use Graph::Easy;
use IPC::Run qw( run binary );
use Text::Tradition::Collation::Reading;
+use Text::Tradition::Collation::Path;
+use XML::LibXML;
use Moose;
has 'graph' => (
path => 'edge',
readings => 'nodes',
paths => 'edges',
+ relationships => 'edges',
},
default => sub { Graph::Easy->new( undirected => 0 ) },
);
predicate => 'has_svg',
);
-has 'graphviz' => (
- is => 'ro',
- isa => 'Str',
- writer => '_save_graphviz',
- predicate => 'has_graphviz',
- );
-
has 'graphml' => (
is => 'ro',
- isa => 'XML::LibXML::Document',
+ isa => 'Str',
writer => '_save_graphml',
predicate => 'has_graphml',
);
default => 'base text',
);
+has 'collapsed' => (
+ is => 'rw',
+ isa => 'Bool',
+ );
+
+has 'linear' => (
+ is => 'rw',
+ isa => 'Bool',
+ default => 1,
+ );
+
+has 'ac_label' => (
+ is => 'rw',
+ isa => 'Str',
+ default => ' (a.c.)',
+ );
+
+
# The collation can be created two ways:
# 1. Collate a set of witnesses (with CollateX I guess) and process
# the results as in 2.
sub BUILD {
my( $self, $args ) = @_;
$self->graph->use_class('node', 'Text::Tradition::Collation::Reading');
+ $self->graph->use_class('edge', 'Text::Tradition::Collation::Path');
# Pass through any graph-specific options.
my $shape = exists( $args->{'shape'} ) ? $args->{'shape'} : 'ellipse';
$target = $self->reading( $target )
unless ref( $target ) eq 'Text::Tradition::Collation::Reading';
foreach my $path ( $source->edges_to( $target ) ) {
- if( $path->label eq $wit ) {
+ if( $path->label eq $wit && $path->class eq 'edge.path' ) {
return;
}
}
$self->$orig( @_ );
};
+# Wrapper around paths
+around paths => sub {
+ my $orig = shift;
+ my $self = shift;
+
+ my @result = grep { $_->class eq 'edge.path' } $self->$orig( @_ );
+ return @result;
+};
+
+around relationships => sub {
+ my $orig = shift;
+ my $self = shift;
+ my @result = grep { $_->class eq 'edge.relationship' } $self->$orig( @_ );
+ return @result;
+};
+
# Wrapper around merge_nodes
sub merge_readings {
return $self->graph->merge_nodes( @_ );
}
+# Extra graph-alike utility
+sub has_path {
+ my( $self, $source, $target, $label ) = @_;
+ my @paths = $source->edges_to( $target );
+ my @relevant = grep { $_->label eq $label } @paths;
+ return scalar @paths;
+}
+
+## Dealing with relationships between readings. This is a different
+## sort of graph edge.
+
+sub add_relationship {
+ my( $self, $type, $source, $target, $global ) = @_;
+
+ # Make sure there is not another relationship between these two
+ # readings already
+ $source = $self->reading( $source )
+ unless ref( $source ) eq 'Text::Tradition::Collation::Reading';
+ $target = $self->reading( $target )
+ unless ref( $target ) eq 'Text::Tradition::Collation::Reading';
+ foreach my $rel ( $source->edges_to( $target ) ) {
+ if( $rel->label eq $type && $rel->class eq 'edge.relationship' ) {
+ return;
+ }
+ }
+
+ my $rel = Text::Tradition::Collation::Relationship->new(
+ 'sort' => $type,
+ 'global' => $global,
+ 'orig_relation' => [ $source, $target ],
+ );
+ print STDERR sprintf( "Setting relationship %s between readings %s (%s)"
+ . " and %s (%s)\n", $type,
+ $source->label, $source->name,
+ $target->label, $target->name );
+ $self->graph->add_edge( $source, $target, $rel );
+ if( $global ) {
+ # Look for all readings with the source label, and if there are
+ # colocated readings with the target label, join them too.
+ foreach my $r ( $self->readings() ) {
+ next unless $r->label eq $source->label;
+ my @colocated = grep { $_->label eq $target->label }
+ $self->same_position_as( $r );
+ if( @colocated ) {
+ warn "Multiple readings with same label at same position!"
+ if @colocated > 1;
+ my $dup_rel = Text::Tradition::Collation::Relationship->new(
+ 'sort' => $type,
+ 'global' => $global,
+ 'orig_relation' => [ $source, $target ],
+ );
+ $self->graph->add_edge( $r, $colocated[0], $dup_rel );
+ }
+ }
+ }
+}
+
=head2 Output method(s)
=over
my( $self, $recalc ) = @_;
return $self->svg if $self->has_svg;
- $self->_save_graphviz( $self->graph->as_graphviz() )
- unless( $self->has_graphviz && !$recalc );
+ $self->collapse_graph_paths();
my @cmd = qw/dot -Tsvg/;
my( $svg, $err );
- my $in = $self->graphviz;
+ my $in = $self->as_dot();
run( \@cmd, \$in, ">", binary(), \$svg );
- $self->{'svg'} = $svg;
+ $self->_save_svg( $svg );
+ $self->expand_graph_paths();
return $svg;
}
+=item B<as_dot>
+
+print $graph->as_dot( $view, $recalculate );
+
+Returns a string that is the collation graph expressed in dot
+(i.e. GraphViz) format. The 'view' argument determines what kind of
+graph is produced.
+ * 'path': a graph of witness paths through the collation (DEFAULT)
+ * 'relationship': a graph of how collation readings relate to
+ each other
+
+=cut
+
+sub as_dot {
+ my( $self, $view ) = @_;
+ $view = 'path' unless $view;
+ # TODO consider making some of these things configurable
+ my $dot = sprintf( "digraph %s {\n", $self->tradition->name );
+ $dot .= "\tedge [ arrowhead=open ];\n";
+ $dot .= "\tgraph [ rankdir=LR ];\n";
+ $dot .= sprintf( "\tnode [ fontsize=%d, fillcolor=%s, style=%s, shape=%s ];\n",
+ 11, "white", "filled", $self->graph->get_attribute( 'node', 'shape' ) );
+
+ foreach my $reading ( $self->readings ) {
+ next if $reading->name eq $reading->label;
+ $dot .= sprintf( "\t\"%s\" [ label=\"%s\" ]\n", $reading->name, $reading->label );
+ }
+
+ my @edges = $view eq 'relationship' ? $self->relationships : $self->paths;
+ foreach my $edge ( @edges ) {
+ $dot .= sprintf( "\t\"%s\" -> \"%s\" [ color=\"%s\", fontcolor=\"%s\", label=\"%s\" ]\n",
+ $edge->from->name, $edge->to->name, '#000000', '#000000', $edge->label );
+ }
+
+ $dot .= "}\n";
+ return $dot;
+}
+
=item B<as_graphml>
print $graph->as_graphml( $recalculate )
$root->setAttributeNS( $xsi_ns, 'schemaLocation', $graphml_schema );
# Add the data keys for nodes
- my @node_data = ( 'name', 'token', 'identical', 'position' );
- foreach my $ndi ( 0 .. $#node_data ) {
+ my @node_data = ( 'name', 'reading', 'identical', 'position' );
+ # HACKY HACKY HACK Relationship data
+ my %node_data_keys;
+ my $ndi = 0;
+ foreach my $datum ( @node_data ) {
+ $node_data_keys{$datum} = 'dn'.$ndi++;
my $key = $root->addNewChild( $graphml_ns, 'key' );
- $key->setAttribute( 'attr.name', $node_data[$ndi] );
+ $key->setAttribute( 'attr.name', $datum );
$key->setAttribute( 'attr.type', 'string' );
$key->setAttribute( 'for', 'node' );
- $key->setAttribute( 'id', 'd'.$ndi );
+ $key->setAttribute( 'id', $node_data_keys{$datum} );
}
- # Add the data keys for edges
- my %wit_hash;
- my $wit_ctr = 0;
- foreach my $wit ( $self->getWitnessList ) {
- my $wit_key = 'w' . $wit_ctr++;
- $wit_hash{$wit} = $wit_key;
+ # Add the data keys for edges, i.e. witnesses
+ my $edi = 0;
+ my %edge_data_keys;
+ foreach my $edge_key( qw/ witness_main witness_ante_corr relationship / ) {
+ $edge_data_keys{$edge_key} = 'de'.$edi++;
my $key = $root->addNewChild( $graphml_ns, 'key' );
- $key->setAttribute( 'attr.name', $wit );
+ $key->setAttribute( 'attr.name', $edge_key );
$key->setAttribute( 'attr.type', 'string' );
$key->setAttribute( 'for', 'edge' );
- $key->setAttribute( 'id', $wit_key );
+ $key->setAttribute( 'id', $edge_data_keys{$edge_key} );
}
-
+
# Add the graph, its nodes, and its edges
my $graph = $root->addNewChild( $graphml_ns, 'graph' );
$graph->setAttribute( 'edgedefault', 'directed' );
$graph->setAttribute( 'id', 'g0' ); # TODO make this meaningful
$graph->setAttribute( 'parse.edgeids', 'canonical' );
- $graph->setAttribute( 'parse.edges', $self->edges() );
+ $graph->setAttribute( 'parse.edges', scalar($self->paths) );
$graph->setAttribute( 'parse.nodeids', 'canonical' );
- $graph->setAttribute( 'parse.nodes', $self->nodes() );
+ $graph->setAttribute( 'parse.nodes', scalar($self->readings) );
$graph->setAttribute( 'parse.order', 'nodesfirst' );
my $node_ctr = 0;
my %node_hash;
- foreach my $n ( $self->readings ) {
+ foreach my $n ( sort { $a->name cmp $b->name } $self->readings ) {
my %this_node_data = ();
- foreach my $ndi ( 0 .. $#node_data ) {
- my $value;
- $this_node_data{'d'.$ndi} = $n->name if $node_data[$ndi] eq 'name';
- $this_node_data{'d'.$ndi} = $n->label
- if $node_data[$ndi] eq 'token';
- $this_node_data{'d'.$ndi} = $n->primary->name if $n->has_primary;
- $this_node_data{'d'.$ndi} =
- $self->{'positions'}->node_position( $n )
- if $node_data[$ndi] eq 'position';
+ foreach my $datum ( @node_data ) {
+ my $key = $node_data_keys{$datum};
+ if( $datum eq 'name' ) {
+ $this_node_data{$key} = $n->name;
+ } elsif( $datum eq 'reading' ) {
+ $this_node_data{$key} = $n->label;
+ } elsif( $datum eq 'identical' && $n->has_primary ) {
+ $this_node_data{$key} = $n->primary->name;
+ } elsif( $datum eq 'position' ) {
+ $this_node_data{$key} = $n->position;
+ }
}
my $node_el = $graph->addNewChild( $graphml_ns, 'node' );
my $node_xmlid = 'n' . $node_ctr++;
foreach my $dk ( keys %this_node_data ) {
my $d_el = $node_el->addNewChild( $graphml_ns, 'data' );
$d_el->setAttribute( 'key', $dk );
- $d_el->appendTextChild( $this_node_data{$dk} );
+ $d_el->appendText( $this_node_data{$dk} );
}
}
- foreach my $e ( $self->edges() ) {
- my( $name, $from, $to ) = ( $e->name,
- $node_hash{ $e->from()->name() },
- $node_hash{ $e->to()->name() } );
+ # Add the path edges
+ my $edge_ctr = 0;
+ foreach my $e ( sort { $a->from->name cmp $b->from->name } $self->graph->edges() ) {
+ my( $name, $from, $to ) = ( 'e'.$edge_ctr++,
+ $node_hash{ $e->from->name() },
+ $node_hash{ $e->to->name() } );
my $edge_el = $graph->addNewChild( $graphml_ns, 'edge' );
$edge_el->setAttribute( 'source', $from );
$edge_el->setAttribute( 'target', $to );
$edge_el->setAttribute( 'id', $name );
- # TODO Got to add the witnesses
+ if( $e->class() eq 'edge.path' ) {
+ # It's a witness path, so add the witness
+ my $base = $e->label;
+ my $key = $edge_data_keys{'witness_main'};
+ # TODO kind of hacky
+ if( $e->label =~ /^(.*?)\s+(\(a\.c\.\))$/ ) {
+ $base = $1;
+ $key = $edge_data_keys{'witness_ante_corr'};
+ }
+ my $wit_el = $edge_el->addNewChild( $graphml_ns, 'data' );
+ $wit_el->setAttribute( 'key', $key );
+ $wit_el->appendText( $base );
+ } else {
+ # It's a relationship
+ my $rel_el = $edge_el->addNewChild( $graphml_ns, 'data' );
+ $rel_el->setAttribute( 'key', $edge_data_keys{'relationship'} );
+ $rel_el->appendText( $e->label() );
+ }
}
# Return the thing
- $self->_save_graphml( $graphml );
- return $graphml;
+ $self->_save_graphml( $graphml->toString(1) );
+ return $graphml->toString(1);
+}
+
+sub _make_xml_attr {
+ my $str = shift;
+ $str =~ s/\s/_/g;
+ $str =~ s/\W//g;
+ $str =~ "a$str" if $str =~ /^\d/;
+ return $str;
+}
+
+sub collapse_graph_paths {
+ my $self = shift;
+ # Our collation graph has an path per witness. This is great for
+ # calculation purposes, but terrible for display. Thus we want to
+ # display only one path between any two nodes.
+
+ return if $self->collapsed;
+
+ print STDERR "Collapsing witness paths in graph...\n";
+
+ # Don't list out every witness if we have more than half to list.
+ my $majority = int( scalar( @{$self->tradition->witnesses} ) / 2 ) + 1;
+ # But don't compress if there are only a few witnesses.
+ $majority = 4 if $majority < 4;
+ foreach my $node( $self->readings ) {
+ my $newlabels = {};
+ # We will visit each node, so we only look ahead.
+ foreach my $edge ( $node->outgoing() ) {
+ next unless $edge->class eq 'edge.path';
+ add_hash_entry( $newlabels, $edge->to->name, $edge->name );
+ $self->del_path( $edge );
+ }
+
+ foreach my $newdest ( keys %$newlabels ) {
+ my $label;
+ my @compressed_wits = ();
+ if( @{$newlabels->{$newdest}} < $majority ) {
+ $label = join( ', ', sort( @{$newlabels->{$newdest}} ) );
+ } else {
+ ## TODO FIX THIS HACK
+ my @aclabels;
+ foreach my $wit ( @{$newlabels->{$newdest}} ) {
+ if( $wit =~ /^(.*?)(\s*\(?a\.\s*c\.\)?)$/ ) {
+ push( @aclabels, $wit );
+ } else {
+ push( @compressed_wits, $wit );
+ }
+ }
+ $label = join( ', ', 'majority', sort( @aclabels ) );
+ }
+
+ my $newpath =
+ $self->add_path( $node, $self->reading( $newdest ), $label );
+ if( @compressed_wits ) {
+ $newpath->hidden_witnesses( \@compressed_wits );
+ }
+ }
+ }
+
+ $self->collapsed( 1 );
+}
+
+sub expand_graph_paths {
+ my $self = shift;
+ # Our collation graph has only one path between any two nodes.
+ # This is great for display, but not so great for analysis.
+ # Expand this so that each witness has its own path between any
+ # two reading nodes.
+ return unless $self->collapsed;
+
+ print STDERR "Expanding witness paths in graph...\n";
+ foreach my $path( $self->paths ) {
+ my $from = $path->from;
+ my $to = $path->to;
+ my @wits = split( /, /, $path->label );
+ if( $path->has_hidden_witnesses ) {
+ push( @wits, @{$path->hidden_witnesses} );
+ }
+ $self->del_path( $path );
+ foreach ( @wits ) {
+ $self->add_path( $from, $to, $_ );
+ }
+ }
+ $self->collapsed( 0 );
}
=back
# An alternative to walk_witness_paths, for use when a collation is
-# constructed from a base text and an apparatus. Also modifies the
-# collation graph to remove all 'base text' paths and replace them
-# with real witness paths.
+# constructed from a base text and an apparatus. We have the
+# sequences of readings and just need to add path edges.
-sub walk_and_expand_base {
- my( $self, $end ) = @_;
+sub make_witness_paths {
+ my( $self ) = @_;
my @common_readings;
foreach my $wit ( @{$self->tradition->witnesses} ) {
- my $sig = $wit->sigil;
- my $post_sig;
- $post_sig = $wit->post_correctione
- if $wit->has_post_correctione;
-
- # $DB::single = 1 if $wit->sigil eq 'Vb11';
- my @wit_path = $self->reading_sequence( $self->start, $end, $sig );
- $wit->path( \@wit_path );
- $self->connect_readings_for_witness( $wit );
- @common_readings = _find_common( \@common_readings, \@wit_path );
-
- # If there is a post-correctio, get its path and compare.
- # Add a correction range for each divergence.
- if( $post_sig ) {
- my @corr_wit_path = $self->reading_sequence( $self->start, $end,
- "$sig$post_sig", $sig );
-
- # Map ante-corr readings to their indices
- my %in_orig;
- my $i = 0;
- map { $in_orig{$_->name} = $i++ } @wit_path;
-
- # Look for divergences
- my $diverged = 0;
- my $last_common;
- my @correction;
- $DB::single = 1 if $sig eq 'Vb12';
- foreach my $rdg ( @corr_wit_path ) {
- if( exists( $in_orig{$rdg->name} ) && !$diverged ) {
- # We are reading the same here
- $last_common = $in_orig{$rdg->name};
- } elsif ( exists( $in_orig{$rdg->name} ) ) {
- # We have been diverging but are reading the same again.
- # Add the correction to the witness.
- my $offset = $last_common + 1;
- my $length = $in_orig{$rdg->name} - $offset;
- $wit->add_correction( $offset, $length, @correction );
- $diverged = 0;
- @common_readings = _remove_common( \@common_readings, \@correction );
- @correction = ();
- $last_common = $in_orig{$rdg->name};
- } elsif( $diverged ) {
- # We are in the middle of a divergence.
- push( @correction, $rdg );
- } else {
- # We have started to diverge. Note it.
- $diverged = 1;
- push( @correction, $rdg );
- }
- }
- # Add any divergence that is at the end of the text
- if( $diverged ) {
- $wit->add_correction( $last_common+1, $#wit_path, \@correction );
- }
- }
- }
-
- # Remove any 'base text' paths.
- foreach my $path ( $self->paths ) {
- $self->del_path( $path )
- if $path->label eq $self->baselabel;
+ print STDERR "Making path for " . $wit->sigil . "\n";
+ $self->make_witness_path( $wit );
+ @common_readings = _find_common( \@common_readings, $wit->path );
+ @common_readings = _find_common( \@common_readings, $wit->uncorrected_path );
}
+ return @common_readings;
}
-sub connect_readings_for_witness {
+sub make_witness_path {
my( $self, $wit ) = @_;
my @chain = @{$wit->path};
+ my $sig = $wit->sigil;
foreach my $idx ( 0 .. $#chain-1 ) {
- $self->add_path( $chain[$idx], $chain[$idx+1], $wit->sigil );
+ $self->add_path( $chain[$idx], $chain[$idx+1], $sig );
+ }
+ @chain = @{$wit->uncorrected_path};
+ foreach my $idx( 0 .. $#chain-1 ) {
+ my $source = $chain[$idx];
+ my $target = $chain[$idx+1];
+ $self->add_path( $source, $target, "$sig (a.c.)" )
+ unless $self->has_path( $source, $target, $sig );
}
}
foreach my $wit ( @{$self->tradition->witnesses} ) {
print STDERR "Calculating positions in " . $wit->sigil . "\n";
_update_positions_from_path( $wit->path, @ordered_common );
- _update_positions_from_path( $wit->corrected_path, @ordered_common )
- if $wit->has_post_correctione;
+ _update_positions_from_path( $wit->uncorrected_path, @ordered_common )
+ if $wit->has_ante_corr;
}
# DEBUG
# that corresponds to its eventual position identifier. Common
# nodes always start a new row, and are thus always in the first
# column.
-
my $wit_matrix = [];
my $cn = 0; # We should hit the common readings in order.
my $row = [];
foreach my $di ( 1..scalar(@{$wit_matrix->[$li-1]}) ) {
my $reading = $wit_matrix->[$li-1]->[$di-1];
my $position = "$li,$di";
+
# If we have seen this node before, we need to compare
# its position with what went before.
unless( $reading->has_position &&
return values( %h );
}
+sub add_hash_entry {
+ my( $hash, $key, $entry ) = @_;
+ if( exists $hash->{$key} ) {
+ push( @{$hash->{$key}}, $entry );
+ } else {
+ $hash->{$key} = [ $entry ];
+ }
+}
+
no Moose;
__PACKAGE__->meta->make_immutable;
projects / scpubgit/stemmatology.git / blobdiff