1AnyEvent::Impl::IOAsyncU(s3e)r Contributed Perl DocumentaAtniyoEnvent::Impl::IOAsync(3)
2
3
4
6 AnyEvent::Impl::IOAsync - AnyEvent adaptor for IO::Async
7
9 use AnyEvent;
10 use IO::Async::Loop;
11
12 # optionally set another event loop
13 use AnyEvent::Impl::IOAsync;
14 my $loop = new IO::Async::Loop;
15 AnyEvent::Impl::IOAsync::set_loop $loop;
16
18 This module provides support for IO::Async as AnyEvent backend. It
19 supports I/O, timers, signals and child process watchers. Idle watchers
20 are emulated. I/O watchers need to dup their fh because IO::Async only
21 supports IO handles, not plain file descriptors.
22
24 The only user-servicible part in this module is the "set_loop" function
25 and $LOOP variable:
26
27 AnyEvent::Impl::IOAsync::set_loop $new_loop
28 Unfortunately, IO::Async has no concept of a default loop. Modules
29 using IO::Async must be told by their caller which loop to use,
30 which makes it impossible to transparently use IO::Async from a
31 module.
32
33 This module is no exception. It creates a new IO::Async::Loop
34 object when it is loaded. This might not be the right loop object,
35 though, and thus you can replace it by a call to this function with
36 the loop object of your choice.
37
38 Note that switching loops while watchers are already initialised
39 can have unexpected effects, and is not supported unless you can
40 live witht he consequences.
41
42 $AnyEvent::Impl::IOAsync::LOOP
43 This variable always contains the IO::Async::Loop object used by
44 this AnyEvent backend. See above for more info.
45
46 Storing the "default" loop makes this module a possible arbiter for
47 other modules that want to use IO::Async transparently. It's
48 advised to directly refer to this variable each time you want to
49 use it, without making a local copy.
50
52 This section had a long list of problems and shortcomings that made it
53 almost impossible to support IO::Async. With version 0.33 of IO::Async,
54 however, most of these have been fixed, so IO::Async can now be used as
55 easily as many other loops.
56
57 There are a few remaining problems that require emulation or
58 workarounds:
59
60 No support for multiple watchers per event
61 In most (all? documentation?) cases you cannot have multiple
62 watchers for the same event (what's the point of having all these
63 fancy notifier classes when you cannot have multiple notifiers for
64 the same event? That's like only allowing one timer per second or
65 so...).
66
67 For I/O watchers, AnyEvent has to dup() every file handle, as
68 IO::Async fails to support the same or different file handles
69 pointing to the same fd (the good thing is that it is documented,
70 but why not fix it instead?).
71
72 Apart from these fatal flaws, there are a number of unpleasent
73 properties that just need some mentioning:
74
75 Confusing and misleading names
76 Another rather negative point about this module family is its name,
77 which is deeply confusing: Despite the "async" in the name,
78 IO::Async only does synchronous I/O, there is nothing
79 "asynchronous" about it whatsoever (when I first heard about it, I
80 thought, "wow, a second async I/O module, what does it do compared
81 to IO::AIO", and was somehow set back when I learned that the only
82 "async" aspect of it is the name).
83
84 Inconsistent, incomplete and convoluted API
85 Implementing AnyEvent's rather simple timers on top of IO::Async's
86 timers was a nightmare (try implementing a timer with configurable
87 interval and delay value...).
88
89 The method naming is chaotic: "watch_child" creates a child
90 watcher, but "watch_io" is an internal method; "detach_signal"
91 removes a signal watcher, but "detach_child" forks a subprocess and
92 so on).
93
94 Unpleasant surprises on GNU/Linux
95 When you develop your program on FreeBSD and run it on GNU/Linux,
96 you might have unpleasant surprises, as IO::Async::Loop will by
97 default use IO::Async::Loop::Epoll, which is incompatible with
98 "fork", so your network server will run into spurious and very hard
99 to debug problems under heavy load, as IO::Async forks a lot of
100 processes, e.g. for DNS resolution. It would be better if IO::Async
101 would only load "safe" backends by default (or fix the epoll
102 backend to work in the presence of fork, which admittedly is hard -
103 EV does it for you, and also does not use unsafe backends by
104 default).
105
106 On the positive side, performance with IO::Async is quite good even in
107 my very demanding eyes.
108
110 AnyEvent, IO::Async.
111
113 Marc Lehmann <schmorp@schmorp.de>
114 http://anyevent.schmorp.de
115
116 Paul Evans <leonerd@leonerd.org.uk>
117 Rewrote the backend for IO::Async version 0.33.
118
119
120
121perl v5.36.0 2022-07-22 AnyEvent::Impl::IOAsync(3)