NAME

Algorithm::Dependency - Base class for implementing various dependency trees

VERSION

version 1.112

SYNOPSIS

Typical Usage: Ordering based on dependency requirements

  use Algorithm::Dependency::Ordered;
  use Algorithm::Dependency::Source::HoA;
  
  my $deps = {
    core  => [ ],
    a     => [ 'core' ],
    b     => [ 'a' ]
    this  => [ ],
    that  => [ ],
  };
  my $deps_source = Algorithm::Dependency::Source::HoA->new( $deps );

  my $dep = Algorithm::Dependency::Ordered->new(
    source   => $deps_source,
    selected => [ 'this', 'that' ], # Items we have processed elsewhere or have already satisfied
  )
  or die 'Failed to set up dependency algorithm';

  my $also = $dep->schedule_all();
  # Returns: ['core', 'a', 'b'] -- ie: installation-order. Whereas using base
  # Algorithm::Dependency would return sorted ['a', 'b', 'core']

  my $also = $dep->schedule( 'b' );
  # Returns: ['core', 'a', 'b'] -- installation order, including ourselves

  my $also = $dep->depends( 'b' );
  # Returns: ['a', 'core'] -- sorted order, not including ourselves

Base Classes

  use Algorithm::Dependency;
  use Algorithm::Dependency::Source::File;
  
  # Load the data from a simple text file
  my $data_source = Algorithm::Dependency::Source::File->new( 'foo.txt' );
  
  # Create the dependency object, and indicate the items that are already
  # selected/installed/etc in the database
  my $dep = Algorithm::Dependency->new(
      source   => $data_source,
      selected => [ 'This', 'That' ]
  ) or die 'Failed to set up dependency algorithm';
  
  # For the item 'Foo', find out the other things we also have to select.
  # This WON'T include the item we selected, 'Foo'.
  my $also = $dep->depends( 'Foo' );
  print $also
        ? "By selecting 'Foo', you are also selecting the following items: "
                . join( ', ', @$also )
        : "Nothing else to select for 'Foo'";
  
  # Find out the order we need to act on the items in.
  # This WILL include the item we selected, 'Foo'.
  my $schedule = $dep->schedule( 'Foo' );

DESCRIPTION

Algorithm::Dependency is a framework for creating simple read-only dependency hierarchies, where you have a set of items that rely on other items in the set, and require actions on them as well.

Despite the most visible of these being software installation systems like the CPAN installer, or Debian apt-get, they are useful in other situations. This module intentionally uses implementation-neutral words, to avoid confusion.

Terminology

The term ITEM refers to a single entity, such as a single software package, in the overall set of possible entities. Internally, this is a fairly simple object. See Algorithm::Dependency::Item for details.

The term SELECT means that a particular item, for your purposes, has already been acted up in the required way. For example, if the software package had already been installed, and didn't need to be re-installed, it would be SELECTED.

The term SOURCE refers to a location that contains the master set of items. This will be very application specific, and might be a flat file, some form of database, the list of files in a folder, or generated dynamically.

General Description

Algorithm::Dependency implements algorithms relating to dependency hierarchies. To use this framework, all you need is a source for the master list of all the items, and a list of those already selected. If your dependency hierarchy doesn't require the concept of items that are already selected, simply don't pass anything to the constructor for it.

Please note that the class Algorithm::Dependency does NOT implement an ordering, for speed and simplicity reasons. That is, the schedule it provides is not in any particular order. If item 'A' depends on item 'B', it will not place B before A in the schedule. This makes it unsuitable for things like software installers, as they typically would need B to be installed before A, or the installation of A would fail.

For dependency hierarchies requiring the items to be acted on in a particular order, either top down or bottom up, see Algorithm::Dependency::Ordered. It should be more applicable for your needs. This is the the subclass you would probably use to implement a simple ( non-versioned ) package installation system. Please note that an ordered hierarchy has additional constraints. For example, circular dependencies ARE legal in a non-ordered hierarchy, but ARE NOT legal in an ordered hierarchy.

Extending

A module for creating a source from a simple flat file is included. For details see Algorithm::Dependency::Source::File. Information on creating a source for your particular use is in Algorithm::Dependency::Source.

METHODS

new %args

The constructor creates a new context object for the dependency algorithms to act in. It takes as argument a series of options for creating the object.

source => $Source

The only compulsory option is the source of the dependency items. This is an object of a subclass of Algorithm::Dependency::Source. In practical terms, this means you will create the source object before creating the Algorithm::Dependency object.

selected => [ 'A', 'B', 'C', etc... ]

The selected option provides a list of those items that have already been 'selected', acted upon, installed, or whatever. If another item depends on one in this list, we don't have to include it in the output of the schedule or depends methods.

ignore_orphans => 1

Normally, the item source is expected to be largely perfect and error free. An 'orphan' is an item name that appears as a dependency of another item, but doesn't exist, or has been deleted.

By providing the ignore_orphans flag, orphans are simply ignored. Without the ignore_orphans flag, an error will be returned if an orphan is found.

The new constructor returns a new Algorithm::Dependency object on success, or undef on error.

source

The source method retrieves the Algorithm::Dependency::Source object for the algorithm context.

selected_list

The selected_list method returns, as a list and in alphabetical order, the list of the names of the selected items.

selected $name

Given an item name, the selected method will return true if the item is selected, false is not, or undef if the item does not exist, or an error occurs.

item $name

The item method fetches and returns the item object, as specified by the name argument.

Returns an Algorithm::Dependency::Item object on success, or undef if an item does not exist for the argument provided.

depends $name1, ..., $nameN

Given a list of one or more item names, the depends method will return a reference to an array containing a list of the names of all the OTHER items that also have to be selected to meet dependencies.

That is, if item A depends on B and C then the depends method would return a reference to an array with B and C. ( [ 'B', 'C' ] )

If multiple item names are provided, the same applies. The list returned will not contain duplicates.

The method returns a reference to an array of item names on success, a reference to an empty array if no other items are needed, or undef on error.

NOTE: The result of depends is ordered by an internal sort irrespective of the ordering provided by the dependency handler. Use Algorithm::Dependency::Ordered and schedule to use the most common ordering (process sequence)

schedule $name1, ..., $nameN

Given a list of one or more item names, the depends method will return, as a reference to an array, the ordered list of items you should act upon in whichever order this particular dependency handler uses - see Algorithm::Dependency::Ordered for one that implements the most common ordering (process sequence).

This would be the original names provided, plus those added to satisfy dependencies, in the preferred order of action. For the normal algorithm, where order it not important, this is alphabetical order. This makes it easier for someone watching a program operate on the items to determine how far you are through the task and makes any logs easier to read.

If any of the names you provided in the arguments is already selected, it will not be included in the list.

The method returns a reference to an array of item names on success, a reference to an empty array if no items need to be acted upon, or undef on error.

schedule_all;

The schedule_all method acts the same as the schedule method, but returns a schedule that selected all the so-far unselected items.

TO DO

Add the check_source method, to verify the integrity of the source.

Possibly add Algorithm::Dependency::Versions, to implement an ordered dependency tree with versions, like for perl modules.

Currently readonly. Make the whole thing writable, so the module can be used as the core of an actual dependency application, as opposed to just being a tool.

SEE ALSO

Algorithm::Dependency::Ordered, Algorithm::Dependency::Item, Algorithm::Dependency::Source, Algorithm::Dependency::Source::File

SUPPORT

Bugs may be submitted through the RT bug tracker (or bug-Algorithm-Dependency@rt.cpan.org).

AUTHOR

Adam Kennedy <adamk@cpan.org>

CONTRIBUTORS

  • Adam Kennedy <adam@ali.as>

  • Karen Etheridge <ether@cpan.org>

  • Mark Murawski <markm@intellasoft.net>

COPYRIGHT AND LICENSE

This software is copyright (c) 2003 by Adam Kennedy.

This is free software; you can redistribute it and/or modify it under the same terms as the Perl 5 programming language system itself.