Object::ForkAware(3pm) | User Contributed Perl Documentation | Object::ForkAware(3pm) |
Object::ForkAware - Make an object aware of process forks and threads, recreating itself as needed
version 0.005
use Object::ForkAware; my $client = Object::ForkAware->new( create => sub { MyClient->new(server => 'foo.com', port => '1234') }, ); # do things with object as normal... $client->send(...); # later, we fork for some reason if (fork == 0) { # child process $client->send(...); } # no boom happens! fork is detected and client object is regenerated
If you've ever had an object representing a network connection to some server, or something else containing a socket, a filehandle, etc, and used it in a program that forks, and then forgot to close and reopen your socket/handle etc in the new process, you'll know what chaos can ensue. Depending on the type of connection, you can have multiple processes trying to write to the same resource at once, or simultaneous reads getting each other's data, dogs and cats living together... It's horrible, and it's an easy problem to run into.
This module invisibly wraps your object and makes it fork-aware, automatically checking $$ on every access and recreating the object if the process id changes. (The object is also thread-aware; if the thread id changes, the object is recreated in the same manner.)
The object can be safely used with type checks and various type constraint mechanisms, as "isa" and "can" respond as if they were being called against the contained object itself.
You can also ensure that a fork never happens, by making use of the optional "on_fork" handler:
my $client = Object::ForkAware->new( create => sub { MyClient->new(server => 'foo.com', port => '1234') }, on_fork => sub { die 'fork detected!' }, );
Or, if regenerating the object needs to be done differently than the initial creation:
my $client = Object::ForkAware->new( create => sub { MyClient->new(server => 'foo.com', port => '1234') }, on_fork => sub { MyClient->new(server => 'other.foo.com' }, );
Provides an instance of this class. Available options are:
There are no other public methods. All method calls on the object will be passed through to the containing object, after checking $$ (or "threads->tid") and possibly recreating the object via the provided "create" (or "on_fork") sub.
Using the Object::ForkAware object with an operator that the containing object has overloaded will not work; behaviour is as if there was no operator overloading. Partial support is possible, but is not yet implemented.
The concept for this module came about through a conversation with Matt S. Trout ("mst@shadowcat.co.uk") after experiencing the issue described in the synopsis on a prefork job-processing daemon.
Some of the pid detection logic was inspired by the wonderful DBIx::Connector.
Bugs may be submitted through the RT bug tracker <https://rt.cpan.org/Public/Dist/Display.html?Name=Object-ForkAware> (or bug-Object-ForkAware@rt.cpan.org <mailto:bug-Object-ForkAware@rt.cpan.org>).
I am also usually active on irc, as 'ether' at "irc.perl.org".
Karen Etheridge <ether@cpan.org>
Graham Knop <haarg@haarg.org>
This software is copyright (c) 2013 by Karen Etheridge.
This is free software; you can redistribute it and/or modify it under the same terms as the Perl 5 programming language system itself.
2021-01-09 | perl v5.32.0 |