| Pixie documentation | Contained in the Pixie distribution. |
Pixie::Store -- Abstract interface to physical storage
In a deploy script:
use Pixie::Store::DBI;
# Setup the datastore.
Pixie::Store::DBI->deploy('dbi:mysql:dbname=foo',
user => 'wibble',
pass => 'plib',
object_table => 'object');
In a pixie client:
use Pixie::Store::MySubclass;
use Pixie;
my $pixie = Pixie->connect('prefix:myspec',
user => 'bill',
pass => 'flobadob');
Pixie::Store provides pixie with an abstracted interface to the
physical storage used to actually store the objects that Pixie
manages. It is not a 'public' class; most Pixie users will never have
to touch it except maybe to call the deploy method of an
appropriate subclass.
However, if you want to add another storage medium to Pixie, start here. (If you want to add specific methods for storing in a particular RDBMS, you should take a look at DBIx::AnyDBD before diving into Pixie::Store::DBI::Default and its woefully underdocumented friends.
There is no public interface to Pixie::Store. However, where
appropriate, Pixie::Store subclasses may implement a deploy method
which should be responsible for setting up a suitable storage
structure which can be connected to later.
Pixie::Store implements almost no methods itself, except for a 'connect' factory method, which takes a 'storage spec' (similar in form to the classic DBI data source spec), works out which concrete subclass to use for the real connection, loads it if necessary and uses that to build a store object.
But Pixie proper depends on the following methods existing and working as described.
Makes the actual connection and returns an object of the appropriate
class. The only fixed part of the interface is that the storage spec
shall come first, and the only fixed part of that is that storage
specs tend to look like 'id:...'. The 'id:' tag is used by
Pixie::Store::connect to identify which subclass to
instantiate. The Typemap has more details of how that works.
Empties the datastore, removes all stored objects and any associated metadata. Use with caution. (It is remarkably handy when one is writing test scripts though...)
Take the FLATTENED_OBJECT and stash it where it can be found via the given OID. The FLATTENED_OBJECT is guaranteed to be an arbitrarily long string of bytes (just to make life easy...). An OID is a string of up to 255 characters. Overwrites any existing entry at that OID.
Returns the object associated with the given OID if it exists; returns undef/the empty list if no object can be found and throws an exception if it finds more than one object associated with that OID. (OIDs are supposed to be unique after all).
Deletes the object associated with OID. Returns true if an object existed, or false if there was no such object.
Possibly misnamed. Locks the database so that nobody else can interfere. (Actually, it is often implemented as 'begin transaction'...).
Again, possibly misnamed. Ensures that all the changes that have been inserted really have been inserted, and frees the database for other users. Should possibly be called 'commit'.
Rolls the database back to the state it was in at the last 'lock'. Not misnamed. (Hurrah).
Once you have subclassed Pixie::Store you need to let it know about
your new subclass so it can make connect work. To do that, pick an
appropriate prefix string to identify your subclass and add something
like the following -- after the use base 'Pixie::Store';
part, or things will break -- to your code:
$Pixie::Store::typemap{prefix} = __PACKAGE__;
Once you have done this, the code given in the synopsis should work, as if by magic.
James Duncan <james@fotango.com>, Piers Cawley <pdcawley@bofh.org.uk> and Leon Brocard <acme@astray.com>.
Copyright 2002 Fotango Ltd
This software is released under the same license as Perl itself.
| Pixie documentation | Contained in the Pixie distribution. |
package Pixie::Store; use strict; our $VERSION="2.06"; my %typemap = ( memory => 'Pixie::Store::Memory', bdb => 'Pixie::Store::BerkeleyDB', dbi => 'Pixie::Store::DBI', ); use Scalar::Util qw/weaken/; use Carp; #use overload # '""' => 'as_string'; sub as_string { my $proto = shift; my $str; if (ref $proto) { $str .= ref($proto); $str .= ": $proto->{spec}" if $proto->{spec}; } else { $str .= $proto; } return $str; } sub connect { my $proto = shift; my $spec = shift; my($type, $path) = split(':', $spec, 2); $type = lc($type); die "Invalid database spec" unless exists $typemap{$type}; eval "require " . $typemap{$type}; die $@ if $@; my $self = $typemap{$type}->connect($path,@_); $self->{spec} = $spec; return $self; } sub object_graph_for { my $self = shift; my $pixie = shift; my $graph = $pixie->get_object_named('PIXIE::Node Graph'); unless ($graph) { $graph = Pixie::ObjectGraph->new; $pixie->bind_name('PIXIE::Node Graph' => $graph); } return $graph; } sub remove_from_store { my $self = shift; my($oid) = @_; $self->remove_from_rootset($oid) ->_delete($oid); } # Low level locking sub locked_set { my $self = shift; return $self->{locked_set} ||= {}; } sub lock_object_for { my $self = shift; my($oid, $locker) = @_; $self->locked_set->{ $oid } = $locker->_oid; return 1; } sub unlock_object_for { my $self = shift; my $oid = shift; delete $self->locked_set->{$oid}; } sub release_all_locks { my $self = shift; my $locked_set = $self->locked_set; $self->unlock_object_for(@$_) for map {[$_ => $locked_set->{$_}]} keys %$locked_set; return $self; } sub add_to_rootset { my $self = shift; my $thing = shift; $self->_add_to_rootset($thing) unless $self->is_hidden($thing); } sub is_hidden { my $self = shift; my $thing = shift; $thing->PIXIE::oid =~ /^<NAME:PIXIE::/; } sub _add_to_rootset { $_[0]->subclass_responsibility(@_) } sub remove_from_rootset { $_[0]->subclass_responsibility($_[0]); } sub rootset { $_[0]->subclass_responsibility; } sub lock { $_[0]->subclass_responsibility(@_) } sub unlock { $_[0]->subclass_responsibility(@_) } sub rollback { $_[0]->subclass_responsibility(@_) } sub lock_for_GC { my $self = shift; $self->lock; } sub unlock_after_GC { my $self = shift; $self->unlock; } sub subclass_responsibility { my $self = shift; require Carp; Carp::carp( (caller(1))[3], " not implemented for ", ref($self) ); return wantarray ? @_ : $_[-1]; } sub DESTROY { my $self = shift; $self->release_all_locks; } 1; __END__