MyCPAN::Indexer::Tutorial - How the backpan_indexer.pl pieces fit together
MyCPAN::Indexer system lets you plug in different components to control major portions of the process of examining Perl distributions and collating the results.
It's up to each component to obey its interface and do that parts the other components expect it to do.
The idea is to decouple some of these bits as much as possible.
backpan_indexer.pl does its work,
it stores information about its components in an anonymous hash called
The different components have access to this hash.
(To Do: this is some pretty bad design smell,
but that's how it is right now).
Specific implementations will impose other requirements not listed in this tutorial.
The application is the bit that you write when you want to do something very specialized with a different process. The application object controls the big picture.
the module version,
the script version.
The coordinator is just a way for the components to talk to each other. The application starts up, creates a coordinator object, and stores it. The application gives a reference to the coordinator to every component.
When the application creates components, it tells each of about the coordinator. Each component can talk to the coordinator to get to parts of the application it doesn't directly know about. Each component tells the coordinator about itself so the coordinator can talk to any component.
The coordinator maintains the "notes", which are arbitrary bits of information that components use to pass information around.
Most of the work to examine a Perl distribution is in
When it gets down to it,
MyCPAN knows about Perl distributions is in there.
It has a
run() method which handles the examination.
It kicks off
which figures out what to do by getting a list of steps from
This technique is common throughout
One method returns a list of methods to run.
a subclass can control the process by overriding the method that returns the steps.
The basic class is
MyCPAN::Indexer::TestCensus is an example of another indexing class.
The Queue class is responsible for getting the list of distributions to process.
get_queue and passes it a ConfigReader::Simple object.
get_queue does whatever it needs to do,
then returns an array reference of file paths to process.
Each path should represent a single Perl distribution.
queue - a reference to the array reference returned by get_queue.
To Do: The Queue class should really be an iterator of some sort. Instead of returning an array (which it can't change), return an iterator.
The Worker class returns the anonymous subroutine that the interface class calls for each of its cycles. Inside that code reference, do the actual indexing work, including saving the results.
get_task with a reference to its
child_task - a reference to the code reference returned by get_task.
To Do: There should be a storage class which the worker class hands the results to.
The Reporter class implements the bits to store the result of the Worker class.
get_reporter with a reference to its
get_reporter( $info )
reporter - the code ref to handle storing the information
Expects in config:
The Dispatcher class implements the bits to hand out work to the worker class. The Interface class, discussed next, repeatedly calls the interface_callback code ref the Dispatcher class provides.
dispatcher - the dispatcher object, with start and finish methods interface_callback - a code ref to call repeatedly in the Interface class
child_task - the code ref that handles indexing a single dist queue - the array ref of dist paths
The Interface class really has two jobs. It makes the live reporting interface while
backpan_indexer.pl runs, at it repeatedly calls the dispatcher to start new work.
interface_callback - a code ref to call repeatedly in the Interface class
This code is in Github:
brian d foy,
Copyright (c) 2008-2009, brian d foy, All Rights Reserved.
You may redistribute this under the same terms as Perl itself.