VERSION

This document describes MCE::Relay version 1.608

SYNOPSIS

use MCE::Flow;
my $file = shift || \*STDIN;
## Line Count #######################################
mce_flow_f {
use_slurpio => 1, init_relay => 0,
},
sub {
my ($mce, $slurp_ref, $chunk_id) = @_;
my $line_count = ($$slurp_ref =~ tr/\n//);
## Receive and pass on updated information.
my $lines_read = MCE::relay { $_ += $line_count };
}, $file;
my $total_lines = MCE->relay_final;
print {*STDERR} "$total_lines\n";
## Orderly Action ###################################
mce_flow_f {
use_slurpio => 1, init_relay => 0,
},
sub {
my ($mce, $slurp_ref, $chunk_id) = @_;
## Exclusive access to STDOUT. Relays 0.
MCE::relay { print $$slurp_ref };
}, $file;

DESCRIPTION

This module enables workers to receive and pass on information orderly with zero involvement by the manager process while running. The module is loaded automatically when init_relay is specified.

All workers must participate when relaying data. Calling relay more than once is not recommended inside the block. Doing so will stall the application.

Relaying is not met for passing big data. The last worker will likely stall if exceeding the buffer size for the socket. Not exceeding 8 KiB is safe across all platforms.

API DOCUMENTATION

MCE->relay ( sub { code } )

MCE::relay { code }

Relay is enabled by specifying the init_relay option which takes a hash or array reference, or a scalar value. Relaying is orderly and driven by chunk_id when processing data, otherwise task_wid. Omitting the code block (e.g. MCE::relay) relays forward.

Below, relaying multiple values via a HASH reference.

   use MCE::Flow max_workers => 4;
   mce_flow {
      init_relay => { p => 0, e => 0 },
   },
   sub {
      my $wid = MCE->wid;
      ## do work
      my $pass = $wid % 3;
      my $errs = $wid % 2;
      ## relay
      my %last_rpt = MCE::relay { $_->{p} += $pass; $_->{e} += $errs };
      MCE->print("$wid: passed $pass, errors $errs\n");
      return;
   };
   my %results = MCE->relay_final;
   print "   passed $results{p}, errors $results{e} final\n\n";
   -- Output
   1: passed 1, errors 1
   2: passed 2, errors 0
   3: passed 0, errors 1
   4: passed 1, errors 0
      passed 4, errors 2 final

Or multiple values via an ARRAY reference.

   use MCE::Flow max_workers => 4;
   mce_flow {
      init_relay => [ 0, 0 ],
   },
   sub {
      my $wid = MCE->wid;
      ## do work
      my $pass = $wid % 3;
      my $errs = $wid % 2;
      ## relay
      my @last_rpt = MCE::relay { $_->[0] += $pass; $_->[1] += $errs };
      MCE->print("$wid: passed $pass, errors $errs\n");
      return;
   };
   my ($pass, $errs) = MCE->relay_final;
   print "   passed $pass, errors $errs final\n\n";
   -- Output
   1: passed 1, errors 1
   2: passed 2, errors 0
   3: passed 0, errors 1
   4: passed 1, errors 0
      passed 4, errors 2 final

Or simply a scalar value.

   use MCE::Flow max_workers => 4;
   mce_flow {
      init_relay => 0,
   },
   sub {
      my $wid = MCE->wid;
      ## do work
      my $bytes_read = 1000 + ((MCE->wid % 3) * 3);
      ## relay
      my $last_offset = MCE::relay { $_ += $bytes_read };
      ## output
      MCE->print("$wid: $bytes_read\n");
      return;
   };
   my $total = MCE->relay_final;
   print "   $total size\n\n";
   -- Output
   1: 1003
   2: 1006
   3: 1000
   4: 1003
      4012 size

MCE->relay_final ( void )

Call this method to obtain the final relay values after running. See included example findnull.pl for another use case.

   use MCE max_workers => 4;
   my $mce = MCE->new(
      init_relay => [ 0, 100 ],       ## initial values (two counters)
      user_func => sub {
         my ($mce) = @_;
         ## do work
         my ($acc1, $acc2) = (10, 20);
         ## relay to next worker
         MCE::relay { $_->[0] += $acc1; $_->[1] += $acc2 };
         return;
      }
   )->run;
   my ($cnt1, $cnt2) = $mce->relay_final;
   print "$cnt1 : $cnt2\n";
   -- Output
   40 : 180

MCE->relay_recv ( void )

The relay_recv method allows one to perform an exclusive action prior to relaying. Below, the user_func is taken from the cat.pl example. Relaying is chunk_id driven (or task_wid when not processing input), thus orderly.

   user_func => sub {
      my ($mce, $chunk_ref, $chunk_id) = @_;
      if ($n_flag) {
         ## Relays the total lines read.
         my $output = ''; my $line_count = ($$chunk_ref =~ tr/\n//);
         my $lines_read = MCE::relay { $_ += $line_count };
         open my $fh, '<', $chunk_ref;
         $output .= sprintf "%6d\t%s", ++$lines_read, $_ while (<$fh>);
         close $fh;
         $output .= ":$chunk_id";
         MCE->do('display_chunk', $output);
      }
      else {
         ## The following is another way to have ordered output. Workers
         ## write directly to STDOUT exclusively without any involvement
         ## from the manager process. The statements between relay_recv
         ## and relay run serially and most important orderly.
         ## STDERR/OUT flush automatically inside worker threads and
         ## processes. Disable buffering on file handles otherwise.
         MCE->relay_recv;             ## my $val = MCE->relay_recv;
                                      ## relay simply forwards 0 below
         print $$chunk_ref;           ## exclusive access to STDOUT
                                      ## important, flush immediately
         MCE->relay;
      }
      return;
   }

INDEX

MCE

AUTHOR

Mario E. Roy, <marioeroy AT gmail DOT com>