start handling CTE output files

[scpubgit/stemmatology.git] / lib / Text / Tradition / Collation.pm

package Text::Tradition::Collation;

use Encode qw( decode_utf8 );
use File::Temp;
use Graph::Easy;
use IPC::Run qw( run binary );
use Text::CSV_XS;
use Text::Tradition::Collation::Path;
use Text::Tradition::Collation::Reading;
use Text::Tradition::Collation::Relationship;
use Text::Tradition::Collation::Segment;
use XML::LibXML;
use Moose;

has 'graph' => (
    is => 'ro',
    isa => 'Graph::Easy',
    handles => {
        add_reading => 'add_node',
        add_lacuna => 'add_node',
        del_reading => 'del_node',
        del_segment => 'del_node',
        add_path => 'add_edge',
        del_path => 'del_edge',
        reading => 'node',
        path => 'edge',
        readings => 'nodes',
        segments => 'nodes',
        paths => 'edges',
        relationships => 'edges',
    },
    default => sub { Graph::Easy->new( undirected => 0 ) },
    );
                

has 'tradition' => (  # TODO should this not be ro?
    is => 'rw',
    isa => 'Text::Tradition',
    );

has 'svg' => (
    is => 'ro',
    isa => 'Str',
    writer => '_save_svg',
    predicate => 'has_svg',
    );

has 'graphml' => (
    is => 'ro',
    isa => 'Str',
    writer => '_save_graphml',
    predicate => 'has_graphml',
    );

has 'csv' => (
    is => 'ro',
    isa => 'Str',
    writer => '_save_csv',
    predicate => 'has_csv',
    );

# Keeps track of the lemmas within the collation.  At most one lemma
# per position in the graph.
has 'lemmata' => (
    is => 'ro',
    isa => 'HashRef[Maybe[Str]]',
    default => sub { {} },
    );

has 'wit_list_separator' => (
    is => 'rw',
    isa => 'Str',
    default => ', ',
    );

has 'baselabel' => (
    is => 'rw',
    isa => 'Str',
    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.
# 2. Read a pre-prepared collation in one of a variety of formats,
#    and make the graph from that.

# The graph itself will (for now) be immutable, and the positions
# within the graph will also be immutable.  We need to calculate those
# positions upon graph construction.  The equivalences between graph
# nodes will be mutable, entirely determined by the user (or possibly
# by some semantic pre-processing provided by the user.)  So the
# constructor should just make an empty equivalences object.  The
# constructor will also need to make the witness objects, if we didn't
# come through option 1.

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';
    $self->graph->set_attribute( 'node', 'shape', $shape );
}

around add_lacuna => sub {
    my $orig = shift;
    my $self = shift;
    my $id = shift @_;
    my $l = $self->$orig( '#LACUNA_' . $id . '#' );
    $l->is_lacuna( 1 );
    return $l;
};

# Wrapper around add_path 

around add_path => sub {
    my $orig = shift;
    my $self = shift;

    # Make sure there are three arguments
    unless( @_ == 3 ) {
        warn "Call add_path with args source, target, witness";
        return;
    }
    # Make sure the proposed path does not yet exist
    # NOTE 'reading' will currently return readings and segments
    my( $source, $target, $wit ) = @_;
    $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 $path ( $source->edges_to( $target ) ) {
        if( $path->label eq $wit && $path->class eq 'edge.path' ) {
            return;
        }
    }
    # Do the deed
    $self->$orig( @_ );
};

# Wrapper around paths
around paths => sub {
    my $orig = shift;
    my $self = shift;

    my @result = grep { $_->sub_class eq 'path' } $self->$orig( @_ );
    return @result;
};

around relationships => sub {
    my $orig = shift;
    my $self = shift;
    my @result = grep { $_->sub_class eq 'relationship' } $self->$orig( @_ );
    return @result;
};

around readings => sub {
    my $orig = shift;
    my $self = shift;
    my @result = grep { $_->sub_class ne 'segment' } $self->$orig( @_ );
    return @result;
};

around segments => sub {
    my $orig = shift;
    my $self = shift;
    my @result = grep { $_->sub_class eq 'segment' } $self->$orig( @_ );
    return @result;
};

# Wrapper around merge_nodes

sub merge_readings {
    my $self = shift;
    my $first_node = shift;
    my $second_node = shift;
    $first_node->merge_from( $second_node );
    unshift( @_, $first_node, $second_node );
    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 @relevant;
}

## Dealing with groups of readings, i.e. segments.

sub add_segment {
    my( $self, @items ) = @_;
    my $segment = Text::Tradition::Collation::Segment->new( 'members' => \@items );
    return $segment;
}

## Dealing with relationships between readings.  This is a different
## sort of graph edge.  Return a success/failure value and a list of
## node pairs that have been linked.

sub add_relationship {
    my( $self, $source, $target, $options ) = @_;

    # Make sure there is not another relationship between these two
    # readings or segments already
    $source = $self->reading( $source )
        unless ref( $source ) && $source->isa( 'Graph::Easy::Node' );
    $target = $self->reading( $target )
        unless ref( $target ) && $target->isa( 'Graph::Easy::Node' );
    foreach my $rel ( $source->edges_to( $target ), $target->edges_to( $source ) ) {
        if( $rel->class eq 'edge.relationship' ) {
            return ( undef, "Relationship already exists between these readings" );
        }
    }
    if( $options->{'equal_rank'} && !relationship_valid( $source, $target ) ) {
        return ( undef, 'Relationship creates witness loop' );
    }

    # TODO Think about positional hilarity if relationships are added after positions
    # are assigned.
    
    my @joined = ( [ $source->name, $target->name ] );  # Keep track of the nodes we join.
    
    $options->{'this_relation'} = [ $source, $target ];
    my $rel;
    eval { $rel = Text::Tradition::Collation::Relationship->new( %$options ) };
    if( $@ ) {
       return ( undef, $@ );
    }
    $self->graph->add_edge( $source, $target, $rel );
    
    # TODO Handle global relationship setting

    return( 1, @joined );
}

sub relationship_valid {
    my( $source, $target ) = @_;
    # Check that linking the source and target in a relationship won't lead
    # to a path loop for any witness.
    my @proposed_related = ( $source, $target );
    push( @proposed_related, $source->related_readings );
    push( @proposed_related, $target->related_readings );
    my %pr_ids;
    map { $pr_ids{ $_->name } = 1 } @proposed_related;
    # The lists of 'in' and 'out' should not have any element that appears
    # in 'proposed_related'.
    foreach my $pr ( @proposed_related ) {
        foreach my $e ( $pr->incoming ) {
            if( exists $pr_ids{ $e->from->name } ) {
                return 0;
            }
        }
        foreach my $e ( $pr->outgoing ) {
            if( exists $pr_ids{ $e->to->name } ) {
                return 0;
            }
        }
    }
    return 1;
}

=head2 Output method(s)

=over

=item B<as_svg>

print $graph->as_svg( $recalculate );

Returns an SVG string that represents the graph.  Uses GraphViz to do
this, because Graph::Easy doesn\'t cope well with long graphs. Unless
$recalculate is passed (and is a true value), the method will return a
cached copy of the SVG after the first call to the method.

=cut

sub as_svg {
    my( $self, $recalc ) = @_;
    return $self->svg if $self->has_svg;
    
    $self->collapse_graph_paths();
    
    my @cmd = qw/dot -Tsvg/;
    my( $svg, $err );
    my $dotfile = File::Temp->new();
    ## TODO REMOVE
    # $dotfile->unlink_on_destroy(0);
    binmode $dotfile, ':utf8';
    print $dotfile $self->as_dot();
    push( @cmd, $dotfile->filename );
    run( \@cmd, ">", binary(), \$svg );
    $svg = decode_utf8( $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 ) {
        # Need not output nodes without separate labels
        next if $reading->name eq $reading->label;
        # TODO output readings or segments, but not both
        next if $reading->class eq 'node.segment';
        $dot .= sprintf( "\t\"%s\" [ label=\"%s\" ];\n", $reading->name, $reading->label );
    }

    my @edges = $view eq 'relationship' ? $self->relationships : $self->paths;
    foreach my $edge ( @edges ) {
        my %variables = ( 'color' => '#000000',
                          'fontcolor' => '#000000',
                          'label' => $edge->label,
            );
        my $varopts = join( ', ', map { $_.'="'.$variables{$_}.'"' } sort keys %variables );
        $dot .= sprintf( "\t\"%s\" -> \"%s\" [ %s ];\n",
                         $edge->from->name, $edge->to->name, $varopts );
    }
    $dot .= "}\n";
    return $dot;
}

=item B<as_graphml>

print $graph->as_graphml( $recalculate )

Returns a GraphML representation of the collation graph, with
transposition information and position information. Unless
$recalculate is passed (and is a true value), the method will return a
cached copy of the SVG after the first call to the method.

=cut

sub as_graphml {
    my( $self, $recalc ) = @_;
    return $self->graphml if $self->has_graphml;

    # Some namespaces
    my $graphml_ns = 'http://graphml.graphdrawing.org/xmlns';
    my $xsi_ns = 'http://www.w3.org/2001/XMLSchema-instance';
    my $graphml_schema = 'http://graphml.graphdrawing.org/xmlns ' .
        'http://graphml.graphdrawing.org/xmlns/1.0/graphml.xsd';

    # Create the document and root node
    my $graphml = XML::LibXML->createDocument( "1.0", "UTF-8" );
    my $root = $graphml->createElementNS( $graphml_ns, 'graphml' );
    $graphml->setDocumentElement( $root );
    $root->setNamespace( $xsi_ns, 'xsi', 0 );
    $root->setAttributeNS( $xsi_ns, 'schemaLocation', $graphml_schema );

    # TODO Add some global graph data

    # Add the data keys for nodes
    my %node_data_keys;
    my $ndi = 0;
    foreach my $datum ( qw/ name reading identical rank class / ) {
        $node_data_keys{$datum} = 'dn'.$ndi++;
        my $key = $root->addNewChild( $graphml_ns, 'key' );
        $key->setAttribute( 'attr.name', $datum );
        $key->setAttribute( 'attr.type', 'string' );
        $key->setAttribute( 'for', 'node' );
        $key->setAttribute( 'id', $node_data_keys{$datum} );
    }

    # 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 class / ) {
        $edge_data_keys{$edge_key} = 'de'.$edi++;
        my $key = $root->addNewChild( $graphml_ns, 'key' );
        $key->setAttribute( 'attr.name', $edge_key );
        $key->setAttribute( 'attr.type', 'string' );
        $key->setAttribute( 'for', 'edge' );
        $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' );
    $graph->setAttribute( 'parse.edges', scalar($self->paths) );
    $graph->setAttribute( 'parse.nodeids', 'canonical' );
    $graph->setAttribute( 'parse.nodes', scalar($self->readings) );
    $graph->setAttribute( 'parse.order', 'nodesfirst' );

    my $node_ctr = 0;
    my %node_hash;
    # Add our readings to the graph
    foreach my $n ( sort { $a->name cmp $b->name } $self->readings ) {
        my $node_el = $graph->addNewChild( $graphml_ns, 'node' );
        my $node_xmlid = 'n' . $node_ctr++;
        $node_hash{ $n->name } = $node_xmlid;
        $node_el->setAttribute( 'id', $node_xmlid );
        _add_graphml_data( $node_el, $node_data_keys{'name'}, $n->name );
        _add_graphml_data( $node_el, $node_data_keys{'reading'}, $n->label );
        _add_graphml_data( $node_el, $node_data_keys{'rank'}, $n->rank )
            if $n->has_rank;
        _add_graphml_data( $node_el, $node_data_keys{'class'}, $n->sub_class );
        _add_graphml_data( $node_el, $node_data_keys{'identical'}, $n->primary->name )
            if $n->has_primary;
    }

    # Add any segments we have
    foreach my $n ( sort { $a->name cmp $b->name } $self->segments ) {
        my $node_el = $graph->addNewChild( $graphml_ns, 'node' );
        my $node_xmlid = 'n' . $node_ctr++;
        $node_hash{ $n->name } = $node_xmlid;
        $node_el->setAttribute( 'id', $node_xmlid );
        _add_graphml_data( $node_el, $node_data_keys{'class'}, $n->sub_class );
        _add_graphml_data( $node_el, $node_data_keys{'name'}, $n->name );
    }

    # Add the path, relationship, and segment 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 );
        # Add the edge class
        _add_graphml_data( $edge_el, $edge_data_keys{'class'}, $e->sub_class );
        if( $e->sub_class eq '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'};
            }
            _add_graphml_data( $edge_el, $key, $base );
        } elsif( $e->sub_class eq 'relationship' ) {
            # It's a relationship
            _add_graphml_data( $edge_el, $edge_data_keys{'relationship'}, $e->label );
        } # else a segment, nothing to record but source, target, class
    }

    # Return the thing
    $self->_save_graphml( $graphml->toString(1) );
    return $graphml->toString(1);
}

sub _add_graphml_data {
    my( $el, $key, $value ) = @_;
    my $data_el = $el->addNewChild( $el->namespaceURI, 'data' );
    return unless defined $value;
    $data_el->setAttribute( 'key', $key );
    $data_el->appendText( $value );
}

=item B<as_csv>

print $graph->as_csv( $recalculate )

Returns a CSV alignment table representation of the collation graph, one
row per witness (or witness uncorrected.) Unless $recalculate is passed
(and is a true value), the method will return a cached copy of the CSV
after the first call to the method.

=cut

sub as_csv {
    my( $self, $recalc ) = @_;
    return $self->csv if $self->has_csv;
    my $table = $self->make_alignment_table;
    my $csv = Text::CSV_XS->new( { binary => 1, quote_null => 0 } );    
    my @result;
    foreach my $row ( @$table ) {
        $csv->combine( @$row );
        push( @result, decode_utf8( $csv->string ) );
    }
    $self->_save_csv( join( "\n", @result ) );
    return $self->csv;
}

sub make_alignment_table {
    my $self = shift;
    unless( $self->linear ) {
        warn "Need a linear graph in order to make an alignment table";
        return;
    }
    my $table;
    my @all_pos = sort { $a <=> $b } $self->possible_positions;
    foreach my $wit ( $self->tradition->witnesses ) {
        # print STDERR "Making witness row(s) for " . $wit->sigil . "\n";
        my @row = _make_witness_row( $wit->path, \@all_pos );
        unshift( @row, $wit->sigil );
        push( @$table, \@row );
        if( $wit->has_ante_corr ) {
            my @ac_row = _make_witness_row( $wit->uncorrected_path, \@all_pos );
            unshift( @ac_row, $wit->sigil . $self->ac_label );
            push( @$table, \@ac_row );
        }           
    }
    # Return a table where the witnesses read in columns rather than rows.
    my $turned = _turn_table( $table );
    return $turned;
}

sub _make_witness_row {
    my( $path, $positions ) = @_;
    my %char_hash;
    map { $char_hash{$_} = undef } @$positions;
    foreach my $rdg ( @$path ) {
        my $rtext = $rdg->text;
        $rtext = '#LACUNA#' if $rdg->is_lacuna;
        $char_hash{$rdg->rank} = $rtext;
    }
    my @row = map { $char_hash{$_} } @$positions;
    # Fill in lacuna markers for undef spots in the row
    my $last_el = shift @row;
    my @filled_row = ( $last_el );
    foreach my $el ( @row ) {
        if( $last_el && $last_el eq '#LACUNA#' && !defined $el ) {
            $el = '#LACUNA#';
        }
        push( @filled_row, $el );
        $last_el = $el;
    }
    return @filled_row;
}

# Helper to turn the witnesses along columns rather than rows.  Assumes
# equal-sized rows.
sub _turn_table {
    my( $table ) = @_;
    my $result = [];
    return $result unless scalar @$table;
    my $nrows = scalar @{$table->[0]};
    foreach my $idx ( 0 .. $nrows - 1 ) {
        foreach my $wit ( 0 .. $#{$table} ) {
            $result->[$idx]->[$wit] = $table->[$wit]->[$idx];
        }
    }
    return $result;        
}


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 $label;
            my @compressed_wits = @{$newlabels->{$newdest}};
            if( @compressed_wits < $majority ) {
                $label = join( ', ', sort( @{$newlabels->{$newdest}} ) );
            } else {
                ## TODO FIX THIS HACK
                my @aclabels;
                foreach my $wit ( @compressed_wits ) {
                    push( @aclabels, $wit ) if( $wit =~ /^(.*?)(\s*\(?a\.\s*c\.\)?)$/ );
                }
                $label = join( ', ', 'majority', sort( @aclabels ) );
            }
            
            my $newpath = $self->add_path( $node, $self->reading( $newdest ), $label );
            $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;
        warn sprintf( "No hidden witnesses on %s -> %s ?", $from->name, $to->name )
            unless $path->has_hidden_witnesses;
        my @wits = @{$path->hidden_witnesses};
        $self->del_path( $path );
        foreach ( @wits ) {
            $self->add_path( $from, $to, $_ );
        }
    }
    $self->collapsed( 0 );
}

=back

=head2 Navigation methods

=over

=item B<start>

my $beginning = $collation->start();

Returns the beginning of the collation, a meta-reading with label '#START#'.

=cut

sub start {
    # Return the beginning reading of the graph.
    my $self = shift;
    my( $new_start ) = @_;
    if( $new_start ) {
        $self->del_reading( '#START#' );
        $self->graph->rename_node( $new_start, '#START#' );
    }
    # Make sure the start node has a start position.
    unless( $self->reading( '#START#' )->has_rank ) {
        $self->reading( '#START#' )->rank( '0' );
    }
    return $self->reading('#START#');
}

=item B<end>

my $end = $collation->end();

Returns the end of the collation, a meta-reading with label '#END#'.

=cut

sub end {
    my $self = shift;
    my( $new_end ) = @_;
    if( $new_end ) {
        $self->del_reading( '#END#' );
        $self->graph->rename_node( $new_end, '#END#' );
    }
    return $self->reading('#END#');
}

=item B<reading_sequence>

my @readings = $graph->reading_sequence( $first, $last, $path[, $alt_path] );

Returns the ordered list of readings, starting with $first and ending
with $last, along the given witness path.  If no path is specified,
assume that the path is that of the base text (if any.)

=cut

# TODO Think about returning some lazy-eval iterator.

sub reading_sequence {
    my( $self, $start, $end, $witness, $backup ) = @_;

    $witness = $self->baselabel unless $witness;
    my @readings = ( $start );
    my %seen;
    my $n = $start;
    while( $n && $n ne $end ) {
        if( exists( $seen{$n->name()} ) ) {
            warn "Detected loop at " . $n->name();
            last;
        }
        $seen{$n->name()} = 1;
        
        my $next = $self->next_reading( $n, $witness, $backup );
        warn "Did not find any path for $witness from reading " . $n->name
            unless $next;
        push( @readings, $next );
        $n = $next;
    }
    # Check that the last reading is our end reading.
    my $last = $readings[$#readings];
    warn "Last reading found from " . $start->label() .
        " for witness $witness is not the end!"
        unless $last eq $end;
    
    return @readings;
}

=item B<next_reading>

my $next_reading = $graph->next_reading( $reading, $witpath );

Returns the reading that follows the given reading along the given witness
path.  

=cut

sub next_reading {
    # Return the successor via the corresponding path.
    my $self = shift;
    return $self->_find_linked_reading( 'next', @_ );
}

=item B<prior_reading>

my $prior_reading = $graph->prior_reading( $reading, $witpath );

Returns the reading that precedes the given reading along the given witness
path.  

=cut

sub prior_reading {
    # Return the predecessor via the corresponding path.
    my $self = shift;
    return $self->_find_linked_reading( 'prior', @_ );
}

sub _find_linked_reading {
    my( $self, $direction, $node, $path, $alt_path ) = @_;
    my @linked_paths = $direction eq 'next' 
        ? $node->outgoing() : $node->incoming();
    return undef unless scalar( @linked_paths );
    
    # We have to find the linked path that contains all of the
    # witnesses supplied in $path.
    my( @path_wits, @alt_path_wits );
    @path_wits = $self->witnesses_of_label( $path ) if $path;
    @alt_path_wits = $self->witnesses_of_label( $alt_path ) if $alt_path;
    my $base_le;
    my $alt_le;
    foreach my $le ( @linked_paths ) {
        if( $le->name eq $self->baselabel ) {
            $base_le = $le;
        } else {
            my @le_wits = $self->witnesses_of_label( $le->name );
            if( _is_within( \@path_wits, \@le_wits ) ) {
                # This is the right path.
                return $direction eq 'next' ? $le->to() : $le->from();
            } elsif( _is_within( \@alt_path_wits, \@le_wits ) ) {
                $alt_le = $le;
            }
        }
    }
    # Got this far? Return the alternate path if it exists.
    return $direction eq 'next' ? $alt_le->to() : $alt_le->from()
        if $alt_le;

    # Got this far? Return the base path if it exists.
    return $direction eq 'next' ? $base_le->to() : $base_le->from()
        if $base_le;

    # Got this far? We have no appropriate path.
    warn "Could not find $direction node from " . $node->label 
        . " along path $path";
    return undef;
}

# Some set logic.
sub _is_within {
    my( $set1, $set2 ) = @_;
    my $ret = @$set1; # will be 0, i.e. false, if set1 is empty
    foreach my $el ( @$set1 ) {
        $ret = 0 unless grep { /^\Q$el\E$/ } @$set2;
    }
    return $ret;
}


## INITIALIZATION METHODS - for use by parsers
# Walk the paths for each witness in the graph, and return the nodes
# that the graph has in common.  If $using_base is true, some 
# different logic is needed.
# NOTE This does not create paths; it merely finds common readings.

sub walk_witness_paths {
    my( $self ) = @_;
    # For each witness, walk the path through the graph.
    # Then we need to find the common nodes.  
    # TODO This method is going to fall down if we have a very gappy 
    # text in the collation.
    my $paths = {};
    my @common_readings;
    foreach my $wit ( $self->tradition->witnesses ) {
        my $curr_reading = $self->start;
        my @wit_path = $self->reading_sequence( $self->start, $self->end, 
                                                $wit->sigil );
        $wit->path( \@wit_path );

        # Detect the common readings.
        @common_readings = _find_common( \@common_readings, \@wit_path );
    }

    # Mark all the nodes as either common or not.
    foreach my $cn ( @common_readings ) {
        print STDERR "Setting " . $cn->name . " / " . $cn->label 
            . " as common node\n";
        $cn->make_common;
    }
    foreach my $n ( $self->readings() ) {
        $n->make_variant unless $n->is_common;
    }
    # Return an array of the common nodes in order.
    return @common_readings;
}

sub _find_common {
    my( $common_readings, $new_path ) = @_;
    my @cr;
    if( @$common_readings ) {
        foreach my $n ( @$new_path ) {
            push( @cr, $n ) if grep { $_ eq $n } @$common_readings;
        }
    } else {
        push( @cr, @$new_path );
    }
    return @cr;
}

sub _remove_common {
    my( $common_readings, $divergence ) = @_;
    my @cr;
    my %diverged;
    map { $diverged{$_->name} = 1 } @$divergence;
    foreach( @$common_readings ) {
        push( @cr, $_ ) unless $diverged{$_->name};
    }
    return @cr;
}


# For use when a collation is constructed from a base text and an apparatus.
# We have the sequences of readings and just need to add path edges.

sub make_witness_paths {
    my( $self ) = @_;
    foreach my $wit ( $self->tradition->witnesses ) {
        print STDERR "Making path for " . $wit->sigil . "\n";
        $self->make_witness_path( $wit );
    }
}

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], $sig );
    }
    if( $wit->has_ante_corr ) {
        @chain = @{$wit->uncorrected_path};
        foreach my $idx( 0 .. $#chain-1 ) {
            my $source = $chain[$idx];
            my $target = $chain[$idx+1];
            $self->add_path( $source, $target, $sig.$self->ac_label )
                unless $self->has_path( $source, $target, $sig );
        }
    }
}

sub calculate_ranks {
    my $self = shift;
    # Walk a version of the graph where every node linked by a relationship 
    # edge is fundamentally the same node, and do a topological ranking on
    # the nodes in this graph.
    my $topo_graph = Graph::Easy->new();
    my %rel_containers;
    my $rel_ctr = 0;
    # Add the nodes
    foreach my $r ( $self->readings ) {
        next if exists $rel_containers{$r->name};
        my @rels = $r->related_readings( 'colocated' );
        if( @rels ) {
            # Make a relationship container.
            push( @rels, $r );
            my $rn = $topo_graph->add_node( 'rel_container_' . $rel_ctr++ );
            foreach( @rels ) {
                $rel_containers{$_->name} = $rn;
            }
        } else {
            # Add a new node to mirror the old node.
            $rel_containers{$r->name} = $topo_graph->add_node( $r->name );
        }
    }

    # Add the edges. Need only one edge between any pair of nodes.
    foreach my $r ( $self->readings ) {
        foreach my $n ( $r->neighbor_readings( 'forward' ) ) {
            $topo_graph->add_edge_once( $rel_containers{$r->name}, 
                                        $rel_containers{$n->name} );
        }
    }
    
    # Now do the rankings, starting with the start node.
    my $topo_start = $rel_containers{$self->start->name};
    my $node_ranks = { $topo_start->name => 0 };
    my @curr_origin = ( $topo_start );
    # A little iterative function.
    while( @curr_origin ) {
        @curr_origin = _assign_rank( $node_ranks, @curr_origin );
    }
    # Transfer our rankings from the topological graph to the real one.
    foreach my $r ( $self->readings ) {
        $r->rank( $node_ranks->{$rel_containers{$r->name}->name} );
    }
}

sub _assign_rank {
    my( $node_ranks, @current_nodes ) = @_;
    # 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.
    my @next_nodes;
    foreach my $c ( @current_nodes ) {
        warn "Current reading " . $c->name . "has no rank!"
            unless exists $node_ranks->{$c->name};
        # print STDERR "Looking at child of node " . $c->name . ", rank " 
        #     . $node_ranks->{$c->name} . "\n";
        my @children = map { $_->to } $c->outgoing;
        foreach my $child ( @children ) {
            next if exists $node_ranks->{$child->name};
            my $highest_rank = -1;
            my $skip = 0;
            my @parents = map { $_->from } $child->incoming;
            foreach my $parent ( @parents ) {
                if( exists $node_ranks->{$parent->name} ) {
                    $highest_rank = $node_ranks->{$parent->name} 
                        if $highest_rank <= $node_ranks->{$parent->name};
                } else {
                    $skip = 1;
                    last;
                }
            }
            next if $skip;
            # print STDERR "Assigning rank " . ( $highest_rank + 1 ) . " to node " . $child->name . "\n";
            $node_ranks->{$child->name} = $highest_rank + 1;
            push( @next_nodes, $child );
        }
    }
    return @next_nodes;
}

sub possible_positions {
    my $self = shift;
    my %all_pos;
    map { $all_pos{ $_->rank } = 1 } $self->readings;
    return keys %all_pos;
}

# TODO think about indexing this.
sub readings_at_position {
    my( $self, $position, $strict ) = @_;
    my @answer;
    foreach my $r ( $self->readings ) {
        push( @answer, $r ) if $r->is_at_position( $position, $strict );
    }
    return @answer;
}

## Lemmatizer functions

sub init_lemmata {
    my $self = shift;

    foreach my $position ( $self->possible_positions ) {
        $self->lemmata->{$position} = undef;
    }

    foreach my $cr ( $self->common_readings ) {
        $self->lemmata->{$cr->position->maxref} = $cr->name;
    }
}

sub common_readings {
    my $self = shift;
    my @common = grep { $_->is_common } $self->readings();
    return sort { $a->rank <=> $b->rank } @common;
}
    
=item B<lemma_readings>

my @state = $graph->lemma_readings( @readings_delemmatized );

Takes a list of readings that have just been delemmatized, and returns
a set of tuples of the form ['reading', 'state'] that indicates what
changes need to be made to the graph.

=over

=item * 

A state of 1 means 'lemmatize this reading'

=item * 

A state of 0 means 'delemmatize this reading'

=item * 

A state of undef means 'an ellipsis belongs in the text here because
no decision has been made / an earlier decision was backed out'

=back

=cut

sub lemma_readings {
    my( $self, @toggled_off_nodes ) = @_;

    # First get the positions of those nodes which have been
    # toggled off.
    my $positions_off = {};
    map { $positions_off->{ $_->position->reference } = $_->name } 
        @toggled_off_nodes;

    # Now for each position, we have to see if a node is on, and we
    # have to see if a node has been turned off.  The lemmata hash
    # should contain fixed positions, range positions whose node was
    # just turned off, and range positions whose node is on.
    my @answer;
    my %fixed_positions;
    # TODO One of these is probably redundant.
    map { $fixed_positions{$_} = 0 } keys %{$self->lemmata};
    map { $fixed_positions{$_} = 0 } keys %{$positions_off};
    map { $fixed_positions{$_} = 1 } $self->possible_positions;
    foreach my $pos ( sort { Text::Tradition::Collation::Position::str_cmp( $a, $b ) } keys %fixed_positions ) {
        # Find the state of this position.  If there is an active node,
        # its name will be the state; otherwise the state will be 0 
        # (nothing at this position) or undef (ellipsis at this position)
        my $active = undef;
        $active = $self->lemmata->{$pos} if exists $self->lemmata->{$pos};
        
        # Is there a formerly active node that was toggled off?
        if( exists( $positions_off->{$pos} ) ) {
            my $off_node = $positions_off->{$pos};
            if( $active && $active ne $off_node) {
                push( @answer, [ $off_node, 0 ], [ $active, 1 ] );
            } else {
                unless( $fixed_positions{$pos} ) {
                    $active = 0;
                    delete $self->lemmata->{$pos};
                }
                push( @answer, [ $off_node, $active ] );
            }

        # No formerly active node, so we just see if there is a currently
        # active one.
        } elsif( $active ) {
            # Push the active node, whatever it is.
            push( @answer, [ $active, 1 ] );
        } else {
            # Push the state that is there. Arbitrarily use the first node
            # at that position.
            my @pos_nodes = $self->readings_at_position( $pos );
            push( @answer, [ $pos_nodes[0]->name, $self->lemmata->{$pos} ] );
            delete $self->lemmata->{$pos} unless $fixed_positions{$pos};
        }
    }

    return @answer;
}

=item B<toggle_reading>

my @readings_delemmatized = $graph->toggle_reading( $reading_name );

Takes a reading node name, and either lemmatizes or de-lemmatizes
it. Returns a list of all readings that are de-lemmatized as a result
of the toggle.

=cut

sub toggle_reading {
    my( $self, $rname ) = @_;
    
    return unless $rname;
    my $reading = $self->reading( $rname );
    if( !$reading || $reading->is_common() ) {
        # Do nothing, it's a common node.
        return;
    } 
    
    my $pos = $reading->position;
    my $fixed = $reading->position->fixed;
    my $old_state = $self->lemmata->{$pos->reference};

    my @readings_off;
    if( $old_state && $old_state eq $rname ) {
        # Turn off the node. We turn on no others by default.
        push( @readings_off, $reading );
    } else {
        # Turn on the node.
        $self->lemmata->{$pos->reference} = $rname;
        # Any other 'on' readings in the same position should be off
        # if we have a fixed position.
        push( @readings_off, $self->same_position_as( $reading, 1 ) )
            if $pos->fixed;
        # Any node that is an identical transposed one should be off.
        push( @readings_off, $reading->identical_readings );
    }
    @readings_off = unique_list( @readings_off );
        
    # Turn off the readings that need to be turned off.
    my @readings_delemmatized;
    foreach my $n ( @readings_off ) {
        my $npos = $n->position;
        my $state = undef;
        $state = $self->lemmata->{$npos->reference}
            if defined $self->lemmata->{$npos->reference};
        if( $state && $state eq $n->name ) { 
            # this reading is still on, so turn it off
            push( @readings_delemmatized, $n );
            my $new_state = undef;
            if( $npos->fixed && $n eq $reading ) {
                # This is the reading that was clicked, so if there are no
                # other readings there and this is a fixed position, turn off 
                # the position.  In all other cases, restore the ellipsis.
                my @other_n = $self->same_position_as( $n ); # TODO do we need strict?
                $new_state = 0 unless @other_n;
            }
            $self->lemmata->{$npos->reference} = $new_state;
        } elsif( $old_state && $old_state eq $n->name ) { 
            # another reading has already been turned on here
            push( @readings_delemmatized, $n );
        } # else some other reading was on anyway, so pass.
    }
    return @readings_delemmatized;
}

sub same_position_as {
    my( $self, $reading, $strict ) = @_;
    my $pos = $reading->position;
    my %onpath = ( $reading->name => 1 );
    # TODO This might not always be sufficient.  We really want to
    # exclude all readings on this one's path between its two
    # common points.
    map { $onpath{$_->name} = 1 } $reading->neighbor_readings;
    my @same = grep { !$onpath{$_->name} } 
        $self->readings_at_position( $reading->position, $strict );
    return @same;
}

# Return the string that joins together a list of witnesses for
# display on a single path.
sub path_label {
    my $self = shift;
    return join( $self->wit_list_separator, @_ );
}

sub witnesses_of_label {
    my( $self, $label ) = @_;
    my $regex = $self->wit_list_separator;
    my @answer = split( /\Q$regex\E/, $label );
    return @answer;
}    

sub unique_list {
    my( @list ) = @_;
    my %h;
    map { $h{$_->name} = $_ } @list;
    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;