The Perl Toolchain Summit needs more sponsors. If your company depends on Perl, please support this very important event.

=head1 NAME

Net::SIP::Dispatcher::Eventloop - simple event loop for L<Net::SIP::Dispatcher>

=head1 SYNOPSIS

  my $loop = Net::SIP::Dispatcher::Eventloop->new;
  $loop->addFD( $fd, $callback );
  $loop->add_timer( 10,$callback );
  $loop->loop;

=head1 DESCRIPTION

The package implements a simple event loop. It's not optimized
for speed but it is provided as a simple implementation in case the
users application does not has an event loop yet.

Because the eventloop object can be given in the constructor
of L<Net::SIP::Dispatcher> you might provide an alternative implementation,
which implemented the described methods.

=head1 CONSTRUCTOR

=over 4

=item new

Creates new event loop, returns created object

=back

=head1 METHODS

=over 4

=item addFD ( HANDLE, CALLBACK, ?NAME )

Adds file handle HANDLE to the event loop, so that CALLBACK gets triggered
if HANDLE is readable. CALLBACK is a callback accepted by B<invoke_callback>
in L<Net::SIP::Util>. The callback will be invoked with HANDLE as an additional
argument.

NAME can be used to aid debugging, it will be shown in the debug messages
once the FD gets ready.

If there was already a callback for HANDLE it gets replaced by the new one.

IMPORTANT NOTE: CALLBACK gets triggered if HANDLE *is* readable inside the
loop, not if HANDLE *gets* readable. Unlike with L<Event::Lib> or similar
the CALLBACK is not triggered by the edge, but by the level (like poll(2)
or select(2)). So if 2 bytes come in at the handle and one reads only
1 byte in the callback the callback gets triggered again for more data.
You have to watch this, if you want to integrate L<Net::SIP> with your
existing event loop.

=item delFD ( HANDLE )

Removes HANDLE from loop, e.g. no more checking for readability will be done.

=item add_timer( WHEN, CALLBACK, [ REPEAT ] )

Adds timer which gets triggered at WHEN or C<< now + WHEN >>. Depending on
the value of WHEN it gets interpreted as the number of seconds since 1970-01-01
(when it's really big) or as a relative time (when it's not that big).
WHEN can be floating point for subseconds resolution.

CALLBACK is a callback accepted by B<invoke_callback> in L<Net::SIP::Util>.
It gets invoked with the timer object (see later) as an additional argument,
which has a method B<cancel> for canceling the (repeating) timer.

REPEAT is the number of seconds between each invocation of the timer. If greater
then 0 (subsection resulution possible) the callback will be called each
REPEAT seconds, after it was called the first time at WHEN.

The method will return an object which has a method B<cancel> which can be
used to cancel the timer before it gets triggered (or gets triggered the next
time in case of repeating timers).

=item looptime

Returns the current loop time in subseconds resolution (using B<gettimeofday>
from L<Time::HiRes>). This is not the current time, but the time, when the
last event in the loop occured (e.g. when the select(2) call returned)

=item loop ( [ TIMEOUT, \@STOPVAR ] )

The main loop, e.g. continuiosly checks timers and file handles for the
events and calls callbacks.

If TIMEOUT is given it will run the loop for at most TIMEOUT seconds,
then the method will return. Undefined TIMEOUT means that it will never
return because of timeout and TIMEOUT of 0 means that it will check all
timers and handles only once and then return.

@STOPVAR is a list of scalar references. These scalars are expected to
be changed from the callbacks, so it will check after each loop cycle, e.g.
after all callbacks are called (timers and handles) if any of these scalars
is TRUE, in which case it will end the loop.

The behavior with STOPVAR cannot be found in most existing event loops.
If you want to integrate L<Net::SIP> with your own event loop you migth
simply wrap all callbacks given in B<addFD> and B<add_timer> in another
callback which at the end checks the stopvars and terminates the 3rd-party loop
in a loop-specific way.

=back