… | |
… | |
176 | =head2 I/O WATCHERS |
176 | =head2 I/O WATCHERS |
177 | |
177 | |
178 | You can create an I/O watcher by calling the C<< AnyEvent->io >> method |
178 | You can create an I/O watcher by calling the C<< AnyEvent->io >> method |
179 | with the following mandatory key-value pairs as arguments: |
179 | with the following mandatory key-value pairs as arguments: |
180 | |
180 | |
181 | C<fh> is the Perl I<file handle> (I<not> file descriptor) to watch |
181 | C<fh> is the Perl I<file handle> (or a naked file descriptor) to watch |
182 | for events (AnyEvent might or might not keep a reference to this file |
182 | for events (AnyEvent might or might not keep a reference to this file |
183 | handle). Note that only file handles pointing to things for which |
183 | handle). Note that only file handles pointing to things for which |
184 | non-blocking operation makes sense are allowed. This includes sockets, |
184 | non-blocking operation makes sense are allowed. This includes sockets, |
185 | most character devices, pipes, fifos and so on, but not for example files |
185 | most character devices, pipes, fifos and so on, but not for example files |
186 | or block devices. |
186 | or block devices. |
… | |
… | |
392 | |
392 | |
393 | There is a slight catch to child watchers, however: you usually start them |
393 | There is a slight catch to child watchers, however: you usually start them |
394 | I<after> the child process was created, and this means the process could |
394 | I<after> the child process was created, and this means the process could |
395 | have exited already (and no SIGCHLD will be sent anymore). |
395 | have exited already (and no SIGCHLD will be sent anymore). |
396 | |
396 | |
397 | Not all event models handle this correctly (POE doesn't), but even for |
397 | Not all event models handle this correctly (neither POE nor IO::Async do, |
|
|
398 | see their AnyEvent::Impl manpages for details), but even for event models |
398 | event models that I<do> handle this correctly, they usually need to be |
399 | that I<do> handle this correctly, they usually need to be loaded before |
399 | loaded before the process exits (i.e. before you fork in the first place). |
400 | the process exits (i.e. before you fork in the first place). AnyEvent's |
|
|
401 | pure perl event loop handles all cases correctly regardless of when you |
|
|
402 | start the watcher. |
400 | |
403 | |
401 | This means you cannot create a child watcher as the very first thing in an |
404 | This means you cannot create a child watcher as the very first |
402 | AnyEvent program, you I<have> to create at least one watcher before you |
405 | thing in an AnyEvent program, you I<have> to create at least one |
403 | C<fork> the child (alternatively, you can call C<AnyEvent::detect>). |
406 | watcher before you C<fork> the child (alternatively, you can call |
|
|
407 | C<AnyEvent::detect>). |
404 | |
408 | |
405 | Example: fork a process and wait for it |
409 | Example: fork a process and wait for it |
406 | |
410 | |
407 | my $done = AnyEvent->condvar; |
411 | my $done = AnyEvent->condvar; |
408 | |
412 | |
… | |
… | |
595 | |
599 | |
596 | =item $cv->begin ([group callback]) |
600 | =item $cv->begin ([group callback]) |
597 | |
601 | |
598 | =item $cv->end |
602 | =item $cv->end |
599 | |
603 | |
600 | These two methods are EXPERIMENTAL and MIGHT CHANGE. |
|
|
601 | |
|
|
602 | These two methods can be used to combine many transactions/events into |
604 | These two methods can be used to combine many transactions/events into |
603 | one. For example, a function that pings many hosts in parallel might want |
605 | one. For example, a function that pings many hosts in parallel might want |
604 | to use a condition variable for the whole process. |
606 | to use a condition variable for the whole process. |
605 | |
607 | |
606 | Every call to C<< ->begin >> will increment a counter, and every call to |
608 | Every call to C<< ->begin >> will increment a counter, and every call to |
607 | C<< ->end >> will decrement it. If the counter reaches C<0> in C<< ->end |
609 | C<< ->end >> will decrement it. If the counter reaches C<0> in C<< ->end |
608 | >>, the (last) callback passed to C<begin> will be executed. That callback |
610 | >>, the (last) callback passed to C<begin> will be executed. That callback |
609 | is I<supposed> to call C<< ->send >>, but that is not required. If no |
611 | is I<supposed> to call C<< ->send >>, but that is not required. If no |
610 | callback was set, C<send> will be called without any arguments. |
612 | callback was set, C<send> will be called without any arguments. |
611 | |
613 | |
612 | Let's clarify this with the ping example: |
614 | You can think of C<< $cv->send >> giving you an OR condition (one call |
|
|
615 | sends), while C<< $cv->begin >> and C<< $cv->end >> giving you an AND |
|
|
616 | condition (all C<begin> calls must be C<end>'ed before the condvar sends). |
|
|
617 | |
|
|
618 | Let's start with a simple example: you have two I/O watchers (for example, |
|
|
619 | STDOUT and STDERR for a program), and you want to wait for both streams to |
|
|
620 | close before activating a condvar: |
|
|
621 | |
|
|
622 | my $cv = AnyEvent->condvar; |
|
|
623 | |
|
|
624 | $cv->begin; # first watcher |
|
|
625 | my $w1 = AnyEvent->io (fh => $fh1, cb => sub { |
|
|
626 | defined sysread $fh1, my $buf, 4096 |
|
|
627 | or $cv->end; |
|
|
628 | }); |
|
|
629 | |
|
|
630 | $cv->begin; # second watcher |
|
|
631 | my $w2 = AnyEvent->io (fh => $fh2, cb => sub { |
|
|
632 | defined sysread $fh2, my $buf, 4096 |
|
|
633 | or $cv->end; |
|
|
634 | }); |
|
|
635 | |
|
|
636 | $cv->recv; |
|
|
637 | |
|
|
638 | This works because for every event source (EOF on file handle), there is |
|
|
639 | one call to C<begin>, so the condvar waits for all calls to C<end> before |
|
|
640 | sending. |
|
|
641 | |
|
|
642 | The ping example mentioned above is slightly more complicated, as the |
|
|
643 | there are results to be passwd back, and the number of tasks that are |
|
|
644 | begung can potentially be zero: |
613 | |
645 | |
614 | my $cv = AnyEvent->condvar; |
646 | my $cv = AnyEvent->condvar; |
615 | |
647 | |
616 | my %result; |
648 | my %result; |
617 | $cv->begin (sub { $cv->send (\%result) }); |
649 | $cv->begin (sub { $cv->send (\%result) }); |
… | |
… | |
637 | loop, which serves two important purposes: first, it sets the callback |
669 | loop, which serves two important purposes: first, it sets the callback |
638 | to be called once the counter reaches C<0>, and second, it ensures that |
670 | to be called once the counter reaches C<0>, and second, it ensures that |
639 | C<send> is called even when C<no> hosts are being pinged (the loop |
671 | C<send> is called even when C<no> hosts are being pinged (the loop |
640 | doesn't execute once). |
672 | doesn't execute once). |
641 | |
673 | |
642 | This is the general pattern when you "fan out" into multiple subrequests: |
674 | This is the general pattern when you "fan out" into multiple (but |
643 | use an outer C<begin>/C<end> pair to set the callback and ensure C<end> |
675 | potentially none) subrequests: use an outer C<begin>/C<end> pair to set |
644 | is called at least once, and then, for each subrequest you start, call |
676 | the callback and ensure C<end> is called at least once, and then, for each |
645 | C<begin> and for each subrequest you finish, call C<end>. |
677 | subrequest you start, call C<begin> and for each subrequest you finish, |
|
|
678 | call C<end>. |
646 | |
679 | |
647 | =back |
680 | =back |
648 | |
681 | |
649 | =head3 METHODS FOR CONSUMERS |
682 | =head3 METHODS FOR CONSUMERS |
650 | |
683 | |
… | |
… | |
730 | AnyEvent::Impl::Tk based on Tk, very bad choice. |
763 | AnyEvent::Impl::Tk based on Tk, very bad choice. |
731 | AnyEvent::Impl::Qt based on Qt, cannot be autoprobed (see its docs). |
764 | AnyEvent::Impl::Qt based on Qt, cannot be autoprobed (see its docs). |
732 | AnyEvent::Impl::EventLib based on Event::Lib, leaks memory and worse. |
765 | AnyEvent::Impl::EventLib based on Event::Lib, leaks memory and worse. |
733 | AnyEvent::Impl::POE based on POE, not generic enough for full support. |
766 | AnyEvent::Impl::POE based on POE, not generic enough for full support. |
734 | |
767 | |
|
|
768 | # warning, support for IO::Async is only partial, as it is too broken |
|
|
769 | # and limited toe ven support the AnyEvent API. See AnyEvent::Impl::Async. |
|
|
770 | AnyEvent::Impl::IOAsync based on IO::Async, cannot be autoprobed (see its docs). |
|
|
771 | |
735 | There is no support for WxWidgets, as WxWidgets has no support for |
772 | There is no support for WxWidgets, as WxWidgets has no support for |
736 | watching file handles. However, you can use WxWidgets through the |
773 | watching file handles. However, you can use WxWidgets through the |
737 | POE Adaptor, as POE has a Wx backend that simply polls 20 times per |
774 | POE Adaptor, as POE has a Wx backend that simply polls 20 times per |
738 | second, which was considered to be too horrible to even consider for |
775 | second, which was considered to be too horrible to even consider for |
739 | AnyEvent. Likewise, other POE backends can be used by AnyEvent by using |
776 | AnyEvent. Likewise, other POE backends can be used by AnyEvent by using |
… | |
… | |
831 | |
868 | |
832 | |
869 | |
833 | =head1 OTHER MODULES |
870 | =head1 OTHER MODULES |
834 | |
871 | |
835 | The following is a non-exhaustive list of additional modules that use |
872 | The following is a non-exhaustive list of additional modules that use |
836 | AnyEvent and can therefore be mixed easily with other AnyEvent modules |
873 | AnyEvent as a client and can therefore be mixed easily with other AnyEvent |
837 | in the same program. Some of the modules come with AnyEvent, some are |
874 | modules and other event loops in the same program. Some of the modules |
838 | available via CPAN. |
875 | come with AnyEvent, most are available via CPAN. |
839 | |
876 | |
840 | =over 4 |
877 | =over 4 |
841 | |
878 | |
842 | =item L<AnyEvent::Util> |
879 | =item L<AnyEvent::Util> |
843 | |
880 | |
… | |
… | |
852 | |
889 | |
853 | =item L<AnyEvent::Handle> |
890 | =item L<AnyEvent::Handle> |
854 | |
891 | |
855 | Provide read and write buffers, manages watchers for reads and writes, |
892 | Provide read and write buffers, manages watchers for reads and writes, |
856 | supports raw and formatted I/O, I/O queued and fully transparent and |
893 | supports raw and formatted I/O, I/O queued and fully transparent and |
857 | non-blocking SSL/TLS. |
894 | non-blocking SSL/TLS (via L<AnyEvent::TLS>. |
858 | |
895 | |
859 | =item L<AnyEvent::DNS> |
896 | =item L<AnyEvent::DNS> |
860 | |
897 | |
861 | Provides rich asynchronous DNS resolver capabilities. |
898 | Provides rich asynchronous DNS resolver capabilities. |
862 | |
899 | |
… | |
… | |
890 | |
927 | |
891 | =item L<AnyEvent::GPSD> |
928 | =item L<AnyEvent::GPSD> |
892 | |
929 | |
893 | A non-blocking interface to gpsd, a daemon delivering GPS information. |
930 | A non-blocking interface to gpsd, a daemon delivering GPS information. |
894 | |
931 | |
|
|
932 | =item L<AnyEvent::IRC> |
|
|
933 | |
|
|
934 | AnyEvent based IRC client module family (replacing the older Net::IRC3). |
|
|
935 | |
|
|
936 | =item L<AnyEvent::XMPP> |
|
|
937 | |
|
|
938 | AnyEvent based XMPP (Jabber protocol) module family (replacing the older |
|
|
939 | Net::XMPP2>. |
|
|
940 | |
895 | =item L<AnyEvent::IGS> |
941 | =item L<AnyEvent::IGS> |
896 | |
942 | |
897 | A non-blocking interface to the Internet Go Server protocol (used by |
943 | A non-blocking interface to the Internet Go Server protocol (used by |
898 | L<App::IGS>). |
944 | L<App::IGS>). |
899 | |
945 | |
900 | =item L<AnyEvent::IRC> |
|
|
901 | |
|
|
902 | AnyEvent based IRC client module family (replacing the older Net::IRC3). |
|
|
903 | |
|
|
904 | =item L<Net::XMPP2> |
|
|
905 | |
|
|
906 | AnyEvent based XMPP (Jabber protocol) module family. |
|
|
907 | |
|
|
908 | =item L<Net::FCP> |
946 | =item L<Net::FCP> |
909 | |
947 | |
910 | AnyEvent-based implementation of the Freenet Client Protocol, birthplace |
948 | AnyEvent-based implementation of the Freenet Client Protocol, birthplace |
911 | of AnyEvent. |
949 | of AnyEvent. |
912 | |
950 | |
… | |
… | |
916 | |
954 | |
917 | =item L<Coro> |
955 | =item L<Coro> |
918 | |
956 | |
919 | Has special support for AnyEvent via L<Coro::AnyEvent>. |
957 | Has special support for AnyEvent via L<Coro::AnyEvent>. |
920 | |
958 | |
921 | =item L<IO::Lambda> |
|
|
922 | |
|
|
923 | The lambda approach to I/O - don't ask, look there. Can use AnyEvent. |
|
|
924 | |
|
|
925 | =back |
959 | =back |
926 | |
960 | |
927 | =cut |
961 | =cut |
928 | |
962 | |
929 | package AnyEvent; |
963 | package AnyEvent; |
… | |
… | |
931 | no warnings; |
965 | no warnings; |
932 | use strict qw(vars subs); |
966 | use strict qw(vars subs); |
933 | |
967 | |
934 | use Carp; |
968 | use Carp; |
935 | |
969 | |
936 | our $VERSION = 4.411; |
970 | our $VERSION = 4.8; |
937 | our $MODEL; |
971 | our $MODEL; |
938 | |
972 | |
939 | our $AUTOLOAD; |
973 | our $AUTOLOAD; |
940 | our @ISA; |
974 | our @ISA; |
941 | |
975 | |
942 | our @REGISTRY; |
976 | our @REGISTRY; |
943 | |
977 | |
944 | our $WIN32; |
978 | our $WIN32; |
945 | |
979 | |
946 | BEGIN { |
980 | BEGIN { |
947 | my $win32 = ! ! ($^O =~ /mswin32/i); |
981 | eval "sub WIN32(){ " . (($^O =~ /mswin32/i)*1) ." }"; |
948 | eval "sub WIN32(){ $win32 }"; |
982 | eval "sub TAINT(){ " . (${^TAINT}*1) . " }"; |
|
|
983 | |
|
|
984 | delete @ENV{grep /^PERL_ANYEVENT_/, keys %ENV} |
|
|
985 | if ${^TAINT}; |
949 | } |
986 | } |
950 | |
987 | |
951 | our $verbose = $ENV{PERL_ANYEVENT_VERBOSE}*1; |
988 | our $verbose = $ENV{PERL_ANYEVENT_VERBOSE}*1; |
952 | |
989 | |
953 | our %PROTOCOL; # (ipv4|ipv6) => (1|2), higher numbers are preferred |
990 | our %PROTOCOL; # (ipv4|ipv6) => (1|2), higher numbers are preferred |
… | |
… | |
971 | [Event::Lib:: => AnyEvent::Impl::EventLib::], # too buggy |
1008 | [Event::Lib:: => AnyEvent::Impl::EventLib::], # too buggy |
972 | [Qt:: => AnyEvent::Impl::Qt::], # requires special main program |
1009 | [Qt:: => AnyEvent::Impl::Qt::], # requires special main program |
973 | [POE::Kernel:: => AnyEvent::Impl::POE::], # lasciate ogni speranza |
1010 | [POE::Kernel:: => AnyEvent::Impl::POE::], # lasciate ogni speranza |
974 | [Wx:: => AnyEvent::Impl::POE::], |
1011 | [Wx:: => AnyEvent::Impl::POE::], |
975 | [Prima:: => AnyEvent::Impl::POE::], |
1012 | [Prima:: => AnyEvent::Impl::POE::], |
|
|
1013 | # IO::Async is just too broken - we would need workaorunds for its |
|
|
1014 | # byzantine signal and broken child handling, among others. |
|
|
1015 | # IO::Async is rather hard to detect, as it doesn't have any |
|
|
1016 | # obvious default class. |
|
|
1017 | # [IO::Async:: => AnyEvent::Impl::IOAsync::], # requires special main program |
|
|
1018 | # [IO::Async::Loop:: => AnyEvent::Impl::IOAsync::], # requires special main program |
|
|
1019 | # [IO::Async::Notifier:: => AnyEvent::Impl::IOAsync::], # requires special main program |
976 | ); |
1020 | ); |
977 | |
1021 | |
978 | our %method = map +($_ => 1), |
1022 | our %method = map +($_ => 1), |
979 | qw(io timer time now now_update signal child idle condvar one_event DESTROY); |
1023 | qw(io timer time now now_update signal child idle condvar one_event DESTROY); |
980 | |
1024 | |
… | |
… | |
1072 | } |
1116 | } |
1073 | |
1117 | |
1074 | # utility function to dup a filehandle. this is used by many backends |
1118 | # utility function to dup a filehandle. this is used by many backends |
1075 | # to support binding more than one watcher per filehandle (they usually |
1119 | # to support binding more than one watcher per filehandle (they usually |
1076 | # allow only one watcher per fd, so we dup it to get a different one). |
1120 | # allow only one watcher per fd, so we dup it to get a different one). |
1077 | sub _dupfh($$$$) { |
1121 | sub _dupfh($$;$$) { |
1078 | my ($poll, $fh, $r, $w) = @_; |
1122 | my ($poll, $fh, $r, $w) = @_; |
1079 | |
1123 | |
1080 | # cygwin requires the fh mode to be matching, unix doesn't |
1124 | # cygwin requires the fh mode to be matching, unix doesn't |
1081 | my ($rw, $mode) = $poll eq "r" ? ($r, "<") |
1125 | my ($rw, $mode) = $poll eq "r" ? ($r, "<") : ($w, ">"); |
1082 | : $poll eq "w" ? ($w, ">") |
|
|
1083 | : Carp::croak "AnyEvent->io requires poll set to either 'r' or 'w'"; |
|
|
1084 | |
1126 | |
1085 | open my $fh2, "$mode&" . fileno $fh |
1127 | open my $fh2, "$mode&", $fh |
1086 | or die "cannot dup() filehandle: $!,"; |
1128 | or die "AnyEvent->io: cannot dup() filehandle in mode '$poll': $!,"; |
1087 | |
1129 | |
1088 | # we assume CLOEXEC is already set by perl in all important cases |
1130 | # we assume CLOEXEC is already set by perl in all important cases |
1089 | |
1131 | |
1090 | ($fh2, $rw) |
1132 | ($fh2, $rw) |
1091 | } |
1133 | } |
… | |
… | |
1337 | so on. |
1379 | so on. |
1338 | |
1380 | |
1339 | =head1 ENVIRONMENT VARIABLES |
1381 | =head1 ENVIRONMENT VARIABLES |
1340 | |
1382 | |
1341 | The following environment variables are used by this module or its |
1383 | The following environment variables are used by this module or its |
1342 | submodules: |
1384 | submodules. |
|
|
1385 | |
|
|
1386 | Note that AnyEvent will remove I<all> environment variables starting with |
|
|
1387 | C<PERL_ANYEVENT_> from C<%ENV> when it is loaded while taint mode is |
|
|
1388 | enabled. |
1343 | |
1389 | |
1344 | =over 4 |
1390 | =over 4 |
1345 | |
1391 | |
1346 | =item C<PERL_ANYEVENT_VERBOSE> |
1392 | =item C<PERL_ANYEVENT_VERBOSE> |
1347 | |
1393 | |
… | |
… | |
1359 | =item C<PERL_ANYEVENT_STRICT> |
1405 | =item C<PERL_ANYEVENT_STRICT> |
1360 | |
1406 | |
1361 | AnyEvent does not do much argument checking by default, as thorough |
1407 | AnyEvent does not do much argument checking by default, as thorough |
1362 | argument checking is very costly. Setting this variable to a true value |
1408 | argument checking is very costly. Setting this variable to a true value |
1363 | will cause AnyEvent to load C<AnyEvent::Strict> and then to thoroughly |
1409 | will cause AnyEvent to load C<AnyEvent::Strict> and then to thoroughly |
1364 | check the arguments passed to most method calls. If it finds any problems |
1410 | check the arguments passed to most method calls. If it finds any problems, |
1365 | it will croak. |
1411 | it will croak. |
1366 | |
1412 | |
1367 | In other words, enables "strict" mode. |
1413 | In other words, enables "strict" mode. |
1368 | |
1414 | |
1369 | Unlike C<use strict>, it is definitely recommended ot keep it off in |
1415 | Unlike C<use strict>, it is definitely recommended to keep it off in |
1370 | production. Keeping C<PERL_ANYEVENT_STRICT=1> in your environment while |
1416 | production. Keeping C<PERL_ANYEVENT_STRICT=1> in your environment while |
1371 | developing programs can be very useful, however. |
1417 | developing programs can be very useful, however. |
1372 | |
1418 | |
1373 | =item C<PERL_ANYEVENT_MODEL> |
1419 | =item C<PERL_ANYEVENT_MODEL> |
1374 | |
1420 | |
… | |
… | |
1419 | |
1465 | |
1420 | =item C<PERL_ANYEVENT_MAX_FORKS> |
1466 | =item C<PERL_ANYEVENT_MAX_FORKS> |
1421 | |
1467 | |
1422 | The maximum number of child processes that C<AnyEvent::Util::fork_call> |
1468 | The maximum number of child processes that C<AnyEvent::Util::fork_call> |
1423 | will create in parallel. |
1469 | will create in parallel. |
|
|
1470 | |
|
|
1471 | =item C<PERL_ANYEVENT_MAX_OUTSTANDING_DNS> |
|
|
1472 | |
|
|
1473 | The default value for the C<max_outstanding> parameter for the default DNS |
|
|
1474 | resolver - this is the maximum number of parallel DNS requests that are |
|
|
1475 | sent to the DNS server. |
|
|
1476 | |
|
|
1477 | =item C<PERL_ANYEVENT_RESOLV_CONF> |
|
|
1478 | |
|
|
1479 | The file to use instead of F</etc/resolv.conf> (or OS-specific |
|
|
1480 | configuration) in the default resolver. When set to the empty string, no |
|
|
1481 | default config will be used. |
|
|
1482 | |
|
|
1483 | =item C<PERL_ANYEVENT_CA_FILE>, C<PERL_ANYEVENT_CA_PATH>. |
|
|
1484 | |
|
|
1485 | When neither C<ca_file> nor C<ca_path> was specified during |
|
|
1486 | L<AnyEvent::TLS> context creation, and either of these environment |
|
|
1487 | variables exist, they will be used to specify CA certificate locations |
|
|
1488 | instead of a system-dependent default. |
1424 | |
1489 | |
1425 | =back |
1490 | =back |
1426 | |
1491 | |
1427 | =head1 SUPPLYING YOUR OWN EVENT MODEL INTERFACE |
1492 | =head1 SUPPLYING YOUR OWN EVENT MODEL INTERFACE |
1428 | |
1493 | |
… | |
… | |
1673 | EV/Any 100000 224 2.88 0.34 0.27 EV + AnyEvent watchers |
1738 | EV/Any 100000 224 2.88 0.34 0.27 EV + AnyEvent watchers |
1674 | CoroEV/Any 100000 224 2.85 0.35 0.28 coroutines + Coro::Signal |
1739 | CoroEV/Any 100000 224 2.85 0.35 0.28 coroutines + Coro::Signal |
1675 | Perl/Any 100000 452 4.13 0.73 0.95 pure perl implementation |
1740 | Perl/Any 100000 452 4.13 0.73 0.95 pure perl implementation |
1676 | Event/Event 16000 517 32.20 31.80 0.81 Event native interface |
1741 | Event/Event 16000 517 32.20 31.80 0.81 Event native interface |
1677 | Event/Any 16000 590 35.85 31.55 1.06 Event + AnyEvent watchers |
1742 | Event/Any 16000 590 35.85 31.55 1.06 Event + AnyEvent watchers |
|
|
1743 | IOAsync/Any 16000 989 38.10 32.77 11.13 via IO::Async::Loop::IO_Poll |
|
|
1744 | IOAsync/Any 16000 990 37.59 29.50 10.61 via IO::Async::Loop::Epoll |
1678 | Glib/Any 16000 1357 102.33 12.31 51.00 quadratic behaviour |
1745 | Glib/Any 16000 1357 102.33 12.31 51.00 quadratic behaviour |
1679 | Tk/Any 2000 1860 27.20 66.31 14.00 SEGV with >> 2000 watchers |
1746 | Tk/Any 2000 1860 27.20 66.31 14.00 SEGV with >> 2000 watchers |
1680 | POE/Event 2000 6328 109.99 751.67 14.02 via POE::Loop::Event |
1747 | POE/Event 2000 6328 109.99 751.67 14.02 via POE::Loop::Event |
1681 | POE/Select 2000 6027 94.54 809.13 579.80 via POE::Loop::Select |
1748 | POE/Select 2000 6027 94.54 809.13 579.80 via POE::Loop::Select |
1682 | |
1749 | |
… | |
… | |
1711 | performance becomes really bad with lots of file descriptors (and few of |
1778 | performance becomes really bad with lots of file descriptors (and few of |
1712 | them active), of course, but this was not subject of this benchmark. |
1779 | them active), of course, but this was not subject of this benchmark. |
1713 | |
1780 | |
1714 | The C<Event> module has a relatively high setup and callback invocation |
1781 | The C<Event> module has a relatively high setup and callback invocation |
1715 | cost, but overall scores in on the third place. |
1782 | cost, but overall scores in on the third place. |
|
|
1783 | |
|
|
1784 | C<IO::Async> performs admirably well, about on par with C<Event>, even |
|
|
1785 | when using its pure perl backend. |
1716 | |
1786 | |
1717 | C<Glib>'s memory usage is quite a bit higher, but it features a |
1787 | C<Glib>'s memory usage is quite a bit higher, but it features a |
1718 | faster callback invocation and overall ends up in the same class as |
1788 | faster callback invocation and overall ends up in the same class as |
1719 | C<Event>. However, Glib scales extremely badly, doubling the number of |
1789 | C<Event>. However, Glib scales extremely badly, doubling the number of |
1720 | watchers increases the processing time by more than a factor of four, |
1790 | watchers increases the processing time by more than a factor of four, |
… | |
… | |
1798 | it to another server. This includes deleting the old timeout and creating |
1868 | it to another server. This includes deleting the old timeout and creating |
1799 | a new one that moves the timeout into the future. |
1869 | a new one that moves the timeout into the future. |
1800 | |
1870 | |
1801 | =head3 Results |
1871 | =head3 Results |
1802 | |
1872 | |
1803 | name sockets create request |
1873 | name sockets create request |
1804 | EV 20000 69.01 11.16 |
1874 | EV 20000 69.01 11.16 |
1805 | Perl 20000 73.32 35.87 |
1875 | Perl 20000 73.32 35.87 |
|
|
1876 | IOAsync 20000 157.00 98.14 epoll |
|
|
1877 | IOAsync 20000 159.31 616.06 poll |
1806 | Event 20000 212.62 257.32 |
1878 | Event 20000 212.62 257.32 |
1807 | Glib 20000 651.16 1896.30 |
1879 | Glib 20000 651.16 1896.30 |
1808 | POE 20000 349.67 12317.24 uses POE::Loop::Event |
1880 | POE 20000 349.67 12317.24 uses POE::Loop::Event |
1809 | |
1881 | |
1810 | =head3 Discussion |
1882 | =head3 Discussion |
1811 | |
1883 | |
1812 | This benchmark I<does> measure scalability and overall performance of the |
1884 | This benchmark I<does> measure scalability and overall performance of the |
1813 | particular event loop. |
1885 | particular event loop. |
… | |
… | |
1815 | EV is again fastest. Since it is using epoll on my system, the setup time |
1887 | EV is again fastest. Since it is using epoll on my system, the setup time |
1816 | is relatively high, though. |
1888 | is relatively high, though. |
1817 | |
1889 | |
1818 | Perl surprisingly comes second. It is much faster than the C-based event |
1890 | Perl surprisingly comes second. It is much faster than the C-based event |
1819 | loops Event and Glib. |
1891 | loops Event and Glib. |
|
|
1892 | |
|
|
1893 | IO::Async performs very well when using its epoll backend, and still quite |
|
|
1894 | good compared to Glib when using its pure perl backend. |
1820 | |
1895 | |
1821 | Event suffers from high setup time as well (look at its code and you will |
1896 | Event suffers from high setup time as well (look at its code and you will |
1822 | understand why). Callback invocation also has a high overhead compared to |
1897 | understand why). Callback invocation also has a high overhead compared to |
1823 | the C<< $_->() for .. >>-style loop that the Perl event loop uses. Event |
1898 | the C<< $_->() for .. >>-style loop that the Perl event loop uses. Event |
1824 | uses select or poll in basically all documented configurations. |
1899 | uses select or poll in basically all documented configurations. |
… | |
… | |
1887 | =item * C-based event loops perform very well with small number of |
1962 | =item * C-based event loops perform very well with small number of |
1888 | watchers, as the management overhead dominates. |
1963 | watchers, as the management overhead dominates. |
1889 | |
1964 | |
1890 | =back |
1965 | =back |
1891 | |
1966 | |
|
|
1967 | =head2 THE IO::Lambda BENCHMARK |
|
|
1968 | |
|
|
1969 | Recently I was told about the benchmark in the IO::Lambda manpage, which |
|
|
1970 | could be misinterpreted to make AnyEvent look bad. In fact, the benchmark |
|
|
1971 | simply compares IO::Lambda with POE, and IO::Lambda looks better (which |
|
|
1972 | shouldn't come as a surprise to anybody). As such, the benchmark is |
|
|
1973 | fine, and mostly shows that the AnyEvent backend from IO::Lambda isn't |
|
|
1974 | very optimal. But how would AnyEvent compare when used without the extra |
|
|
1975 | baggage? To explore this, I wrote the equivalent benchmark for AnyEvent. |
|
|
1976 | |
|
|
1977 | The benchmark itself creates an echo-server, and then, for 500 times, |
|
|
1978 | connects to the echo server, sends a line, waits for the reply, and then |
|
|
1979 | creates the next connection. This is a rather bad benchmark, as it doesn't |
|
|
1980 | test the efficiency of the framework or much non-blocking I/O, but it is a |
|
|
1981 | benchmark nevertheless. |
|
|
1982 | |
|
|
1983 | name runtime |
|
|
1984 | Lambda/select 0.330 sec |
|
|
1985 | + optimized 0.122 sec |
|
|
1986 | Lambda/AnyEvent 0.327 sec |
|
|
1987 | + optimized 0.138 sec |
|
|
1988 | Raw sockets/select 0.077 sec |
|
|
1989 | POE/select, components 0.662 sec |
|
|
1990 | POE/select, raw sockets 0.226 sec |
|
|
1991 | POE/select, optimized 0.404 sec |
|
|
1992 | |
|
|
1993 | AnyEvent/select/nb 0.085 sec |
|
|
1994 | AnyEvent/EV/nb 0.068 sec |
|
|
1995 | +state machine 0.134 sec |
|
|
1996 | |
|
|
1997 | The benchmark is also a bit unfair (my fault): the IO::Lambda/POE |
|
|
1998 | benchmarks actually make blocking connects and use 100% blocking I/O, |
|
|
1999 | defeating the purpose of an event-based solution. All of the newly |
|
|
2000 | written AnyEvent benchmarks use 100% non-blocking connects (using |
|
|
2001 | AnyEvent::Socket::tcp_connect and the asynchronous pure perl DNS |
|
|
2002 | resolver), so AnyEvent is at a disadvantage here, as non-blocking connects |
|
|
2003 | generally require a lot more bookkeeping and event handling than blocking |
|
|
2004 | connects (which involve a single syscall only). |
|
|
2005 | |
|
|
2006 | The last AnyEvent benchmark additionally uses L<AnyEvent::Handle>, which |
|
|
2007 | offers similar expressive power as POE and IO::Lambda, using conventional |
|
|
2008 | Perl syntax. This means that both the echo server and the client are 100% |
|
|
2009 | non-blocking, further placing it at a disadvantage. |
|
|
2010 | |
|
|
2011 | As you can see, the AnyEvent + EV combination even beats the |
|
|
2012 | hand-optimised "raw sockets benchmark", while AnyEvent + its pure perl |
|
|
2013 | backend easily beats IO::Lambda and POE. |
|
|
2014 | |
|
|
2015 | And even the 100% non-blocking version written using the high-level (and |
|
|
2016 | slow :) L<AnyEvent::Handle> abstraction beats both POE and IO::Lambda by a |
|
|
2017 | large margin, even though it does all of DNS, tcp-connect and socket I/O |
|
|
2018 | in a non-blocking way. |
|
|
2019 | |
|
|
2020 | The two AnyEvent benchmarks programs can be found as F<eg/ae0.pl> and |
|
|
2021 | F<eg/ae2.pl> in the AnyEvent distribution, the remaining benchmarks are |
|
|
2022 | part of the IO::lambda distribution and were used without any changes. |
|
|
2023 | |
1892 | |
2024 | |
1893 | =head1 SIGNALS |
2025 | =head1 SIGNALS |
1894 | |
2026 | |
1895 | AnyEvent currently installs handlers for these signals: |
2027 | AnyEvent currently installs handlers for these signals: |
1896 | |
2028 | |
… | |
… | |
1899 | =item SIGCHLD |
2031 | =item SIGCHLD |
1900 | |
2032 | |
1901 | A handler for C<SIGCHLD> is installed by AnyEvent's child watcher |
2033 | A handler for C<SIGCHLD> is installed by AnyEvent's child watcher |
1902 | emulation for event loops that do not support them natively. Also, some |
2034 | emulation for event loops that do not support them natively. Also, some |
1903 | event loops install a similar handler. |
2035 | event loops install a similar handler. |
|
|
2036 | |
|
|
2037 | If, when AnyEvent is loaded, SIGCHLD is set to IGNORE, then AnyEvent will |
|
|
2038 | reset it to default, to avoid losing child exit statuses. |
1904 | |
2039 | |
1905 | =item SIGPIPE |
2040 | =item SIGPIPE |
1906 | |
2041 | |
1907 | A no-op handler is installed for C<SIGPIPE> when C<$SIG{PIPE}> is C<undef> |
2042 | A no-op handler is installed for C<SIGPIPE> when C<$SIG{PIPE}> is C<undef> |
1908 | when AnyEvent gets loaded. |
2043 | when AnyEvent gets loaded. |
… | |
… | |
1920 | |
2055 | |
1921 | =back |
2056 | =back |
1922 | |
2057 | |
1923 | =cut |
2058 | =cut |
1924 | |
2059 | |
|
|
2060 | undef $SIG{CHLD} |
|
|
2061 | if $SIG{CHLD} eq 'IGNORE'; |
|
|
2062 | |
1925 | $SIG{PIPE} = sub { } |
2063 | $SIG{PIPE} = sub { } |
1926 | unless defined $SIG{PIPE}; |
2064 | unless defined $SIG{PIPE}; |
1927 | |
|
|
1928 | |
2065 | |
1929 | =head1 FORK |
2066 | =head1 FORK |
1930 | |
2067 | |
1931 | Most event libraries are not fork-safe. The ones who are usually are |
2068 | Most event libraries are not fork-safe. The ones who are usually are |
1932 | because they rely on inefficient but fork-safe C<select> or C<poll> |
2069 | because they rely on inefficient but fork-safe C<select> or C<poll> |
… | |
… | |
1955 | Similar considerations apply to $ENV{PERL_ANYEVENT_VERBOSE}, as that can |
2092 | Similar considerations apply to $ENV{PERL_ANYEVENT_VERBOSE}, as that can |
1956 | be used to probe what backend is used and gain other information (which is |
2093 | be used to probe what backend is used and gain other information (which is |
1957 | probably even less useful to an attacker than PERL_ANYEVENT_MODEL), and |
2094 | probably even less useful to an attacker than PERL_ANYEVENT_MODEL), and |
1958 | $ENV{PERL_ANYEVENT_STRICT}. |
2095 | $ENV{PERL_ANYEVENT_STRICT}. |
1959 | |
2096 | |
|
|
2097 | Note that AnyEvent will remove I<all> environment variables starting with |
|
|
2098 | C<PERL_ANYEVENT_> from C<%ENV> when it is loaded while taint mode is |
|
|
2099 | enabled. |
|
|
2100 | |
1960 | |
2101 | |
1961 | =head1 BUGS |
2102 | =head1 BUGS |
1962 | |
2103 | |
1963 | Perl 5.8 has numerous memleaks that sometimes hit this module and are hard |
2104 | Perl 5.8 has numerous memleaks that sometimes hit this module and are hard |
1964 | to work around. If you suffer from memleaks, first upgrade to Perl 5.10 |
2105 | to work around. If you suffer from memleaks, first upgrade to Perl 5.10 |
… | |
… | |
1975 | L<Glib>, L<Tk>, L<Event::Lib>, L<Qt>, L<POE>. |
2116 | L<Glib>, L<Tk>, L<Event::Lib>, L<Qt>, L<POE>. |
1976 | |
2117 | |
1977 | Implementations: L<AnyEvent::Impl::EV>, L<AnyEvent::Impl::Event>, |
2118 | Implementations: L<AnyEvent::Impl::EV>, L<AnyEvent::Impl::Event>, |
1978 | L<AnyEvent::Impl::Glib>, L<AnyEvent::Impl::Tk>, L<AnyEvent::Impl::Perl>, |
2119 | L<AnyEvent::Impl::Glib>, L<AnyEvent::Impl::Tk>, L<AnyEvent::Impl::Perl>, |
1979 | L<AnyEvent::Impl::EventLib>, L<AnyEvent::Impl::Qt>, |
2120 | L<AnyEvent::Impl::EventLib>, L<AnyEvent::Impl::Qt>, |
1980 | L<AnyEvent::Impl::POE>. |
2121 | L<AnyEvent::Impl::POE>, L<AnyEvent::Impl::IOAsync>. |
1981 | |
2122 | |
1982 | Non-blocking file handles, sockets, TCP clients and |
2123 | Non-blocking file handles, sockets, TCP clients and |
1983 | servers: L<AnyEvent::Handle>, L<AnyEvent::Socket>. |
2124 | servers: L<AnyEvent::Handle>, L<AnyEvent::Socket>, L<AnyEvent::TLS>. |
1984 | |
2125 | |
1985 | Asynchronous DNS: L<AnyEvent::DNS>. |
2126 | Asynchronous DNS: L<AnyEvent::DNS>. |
1986 | |
2127 | |
1987 | Coroutine support: L<Coro>, L<Coro::AnyEvent>, L<Coro::EV>, L<Coro::Event>, |
2128 | Coroutine support: L<Coro>, L<Coro::AnyEvent>, L<Coro::EV>, |
|
|
2129 | L<Coro::Event>, |
1988 | |
2130 | |
1989 | Nontrivial usage examples: L<Net::FCP>, L<Net::XMPP2>, L<AnyEvent::DNS>. |
2131 | Nontrivial usage examples: L<AnyEvent::GPSD>, L<AnyEvent::XMPP>, |
|
|
2132 | L<AnyEvent::HTTP>. |
1990 | |
2133 | |
1991 | |
2134 | |
1992 | =head1 AUTHOR |
2135 | =head1 AUTHOR |
1993 | |
2136 | |
1994 | Marc Lehmann <schmorp@schmorp.de> |
2137 | Marc Lehmann <schmorp@schmorp.de> |