1MCE::Core(3) User Contributed Perl Documentation MCE::Core(3)
2
6 MCE::Core - Documentation describing the core MCE API
7
9 This document describes MCE::Core version 1.879
10
12 This is a simplistic use case of MCE running with 5 workers.
13 ## Construction using the Core API
15 use MCE;
17 my $mce = MCE->new(
19 max_workers => 5,
20 user_func => sub {
21 my ($mce) = @_;
22 $mce->say("Hello from " . $mce->wid);
23 }
24 );
25 $mce->run;
27 ## Construction using a MCE model
29 use MCE::Flow max_workers => 5;
31 mce_flow sub {
33 my ($mce) = @_;
34 MCE->say("Hello from " . MCE->wid);
35 };
36 -- Output
38 Hello from 2
40 Hello from 4
41 Hello from 5
42 Hello from 1
43 Hello from 3
44 MCE->new ( [ options ] )
46 Below, a new instance is configured with all available options.
47 use MCE;
49 my $mce = MCE->new(
51 max_workers => 8, ## Default 1
53 # Number of workers to spawn.
55 # MCE sets an upper-limit of 8 for 'auto'. MCE 1.521+.
57 # max_workers => 'auto', ## # of lcores, 8 maximum
58 # max_workers => 'auto-1', ## 7 on HW with 16 lcores
59 # max_workers => 'auto-1', ## 3 on HW with 4 lcores
60 # Specify a percentage. MCE 1.875+.
62 # max_workers => '25%', ## 4 on HW with 16 lcores
63 # max_workers => '50%', ## 8 on HW with 16 lcores
64 # Run on all logical cores.
66 # max_workers => MCE::Util::get_ncpu(),
67 chunk_size => 2000, ## Default 1
69 # Can also take a suffix; k (kibiBytes) or m (mebiBytes).
71 # The default is 1 when using the Core API and 'auto' for
72 # MCE Models. For arrays or queues, chunk_size means the
73 # number of records per chunk. For iterators, MCE will not
74 # use chunk_size, though the iterator may use it to determine
75 # how much to return per iteration. For files, smaller than or
76 # equal to 8192 is the number of records. Greater than 8192
77 # is the number of bytes. MCE reads until the end of record
78 # before calling user_func.
79 # chunk_size => 1, ## Consists of 1 record
81 # chunk_size => 1000, ## Consists of 1000 records
82 # chunk_size => '16k', ## Approximate 16 kibiBytes (KiB)
83 # chunk_size => '20m', ## Approximate 20 mebiBytes (MiB)
84 tmp_dir => $tmp_dir, ## Default $MCE::Signal::tmp_dir
86 # Default is $MCE::Signal::tmp_dir which points to
88 # $ENV{TEMP} if defined. Otherwise, tmp_dir points
89 # to a location under /tmp.
90 freeze => \&encode_sereal, ## Default \&Storable::freeze
92 thaw => \&decode_sereal, ## Default \&Storable::thaw
93 # Release 1.412 allows freeze and thaw to be overridden.
95 # Simply include a serialization module prior to loading
96 # MCE. Configure freeze/thaw options.
97 # use Sereal qw( encode_sereal decode_sereal );
99 # use CBOR::XS qw( encode_cbor decode_cbor );
100 # use JSON::XS qw( encode_json decode_json );
101 #
102 # use MCE;
103 gather => \@a, ## Default undef
105 # Release 1.5 allows for gathering of data to an array or
107 # hash reference, a MCE::Queue/Thread::Queue object, or code
108 # reference. One invokes gathering by calling the gather
109 # method as often as needed.
110 # gather => \@array,
112 # gather => \%hash,
113 # gather => $queue,
114 # gather => \&order,
115 init_relay => 0, ## Default undef
117 # For specifying the initial relay value. Allowed values
119 # are array_ref, hash_ref, or scalar. The MCE::Relay module
120 # is loaded automatically when specified.
121 # init_relay => \@array,
123 # init_relay => \%hash,
124 # init_relay => scalar,
125 input_data => $input_file, ## Default undef
127 RS => "\n>", ## Default undef
128 # input_data => '/path/to/file' ## Process file
130 # input_data => \@array ## Process array
131 # input_data => \*FILE_HNDL ## Process file handle
132 # input_data => $io ## Process IO::All { File, Pipe, STDIO }
133 # input_data => \$scalar ## Treated like a file
134 # input_data => \&iterator ## User specified iterator
135 # The RS option (for input record separator) applies to files
137 # and file handles.
138 # MCE applies additional logic when RS begins with a newline
140 # character; e.g. RS => "\n>". It trims away characters after
141 # the newline and prepends them to the next record.
142 #
143 # Typically, the left side is what happens for $/ = "\n>".
144 # The right side is what user_func receives.
145 #
146 # All records begin with > and end with \n
147 # Record 1: >seq1 ... \n> (to) >seq1 ... \n
148 # Record 2: seq2 ... \n> >seq2 ... \n
149 # Record 3: seq3 ... \n> >seq3 ... \n
150 # Last Rec: seqN ... \n >seqN ... \n
151 loop_timeout => 20, ## Default 0
153 # Added in 1.7, enables the manager process to timeout of a read
155 # operation on channel 0 (UNIX platforms only). The manager process
156 # decrements the total workers running for any worker which have
157 # died in an uncontrollable manner. Specify this option if on
158 # occassion a worker dies unexpectedly (i.e. from an XS module).
159 # Option works with init_relay on UNIX platforms since MCE 1.844.
161 # A number smaller than 5 is silently increased to 5.
162 max_retries => 2, ## Default 0
164 # This option, added in 1.7, causes MCE to retry a failed
166 # chunk from a worker dying while processing input data or
167 # sequence of numbers.
168 parallel_io => 1, ## Default 0
170 posix_exit => 1, ## Default 0
171 use_slurpio => 1, ## Default 0
172 # The parallel_io option enables parallel reads during large
174 # slurpio, useful when reading from fast storage. Do not enable
175 # parallel_io when running MCE on many nodes with input coming
176 # from shared storage.
177 # Set posix_exit to avoid all END and destructor processing.
179 # Constructing MCE inside a thread implies 1 or if present CGI,
180 # FCGI, Coro, Curses, Gearman::Util, Gearman::XS, LWP::UserAgent,
181 # Mojo::IOLoop, STFL, Tk, Wx, or Win32::GUI.
182 # Enable slurpio to pass the raw chunk (scalar ref) to the user
184 # function when reading input files.
185 use_threads => 1, ## Auto 0 or 1
187 # By default MCE spawns child processes on UNIX platforms and
189 # threads on Windows (i.e. $^O eq 'MSWin32').
190 # MCE supports threads via two threading libraries if threads
192 # is preferred over child processes. The use of threads requires
193 # a thread library prior to loading MCE, causing the use_threads
194 # option to default to 1. Specify 0 for child processes.
195 #
196 # use threads; use forks;
197 # use threads::shared; use forks::shared;
198 # use MCE (or) use MCE; (or) use MCE;
199 spawn_delay => 0.045, ## Default undef
201 submit_delay => 0.015, ## Default undef
202 job_delay => 0.060, ## Default undef
203 # Time to wait in fractional seconds after spawning a worker,
205 # after submitting parameters to worker (MCE->run, MCE->process),
206 # and worker running (one time staggered delay).
207 # Specify job_delay to stagger workers connecting to a database.
209 on_post_exit => \&on_post_exit, ## Default undef
211 on_post_run => \&on_post_run, ## Default undef
212 # Execute the code block after a worker exits or dies.
214 # (i.e. MCE->exit, exit, die)
215 # Execute the code block after running.
217 # (i.e. MCE->process, MCE->run)
218 progress => sub { ... }, ## Default undef
220 # A code block for receiving info on the progress made.
222 # See section labeled "MCE PROGRESS DEMONSTRATIONS" at the
223 # end of this document.
224 user_args => { env => 'test' }, ## Default undef
226 # MCE release 1.4 added a new parameter to allow one to
228 # specify arbitrary arguments such as a string, an ARRAY
229 # or HASH reference. Workers can access this directly.
230 # (i.e. my $args = $mce->{user_args} or MCE->user_args)
231 user_begin => \&user_begin, ## Default undef
233 user_func => \&user_func, ## Default undef
234 user_end => \&user_end, ## Default undef
235 # Think of user_begin, user_func, and user_end as in
237 # the awk scripting language:
238 # awk 'BEGIN { begin } { func } { func } ... END { end }'
239 # MCE workers call user_begin once at the start of a job,
241 # then user_func repeatedly until no chunks remain.
242 # Afterwards, user_end is called.
243 user_error => \&user_error, ## Default undef
245 user_output => \&user_output, ## Default undef
246 # MCE will forward data to user_error/user_output,
248 # when defined, for the following methods.
249 # MCE->sendto(\*STDERR, "sent to user_error\n");
251 # MCE->printf(\*STDERR, "%s\n", "sent to user_error");
252 # MCE->print(\*STDERR, "sent to user_error\n");
253 # MCE->say(\*STDERR, "sent to user_error");
254 # MCE->sendto(\*STDOUT, "sent to user_output\n");
256 # MCE->printf("%s\n", "sent to user_output");
257 # MCE->print("sent to user_output\n");
258 # MCE->say("sent to user_output");
259 stderr_file => 'err_file', ## Default STDERR
261 stdout_file => 'out_file', ## Default STDOUT
262 # Or to file; user_error and user_output take precedence.
264 flush_file => 0, ## Default 1
266 flush_stderr => 0, ## Default 1
267 flush_stdout => 0, ## Default 1
268 # Flush sendto file, standard error, or standard output.
270 interval => {
272 delay => 0.007 [, max_nodes => 4, node_id => 1 ]
273 },
274 # For use with the yield method introduced in MCE 1.5.
276 # Both max_nodes & node_id are optional and default to 1.
277 # Delay is the amount of time between intervals.
278 # interval => 0.007 ## Shorter; MCE 1.506+
280 sequence => { ## Default undef
282 begin => -1, end => 1 [, step => 0.1 [, format => "%4.1f" ] ]
283 },
284 bounds_only => 1, ## Default undef
286 # For looping through a sequence of numbers in parallel.
288 # STEP, if omitted, defaults to 1 if BEGIN is smaller than
289 # END or -1 if BEGIN is greater than END. The FORMAT string
290 # is passed to sprintf behind the scene (% may be omitted).
291 # e.g. $seq_n_formatted = sprintf("%4.1f", $seq_n);
292 # Do not specify both options; input_data and sequence.
294 # Release 1.4 allows one to specify an array reference.
295 # e.g. sequence => [ -1, 1, 0.1, "%4.1f" ]
296 # The bounds_only => 1 option will compute the 'begin' and
298 # 'end' items only for the chunk and not the items in between
299 # (hence boundaries only). This option has no effect when
300 # sequence is not specified or chunk_size equals 1.
301 # my $begin = $chunk_ref->[0]; my $end = $chunk_ref->[1];
303 task_end => \&task_end, ## Default undef
305 # This is called by the manager process after the task
307 # has completed processing. MCE 1.5 allows this option
308 # to be specified at the top level.
309 task_name => 'string', ## Default 'MCE'
311 # Added in MCE 1.5 and mainly beneficial for user_tasks.
313 # One may specify a unique name per each sub-task.
314 # The string is passed as the 3rd arg to task_end.
315 user_tasks => [ ## Default undef
317 { ... }, ## Options for task 0
318 { ... }, ## Options for task 1
319 { ... }, ## Options for task 2
320 ],
321 # Takes a list of hash references, each allowing up to 17
323 # options. All other MCE options are ignored. The init_relay,
324 # input_data, RS, and use_slurpio options are applicable to
325 # the first task only.
326 # max_workers, chunk_size, input_data, interval, sequence,
328 # bounds_only, user_args, user_begin, user_end, user_func,
329 # gather, task_end, task_name, use_slurpio, use_threads,
330 # init_relay, RS
331 # Options not specified here will default to same option
333 # specified at the top level.
334 );
335 EXPORT_CONST, CONST
337 There are 3 constants which are exportable. Using the constants in lieu
338 of 0,1,2 makes it more legible when accessing the user_func arguments
339 directly.
340 SELF CHUNK CID - MCE CONSTANTS
342 Exports SELF => 0, CHUNK => 1, and CID => 2.
344 use MCE export_const => 1;
346 use MCE const => 1; ## Shorter; MCE 1.415+
347 user_func => sub {
349 # my ($mce, $chunk_ref, $chunk_id) = @_;
350 print "Hello from ", $_[SELF]->wid, "\n";
351 }
352 MCE 1.5 allows all public method to be called directly.
354 use MCE;
356 user_func => sub {
358 # my ($mce, $chunk_ref, $chunk_id) = @_;
359 print "Hello from ", MCE->wid, "\n";
360 }
361 OVERRIDING DEFAULTS
363 The following list options which may be overridden when loading the
364 module.
365 use Sereal qw( encode_sereal decode_sereal );
367 use CBOR::XS qw( encode_cbor decode_cbor );
368 use JSON::XS qw( encode_json decode_json );
369 use MCE
371 max_workers => 4, ## Default 1
372 chunk_size => 100, ## Default 1
373 tmp_dir => "/path/to/app/tmp", ## $MCE::Signal::tmp_dir
374 freeze => \&encode_sereal, ## \&Storable::freeze
375 thaw => \&decode_sereal ## \&Storable::thaw
376 ;
377 my $mce = MCE->new( ... );
379 From MCE 1.8 onwards, Sereal 3.015+ is loaded automatically if
381 available. Specify "Sereal => 0" to use Storable instead.
382 use MCE Sereal => 0;
384 RUNNING
386 Run calls spawn, submits the job; workers call user_begin, user_func,
387 and user_end. Run shuts down workers afterwards. Call spawn whenever
388 the need arises for large data structures prior to running.
389 $mce->spawn; ## Call early if desired
391 $mce->run; ## Call run or process below
393 ## Acquire data arrays and/or input_files. Workers persist after
395 ## processing.
396 $mce->process(\@input_data_1); ## Process array
398 $mce->process(\@input_data_2);
399 $mce->process(\@input_data_n);
400 $mce->process(\%input_hash_1); ## Process hash, current API
402 $mce->process(\%input_hash_2); ## available since 1.828
403 $mce->process(\%input_hash_n);
404 $mce->process('input_file_1'); ## Process file
406 $mce->process('input_file_2');
407 $mce->process('input_file_n');
408 $mce->shutdown; ## Shutdown workers
410 SYNTAX for ON_POST_EXIT
412 Often times, one may want to capture the exit status. The on_post_exit
413 option, if defined, is executed immediately by the manager process
414 after a worker exits via exit (children only), MCE->exit (children and
415 threads), or die.
416 The format of $e->{pid} is PID_123 for children and THR_123 for
418 threads.
419 my $restart_flag = 1;
421 sub on_post_exit {
423 my ($mce, $e) = @_;
424 ## Display all possible hash elements.
426 print "$e->{wid}: $e->{pid}: $e->{status}: $e->{msg}: $e->{id}\n";
427 ## Restart this worker if desired.
429 if ($restart_flag && $e->{wid} == 2) {
430 $mce->restart_worker;
431 $restart_flag = 0;
432 }
433 }
434 sub user_func {
436 my ($mce) = @_;
437 MCE->exit(0, 'msg_foo', 1000 + MCE->wid); ## Args, not necessary
438 }
439 my $mce = MCE->new(
441 on_post_exit => \&on_post_exit,
442 user_func => \&user_func,
443 max_workers => 3
444 );
445 $mce->run;
447 -- Output (child processes)
449 2: PID_33223: 0: msg_foo: 1002
451 1: PID_33222: 0: msg_foo: 1001
452 3: PID_33224: 0: msg_foo: 1003
453 2: PID_33225: 0: msg_foo: 1002
454 -- Output (running with threads)
456 3: TID_3: 0: msg_foo: 1003
458 2: TID_2: 0: msg_foo: 1002
459 1: TID_1: 0: msg_foo: 1001
460 2: TID_4: 0: msg_foo: 1002
461 SYNTAX for ON_POST_RUN
463 The on_post_run option, if defined, is executed immediately by the
464 manager process after running MCE->process or MCE->run. This option
465 receives an array reference of hashes.
466 The difference between on_post_exit and on_post_run is that the former
468 is called immediately whereas the latter is called after all workers
469 have completed running.
470 sub on_post_run {
472 my ($mce, $status_ref) = @_;
473 foreach my $e ( @{ $status_ref } ) {
474 ## Display all possible hash elements.
475 print "$e->{wid}: $e->{pid}: $e->{status}: $e->{msg}: $e->{id}\n";
476 }
477 }
478 sub user_func {
480 my ($mce) = @_;
481 MCE->exit(0, 'msg_foo', 1000 + MCE->wid); ## Args, not necessary
482 }
483 my $mce = MCE->new(
485 on_post_run => \&on_post_run,
486 user_func => \&user_func,
487 max_workers => 3
488 );
489 $mce->run;
491 -- Output (child processes)
493 3: PID_33174: 0: msg_foo: 1003
495 1: PID_33172: 0: msg_foo: 1001
496 2: PID_33173: 0: msg_foo: 1002
497 -- Output (running with threads)
499 2: TID_2: 0: msg_foo: 1002
501 3: TID_3: 0: msg_foo: 1003
502 1: TID_1: 0: msg_foo: 1001
503 SYNTAX for INPUT_DATA
505 MCE supports many ways to specify input_data. Support for iterators was
506 added in MCE 1.505. The RS option allows one to specify the record
507 separator when processing files.
508 MCE is a chunking engine. Therefore, chunk_size is applicable to
510 input_data. Specifying 1 for use_slurpio causes user_func to receive a
511 scalar reference containing the raw data (applicable to files only)
512 instead of an array reference.
513 "IO::All" { File, Pipe, STDIO } is supported since MCE 1.845.
515 input_data => '/path/to/file', ## process file
517 input_data => \@array, ## process array
518 input_data => \%hash, ## process hash, API since 1.828
519 input_data => \*FILE_HNDL, ## process file handle
520 input_data => $fh, ## open $fh, "<", "file"
521 input_data => $fh, ## IO::File "file", "r"
522 input_data => $fh, ## IO::Uncompress::Gunzip "file.gz"
523 input_data => $io, ## IO::All { File, Pipe, STDIO }
524 input_data => \$scalar, ## treated like a file
525 input_data => \&iterator, ## user specified iterator
526 chunk_size => 1, ## >1 means looping inside user_func
528 use_slurpio => 1, ## $chunk_ref is a scalar ref
529 RS => "\n>", ## input record separator
530 The chunk_size value determines the chunking mode to use when
532 processing files. Otherwise, chunk_size is the number of elements for
533 arrays. For files, a chunk size value of <= 8192 is how many records to
534 read. Greater than 8192 is how many bytes to read. MCE appends (the
535 rest) up to the next record separator.
536 chunk_size => 8192, ## Consists of 8192 records
538 chunk_size => 8193, ## Approximate 8193 bytes for files
539 chunk_size => 1, ## Consists of 1 record or element
541 chunk_size => 1000, ## Consists of 1000 records
542 chunk_size => '16k', ## Approximate 16 kibiBytes (KiB)
543 chunk_size => '20m', ## Approximate 20 mebiBytes (MiB)
544 The construction for user_func when chunk_size > 1 and assuming
546 use_slurpio equals 0.
547 user_func => sub {
549 my ($mce, $chunk_ref, $chunk_id) = @_;
550 ## $_ is $chunk_ref->[0] when chunk_size equals 1
552 ## $_ is $chunk_ref otherwise; $_ can be used below
553 for my $record ( @{ $chunk_ref } ) {
555 print "$chunk_id: $record\n";
556 }
557 }
558 # input_data => \%hash
560 # current API available since 1.828
561 user_func => sub {
563 my ($mce, $chunk_ref, $chunk_id) = @_;
564 ## $_ points to $chunk_ref regardless of chunk_size
566 for my $key ( keys %{ $chunk_ref } ) {
568 print "$key: ", $chunk_ref->{$key}, "\n";
569 }
570 }
571 Specifying a value for input_data is straight forward for arrays and
573 files. The next several examples specify an iterator reference for
574 input_data.
575 use MCE;
577 ## A factory function which creates a closure (the iterator itself)
579 ## for generating a sequence of numbers. The external variables
580 ## ($n, $max, $step) are used for keeping state across successive
581 ## calls to the closure. The iterator simply returns when $n > max.
582 sub input_iterator {
584 my ($n, $max, $step) = @_;
585 return sub {
587 return if $n > $max;
588 my $current = $n;
590 $n += $step;
591 return $current;
593 };
594 }
595 ## Run user_func in parallel. Input data can be specified during
597 ## the construction or as an argument to the process method.
598 my $mce = MCE->new(
600 # input_data => input_iterator(10, 30, 2),
602 chunk_size => 1, max_workers => 4,
603 user_func => sub {
605 my ($mce, $chunk_ref, $chunk_id) = @_;
606 MCE->print("$_: ", $_ * 2, "\n");
607 }
608 )->spawn;
610 $mce->process( input_iterator(10, 30, 2) );
612 -- Output Note that output order is not guaranteed
614 Take a look at iterator.pl for ordered output
615 10: 20
617 12: 24
618 16: 32
619 20: 40
620 14: 28
621 22: 44
622 18: 36
623 24: 48
624 26: 52
625 28: 56
626 30: 60
627 The following example queries the DB for the next 1000 rows. Notice the
629 use of fetchall_arrayref. The iterator function itself receives one
630 argument which is chunk_size (added in MCE 1.510) to determine how much
631 to return per iteration. The default is 1 for the Core API and MCE
632 Models.
633 use DBI;
635 use MCE;
636 sub db_iter {
638 my $dsn = "DBI:Oracle:host=db_server;port=db_port;sid=db_name";
640 my $dbh = DBI->connect($dsn, 'db_user', 'db_passwd') ||
642 die "Could not connect to database: $DBI::errstr";
643 my $sth = $dbh->prepare('select color, desc from table');
645 $sth->execute;
647 return sub {
649 my ($chunk_size) = @_;
650 if (my $aref = $sth->fetchall_arrayref(undef, $chunk_size)) {
652 return @{ $aref };
653 }
654 return;
656 };
657 }
658 ## Let's enumerate column indexes for easy column retrieval.
660 my ($i_color, $i_desc) = (0 .. 1);
661 my $mce = MCE->new(
663 max_workers => 3, chunk_size => 1000,
664 input_data => db_iter(),
665 user_func => sub {
667 my ($mce, $chunk_ref, $chunk_id) = @_;
668 my $ret = '';
669 foreach my $row (@{ $chunk_ref }) {
671 $ret .= $row->[$i_color] .": ". $row->[$i_desc] ."\n";
672 }
673 MCE->print($ret);
675 }
676 );
677 $mce->run;
679 There are many modules on CPAN which return an iterator reference.
681 Showing one such example below. The demonstration ensures MCE workers
682 are spawned before obtaining the iterator. Note the worker_id value
683 (left column) in the output.
684 use Path::Iterator::Rule;
686 use MCE;
687 my $start_dir = shift
689 or die "Please specify a starting directory";
690 -d $start_dir
692 or die "Cannot open ($start_dir): No such file or directory";
693 my $mce = MCE->new(
695 max_workers => 'auto',
696 user_func => sub { MCE->say( MCE->wid . ": $_" ) }
697 )->spawn;
698 my $rule = Path::Iterator::Rule->new->file->name( qr/[.](pm)$/ );
700 my $iterator = $rule->iter(
702 $start_dir, { follow_symlinks => 0, depthfirst => 1 }
703 );
704 $mce->process( $iterator );
706 -- Output
708 8: lib/MCE/Core/Input/Generator.pm
710 5: lib/MCE/Core/Input/Handle.pm
711 6: lib/MCE/Core/Input/Iterator.pm
712 2: lib/MCE/Core/Input/Request.pm
713 3: lib/MCE/Core/Manager.pm
714 4: lib/MCE/Core/Input/Sequence.pm
715 7: lib/MCE/Core/Validation.pm
716 1: lib/MCE/Core/Worker.pm
717 8: lib/MCE/Flow.pm
718 5: lib/MCE/Grep.pm
719 6: lib/MCE/Loop.pm
720 2: lib/MCE/Map.pm
721 3: lib/MCE/Queue.pm
722 4: lib/MCE/Signal.pm
723 7: lib/MCE/Stream.pm
724 1: lib/MCE/Subs.pm
725 8: lib/MCE/Util.pm
726 5: lib/MCE.pm
727 Although MCE supports arrays, extra measures are needed to use a "lazy"
729 array as input data. The reason for this is that MCE needs the size of
730 the array before processing which may be unknown for lazy arrays.
731 Therefore, closures provides an excellent mechanism for this.
732 The code block belonging to the lazy array must return undef after
734 exhausting its input data. Otherwise, the process will never end.
735 use Tie::Array::Lazy;
737 use MCE;
738 tie my @a, 'Tie::Array::Lazy', [], sub {
740 my $i = $_[0]->index;
741 return ($i < 10) ? $i : undef;
743 };
744 sub make_iterator {
746 my $i = 0; my $a_ref = shift;
747 return sub {
749 return $a_ref->[$i++];
750 };
751 }
752 my $mce = MCE->new(
754 max_workers => 4, input_data => make_iterator(\@a),
755 user_func => sub {
757 my ($mce, $chunk_ref, $chunk_id) = @_;
758 MCE->say($_);
759 }
760 )->run;
762 -- Output
764 0
766 1
767 2
768 3
769 4
770 6
771 7
772 8
773 5
774 9
775 The following demonstrates how to retrieve a chunk from the lazy array
777 per each successive call. Here, undef is sent by the iterator block
778 when $i is greater than $max. Iterators may optionally use chunk_size
779 to determine how much to return per iteration.
780 use Tie::Array::Lazy;
782 use MCE;
783 tie my @a, 'Tie::Array::Lazy', [], sub {
785 $_[0]->index;
786 };
787 sub make_iterator {
789 my $j = 0; my ($a_ref, $max) = @_;
790 return sub {
792 my ($chunk_size) = @_;
793 my $i = $j; $j += $chunk_size;
794 return if $i > $max;
796 return $j <= $max ? @$a_ref[$i .. $j - 1] : @$a_ref[$i .. $max];
797 };
798 }
799 my $mce = MCE->new(
801 chunk_size => 15, max_workers => 4,
802 input_data => make_iterator(\@a, 100),
803 user_func => sub {
805 my ($mce, $chunk_ref, $chunk_id) = @_;
806 MCE->say("$chunk_id: " . join(' ', @{ $chunk_ref }));
807 }
808 )->run;
810 -- Output
812 1: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
814 2: 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29
815 3: 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44
816 4: 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59
817 5: 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74
818 6: 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89
819 7: 90 91 92 93 94 95 96 97 98 99 100
820 SYNTAX for SEQUENCE
822 The 1.3 release and above allows workers to loop through a sequence of
823 numbers computed mathematically without the overhead of an array. The
824 sequence can be specified separately per each user_task entry unlike
825 input_data which is applicable to the first task only.
826 See the seq_demo.pl example, included with this distribution, on
828 applying sequences with the user_tasks option.
829 Sequence can be defined using an array or a hash reference.
831 use MCE;
833 my $mce = MCE->new(
835 max_workers => 3,
836 # sequence => [ 10, 19, 0.7, "%4.1f" ], ## up to 4 options
838 sequence => {
840 begin => 10, end => 19, step => 0.7, format => "%4.1f"
841 },
842 user_func => sub {
844 my ($mce, $n, $chunk_id) = @_;
845 print $n, " from ", MCE->wid, " id ", $chunk_id, "\n";
846 }
847 );
848 $mce->run;
850 -- Output (sorted afterwards, notice wid and chunk_id in output)
852 10.0 from 1 id 1
854 10.7 from 2 id 2
855 11.4 from 3 id 3
856 12.1 from 1 id 4
857 12.8 from 2 id 5
858 13.5 from 3 id 6
859 14.2 from 1 id 7
860 14.9 from 2 id 8
861 15.6 from 3 id 9
862 16.3 from 1 id 10
863 17.0 from 2 id 11
864 17.7 from 3 id 12
865 18.4 from 1 id 13
866 The 1.5 release includes a new option (bounds_only). This option tells
868 the sequence engine to compute 'begin' and 'end' items only, for the
869 chunk, and not the items in between (hence boundaries only). This
870 option applies to sequence only and has no effect when chunk_size
871 equals 1.
872 The time to run is 0.006s below. This becomes 0.827s without the
874 bounds_only option due to computing all items in between, thus creating
875 a very large array. Basically, specify bounds_only => 1 when boundaries
876 is all you need for looping inside the block; e.g. Monte Carlo
877 simulations.
878 Time was measured using 1 worker to emphasize the difference.
880 use MCE;
882 my $mce = MCE->new(
884 max_workers => 1, chunk_size => 1_250_000,
885 sequence => { begin => 1, end => 10_000_000 },
887 bounds_only => 1,
888 ## For sequence, the input scalar $_ points to $chunk_ref
890 ## when chunk_size > 1, otherwise $chunk_ref->[0].
891 ##
892 ## user_func => sub {
893 ## my $begin = $_->[0]; my $end = $_->[-1];
894 ##
895 ## for ($begin .. $end) {
896 ## ...
897 ## }
898 ## },
899 user_func => sub {
901 my ($mce, $chunk_ref, $chunk_id) = @_;
902 ## $chunk_ref contains 2 items, not 1_250_000
903 my $begin = $chunk_ref->[ 0];
905 my $end = $chunk_ref->[-1]; ## or $chunk_ref->[1]
906 MCE->printf("%7d .. %8d\n", $begin, $end);
908 }
909 );
910 $mce->run;
912 -- Output
914 1 .. 1250000
916 1250001 .. 2500000
917 2500001 .. 3750000
918 3750001 .. 5000000
919 5000001 .. 6250000
920 6250001 .. 7500000
921 7500001 .. 8750000
922 8750001 .. 10000000
923 SYNTAX for MAX_RETRIES
925 The max_retries option, added in 1.7, allows MCE to retry a failed
926 chunk from a worker dying while processing input data or a sequence of
927 numbers.
928 When max_retries is set, MCE configures the on_post_exit option
930 automatically using the following code before running. Specify
931 on_post_exit explicitly for any further tailoring. The restart_worker
932 line is necessary, obviously.
933 on_post_exit => sub {
935 my ( $mce, $e, $retry_cnt ) = @_;
936 if ( $e->{id} ) {
938 my $cnt = $retry_cnt + 1;
939 my $msg = "Error: chunk $e->{id} failed";
940 if ( defined $mce->{init_relay} ) {
942 print {*STDERR} "$msg, retrying chunk attempt # $cnt\n"
943 if ( $retry_cnt < $mce->{max_retries} );
944 }
945 else {
946 ( $retry_cnt < $mce->{max_retries} )
947 ? print {*STDERR} "$msg, retrying chunk attempt # $cnt\n"
948 : print {*STDERR} "$msg\n";
949 }
950 $mce->restart_worker;
952 }
953 }
954 We let MCE handle on_post_exit automatically below, which is
956 essentially the same code shown above. For max_retries to work, the
957 worker must die, abnormally included, or call MCE->exit. Notice that we
958 pass the chunk_id value for the 3rd argument to MCE->exit (defaults to
959 chunk_id if omitted since MCE 1.844).
960 ## max_retries demonstration
962 use strict;
964 use warnings;
965 use MCE;
967 sub user_func {
969 my ( $mce, $chunk_ref, $chunk_id ) = @_;
970 # die "Died : chunk_id = 3\n" if $chunk_id == 3;
972 MCE->exit(1, undef, $chunk_id) if $chunk_id == 3;
973 print "$chunk_id\n";
975 }
976 my $mce = MCE->new(
978 max_workers => 1,
979 max_retries => 2,
980 user_func => \&user_func,
981 )->spawn;
982 my $input_data = [ 0..7 ];
984 $mce->process( { chunk_size => 1 }, $input_data );
986 $mce->shutdown;
987 -- Output
989 1
991 2
992 Error: chunk 3 failed, retrying chunk attempt # 1
993 Error: chunk 3 failed, retrying chunk attempt # 2
994 Error: chunk 3 failed
995 4
996 5
997 6
998 7
999 8
1000 Orderly output with max_retries is possible since MCE 1.844. Below,
1002 chunk 3 succeeds whereas chunk 5 fails due to exceeding the number of
1003 retries. Be sure to call MCE::relay inside "user_func" and near the end
1004 of the block.
1005 ## max_retries demonstration with init_relay
1007 use strict;
1009 use warnings;
1010 use MCE;
1012 use MCE::Shared;
1013 tie my $retries1, 'MCE::Shared', 0;
1015 tie my $retries2, 'MCE::Shared', 0;
1016 MCE->new(
1018 max_workers => 4,
1019 input_data => [ 1..7 ],
1020 chunk_size => 1,
1021 max_retries => 2,
1023 init_relay => 0,
1024 user_func => sub {
1026 if ( MCE->chunk_id == 3 ) {
1027 MCE->exit if ++$retries1 <= 2;
1028 }
1029 if ( MCE->chunk_id == 5 ) {
1030 MCE->exit if ++$retries2 <= 3;
1031 }
1032 MCE::relay {
1033 $_ += 1;
1034 print MCE->chunk_id, "\n";
1035 };
1036 }
1037 )->run;
1038 print "final: ", MCE::relay_final(), "\n";
1040 -- Output
1042 1
1044 2
1045 Error: chunk 3 failed, retrying chunk attempt # 1
1046 Error: chunk 5 failed, retrying chunk attempt # 1
1047 Error: chunk 3 failed, retrying chunk attempt # 2
1048 Error: chunk 5 failed, retrying chunk attempt # 2
1049 3
1050 4
1051 Error: chunk 5 failed
1052 6
1053 7
1054 final: 6
1055 SYNTAX for USER_BEGIN and USER_END
1057 The user_begin and user_end options, if specified, behave similarly to
1058 awk 'BEGIN { begin } { func } { func } ... END { end }'. These are
1059 called once per worker during each run.
1060 MCE 1.510 passes 2 additional parameters ($task_id and $task_name).
1062 sub user_begin { ## Called once at the beginning
1064 my ($mce, $task_id, $task_name) = @_;
1065 $mce->{wk_total_rows} = 0;
1066 }
1067 sub user_func { ## Called while processing
1069 my $mce = shift;
1070 $mce->{wk_total_rows} += 1;
1071 }
1072 sub user_end { ## Called once at the end
1074 my ($mce, $task_id, $task_name) = @_;
1075 printf "## %d: Processed %d rows\n",
1076 MCE->wid, $mce->{wk_total_rows};
1077 }
1078 my $mce = MCE->new(
1080 user_begin => \&user_begin,
1081 user_func => \&user_func,
1082 user_end => \&user_end
1083 );
1084 $mce->run;
1086 SYNTAX for USER_FUNC with USE_SLURPIO => 0
1088 When processing input data, MCE can pass an array of rows or a slurped
1089 chunk. Below, a reference to an array containing the chunk data is
1090 processed.
1091 e.g. $chunk_ref = [ record1, record2, record3, ... ]
1093 sub user_func {
1095 my ($mce, $chunk_ref, $chunk_id) = @_;
1097 foreach my $row ( @{ $chunk_ref } ) {
1099 $mce->{wk_total_rows} += 1;
1100 print $row;
1101 }
1102 }
1103 my $mce = MCE->new(
1105 chunk_size => 100,
1106 input_data => "/path/to/file",
1107 user_func => \&user_func,
1108 use_slurpio => 0
1109 );
1110 $mce->run;
1112 SYNTAX for USER_FUNC with USE_SLURPIO => 1
1114 Here, a reference to a scalar containing the raw chunk data is
1115 processed.
1116 sub user_func {
1118 my ($mce, $chunk_ref, $chunk_id) = @_;
1120 my $count = () = $$chunk_ref =~ /abc/;
1122 }
1123 my $mce = MCE->new(
1125 chunk_size => 16000,
1126 input_data => "/path/to/file",
1127 user_func => \&user_func,
1128 use_slurpio => 1
1129 );
1130 $mce->run;
1132 SYNTAX for USER_ERROR and USER_OUTPUT
1134 Output from MCE->sendto('STDERR/STDOUT', ...), MCE->printf, MCE->print,
1135 and MCE->say can be intercepted by specifying the user_error and
1136 user_output options. MCE on receiving output will forward to user_error
1137 or user_output in a serialized fashion.
1138 Handy when wanting to filter, modify, and/or direct the output
1140 elsewhere.
1141 sub user_error { ## Redirect STDERR to STDOUT
1143 my $error = shift;
1144 print {*STDOUT} $error;
1145 }
1146 sub user_output { ## Redirect STDOUT to STDERR
1148 my $output = shift;
1149 print {*STDERR} $output;
1150 }
1151 sub user_func {
1153 my ($mce, $chunk_ref, $chunk_id) = @_;
1154 my $count = 0;
1155 foreach my $row ( @{ $chunk_ref } ) {
1157 MCE->print($row);
1158 $count += 1;
1159 }
1160 MCE->print(\*STDERR, "$chunk_id: processed $count rows\n");
1162 }
1163 my $mce = MCE->new(
1165 chunk_size => 1000,
1166 input_data => "/path/to/file",
1167 user_error => \&user_error,
1168 user_output => \&user_output,
1169 user_func => \&user_func
1170 );
1171 $mce->run;
1173 SYNTAX for USER_TASKS and TASK_END
1175 This option takes an array of tasks. Each task allows up to 17 options.
1176 The init_relay, input_data, RS, and use_slurpio options may be defined
1177 inside the first task or at the top level, otherwise ignored under
1178 other sub-tasks.
1179 max_workers, chunk_size, input_data, interval, sequence,
1181 bounds_only, user_args, user_begin, user_end, user_func,
1182 gather, task_end, task_name, use_slurpio, use_threads,
1183 init_relay, RS
1184 Sequence and chunk_size were added in 1.3. User_args was introduced in
1186 1.4. Name and input_data are new options allowed in 1.5. In addition,
1187 one can specify task_end at the top level. Task_end also receives 2
1188 additional arguments $task_id and $task_name (shown below).
1189 Options not specified here will default to the same option specified at
1191 the top level. The task_end option is called by the manager process
1192 when all workers for that sub-task have completed processing.
1193 Forking and threading can be intermixed among tasks unless running
1195 Cygwin. The run method will continue running until all workers have
1196 completed processing.
1197 use threads;
1199 use threads::shared;
1200 use MCE;
1202 sub parallel_task1 { sleep 2; }
1204 sub parallel_task2 { sleep 1; }
1205 my $mce = MCE->new(
1207 task_end => sub {
1209 my ($mce, $task_id, $task_name) = @_;
1210 print "Task [$task_id -- $task_name] completed processing\n";
1211 },
1212 user_tasks => [{
1214 task_name => 'foo',
1215 max_workers => 2,
1216 user_func => \¶llel_task1,
1217 use_threads => 0 ## Not using threads
1218 },{
1220 task_name => 'bar',
1221 max_workers => 4,
1222 user_func => \¶llel_task2,
1223 use_threads => 1 ## Yes, threads
1224 }]
1226 );
1227 $mce->run;
1229 -- Output
1231 Task [1 -- bar] completed processing
1233 Task [0 -- foo] completed processing
1234
1236 Beginning with MCE 1.5, the input scalar $_ is localized prior to
1237 calling user_func for input_data and sequence of numbers. The following
1238 applies.
1239 use_slurpio => 1
1241 $_ is a reference to the buffer e.g. $_ = \$_buffer;
1242 $_ is a reference regardless of whether chunk_size is 1 or greater
1243 user_func => sub {
1245 # my ($mce, $chunk_ref, $chunk_id) = @_;
1246 print ${ $_ }; ## $_ is same as $chunk_ref
1247 }
1248 chunk_size is greater than 1, use_slurpio => 0
1250 $_ is a reference to an array. $_ = \@_records; $_ = \@_seq_n;
1251 $_ is same as $chunk_ref or $_[CHUNK]
1252 user_func => sub {
1254 # my ($mce, $chunk_ref, $chunk_id) = @_;
1255 for my $row ( @{ $_ } ) {
1256 print $row, "\n";
1257 }
1258 }
1259 use MCE const => 1;
1261 user_func => sub {
1263 # my ($mce, $chunk_ref, $chunk_id) = @_;
1264 for my $row ( @{ $_[CHUNK] } ) {
1265 print $row, "\n";
1266 }
1267 }
1268 chunk_size equals 1, use_slurpio => 0
1270 $_ contains the actual value. $_ = $_buffer; $_ = $seq_n;
1271 ## Note that $_ and $chunk_ref are not the same below.
1273 ## $chunk_ref is a reference to an array.
1274 user_func => sub {
1276 # my ($mce, $chunk_ref, $chunk_id) = @_;
1277 print $_, "\n; ## Same as $chunk_ref->[0];
1278 }
1279 $mce->foreach("/path/to/file", sub {
1281 # my ($mce, $chunk_ref, $chunk_id) = @_;
1282 print $_; ## Same as $chunk_ref->[0];
1283 });
1284 ## However, that is not the case for the forseq method.
1286 ## Both $_ and $n_seq are the same when chunk_size => 1.
1287 $mce->forseq([ 1, 9 ], sub {
1289 # my ($mce, $n_seq, $chunk_id) = @_;
1290 print $_, "\n"; ## Same as $n_seq
1291 });
1292 Sequence can also be specified using an array reference. The below
1294 is the same as the example afterwards.
1295 $mce->forseq( { begin => 10, end => 40, step => 2 }, ... );
1297 The code block receives an array containing the next 5 sequences.
1299 Chunk 1 (chunk_id 1) contains 10,12,14,16,18. $n_seq is a reference
1300 to an array, same as $_, due to chunk_size being greater than 1.
1301 $mce->forseq( [ 10, 40000, 2 ], { chunk_size => 5 }, sub {
1303 # my ($mce, $n_seq, $chunk_id) = @_;
1304 my @result;
1305 for my $n ( @{ $_ } ) {
1306 ... do work, append to result for 5
1307 }
1308 ... do something with result afterwards
1309 });
1310
1312 The methods listed below are callable by the main process and workers.
1313 MCE->abort ( void )
1315 $mce->abort ( void )
1316 The 'abort' method is applicable when processing input_data only. This
1317 causes all workers to abort after processing the current chunk.
1318 Workers write the next offset position to the queue socket for the next
1320 available worker. In essence, the 'abort' method writes the last offset
1321 position. Workers, on requesting the next offset position, will think
1322 the end of input_data has been reached and leave the chunking loop.
1323 MCE->abort;
1325 $mce->abort;
1326 MCE->chunk_id ( void )
1328 $mce->chunk_id ( void )
1329 Returns the chunk_id for the current chunk. The value starts at 1.
1330 Chunking applies to input_data or sequence. The value is 0 for the
1331 manager process.
1332 my $chunk_id = MCE->chunk_id;
1334 my $chunk_id = $mce->chunk_id;
1335 MCE->chunk_size ( void )
1337 $mce->chunk_size ( void )
1338 Getter method for chunk_size used by MCE.
1339 my $chunk_size = MCE->chunk_size;
1341 my $chunk_size = $mce->chunk_size;
1342 MCE->do ( 'callback_func' [, $arg1, ... ] )
1344 $mce->do ( 'callback_func' [, $arg1, ... ] )
1345 MCE serializes data transfers from a worker process via helper
1346 functions do & sendto to the manager process. The callback function can
1347 optionally return a reply. Support for calling by the manager process
1348 was enabled in MCE 1.839.
1349 [ $reply = ] MCE->do('callback' [, $arg1, ... ]);
1351 Passing args to a callback function using references & scalar.
1353 sub callback {
1355 my ($array_ref, $hash_ref, $scalar_ref, $scalar) = @_;
1356 ...
1357 }
1358 MCE->do('main::callback', \@a, \%h, \$s, 'foo');
1360 MCE->do('callback', \@a, \%h, \$s, 'foo');
1361 MCE knows if wanting a void, list, hash, or a scalar return value.
1363 MCE->do('callback' [, $arg1, ... ]);
1365 my @array = MCE->do('callback' [, $arg1, ... ]);
1367 my %hash = MCE->do('callback' [, $arg1, ... ]);
1368 my $scalar = MCE->do('callback' [, $arg1, ... ]);
1369 MCE->freeze ( $object_ref )
1371 $mce->freeze ( $object_ref )
1372 Calls the internal freeze method to serialize an object. The default
1373 serialization routines are handled by Sereal if available or Storable.
1374 my $frozen = MCE->freeze([ 0, 2, 4 ]);
1376 my $frozen = $mce->freeze([ 0, 2, 4 ]);
1377 MCE->max_retries ( void )
1379 $mce->max_retries ( void )
1380 Getter method for max_retries used by MCE.
1381 my $max_retries = MCE->max_retries;
1383 my $max_retries = $mce->max_retries;
1384 MCE->max_workers ( void )
1386 $mce->max_workers ( void )
1387 Getter method for max_workers used by MCE.
1388 my $max_workers = MCE->max_workers;
1390 my $max_workers = $mce->max_workers;
1391 MCE->pid ( void )
1393 $mce->pid ( void )
1394 Returns the Process ID. Threads have thread ID attached to the value.
1395 my $pid = MCE->pid; ## 16180 (pid) ; 16180.2 (pid.tid)
1397 my $pid = $mce->pid;
1398 MCE->printf ( $format, $list [, ... ] )
1400 MCE->print ( $list [, ... ] )
1401 MCE->say ( $list [, ... ] )
1402 $mce->printf ( $format, $list [, ... ] )
1403 $mce->print ( $list [, ... ] )
1404 $mce->say ( $list [, ... ] )
1405 Use the printf, print, and say methods when wanting to serialize output
1406 among workers and the manager process. These are sugar syntax for the
1407 sendto method. These behave similar to the native subroutines in Perl
1408 with the exception that barewords must be passed as a reference and
1409 require the comma after it including file handles.
1410 Say is like print, but implicitly appends a newline.
1412 MCE->printf(\*STDOUT, "%s: %d\n", $name, $age);
1414 MCE->printf($fh, "%s: %d\n", $name, $age);
1415 MCE->printf("%s: %d\n", $name, $age);
1416 MCE->print(\*STDERR, "$error_msg\n");
1418 MCE->print($fh, $log_msg."\n");
1419 MCE->print("$output_msg\n");
1420 MCE->say(\*STDERR, $error_msg);
1422 MCE->say($fh, $log_msg);
1423 MCE->say($output_msg);
1424 Caveat: Use the following syntax when passing a reference not a glob or
1426 file handle. Otherwise, MCE will error indicating the first argument is
1427 not a glob reference.
1428 MCE->print(\*STDOUT, \@array, "\n");
1430 MCE->print("", \@array, "\n"); ## ok
1431 Sending to "IO::All" { File, Pipe, STDIO } is supported since MCE
1433 1.845.
1434 use IO::All;
1436 my $out = io->stdout;
1438 my $err = io->stderr;
1439 MCE->printf($out, "%s\n", "sent to stdout");
1441 MCE->printf($err, "%s\n", "sent to stderr");
1442 MCE->print($out, "sent to stdout\n");
1444 MCE->print($err, "sent to stderr\n");
1445 MCE->say($out, "sent to stdout");
1447 MCE->say($err, "sent to stderr");
1448 MCE->sess_dir ( void )
1450 $mce->sess_dir ( void )
1451 Returns the session directory used by the MCE instance. This is defined
1452 during spawning and removed during shutdown.
1453 my $sess_dir = MCE->sess_dir;
1455 my $sess_dir = $mce->sess_dir;
1456 MCE->task_id ( void )
1458 $mce->task_id ( void )
1459 Returns the task ID. This applies to the user_tasks option (starts at
1460 0).
1461 my $task_id = MCE->task_id;
1463 my $task_id = $mce->task_id;
1464 MCE->task_name ( void )
1466 $mce->task_name ( void )
1467 Returns the task_name value specified via the task_name option when
1468 configuring MCE.
1469 my $task_name = MCE->task_name;
1471 my $task_name = $mce->task_name;
1472 MCE->task_wid ( void )
1474 $mce->task_wid ( void )
1475 Returns the task worker ID (applies to user_tasks). The value starts at
1476 1 per each task configured within user_tasks. The value is 0 for the
1477 manager process.
1478 my $task_wid = MCE->task_wid;
1480 my $task_wid = $mce->task_wid;
1481 MCE->thaw ( $frozen )
1483 $mce->thaw ( $frozen )
1484 Calls the internal thaw method to un-serialize the frozen object.
1485 my $object_ref = MCE->thaw($frozen);
1487 my $object_ref = $mce->thaw($frozen);
1488 MCE->tmp_dir ( void )
1490 $mce->tmp_dir ( void )
1491 Returns the temporary directory used by MCE.
1492 my $tmp_dir = MCE->tmp_dir;
1494 my $tmp_dir = $mce->tmp_dir;
1495 MCE->user_args ( void )
1497 $mce->user_args ( void )
1498 Returns the arguments specified via the user_args option.
1499 my ($arg1, $arg2, $arg3) = MCE->user_args;
1501 my ($arg1, $arg2, $arg3) = $mce->user_args;
1502 MCE->wid ( void )
1504 $mce->wid ( void )
1505 Returns the MCE worker ID. Starts at 1 per each MCE instance. The value
1506 is 0 for the manager process.
1507 my $wid = MCE->wid;
1509 my $wid = $mce->wid;
1510
1512 Methods listed below are callable by the main process only.
1513 MCE->forchunk ( $input_data [, { options } ], sub { ... } )
1515 MCE->foreach ( $input_data [, { options } ], sub { ... } )
1516 MCE->forseq ( $sequence_spec [, { options } ], sub { ... } )
1517 $mce->forchunk ( $input_data [, { options } ], sub { ... } )
1518 $mce->foreach ( $input_data [, { options } ], sub { ... } )
1519 $mce->forseq ( $sequence_spec [, { options } ], sub { ... } )
1520 Forchunk, foreach, and forseq are sugar methods and described in
1521 MCE::Candy. Stubs exist in MCE which load MCE::Candy automatically.
1522 MCE->process ( $input_data [, { options } ] )
1524 $mce->process ( $input_data [, { options } ] )
1525 The process method will spawn workers automatically if not already
1526 spawned. It will set input_data => $input_data. It calls run(0) to not
1527 auto-shutdown workers. Specifying options is optional.
1528 Allowable options { key => value, ... } are:
1530 chunk_size input_data job_delay spawn_delay submit_delay
1532 flush_file flush_stderr flush_stdout stderr_file stdout_file
1533 on_post_exit on_post_run sequence user_args user_begin user_end
1534 user_func user_error user_output use_slurpio RS
1535 Options remain persistent going forward unless changed. Setting
1537 user_begin, user_end, or user_func will cause already spawned workers
1538 to shut down and re-spawn automatically. Therefore, define these during
1539 instantiation.
1540 The below will cause workers to re-spawn after running.
1542 my $mce = MCE->new( max_workers => 'auto' );
1544 $mce->process( {
1546 user_begin => sub { ## connect to DB },
1547 user_func => sub { ## process each row },
1548 user_end => sub { ## close handle to DB },
1549 }, \@input_data );
1550 $mce->process( {
1552 user_begin => sub { ## connect to DB },
1553 user_func => sub { ## process each file },
1554 user_end => sub { ## close handle to DB },
1555 }, "/list/of/files" );
1556 Do the following if wanting workers to persist between jobs.
1558 use MCE max_workers => 'auto';
1560 my $mce = MCE->new(
1562 user_begin => sub { ## connect to DB },
1563 user_func => sub { ## process each chunk or row or host },
1564 user_end => sub { ## close handle to DB },
1565 );
1566 $mce->spawn; ## Spawn early if desired
1568 $mce->process("/one/very_big_file/_mce_/will_chunk_in_parallel");
1570 $mce->process(\@array_of_files_to_grep);
1571 $mce->process("/path/to/host/list");
1572 $mce->process($array_ref);
1574 $mce->process($array_ref, { stdout_file => $output_file });
1575 ## This was not allowed before. Fixed in 1.415.
1577 $mce->process({ sequence => { begin => 10, end => 90, step 2 } });
1578 $mce->process({ sequence => [ 10, 90, 2 ] });
1579 $mce->shutdown;
1581 MCE->relay_final ( void )
1583 $mce->relay_final ( void )
1584 The relay methods are described in MCE::Relay. Relay capabilities are
1585 enabled by specifying the "init_relay" MCE option.
1586 MCE->restart_worker ( void )
1588 $mce->restart_worker ( void )
1589 One can restart a worker who has died or exited. The job never ends
1590 below due to restarting each time. Recommended is to call MCE->exit or
1591 $mce->exit instead of the native exit function for better handling,
1592 especially under the Windows environment.
1593 The $e->{wid} argument is no longer necessary starting with the 1.5
1595 release.
1596 Press [ctrl-c] to terminate the script.
1598 my $mce = MCE->new(
1600 on_post_exit => sub {
1602 my ($mce, $e) = @_;
1603 print "$e->{wid}: $e->{pid}: status $e->{status}: $e->{msg}";
1604 # $mce->restart_worker($e->{wid}); ## MCE-1.415 and below
1605 $mce->restart_worker; ## MCE-1.500 and above
1606 },
1607 user_begin => sub {
1609 my ($mce, $task_id, $task_name) = @_;
1610 ## Not interested in die messages going to STDERR,
1611 ## because the die handler calls MCE->exit(255, $_[0]).
1612 close STDERR;
1613 },
1614 user_tasks => [{
1616 max_workers => 5,
1617 user_func => sub {
1618 my ($mce) = @_; sleep MCE->wid;
1619 MCE->exit(3, "exited from " . MCE->wid . "\n");
1620 }
1621 },{
1622 max_workers => 4,
1623 user_func => sub {
1624 my ($mce) = @_; sleep MCE->wid;
1625 die("died from " . MCE->wid . "\n");
1626 }
1627 }]
1628 );
1629 $mce->run;
1631 -- Output
1633 1: PID_85388: status 3: exited from 1
1635 2: PID_85389: status 3: exited from 2
1636 1: PID_85397: status 3: exited from 1
1637 3: PID_85390: status 3: exited from 3
1638 1: PID_85399: status 3: exited from 1
1639 4: PID_85391: status 3: exited from 4
1640 2: PID_85398: status 3: exited from 2
1641 1: PID_85401: status 3: exited from 1
1642 5: PID_85392: status 3: exited from 5
1643 1: PID_85404: status 3: exited from 1
1644 6: PID_85393: status 255: died from 6
1645 3: PID_85400: status 3: exited from 3
1646 2: PID_85403: status 3: exited from 2
1647 1: PID_85406: status 3: exited from 1
1648 7: PID_85394: status 255: died from 7
1649 1: PID_85410: status 3: exited from 1
1650 8: PID_85395: status 255: died from 8
1651 4: PID_85402: status 3: exited from 4
1652 2: PID_85409: status 3: exited from 2
1653 1: PID_85412: status 3: exited from 1
1654 9: PID_85396: status 255: died from 9
1655 3: PID_85408: status 3: exited from 3
1656 1: PID_85416: status 3: exited from 1
1657 ...
1659 MCE->run ( [ $auto_shutdown [, { options } ] ] )
1661 $mce->run ( [ $auto_shutdown [, { options } ] ] )
1662 The run method, by default, spawns workers, processes once, and shuts
1663 down afterwards. Specify 0 for $auto_shutdown when wanting workers to
1664 persist after running (default 1).
1665 Specifying options is optional. Valid options are the same as for the
1667 process method.
1668 my $mce = MCE->new( ... );
1670 ## Disables auto-shutdown
1672 $mce->run(0);
1673 MCE->send ( $data_ref )
1675 $mce->send ( $data_ref )
1676 The 'send' method is useful when wanting to spawn workers early to
1677 minimize memory consumption and afterwards send data individually to
1678 each worker. One cannot send more than the total workers spawned.
1679 Workers store the received data as $mce->{user_data}.
1680 The data which can be sent is restricted to an ARRAY, HASH, or PDL
1682 reference. Workers begin processing immediately after receiving data.
1683 Workers set $mce->{user_data} to undef after processing. One cannot
1684 specify input_data, sequence, or user_tasks when using the "send"
1685 method.
1686 Passing any options e.g. run(0, { options }) is ignored due to workers
1688 running immediately after receiving user data. There is no guarantee to
1689 which worker will receive data first. It depends on which worker is
1690 available awaiting data.
1691 use MCE;
1693 my $mce = MCE->new(
1695 max_workers => 5,
1696 user_func => sub {
1698 my ($mce) = @_;
1699 my $data = $mce->{user_data};
1700 my $first_name = $data->{first_name};
1701 print MCE->wid, ": Hello from $first_name\n";
1702 }
1703 );
1704 $mce->spawn; ## Optional, send will spawn if necessary.
1706 $mce->send( { first_name => "Theresa" } );
1708 $mce->send( { first_name => "Francis" } );
1709 $mce->send( { first_name => "Padre" } );
1710 $mce->send( { first_name => "Anthony" } );
1711 $mce->run; ## Wait for workers to complete processing.
1713 -- Output
1715 2: Hello from Theresa
1717 5: Hello from Anthony
1718 3: Hello from Francis
1719 4: Hello from Padre
1720 MCE->shutdown ( void )
1722 $mce->shutdown ( void )
1723 The run method will automatically spawn workers, run once, and shutdown
1724 workers automatically. Workers persist after running below. Shutdown
1725 may be called as needed or prior to exiting.
1726 my $mce = MCE->new( ... );
1728 $mce->spawn;
1730 $mce->process(\@input_data_1); ## Processing multiple arrays
1732 $mce->process(\@input_data_2);
1733 $mce->process(\@input_data_n);
1734 $mce->shutdown;
1736 $mce->process('input_file_1'); ## Processing multiple files
1738 $mce->process('input_file_2');
1739 $mce->process('input_file_n');
1740 $mce->shutdown;
1742 MCE->spawn ( void )
1744 $mce->spawn ( void )
1745 Workers are normally spawned automatically. The spawn method allows one
1746 to spawn workers early if so desired.
1747 my $mce = MCE->new( ... );
1749 $mce->spawn;
1751 MCE->status ( void )
1753 $mce->status ( void )
1754 The greatest exit status is saved among workers while running. Look at
1755 the on_post_exit or on_post_run options for callback support.
1756 my $mce = MCE->new( ... );
1758 $mce->run;
1760 my $exit_status = $mce->status;
1762
1764 Methods listed below are callable by workers only.
1765 MCE->exit ( [ $status [, $message [, $id ] ] ] )
1767 $mce->exit ( [ $status [, $message [, $id ] ] ] )
1768 A worker exits from MCE entirely. $id (optional) can be used for
1769 passing the primary key or a string along with the message. Look at the
1770 on_post_exit or on_post_run options for callback support.
1771 MCE->exit; ## default 0
1773 MCE->exit(1);
1774 MCE->exit(2, 'chunk failed', $chunk_id);
1775 MCE->exit(0, 'msg_foo', 'id_1000');
1776 MCE->gather ( $arg1, [, $arg2, ... ] )
1778 $mce->gather ( $arg1, [, $arg2, ... ] )
1779 A worker can submit data to the location specified via the gather
1780 option by calling this method. See MCE::Flow and MCE::Loop for
1781 additional use-case.
1782 use MCE;
1784 my @hosts = qw(
1786 hosta hostb hostc hostd hoste
1787 );
1788 my $mce = MCE->new(
1790 chunk_size => 1, max_workers => 3,
1791 user_func => sub {
1793 # my ($mce, $chunk_ref, $chunk_id) = @_;
1794 my ($output, $error, $status); my $host = $_;
1795 ## Do something with $host;
1797 $output = "Worker ". MCE->wid .": Hello from $host";
1798 if (MCE->chunk_id % 3 == 0) {
1800 ## Simulating an error condition
1801 local $? = 1; $status = $?;
1802 $error = "Error from $host"
1803 }
1804 else {
1805 $status = 0;
1806 }
1807 ## Ensure unique keys (key, value) when gathering to a
1809 ## hash.
1810 MCE->gather("$host.out", $output, "$host.sta", $status);
1811 MCE->gather("$host.err", $error) if (defined $error);
1812 }
1813 );
1814 my %h; $mce->process(\@hosts, { gather => \%h });
1816 foreach my $host (@hosts) {
1818 print $h{"$host.out"}, "\n";
1819 print $h{"$host.err"}, "\n" if (exists $h{"$host.err"});
1820 print "Exit status: ", $h{"$host.sta"}, "\n\n";
1821 }
1822 -- Output
1824 Worker 2: Hello from hosta
1826 Exit status: 0
1827 Worker 1: Hello from hostb
1829 Exit status: 0
1830 Worker 3: Hello from hostc
1832 Error from hostc
1833 Exit status: 1
1834 Worker 2: Hello from hostd
1836 Exit status: 0
1837 Worker 1: Hello from hoste
1839 Exit status: 0
1840 MCE->last ( void )
1842 $mce->last ( void )
1843 Worker leaves the chunking loop or user_func block immediately.
1844 Callable from inside foreach, forchunk, forseq, and user_func.
1845 use MCE;
1847 my $mce = MCE->new(
1849 max_workers => 5
1850 );
1851 my @list = (1 .. 80);
1853 $mce->forchunk(\@list, { chunk_size => 2 }, sub {
1855 my ($mce, $chunk_ref, $chunk_id) = @_;
1857 MCE->last if ($chunk_id > 4);
1858 my @output = ();
1860 foreach my $rec ( @{ $chunk_ref } ) {
1862 push @output, $rec, "\n";
1863 }
1864 MCE->print(@output);
1866 });
1867 -- Output (each chunk above consists of 2 elements)
1869 3
1871 4
1872 1
1873 2
1874 7
1875 8
1876 5
1877 6
1878 MCE->next ( void )
1880 $mce->next ( void )
1881 Worker starts the next iteration of the chunking loop. Callable from
1882 inside foreach, forchunk, forseq, and user_func.
1883 use MCE;
1885 my $mce = MCE->new(
1887 max_workers => 5
1888 );
1889 my @list = (1 .. 80);
1891 $mce->forchunk(\@list, { chunk_size => 4 }, sub {
1893 my ($mce, $chunk_ref, $chunk_id) = @_;
1895 MCE->next if ($chunk_id < 20);
1896 my @output = ();
1898 foreach my $rec ( @{ $chunk_ref } ) {
1900 push @output, $rec, "\n";
1901 }
1902 MCE->print(@output);
1904 });
1905 -- Output (each chunk above consists of 4 elements)
1907 77
1909 78
1910 79
1911 80
1912 MCE::relay { code }
1914 MCE->relay ( sub { code } )
1915 MCE->relay_recv ( void )
1916 $mce->relay ( sub { code } )
1917 $mce->relay_recv ( void )
1918 The relay methods are described in MCE::Relay. Relay capabilities are
1919 enabled by specifying the "init_relay" MCE option.
1920 MCE->sendto ( $to, $arg1, ... )
1922 $mce->sendto ( $to, $arg1, ... )
1923 The sendto method is called by workers for serializing data to standard
1924 output, standard error, or end of file. The action is done by the
1925 manager process.
1926 Release 1.00x supported 1 data argument, not more.
1928 MCE->sendto('file', \@array, '/path/to/file');
1930 MCE->sendto('file', \$scalar, '/path/to/file');
1931 MCE->sendto('file', $scalar, '/path/to/file');
1932 MCE->sendto('STDERR', \@array);
1934 MCE->sendto('STDERR', \$scalar);
1935 MCE->sendto('STDERR', $scalar);
1936 MCE->sendto('STDOUT', \@array);
1938 MCE->sendto('STDOUT', \$scalar);
1939 MCE->sendto('STDOUT', $scalar);
1940 Release 1.100 added the ability to pass multiple arguments. Notice the
1942 syntax change for sending to a file. Passing a reference to an array is
1943 no longer necessary.
1944 MCE->sendto('file:/path/to/file', $arg1 [, $arg2, ... ]);
1946 MCE->sendto('STDERR', $arg1 [, $arg2, ... ]);
1947 MCE->sendto('STDOUT', $arg1 [, $arg2, ... ]);
1948 MCE->sendto('STDOUT', @a, "\n", %h, "\n", $s, "\n");
1950 To retain 1.00x compatibility, sendto outputs the content when a single
1952 data reference is specified. Otherwise, the reference for \@array or
1953 \$scalar is shown in 1.500, not the content.
1954 MCE->sendto('STDERR', \@array); ## 1.00x behavior, content
1956 MCE->sendto('STDOUT', \$scalar);
1957 MCE->sendto('file:/path/to/file', \@array);
1958 ## Output matches the print statement
1960 MCE->sendto(\*STDERR, \@array); ## 1.500 behavior, reference
1962 MCE->sendto(\*STDOUT, \$scalar);
1963 MCE->sendto($fh, \@array);
1964 MCE->sendto('STDOUT', \@array, "\n", \$scalar, "\n");
1966 print {*STDOUT} \@array, "\n", \$scalar, "\n";
1967 MCE 1.500 added support for sending to a glob reference, file
1969 descriptor, and file handle.
1970 MCE->sendto(\*STDERR, "foo\n", \@array, \$scalar, "\n");
1972 MCE->sendto('fd:2', "foo\n", \@array, \$scalar, "\n");
1973 MCE->sendto($fh, "foo\n", \@array, \$scalar, "\n");
1974 MCE->sync ( void )
1976 $mce->sync ( void )
1977 A barrier sync operation means any worker must stop at this point until
1978 all workers reach this barrier. Barrier syncing is useful for many
1979 computer algorithms.
1980 Barrier synchronization is supported for task 0 only or omitting
1982 user_tasks. All workers assigned task_id 0 must call sync whenever
1983 barrier syncing.
1984 use MCE;
1986 sub user_func {
1988 my ($mce) = @_;
1990 my $wid = MCE->wid;
1991 MCE->sendto("STDOUT", "a: $wid\n"); ## MCE 1.0+
1993 MCE->sync;
1994 MCE->sendto(\*STDOUT, "b: $wid\n"); ## MCE 1.5+
1996 MCE->sync;
1997 MCE->print("c: $wid\n"); ## MCE 1.5+
1999 MCE->sync;
2000 return;
2002 }
2003 my $mce = MCE->new(
2005 max_workers => 4, user_func => \&user_func
2006 )->run;
2007 -- Output (without barrier synchronization)
2009 a: 1
2011 a: 2
2012 b: 1
2013 b: 2
2014 c: 1
2015 c: 2
2016 a: 3
2017 b: 3
2018 c: 3
2019 a: 4
2020 b: 4
2021 c: 4
2022 -- Output (with barrier synchronization)
2024 a: 1
2026 a: 2
2027 a: 4
2028 a: 3
2029 b: 2
2030 b: 1
2031 b: 3
2032 b: 4
2033 c: 1
2034 c: 4
2035 c: 2
2036 c: 3
2037 Consider the following example. The MCE->sync operation is done inside
2039 a loop along with MCE->do. A stall may occur for workers calling sync
2040 the 2nd or 3rd time while other workers are sending results via MCE->do
2041 or MCE->sendto.
2042 It requires another semaphore lock in MCE to solve this which was not
2044 done in order to keep resources low. Therefore, please keep this in
2045 mind when mixing MCE->sync with MCE->do or output serialization methods
2046 inside a loop.
2047 sub user_func {
2049 my ($mce) = @_;
2051 my @result;
2052 for (1 .. 3) {
2054 ... compute algorithm ...
2055 MCE->sync;
2057 ... compute algorithm ...
2059 MCE->sync;
2061 MCE->do('aggregate_result', \@result); ## or MCE->sendto
2063 MCE->sync; ## The sync operation is also needed here to
2065 ## prevent MCE from stalling.
2066 }
2067 }
2068 MCE->yield ( void )
2070 $mce->yield ( void )
2071 There may be on occasion when the MCE driven app is too fast. The
2072 interval option combined with the yield method, both introduced with
2073 MCE 1.5, allows one to throttle the app. It adds a "grace" factor to
2074 the design.
2075 A use case is an app configured with 100 workers running on a 24
2077 logical way box. Data is polled from a database containing over 2.5
2078 million rows. Workers chunk away at 300 rows per chunk performing SNMP
2079 gets (300 sockets per worker) polling 25 metrics from each device. With
2080 this scenario, the load on the box may rise beyond 90+. In addition,
2081 IP_Tables may reach its contention point causing the entire application
2082 to fail.
2083 The scenario above is solved by simply having workers yield among
2085 themselves in a synchronized fashion. A delay of 0.007 seconds between
2086 intervals is all that's needed. The load on the box will hover between
2087 23 ~ 27 for the duration of the run. Polling completes in under 17
2088 minutes time. This is quite fast considering the app polls 62.5 million
2089 metrics combined. The math equates to 3,676,470 per minute or rather
2090 61,275 per second from a single box.
2091 ## Both max_nodes and node_id are optional (default 1).
2093 interval => {
2095 delay => 0.007, max_nodes => $max_nodes, node_id => $node_id
2096 }
2097 A 4 node setup can poll 10 million devices without the additional
2099 overhead of a distribution agent. The difference between the 4 nodes
2100 are simply node_id and the where clause used to query the database. The
2101 mac addresses are random such that the data divides equally to any
2102 power of 2. The distribution key lies in the mac address itself. In
2103 fact, the 2nd character from the right is sufficient for maximizing on
2104 the power of randomness for equal distribution.
2105 Query NodeID 1: ... AND substr(MAC, -2, 1) IN ('0', '1', '2', '3')
2107 Query NodeID 2: ... AND substr(MAC, -2, 1) IN ('4', '5', '6', '7')
2108 Query NodeID 3: ... AND substr(MAC, -2, 1) IN ('8', '9', 'A', 'B')
2109 Query NodeID 4: ... AND substr(MAC, -2, 1) IN ('C', 'D', 'E', 'F')
2110 Below, the user_tasks is configured to simulate 4 nodes. This
2112 demonstration uses 2 workers to minimize the output size. Input is from
2113 the sequence option.
2114 use Time::HiRes qw(time);
2116 use MCE;
2117 my $d = shift || 0.1;
2119 local $| = 1;
2121 sub create_task {
2123 my ($node_id) = @_;
2125 my $seq_size = 6;
2127 my $seq_start = ($node_id - 1) * $seq_size + 1;
2128 my $seq_end = $seq_start + $seq_size - 1;
2129 return {
2131 max_workers => 2, sequence => [ $seq_start, $seq_end ],
2132 interval => { delay => $d, max_nodes => 4, node_id => $node_id }
2133 };
2134 }
2135 sub user_begin {
2137 my ($mce, $task_id, $task_name) = @_;
2139 ## The yield method causes this worker to wait for its next time
2141 ## interval slot before running. Yield has no effect without the
2142 ## 'interval' option.
2143 ## Yielding is beneficial inside a user_begin block. A use case
2145 ## is staggering database connections among workers in order
2146 ## to not impact the DB server.
2147 MCE->yield;
2149 MCE->printf(
2151 "Node %2d: %0.5f -- Worker %2d: %12s -- Started\n",
2152 MCE->task_id + 1, time, MCE->task_wid, ''
2153 );
2154 return;
2156 }
2157 {
2159 my $prev_time = time;
2160 sub user_func {
2162 my ($mce, $seq_n, $chunk_id) = @_;
2164 ## Yield simply waits for the next time interval.
2166 MCE->yield;
2167 ## Calculate how long this worker has waited.
2169 my $curr_time = time;
2170 my $time_waited = $curr_time - $prev_time;
2171 $prev_time = $curr_time;
2173 MCE->printf(
2175 "Node %2d: %0.5f -- Worker %2d: %12.5f -- Seq_N %3d\n",
2176 MCE->task_id + 1, time, MCE->task_wid, $time_waited, $seq_n
2177 );
2178 return;
2180 }
2181 }
2182 ## Simulate a 4 node environment passing node_id to create_task.
2184 print "Node_ID Current_Time Worker_ID Time_Waited Comment\n";
2186 MCE->new(
2188 user_begin => \&user_begin,
2189 user_func => \&user_func,
2190 user_tasks => [
2192 create_task(1),
2193 create_task(2),
2194 create_task(3),
2195 create_task(4)
2196 ]
2197 )->run;
2199 -- Output (notice Current_Time below, stays 0.10 apart)
2201 Node_ID Current_Time Worker_ID Time_Waited Comment
2203 Node 1: 1374807976.74634 -- Worker 1: -- Started
2204 Node 2: 1374807976.84634 -- Worker 1: -- Started
2205 Node 3: 1374807976.94638 -- Worker 1: -- Started
2206 Node 4: 1374807977.04639 -- Worker 1: -- Started
2207 Node 1: 1374807977.14634 -- Worker 2: -- Started
2208 Node 2: 1374807977.24640 -- Worker 2: -- Started
2209 Node 3: 1374807977.34649 -- Worker 2: -- Started
2210 Node 4: 1374807977.44657 -- Worker 2: -- Started
2211 Node 1: 1374807977.54636 -- Worker 1: 0.90037 -- Seq_N 1
2212 Node 2: 1374807977.64638 -- Worker 1: 1.00040 -- Seq_N 7
2213 Node 3: 1374807977.74642 -- Worker 1: 1.10043 -- Seq_N 13
2214 Node 4: 1374807977.84643 -- Worker 1: 1.20045 -- Seq_N 19
2215 Node 1: 1374807977.94636 -- Worker 2: 1.30037 -- Seq_N 2
2216 Node 2: 1374807978.04638 -- Worker 2: 1.40040 -- Seq_N 8
2217 Node 3: 1374807978.14641 -- Worker 2: 1.50042 -- Seq_N 14
2218 Node 4: 1374807978.24644 -- Worker 2: 1.60045 -- Seq_N 20
2219 Node 1: 1374807978.34628 -- Worker 1: 0.79996 -- Seq_N 3
2220 Node 2: 1374807978.44631 -- Worker 1: 0.79996 -- Seq_N 9
2221 Node 3: 1374807978.54634 -- Worker 1: 0.79996 -- Seq_N 15
2222 Node 4: 1374807978.64636 -- Worker 1: 0.79997 -- Seq_N 21
2223 Node 1: 1374807978.74628 -- Worker 2: 0.79996 -- Seq_N 4
2224 Node 2: 1374807978.84632 -- Worker 2: 0.79997 -- Seq_N 10
2225 Node 3: 1374807978.94634 -- Worker 2: 0.79996 -- Seq_N 16
2226 Node 4: 1374807979.04636 -- Worker 2: 0.79996 -- Seq_N 22
2227 Node 1: 1374807979.14628 -- Worker 1: 0.80001 -- Seq_N 5
2228 Node 2: 1374807979.24631 -- Worker 1: 0.80000 -- Seq_N 11
2229 Node 3: 1374807979.34634 -- Worker 1: 0.80001 -- Seq_N 17
2230 Node 4: 1374807979.44636 -- Worker 1: 0.80000 -- Seq_N 23
2231 Node 1: 1374807979.54628 -- Worker 2: 0.80000 -- Seq_N 6
2232 Node 2: 1374807979.64631 -- Worker 2: 0.80000 -- Seq_N 12
2233 Node 3: 1374807979.74633 -- Worker 2: 0.80000 -- Seq_N 18
2234 Node 4: 1374807979.84636 -- Worker 2: 0.80000 -- Seq_N 24
2235 The interval.pl example above is included with MCE.
2237
2239 The "progress" option takes a code block for receiving info on the
2240 progress made while processing input data; e.g. "input_data" or
2241 "sequence". To make this work, one provides the "progress" option a
2242 closure block like so, passing along the size of the input_data; e.g
2243 "scalar @array" or "-s /path/to/file".
2244 Current API available since 1.813.
2246 A worker, upon completing processing its chunk, notifies the manager-
2248 process with the size of the chunk. That could be the number of rows or
2249 literally the size of the chunk when processing an input file. The
2250 manager-process accumulates the size before calling the code block
2251 associated with the "progress" option.
2252 When running many tasks simultaneously, via "user_tasks", the call is
2254 initiated by workers at level 0 only or rather the first task, not
2255 shown here.
2256 use Time::HiRes 'sleep';
2258 use MCE;
2259 sub make_progress {
2261 my ($total_size) = @_;
2262 return sub {
2263 my ($completed_size) = @_;
2264 printf "%0.1f%%\n", $completed_size / $total_size * 100;
2265 };
2266 }
2267 my @input = (1..150);
2269 MCE->new(
2271 chunk_size => 10,
2272 max_workers => 4,
2273 input_data => \@input,
2274 progress => make_progress( scalar @input ),
2275 user_func => sub { sleep 1.5 }
2276 )->run();
2277 -- Output
2279 6.7%
2281 13.3%
2282 20.0%
2283 26.7%
2284 33.3%
2285 40.0%
2286 46.7%
2287 53.3%
2288 60.0%
2289 66.7%
2290 73.3%
2291 80.0%
2292 86.7%
2293 93.3%
2294 100.0%
2295 Next is the code using MCE::Flow and ProgressBar::Stack to do the same
2297 thing, practically.
2298 use Time::HiRes 'sleep';
2300 use ProgressBar::Stack;
2301 use MCE::Flow;
2302 sub make_progress {
2304 my ($total_size) = @_;
2305 init_progress();
2306 return sub {
2307 my ($completed_size) = @_;
2308 update_progress sprintf("%0.1f", $completed_size / $total_size * 100);
2309 };
2310 }
2311 my @input = (1..150);
2313 MCE::Flow->init(
2315 chunk_size => 10,
2316 max_workers => 4,
2317 progress => make_progress( scalar @input )
2318 );
2319 MCE::Flow->run( sub { sleep 1.5 }, \@input );
2321 MCE::Flow->finish();
2322 print "\n";
2324 -- Output
2326 [################ ] 80.0% ETA: 0:01
2328 For sequence of numbers, using the "sequence" option, one must account
2330 for "step_size", typically set to 1 automatically.
2331 use Time::HiRes 'sleep';
2333 use MCE;
2334 sub make_progress {
2336 my ($total_size) = @_;
2337 return sub {
2338 my ($completed_size) = @_;
2339 printf "%0.1f%%\n", $completed_size / $total_size * 100;
2340 };
2341 }
2342 MCE->new(
2344 chunk_size => 10,
2345 max_workers => 4,
2346 sequence => [ 1, 100, 2 ],
2347 progress => make_progress( int( 100 / 2 + 0.5 ) ),
2348 user_func => sub { sleep 1.5 }
2349 )->run();
2350 -- Output
2352 20.0%
2354 40.0%
2355 60.0%
2356 80.0%
2357 100.0%
2358 Changing "chunk_size" to 1 means workers notify the manager process
2360 more often, thus increasing granularity. Take a look at the output.
2361 2.0%
2363 4.0%
2364 6.0%
2365 8.0%
2366 10.0%
2367 ...
2368 92.0%
2369 94.0%
2370 96.0%
2371 98.0%
2372 100.0%
2373 Here is the same thing using MCE::Flow together with
2375 ProgressBar::Stack.
2376 use Time::HiRes 'sleep';
2378 use ProgressBar::Stack;
2379 use MCE::Flow;
2380 sub make_progress {
2382 my ($total_size) = @_;
2383 init_progress();
2384 return sub {
2385 my ($completed_size) = @_;
2386 update_progress sprintf("%0.1f", $completed_size / $total_size * 100);
2387 };
2388 }
2389 MCE::Flow->init(
2391 chunk_size => 1,
2392 max_workers => 4,
2393 progress => make_progress( int( 100 / 2 + 0.5 ) )
2394 );
2395 MCE::Flow->run_seq( sub { sleep 0.5 }, 1, 100, 2 );
2397 MCE::Flow->finish();
2398 print "\n";
2400 -- Output
2402 [######### ] 48.0% ETA: 0:03
2404 For files and file handles, workers send the actual size of the data
2406 read versus counting rows.
2407 use Time::HiRes 'sleep';
2409 use MCE;
2410 sub make_progress {
2412 my ($total_size) = @_;
2413 return sub {
2414 my ($completed_size) = @_;
2415 printf "%0.1f%%\n", $completed_size / $total_size * 100;
2416 };
2417 }
2418 my $input_file = "/path/to/input_file.txt";
2420 MCE->new(
2422 chunk_size => 5,
2423 max_workers => 4,
2424 input_data => $input_file,
2425 progress => make_progress( -s $input_file ),
2426 user_func => sub { sleep 0.03 }
2427 )->run();
2428 For consistency, here is the example using MCE::Flow, again with
2430 ProgressBar::Stack.
2431 use Time::HiRes 'sleep';
2433 use ProgressBar::Stack;
2434 use MCE::Flow;
2435 sub make_progress {
2437 my ($total_size) = @_;
2438 init_progress();
2439 return sub {
2440 my ($completed_size) = @_;
2441 update_progress sprintf("%0.1f", $completed_size / $total_size * 100);
2442 };
2443 }
2444 my $input_file = "/path/to/input_file.txt";
2446 MCE::Flow->init(
2448 chunk_size => 5,
2449 max_workers => 4,
2450 progress => make_progress( -s $input_file )
2451 );
2452 MCE::Flow->run_file( sub { sleep 0.03 }, $input_file );
2454 MCE::Flow->finish();
2455 The next demonstration processes three arrays consecutively. For this
2457 one, MCE workers persist after running. This needs MCE 1.814 or later
2458 to run. Otherwise, the progress output is not shown in MCE 1.813.
2459 use Time::HiRes 'sleep';
2461 use ProgressBar::Stack;
2462 use MCE;
2463 sub make_progress {
2465 my ($total_size, $message) = @_;
2466 init_progress();
2467 return sub {
2468 my ($completed_size) = @_;
2469 update_progress(
2470 sprintf("%0.1f", $completed_size / $total_size * 100),
2471 $message
2472 );
2473 };
2474 }
2475 my $mce = MCE->new(
2477 chunk_size => 10,
2478 max_workers => 4,
2479 user_func => sub { sleep 0.5 }
2480 )->spawn();
2481 my @a1 = ( 1 .. 200 );
2483 my @a2 = ( 1 .. 500 );
2484 my @a3 = ( 1 .. 300 );
2485 $mce->process({ progress => make_progress(scalar(@a1), "array 1") }, \@a1);
2487 print "\n";
2489 $mce->process({ progress => make_progress(scalar(@a2), "array 2") }, \@a2);
2491 print "\n";
2493 $mce->process({ progress => make_progress(scalar(@a3), "array 3") }, \@a3);
2495 print "\n";
2497 $mce->shutdown;
2499 -- Output
2501 [####################] 100.0% ETA: 0:00 array 1
2503 [####################] 100.0% ETA: 0:00 array 2
2504 [####################] 100.0% ETA: 0:00 array 3
2505 When size is not know, such as reading from "STDIN", the only thing one
2507 can do is report the size completed thus far.
2508 # 1 kibibyte equals 1024 bytes
2510 progress => sub {
2512 my ($completed_size) = @_;
2513 printf "%0.1f kibibytes\n", $completed_size / 1024;
2514 }
2515
2517 • MCE::Examples
2518
2520 MCE
2521
2523 Mario E. Roy, <marioeroy AT gmail DOT com>
2524perl v5.36.0 2022-07-22 MCE::Core(3)