| Event::RPC(3pm) | User Contributed Perl Documentation | Event::RPC(3pm) |
Event::RPC - Event based transparent Client/Server RPC framework
#-- Server Code
use Event::RPC::Server;
use My::TestModule;
my $server = Event::RPC::Server->new (
port => 5555,
classes => { "My::TestModule" => { ... } },
);
$server->start;
----------------------------------------------------------
#-- Client Code
use Event::RPC::Client;
my $client = Event::RPC::Client->new (
server => "localhost",
port => 5555,
);
$client->connect;
#-- Call methods of My::TestModule on the server
my $obj = My::TestModule->new ( foo => "bar" );
my $foo = $obj->get_foo;
Event::RPC supports you in developing Event based networking client/server applications with transparent object/method access from the client to the server. Network communication is optionally encrypted using IO::Socket::SSL. Several event loop managers are supported due to an extensible API. Currently Event, Glib and AnyEvent are implemented. The latter lets you use nearly every event loop implementation available for Perl. AnyEvent was invented after Event::RPC was created and thus Event::RPC started using it's own abstraction model.
Event::RPC consists of a server and a client library. The server exports a list of classes and methods, which are allowed to be called over the network. More specific it acts as a proxy for objects created on the server side (on demand of the connected clients) which handles client side methods calls with transport of method arguments and return values.
The object proxy handles refcounting and destruction of objects created by clients properly. Objects as method parameters and return values are handled as well (although with some limitations, see below).
For the client the whole thing is totally transparent - once connected to the server it doesn't know whether it calls methods on local or remote objects.
Also the methods on the server newer know whether they are called locally or from a connected client. Your application logic is not affected by Event::RPC at all, at least if it has a rudimentary clean OO design.
For details on implementing servers and clients please refer to the man pages of Event::RPC::Server and Event::RPC::Client.
Event::RPC needs either one of the following modules on the server (they're not necessary on the client):
Event Glib AnyEvent
They're needed for event handling resp. mainloop implementation. If you like to use SSL encryption you need to install
IO::Socket::SSL
Event::RPC needs minimum one of the following modules for data serialisation:
Sereal (::Decoder and ::Encoder) CBOR::XS JSON::XS Storable
Server and client negotiate automatically which serialiser to use to achieve maximum compatibility.
Some words about the Storable module: it's known to be insecure, so Event::RPC uses it as the last option. You can even prevent Event::RPC from using it at all (even when it's installed, which is the case for nearly every Perl installation) - check manpages of Event::Server and Event::Client about the details.
In case you use Storable take care that both client and server use exactly the same version of the Storable module! Otherwise Event::RPC client/server communication will fail badly.
You get the latest installation tarballs and online documentation at this location:
http://www.exit1.org/Event-RPC/
If your system meets the requirements mentioned above, installation is just:
perl Makefile.PL make test make install
To test a specific Event loop implementation, export the variable EVENT_RPC_LOOP:
export EVENT_RPC_LOOP=Event::RPC::Loop::Glib make test
Otherwise Event::RPC will fallback to the most appropriate module installed on your system.
The tarball includes an examples/ directory which contains two programs:
server.pl client.pl
Just execute them with --help to get the usage. They do some very simple communication but are good to test your setup, in particular in a mixed environment.
Although the classes and objects on the server are accessed transparently by the client there are some limitations should be aware of. With a clean object oriented design these should be no problem in real applications:
All objects reside on the server and they keep there! The client just has specially wrapped proxy objects, which trigger the necessary magic to access the object's methods on the server. Complete objects are never transferred from the server to the client, so something like this does not work:
$object->{data} = "changed data";
(assuming $object is a hash ref on the server).
Only method calls are transferred to the server, so even for "simple" data manipulation a method call is necessary:
$object->set_data ("changed data");
As well for reading an object attribute. Accessing a hash key will fail:
my $data = $object->{data};
Instead call a method which returns the 'data' member:
my $data = $object->get_data;
Event::RPC handles methods which return objects. The only requirement is that they are declared as a Object returner on the server (refer to Event::RPC::Server for details), but not if the object is hidden inside a deep complex data structure.
An array or hash ref of objects is Ok, but not more. This would require to much expensive runtime data inspection.
Object receiving parameters are more restrictive, since even hiding them inside one array or hash ref is not allowed. They must be passed as a direkt argument of the method subroutine.
Jörn Reder <joern AT zyn.de>
Copyright (C) 2005-2015 by Jörn Reder <joern AT zyn.de>.
This library is free software; you can redistribute it and/or modify it under the same terms as Perl itself.
| 2018-07-21 | perl v5.26.2 |