[ViewVC] Diff of: cvs/AnyEvent-Fork-RPC/RPC.pm

Comparing AnyEvent-Fork-RPC/RPC.pm (file contents):
Revision 1.9 by root, Wed Apr 17 21:48:35 2013 UTC vs.
Revision 1.19 by root, Thu Apr 18 20:27:02 2013 UTC

…		…
1	=head1 NAME	1	=head1 NAME
2		2
3	AnyEvent::Fork::RPC - simple RPC extension for AnyEvent::Fork	3	AnyEvent::Fork::RPC - simple RPC extension for AnyEvent::Fork
		4
		5	THE API IS NOT FINISHED, CONSIDER THIS A TECHNOLOGY DEMO
4		6
5	=head1 SYNOPSIS	7	=head1 SYNOPSIS
6		8
7	use AnyEvent::Fork::RPC;	9	use AnyEvent::Fork::RPC;
8	# use AnyEvent::Fork is not needed	10	# use AnyEvent::Fork is not needed
12	->require ("MyModule")	14	->require ("MyModule")
13	->AnyEvent::Fork::RPC::run (	15	->AnyEvent::Fork::RPC::run (
14	"MyModule::server",	16	"MyModule::server",
15	);	17	);
16		18
		19	use AnyEvent;
		20
17	my $cv = AE::cv;	21	my $cv = AE::cv;
18		22
19	$rpc->(1, 2, 3, sub {	23	$rpc->(1, 2, 3, sub {
20	print "MyModule::server returned @_\n";	24	print "MyModule::server returned @_\n";
21	$cv->send;	25	$cv->send;
…		…
39	Loading this module also always loads L<AnyEvent::Fork>, so you can make a	43	Loading this module also always loads L<AnyEvent::Fork>, so you can make a
40	separate C<use AnyEvent::Fork> if you wish, but you don't have to.	44	separate C<use AnyEvent::Fork> if you wish, but you don't have to.
41		45
42	=head1 EXAMPLES	46	=head1 EXAMPLES
43		47
44	=head2 Synchronous Backend	48	=head2 Example 1: Synchronous Backend
45		49
46	Here is a simple example that implements a backend that executes C<unlink>	50	Here is a simple example that implements a backend that executes C<unlink>
47	and C<rmdir> calls, and reports their status back. It also reports the	51	and C<rmdir> calls, and reports their status back. It also reports the
48	number of requests it has processed every three requests, which is clearly	52	number of requests it has processed every three requests, which is clearly
49	silly, but illustrates the use of events.	53	silly, but illustrates the use of events.
50		54
51	First the parent process:	55	First the parent process:
52		56
53	use AnyEvent;	57	use AnyEvent;
54	use AnyEvent::Fork;
55	use AnyEvent::Fork::RPC;	58	use AnyEvent::Fork::RPC;
56		59
57	my $done = AE::cv;	60	my $done = AE::cv;
58		61
59	my $rpc = AnyEvent::Fork	62	my $rpc = AnyEvent::Fork
…		…
137		140
138	And as a final remark, there is a fine module on CPAN that can	141	And as a final remark, there is a fine module on CPAN that can
139	asynchronously C<rmdir> and C<unlink> and a lot more, and more efficiently	142	asynchronously C<rmdir> and C<unlink> and a lot more, and more efficiently
140	than this example, namely L<IO::AIO>.	143	than this example, namely L<IO::AIO>.
141		144
		145	=head3 Example 1a: the same with the asynchronous backend
		146
		147	This example only shows what needs to be changed to use the async backend
		148	instead. Doing this is not very useful, the purpose of this example is
		149	to show the minimum amount of change that is required to go from the
		150	synchronous to the asynchronous backend.
		151
		152	To use the async backend in the previous example, you need to add the
		153	C<async> parameter to the C<AnyEvent::Fork::RPC::run> call:
		154
		155	->AnyEvent::Fork::RPC::run ("MyWorker::run",
		156	async => 1,
		157	...
		158
		159	And since the function call protocol is now changed, you need to adopt
		160	C<MyWorker::run> to the async API.
		161
		162	First, you need to accept the extra initial C<$done> callback:
		163
		164	sub run {
		165	my ($done, $cmd, $path) = @_;
		166
		167	And since a response is now generated when C<$done> is called, as opposed
		168	to when the function returns, we need to call the C<$done> function with
		169	the status:
		170
		171	$done->($status or (0, "$!"));
		172
		173	A few remarks are in order. First, it's quite pointless to use the async
		174	backend for this example - but it I<is> possible. Second, you can call
		175	C<$done> before or after returning from the function. Third, having both
		176	returned from the function and having called the C<$done> callback, the
		177	child process may exit at any time, so you should call C<$done> only when
		178	you really I<are> done.
		179
		180	=head2 Example 2: Asynchronous Backend
		181
		182	This example implements multiple count-downs in the child, using
		183	L<AnyEvent> timers. While this is a bit silly (one could use timers in te
		184	parent just as well), it illustrates the ability to use AnyEvent in the
		185	child and the fact that responses can arrive in a different order then the
		186	requests.
		187
		188	It also shows how to embed the actual child code into a C<__DATA__>
		189	section, so it doesn't need any external files at all.
		190
		191	And when your parent process is often busy, and you have stricter timing
		192	requirements, then running timers in a child process suddenly doesn't look
		193	so silly anymore.
		194
		195	Without further ado, here is the code:
		196
		197	use AnyEvent;
		198	use AnyEvent::Fork::RPC;
		199
		200	my $done = AE::cv;
		201
		202	my $rpc = AnyEvent::Fork
		203	->new
		204	->require ("AnyEvent::Fork::RPC::Async")
		205	->eval (do { local $/; <DATA> })
		206	->AnyEvent::Fork::RPC::run ("run",
		207	async => 1,
		208	on_error => sub { warn "FATAL: $_[0]"; exit 1 },
		209	on_event => sub { print $_[0] },
		210	on_destroy => $done,
		211	);
		212
		213	for my $count (3, 2, 1) {
		214	$rpc->($count, sub {
		215	warn "job $count finished\n";
		216	});
		217	}
		218
		219	undef $rpc;
		220
		221	$done->recv;
		222
		223	__DATA__
		224
		225	# this ends up in main, as we don't use a package declaration
		226
		227	use AnyEvent;
		228
		229	sub run {
		230	my ($done, $count) = @_;
		231
		232	my $n;
		233
		234	AnyEvent::Fork::RPC::event "starting to count up to $count\n";
		235
		236	my $w; $w = AE::timer 1, 1, sub {
		237	++$n;
		238
		239	AnyEvent::Fork::RPC::event "count $n of $count\n";
		240
		241	if ($n == $count) {
		242	undef $w;
		243	$done->();
		244	}
		245	};
		246	}
		247
		248	The parent part (the one before the C<__DATA__> section) isn't very
		249	different from the earlier examples. It sets async mode, preloads
		250	the backend module (so the C<AnyEvent::Fork::RPC::event> function is
		251	declared), uses a slightly different C<on_event> handler (which we use
		252	simply for logging purposes) and then, instead of loading a module with
		253	the actual worker code, it C<eval>'s the code from the data section in the
		254	child process.
		255
		256	It then starts three countdowns, from 3 to 1 seconds downwards, destroys
		257	the rpc object so the example finishes eventually, and then just waits for
		258	the stuff to trickle in.
		259
		260	The worker code uses the event function to log some progress messages, but
		261	mostly just creates a recurring one-second timer.
		262
		263	The timer callback increments a counter, logs a message, and eventually,
		264	when the count has been reached, calls the finish callback.
		265
		266	On my system, this results in the following output. Since all timers fire
		267	at roughly the same time, the actual order isn't guaranteed, but the order
		268	shown is very likely what you would get, too.
		269
		270	starting to count up to 3
		271	starting to count up to 2
		272	starting to count up to 1
		273	count 1 of 3
		274	count 1 of 2
		275	count 1 of 1
		276	job 1 finished
		277	count 2 of 2
		278	job 2 finished
		279	count 2 of 3
		280	count 3 of 3
		281	job 3 finished
		282
		283	While the overall ordering isn't guaranteed, the async backend still
		284	guarantees that events and responses are delivered to the parent process
		285	in the exact same ordering as they were generated in the child process.
		286
		287	And unless your system is I<very> busy, it should clearly show that the
		288	job started last will finish first, as it has the lowest count.
		289
		290	This concludes the async example. Since L<AnyEvent::Fork> does not
		291	actually fork, you are free to use about any module in the child, not just
		292	L<AnyEvent>, but also L<IO::AIO>, or L<Tk> for example.
		293
142	=head1 PARENT PROCESS USAGE	294	=head1 PARENT PROCESS USAGE
143		295
144	This module exports nothing, and only implements a single function:	296	This module exports nothing, and only implements a single function:
145		297
146	=over 4	298	=over 4
…		…
224		376
225	The default server used in the child does all I/O blockingly, and only	377	The default server used in the child does all I/O blockingly, and only
226	allows a single RPC call to execute concurrently.	378	allows a single RPC call to execute concurrently.
227		379
228	Setting C<async> to a true value switches to another implementation that	380	Setting C<async> to a true value switches to another implementation that
229	uses L<AnyEvent> in the child and allows multiple concurrent RPC calls.	381	uses L<AnyEvent> in the child and allows multiple concurrent RPC calls (it
		382	does not support recursion in the event loop however, blocking condvar
		383	calls will fail).
230		384
231	The actual API in the child is documented in the section that describes	385	The actual API in the child is documented in the section that describes
232	the calling semantics of the returned C<$rpc> function.	386	the calling semantics of the returned C<$rpc> function.
233		387
234	If you want to pre-load the actual back-end modules to enable memory	388	If you want to pre-load the actual back-end modules to enable memory
…		…
236	synchronous, and C<AnyEvent::Fork::RPC::Async> for asynchronous mode.	390	synchronous, and C<AnyEvent::Fork::RPC::Async> for asynchronous mode.
237		391
238	If you use a template process and want to fork both sync and async	392	If you use a template process and want to fork both sync and async
239	children, then it is permissible to load both modules.	393	children, then it is permissible to load both modules.
240		394
241	=item serialiser => $string (default: '(sub { pack "(w/a)", @_ }, sub { unpack "(w/a)", shift })')	395	=item serialiser => $string (default: $AnyEvent::Fork::RPC::STRING_SERIALISER)
242		396
243	All arguments, result data and event data have to be serialised to be	397	All arguments, result data and event data have to be serialised to be
244	transferred between the processes. For this, they have to be frozen and	398	transferred between the processes. For this, they have to be frozen and
245	thawed in both parent and child processes.	399	thawed in both parent and child processes.
246		400
247	By default, only octet strings can be passed between the processes, which	401	By default, only octet strings can be passed between the processes, which
248	is reasonably fast and efficient.	402	is reasonably fast and efficient and requires no extra modules.
249		403
250	For more complicated use cases, you can provide your own freeze and thaw	404	For more complicated use cases, you can provide your own freeze and thaw
251	functions, by specifying a string with perl source code. It's supposed to	405	functions, by specifying a string with perl source code. It's supposed to
252	return two code references when evaluated: the first receives a list of	406	return two code references when evaluated: the first receives a list of
253	perl values and must return an octet string. The second receives the octet	407	perl values and must return an octet string. The second receives the octet
…		…
255		409
256	If you need an external module for serialisation, then you can either	410	If you need an external module for serialisation, then you can either
257	pre-load it into your L<AnyEvent::Fork> process, or you can add a C<use>	411	pre-load it into your L<AnyEvent::Fork> process, or you can add a C<use>
258	or C<require> statement into the serialiser string. Or both.	412	or C<require> statement into the serialiser string. Or both.
259		413
		414	Here are some examples - some of them are also available as global
		415	variables that make them easier to use.
		416
		417	=over 4
		418
		419	=item octet strings - C<$AnyEvent::Fork::RPC::STRING_SERIALISER>
		420
		421	This serialiser concatenates length-prefixes octet strings, and is the
		422	default.
		423
		424	Implementation:
		425
		426	(
		427	sub { pack "(w/a)", @_ },
		428	sub { unpack "(w/a)", shift }
		429	)
		430
		431	=item json - C<$AnyEvent::Fork::RPC::JSON_SERIALISER>
		432
		433	This serialiser creates JSON arrays - you have to make sure the L<JSON>
		434	module is installed for this serialiser to work. It can be beneficial for
		435	sharing when you preload the L<JSON> module in a template process.
		436
		437	L<JSON> (with L<JSON::XS> installed) is slower than the octet string
		438	serialiser, but usually much faster than L<Storable>, unless big chunks of
		439	binary data need to be transferred.
		440
		441	Implementation:
		442
		443	use JSON ();
		444	(
		445	sub { JSON::encode_json \@_ },
		446	sub { @{ JSON::decode_json shift } }
		447	)
		448
		449	=item storable - C<$AnyEvent::Fork::RPC::STORABLE_SERIALISER>
		450
		451	This serialiser uses L<Storable>, which means it has high chance of
		452	serialising just about anything you throw at it, at the cost of having
		453	very high overhead per operation. It also comes with perl.
		454
		455	Implementation:
		456
		457	use Storable ();
		458	(
		459	sub { Storable::freeze \@_ },
		460	sub { @{ Storable::thaw shift } }
		461	)
		462
		463	=back
		464
260	=back	465	=back
261		466
262	See the examples section earlier in this document for some actual	467	See the examples section earlier in this document for some actual
263	examples.	468	examples.
264		469
265	=cut	470	=cut
266		471
267	our $STRING_SERIALISER = '(sub { pack "(w/a)", @_ }, sub { unpack "(w/a)", shift })';	472	our $STRING_SERIALISER = '(sub { pack "(w/a)", @_ }, sub { unpack "(w/a)", shift })';
		473	our $JSON_SERIALISER = 'use JSON (); (sub { JSON::encode_json \@_ }, sub { @{ JSON::decode_json shift } })';
		474	our $STORABLE_SERIALISER = 'use Storable (); (sub { Storable::freeze \@_ }, sub { @{ Storable::thaw shift } })';
268		475
269	sub run {	476	sub run {
270	my ($self, $function, %arg) = @_;	477	my ($self, $function, %arg) = @_;
271		478
272	my $serialiser = delete $arg{serialiser} \|\| $STRING_SERIALISER;	479	my $serialiser = delete $arg{serialiser} \|\| $STRING_SERIALISER;
…		…
341	}	548	}
342	} elsif (defined $len) {	549	} elsif (defined $len) {
343	undef $rw; undef $ww; # it ends here	550	undef $rw; undef $ww; # it ends here
344		551
345	if (@rcb \|\| %rcb) {	552	if (@rcb \|\| %rcb) {
346	use Data::Dump;ddx[\@rcb,\%rcb];#d#
347	$on_error->("unexpected eof");	553	$on_error->("unexpected eof");
348	} else {	554	} else {
349	$on_destroy->();	555	$on_destroy->();
350	}	556	}
351	} elsif ($! != Errno::EAGAIN && $! != Errno::EWOULDBLOCK) {	557	} elsif ($! != Errno::EAGAIN && $! != Errno::EWOULDBLOCK) {
…		…
357	$ww \|\|= AE::io $fh, 1, $wcb;	563	$ww \|\|= AE::io $fh, 1, $wcb;
358	});	564	});
359		565
360	my $guard = Guard::guard {	566	my $guard = Guard::guard {
361	$shutdown = 1;	567	$shutdown = 1;
362	$ww \|\|= $fh && AE::io $fh, 1, $wcb;	568
		569	shutdown $fh, 1 if $fh && !$ww;
363	};	570	};
364		571
365	my $id;	572	my $id;
366		573
367	$arg{async}	574	$arg{async}
…		…
431	See the examples section earlier in this document for some actual	638	See the examples section earlier in this document for some actual
432	examples.	639	examples.
433		640
434	=back	641	=back
435		642
		643	=head1 ADVANCED TOPICS
		644
		645	=head2 Choosing a backend
		646
		647	So how do you decide which backend to use? Well, that's your problem to
		648	solve, but here are some thoughts on the matter:
		649
		650	=over 4
		651
		652	=item Synchronous
		653
		654	The synchronous backend does not rely on any external modules (well,
		655	except L<common::sense>, which works around a bug in how perl's warning
		656	system works). This keeps the process very small, for example, on my
		657	system, an empty perl interpreter uses 1492kB RSS, which becomes 2020kB
		658	after C<use warnings; use strict> (for people who grew up with C64s around
		659	them this is probably shocking every single time they see it). The worker
		660	process in the first example in this document uses 1792kB.
		661
		662	Since the calls are done synchronously, slow jobs will keep newer jobs
		663	from executing.
		664
		665	The synchronous backend also has no overhead due to running an event loop
		666	- reading requests is therefore very efficient, while writing responses is
		667	less so, as every response results in a write syscall.
		668
		669	If the parent process is busy and a bit slow reading responses, the child
		670	waits instead of processing further requests. This also limits the amount
		671	of memory needed for buffering, as never more than one response has to be
		672	buffered.
		673
		674	The API in the child is simple - you just have to define a function that
		675	does something and returns something.
		676
		677	It's hard to use modules or code that relies on an event loop, as the
		678	child cannot execute anything while it waits for more input.
		679
		680	=item Asynchronous
		681
		682	The asynchronous backend relies on L<AnyEvent>, which tries to be small,
		683	but still comes at a price: On my system, the worker from example 1a uses
		684	3420kB RSS (for L<AnyEvent>, which loads L<EV>, which needs L<XSLoader>
		685	which in turn loads a lot of other modules such as L<warnings>, L<strict>,
		686	L<vars>, L<Exporter>...).
		687
		688	It batches requests and responses reasonably efficiently, doing only as
		689	few reads and writes as needed, but needs to poll for events via the event
		690	loop.
		691
		692	Responses are queued when the parent process is busy. This means the child
		693	can continue to execute any queued requests. It also means that a child
		694	might queue a lot of responses in memory when it generates them and the
		695	parent process is slow accepting them.
		696
		697	The API is not a straightforward RPC pattern - you have to call a
		698	"done" callback to pass return values and signal completion. Also, more
		699	importantly, the API starts jobs as fast as possible - when 1000 jobs
		700	are queued and the jobs are slow, they will all run concurrently. The
		701	child must implement some queueing/limiting mechanism if this causes
		702	problems. Alternatively, the parent could limit the amount of rpc calls
		703	that are outstanding.
		704
		705	Blocking use of condvars is not supported.
		706
		707	Using event-based modules such as L<IO::AIO>, L<Gtk2>, L<Tk> and so on is
		708	easy.
		709
		710	=back
		711
		712	=head2 Passing file descriptors
		713
		714	Unlike L<AnyEvent::Fork>, this module has no in-built file handle or file
		715	descriptor passing abilities.
		716
		717	The reason is that passing file descriptors is extraordinary tricky
		718	business, and conflicts with efficient batching of messages.
		719
		720	There still is a method you can use: Create a
		721	C<AnyEvent::Util::portable_socketpair> and C<send_fh> one half of it to
		722	the process before you pass control to C<AnyEvent::Fork::RPC::run>.
		723
		724	Whenever you want to pass a file descriptor, send an rpc request to the
		725	child process (so it expects the descriptor), then send it over the other
		726	half of the socketpair. The child should fetch the descriptor from the
		727	half it has passed earlier.
		728
		729	Here is some (untested) pseudocode to that effect:
		730
		731	use AnyEvent::Util;
		732	use AnyEvent::Fork::RPC;
		733	use IO::FDPass;
		734
		735	my ($s1, $s2) = AnyEvent::Util::portable_socketpair;
		736
		737	my $rpc = AnyEvent::Fork
		738	->new
		739	->send_fh ($s2)
		740	->require ("MyWorker")
		741	->AnyEvent::Fork::RPC::run ("MyWorker::run"
		742	init => "MyWorker::init",
		743	);
		744
		745	undef $s2; # no need to keep it around
		746
		747	# pass an fd
		748	$rpc->("i'll send some fd now, please expect it!", my $cv = AE::cv);
		749
		750	IO::FDPass fileno $s1, fileno $handle_to_pass;
		751
		752	$cv->recv;
		753
		754	The MyWorker module could look like this:
		755
		756	package MyWorker;
		757
		758	use IO::FDPass;
		759
		760	my $s2;
		761
		762	sub init {
		763	$s2 = $_[0];
		764	}
		765
		766	sub run {
		767	if ($_[0] eq "i'll send some fd now, please expect it!") {
		768	my $fd = IO::FDPass::recv fileno $s2;
		769	...
		770	}
		771	}
		772
		773	Of course, this might be blocking if you pass a lot of file descriptors,
		774	so you might want to look into L<AnyEvent::FDpasser> which can handle the
		775	gory details.
		776
436	=head1 SEE ALSO	777	=head1 SEE ALSO
437		778
438	L<AnyEvent::Fork> (to create the processes in the first place),	779	L<AnyEvent::Fork>, to create the processes in the first place.
		780
439	L<AnyEvent::Fork::Pool> (to manage whole pools of processes).	781	L<AnyEvent::Fork::Pool>, to manage whole pools of processes.
440		782
441	=head1 AUTHOR AND CONTACT INFORMATION	783	=head1 AUTHOR AND CONTACT INFORMATION
442		784
443	Marc Lehmann <schmorp@schmorp.de>	785	Marc Lehmann <schmorp@schmorp.de>
444	http://software.schmorp.de/pkg/AnyEvent-Fork-RPC	786	http://software.schmorp.de/pkg/AnyEvent-Fork-RPC

Diff Legend

-–
+Removed lines
-+
+Added lines
-<
+Changed lines
->
+Changed lines

Comparing AnyEvent-Fork-RPC/RPC.pm (file contents): Revision 1.9 by root, Wed Apr 17 21:48:35 2013 UTC vs. Revision 1.19 by root, Thu Apr 18 20:27:02 2013 UTC

Diff Legend

Comparing AnyEvent-Fork-RPC/RPC.pm (file contents):
Revision 1.9 by root, Wed Apr 17 21:48:35 2013 UTC vs.
Revision 1.19 by root, Thu Apr 18 20:27:02 2013 UTC