X-Git-Url: http://git.shadowcat.co.uk/gitweb/gitweb.cgi?a=blobdiff_plain;f=lib%2FText%2FTradition%2FCollation%2FRelationshipStore.pm;h=4342bd98b79b209a991c37f034e8460a126a7095;hb=428bcf0bc79f77a7857b21ef881708faa792e33a;hp=91a63afc004fca345f7975edbc344a3d518fbfdb;hpb=359944f792e53a70d22ddb199c10ef25c269e154;p=scpubgit%2Fstemmatology.git diff --git a/lib/Text/Tradition/Collation/RelationshipStore.pm b/lib/Text/Tradition/Collation/RelationshipStore.pm index 91a63af..4342bd9 100644 --- a/lib/Text/Tradition/Collation/RelationshipStore.pm +++ b/lib/Text/Tradition/Collation/RelationshipStore.pm @@ -31,10 +31,10 @@ use_ok( 'Text::Tradition::Collation::RelationshipStore' ); my $cxfile = 't/data/Collatex-16.xml'; my $t = Text::Tradition->new( - 'name' => 'inline', - 'input' => 'CollateX', - 'file' => $cxfile, - ); + 'name' => 'inline', + 'input' => 'CollateX', + 'file' => $cxfile, + ); my $c = $t->collation; my @v1 = $c->add_relationship( 'n21', 'n22', { 'type' => 'lexical' } ); @@ -97,6 +97,7 @@ has 'equivalence_graph' => ( is => 'ro', isa => 'Graph', default => sub { Graph->new() }, + writer => '_reset_equivalence', ); has '_node_equivalences' => ( @@ -106,6 +107,7 @@ has '_node_equivalences' => ( equivalence => 'get', set_equivalence => 'set', remove_equivalence => 'delete', + _clear_equivalence => 'clear', }, ); @@ -116,6 +118,7 @@ has '_equivalence_readings' => ( eqreadings => 'get', set_eqreadings => 'set', remove_eqreadings => 'delete', + _clear_eqreadings => 'clear', }, ); @@ -133,7 +136,6 @@ around delete_reading => sub { my $orig = shift; my $self = shift; - $DB::single = 1; $self->_remove_equivalence_node( @_ ); $self->$orig( @_ ); }; @@ -165,7 +167,7 @@ sub _set_relationship { my( $self, $relationship, @vector ) = @_; $self->graph->add_edge( @vector ); $self->graph->set_edge_attribute( @vector, 'object', $relationship ); - $self->make_equivalence( @vector ) if $relationship->colocated; + $self->_make_equivalence( @vector ) if $relationship->colocated; } =head2 create @@ -194,18 +196,9 @@ sub create { } } - # Check to see if a nonlocal relationship is defined for the two readings - $rel = $self->scoped_relationship( $options->{'reading_a'}, - $options->{'reading_b'} ); - if( $rel && $rel->type eq $options->{'type'} ) { - return $rel; - } elsif( $rel ) { - throw( sprintf( "Relationship of type %s with scope %s already defined for readings %s and %s", $rel->type, $rel->scope, $options->{'reading_a'}, $options->{'reading_b'} ) ); - } else { - $rel = Text::Tradition::Collation::Relationship->new( $options ); - $self->add_scoped_relationship( $rel ) if $rel->nonlocal; - return $rel; - } + $rel = Text::Tradition::Collation::Relationship->new( $options ); + $self->add_scoped_relationship( $rel ) if $rel->nonlocal; + return $rel; } =head2 add_scoped_relationship( $rel ) @@ -239,8 +232,12 @@ between the two reading strings. Returns undef if there is no general relationsh sub scoped_relationship { my( $self, $rdga, $rdgb ) = @_; my( $first, $second ) = sort( $rdga, $rdgb ); + my( $lcfirst, $lcsecond ) = sort( lc( $rdga ), lc( $rdgb ) ); if( exists $self->scopedrels->{$first}->{$second} ) { return $self->scopedrels->{$first}->{$second}; + } elsif( exists $self->scopedrels->{$lcfirst}->{$lcsecond} ) { + my $rel = $self->scopedrels->{$lcfirst}->{$lcsecond}; + return $rel->type ne 'orthographic' ? $rel : undef; } else { return undef; } @@ -258,103 +255,134 @@ add_relationship. =begin testing +use Test::Warn; use Text::Tradition; use TryCatch; -my $t1 = Text::Tradition->new( 'input' => 'Self', 'file' => 't/data/legendfrag.xml' ); -# Test 1: try to equate nodes that are prevented with an intermediate collation +my $t1; +warning_is { + $t1 = Text::Tradition->new( 'input' => 'Self', 'file' => 't/data/legendfrag.xml' ); +} 'DROPPING r14.2 -> r8.1: Cannot set relationship on a meta reading', + "Got expected relationship drop warning on parse"; + +# Test 1.1: try to equate nodes that are prevented with an intermediate collation ok( $t1, "Parsed test fragment file" ); my $c1 = $t1->collation; -my $trel = $c1->get_relationship( '9,2', '9,3' ); +my $trel = $c1->get_relationship( 'r9.2', 'r9.3' ); is( ref( $trel ), 'Text::Tradition::Collation::Relationship', "Troublesome relationship exists" ); is( $trel->type, 'collated', "Troublesome relationship is a collation" ); # Try to make the link we want try { - $c1->add_relationship( '8,6', '10,3', { 'type' => 'orthographic' } ); + $c1->add_relationship( 'r8.6', 'r10.3', { 'type' => 'orthographic' } ); ok( 1, "Added cross-collation relationship as expected" ); -} catch { - ok( 0, "Existing collation blocked equivalence relationship" ); +} catch( Text::Tradition::Error $e ) { + ok( 0, "Existing collation blocked equivalence relationship: " . $e->message ); } try { $c1->calculate_ranks(); ok( 1, "Successfully calculated ranks" ); -} catch { - ok( 0, "Collation now has a cycle" ); +} catch ( Text::Tradition::Error $e ) { + ok( 0, "Collation now has a cycle: " . $e->message ); } -# Now attempt merge of an identical reading +# Test 1.2: attempt merge of an identical reading try { - $c1->merge_readings( '9,3', '11,5' ); + $c1->merge_readings( 'r9.3', 'r11.5' ); ok( 1, "Successfully merged reading 'pontifex'" ); } catch ( Text::Tradition::Error $e ) { ok( 0, "Merge of mergeable readings failed: $e->message" ); } -# Test 2: try to equate nodes that are prevented with a real intermediate +# Test 1.3: attempt relationship with a meta reading (should fail) +try { + $c1->add_relationship( 'r8.1', 'r9.2', { 'type' => 'collated' } ); + ok( 0, "Allowed a meta-reading to be used in a relationship" ); +} catch ( Text::Tradition::Error $e ) { + is( $e->message, 'Cannot set relationship on a meta reading', + "Relationship link prevented for a meta reading" ); +} + +# Test 1.4: try to break a relationship near a meta reading +$c1->add_relationship( 'r7.6', 'r7.3', { type => 'orthographic' } ); +try { + $c1->del_relationship( 'r7.6', 'r7.7' ); + $c1->del_relationship( 'r7.6', 'r7.3' ); + ok( 1, "Relationship broken with a meta reading as neighbor" ); +} catch { + ok( 0, "Relationship deletion failed with a meta reading as neighbor" ); +} + +# Test 2.1: try to equate nodes that are prevented with a real intermediate # equivalence -my $t2 = Text::Tradition->new( 'input' => 'Self', 'file' => 't/data/legendfrag.xml' ); +my $t2; +warning_is { + $t2 = Text::Tradition->new( 'input' => 'Self', 'file' => 't/data/legendfrag.xml' ); +} 'DROPPING r14.2 -> r8.1: Cannot set relationship on a meta reading', + "Got expected relationship drop warning on parse"; my $c2 = $t2->collation; -$c2->add_relationship( '9,2', '9,3', { 'type' => 'lexical' } ); -my $trel2 = $c2->get_relationship( '9,2', '9,3' ); +$c2->add_relationship( 'r9.2', 'r9.3', { 'type' => 'lexical' } ); +my $trel2 = $c2->get_relationship( 'r9.2', 'r9.3' ); is( ref( $trel2 ), 'Text::Tradition::Collation::Relationship', "Created blocking relationship" ); is( $trel2->type, 'lexical', "Blocking relationship is not a collation" ); # This time the link ought to fail try { - $c2->add_relationship( '8,6', '10,3', { 'type' => 'orthographic' } ); + $c2->add_relationship( 'r8.6', 'r10.3', { 'type' => 'orthographic' } ); ok( 0, "Added cross-equivalent bad relationship" ); -} catch { - ok( 1, "Existing equivalence blocked crossing relationship" ); +} catch ( Text::Tradition::Error $e ) { + like( $e->message, qr/witness loop/, + "Existing equivalence blocked crossing relationship" ); } try { $c2->calculate_ranks(); ok( 1, "Successfully calculated ranks" ); -} catch { - ok( 0, "Collation now has a cycle" ); +} catch ( Text::Tradition::Error $e ) { + ok( 0, "Collation now has a cycle: " . $e->message ); } -# Test 3: make a straightforward pair of transpositions. +# Test 3.1: make a straightforward pair of transpositions. my $t3 = Text::Tradition->new( 'input' => 'Self', 'file' => 't/data/lf2.xml' ); # Test 1: try to equate nodes that are prevented with an intermediate collation my $c3 = $t3->collation; try { - $c3->add_relationship( '36,4', '38,3', { 'type' => 'transposition' } ); + $c3->add_relationship( 'r36.4', 'r38.3', { 'type' => 'transposition' } ); ok( 1, "Added straightforward transposition" ); -} catch { - ok( 0, "Failed to add normal transposition" ); +} catch ( Text::Tradition::Error $e ) { + ok( 0, "Failed to add normal transposition: " . $e->message ); } try { - $c3->add_relationship( '36,3', '38,2', { 'type' => 'transposition' } ); + $c3->add_relationship( 'r36.3', 'r38.2', { 'type' => 'transposition' } ); ok( 1, "Added straightforward transposition complement" ); -} catch { - ok( 0, "Failed to add normal transposition complement" ); +} catch ( Text::Tradition::Error $e ) { + ok( 0, "Failed to add normal transposition complement: " . $e->message ); } -# Test 4: try to make a transposition that could be a parallel. +# Test 3.2: try to make a transposition that could be a parallel. try { - $c3->add_relationship( '28,2', '29,2', { 'type' => 'transposition' } ); + $c3->add_relationship( 'r28.2', 'r29.2', { 'type' => 'transposition' } ); ok( 0, "Added bad colocated transposition" ); -} catch { - ok( 1, "Prevented bad colocated transposition" ); +} catch ( Text::Tradition::Error $e ) { + like( $e->message, qr/Readings appear to be colocated/, + "Prevented bad colocated transposition" ); } -# Test 5: make the parallel, and then make the transposition again. +# Test 3.3: make the parallel, and then make the transposition again. try { - $c3->add_relationship( '28,3', '29,3', { 'type' => 'orthographic' } ); + $c3->add_relationship( 'r28.3', 'r29.3', { 'type' => 'orthographic' } ); ok( 1, "Equated identical readings for transposition" ); -} catch { - ok( 0, "Failed to equate identical readings" ); +} catch ( Text::Tradition::Error $e ) { + ok( 0, "Failed to equate identical readings: " . $e->message ); } try { - $c3->add_relationship( '28,2', '29,2', { 'type' => 'transposition' } ); + $c3->add_relationship( 'r28.2', 'r29.2', { 'type' => 'transposition' } ); ok( 1, "Added straightforward transposition complement" ); -} catch { - ok( 0, "Failed to add normal transposition complement" ); +} catch ( Text::Tradition::Error $e ) { + ok( 0, "Failed to add normal transposition complement: " . $e->message ); } =end testing @@ -364,8 +392,11 @@ try { sub add_relationship { my( $self, $source, $target, $options ) = @_; my $c = $self->collation; - + my $sourceobj = $c->reading( $source ); + my $targetobj = $c->reading( $target ); throw( "Adding self relationship at $source" ) if $source eq $target; + throw( "Cannot set relationship on a meta reading" ) + if( $sourceobj->is_meta || $targetobj->is_meta ); my $relationship; my $thispaironly; my $droppedcolls = []; @@ -385,34 +416,28 @@ sub add_relationship { } # Try to create the relationship object. - $options->{'reading_a'} = $c->reading( $source )->text; - $options->{'reading_b'} = $c->reading( $target )->text; + $options->{'reading_a'} = $sourceobj->text; + $options->{'reading_b'} = $targetobj->text; $options->{'orig_a'} = $source; $options->{'orig_b'} = $target; if( $options->{'scope'} ne 'local' ) { # Is there a relationship with this a & b already? # Case-insensitive for non-orthographics. - my $rdga = $options->{'type'} eq 'orthographic' - ? $options->{'reading_a'} : lc( $options->{'reading_a'} ); - my $rdgb = $options->{'type'} eq 'orthographic' - ? $options->{'reading_b'} : lc( $options->{'reading_b'} ); + my $rdga = $options->{'reading_a'}; + my $rdgb = $options->{'reading_b'}; my $otherrel = $self->scoped_relationship( $rdga, $rdgb ); if( $otherrel && $otherrel->type eq $options->{type} && $otherrel->scope eq $options->{scope} ) { - warn "Applying existing scoped relationship"; + warn "Applying existing scoped relationship for $rdga / $rdgb"; $relationship = $otherrel; + } elsif( $otherrel ) { + throw( "Conflicting scoped relationship for $rdga / $rdgb at $source / $target" ); } } $relationship = $self->create( $options ) unless $relationship; # Will throw on error } - # Find all the pairs for which we need to set the relationship. - my @vectors; - if( $relationship->colocated && $relationship->nonlocal && !$thispaironly ) { - push( @vectors, $self->_find_applicable( $relationship ) ); - } - # Now set the relationship(s). my @pairs_set; my $rel = $self->get_relationship( $source, $target ); @@ -435,19 +460,58 @@ sub add_relationship { $self->_set_relationship( $relationship, $source, $target ) unless $skip; push( @pairs_set, [ $source, $target ] ); - # Set any additional relationships that might be in @vectors. - foreach my $v ( @vectors ) { - next if $v->[0] eq $source && $v->[1] eq $target; - next if $v->[1] eq $source && $v->[0] eq $target; - my @added = $self->add_relationship( @$v, $relationship ); - push( @pairs_set, @added ); + # Find all the pairs for which we need to set the relationship. + if( $relationship->colocated && $relationship->nonlocal && !$thispaironly ) { + push( @pairs_set, $self->add_global_relationship( $relationship ) ); } - # Finally, restore whatever collations we can, and return. $self->_restore_collations( @$droppedcolls ); return @pairs_set; } +=head2 add_global_relationship( $options, $skipvector ) + +Adds the relationship specified wherever the relevant readings appear together +in the graph. Options as in add_relationship above. + +=cut + +sub add_global_relationship { + my( $self, $options ) = @_; + # First see if we are dealing with a relationship object already + my $relationship; + if( ref( $options ) eq 'Text::Tradition::Collation::Relationship' ) { + $relationship = $options; + } else { + # Then see if a scoped relationship already applies for the words. + my $scopedrel = $self->scoped_relationship( + $options->{reading_a}, $options->{reading_b} ); + $relationship = $scopedrel ? $scopedrel + : $self->create( $options ); + } + # Sanity checking + throw( "Relationship passed to add_global is not global" ) + unless $relationship->nonlocal; + throw( "Relationship passed to add_global is not a valid global type" ) + unless $relationship->colocated && $relationship->type ne 'collated'; + + # Apply the relationship wherever it is valid + my @pairs_set; + foreach my $v ( $self->_find_applicable( $relationship ) ) { + my $exists = $self->get_relationship( @$v ); + if( $exists && $exists->type ne 'collated' ) { + throw( "Found conflicting relationship at @$v" ) + unless $exists->type eq $relationship->type + && $exists->scope eq $relationship->scope; + } else { + my @added = $self->add_relationship( @$v, $relationship ); + push( @pairs_set, @added ); + } + } + return @pairs_set; +} + + =head2 del_scoped_relationship( $reading_a, $reading_b ) Returns the general (document-level or global) relationship that has been defined @@ -533,7 +597,7 @@ sub del_relationship { sub _remove_relationship { my( $self, $equiv, @vector ) = @_; $self->graph->delete_edge( @vector ); - $self->break_equivalence( @vector ) if $equiv; + $self->_break_equivalence( @vector ) if $equiv; } =head2 relationship_valid( $source, $target, $type ) @@ -548,7 +612,9 @@ sub relationship_valid { my( $self, $source, $target, $rel, $mustdrop ) = @_; $mustdrop = [] unless $mustdrop; # in case we were passed nothing my $c = $self->collation; - if ( $rel eq 'transposition' || $rel eq 'repetition' ) { + ## Assume validity is okay if we are initializing from scratch. + return ( 1, "initializing" ) unless $c->tradition->_initialized; + if ( $rel eq 'transposition' || $rel eq 'repetition' ) { # Check that the two readings do (for a repetition) or do not (for # a transposition) appear in the same witness. # TODO this might be called before witness paths are set... @@ -586,9 +652,7 @@ sub relationship_valid { push( @$mustdrop, $self->_drop_collations( $source ) ); push( @$mustdrop, $self->_drop_collations( $target ) ); if( $c->end->has_rank ) { - my $cpred = $c->common_predecessor( $source, $target ); - my $csucc = $c->common_successor( $source, $target ); - foreach my $rk ( $cpred->rank+1 .. $csucc->rank-1 ) { + foreach my $rk ( $sourcerank .. $targetrank ) { map { push( @$mustdrop, $self->_drop_collations( $_->id ) ) } $c->readings_at_rank( $rk ); } @@ -681,6 +745,9 @@ sub related_readings { # Backwards compat if( $filter eq 'colocated' ) { $filter = sub { $_[0]->colocated }; + } elsif( !ref( $filter ) ) { + my $type = $filter; + $filter = sub { $_[0]->type eq $type }; } my %found = ( $reading => 1 ); my $check = [ $reading ]; @@ -736,7 +803,7 @@ sub merge_readings { $rel = $self->get_relationship( @$edge ); $self->_set_relationship( $rel, @vector ); } - $self->make_equivalence( $deleted, $kept ); + $self->_make_equivalence( $deleted, $kept ); } ### Equivalence logic @@ -746,14 +813,14 @@ sub _remove_equivalence_node { my $group = $self->equivalence( $node ); my $nodelist = $self->eqreadings( $group ); if( @$nodelist == 1 && $nodelist->[0] eq $node ) { - #print STDERR "Removing equivalence $group for $node\n"; + $self->equivalence_graph->delete_vertex( $group ); $self->remove_eqreadings( $group ); + $self->remove_equivalence( $group ); } elsif( @$nodelist == 1 ) { - warn "DATA INCONSISTENCY in equivalence graph: " . $nodelist->[0] . - " in group that should have only $node"; + throw( "DATA INCONSISTENCY in equivalence graph: " . $nodelist->[0] . + " in group that should have only $node" ); } else { - #print STDERR "Removing $node from equivalence $group\n"; - my @newlist = grep { $_ ne $node } @$nodelist; + my @newlist = grep { $_ ne $node } @$nodelist; $self->set_eqreadings( $group, \@newlist ); $self->remove_equivalence( $node ); } @@ -761,7 +828,8 @@ sub _remove_equivalence_node { =head2 add_equivalence_edge -Return the relationship object, if any, that exists between two readings. +Add an edge in the equivalence graph corresponding to $source -> $target in the +collation. Should only be called by Collation. =cut @@ -769,13 +837,13 @@ sub add_equivalence_edge { my( $self, $source, $target ) = @_; my $seq = $self->equivalence( $source ); my $teq = $self->equivalence( $target ); - #print STDERR "Adding equivalence edge $seq -> $teq for $source -> $target\n"; $self->equivalence_graph->add_edge( $seq, $teq ); } -=head2 add_equivalence_edge +=head2 delete_equivalence_edge -Return the relationship object, if any, that exists between two readings. +Remove an edge in the equivalence graph corresponding to $source -> $target in the +collation. Should only be called by Collation. =cut @@ -783,7 +851,6 @@ sub delete_equivalence_edge { my( $self, $source, $target ) = @_; my $seq = $self->equivalence( $source ); my $teq = $self->equivalence( $target ); - #print STDERR "Deleting equivalence edge $seq -> $teq for $source -> $target\n"; $self->equivalence_graph->delete_edge( $seq, $teq ); } @@ -793,23 +860,16 @@ sub _is_disconnected { || scalar $self->equivalence_graph->successorless_vertices > 1 ); } -=head2 make_equivalence - -Equate two readings in the equivalence graph. Should only be called internally. - -=cut - -sub make_equivalence { +# Equate two readings in the equivalence graph +sub _make_equivalence { my( $self, $source, $target ) = @_; # Get the source equivalent readings my $seq = $self->equivalence( $source ); my $teq = $self->equivalence( $target ); # Nothing to do if they are already equivalent... return if $seq eq $teq; - #print STDERR "Making equivalence for $source -> $target\n"; my $sourcepool = $self->eqreadings( $seq ); # and add them to the target readings. - # print STDERR "Moving readings '@$sourcepool' from group $seq to $teq\n"; push( @{$self->eqreadings( $teq )}, @$sourcepool ); map { $self->set_equivalence( $_, $teq ) } @$sourcepool; # Then merge the nodes in the equivalence graph. @@ -820,14 +880,15 @@ sub make_equivalence { $self->equivalence_graph->add_edge( $teq, $succ ); } $self->equivalence_graph->delete_vertex( $seq ); -# throw( "Graph got disconnected making $source / $target equivalence" ) -# if $self->_is_disconnected; + # TODO enable this after collation parsing is done + throw( "Graph got disconnected making $source / $target equivalence" ) + if $self->_is_disconnected && $self->collation->tradition->_initialized; } =head2 test_equivalence -Test whether, if two readings were equated with a relationship, the graph would -still be valid. +Test whether, if two readings were equated with a 'colocated' relationship, +the graph would still be valid. =cut @@ -885,13 +946,8 @@ sub test_equivalence { return $ret; } -=head2 break_equivalence - -Unmake an equivalence link between two readings. Should only be called internally. - -=cut - -sub break_equivalence { +# Unmake an equivalence link between two readings. Should only be called internally. +sub _break_equivalence { my( $self, $source, $target ) = @_; # This is the hard one. Need to reconstruct the equivalence groups without @@ -901,19 +957,15 @@ sub break_equivalence { map { $tng{$_} = 1 } $self->_find_equiv_without( $target, $source ); # If these groups intersect, they are still connected; do nothing. foreach my $el ( keys %tng ) { - if( exists $sng{$el} ) { - #print STDERR "Equivalence break $source / $target is a noop\n"; - return; - } + return if( exists $sng{$el} ); } - #print STDERR "Breaking equivalence $source / $target\n"; # If they don't intersect, then we split the nodes in the graph and in # the hashes. First figure out which group has which name - my $oldgroup = $self->equivalence( $source ); # eq for $target - my $swapped = $oldgroup eq $source; - my $newgroup = $swapped ? $target : $source; + my $oldgroup = $self->equivalence( $source ); # same as $target + my $keepsource = $sng{$oldgroup}; + my $newgroup = $keepsource ? $target : $source; my( $oldmembers, $newmembers ); - if( $swapped ) { + if( $keepsource ) { $oldmembers = [ keys %sng ]; $newmembers = [ keys %tng ]; } else { @@ -923,6 +975,9 @@ sub break_equivalence { # First alter the old group in the hash $self->set_eqreadings( $oldgroup, $oldmembers ); + foreach my $el ( @$oldmembers ) { + $self->set_equivalence( $el, $oldgroup ); + } # then add the new group back to the hash with its new key $self->set_eqreadings( $newgroup, $newmembers ); @@ -936,9 +991,11 @@ sub break_equivalence { my $c = $self->collation; foreach my $rdg ( @$newmembers ) { foreach my $rp ( $c->sequence->predecessors( $rdg ) ) { + next unless $self->equivalence( $rp ); $self->equivalence_graph->add_edge( $self->equivalence( $rp ), $newgroup ); } foreach my $rs ( $c->sequence->successors( $rdg ) ) { + next unless $self->equivalence( $rs ); $self->equivalence_graph->add_edge( $newgroup, $self->equivalence( $rs ) ); } } @@ -947,9 +1004,11 @@ sub break_equivalence { my( %old_pred, %old_succ ); foreach my $rdg ( @$oldmembers ) { foreach my $rp ( $c->sequence->predecessors( $rdg ) ) { + next unless $self->equivalence( $rp ); $old_pred{$self->equivalence( $rp )} = 1; } foreach my $rs ( $c->sequence->successors( $rdg ) ) { + next unless $self->equivalence( $rs ); $old_succ{$self->equivalence( $rs )} = 1; } } @@ -963,8 +1022,9 @@ sub break_equivalence { $self->equivalence_graph->delete_edge( $oldgroup, $s ); } } -# throw( "Graph got disconnected breaking $source / $target equivalence" ) -# if $self->_is_disconnected; + # TODO enable this after collation parsing is done + throw( "Graph got disconnected breaking $source / $target equivalence" ) + if $self->_is_disconnected && $self->collation->tradition->_initialized; } sub _find_equiv_without { @@ -988,6 +1048,101 @@ sub _find_equiv_without { return keys %found; } +=head2 rebuild_equivalence + +(Re)build the equivalence graph from scratch. Dumps the graph, makes a new one, +adds all readings and edges, then makes an equivalence for all relationships. + +=cut + +sub rebuild_equivalence { + my $self = shift; + my $newgraph = Graph->new(); + # Set this as the new equivalence graph + $self->_reset_equivalence( $newgraph ); + # Clear out the data hashes + $self->_clear_equivalence; + $self->_clear_eqreadings; + + $self->collation->tradition->_init_done(0); + # Add the readings + foreach my $r ( $self->collation->readings ) { + my $rid = $r->id; + $newgraph->add_vertex( $rid ); + $self->set_equivalence( $rid, $rid ); + $self->set_eqreadings( $rid, [ $rid ] ); + } + + # Now add the edges + foreach my $e ( $self->collation->paths ) { + $self->add_equivalence_edge( @$e ); + } + + # Now equate the colocated readings. This does no testing; + # it assumes that all preexisting relationships are valid. + foreach my $rel ( $self->relationships ) { + my $relobj = $self->get_relationship( $rel ); + next unless $relobj && $relobj->colocated; + $self->_make_equivalence( @$rel ); + } + $self->collation->tradition->_init_done(1); +} + +=head2 equivalence_ranks + +Rank all vertices in the equivalence graph, and return a hash reference with +vertex => rank mapping. + +=cut + +sub equivalence_ranks { + my $self = shift; + my $eqstart = $self->equivalence( $self->collation->start ); + my $eqranks = { $eqstart => 0 }; + my $rankeqs = { 0 => [ $eqstart ] }; + my @curr_origin = ( $eqstart ); + # A little iterative function. + while( @curr_origin ) { + @curr_origin = $self->_assign_rank( $eqranks, $rankeqs, @curr_origin ); + } + return( $eqranks, $rankeqs ); +} + +sub _assign_rank { + my( $self, $node_ranks, $rank_nodes, @current_nodes ) = @_; + my $graph = $self->equivalence_graph; + # Look at each of the children of @current_nodes. If all the child's + # parents have a rank, assign it the highest rank + 1 and add it to + # @next_nodes. Otherwise skip it; we will return when the highest-ranked + # parent gets a rank. + my @next_nodes; + foreach my $c ( @current_nodes ) { + warn "Current reading $c has no rank!" + unless exists $node_ranks->{$c}; + foreach my $child ( $graph->successors( $c ) ) { + next if exists $node_ranks->{$child}; + my $highest_rank = -1; + my $skip = 0; + foreach my $parent ( $graph->predecessors( $child ) ) { + if( exists $node_ranks->{$parent} ) { + $highest_rank = $node_ranks->{$parent} + if $highest_rank <= $node_ranks->{$parent}; + } else { + $skip = 1; + last; + } + } + next if $skip; + my $c_rank = $highest_rank + 1; + # print STDERR "Assigning rank $c_rank to node $child \n"; + $node_ranks->{$child} = $c_rank if $node_ranks; + push( @{$rank_nodes->{$c_rank}}, $child ) if $rank_nodes; + push( @next_nodes, $child ); + } + } + return @next_nodes; +} + ### Output logic sub _as_graphml {