View on
MetaCPAN is shutting down
For details read Perl NOC. After June 25th this page will redirect to
David Muir Sharnoff > Proc-Parallel > Proc::Parallel::Unreliable



Annotate this POD

View/Report Bugs


Proc::Parallel::Unreliable - maintain a pool of unreliable slave processes


 use Proc::Parallel::Unreliable;

        input           => $input_filehandle,
        output          => $output_callback_function,
        cb              => $slave_process_start_function,
        count           => $number_of_slaves_to_run,
        timeout         => $maximum_time_to_process_a_line,
        input_id        => $callback_to_identify_input,
        input_hlines    => $number_of_header_lines_in_input_stream,
        output_hlines   => $number_of_header_lines_in_output_stream,
        output_id       => $callback_to_identify_output,
        items           => $number_of_buffer_lines_per_slave,
        track_items     => $identify_items_flag,
        debug           => $debug_level,
        pidfile         => $process_id_file_to_update,


This module deals with a particular situation: needing to run code that may crash or hang while processing a stream of data.

It reads input, farms the input out to one or more slave process, collects the output from the slave processes, and writes a single stream of output.

Each line of input must be self-contained because it can be randomly sent to any of the slaves.

It has two modes: tracking items, or not tracking items. When it tracks items, it uses the input_id callback to identify the input items and the output_id callback to identify the results it gets back from the slaves. When tracking items, the slave processes must produce output for each input item: they'll be killed off if they don't.

When not tracking items, output is sent to the slave processes until the buffers fill up. It's not as elegant.

If the input stream or the output stream has headers, the input stream headers will be reproduced for each slave and the output stream headers will be suppresssed except for one slave. Use the input_hlines and output_hlines parameters for this.

Output is written with a callbac: output.

The process id file, pidfile will be updated to show the current set of processes.


        output          => sub {
                                print $output $o;
        input           => $input,
        cb              => sub {
                                my ($my_input_fd, $my_output_fd) = @_;
        count           => 3,
        timeout         => 60,
        input_id        => sub {
                                my ($iref) = @_;
                                $$iref =~ /^(\d+)/
                                return $1;
        input_hlines    => 0,
        output_hlines   => 1,
        output_id       => sub {
                                my ($oref) = @_;
                                $$iref =~ /^(\d+)/
                                return $1;
        items           => 4,
        track_items     => 1,
        debug           => $debug,
        pidfile         => "/var/run/",


This package may be used and redistributed under the terms of either the Artistic 2.0 or LGPL 2.1 license.

syntax highlighting: