use Moose;
use Method::Signatures::Simple;
use Data::Perl::Stream::Array;
+use Data::Perl::Collection::Set;
+use Scalar::Util qw(weaken refaddr);
has _store => (is => 'ro', required => 1, init_arg => 'store');
-has _class => (is => 'ro', predicate => '_has_class');
+has _class => (is => 'ro', predicate => '_has_class', init_arg => 'class');
has _set_over => (is => 'ro', required => 1, init_arg => 'set_over');
## member cache (all members)
has _member_cache => (
- is => 'rw', lazy_build => 1,
+ is => 'ro', lazy_build => 1,
predicate => '_member_cache_built',
+ writer => '_set_member_cache',
);
method _build__member_cache {
while (my ($raw) = $stream->next) {
my $obj = do {
if (my ($obj) = $self->_key_cache_get_raw($raw)) {
- $self->_merge($obj, $raw)
+ # can't just $self->_merge($obj, $raw) since $obj might have changed
+ $self->_refresh($obj, $raw)
} else {
$self->_add_to_key_cache($self->_inflate($raw))
}
};
push @cache, $obj;
}
+ $self->_notify_observers(all_members => \@cache);
\@cache
}
: ()
}
+method _all_key_cache_members {
+ values %{$self->_key_cache}
+}
+
+method _set_key_cache_members ($members) {
+ %{$self->_key_cache} = (map +($self->_object_to_id($_) => $_), @$members);
+ return
+}
+
+## observers
+
+has _observer_callbacks => (
+ is => 'ro', default => sub { {} },
+);
+
+method _notify_observers ($event, $payload) {
+ my $oc = $self->_observer_callbacks;
+ foreach my $refaddr (keys %$oc) {
+ my ($obj, $cb) = @{$oc->{$refaddr}};
+ unless (defined $obj) { # weak ref was garbage collected
+ delete $oc->{$refaddr};
+ next;
+ }
+ $obj->$cb($self, $event, $payload);
+ }
+ $payload
+}
+
+method _register_observer ($obj, $cb) {
+ my $entry = [ $obj, $cb ];
+ weaken($entry->[0]);
+ $self->_observer_callbacks->{refaddr($obj)} = $entry;
+ return
+}
+
+method _setup_observation_of ($other) {
+ $other->_register_observer($self, method ($from, $event, $payload) {
+ if ($event eq 'add' or $event eq 'get') {
+ $self->_add_to_caches($payload);
+ } elsif ($event eq 'remove') {
+ $self->_remove_from_caches($payload);
+ } elsif ($event eq 'all_members') {
+ # separate arrayref since future add will trigger push()
+ $self->_set_caches([ @$payload ]);
+ }
+ });
+ return
+}
+
## thunking between the store representation and the set representation
#
# _inflate is raw data -> final repr
# (this is used for pk-generated values and later lazy loading)
#
# _deflate_spec is attributes of final repr -> raw data
+# _merge_spec is final repr + extra attributes and update repr
method _inflate ($raw) {
bless($raw, $self->_class) if $self->_has_class;
$obj
}
+method _refresh ($obj, $raw) {
+ # if $obj has been changed but not flushed we'd destroy data doing
+ # a blind merge - but if $obj has change tracking of some sort then
+ # we -could- do something safely, so this method exists to be mangled
+ # by subclasses
+ $obj
+}
+
method _deflate_spec ($spec) {
$spec
}
+method _merge_spec ($obj, $spec) {
+ @{$obj}{keys %$spec} = values %$spec;
+ $obj
+}
+
## methods to get ids
method _raw_to_id ($raw) {
## array-ish operations - i.e. get all members
method _new_raw_stream {
- $self->_store->new_select_command([])->execute
+ $self->_store->new_select_command({})->execute
}
method flatten {
Data::Perl::Stream::Array->new(array => $self->_member_cache);
}
+# theoretically inefficient except that if we're being asked this then
+# either the data should have been pre-loaded or we're going to get all
+# elements anyway
+
+method count {
+ scalar $self->flatten
+}
+
+method map ($sub) {
+ Data::Perl::Collection::Set->new(
+ members => [ map $sub->($_), $self->flatten ]
+ )
+}
+
+method _set_caches ($members) {
+ $self->_set_member_cache($members);
+ $self->_set_key_cache_members($members);
+ return
+}
+
## load single row
method get ($spec) {
return $got
}
if (my ($raw) = $self->_get_from_store($self->_deflate_spec($spec))) {
- return $self->_add_to_key_cache($self->_inflate($raw))
+ return $self->_notify_observers(
+ get => $self->_add_to_key_cache($self->_inflate($raw))
+ );
}
return undef # we aren't handling cache misses here yet
}
$self->_store->new_select_single_command($raw)->execute
}
-## add to set
+## add member
method add ($new) {
$self->_add_to_store($new);
$self->_add_to_caches($new);
+ $self->_notify_observers(add => $new);
$new
}
$new
}
-## remove from set
+## remove member
method remove ($old) {
$self->_remove_from_store($old);
$self->_remove_from_caches($old);
+ $self->_notify_observers(remove => $old);
$old
}
method _remove_from_store ($old) {
- $self->_store->new_delete_command($self->_deflate($old))->execute
+ $self->_store->new_delete_single_command($self->_deflate($old))->execute
}
method _remove_from_caches ($old) {
$old
}
-## update
+## update member
method _update_in_store ($obj) {
# this is currently a call command but we should think about it
# being a row command so that we can have RETURNING or other
# mechanisms handle things like set-on-update datetime values
- $self->_store->new_update_command($self->_deflate($obj))->execute
+ $self->_store->new_update_single_command($self->_deflate($obj))->execute
+}
+
+# I do wonder if we needed _merge_spec or if we'd be better off with
+# just using the raw merge routine ...
+
+method _update_set_in_store ($spec) {
+ $self->_store->new_update_command($self->_deflate_spec($spec))->execute;
+ if ($self->_member_cache_built) {
+ my $cache = $self->_member_cache;
+ foreach my $obj (@{$cache}) {
+ $self->_merge_spec($obj, $spec);
+ }
+ $self->_notify_observers(all_members => $cache);
+ }
+ return
+}
+
+method _remove_set_from_store {
+ $self->_store->new_delete_command->execute;
+ $self->_set_caches([]);
+ $self->_notify_observers(all_members => []);
+ return
}
1;
projects / dbsrgits/DBIx-Data-Store-old.git / blobdiff