load XML::LibXML only when required; handle global relationships more correctly;...
[scpubgit/stemmatology.git] / lib / Text / Tradition / Collation / RelationshipStore.pm
index bbfc50d..4342bd9 100644 (file)
@@ -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' } );
@@ -196,18 +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 )
@@ -241,8 +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;
        }
@@ -260,10 +255,16 @@ add_relationship.
 
 =begin testing
 
+use Test::Warn;
 use Text::Tradition;
 use TryCatch;
 
-my $t1 = Text::Tradition->new( 'input' => 'Self', 'file' => 't/data/legendfrag.xml' );
+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;
@@ -305,9 +306,23 @@ try {
                "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( 'r9.2', 'r9.3', { 'type' => 'lexical' } );
 my $trel2 = $c2->get_relationship( 'r9.2', 'r9.3' );
@@ -408,27 +423,21 @@ sub add_relationship {
        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 );
@@ -451,19 +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 
@@ -564,7 +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...
@@ -695,6 +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 ];
@@ -760,10 +813,12 @@ sub _remove_equivalence_node {
        my $group = $self->equivalence( $node );
        my $nodelist = $self->eqreadings( $group );
        if( @$nodelist == 1 && $nodelist->[0] eq $node ) {
+               $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 {
                my @newlist = grep { $_ ne $node } @$nodelist;
                $self->set_eqreadings( $group, \@newlist );
@@ -826,8 +881,8 @@ sub _make_equivalence {
        }
        $self->equivalence_graph->delete_vertex( $seq );
        # TODO enable this after collation parsing is done
-#      throw( "Graph got disconnected making $source / $target equivalence" )
-#              if $self->_is_disconnected;
+       throw( "Graph got disconnected making $source / $target equivalence" )
+               if $self->_is_disconnected && $self->collation->tradition->_initialized;
 }
 
 =head2 test_equivalence
@@ -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;
                }
        }
@@ -964,8 +1023,8 @@ sub _break_equivalence {
                }
        }
        # TODO enable this after collation parsing is done
-#      throw( "Graph got disconnected breaking $source / $target equivalence" )
-#              if $self->_is_disconnected;
+       throw( "Graph got disconnected breaking $source / $target equivalence" )
+               if $self->_is_disconnected && $self->collation->tradition->_initialized;
 }
 
 sub _find_equiv_without {
@@ -1005,6 +1064,7 @@ sub rebuild_equivalence {
        $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;
@@ -1025,6 +1085,62 @@ sub rebuild_equivalence {
                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