X-Git-Url: http://git.shadowcat.co.uk/gitweb/gitweb.cgi?a=blobdiff_plain;f=lib%2FText%2FTradition%2FCollation.pm;h=7606cc2dfee068c0518c7dcb89bdf744baf7a0d5;hb=e867486f69f12dc06304594022c298935d1c7fb9;hp=03aed2a0ea1884c8bf6b91a13c905ede37f7ba4c;hpb=a0093bf226790f5f78b96b7d94aa37b261b3fc17;p=scpubgit%2Fstemmatology.git diff --git a/lib/Text/Tradition/Collation.pm b/lib/Text/Tradition/Collation.pm index 03aed2a..7606cc2 100644 --- a/lib/Text/Tradition/Collation.pm +++ b/lib/Text/Tradition/Collation.pm @@ -1,9 +1,14 @@ package Text::Tradition::Collation; +use Encode qw( decode_utf8 ); +use File::Temp; +use Graph; use Graph::Easy; use IPC::Run qw( run binary ); -use Text::Tradition::Collation::Reading; +use Text::CSV_XS; use Text::Tradition::Collation::Path; +use Text::Tradition::Collation::Reading; +use Text::Tradition::Collation::Relationship; use XML::LibXML; use Moose; @@ -11,21 +16,22 @@ has 'graph' => ( is => 'ro', isa => 'Graph::Easy', handles => { - add_reading => 'add_node', - del_reading => 'del_node', - add_path => 'add_edge', - del_path => 'del_edge', - reading => 'node', - path => 'edge', - readings => 'nodes', - paths => 'edges', - relationships => 'edges', + add_reading => 'add_node', + add_lacuna => 'add_node', + del_reading => 'del_node', + add_path => 'add_edge', + del_path => 'del_edge', + reading => 'node', + path => 'edge', + readings => 'nodes', + paths => 'edges', + relationships => 'edges', }, default => sub { Graph::Easy->new( undirected => 0 ) }, ); - + -has 'tradition' => ( +has 'tradition' => ( # TODO should this not be ro? is => 'rw', isa => 'Text::Tradition', ); @@ -44,6 +50,13 @@ has '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' => ( @@ -75,6 +88,12 @@ has 'linear' => ( 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 @@ -99,8 +118,21 @@ sub BUILD { # Pass through any graph-specific options. my $shape = exists( $args->{'shape'} ) ? $args->{'shape'} : 'ellipse'; $self->graph->set_attribute( 'node', 'shape', $shape ); + + # Start and end points for all texts + $self->start( 'INIT' ); + $self->end( 'INIT' ); } +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 { @@ -109,19 +141,20 @@ around add_path => sub { # Make sure there are three arguments unless( @_ == 3 ) { - warn "Call add_path with args source, target, witness"; - return; + 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'; + unless ref( $source ) eq 'Text::Tradition::Collation::Reading'; $target = $self->reading( $target ) - unless ref( $target ) eq 'Text::Tradition::Collation::Reading'; + unless ref( $target ) eq 'Text::Tradition::Collation::Reading'; foreach my $path ( $source->edges_to( $target ) ) { - if( $path->label eq $wit ) { - return; - } + if( $path->label eq $wit && $path->class eq 'edge.path' ) { + return; + } } # Do the deed $self->$orig( @_ ); @@ -132,19 +165,18 @@ around paths => sub { my $orig = shift; my $self = shift; - my @result = grep { $_->class eq 'edge.path' } $self->$orig( @_ ); + my @result = grep { $_->sub_class eq 'path' } $self->$orig( @_ ); return @result; }; around relationships => sub { my $orig = shift; my $self = shift; - my @result = grep { $_->class eq 'edge.relationship' } $self->$orig( @_ ); + my @result = grep { $_->sub_class eq 'relationship' } $self->$orig( @_ ); return @result; }; # Wrapper around merge_nodes - sub merge_readings { my $self = shift; my $first_node = shift; @@ -159,43 +191,73 @@ sub has_path { my( $self, $source, $target, $label ) = @_; my @paths = $source->edges_to( $target ); my @relevant = grep { $_->label eq $label } @paths; - return scalar @paths; + return scalar @relevant; } ## Dealing with relationships between readings. This is a different -## sort of graph edge. +## sort of graph edge. Return a success/failure value and a list of +## node pairs that have been linked. sub add_relationship { - my( $self, $type, $source, $target, $global ) = @_; - 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 ); + my( $self, $source, $target, $options ) = @_; + + # Make sure there is not another relationship between these two + # readings 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 ); - 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 ); - } - } + + # 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) @@ -221,8 +283,14 @@ sub as_svg { my @cmd = qw/dot -Tsvg/; my( $svg, $err ); - my $in = $self->as_dot(); - run( \@cmd, \$in, ">", binary(), \$svg ); + 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; @@ -249,19 +317,24 @@ sub as_dot { $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' ) ); + 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 ); + # Need not output nodes without separate labels + 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 ); + 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; } @@ -285,7 +358,7 @@ sub as_graphml { 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'; + 'http://graphml.graphdrawing.org/xmlns/1.0/graphml.xsd'; # Create the document and root node my $graphml = XML::LibXML->createDocument( "1.0", "UTF-8" ); @@ -294,107 +367,238 @@ sub as_graphml { $root->setNamespace( $xsi_ns, 'xsi', 0 ); $root->setAttributeNS( $xsi_ns, 'schemaLocation', $graphml_schema ); + # Add the data keys for the graph + my %graph_data_keys; + my $gdi = 0; + my @graph_attributes = qw/ wit_list_separator baselabel linear ac_label /; + foreach my $datum ( @graph_attributes ) { + $graph_data_keys{$datum} = 'dg'.$gdi++; + my $key = $root->addNewChild( $graphml_ns, 'key' ); + $key->setAttribute( 'attr.name', $datum ); + $key->setAttribute( 'attr.type', $key eq 'linear' ? 'boolean' : 'string' ); + $key->setAttribute( 'for', 'graph' ); + $key->setAttribute( 'id', $graph_data_keys{$datum} ); + } + # Add the data keys for nodes - my @node_data = ( 'name', 'reading', 'identical', 'position' ); - foreach my $ndi ( 0 .. $#node_data ) { - my $key = $root->addNewChild( $graphml_ns, 'key' ); - $key->setAttribute( 'attr.name', $node_data[$ndi] ); - $key->setAttribute( 'attr.type', 'string' ); - $key->setAttribute( 'for', 'node' ); - $key->setAttribute( 'id', 'd'.$ndi ); + 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 %wit_hash; - my $wit_ctr = 0; - foreach my $wit ( @{$self->tradition->witnesses} ) { - my $wit_key = 'w' . $wit_ctr++; - $wit_hash{$wit->sigil} = $wit_key; - my $key = $root->addNewChild( $graphml_ns, 'key' ); - $key->setAttribute( 'attr.name', _make_xml_attr( $wit->sigil ) ); - $key->setAttribute( 'attr.type', 'string' ); - $key->setAttribute( 'for', 'edge' ); - $key->setAttribute( 'id', $wit_key ); - my $ackey = $root->addNewChild( $graphml_ns, 'key' ); - $ackey->setAttribute( 'attr.name', _make_xml_attr( $wit->sigil ) . "_ante_corr" ); - $ackey->setAttribute( 'attr.type', 'string' ); - $ackey->setAttribute( 'for', 'edge' ); - $ackey->setAttribute( 'id', $wit_key . "a" ); + my $edi = 0; + my %edge_data_keys; + my @string_keys = qw/ class witness relationship /; + my @bool_keys = qw/ extra equal_rank non_correctable non_independent /; + foreach my $edge_key( @string_keys ) { + $edge_data_keys{$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} ); } - + foreach my $edge_key( @bool_keys ) { + $edge_data_keys{$edge_key} = 'de'.$edi++; + my $key = $root->addNewChild( $graphml_ns, 'key' ); + $key->setAttribute( 'attr.name', $edge_key ); + $key->setAttribute( 'attr.type', 'boolean' ); + $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( 'id', $self->tradition->name ); $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' ); + + # Collation attribute data + foreach my $datum ( @graph_attributes ) { + _add_graphml_data( $graph, $graph_data_keys{$datum}, $self->$datum ); + } my $node_ctr = 0; my %node_hash; - foreach my $n ( $self->readings ) { - my %this_node_data = (); - foreach my $ndi ( 0 .. $#node_data ) { - my $key = $node_data[$ndi]; - if( $key eq 'name' ) { - $this_node_data{'d'.$ndi} = $n->name; - } elsif( $key eq 'token' ) { - $this_node_data{'d'.$ndi} = $n->label; - } elsif( $key eq 'identical' && $n->has_primary ) { - $this_node_data{'d'.$ndi} = $n->primary->name; - } elsif( $key eq 'position' ) { - $this_node_data{'d'.$ndi} = $n->position; - } - } - 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 ); - - foreach my $dk ( keys %this_node_data ) { - my $d_el = $node_el->addNewChild( $graphml_ns, 'data' ); - $d_el->setAttribute( 'key', $dk ); - $d_el->appendText( $this_node_data{$dk} ); - } + # 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 && $n->primary ne $n; } + # Add the path and relationship edges my $edge_ctr = 0; - foreach my $e ( $self->paths() ) { - 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 witness - my $base = $e->label; - my $ante_corr; - if( $e->label =~ /^(.*?)\s+(\(a\.c\.\))$/ ) { - ( $base, $ante_corr ) = ( $1, $2 ); - } - my $key = $wit_hash{$base}; - $key .= "a" if $ante_corr; - my $wit_el = $edge_el->addNewChild( $graphml_ns, 'data' ); - $wit_el->setAttribute( 'key', $key ); - $wit_el->appendText( $e->label ); + 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 ); + + # For some classes we have extra information to save. + 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'}; + # Is this an ante-corr witness? + my $aclabel = $self->ac_label; + if( $e->label =~ /^(.*)\Q$aclabel\E$/ ) { + # Keep the base witness + $base = $1; + # ...and record that this is an 'extra' reading path + _add_graphml_data( $edge_el, $edge_data_keys{'extra'}, 'true' ); + } + _add_graphml_data( $edge_el, $edge_data_keys{'witness'}, $base ); + } elsif( $e->sub_class eq 'relationship' ) { + # It's a relationship, so save the relationship data + _add_graphml_data( $edge_el, $edge_data_keys{'relationship'}, $e->label ); + _add_graphml_data( $edge_el, $edge_data_keys{'equal_rank'}, $e->equal_rank ); + _add_graphml_data( $edge_el, $edge_data_keys{'non_correctable'}, $e->non_correctable ); + _add_graphml_data( $edge_el, $edge_data_keys{'non_independent'}, $e->non_independent ); + } + } + + # Save and return the thing + my $result = decode_utf8( $graphml->toString(1) ); + $self->_save_graphml( $result ); + return $result; +} + +sub _add_graphml_data { + my( $el, $key, $value ) = @_; + return unless defined $value; + my $data_el = $el->addNewChild( $el->namespaceURI, 'data' ); + $data_el->setAttribute( 'key', $key ); + $data_el->appendText( $value ); +} + +=item B + +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; +} + +# Make an alignment table - $noderefs controls whether the objects +# in the table are the nodes or simply their readings. + +sub make_alignment_table { + my( $self, $noderefs ) = @_; + 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, $noderefs ); + unshift( @row, $wit->sigil ); + push( @$table, \@row ); + if( $wit->has_ante_corr ) { + my @ac_row = _make_witness_row( $wit->uncorrected_path, \@all_pos, $noderefs ); + unshift( @ac_row, $wit->sigil . $self->ac_label ); + push( @$table, \@ac_row ); + } } - # Return the thing - $self->_save_graphml( $graphml->toString(1) ); - return $graphml->toString(1); + # 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, $noderefs ) = @_; + 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} = $noderefs ? $rdg : $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 we are using node reference, make the lacuna node appear many times + # in the table. If not, use the lacuna tag. + if( $last_el && _el_is_lacuna( $last_el ) && !defined $el ) { + $el = $noderefs ? $last_el : '#LACUNA#'; + } + push( @filled_row, $el ); + $last_el = $el; + } + return @filled_row; } -sub _make_xml_attr { - my $str = shift; - $str =~ s/\s/_/g; - $str =~ s/\W//g; - $str =~ "a$str" if $str =~ /^\d/; - return $str; +# Tiny utility function to say if a table element is a lacuna +sub _el_is_lacuna { + my $el = shift; + return 1 if $el eq '#LACUNA#'; + return 1 if ref( $el ) eq 'Text::Tradition::Collation::Reading' + && $el->is_lacuna; + return 0; } +# 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 @@ -406,42 +610,35 @@ sub collapse_graph_paths { 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; + 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 ); - } - } + 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 ); @@ -457,16 +654,15 @@ sub expand_graph_paths { 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, $_ ); - } + 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 ); } @@ -487,13 +683,49 @@ Returns the beginning of the collation, a meta-reading with label '#START#'. sub start { # Return the beginning reading of the graph. + my( $self, $new_start ) = @_; + my $start = $self->reading( '#START#' ); + if( ref( $new_start ) eq 'Text::Tradition::Collation::Reading' ) { + # Replace the existing start node. + $self->del_reading( '#START#' ); + $self->graph->rename_node( $new_start, '#START#' ); + $start = $new_start; + } elsif ( $new_start && $new_start eq 'INIT' ) { + # Make a new start node. + $start = $self->add_reading( '#START#' ); + } + # Make sure the start node is a meta node + $start->is_meta( 1 ); + # Make sure the start node has a start position. + unless( $start->has_rank ) { + $start->rank( '0' ); + } + return $start; +} + +=item B + +my $end = $collation->end(); + +Returns the end of the collation, a meta-reading with label '#END#'. + +=cut + +sub end { my $self = shift; - my( $new_start ) = @_; - if( $new_start ) { - $self->del_reading( '#START#' ); - $self->graph->rename_node( $new_start, '#START#' ); + my( $new_end ) = @_; + my $end = $self->reading( '#END#' ); + if( ref( $new_end ) eq 'Text::Tradition::Collation::Reading' ) { + $self->del_reading( '#END#' ); + $self->graph->rename_node( $new_end, '#END#' ); + $end = $new_end + } elsif ( $new_end && $new_end eq 'INIT' ) { + # Make a new start node. + $end = $self->add_reading( '#END#' ); } - return $self->reading('#START#'); + # Make sure the start node is a meta node + $end->is_meta( 1 ); + return $end; } =item B @@ -506,6 +738,8 @@ 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 ) = @_; @@ -514,23 +748,23 @@ sub reading_sequence { 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; + 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; + " for witness $witness is not the end!" + unless $last eq $end; return @readings; } @@ -568,7 +802,7 @@ sub prior_reading { sub _find_linked_reading { my( $self, $direction, $node, $path, $alt_path ) = @_; my @linked_paths = $direction eq 'next' - ? $node->outgoing() : $node->incoming(); + ? $node->outgoing() : $node->incoming(); return undef unless scalar( @linked_paths ); # We have to find the linked path that contains all of the @@ -579,29 +813,29 @@ sub _find_linked_reading { 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; - } - } + 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; + 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; + if $base_le; # Got this far? We have no appropriate path. warn "Could not find $direction node from " . $node->label - . " along path $path"; + . " along path $path"; return undef; } @@ -610,7 +844,7 @@ 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; + $ret = 0 unless grep { /^\Q$el\E$/ } @$set2; } return $ret; } @@ -620,33 +854,34 @@ sub _is_within { # 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, $end ) = @_; + 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, $end, - $wit->sigil ); - $wit->path( \@wit_path ); - - # Detect the common readings. - @common_readings = _find_common( \@common_readings, \@wit_path ); + 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; + 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; + $n->make_variant unless $n->is_common; } # Return an array of the common nodes in order. return @common_readings; @@ -656,11 +891,11 @@ 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; - } + foreach my $n ( @$new_path ) { + push( @cr, $n ) if grep { $_ eq $n } @$common_readings; + } } else { - push( @cr, @$new_path ); + push( @cr, @$new_path ); } return @cr; } @@ -671,27 +906,21 @@ sub _remove_common { my %diverged; map { $diverged{$_->name} = 1 } @$divergence; foreach( @$common_readings ) { - push( @cr, $_ ) unless $diverged{$_->name}; + push( @cr, $_ ) unless $diverged{$_->name}; } return @cr; } -# An alternative to walk_witness_paths, 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. +# 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 ) = @_; - - my @common_readings; - foreach my $wit ( @{$self->tradition->witnesses} ) { - 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 ); + foreach my $wit ( $self->tradition->witnesses ) { + print STDERR "Making path for " . $wit->sigil . "\n"; + $self->make_witness_path( $wit ); } - return @common_readings; } sub make_witness_path { @@ -699,125 +928,138 @@ sub make_witness_path { my @chain = @{$wit->path}; my $sig = $wit->sigil; foreach my $idx ( 0 .. $#chain-1 ) { - $self->add_path( $chain[$idx], $chain[$idx+1], $sig ); + $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 ); + 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 common_readings { +sub calculate_ranks { my $self = shift; - my @common = grep { $_->is_common } $self->readings(); - return sort { _cmp_position( $a->position, $b->position ) } @common; -} - -# Calculate the relative positions of nodes in the graph, if they -# were not given to us. -sub calculate_positions { - my( $self, @ordered_common ) = @_; - - # We have to calculate the position identifiers for each word, - # keyed on the common nodes. This will be 'fun'. The end result - # is a hash per witness, whose key is the word node and whose - # value is its position in the text. Common nodes are always N,1 - # so have identical positions in each text. + # 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->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 = 'rel_container_' . $rel_ctr++; + $topo_graph->add_vertex( $rn ); + foreach( @rels ) { + $rel_containers{$_->name} = $rn; + } + } else { + # Add a new node to mirror the old node. + $rel_containers{$r->name} = $r->name; + $topo_graph->add_vertex( $r->name ); + } + } - my $node_pos = {}; - 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->uncorrected_path, @ordered_common ) - if $wit->has_ante_corr; + # Add the edges. Need only one edge between any pair of nodes. + foreach my $r ( $self->readings ) { + foreach my $n ( $r->neighbor_readings( 'forward' ) ) { + my( $tfrom, $tto ) = ( $rel_containers{$r->name}, + $rel_containers{$n->name} ); + $topo_graph->add_edge( $tfrom, $tto ) + unless $topo_graph->has_edge( $tfrom, $tto ); + } } - # DEBUG - foreach my $r ( $self->readings() ) { - print STDERR "Reading " . $r->name . "/" . $r->label . " has no position\n" - unless( $r->has_position ); - } - - $self->init_lemmata(); -} - -sub _update_positions_from_path { - my( $path, @ordered_common ) = @_; - - # First we walk the given path, making a matrix for the witness - # 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 $wn ( @{$path} ) { - if( $wn eq $ordered_common[$cn] ) { - # Set up to look for the next common node, and - # start a new row of words. - $cn++; - push( @$wit_matrix, $row ) if scalar( @$row ); - $row = []; - } - push( @$row, $wn ); + # Now do the rankings, starting with the start node. + my $topo_start = $rel_containers{$self->start->name}; + my $node_ranks = { $topo_start => 0 }; + my @curr_origin = ( $topo_start ); + # A little iterative function. + while( @curr_origin ) { + @curr_origin = _assign_rank( $topo_graph, $node_ranks, @curr_origin ); } - push( @$wit_matrix, $row ); # Push the last row onto the matrix - - # Now we have a matrix per witness, so that each row in the - # matrix begins with a common node, and continues with all the - # variant words that appear in the witness. We turn this into - # real positions in row,cell format. But we need some - # trickery in order to make sure that each node gets assigned - # to only one position. - - foreach my $li ( 1..scalar(@$wit_matrix) ) { - 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 && - _cmp_position( $position, $reading->position ) < 1 ) { - # The new position ID replaces the old one. - $reading->position( $position ); - } # otherwise, the old position needs to stay. - } + # Transfer our rankings from the topological graph to the real one. + foreach my $r ( $self->readings ) { + $r->rank( $node_ranks->{$rel_containers{$r->name}} ); } } -sub _cmp_position { - my( $a, $b ) = @_; - if ( $a && $b ) { - my @pos_a = split(/,/, $a ); - my @pos_b = split(/,/, $b ); +sub _assign_rank { + my( $graph, $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; 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}; + # print STDERR "Looking at child of node $c, rank " + # . $node_ranks->{$c} . "\n"; + 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; + push( @next_nodes, $child ); + } + } + return @next_nodes; +} - my $big_cmp = $pos_a[0] <=> $pos_b[0]; - return $big_cmp if $big_cmp; - # else - return $pos_a[1] <=> $pos_b[1]; - } elsif ( $b ) { # a is undefined - return -1; - } elsif ( $a ) { # b is undefined - return 1; +# Another method to make up for rough collation methods. If the same reading +# appears multiple times at the same rank, collapse the nodes. +sub flatten_ranks { + my $self = shift; + my %unique_rank_rdg; + foreach my $rdg ( $self->readings ) { + next unless $rdg->has_rank; + my $key = $rdg->rank . "||" . $rdg->text; + if( exists $unique_rank_rdg{$key} ) { + # Combine! + print STDERR "Combining readings at same rank: $key\n"; + $self->merge_readings( $unique_rank_rdg{$key}, $rdg ); + } else { + $unique_rank_rdg{$key} = $rdg; + } } - return 0; # they are both undefined } -sub all_positions { + +sub possible_positions { my $self = shift; - my %positions = (); - map { $positions{$_->position} = 1 } $self->readings; - my @answer = sort { _cmp_position( $a, $b ) } keys( %positions ); - return @answer; + 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, $pos ) = @_; - my @answer = grep { $_->position eq $pos } $self->readings; + my( $self, $position, $strict ) = @_; + my @answer; + foreach my $r ( $self->readings ) { + push( @answer, $r ) if $r->is_at_position( $position, $strict ); + } return @answer; } @@ -825,15 +1067,21 @@ sub readings_at_position { sub init_lemmata { my $self = shift; - - foreach my $position ( $self->all_positions ) { - $self->lemmata->{$position} = undef; + + foreach my $position ( $self->possible_positions ) { + $self->lemmata->{$position} = undef; } foreach my $cr ( $self->common_readings ) { - $self->lemmata->{$cr->position} = $cr->name; + $self->lemmata->{$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 @@ -868,39 +1116,53 @@ sub lemma_readings { # First get the positions of those nodes which have been # toggled off. my $positions_off = {}; - map { $positions_off->{ $_->position } = $_->name } @toggled_off_nodes; + 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. + # 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; - foreach my $pos ( $self->all_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 = $self->lemmata->{$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 { - 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} ] ); - } + 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; } @@ -920,54 +1182,67 @@ sub toggle_reading { return unless $rname; my $reading = $self->reading( $rname ); if( !$reading || $reading->is_common() ) { - # Do nothing, it's a common node. - return; + # Do nothing, it's a common node. + return; } my $pos = $reading->position; - my $old_state = $self->lemmata->{$pos}; + 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 ); + # Turn off the node. We turn on no others by default. + push( @readings_off, $reading ); } else { - # Turn on the node. - $self->lemmata->{$pos} = $rname; - # Any other 'on' readings in the same position should be off. - push( @readings_off, $self->same_position_as( $reading ) ); - # Any node that is an identical transposed one should be off. - push( @readings_off, $reading->identical_readings ); + # 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 $state = $self->lemmata->{$n->position}; - if( $state && $state eq $n->name ) { - # this reading is still on, so turn it off - push( @readings_delemmatized, $n ); - my $new_state = undef; - if( $n eq $reading ) { - # This is the reading that was clicked, so if there are no - # other readings there, turn off the position. In all other - # cases, restore the ellipsis. - my @other_n = $self->same_position_as( $n ); - $new_state = 0 unless @other_n; - } - $self->lemmata->{$n->position} = $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. + 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 ) = @_; + my( $self, $reading, $strict ) = @_; my $pos = $reading->position; - my @same = grep { $_ ne $reading } $self->readings_at_position( $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; } @@ -995,11 +1270,21 @@ sub unique_list { sub add_hash_entry { my( $hash, $key, $entry ) = @_; if( exists $hash->{$key} ) { - push( @{$hash->{$key}}, $entry ); + push( @{$hash->{$key}}, $entry ); } else { - $hash->{$key} = [ $entry ]; + $hash->{$key} = [ $entry ]; } } no Moose; __PACKAGE__->meta->make_immutable; + +=head1 BUGS / TODO + +=over + +=item * Rationalize edge classes + +=item * Port the internal graph from Graph::Easy to Graph + +=back