1Parallel::Iterator(3) User Contributed Perl DocumentationParallel::Iterator(3)
2
3
4

NAME

6       Parallel::Iterator - Simple parallel execution
7

VERSION

9       This document describes Parallel::Iterator version 1.00
10

SYNOPSIS

12           use Parallel::Iterator qw( iterate );
13
14           # A very expensive way to double 100 numbers...
15
16           my @nums = ( 1 .. 100 );
17
18           my $iter = iterate( sub {
19               my ( $id, $job ) = @_;
20               return $job * 2;
21           }, \@nums );
22
23           my @out = ();
24           while ( my ( $index, $value ) = $iter->() ) {
25               $out[$index] = $value;
26           }
27

DESCRIPTION

29       The "map" function applies a user supplied transformation function to
30       each element in a list, returning a new list containing the transformed
31       elements.
32
33       This module provides a 'parallel map'. Multiple worker processes are
34       forked so that many instances of the transformation function may be
35       executed simultaneously.
36
37       For time consuming operations, particularly operations that spend most
38       of their time waiting for I/O, this is a big performance win. It also
39       provides a simple idiom to make effective use of multi CPU systems.
40
41       There is, however, a considerable overhead associated with forking, so
42       the example in the synopsis (doubling a list of numbers) is not a
43       sensible use of this module.
44
45   Example
46       Imagine you have an array of URLs to fetch:
47
48           my @urls = qw(
49               http://google.com/
50               http://hexten.net/
51               http://search.cpan.org/
52               ... and lots more ...
53           );
54
55       Write a function that retrieves a URL and returns its contents or undef
56       if it can't be fetched:
57
58           sub fetch {
59               my $url = shift;
60               my $resp = $ua->get($url);
61               return unless $resp->is_success;
62               return $resp->content;
63           };
64
65       Now write a function to synthesize a special kind of iterator:
66
67           sub list_iter {
68               my @ar = @_;
69               my $pos = 0;
70               return sub {
71                   return if $pos >= @ar;
72                   my @r = ( $pos, $ar[$pos] );  # Note: returns ( index, value )
73                   $pos++;
74                   return @r;
75               };
76           }
77
78       The returned iterator will return each element of the array in turn and
79       then undef. Actually it returns both the index and the value of each
80       element in the array. Because multiple instances of the transformation
81       function execute in parallel the results won't necessarily come back in
82       order. The array index will later allow us to put completed items in
83       the correct place in an output array.
84
85       Get an iterator for the list of URLs:
86
87           my $url_iter = list_iter( @urls );
88
89       Then wrap it in another iterator which will return the transformed
90       results:
91
92           my $page_iter = iterate( \&fetch, $url_iter );
93
94       Finally loop over the returned iterator storing results:
95
96           my @out = ( );
97           while ( my ( $index, $value ) = $page_iter->() ) {
98               $out[$index] = $value;
99           }
100
101       Behind the scenes your program forked into ten (by default) instances
102       of itself and executed the page requests in parallel.
103
104   Simpler interfaces
105       Having to construct an iterator is a pain so "iterate" is smart enough
106       to do that for you. Instead of passing an iterator just pass a
107       reference to the array:
108
109           my $page_iter = iterate( \&fetch, \@urls );
110
111       If you pass a hash reference the iterator you get back will return key,
112       value pairs:
113
114           my $some_iter = iterate( \&fetch, \%some_hash );
115
116       If the returned iterator is inconvenient you can get back a hash or
117       array instead:
118
119           my @done = iterate_as_array( \&fetch, @urls );
120
121           my %done = iterate_as_hash( \&worker, %jobs );
122
123   How It Works
124       The current process is forked once for each worker. Each forked child
125       is connected to the parent by a pair of pipes. The child's STDIN,
126       STDOUT and STDERR are unaffected.
127
128       Input values are serialised (using Storable) and passed to the workers.
129       Completed work items are serialised and returned.
130
131   Caveats
132       Parallel::Iterator is designed to be simple to use - but the underlying
133       forking of the main process can cause mystifying problems unless you
134       have an understanding of what is going on behind the scenes.
135
136       Worker execution enviroment
137
138       All code apart from the worker subroutine executes in the parent
139       process as normal. The worker executes in a forked instance of the
140       parent process. That means that things like this won't work as
141       expected:
142
143           my %tally = ();
144           my @r = iterate_as_array( sub {
145               my ($id, $name) = @_;
146               $tally{$name}++;       # might not do what you think it does
147               return reverse $name;
148           }, @names );
149
150           # Now print out the tally...
151           while ( my ( $name, $count ) = each %tally ) {
152               printf("%5d : %s\n", $count, $name);
153           }
154
155       Because the worker is a closure it can see the %tally hash from its
156       enclosing scope; but because it's running in a forked clone of the
157       parent process it modifies its own copy of %tally rather than the copy
158       for the parent process.
159
160       That means that after the job terminates the %tally in the parent
161       process will be empty.
162
163       In general you should avoid side effects in your worker subroutines.
164
165       Serialization
166
167       Values are serialised using Storable to pass to the worker subroutine
168       and results from the worker are again serialised before being passed
169       back. Be careful what your values refer to: everything has to be
170       serialised. If there's an indirect way to reach a large object graph
171       Storable will find it and performance will suffer.
172
173       To find out how large your serialised values are serialise one of them
174       and check its size:
175
176           use Storable qw( freeze );
177           my $serialized = freeze $some_obj;
178           print length($serialized), " bytes\n";
179
180       In your tests you may wish to guard against the possibility of a change
181       to the structure of your values resulting in a sudden increase in
182       serialized size:
183
184           ok length(freeze $some_obj) < 1000, "Object too bulky?";
185
186       See the documetation for Storable for other caveats.
187
188       Performance
189
190       Process forking is expensive. Only use Parallel::Iterator in cases
191       where:
192
193       the worker waits for I/O
194           The case of fetching web pages is a good example of this. Fetching
195           a page with LWP::UserAgent may take as long as a few seconds but
196           probably consumes only a few milliseconds of processor time.
197           Running many requests in parallel is a huge win - but be kind to
198           the server you're talking to: don't launch a lot of parallel
199           requests unless it's your server or you know it can handle the
200           load.
201
202       the worker is CPU intensive and you have multiple cores / CPUs
203           If the worker is doing an expensive calculation you can parallelise
204           that across multiple CPU cores. Benchmark first though. There's a
205           considerable overhead associated with Parallel::Iterator; unless
206           your calculations are time consuming that overhead will dwarf
207           whatever time they take.
208

INTERFACE

210   "iterate( [ $options ], $worker, $iterator )"
211       Get an iterator that applies the supplied transformation function to
212       each value returned by the input iterator.
213
214       Instead of an iterator you may pass an array or hash reference and
215       "iterate" will convert it internally into a suitable iterator.
216
217       If you are doing this you may wish to investigate "iterate_as_hash" and
218       "iterate_as_array".
219
220       Options
221
222       A reference to a hash of options may be supplied as the first argument.
223       The following options are supported:
224
225       "workers"
226           The number of concurrent processes to launch. Set this to 0 to
227           disable forking. Defaults to 10 on systems that support fork and 0
228           (disable forking) on those that do not.
229
230       "nowarn"
231           Normally "iterate" will issue a warning and fall back to single
232           process mode on systems on which fork is not available. This option
233           supresses that warning.
234
235       "batch"
236           Ordinarily items are passed to the worker one at a time. If you are
237           processing a large number of items it may be more efficient to
238           process them in batches. Specify the batch size using this option.
239
240           Batching is transparent from the caller's perspective. Internally
241           it modifies the iterators and worker (by wrapping them in
242           additional closures) so that they pack, process and unpack chunks
243           of work.
244
245       "adaptive"
246           Extending the idea of batching a number of work items to amortize
247           the overhead of passing work to and from parallel workers you may
248           also ask "iterate" to heuristically determine the batch size by
249           setting the "adaptive" option to a numeric value.
250
251           The batch size will be computed as
252
253               <number of items seen> / <number of workers> / <adaptive>
254
255           A larger value for "adaptive" will reduce the rate at which the
256           batch size increases. Good values tend to be in the range 1 to 2.
257
258           You can also specify lower and, optionally, upper bounds on the
259           batch size by passing an reference to an array containing ( lower
260           bound, growth ratio, upper bound ). The upper bound may be omitted.
261
262               my $iter = iterate(
263                   { adaptive => [ 5, 2, 100 ] },
264                   $worker, \@stuff );
265
266       "onerror"
267           The action to take when an error is thrown in the iterator.
268           Possible values are 'die', 'warn' or a reference to a subroutine
269           that will be called with the index of the job that threw the
270           exception and the value of $@ thrown.
271
272               iterate( {
273                   onerror => sub {
274                       my ($id, $err) = @_;
275                       $self->log( "Error for index $id: $err" );
276                   },
277                   $worker,
278                   \@jobs
279               );
280
281           The default is 'die'.
282
283   "iterate_as_array"
284       As "iterate" but instead of returning an iterator returns an array
285       containing the collected output from the iterator. In a scalar context
286       returns a reference to the same array.
287
288       For this to work properly the input iterator must return (index, value)
289       pairs. This allows the results to be placed in the correct slots in the
290       output array. The simplest way to do this is to pass an array reference
291       as the input iterator:
292
293           my @output = iterate_as_array( \&some_handler, \@input );
294
295   "iterate_as_hash"
296       As "iterate" but instead of returning an iterator returns a hash
297       containing the collected output from the iterator. In a scalar context
298       returns a reference to the same hash.
299
300       For this to work properly the input iterator must return (key, value)
301       pairs. This allows the results to be placed in the correct slots in the
302       output hash. The simplest way to do this is to pass a hash reference as
303       the input iterator:
304
305           my %output = iterate_as_hash( \&some_handler, \%input );
306

CONFIGURATION AND ENVIRONMENT

308       Parallel::Iterator requires no configuration files or environment
309       variables.
310

DEPENDENCIES

312       None.
313

INCOMPATIBILITIES

315       None reported.
316

BUGS AND LIMITATIONS

318       No bugs have been reported.
319
320       Please report any bugs or feature requests to
321       "bug-parallel-iterator@rt.cpan.org", or through the web interface at
322       <http://rt.cpan.org>.
323

AUTHOR

325       Andy Armstrong  "<andy@hexten.net>"
326

THANKS

328       Aristotle Pagaltzis for the END handling suggestion and patch.
329
331       Copyright (c) 2007, Andy Armstrong "<andy@hexten.net>". All rights
332       reserved.
333
334       This module is free software; you can redistribute it and/or modify it
335       under the same terms as Perl itself. See perlartistic.
336

DISCLAIMER OF WARRANTY

338       BECAUSE THIS SOFTWARE IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY
339       FOR THE SOFTWARE, TO THE EXTENT PERMITTED BY APPLICABLE LAW. EXCEPT
340       WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER
341       PARTIES PROVIDE THE SOFTWARE "AS IS" WITHOUT WARRANTY OF ANY KIND,
342       EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED
343       WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE
344       ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE SOFTWARE IS WITH
345       YOU. SHOULD THE SOFTWARE PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL
346       NECESSARY SERVICING, REPAIR, OR CORRECTION.
347
348       IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
349       WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR
350       REDISTRIBUTE THE SOFTWARE AS PERMITTED BY THE ABOVE LICENCE, BE LIABLE
351       TO YOU FOR DAMAGES, INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL, OR
352       CONSEQUENTIAL DAMAGES ARISING OUT OF THE USE OR INABILITY TO USE THE
353       SOFTWARE (INCLUDING BUT NOT LIMITED TO LOSS OF DATA OR DATA BEING
354       RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD PARTIES OR A
355       FAILURE OF THE SOFTWARE TO OPERATE WITH ANY OTHER SOFTWARE), EVEN IF
356       SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH
357       DAMAGES.
358
359
360
361perl v5.12.3                      2011-05-19             Parallel::Iterator(3)
Impressum