view release on metacpan or  search on metacpan

lib/AnyEvent/Task.pm  view on Meta::CPAN

      my ($dbh, $user) = @_;
      print "username: $user->{username}, email: $user->{email}\n";
      $cv->send;
    });

    $cv->recv;

=head2 Output

    username: jimmy, email: jimmy@example.com



=head1 DESCRIPTION

The synopses make this module look much more complicated than it actually is. In a nutshell, a synchronous worker process is forked off by a server whenever a client asks for one. The client keeps as many of these workers around as it wants and deleg...

Another way of saying that is that L<AnyEvent::Task> is a pre-fork-on-demand server (L<AnyEvent::Task::Server>) combined with a persistent worker-pooled client (L<AnyEvent::Task::Client>).

The examples in the synopses are complete stand-alone programs. Run the server in one window and the client in another. The server will remain running but the client will exit after printing its output. Typically the "client" programs would be embedd...

Note that the client examples don't implement error checking (see the L<ERROR HANDLING> section).

A server is started with C<< AnyEvent::Task::Server->new >>. This constructor should be passed in at least the C<listen> and C<interface> arguments. Keep the returned server object around for as long as you want the server to be running. C<listen> is...

A client is started with C<< AnyEvent::Task::Client->new >>. You only need to pass C<connect> to this constructor which is an array ref containing the host and service options to be passed to L<AnyEvent::Socket>'s C<tcp_connect>. Keep the returned cl...

After the server and client are initialised, each process must enter AnyEvent's "main loop" in some way, possibly just C<< AE::cv->recv >>. The C<run> method on the server object is a convenient short-cut for this.

To acquire a worker process you call the C<checkout> method on the client object. The C<checkout> method doesn't need any arguments, but several optional ones such as C<timeout> are described below. As long as the checkout object is around, this chec...

The checkout object is an object that proxies its method calls to a worker process or a function that does the same. The arguments to this method/function are the arguments you wish to send to the worker process followed by a callback to run when the...

In the event of an exception thrown by the worker process, a timeout, or some other unexpected condition, an error is raised in the dynamic context of the callback (see the L<ERROR HANDLING> section).




=head1 DESIGN

Both client and server are of course built with L<AnyEvent>. However, workers can't use AnyEvent (yet). I've never found a need to do event processing in the worker since if the library you wish to use is already AnyEvent-compatible you can simply us...

Each client maintains a "pool" of connections to worker processes. Every time a checkout is requested, the request is placed into a first-come, first-serve queue. Once a worker process becomes available, it is associated with that checkout until that...

C<timeout> can be passed as a keyword argument to C<checkout>. Once a request is queued up on that checkout, a timer of C<timout> seconds (default is 30, undef means infinity) is started. If the request completes during this timeframe, the timer is c...

Note that since timeouts are associated with a checkout, checkouts can be created before the server is started. As long as the server is running within C<timeout> seconds, no error will be thrown and no requests will be lost. The client will continua...

Because of checkout queuing, the maximum number of worker processes a client will attempt to obtain can be limited with the C<max_workers> argument when creating a client object. If there are more live checkouts than C<max_workers>, the remaining che...

The C<min_workers> argument determines how many "hot-standby" workers should be pre-forked when creating the client. The default is 2 though note that this may change to 0 in the future.





=head1 STARTING THE SERVER

Typically you will want to start the client and server as completely separate processes as shown in the synopses.

Running the server and the client in the same process is technically possible but is highly discouraged since the server will C<fork()> when the client demands a new worker process. In this case, all descriptors in use by the client are duped into th...

Since it's more of a bother than it's worth to run the server and the client in the same process, there is an alternate server constructor, C<AnyEvent::Task::Server::fork_task_server> for when you'd like to fork a dedicated server process. It can be ...

    ## my ($keepalive_pipe, $server_pid) =
    AnyEvent::Task::Server::fork_task_server(
      name => 'hello',
      listen => ['unix/', '/tmp/anyevent-task.socket'],
      interface => sub {
                         return "Hello from PID $$";
                       },
    );

The only differences between this and the regular constructor is that C<fork_task_server> will fork a process which becomes the server and will also install a "keep-alive" pipe between the server and the client. This keep-alive pipe will be used by t...

If C<AnyEvent::Task::Server::fork_task_server> is called in a void context then the reference to this keep-alive pipe is pushed onto C<@AnyEvent::Task::Server::children_sockets>. Otherwise, the keep-alive pipe and the server's PID are returned. Closi...

Since the C<fork_task_server> constructor calls fork and requires using AnyEvent in both the parent and child processes, it is important that you not install any AnyEvent watchers before calling it. The usual caveats about forking AnyEvent processes ...

You should also not call C<fork_task_server> after having started threads since, again, this function calls fork. Forking a threaded process is dangerous because the threads might have userspace data-structures in inconsistent states at the time of t...





=head1 INTERFACE

When creating a server, there are two possible formats for the C<interface> option. The first and most general is a coderef. This coderef will be passed the list of arguments that were sent when the checkout was called in the client process (without ...

As described above, you can use a checkout object as a coderef or as an object with methods. If the checkout is invoked as an object, the method name is prepended to the arguments passed to C<interface>:

    interface => sub {
      my ($method, @args) = @_;
    },

If the checkout is invoked as a coderef, method is omitted:

    interface => sub {
      my (@args) = @_;
    },

The second format possible for C<interface> is a hash ref. This is a simple method dispatch feature where the method invoked on the checkout object is the key used to lookup which coderef to run in the worker:

    interface => {
      method1 => sub {
        my (@args) = @_;
      },
      method2 => sub {
        my (@args) = @_;
      },
    },

Note that since the protocol between the client and the worker process is currently JSON-based, all arguments and return values must be serializable to JSON. This includes most perl scalars like strings, a limited range of numerical types, and hash/l...

Because there isn't any way for the callback to indicate the context it desires, interface subs are always called in scalar context.

A future backwards compatible RPC protocol may use L<Sereal>. Although it's inefficient you can already serialise an object with Sereal manually, send the resulting string over the existing protocol, and then deserialise it in the worker.