in reply to Return all the data from child to parent with Parallel::Forkmanager

Hi Microcebus,

You were close. Notice the arguments given to $pm->start and $pm->finish. Start accepts an optional identifier for the process. That same identifier is used as the 3rd argument for the on_finish callback. In the demonstration that follows, I puposely have read_GTF die with an exception. The output is the same whether threads equals 1 or 4.

Like afoken mentioned here, Parallel::ForkManager must be 0.7.6 or later for the demonstration to run.

use strict; use warnings; use feature 'say'; use Parallel::ForkManager; my $threads = 4; my @name = ('read_genome', 'read_mapfile', 'read_GTF', 'read_RM'); my @task = (\&read_genome, \&read_mapfile, \&read_GTF, \&read_RM); my (@ret, @err); if ($threads == 1) { for my $id (0 .. $#task) { $ret[$id] = eval { $task[$id]->() }; $err[$id] = $@; } } else { my $pm = new Parallel::ForkManager($threads); $pm->set_waitpid_blocking_sleep(0); $pm->run_on_finish( sub { my ($pid, $exit, $id, $signal, $core, $data) = @_; $ret[$id] = delete $data->{ret}; $err[$id] = delete $data->{err}; }); for my $id (0 .. $#task) { $pm->start($id) and next; my $res = eval { $task[$id]->() }; $pm->finish(0, { ret => $res, err => $@ }); } $pm->wait_all_children; } sub read_genome { # do something return { 'aa' => 'bb' }; } sub read_mapfile { # do something return { 'cc' => 'dd' }; } sub read_GTF { # do something die 'exception'; return { 'ee' => 'ff' }; # not reached } sub read_RM { # do something return { 'gg' => 'hh' }; } # use data generated in the subroutines use Data::Dumper; for my $id (0 .. $#task) { say "## ", $name[$id]; if (length $err[$id]) { say "ERROR: ", $err[$id]; next; } say Dumper($ret[$id]); }

Output:

## read_genome $VAR1 = { 'aa' => 'bb' }; ## read_mapfile $VAR1 = { 'cc' => 'dd' }; ## read_GTF ERROR: exception at j1.pl line 54. ## read_RM $VAR1 = { 'gg' => 'hh' };

Regards, Mario

Replies are listed 'Best First'.
Re^2: Return all the data from child to parent with Parallel::Forkmanager
by marioroy (Prior) on Aug 19, 2017 at 04:24 UTC

    Hi again,

    The following provides a MCE::Loop demonstration. This time, am using a hash for storing the returned data. Workers pass data using socket handles inside MCE. Thus, no temp files are made for sending data to the parent process.

    use strict; use warnings; use feature 'say'; use MCE::Loop; my $threads = 4; my @name = ('read_genome', 'read_mapfile', 'read_GTF', 'read_RM'); my @task = (\&read_genome, \&read_mapfile, \&read_GTF, \&read_RM); my %ret; if ($threads == 1) { for my $id (0 .. $#task) { $ret{"$id"} = eval { $task[$id]->() }; $ret{"$id:err"} = $@; } } else { MCE::Loop->init( max_workers => $threads, posix_exit => 1, chunk_size => 1 ); %ret = mce_loop { my $id = $_; my $res = eval { $task[$id]->() }; MCE->gather( $id => $res, "$id:err" => $@ ); } [ 0 .. $#task ]; MCE::Loop->finish; } sub read_genome { # do something return { 'aa' => 'bb' }; } sub read_mapfile { # do something return { 'cc' => 'dd' }; } sub read_GTF { # do something die 'exception'; return { 'ee' => 'ff' }; # not reached } sub read_RM { # do something return { 'gg' => 'hh' }; } # use data generated in the subroutines use Data::Dumper; for my $id (0 .. $#task) { say "## ", $name[$id]; if (length $ret{"$id:err"}) { say "ERROR: ", $ret{"$id:err"}; next; } say Dumper($ret{$id}); }

    Output:

    ## read_genome $VAR1 = { 'aa' => 'bb' }; ## read_mapfile $VAR1 = { 'cc' => 'dd' }; ## read_GTF ERROR: exception at j2.pl line 45, <__ANONIO__> line 2. ## read_RM $VAR1 = { 'gg' => 'hh' };

    Regards, Mario