Google::Spreadsheet::Agent - A Distributed Agent System using Google Spreadsheets
Version 0.01
use Google::Spreadsheet::Agent; my $google_agent = Google::Spreadsheet::Agent->new( agent_name => $goal, page_name => $google_page, debug => $debug, max_selves => $max, bind_key_fields => { 'foo' => 'this_particular_foo' }, prerequisites => [ 'isitdone', 'isthisone' ], subsumed_by => { 'someother_agent.pl' => 3, 'someother_process' => 1 } ); $google_agent->run_my(sub { print STDERR "THIS ONE PASSES!!!"; return 1; }); $google_agent->run_my(sub { print STDERR "THIS ONE FAILS AND EITHER EMAILS OR PRINTS THIS ERROR TO STDERR (depending on debug)!!!"; return; }); $google_agent->run_my(sub { print STDERR "THIS ONE PASSES AND UPDATES THE 'cool' field in the spreadsheet!!!"; return (1, {'cool' => 'really cool'}); });
Google::Spreadsheet::Agent is a framework for creating massively distributed pipelines across many different servers, each using the same google spreadsheet as a control panel. It is extensible, and flexible. It doesnt specify what goals any pipeline should be working towards, or which goals are prerequisites for other goals, but it does provide logic for easily defining these relationships based on your own needs. It does this by providing a subsumption architecture, whereby many small, highly focused agents are written to perform specific goals, and also know what resources they require to perform them. In addition, it is designed from the beginning to support the creation of simple human-computational workflows.
Scripts which use Google::Spreadsheet::Agents must supply the appropriate configuration for it to work. This can be done one of two ways.
This is the easiest way to configure a set of agents using the same configuration. See config/agent.conf.yml.tmpl for a template, with documentation, of what needs to be defined.
You can define a HashRef with all of the key-value pairs defined in config/agent.conf.yml.tmpl and pass that to the constructor. This may be more useful where you want to use other serialization systems (e.g. JSON, XML, etc) to store configuration, which can be manipulated into a HashRef to be passed into the constructor.
This method constructs a new instance of an Google::Spreadsheet::Agent. An instance must specify its name, the name of the Worksheet within the spreadsheet that it is working off, and values for the required key_field(s) within the configuration which will result in a single row being returned from the given spreadsheet. Optionally, you can specify an ArrayRef of prerequisite fields in the spreadsheet which must be true before the agent can run, whether to print out debug information to the terminal, or email the errors using the configured email only on errors (default), the maximum number of agents of this name to allow to run on the given machine, and a HashRef of processes which, if a certain number are already running on the machine, should cause the agent to exit without running. required: agent_name => Str config || config_file (you must supply configuration) page_name => Str bind_key_fields => HashRef { key_field_name => bound_value, ... } optional: prerequisites => [] debug => Bool max_selves => Int subsumed_by => { process_name => max_allowed, ... } This method will throw an exception if bind_key_fields are not supplied for required key_fields, as specified in the configuration. Also, there must be a field in the spreadsheet name for the agent_name. This field will be filled in with the status of the agent for a particular row, e.g. 1 for finished, r:hostname for running, or f:hostname for failure.
This method takes a subroutine codeRef as an argument. It then checks to determine if the agent needs to run for the given bind_key_field(s) specified row (it must have a 1 in the 'ready' field for the row, and the agent_name field must be empty), whether any prerequisite fields are true, whether the agent is subsumed by something else running on the machine, and whether there are not already max_selves other instances of the agent running on the machine. If all of these are true, it then attempts to fill its hostname into the field for the agent_name. If it succeeds, it will then run the code_ref. If it does not succeed (such as if an instance running on another server already chose that job and won the field) it exits. The coderef can do almost anything it wants to do, but it must return one of the following:
This instructs Google::Spreadsheet::Agent to place a 1 (true) value in the field for the agent on the spreadsheet, signifying that it has been completed.
This instructs Google::Spreadsheet::Agent to place F:hostname into the field for the agent on the spreadsheet, signifying that it has failed. It will not run again for this job until the failure is cleared from the spreadsheet (by any other agent).
This does what returning true or false does, as well as allowing specific fields in the spreadsheet to also be modified by the calling code. The HashRef should contain keys only for those fields to be updated (it should not attempt to update the field for the agent_name itself, as this will be ignored).
In addition, the coderef can print to STDOUT and STDERR. If the agent was instantiated in debug mode (true), it will print these to their normal destination. If the agent was instantiated without debug mode (the default), STDOUT and STDERR are captured, and, if the codeRef returned false, emailed to the address specified in the configuration using the same google account that configures access to the google spreadsheet. One thing the agent must try at all costs to avoid is dying during the subref (e.g. use eval for anything that you dont have control over). It should always try to return one of the valid return states so that the spreadsheet status can be updated correctly.
This returns the name of the agent, in case it is needed by the calling code for other reasons.
This returns the debug state specified in the constructor.
This returns the actual Net::Google::Spreadsheet object used by the agent, in case other types of queries, or modifications need to be made that do not fit within this system.
Darin London, <darin.london at duke.edu>
<darin.london at duke.edu>
Please report any bugs or feature requests to bug-google-spreadsheet-agent at rt.cpan.org, or through the web interface at http://rt.cpan.org/NoAuth/ReportBug.html?Queue=Google-Spreadsheet-Agent. I will be notified, and then you'll automatically be notified of progress on your bug as I make changes.
bug-google-spreadsheet-agent at rt.cpan.org
You can find documentation for this module with the perldoc command.
perldoc Google::Spreadsheet::Agent
You can also look for information at:
RT: CPAN's request tracker
http://rt.cpan.org/NoAuth/Bugs.html?Dist=Google-Spreadsheet-Agent
AnnoCPAN: Annotated CPAN documentation
http://annocpan.org/dist/Google-Spreadsheet-Agent
Net::Google::Spreadsheets Moose
Copyright 2009 Darin London.
This program is free software; you can redistribute it and/or modify it under the terms of either: the GNU General Public License as published by the Free Software Foundation; or the Artistic License.
See http://dev.perl.org/licenses/ for more information.
Note, these use an older Namespace IGSP::GoogleAgent instead of Google::Spreadsheet::Agent. RFC: 798154 Code: 798311
To install Google::Spreadsheet::Agent, copy and paste the appropriate command in to your terminal.
cpanm
cpanm Google::Spreadsheet::Agent
CPAN shell
perl -MCPAN -e shell install Google::Spreadsheet::Agent
For more information on module installation, please visit the detailed CPAN module installation guide.