MCE::Stream - Parallel stream model for chaining multiple maps and greps
This document describes MCE::Stream version 1.508
## 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 };
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;
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;
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;
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.
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 ];
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; $_ . "\r\n" }, \$scalar;
Sequence can be defined as a list, an array reference, or a hash reference. The functions require both begin and end values to run. Step and format are optional. The format is passed to sprintf (% may be omitted below).
my ($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 };
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 };
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;
MCE
Mario E. Roy, <marioeroy AT gmail DOT com>
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.
To install MCE, copy and paste the appropriate command in to your terminal.
cpanm
cpanm MCE
CPAN shell
perl -MCPAN -e shell install MCE
For more information on module installation, please visit the detailed CPAN module installation guide.