MCE::Candy(3) Sugar methods and output iterators for Many-Core Engine

VERSION

This document describes MCE::Candy version 1.608

DESCRIPTION

This module provides a collection of sugar methods and helpful output iterators for preserving output order.

FOR SUGAR METHODS

The sugar methods described below were created prior to the 1.5 release which added MCE Models. This module is loaded automatically upon calling a ``for'' method.

$mce->forchunk ( $input_data [, { options } ], sub { ... } )

Forchunk, foreach, and forseq are sugar methods in MCE. Workers are spawned automatically, the code block is executed in parallel, and shutdown is called. Do not call these methods if workers must persist afterwards.

Specifying options is optional. Valid options are the same as for the process method.

   ## Declare a MCE instance.
   my $mce = MCE->new(
      chunk_size  => 20,
      max_workers => $max_workers
   );
   ## Arguments inside the code block are the same as passed to user_func.
   $mce->forchunk(\@input_data, sub {
      my ($mce, $chunk_ref, $chunk_id) = @_;
      foreach ( @{ $chunk_ref } ) {
         MCE->print("$chunk_id: $_\n");
      }
   });
   ## Passing chunk_size as an option.
   $mce->forchunk(\@input_data, { chunk_size => 30 }, sub {
      ...
   });

$mce->foreach ( $input_data [, { options } ], sub { ... } )

Foreach implies chunk_size => 1 and cannot be overwritten. Thus, looping is not necessary inside the block.

   my $mce = MCE->new(
      max_workers => $max_workers
   );
   $mce->foreach(\@input_data, sub {
      my ($mce, $chunk_ref, $chunk_id) = @_;
      my $row = $chunk_ref->[0];
      MCE->print("$chunk_id: $row\n");
   });

$mce->forseq ( $sequence_spec [, { options } ], sub { ... } )

Sequence can be defined using an array or hash reference.

   my $mce = MCE->new(
      max_workers => 3
   );
   $mce->forseq([ 20, 40 ], sub {
      my ($mce, $n, $chunk_id) = @_;
      my $result = `ping 192.168.1.${n}`;
      ...
   });
   $mce->forseq({ begin => 15, end => 10, step => -1 }, sub {
      my ($mce, $n, $chunk_id) = @_;
      print $n, " from ", MCE->wid, "\n";
   });

The $n_seq variable points to an array_ref of sequences. Chunk size defaults to 1 when not specified.

   $mce->forseq([ 20, 80 ], { chunk_size => 10 }, sub {
      my ($mce, $n_seq, $chunk_id) = @_;
      for my $n ( @{ $n_seq } ) {
         my $result = `ping 192.168.1.${n}`;
         ...
      }
   });

OUTPUT ITERATORS WITH INPUT

This module includes 2 output iterators which are useful for preserving output order while gathering data. These cover the 2 general use cases. The chunk_id value must be the first argument to gather. Gather must also not be called more than once inside the block.

gather => MCE::Candy::out_iter_array( \@array )

The example utilizes the Core API with chunking disabled. Basically, setting chunk_size to 1.

   use MCE;
   use MCE::Candy;
   my @results;
   my $mce = MCE->new(
      chunk_size => 1, max_workers => 4,
      gather => MCE::Candy::out_iter_array(\@results),
      user_func => sub {
         my ($mce, $chunk_ref, $chunk_id) = @_;
         $mce->gather($chunk_id, $chunk_ref->[0] * 2);
      }
   );
   $mce->process([ 100 .. 109 ]);
   print "@results", "\n";
   -- Output
   200 202 204 206 208 210 212 214 216 218

Chunking may be desired for thousands or more items. In other words, wanting to reduce the overhead placed on IPC.

   use MCE;
   use MCE::Candy;
   my @results;
   my $mce = MCE->new(
      chunk_size => 100, max_workers => 4,
      gather => MCE::Candy::out_iter_array(\@results),
      user_func => sub {
         my ($mce, $chunk_ref, $chunk_id) = @_;
         my @output;
         foreach my $item (@{ $chunk_ref }) {
            push @output, $item * 2;
         }
         $mce->gather($chunk_id, @output);
      }
   );
   $mce->process([ 100_000 .. 200_000 - 1 ]);
   print scalar @results, "\n";
   -- Output
   100000

gather => MCE::Candy::out_iter_fh( $fh )

Let's change things a bit and use MCE::Flow for the next 2 examples. Chunking is not desired for the first example.

   use MCE::Flow;
   use MCE::Candy;
   open my $fh, '>', '/tmp/foo.txt';
   mce_flow {
      chunk_size => 1, max_workers => 4,
      gather => MCE::Candy::out_iter_fh($fh)
   },
   sub {
      my ($mce, $chunk_ref, $chunk_id) = @_;
      $mce->gather($chunk_id, $chunk_ref->[0] * 2, "\n");
   }, (100 .. 109);
   close $fh;
   -- Output sent to '/tmp/foo.txt'
   200
   202
   204
   206
   208
   210
   212
   214
   216
   218

Chunking is desired for the next example due to processing many thousands.

   use MCE::Flow;
   use MCE::Candy;
   open my $fh, '>', '/tmp/foo.txt';
   mce_flow {
      chunk_size => 100, max_workers => 4,
      gather => MCE::Candy::out_iter_fh( $fh )
   },
   sub {
      my ($mce, $chunk_ref, $chunk_id) = @_;
      my @output;
      foreach my $item (@{ $chunk_ref }) {
         push @output, ($item * 2) . "\n";
      }
      $mce->gather($chunk_id, @output);
   }, (100_000 .. 200_000 - 1);
   close $fh;
   print -s '/tmp/foo.txt', "\n";
   -- Output
   700000

OUTPUT ITERATORS WITHOUT INPUT

Input data is not a requirement for using the output iterators included in this module. The 'chunk_id' value is set uniquely and the same as 'wid' when not processing input data.

gather => MCE::Candy::out_iter_array( \@array )

   use MCE::Flow;
   use MCE::Candy;
   my @results;
   mce_flow {
      max_workers => 'auto', ## Note that 'auto' is never higher than 8
      gather => MCE::Candy::out_iter_array(\@results)
   },
   sub {
      my ($mce) = @_;        ## This line is not necessary
                             ## Calling via module okay; e.g: MCE->method
      ## Do work
      ## Sending a complex data structure is allowed
      ## Output will become orderly by iterator
      $mce->gather( $mce->chunk_id, {
         wid => $mce->wid, result => $mce->wid * 2
      });
   };
   foreach my $href (@results) {
      print $href->{wid} .": ". $href->{result} ."\n";
   }
   -- Output
   1: 2
   2: 4
   3: 6
   4: 8
   5: 10
   6: 12
   7: 14
   8: 16

gather => MCE::Candy::out_iter_fh( $fh )

   use MCE::Flow;
   use MCE::Candy;
   open my $fh, '>', '/tmp/out.txt';
   mce_flow {
      max_workers => 'auto', ## See get_ncpu in <MCE::Util|MCE::Util> 
      gather => MCE::Candy::out_iter_fh($fh)
   },
   sub {
      my $output = "# Worker ID: " . MCE->wid . "\n";
      ## Append results to $output string
      $output .= (MCE->wid * 2) . "\n\n";
      ## Output will become orderly by iterator
      MCE->gather( MCE->wid, $output );
   };
   close $fh;
   -- Output
   # Worker ID: 1
   2
   # Worker ID: 2
   4
   # Worker ID: 3
   6
   # Worker ID: 4
   8
   # Worker ID: 5
   10
   # Worker ID: 6
   12
   # Worker ID: 7
   14
   # Worker ID: 8
   16

INDEX

MCE

AUTHOR

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