++ed by:
AWNCORP CARWASH DAVIDO DRAEGTUN EGOR

12 PAUSE user(s)
5 non-PAUSE user(s).

Mario Roy

NAME

MCE::Stream - Parallel stream model for chaining multiple maps and greps

VERSION

This document describes MCE::Stream version 1.514

SYNOPSIS

   ## Exports mce_stream, mce_stream_f, mce_stream_s
   use MCE::Stream;

   my (@m1, @m2, @m3);

   ## Default mode is map and processed from right-to-left
   @m1 = mce_stream sub { $_ * 3 }, sub { $_ * 2 }, 1..10000;
   mce_stream \@m2, sub { $_ * 3 }, sub { $_ * 2 }, 1..10000;

   ## Native Perl
   @m3 = map { $_ * $_ } grep { $_ % 5 == 0 } 1..10000;

   ## Streaming grep and map in parallel
   mce_stream \@m3,
      { mode => 'map',  code => sub { $_ * $_ } },
      { mode => 'grep', code => sub { $_ % 5 == 0 } }, 1..10000;

   ## Array or array_ref
   my @a = mce_stream sub { $_ * $_ }, 1..10000;
   my @b = mce_stream sub { $_ * $_ }, [ 1..10000 ];

   ## File_path, glob_ref, or scalar_ref
   my @c = mce_stream_f sub { chomp; $_ }, "/path/to/file";
   my @d = mce_stream_f sub { chomp; $_ }, $file_handle;
   my @e = mce_stream_f sub { chomp; $_ }, \$scalar;

   ## Sequence of numbers (begin, end [, step, format])
   my @f = mce_stream_s sub { $_ * $_ }, 1, 10000, 5;
   my @g = mce_stream_s sub { $_ * $_ }, [ 1, 10000, 5 ];

   my @h = mce_stream_s sub { $_ * $_ }, {
      begin => 1, end => 10000, step => 5, format => undef
   };

DESCRIPTION

This module allows one to stream multiple map and/or grep operations in parallel. Code blocks run simultaneously from right-to-left. Chunk data are sent immediately to the next code block during processing. The results are appended immediately as well when passing the reference to the array.

   ## Appends are serialized, even out-of-order ok, but immediately.
   ## Out-of-order chunks are held temporarily until ordered chunks
   ## arrive before appending.

   mce_stream \@a, sub { $_ }, sub { $_ }, sub { $_ }, 1..10000;

   ##                                                    input
   ##                                        chunk1      input
   ##                            chunk3      chunk2      input
   ##                chunk2      chunk2      chunk3      input
   ##   append1      chunk3      chunk1      chunk4      input
   ##   append2      chunk1      chunk5      chunk5      input
   ##   append3      chunk5      chunk4      chunk6      ...
   ##   append4      chunk4      chunk6      ...
   ##   append5      chunk6      ...
   ##   append6      ...
   ##   ...
   ##

MCE incurs a small overhead due to passing of data. Therefore, a fast code block will likely run faster when chaining multiple map functions natively in Perl. However, the overhead quickly diminishes as the complexity of the code increases.

   ## 0.542 secs -- baseline using the native map function
   my @m1 = map { $_ * 4 } map { $_ * 3 } map { $_ * 2 } 1..1000000;

   ## 0.765 secs -- this is quite amazing considering data passing
   my @m2 = mce_stream
         sub { $_ * 4 }, sub { $_ * 3 }, sub { $_ * 2 }, 1..1000000;

   ## 0.592 secs -- appends to @m3 are immediate
   my @m3; mce_stream \@m3,
         sub { $_ * 4 }, sub { $_ * 3 }, sub { $_ * 2 }, 1..1000000;

The mce_stream_s function will provide better times, useful when the input data is simply a range of numbers. Workers generate sequences mathematically among themselves without any interaction from the manager process. Two arguments are required for mce_stream_s (begin, end). Step defaults to 1 if begin is smaller than end, otherwise -1.

   ## 0.447 secs -- numbers are generated mathematically via sequence
   my @m4; mce_stream_s \@m4,
         sub { $_ * 4 }, sub { $_ * 3 }, sub { $_ * 2 }, 1, 1000000;

OVERRIDING DEFAULTS

The following list 6 options which may be overridden when loading the module.

   use Sereal qw(encode_sereal decode_sereal);

   use MCE::Stream
         default_mode => 'grep',               ## Default 'map'
         max_workers  => 8,                    ## Default 'auto'
         chunk_size   => 500,                  ## Default 'auto'
         tmp_dir      => "/path/to/app/tmp",   ## $MCE::Signal::tmp_dir
         freeze       => \&encode_sereal,      ## \&Storable::freeze
         thaw         => \&decode_sereal       ## \&Storable::thaw
   ;

There is a simpler way to enable Sereal with MCE 1.5. The following will attempt to use Sereal if available, otherwise will default back to using Storable for serialization.

   use MCE::Stream Sereal => 1;

   ## Serialization is through Sereal if available.
   my @m2 = mce_stream sub { $_ * $_ }, 1..10000;

CUSTOMIZING MCE

init

The init function accepts a hash of MCE options. The gather and bounds_only options, if specified, will be ignored due to being used internally by the module (not shown below).

   use MCE::Stream;

   MCE::Stream::init {
      chunk_size => 1, max_workers => 4,

      user_begin => sub {
         print "## ", MCE->wid, " started\n";
      },

      user_end => sub {
         print "## ", MCE->wid, " completed\n";
      }
   };

   my @a = mce_stream sub { $_ * $_ }, 1..100;

   print "\n", "@a", "\n";

   -- Output

   ## 1 started
   ## 2 started
   ## 3 started
   ## 4 started
   ## 3 completed
   ## 1 completed
   ## 2 completed
   ## 4 completed

   1 4 9 16 25 36 49 64 81 100 121 144 169 196 225 256 289 324 361
   400 441 484 529 576 625 676 729 784 841 900 961 1024 1089 1156
   1225 1296 1369 1444 1521 1600 1681 1764 1849 1936 2025 2116 2209
   2304 2401 2500 2601 2704 2809 2916 3025 3136 3249 3364 3481 3600
   3721 3844 3969 4096 4225 4356 4489 4624 4761 4900 5041 5184 5329
   5476 5625 5776 5929 6084 6241 6400 6561 6724 6889 7056 7225 7396
   7569 7744 7921 8100 8281 8464 8649 8836 9025 9216 9409 9604 9801
   10000

Like with MCE::Stream::init above, MCE options may be specified using an anonymous hash for the first argument. Notice how both max_workers and task_name can take an anonymous array for setting values individually for each code block. Remember that MCE::Stream processes from right-to-left when setting the individual values.

   use MCE::Stream;

   my @a = mce_stream {
      max_workers => [ 2, 4, 3, ], task_name => [ 'c', 'b', 'a' ],

      user_end => sub {
         my ($task_id, $task_name) = (MCE->task_id, MCE->task_name);
         print "$task_id - $task_name completed\n";
      },

      task_end => sub {
         my ($mce, $task_id, $task_name) = @_;
         MCE->print("$task_id - $task_name ended\n");
      }
   },
   sub { $_ * 4 },             ## 2 workers, named c
   sub { $_ * 3 },             ## 4 workers, named b
   sub { $_ * 2 }, 1..10000;   ## 3 workers, named a

   -- Output

   0 - a completed
   0 - a completed
   0 - a completed
   0 - a ended
   1 - b completed
   1 - b completed
   1 - b completed
   1 - b completed
   1 - b ended
   2 - c completed
   2 - c completed
   2 - c ended

Note that the anonymous hash, for specifying options, also comes first when passing the array reference.

   my @a; mce_stream {
      ...
   }, \@a, sub { ... }, sub { ... }, 1..10000;

API DOCUMENTATION

Scripts using MCE::Stream can be written using the long or short form. The long form becomes relavant when mixing modes. Processing occurs from right-to-left.

   my @m3 = mce_stream
      { mode => 'map',  code => sub { $_ * $_ } },
      { mode => 'grep', code => sub { $_ % 5 == 0 } }, 1..10000;

   my @m4; mce_stream \@m4,
      { mode => 'map',  code => sub { $_ * $_ } },
      { mode => 'grep', code => sub { $_ % 5 == 0 } }, 1..10000;

For multiple grep blocks, the short form can be used. Simply specify the default mode for the module. The two valid values for default_mode is 'grep' and 'map'.

   use MCE::Stream default_mode => 'grep';

   my @f = mce_stream_f sub { /ending$/ }, sub { /^starting/ }, $file;

The following assumes 'map' for default_mode in order to demonstrate all the possibilities of passing input data into the code block.

mce_stream sub { code }, list

Input data can be defined using a list or passing a reference to an array.

   my @a = mce_stream sub { $_ * 2 }, 1..1000;
   my @b = mce_stream sub { $_ * 2 }, [ 1..1000 ];
mce_stream_f sub { code }, file

The fastest of these is the /path/to/file. Workers communicate the next offset position among themselves without any interaction from the manager process.

   my @c = mce_stream_f sub { chomp; $_ . "\r\n" }, "/path/to/file";
   my @d = mce_stream_f sub { chomp; $_ . "\r\n" }, $file_handle;
   my @e = mce_stream_f sub { chomp; $_ . "\r\n" }, \$scalar;
mce_stream_s sub { code }, sequence

Sequence can be defined as a list, an array reference, or a hash reference. The functions require both begin and end values to run. Step and format are optional. The format is passed to sprintf (% may be omitted below).

   my ($beg, $end, $step, $fmt) = (10, 20, 0.1, "%4.1f");

   my @f = mce_stream_s sub { $_ }, $beg, $end, $step, $fmt;
   my @g = mce_stream_s sub { $_ }, [ $beg, $end, $step, $fmt ];

   my @h = mce_stream_s sub { $_ }, {
      begin => $beg, end => $end, step => $step, format => $fmt
   };
mce_stream { input_data => iterator }, sub { code }

An iterator reference can by specified for input data. Notice the anonymous hash as the first argument to mce_stream. The only other way is to specify input_data via MCE::Stream::init. This prevents MCE::Stream from configuring the iterator reference as another user task which will not work.

Iterators are described under "SYNTAX for INPUT_DATA" at MCE::Core.

   MCE::Stream::init {
      input_data => iterator
   };

   my @a = mce_stream sub { $_ * 3 }, sub { $_ * 2 };

MANUAL SHUTDOWN

finish

MCE workers remain persistent as much as possible after running. Shutdown occurs when the script exits. One can manually shutdown MCE by simply calling finish after running. This resets the MCE instance.

   use MCE::Stream;

   MCE::Stream::init {
      chunk_size => 20, max_workers => 'auto'
   };

   my @a = mce_stream { ... } 1..100;

   MCE::Stream::finish;

INDEX

MCE

AUTHOR

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

LICENSE

This program is free software; you can redistribute it and/or modify it under the terms of either: the GNU General Public License as published by the Free Software Foundation; or the Artistic License.

See http://dev.perl.org/licenses/ for more information.




Hosting generously
sponsored by Bytemark