as early as possible. The reason is that modules might create 586 generally speaking, you should load it as early as possible. The reason
272 watchers when they are loaded, and AnyEvent will decide on the event 587 is that modules might create watchers when they are loaded, and AnyEvent
273 model to use as soon as it creates watchers, and it might chose the 588 will decide on the event model to use as soon as it creates watchers,
274 wrong one unless you load the correct one yourself. 589 and it might chose the wrong one unless you load the correct one
590 yourself.
275 591
276 You can chose to use a rather inefficient pure-perl implementation by 592 You can chose to use a rather inefficient pure-perl implementation by
277 loading the "AnyEvent::Impl::Perl" module, but letting AnyEvent chose is 593 loading the "AnyEvent::Impl::Perl" module, which gives you similar
278 generally better. 594 behaviour everywhere, but letting AnyEvent chose is generally better.
595
596OTHER MODULES
597 The following is a non-exhaustive list of additional modules that use
598 AnyEvent and can therefore be mixed easily with other AnyEvent modules
599 in the same program. Some of the modules come with AnyEvent, some are
600 available via CPAN.
601
602 AnyEvent::Util
603 Contains various utility functions that replace often-used but
604 blocking functions such as "inet_aton" by event-/callback-based
605 versions.
606
607 AnyEvent::Handle
608 Provide read and write buffers and manages watchers for reads and
609 writes.
610
611 AnyEvent::Socket
612 Provides various utility functions for (internet protocol) sockets,
613 addresses and name resolution. Also functions to create non-blocking
614 tcp connections or tcp servers, with IPv6 and SRV record support and
615 more.
616
617 AnyEvent::HTTPD
618 Provides a simple web application server framework.
619
620 AnyEvent::DNS
621 Provides rich asynchronous DNS resolver capabilities.
622
623 AnyEvent::FastPing
624 The fastest ping in the west.
625
626 Net::IRC3
627 AnyEvent based IRC client module family.
628
629 Net::XMPP2
630 AnyEvent based XMPP (Jabber protocol) module family.
631
632 Net::FCP
633 AnyEvent-based implementation of the Freenet Client Protocol,
634 birthplace of AnyEvent.
635
636 Event::ExecFlow
637 High level API for event-based execution flow control.
638
639 Coro
640 Has special support for AnyEvent via Coro::AnyEvent.
641
642 AnyEvent::AIO, IO::AIO
643 Truly asynchronous I/O, should be in the toolbox of every event
644 programmer. AnyEvent::AIO transparently fuses IO::AIO and AnyEvent
645 together.
646
647 AnyEvent::BDB, BDB
648 Truly asynchronous Berkeley DB access. AnyEvent::AIO transparently
649 fuses IO::AIO and AnyEvent together.
650
651 IO::Lambda
652 The lambda approach to I/O - don't ask, look there. Can use
653 AnyEvent.
279 654
280SUPPLYING YOUR OWN EVENT MODEL INTERFACE 655SUPPLYING YOUR OWN EVENT MODEL INTERFACE
656 This is an advanced topic that you do not normally need to use AnyEvent
657 in a module. This section is only of use to event loop authors who want
658 to provide AnyEvent compatibility.
659
281 If you need to support another event library which isn't directly 660 If you need to support another event library which isn't directly
282 supported by AnyEvent, you can supply your own interface to it by 661 supported by AnyEvent, you can supply your own interface to it by
283 pushing, before the first watcher gets created, the package name of the 662 pushing, before the first watcher gets created, the package name of the
284 event module and the package name of the interface to use onto 663 event module and the package name of the interface to use onto
285 @AnyEvent::REGISTRY. You can do that before and even without loading 664 @AnyEvent::REGISTRY. You can do that before and even without loading
286 AnyEvent. 665 AnyEvent, so it is reasonably cheap.
287 666
288 Example: 667 Example:
289 668
290 push @AnyEvent::REGISTRY, [urxvt => urxvt::anyevent::]; 669 push @AnyEvent::REGISTRY, [urxvt => urxvt::anyevent::];
291 670
292 This tells AnyEvent to (literally) use the "urxvt::anyevent::" 671 This tells AnyEvent to (literally) use the "urxvt::anyevent::"
293 package/class when it finds the "urxvt" package/module is loaded. When 672 package/class when it finds the "urxvt" package/module is already
673 loaded.
674
294 AnyEvent is loaded and asked to find a suitable event model, it will 675 When AnyEvent is loaded and asked to find a suitable event model, it
295 first check for the presence of urxvt. 676 will first check for the presence of urxvt by trying to "use" the
677 "urxvt::anyevent" module.
296 678
297 The class should provide implementations for all watcher types (see 679 The class should provide implementations for all watcher types. See
298 AnyEvent::Impl::Event (source code), AnyEvent::Impl::Glib (Source code) 680 AnyEvent::Impl::EV (source code), AnyEvent::Impl::Glib (Source code) and
299 and so on for actual examples, use "perldoc -m AnyEvent::Impl::Glib" to 681 so on for actual examples. Use "perldoc -m AnyEvent::Impl::Glib" to see
300 see the sources). 682 the sources.
301 683
684 If you don't provide "signal" and "child" watchers than AnyEvent will
685 provide suitable (hopefully) replacements.
686
302 The above isn't fictitious, the *rxvt-unicode* (a.k.a. urxvt) uses the 687 The above example isn't fictitious, the *rxvt-unicode* (a.k.a. urxvt)
303 above line as-is. An interface isn't included in AnyEvent because it 688 terminal emulator uses the above line as-is. An interface isn't included
304 doesn't make sense outside the embedded interpreter inside 689 in AnyEvent because it doesn't make sense outside the embedded
305 *rxvt-unicode*, and it is updated and maintained as part of the 690 interpreter inside *rxvt-unicode*, and it is updated and maintained as
306 *rxvt-unicode* distribution. 691 part of the *rxvt-unicode* distribution.
307 692
308 *rxvt-unicode* also cheats a bit by not providing blocking access to 693 *rxvt-unicode* also cheats a bit by not providing blocking access to
309 condition variables: code blocking while waiting for a condition will 694 condition variables: code blocking while waiting for a condition will
310 "die". This still works with most modules/usages, and blocking calls 695 "die". This still works with most modules/usages, and blocking calls
311 must not be in an interactive application, so it makes sense. 696 must not be done in an interactive application, so it makes sense.
312 697
313ENVIRONMENT VARIABLES 698ENVIRONMENT VARIABLES
314 The following environment variables are used by this module: 699 The following environment variables are used by this module:
315 700
316 "PERL_ANYEVENT_VERBOSE" when set to 2 or higher, reports which event 701 "PERL_ANYEVENT_VERBOSE"
317 model gets used. 702 By default, AnyEvent will be completely silent except in fatal
703 conditions. You can set this environment variable to make AnyEvent
704 more talkative.
318 705
319EXAMPLE 706 When set to 1 or higher, causes AnyEvent to warn about unexpected
707 conditions, such as not being able to load the event model specified
708 by "PERL_ANYEVENT_MODEL".
709
710 When set to 2 or higher, cause AnyEvent to report to STDERR which
711 event model it chooses.
712
713 "PERL_ANYEVENT_MODEL"
714 This can be used to specify the event model to be used by AnyEvent,
715 before auto detection and -probing kicks in. It must be a string
716 consisting entirely of ASCII letters. The string "AnyEvent::Impl::"
717 gets prepended and the resulting module name is loaded and if the
718 load was successful, used as event model. If it fails to load
719 AnyEvent will proceed with auto detection and -probing.
720
721 This functionality might change in future versions.
722
723 For example, to force the pure perl model (AnyEvent::Impl::Perl) you
724 could start your program like this:
725
726 PERL_ANYEVENT_MODEL=Perl perl ...
727
728 "PERL_ANYEVENT_PROTOCOLS"
729 Used by both AnyEvent::DNS and AnyEvent::Socket to determine
730 preferences for IPv4 or IPv6. The default is unspecified (and might
731 change, or be the result of auto probing).
732
733 Must be set to a comma-separated list of protocols or address
734 families, current supported: "ipv4" and "ipv6". Only protocols
735 mentioned will be used, and preference will be given to protocols
736 mentioned earlier in the list.
737
738 This variable can effectively be used for denial-of-service attacks
739 against local programs (e.g. when setuid), although the impact is
740 likely small, as the program has to handle connection errors
741 already-
742
743 Examples: "PERL_ANYEVENT_PROTOCOLS=ipv4,ipv6" - prefer IPv4 over
744 IPv6, but support both and try to use both.
745 "PERL_ANYEVENT_PROTOCOLS=ipv4" - only support IPv4, never try to
746 resolve or contact IPv6 addresses.
747 "PERL_ANYEVENT_PROTOCOLS=ipv6,ipv4" support either IPv4 or IPv6, but
748 prefer IPv6 over IPv4.
749
750 "PERL_ANYEVENT_EDNS0"
751 Used by AnyEvent::DNS to decide whether to use the EDNS0 extension
752 for DNS. This extension is generally useful to reduce DNS traffic,
753 but some (broken) firewalls drop such DNS packets, which is why it
754 is off by default.
755
756 Setting this variable to 1 will cause AnyEvent::DNS to announce
757 EDNS0 in its DNS requests.
758
759EXAMPLE PROGRAM
320 The following program uses an io watcher to read data from stdin, a 760 The following program uses an I/O watcher to read data from STDIN, a
321 timer to display a message once per second, and a condvar to exit the 761 timer to display a message once per second, and a condition variable to
322 program when the user enters quit: 762 quit the program when the user enters quit:
323 763
324 use AnyEvent; 764 use AnyEvent;
325 765
326 my $cv = AnyEvent->condvar; 766 my $cv = AnyEvent->condvar;
327 767
328 my $io_watcher = AnyEvent->io (fh => \*STDIN, poll => 'r', cb => sub { 768 my $io_watcher = AnyEvent->io (
769 fh => \*STDIN,
770 poll => 'r',
771 cb => sub {
329 warn "io event <$_[0]>\n"; # will always output <r> 772 warn "io event <$_[0]>\n"; # will always output <r>
330 chomp (my $input = <STDIN>); # read a line 773 chomp (my $input = <STDIN>); # read a line
331 warn "read: $input\n"; # output what has been read 774 warn "read: $input\n"; # output what has been read
332 $cv->broadcast if $input =~ /^q/i; # quit program if /^q/i 775 $cv->send if $input =~ /^q/i; # quit program if /^q/i
776 },
333 }); 777 );
334 778
335 my $time_watcher; # can only be used once 779 my $time_watcher; # can only be used once
336 780
337 sub new_timer { 781 sub new_timer {
338 $timer = AnyEvent->timer (after => 1, cb => sub { 782 $timer = AnyEvent->timer (after => 1, cb => sub {
341 }); 785 });
342 } 786 }
343 787
344 new_timer; # create first timer 788 new_timer; # create first timer
345 789
346 $cv->wait; # wait until user enters /^q/i 790 $cv->recv; # wait until user enters /^q/i
347 791
348REAL-WORLD EXAMPLE 792REAL-WORLD EXAMPLE
349 Consider the Net::FCP module. It features (among others) the following 793 Consider the Net::FCP module. It features (among others) the following
350 API calls, which are to freenet what HTTP GET requests are to http: 794 API calls, which are to freenet what HTTP GET requests are to http:
351 795
400 syswrite $txn->{fh}, $txn->{request} 844 syswrite $txn->{fh}, $txn->{request}
401 or die "connection or write error"; 845 or die "connection or write error";
402 $txn->{w} = AnyEvent->io (fh => $txn->{fh}, poll => 'r', cb => sub { $txn->fh_ready_r }); 846 $txn->{w} = AnyEvent->io (fh => $txn->{fh}, poll => 'r', cb => sub { $txn->fh_ready_r });
403 847
404 Again, "fh_ready_r" waits till all data has arrived, and then stores the 848 Again, "fh_ready_r" waits till all data has arrived, and then stores the
405 result and signals any possible waiters that the request ahs finished: 849 result and signals any possible waiters that the request has finished:
406 850
407 sysread $txn->{fh}, $txn->{buf}, length $txn->{$buf}; 851 sysread $txn->{fh}, $txn->{buf}, length $txn->{$buf};
408 852
409 if (end-of-file or data complete) { 853 if (end-of-file or data complete) {
410 $txn->{result} = $txn->{buf}; 854 $txn->{result} = $txn->{buf};
411 $txn->{finished}->broadcast; 855 $txn->{finished}->send;
412 $txb->{cb}->($txn) of $txn->{cb}; # also call callback 856 $txb->{cb}->($txn) of $txn->{cb}; # also call callback
413 } 857 }
414 858
415 The "result" method, finally, just waits for the finished signal (if the 859 The "result" method, finally, just waits for the finished signal (if the
416 request was already finished, it doesn't wait, of course, and returns 860 request was already finished, it doesn't wait, of course, and returns
417 the data: 861 the data:
418 862
419 $txn->{finished}->wait; 863 $txn->{finished}->recv;
420 return $txn->{result}; 864 return $txn->{result};
421 865
422 The actual code goes further and collects all errors ("die"s, 866 The actual code goes further and collects all errors ("die"s,
423 exceptions) that occured during request processing. The "result" method 867 exceptions) that occurred during request processing. The "result" method
424 detects wether an exception as thrown (it is stored inside the $txn 868 detects whether an exception as thrown (it is stored inside the $txn
425 object) and just throws the exception, which means connection errors and 869 object) and just throws the exception, which means connection errors and
426 other problems get reported tot he code that tries to use the result, 870 other problems get reported tot he code that tries to use the result,
427 not in a random callback. 871 not in a random callback.
428 872
429 All of this enables the following usage styles: 873 All of this enables the following usage styles:
459 903
460 my $quit = AnyEvent->condvar; 904 my $quit = AnyEvent->condvar;
461 905
462 $fcp->txn_client_get ($url)->cb (sub { 906 $fcp->txn_client_get ($url)->cb (sub {
463 ... 907 ...
464 $quit->broadcast; 908 $quit->send;
465 }); 909 });
466 910
467 $quit->wait; 911 $quit->recv;
912
913BENCHMARKS
914 To give you an idea of the performance and overheads that AnyEvent adds
915 over the event loops themselves and to give you an impression of the
916 speed of various event loops I prepared some benchmarks.
917
918 BENCHMARKING ANYEVENT OVERHEAD
919 Here is a benchmark of various supported event models used natively and
920 through AnyEvent. The benchmark creates a lot of timers (with a zero
921 timeout) and I/O watchers (watching STDOUT, a pty, to become writable,
922 which it is), lets them fire exactly once and destroys them again.
923
924 Source code for this benchmark is found as eg/bench in the AnyEvent
925 distribution.
926
927 Explanation of the columns
928 *watcher* is the number of event watchers created/destroyed. Since
929 different event models feature vastly different performances, each event
930 loop was given a number of watchers so that overall runtime is
931 acceptable and similar between tested event loop (and keep them from
932 crashing): Glib would probably take thousands of years if asked to
933 process the same number of watchers as EV in this benchmark.
934
935 *bytes* is the number of bytes (as measured by the resident set size,
936 RSS) consumed by each watcher. This method of measuring captures both C
937 and Perl-based overheads.
938
939 *create* is the time, in microseconds (millionths of seconds), that it
940 takes to create a single watcher. The callback is a closure shared
941 between all watchers, to avoid adding memory overhead. That means
942 closure creation and memory usage is not included in the figures.
943
944 *invoke* is the time, in microseconds, used to invoke a simple callback.
945 The callback simply counts down a Perl variable and after it was invoked
946 "watcher" times, it would "->send" a condvar once to signal the end of
947 this phase.
948
949 *destroy* is the time, in microseconds, that it takes to destroy a
950 single watcher.
951
952 Results
953 name watchers bytes create invoke destroy comment
954 EV/EV 400000 244 0.56 0.46 0.31 EV native interface
955 EV/Any 100000 244 2.50 0.46 0.29 EV + AnyEvent watchers
956 CoroEV/Any 100000 244 2.49 0.44 0.29 coroutines + Coro::Signal
957 Perl/Any 100000 513 4.92 0.87 1.12 pure perl implementation
958 Event/Event 16000 516 31.88 31.30 0.85 Event native interface
959 Event/Any 16000 590 35.75 31.42 1.08 Event + AnyEvent watchers
960 Glib/Any 16000 1357 98.22 12.41 54.00 quadratic behaviour
961 Tk/Any 2000 1860 26.97 67.98 14.00 SEGV with >> 2000 watchers
962 POE/Event 2000 6644 108.64 736.02 14.73 via POE::Loop::Event
963 POE/Select 2000 6343 94.13 809.12 565.96 via POE::Loop::Select
964
965 Discussion
966 The benchmark does *not* measure scalability of the event loop very
967 well. For example, a select-based event loop (such as the pure perl one)
968 can never compete with an event loop that uses epoll when the number of
969 file descriptors grows high. In this benchmark, all events become ready
970 at the same time, so select/poll-based implementations get an unnatural
971 speed boost.
972
973 Also, note that the number of watchers usually has a nonlinear effect on
974 overall speed, that is, creating twice as many watchers doesn't take
975 twice the time - usually it takes longer. This puts event loops tested
976 with a higher number of watchers at a disadvantage.
977
978 To put the range of results into perspective, consider that on the
979 benchmark machine, handling an event takes roughly 1600 CPU cycles with
980 EV, 3100 CPU cycles with AnyEvent's pure perl loop and almost 3000000
981 CPU cycles with POE.
982
983 "EV" is the sole leader regarding speed and memory use, which are both
984 maximal/minimal, respectively. Even when going through AnyEvent, it uses
985 far less memory than any other event loop and is still faster than Event
986 natively.
987
988 The pure perl implementation is hit in a few sweet spots (both the
989 constant timeout and the use of a single fd hit optimisations in the
990 perl interpreter and the backend itself). Nevertheless this shows that
991 it adds very little overhead in itself. Like any select-based backend
992 its performance becomes really bad with lots of file descriptors (and
993 few of them active), of course, but this was not subject of this
994 benchmark.
995
996 The "Event" module has a relatively high setup and callback invocation
997 cost, but overall scores in on the third place.
998
999 "Glib"'s memory usage is quite a bit higher, but it features a faster
1000 callback invocation and overall ends up in the same class as "Event".
1001 However, Glib scales extremely badly, doubling the number of watchers
1002 increases the processing time by more than a factor of four, making it
1003 completely unusable when using larger numbers of watchers (note that
1004 only a single file descriptor was used in the benchmark, so
1005 inefficiencies of "poll" do not account for this).
1006
1007 The "Tk" adaptor works relatively well. The fact that it crashes with
1008 more than 2000 watchers is a big setback, however, as correctness takes
1009 precedence over speed. Nevertheless, its performance is surprising, as
1010 the file descriptor is dup()ed for each watcher. This shows that the
1011 dup() employed by some adaptors is not a big performance issue (it does
1012 incur a hidden memory cost inside the kernel which is not reflected in
1013 the figures above).
1014
1015 "POE", regardless of underlying event loop (whether using its pure perl
1016 select-based backend or the Event module, the POE-EV backend couldn't be
1017 tested because it wasn't working) shows abysmal performance and memory
1018 usage with AnyEvent: Watchers use almost 30 times as much memory as EV
1019 watchers, and 10 times as much memory as Event (the high memory
1020 requirements are caused by requiring a session for each watcher).
1021 Watcher invocation speed is almost 900 times slower than with AnyEvent's
1022 pure perl implementation.
1023
1024 The design of the POE adaptor class in AnyEvent can not really account
1025 for the performance issues, though, as session creation overhead is
1026 small compared to execution of the state machine, which is coded pretty
1027 optimally within AnyEvent::Impl::POE (and while everybody agrees that
1028 using multiple sessions is not a good approach, especially regarding
1029 memory usage, even the author of POE could not come up with a faster
1030 design).
1031
1032 Summary
1033 * Using EV through AnyEvent is faster than any other event loop (even
1034 when used without AnyEvent), but most event loops have acceptable
1035 performance with or without AnyEvent.
1036
1037 * The overhead AnyEvent adds is usually much smaller than the overhead
1038 of the actual event loop, only with extremely fast event loops such
1039 as EV adds AnyEvent significant overhead.
1040
1041 * You should avoid POE like the plague if you want performance or
1042 reasonable memory usage.
1043
1044 BENCHMARKING THE LARGE SERVER CASE
1045 This benchmark actually benchmarks the event loop itself. It works by
1046 creating a number of "servers": each server consists of a socket pair, a
1047 timeout watcher that gets reset on activity (but never fires), and an
1048 I/O watcher waiting for input on one side of the socket. Each time the
1049 socket watcher reads a byte it will write that byte to a random other
1050 "server".
1051
1052 The effect is that there will be a lot of I/O watchers, only part of
1053 which are active at any one point (so there is a constant number of
1054 active fds for each loop iteration, but which fds these are is random).
1055 The timeout is reset each time something is read because that reflects
1056 how most timeouts work (and puts extra pressure on the event loops).
1057
1058 In this benchmark, we use 10000 socket pairs (20000 sockets), of which
1059 100 (1%) are active. This mirrors the activity of large servers with
1060 many connections, most of which are idle at any one point in time.
1061
1062 Source code for this benchmark is found as eg/bench2 in the AnyEvent
1063 distribution.
1064
1065 Explanation of the columns
1066 *sockets* is the number of sockets, and twice the number of "servers"
1067 (as each server has a read and write socket end).
1068
1069 *create* is the time it takes to create a socket pair (which is
1070 nontrivial) and two watchers: an I/O watcher and a timeout watcher.
1071
1072 *request*, the most important value, is the time it takes to handle a
1073 single "request", that is, reading the token from the pipe and
1074 forwarding it to another server. This includes deleting the old timeout
1075 and creating a new one that moves the timeout into the future.
1076
1077 Results
1078 name sockets create request
1079 EV 20000 69.01 11.16
1080 Perl 20000 73.32 35.87
1081 Event 20000 212.62 257.32
1082 Glib 20000 651.16 1896.30
1083 POE 20000 349.67 12317.24 uses POE::Loop::Event
1084
1085 Discussion
1086 This benchmark *does* measure scalability and overall performance of the
1087 particular event loop.
1088
1089 EV is again fastest. Since it is using epoll on my system, the setup
1090 time is relatively high, though.
1091
1092 Perl surprisingly comes second. It is much faster than the C-based event
1093 loops Event and Glib.
1094
1095 Event suffers from high setup time as well (look at its code and you
1096 will understand why). Callback invocation also has a high overhead
1097 compared to the "$_->() for .."-style loop that the Perl event loop
1098 uses. Event uses select or poll in basically all documented
1099 configurations.
1100
1101 Glib is hit hard by its quadratic behaviour w.r.t. many watchers. It
1102 clearly fails to perform with many filehandles or in busy servers.
1103
1104 POE is still completely out of the picture, taking over 1000 times as
1105 long as EV, and over 100 times as long as the Perl implementation, even
1106 though it uses a C-based event loop in this case.
1107
1108 Summary
1109 * The pure perl implementation performs extremely well.
1110
1111 * Avoid Glib or POE in large projects where performance matters.
1112
1113 BENCHMARKING SMALL SERVERS
1114 While event loops should scale (and select-based ones do not...) even to
1115 large servers, most programs we (or I :) actually write have only a few
1116 I/O watchers.
1117
1118 In this benchmark, I use the same benchmark program as in the large
1119 server case, but it uses only eight "servers", of which three are active
1120 at any one time. This should reflect performance for a small server
1121 relatively well.
1122
1123 The columns are identical to the previous table.
1124
1125 Results
1126 name sockets create request
1127 EV 16 20.00 6.54
1128 Perl 16 25.75 12.62
1129 Event 16 81.27 35.86
1130 Glib 16 32.63 15.48
1131 POE 16 261.87 276.28 uses POE::Loop::Event
1132
1133 Discussion
1134 The benchmark tries to test the performance of a typical small server.
1135 While knowing how various event loops perform is interesting, keep in
1136 mind that their overhead in this case is usually not as important, due
1137 to the small absolute number of watchers (that is, you need efficiency
1138 and speed most when you have lots of watchers, not when you only have a
1139 few of them).
1140
1141 EV is again fastest.
1142
1143 Perl again comes second. It is noticeably faster than the C-based event
1144 loops Event and Glib, although the difference is too small to really
1145 matter.
1146
1147 POE also performs much better in this case, but is is still far behind
1148 the others.
1149
1150 Summary
1151 * C-based event loops perform very well with small number of watchers,
1152 as the management overhead dominates.
1153
1154FORK
1155 Most event libraries are not fork-safe. The ones who are usually are
1156 because they rely on inefficient but fork-safe "select" or "poll" calls.
1157 Only EV is fully fork-aware.
1158
1159 If you have to fork, you must either do so *before* creating your first
1160 watcher OR you must not use AnyEvent at all in the child.
1161
1162SECURITY CONSIDERATIONS
1163 AnyEvent can be forced to load any event model via
1164 $ENV{PERL_ANYEVENT_MODEL}. While this cannot (to my knowledge) be used
1165 to execute arbitrary code or directly gain access, it can easily be used
1166 to make the program hang or malfunction in subtle ways, as AnyEvent
1167 watchers will not be active when the program uses a different event
1168 model than specified in the variable.
1169
1170 You can make AnyEvent completely ignore this variable by deleting it
1171 before the first watcher gets created, e.g. with a "BEGIN" block:
1172
1173 BEGIN { delete $ENV{PERL_ANYEVENT_MODEL} }
1174
1175 use AnyEvent;
1176
1177 Similar considerations apply to $ENV{PERL_ANYEVENT_VERBOSE}, as that can
1178 be used to probe what backend is used and gain other information (which
1179 is probably even less useful to an attacker than PERL_ANYEVENT_MODEL).
468 1180
469SEE ALSO 1181SEE ALSO
470 Event modules: Coro::EV, EV, EV::Glib, Glib::EV, Coro::Event, Event, 1182 Utility functions: AnyEvent::Util.
471 Glib::Event, Glib, Coro, Tk.
472 1183
473 Implementations: AnyEvent::Impl::CoroEV, AnyEvent::Impl::EV, 1184 Event modules: EV, EV::Glib, Glib::EV, Event, Glib::Event, Glib, Tk,
474 AnyEvent::Impl::CoroEvent, AnyEvent::Impl::Event, AnyEvent::Impl::Glib, 1185 Event::Lib, Qt, POE.
1186
1187 Implementations: AnyEvent::Impl::EV, AnyEvent::Impl::Event,
475 AnyEvent::Impl::Tk, AnyEvent::Impl::Perl. 1188 AnyEvent::Impl::Glib, AnyEvent::Impl::Tk, AnyEvent::Impl::Perl,
1189 AnyEvent::Impl::EventLib, AnyEvent::Impl::Qt, AnyEvent::Impl::POE.
476 1190
1191 Non-blocking file handles, sockets, TCP clients and servers:
1192 AnyEvent::Handle, AnyEvent::Socket.
1193
1194 Asynchronous DNS: AnyEvent::DNS.
1195
1196 Coroutine support: Coro, Coro::AnyEvent, Coro::EV, Coro::Event,
1197
477 Nontrivial usage examples: Net::FCP, Net::XMPP2. 1198 Nontrivial usage examples: Net::FCP, Net::XMPP2, AnyEvent::DNS.
478 1199
1200AUTHOR
1201 Marc Lehmann <schmorp@schmorp.de>
1202 http://home.schmorp.de/
479 1203

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines