Re: Yet another example to get URLs in parallel

[marioroy]

Hi karlgoethebier,

Let's imagine for a minute, the following statement.

$mutex->enter( $shared->set( $_ => $fetch->($_) ) );

1. the worker enters a mutex
   meaning one worker runs solo while inside the mutex
2. then does a fetch on given URL
3. then stores the result into a shared hash
4. finally, leaves the mutex

The statement above is causing MCE workers to run serially, not parallel. I've gone back to your earlier example here and that looks fine. However for this thread, maybe running solo is what karlgoethebier intended and respecting his decision to do so. Surely, he wanted the code to run parallel ;-).

    mce_loop {
        MCE->yield;

        # run parallel
        my $url = $_; my $result = $fetch->($url);

        # run solo to store the result
        $mutex->enter( $shared->set( $url => $result ) );

        # am back to running parallel
        # ...
    }

A mutex isn't needed when IPC involves a single trip, typical for the OO interface.

    mce_loop {
        MCE->yield;

        # run parallel, without a mutex
        $shared->set( $_ => $fetch->($_) );

    }

A mutex is often necessary for a shared hash when constructed via the TIE interface.

    tie my %hash, 'MCE::Shared';
    my $shared = MCE::Shared->hash();
    my $mutex = MCE::Mutex->new();

    $hash{number} = 0;   # 1 trip, store
    $shared->set( number => 0 );   # 1 trip

    # 2 trips fetch and store, needs a mutex
    $mutex->enter( $hash{number} += 2 );

    # 1 trip via the OO interface
    $shared->incrby( number => 2 );

Regards, Mario