| 1 |
=head1 NAME |
| 2 |
|
| 3 |
Coro::State - first class continuations |
| 4 |
|
| 5 |
=head1 SYNOPSIS |
| 6 |
|
| 7 |
use Coro::State; |
| 8 |
|
| 9 |
$new = new Coro::State sub { |
| 10 |
print "in coroutine (called with @_), switching back\n"; |
| 11 |
$new->transfer ($main); |
| 12 |
print "in coroutine again, switching back\n"; |
| 13 |
$new->transfer ($main); |
| 14 |
}, 5; |
| 15 |
|
| 16 |
$main = new Coro::State; |
| 17 |
|
| 18 |
print "in main, switching to coroutine\n"; |
| 19 |
$main->transfer ($new); |
| 20 |
print "back in main, switch to coroutine again\n"; |
| 21 |
$main->transfer ($new); |
| 22 |
print "back in main\n"; |
| 23 |
|
| 24 |
=head1 DESCRIPTION |
| 25 |
|
| 26 |
This module implements coroutines. Coroutines, similar to continuations, |
| 27 |
allow you to run more than one "thread of execution" in parallel. Unlike |
| 28 |
threads, there is no parallelism and only voluntary switching is used so |
| 29 |
locking problems are greatly reduced. |
| 30 |
|
| 31 |
This can be used to implement non-local jumps, exception handling, |
| 32 |
(non-clonable) continuations and more. |
| 33 |
|
| 34 |
This module provides only low-level functionality. See L<Coro> and related |
| 35 |
modules for a higher level process abstraction including scheduling. |
| 36 |
|
| 37 |
=head2 MODEL |
| 38 |
|
| 39 |
Coro::State implements two different coroutine models: Perl and C. The |
| 40 |
C coroutines (called cctx's) are basically simplified perl interpreters |
| 41 |
running/interpreting the Perl coroutines. A single interpreter can run any |
| 42 |
number of Perl coroutines, so usually there are very few C coroutines. |
| 43 |
|
| 44 |
When Perl code calls a C function (e.g. in an extension module) and that |
| 45 |
C function then calls back into Perl or does a coroutine switch the C |
| 46 |
coroutine can no longer execute other Perl coroutines, so it stays tied to |
| 47 |
the specific coroutine until it returns to the original Perl caller, after |
| 48 |
which it is again available to run other Perl coroutines. |
| 49 |
|
| 50 |
The main program always has its own "C coroutine" (which really is |
| 51 |
*the* Perl interpreter running the whole program), so there will always |
| 52 |
be at least one additional C coroutine. You can use the debugger (see |
| 53 |
L<Coro::Debug>) to find out which coroutines are tied to their cctx and |
| 54 |
which aren't. |
| 55 |
|
| 56 |
=head2 MEMORY CONSUMPTION |
| 57 |
|
| 58 |
A newly created coroutine that has not been used only allocates a |
| 59 |
relatively small (a hundred bytes) structure. Only on the first |
| 60 |
C<transfer> will perl allocate stacks (a few kb, 64 bit architetcures |
| 61 |
use twice as much, i.e. a few kb :) and optionally a C stack/coroutine |
| 62 |
(cctx) for coroutines that recurse through C functions. All this is very |
| 63 |
system-dependent. On my x86-pc-linux-gnu system this amounts to about 2k |
| 64 |
per (non-trivial but simple) coroutine. |
| 65 |
|
| 66 |
You can view the actual memory consumption using Coro::Debug. Keep in mind |
| 67 |
that a for loop or other block constructs can easily consume 100-200 bytes |
| 68 |
per nesting level. |
| 69 |
|
| 70 |
=cut |
| 71 |
|
| 72 |
package Coro::State; |
| 73 |
|
| 74 |
use strict; |
| 75 |
no warnings "uninitialized"; |
| 76 |
|
| 77 |
use Carp; |
| 78 |
|
| 79 |
use Time::HiRes (); # currently only used for PerlIO::cede |
| 80 |
|
| 81 |
our $DIEHOOK; |
| 82 |
our $WARNHOOK; |
| 83 |
|
| 84 |
BEGIN { |
| 85 |
$DIEHOOK = sub { }; |
| 86 |
$WARNHOOK = sub { warn $_[0] }; |
| 87 |
} |
| 88 |
|
| 89 |
sub diehook { &$DIEHOOK } |
| 90 |
sub warnhook { &$WARNHOOK } |
| 91 |
|
| 92 |
use XSLoader; |
| 93 |
|
| 94 |
BEGIN { |
| 95 |
our $VERSION = "5.0"; |
| 96 |
|
| 97 |
# must be done here because the xs part expects it to exist |
| 98 |
# it might exist already because Coro::Specific created it. |
| 99 |
$Coro::current ||= { }; |
| 100 |
|
| 101 |
{ |
| 102 |
# save/restore the handlers before/after overwriting %SIG magic |
| 103 |
local $SIG{__DIE__}; |
| 104 |
local $SIG{__WARN__}; |
| 105 |
|
| 106 |
XSLoader::load __PACKAGE__, $VERSION; |
| 107 |
} |
| 108 |
|
| 109 |
# need to do it after overwriting the %SIG magic |
| 110 |
$SIG{__DIE__} ||= \&diehook; |
| 111 |
$SIG{__WARN__} ||= \&warnhook; |
| 112 |
} |
| 113 |
|
| 114 |
use Exporter; |
| 115 |
use base Exporter::; |
| 116 |
|
| 117 |
=head2 GLOBAL VARIABLES |
| 118 |
|
| 119 |
=over 4 |
| 120 |
|
| 121 |
=item $Coro::State::DIEHOOK |
| 122 |
|
| 123 |
This works similarly to C<$SIG{__DIE__}> and is used as the default die |
| 124 |
hook for newly created coroutines. This is useful if you want some generic |
| 125 |
logging function that works for all coroutines that don't set their own |
| 126 |
hook. |
| 127 |
|
| 128 |
When Coro::State is first loaded it will install these handlers for the |
| 129 |
main program, too, unless they have been overwritten already. |
| 130 |
|
| 131 |
The default handlers provided will behave like the built-in ones (as if |
| 132 |
they weren't there). |
| 133 |
|
| 134 |
If you don't want to exit your program on uncaught exceptions, you can |
| 135 |
must not return from your die hook - terminate instead. |
| 136 |
|
| 137 |
Note 1: You I<must> store a valid code reference in these variables, |
| 138 |
C<undef> will I<not> do. |
| 139 |
|
| 140 |
Note 2: The value of this variable will be shared among all coroutines, so |
| 141 |
changing its value will change it in all coroutines that don't have their |
| 142 |
own die handler. |
| 143 |
|
| 144 |
=item $Coro::State::WARNHOOK |
| 145 |
|
| 146 |
Similar to above die hook, but augments C<$SIG{__WARN__}>. |
| 147 |
|
| 148 |
=back |
| 149 |
|
| 150 |
=head2 FUNCTIONS |
| 151 |
|
| 152 |
=over 4 |
| 153 |
|
| 154 |
=item $coro = new Coro::State [$coderef[, @args...]] |
| 155 |
|
| 156 |
Create a new coroutine and return it. The first C<transfer> call to this |
| 157 |
coroutine will start execution at the given coderef, with the given |
| 158 |
arguments. |
| 159 |
|
| 160 |
Note that the arguments will not be copied. Instead, as with normal |
| 161 |
function calls, the coroutine receives passed arguments by reference, so |
| 162 |
make sure you don't change them in unexpected ways. |
| 163 |
|
| 164 |
Returning from such a coroutine is I<NOT> supported. Neither is calling |
| 165 |
C<exit> or throwing an uncaught exception. The following paragraphs |
| 166 |
describe what happens in current versions of Coro. |
| 167 |
|
| 168 |
If the subroutine returns the program will be terminated as if execution |
| 169 |
of the main program ended. |
| 170 |
|
| 171 |
If it throws an exception the program will terminate unless the exception |
| 172 |
is caught, exactly like in the main program. |
| 173 |
|
| 174 |
Calling C<exit> in a coroutine does the same as calling it in the main |
| 175 |
program. |
| 176 |
|
| 177 |
If the coderef is omitted this function will create a new "empty" |
| 178 |
coroutine, i.e. a coroutine that cannot be transfered to but can be used |
| 179 |
to save the current coroutine state in (note that this is dangerous, as no |
| 180 |
reference is taken to ensure that the "current coroutine state" survives, |
| 181 |
the caller is responsible to ensure that the cloned state does not go |
| 182 |
away). |
| 183 |
|
| 184 |
The returned object is an empty hash which can be used for any purpose |
| 185 |
whatsoever, for example when subclassing Coro::State. |
| 186 |
|
| 187 |
Certain variables are "localised" to each coroutine, that is, certain |
| 188 |
"global" variables are actually per coroutine. Not everything that would |
| 189 |
sensibly be localised currently is, and not everything that is localised |
| 190 |
makes sense for every application, and the future might bring changes. |
| 191 |
|
| 192 |
The following global variables can have different values per coroutine, |
| 193 |
and have the stated initial values: |
| 194 |
|
| 195 |
Variable Initial Value |
| 196 |
@_ whatever arguments were passed to the Coro |
| 197 |
$_ undef |
| 198 |
$@ undef |
| 199 |
$/ "\n" |
| 200 |
$SIG{__DIE__} aliased to $Coro::State::DIEHOOK(*) |
| 201 |
$SIG{__WARN__} aliased to $Coro::State::WARNHOOK(*) |
| 202 |
(default fh) *STDOUT |
| 203 |
$1, $2... all regex results are initially undefined |
| 204 |
|
| 205 |
(*) reading the value from %SIG is not supported, but local'ising is. |
| 206 |
|
| 207 |
If you feel that something important is missing then tell me. Also |
| 208 |
remember that every function call that might call C<transfer> (such |
| 209 |
as C<Coro::Channel::put>) might clobber any global and/or special |
| 210 |
variables. Yes, this is by design ;) You can always create your own |
| 211 |
process abstraction model that saves these variables. |
| 212 |
|
| 213 |
The easiest way to do this is to create your own scheduling primitive like |
| 214 |
in the code below, and use it in your coroutines: |
| 215 |
|
| 216 |
sub my_cede { |
| 217 |
local ($;, ...); |
| 218 |
Coro::cede; |
| 219 |
} |
| 220 |
|
| 221 |
=cut |
| 222 |
|
| 223 |
=item $state->throw ([$scalar]) |
| 224 |
|
| 225 |
See L<< Coro->throw >>. |
| 226 |
|
| 227 |
=item $state->call ($coderef) |
| 228 |
|
| 229 |
Try to call the given C<$coderef> in the context of the given state. This |
| 230 |
works even when the state is currently within an XS function, and can |
| 231 |
be very dangerous. You can use it to acquire stack traces etc. (see the |
| 232 |
Coro::Debug module for more details). The coderef MUST NOT EVER transfer |
| 233 |
to another state. |
| 234 |
|
| 235 |
=item $state->eval ($string) |
| 236 |
|
| 237 |
Like C<call>, but eval's the string. Dangerous. |
| 238 |
|
| 239 |
=item $state->swap_defsv |
| 240 |
|
| 241 |
=item $state->swap_defav |
| 242 |
|
| 243 |
Swap the current C<$_> (swap_defsv) or C<@_> (swap_defav) with the |
| 244 |
equivalent in the saved state of C<$state>. This can be used to give the |
| 245 |
coroutine a defined content for C<@_> and C<$_> before transfer'ing to it. |
| 246 |
|
| 247 |
=item $state->trace ($flags) |
| 248 |
|
| 249 |
Internal function to control tracing. I just mention this so you can stay |
| 250 |
away from abusing it. |
| 251 |
|
| 252 |
=item $prev->transfer ($next) |
| 253 |
|
| 254 |
Save the state of the current subroutine in C<$prev> and switch to the |
| 255 |
coroutine saved in C<$next>. |
| 256 |
|
| 257 |
The "state" of a subroutine includes the scope, i.e. lexical variables and |
| 258 |
the current execution state (subroutine, stack). |
| 259 |
|
| 260 |
=item $bytes = $state->rss |
| 261 |
|
| 262 |
Returns the memory allocated by the coroutine (which includes static |
| 263 |
structures, various perl stacks but NOT local variables, arguments or any |
| 264 |
C context data). This is a rough indication of how much memory it might |
| 265 |
use. |
| 266 |
|
| 267 |
=item $state->has_cctx |
| 268 |
|
| 269 |
Returns whether the state currently uses a cctx/C context. An active |
| 270 |
state always has a cctx, as well as the main program. Other states only |
| 271 |
use a cctxts when needed. |
| 272 |
|
| 273 |
=item Coro::State::force_cctx |
| 274 |
|
| 275 |
Forces the allocation of a C context for the currently running coroutine |
| 276 |
(if not already done). Apart from benchmarking there is little point |
| 277 |
in doing so, however. |
| 278 |
|
| 279 |
=item $ncctx = Coro::State::cctx_count |
| 280 |
|
| 281 |
Returns the number of C-level coroutines allocated. If this number is |
| 282 |
very high (more than a dozen) it might help to identify points of C-level |
| 283 |
recursion in your code and moving this into a separate coroutine. |
| 284 |
|
| 285 |
=item $nidle = Coro::State::cctx_idle |
| 286 |
|
| 287 |
Returns the number of allocated but idle (free for reuse) C level |
| 288 |
coroutines. Currently, Coro will limit the number of idle/unused cctxs to |
| 289 |
8. |
| 290 |
|
| 291 |
=item $old = Coro::State::cctx_stacksize [$new_stacksize] |
| 292 |
|
| 293 |
Returns the current C stack size and optionally sets the new I<minimum> |
| 294 |
stack size to C<$new_stacksize> I<long>s. Existing stacks will not |
| 295 |
be changed, but Coro will try to replace smaller stacks as soon as |
| 296 |
possible. Any Coro::State that starts to use a stack after this call is |
| 297 |
guaranteed this minimum stack size. |
| 298 |
|
| 299 |
Please note that Coroutines will only need to use a C-level stack if the |
| 300 |
interpreter recurses or calls a function in a module that calls back into |
| 301 |
the interpreter, so use of this feature is usually never needed. |
| 302 |
|
| 303 |
=item $old = Coro::State::cctx_max_idle [$new_count] |
| 304 |
|
| 305 |
Coro caches C contexts that are not in use currently, as creating them |
| 306 |
from scratch has some overhead. |
| 307 |
|
| 308 |
This function returns the current maximum number of idle C contexts and |
| 309 |
optionally sets the new amount. The count must be at least C<1>, with the |
| 310 |
default being C<4>. |
| 311 |
|
| 312 |
=item @states = Coro::State::list |
| 313 |
|
| 314 |
Returns a list of all states currently allocated. |
| 315 |
|
| 316 |
=item $clone = $state->clone |
| 317 |
|
| 318 |
This exciting method takes a Coro::State object and clones, i.e., creates |
| 319 |
a copy. This makes it possible to restore a state more than once. |
| 320 |
|
| 321 |
Since its only known purpose is intellectual self-gratification, and |
| 322 |
because it is a difficult piece of code, it is not enabled by default, and |
| 323 |
not supported. |
| 324 |
|
| 325 |
Among the games you can play with this is implement a scheme-like call/cc, |
| 326 |
as the following code does (well, with small differences). |
| 327 |
|
| 328 |
sub callcc(&@) { |
| 329 |
my ($func, @arg) = @_; |
| 330 |
|
| 331 |
my $state = new Coro::State; |
| 332 |
my $wrapper; $wrapper = new Coro::State sub { |
| 333 |
my $escape = sub { |
| 334 |
@arg = @_; |
| 335 |
$wrapper->transfer ($state->clone); |
| 336 |
}; |
| 337 |
$escape->($func->($escape, @arg)); |
| 338 |
}; |
| 339 |
|
| 340 |
$state->transfer ($wrapper); |
| 341 |
@arg |
| 342 |
} |
| 343 |
|
| 344 |
Here are a few little-known facts: First, coroutines *are* full/true/real |
| 345 |
continuations. Secondly Coro::State objects (without clone) *are* first |
| 346 |
class continuations. Thirdly, nobody has ever found a use for the full |
| 347 |
power of call/cc that isn't better (faster, easier, more efficient) |
| 348 |
implemented differently, and nobody has yet found a useful control |
| 349 |
construct that can't be implemented without it already, just much faster |
| 350 |
and with fewer resources. |
| 351 |
|
| 352 |
Besides, call/cc is much less useful in a Perl-like dynamic language (with |
| 353 |
references, and its scoping rules) then in, say, scheme. |
| 354 |
|
| 355 |
Now, the known limitations of C<clone>: |
| 356 |
|
| 357 |
It probably only works on perl 5.10; it cannot clone a coroutine inside |
| 358 |
the substition operator (but windows perl can't fork from there either) |
| 359 |
and some other contexts, and C<abort ()> is the preferred mechanism to |
| 360 |
signal errors. It cannot clone a state that has a c context attached |
| 361 |
(implementing clone on the C level is too hard for me to even try), |
| 362 |
which rules out calling call/cc from the main coroutine. It cannot |
| 363 |
clone a context that hasn't even been started yet. It doesn't work with |
| 364 |
C<-DDEBUGGING> (but what does). It probably also leaks, and sometimes |
| 365 |
triggers a few assertions inside Coro. Most of these limitations *are* |
| 366 |
fixable with some effort, but that's pointless just to make a point that |
| 367 |
it could be done. |
| 368 |
|
| 369 |
=cut |
| 370 |
|
| 371 |
# used by Coro::Debug only atm. |
| 372 |
sub debug_desc { |
| 373 |
$_[0]{desc} |
| 374 |
} |
| 375 |
|
| 376 |
# for very deep reasons, we must initialise $Coro::main here. |
| 377 |
|
| 378 |
{ |
| 379 |
package Coro; |
| 380 |
|
| 381 |
our $main; # main coroutine |
| 382 |
our $current; # current coroutine |
| 383 |
|
| 384 |
$main = Coro::State::new Coro::; |
| 385 |
|
| 386 |
$main->{desc} = "[main::]"; |
| 387 |
|
| 388 |
# maybe some other module used Coro::Specific before... |
| 389 |
$main->{_specific} = $current->{_specific} |
| 390 |
if $current; |
| 391 |
|
| 392 |
_set_current $main; |
| 393 |
} |
| 394 |
|
| 395 |
1; |
| 396 |
|
| 397 |
=back |
| 398 |
|
| 399 |
=head1 BUGS |
| 400 |
|
| 401 |
This module is not thread-safe. You must only ever use this module from |
| 402 |
the same thread (this requirement might be removed in the future). |
| 403 |
|
| 404 |
=head1 SEE ALSO |
| 405 |
|
| 406 |
L<Coro>. |
| 407 |
|
| 408 |
=head1 AUTHOR |
| 409 |
|
| 410 |
Marc Lehmann <schmorp@schmorp.de> |
| 411 |
http://home.schmorp.de/ |
| 412 |
|
| 413 |
=cut |
| 414 |
|
