ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/Coro/Coro.pm
(Generate patch)

Comparing Coro/Coro.pm (file contents):
Revision 1.7 by root, Fri Jul 13 13:05:38 2001 UTC vs.
Revision 1.30 by root, Sat Aug 11 19:59:19 2001 UTC

1=head1 NAME 1=head1 NAME
2 2
3Coro - create and manage simple coroutines 3Coro - coroutine process abstraction
4 4
5=head1 SYNOPSIS 5=head1 SYNOPSIS
6 6
7 use Coro; 7 use Coro;
8 8
9 $new = new Coro sub { 9 async {
10 print "in coroutine, switching back\n"; 10 # some asynchronous thread of execution
11 $new->transfer($main);
12 print "in coroutine again, switching back\n";
13 $new->transfer($main);
14 }; 11 };
15 12
16 $main = new Coro; 13 # alternatively create an async process like this:
17 14
18 print "in main, switching to coroutine\n"; 15 sub some_func : Coro {
19 $main->transfer($new); 16 # some more async code
20 print "back in main, switch to coroutine again\n"; 17 }
21 $main->transfer($new); 18
22 print "back in main\n"; 19 cede;
23 20
24=head1 DESCRIPTION 21=head1 DESCRIPTION
25 22
26This module implements coroutines. Coroutines, similar to continuations, 23This module collection manages coroutines. Coroutines are similar to
27allow you to run more than one "thread of execution" in parallel. Unlike 24Threads but don't run in parallel.
28threads this, only voluntary switching is used so locking problems are
29greatly reduced.
30 25
31Although this is the "main" module of the Coro family it provides only 26This module is still experimental, see the BUGS section below.
32low-level functionality. See L<Coro::Process> and related modules for a 27
33more useful process abstraction including scheduling. 28In this module, coroutines are defined as "callchain + lexical variables
29+ @_ + $_ + $@ + $^W + C stack), that is, a coroutine has it's own
30callchain, it's own set of lexicals and it's own set of perl's most
31important global variables.
32
33=cut
34
35package Coro;
36
37use Coro::State;
38
39use base Exporter;
40
41$VERSION = 0.45;
42
43@EXPORT = qw(async cede schedule terminate current);
44@EXPORT_OK = qw($current);
45
46{
47 my @async;
48 my $init;
49
50 # this way of handling attributes simply is NOT scalable ;()
51 sub import {
52 Coro->export_to_level(1, @_);
53 my $old = *{(caller)[0]."::MODIFY_CODE_ATTRIBUTES"}{CODE};
54 *{(caller)[0]."::MODIFY_CODE_ATTRIBUTES"} = sub {
55 my ($package, $ref) = (shift, shift);
56 my @attrs;
57 for (@_) {
58 if ($_ eq "Coro") {
59 push @async, $ref;
60 unless ($init++) {
61 eval q{
62 sub INIT {
63 &async(pop @async) while @async;
64 }
65 };
66 }
67 } else {
68 push @attrs, $_;
69 }
70 }
71 return $old ? $old->($package, $ref, @attrs) : @attrs;
72 };
73 }
74
75}
76
77=item $main
78
79This coroutine represents the main program.
80
81=cut
82
83our $main = new Coro;
84
85=item $current (or as function: current)
86
87The current coroutine (the last coroutine switched to). The initial value is C<$main> (of course).
88
89=cut
90
91# maybe some other module used Coro::Specific before...
92if ($current) {
93 $main->{specific} = $current->{specific};
94}
95
96our $current = $main;
97
98sub current() { $current }
99
100=item $idle
101
102The coroutine to switch to when no other coroutine is running. The default
103implementation prints "FATAL: deadlock detected" and exits.
104
105=cut
106
107# should be done using priorities :(
108our $idle = new Coro sub {
109 print STDERR "FATAL: deadlock detected\n";
110 exit(51);
111};
112
113# this coroutine is necessary because a coroutine
114# cannot destroy itself.
115my @destroy;
116my $manager = new Coro sub {
117 while() {
118 delete ((pop @destroy)->{_coro_state}) while @destroy;
119 &schedule;
120 }
121};
122
123# static methods. not really.
124
125=head2 STATIC METHODS
126
127Static methods are actually functions that operate on the current process only.
34 128
35=over 4 129=over 4
36 130
37=cut 131=item async { ... } [@args...]
38 132
39package Coro; 133Create a new asynchronous process and return it's process object
134(usually unused). When the sub returns the new process is automatically
135terminated.
40 136
41BEGIN { 137 # create a new coroutine that just prints its arguments
42 $VERSION = 0.03; 138 async {
139 print "@_\n";
140 } 1,2,3,4;
43 141
44 require XSLoader; 142The coderef you submit MUST NOT be a closure that refers to variables
45 XSLoader::load Coro, $VERSION; 143in an outer scope. This does NOT work. Pass arguments into it instead.
46}
47 144
48=item $coro = new [$coderef [, @args]]
49
50Create a new coroutine and return it. The first C<transfer> call to this
51coroutine will start execution at the given coderef. If, the subroutine
52returns it will be executed again.
53
54If the coderef is omitted this function will create a new "empty"
55coroutine, i.e. a coroutine that cannot be transfered to but can be used
56to save the current coroutine in.
57
58=cut 145=cut
146
147sub async(&@) {
148 my $pid = new Coro @_;
149 $manager->ready; # this ensures that the stack is cloned from the manager
150 $pid->ready;
151 $pid;
152}
153
154=item schedule
155
156Calls the scheduler. Please note that the current process will not be put
157into the ready queue, so calling this function usually means you will
158never be called again.
159
160=cut
161
162=item cede
163
164"Cede" to other processes. This function puts the current process into the
165ready queue and calls C<schedule>, which has the effect of giving up the
166current "timeslice" to other coroutines of the same or higher priority.
167
168=cut
169
170=item terminate
171
172Terminates the current process.
173
174Future versions of this function will allow result arguments.
175
176=cut
177
178sub terminate {
179 $current->cancel;
180 &schedule;
181 die; # NORETURN
182}
183
184=back
185
186# dynamic methods
187
188=head2 PROCESS METHODS
189
190These are the methods you can call on process objects.
191
192=over 4
193
194=item new Coro \&sub [, @args...]
195
196Create a new process and return it. When the sub returns the process
197automatically terminates. To start the process you must first put it into
198the ready queue by calling the ready method.
199
200The coderef you submit MUST NOT be a closure that refers to variables
201in an outer scope. This does NOT work. Pass arguments into it instead.
202
203=cut
204
205sub _newcoro {
206 terminate &{+shift};
207}
59 208
60sub new { 209sub new {
61 my $class = $_[0]; 210 my $class = shift;
62 my $proc = $_[1] || sub { die "tried to transfer to an empty coroutine" }; 211 bless {
63 bless _newprocess { 212 _coro_state => (new Coro::State $_[0] && \&_newcoro, @_),
64 do {
65 eval { &$proc };
66 if ($@) {
67 $error_msg = $@;
68 $error_coro = _newprocess { };
69 &transfer($error_coro, $error);
70 }
71 } while (1);
72 }, $class; 213 }, $class;
73} 214}
74 215
75=item $prev->transfer($next) 216=item $process->ready
76 217
77Save the state of the current subroutine in C<$prev> and switch to the 218Put the current process into the ready queue.
78coroutine saved in C<$next>.
79 219
80The "state" of a subroutine only ever includes scope, i.e. lexical
81variables and the current execution state. It does not save/restore any
82global variables such as C<$_> or C<$@> or any other special or non
83special variables. So remember that every function call that might call
84C<transfer> (such as C<Coro::Channel::put>) might clobber any global
85and/or special variables. Yes, this is by design ;) You cna always create
86your own process abstraction model that saves these variables.
87
88The easiest way to do this is to create your own scheduling primitive like this:
89
90 sub schedule {
91 local ($_, $@, ...);
92 $old->transfer($new);
93 }
94
95=cut 220=cut
96 221
97# I call the _transfer function from a perl function 222=item $process->cancel
98# because that way perl saves all important things on
99# the stack. Actually, I'd do it from within XS, but
100# I couldn't get it to work.
101sub transfer {
102 _transfer($_[0], $_[1]);
103}
104 223
105=item $error, $error_msg, $error_coro 224Like C<terminate>, but terminates the specified process instead.
106 225
107This coroutine will be called on fatal errors. C<$error_msg> and
108C<$error_coro> return the error message and the error-causing coroutine
109(NOT an object) respectively. This API might change.
110
111=cut 226=cut
112 227
113$error_msg = 228sub cancel {
114$error_coro = undef; 229 push @destroy, $_[0];
230 $manager->ready;
231}
115 232
116$error = _newprocess { 233=back
117 print STDERR "FATAL: $error_msg\nprogram aborted\n"; 234
118 exit 50; 235=cut
119};
120 236
1211; 2371;
122 238
123=back 239=head1 BUGS/LIMITATIONS
124 240
125=head1 BUGS 241 - could be faster, especially when the core would introduce special
126 242 support for coroutines (like it does for threads).
127This module has not yet been extensively tested. 243 - there is still a memleak on coroutine termination that I could not
244 identify. Could be as small as a single SV.
245 - this module is not well-tested.
246 - if variables or arguments "disappear" (become undef) or become
247 corrupted please contact the author so he cen iron out the
248 remaining bugs.
249 - this module is not thread-safe. You must only ever use this module from
250 the same thread (this requirement might be loosened in the future to
251 allow per-thread schedulers, but Coro::State does not yet allow this).
128 252
129=head1 SEE ALSO 253=head1 SEE ALSO
130 254
131L<Coro::Process>, L<Coro::Signal>. 255L<Coro::Channel>, L<Coro::Cont>, L<Coro::Specific>, L<Coro::Semaphore>,
256L<Coro::Signal>, L<Coro::State>, L<Coro::Event>, L<Coro::RWLock>,
257L<Coro::Handle>, L<Coro::Socket>.
132 258
133=head1 AUTHOR 259=head1 AUTHOR
134 260
135 Marc Lehmann <pcg@goof.com> 261 Marc Lehmann <pcg@goof.com>
136 http://www.goof.com/pcg/marc/ 262 http://www.goof.com/pcg/marc/

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines