Coro/Coro/Channel.pm

=head1 NAME

Coro::Channel - message queues

=head1 SYNOPSIS

 use Coro;

 $q1 = new Coro::Channel <maxsize>;

 $q1->put ("xxx");
 print $q1->get;

 die unless $q1->size;

=head1 DESCRIPTION

A Coro::Channel is the equivalent of a unix pipe (and similar to amiga
message ports): you can put things into it on one end and read things out
of it from the other end. If the capacity of the Channel is maxed out
writers will block. Both ends of a Channel can be read/written from by as
many coroutines as you want concurrently.

You don't have to load C<Coro::Channel> manually, it will be loaded
automatically when you C<use Coro> and call the C<new> constructor.

=over 4

=cut

package Coro::Channel;

use common::sense;

use Coro ();
use Coro::Semaphore ();

our $VERSION = 6.47;

sub DATA (){ 0 }
sub SGET (){ 1 }
sub SPUT (){ 2 }

=item $q = new Coro:Channel $maxsize

Create a new channel with the given maximum size (practically unlimited
if C<maxsize> is omitted or zero). Giving a size of one gives you a
traditional channel, i.e. a queue that can store only a single element
(which means there will be no buffering, and C<put> will wait until there
is a corresponding C<get> call). To buffer one element you have to specify
C<2>, and so on.

=cut

sub new {
   # we cheat and set infinity == 2*10**9
   bless [
      [], # initially empty
      (Coro::Semaphore::_alloc 0), # counts data
      (Coro::Semaphore::_alloc +($_[1] || 2_000_000_000) - 1), # counts remaining space
   ]
}

=item $q->put ($scalar)

Put the given scalar into the queue.

=cut

sub put {
   push @{$_[0][DATA]}, $_[1];
   Coro::Semaphore::up   $_[0][SGET];
   Coro::Semaphore::down $_[0][SPUT];
}

=item $q->get

Return the next element from the queue, waiting if necessary.

=cut

sub get {
   Coro::Semaphore::down $_[0][SGET];
   Coro::Semaphore::up   $_[0][SPUT];
   shift @{$_[0][DATA]}
}

=item $q->shutdown

Shuts down the Channel by pushing a virtual end marker onto it: This
changes the behaviour of the Channel when it becomes or is empty to return
C<undef>, almost as if infinitely many C<undef> elements had been put
into the queue.

Specifically, this function wakes up any pending C<get> calls and lets
them return C<undef>, the same on future C<get> calls. C<size> will return
the real number of stored elements, though.

Another way to describe the behaviour is that C<get> calls will not block
when the queue becomes empty but immediately return C<undef>. This means
that calls to C<put> will work normally and the data will be returned on
subsequent C<get> calls.

This method is useful to signal the end of data to any consumers, quite
similar to an end of stream on e.g. a tcp socket: You have one or more
producers that C<put> data into the Channel and one or more consumers who
C<get> them. When all producers have finished producing data, a call to
C<shutdown> signals this fact to any consumers.

A common implementation uses one or more threads that C<get> from
a channel until it returns C<undef>. To clean everything up, first
C<shutdown> the channel, then C<join> the threads.

=cut

sub shutdown {
   Coro::Semaphore::adjust $_[0][SGET], 1_000_000_000;
}

=item $q->size

Return the number of elements waiting to be consumed. Please note that:

  if ($q->size) {
     my $data = $q->get;
     ...
  }

is I<not> a race condition but instead works just fine. Note that the
number of elements that wait can be larger than C<$maxsize>, as it
includes any coroutines waiting to put data into the channel (but not any
shutdown condition).

This means that the number returned is I<precisely> the number of calls
to C<get> that will succeed instantly and return some data. Calling
C<shutdown> has no effect on this number.

=cut

sub size {
   scalar @{$_[0][DATA]}
}

# this is not undocumented by accident - if it breaks, you
# get to keep the pieces
sub adjust {
   Coro::Semaphore::adjust $_[0][SPUT], $_[1];
}

1;

=back

=head1 AUTHOR/SUPPORT/CONTACT

   Marc A. Lehmann <schmorp@schmorp.de>
   http://software.schmorp.de/pkg/Coro.html

=cut

Revision:	1.136
Committed:	Sat Jul 11 01:59:05 2015 UTC (8 years, 10 months ago) by root
Branch:	MAIN
CVS Tags:	rel-6_47
Changes since 1.135:	+1 -1 lines
Log Message:	6.47
#	User	Rev	Content
1	root	1.1	=head1 NAME
2
3			Coro::Channel - message queues
4
5			=head1 SYNOPSIS
6
7	root	1.101	use Coro;
8	root	1.1
9			$q1 = new Coro::Channel <maxsize>;
10
11	root	1.46	$q1->put ("xxx");
12	root	1.1	print $q1->get;
13
14			die unless $q1->size;
15
16			=head1 DESCRIPTION
17
18	root	1.75	A Coro::Channel is the equivalent of a unix pipe (and similar to amiga
19			message ports): you can put things into it on one end and read things out
20			of it from the other end. If the capacity of the Channel is maxed out
21			writers will block. Both ends of a Channel can be read/written from by as
22			many coroutines as you want concurrently.
23	root	1.8
24	root	1.101	You don't have to load C<Coro::Channel> manually, it will be loaded
25			automatically when you C<use Coro> and call the C<new> constructor.
26
27	root	1.1	=over 4
28
29			=cut
30
31			package Coro::Channel;
32
33	root	1.92	use common::sense;
34	root	1.45
35	root	1.3	use Coro ();
36	root	1.71	use Coro::Semaphore ();
37	root	1.1
38	root	1.136	our $VERSION = 6.47;
39	root	1.1
40	root	1.70	sub DATA (){ 0 }
41	root	1.71	sub SGET (){ 1 }
42			sub SPUT (){ 2 }
43	root	1.70
44	root	1.1	=item $q = new Coro:Channel $maxsize
45
46	root	1.75	Create a new channel with the given maximum size (practically unlimited
47	root	1.128	if C<maxsize> is omitted or zero). Giving a size of one gives you a
48			traditional channel, i.e. a queue that can store only a single element
49			(which means there will be no buffering, and C<put> will wait until there
50			is a corresponding C<get> call). To buffer one element you have to specify
51	root	1.75	C<2>, and so on.
52	root	1.1
53			=cut
54
55			sub new {
56	root	1.84	# we cheat and set infinity == 210*9
57	root	1.71	bless [
58	root	1.96	[], # initially empty
59			(Coro::Semaphore::_alloc 0), # counts data
60			(Coro::Semaphore::_alloc +($_[1] \|\| 2_000_000_000) - 1), # counts remaining space
61	root	1.71	]
62	root	1.1	}
63
64	root	1.44	=item $q->put ($scalar)
65	root	1.1
66			Put the given scalar into the queue.
67
68			=cut
69
70			sub put {
71	root	1.70	push @{$_[0][DATA]}, $_[1];
72	root	1.71	Coro::Semaphore::up $_[0][SGET];
73			Coro::Semaphore::down $_[0][SPUT];
74	root	1.1	}
75
76			=item $q->get
77
78			Return the next element from the queue, waiting if necessary.
79
80			=cut
81
82			sub get {
83	root	1.71	Coro::Semaphore::down $_[0][SGET];
84			Coro::Semaphore::up $_[0][SPUT];
85	root	1.70	shift @{$_[0][DATA]}
86	root	1.47	}
87
88	root	1.77	=item $q->shutdown
89
90			Shuts down the Channel by pushing a virtual end marker onto it: This
91			changes the behaviour of the Channel when it becomes or is empty to return
92	root	1.118	C<undef>, almost as if infinitely many C<undef> elements had been put
93	root	1.77	into the queue.
94
95			Specifically, this function wakes up any pending C<get> calls and lets
96			them return C<undef>, the same on future C<get> calls. C<size> will return
97			the real number of stored elements, though.
98
99			Another way to describe the behaviour is that C<get> calls will not block
100			when the queue becomes empty but immediately return C<undef>. This means
101			that calls to C<put> will work normally and the data will be returned on
102			subsequent C<get> calls.
103
104			This method is useful to signal the end of data to any consumers, quite
105			similar to an end of stream on e.g. a tcp socket: You have one or more
106			producers that C<put> data into the Channel and one or more consumers who
107			C<get> them. When all producers have finished producing data, a call to
108			C<shutdown> signals this fact to any consumers.
109
110	root	1.118	A common implementation uses one or more threads that C<get> from
111			a channel until it returns C<undef>. To clean everything up, first
112			C<shutdown> the channel, then C<join> the threads.
113
114	root	1.77	=cut
115
116			sub shutdown {
117	root	1.103	Coro::Semaphore::adjust $_[0][SGET], 1_000_000_000;
118	root	1.77	}
119
120	root	1.1	=item $q->size
121
122			Return the number of elements waiting to be consumed. Please note that:
123
124			if ($q->size) {
125			my $data = $q->get;
126	root	1.77	...
127	root	1.1	}
128
129	root	1.83	is I<not> a race condition but instead works just fine. Note that the
130			number of elements that wait can be larger than C<$maxsize>, as it
131			includes any coroutines waiting to put data into the channel (but not any
132			shutdown condition).
133
134	root	1.96	This means that the number returned is I<precisely> the number of calls
135			to C<get> that will succeed instantly and return some data. Calling
136			C<shutdown> has no effect on this number.
137	root	1.1
138			=cut
139
140			sub size {
141	root	1.70	scalar @{$_[0][DATA]}
142	root	1.1	}
143
144	root	1.96	# this is not undocumented by accident - if it breaks, you
145			# get to keep the pieces
146			sub adjust {
147			Coro::Semaphore::adjust $_[0][SPUT], $_[1];
148			}
149
150	root	1.1	1;
151
152			=back
153
154	root	1.132	=head1 AUTHOR/SUPPORT/CONTACT
155	root	1.1
156	root	1.132	Marc A. Lehmann <schmorp@schmorp.de>
157			http://software.schmorp.de/pkg/Coro.html
158	root	1.1
159			=cut
160