use Text::CSV_XS;
use Text::Tradition::Collation::Reading;
use Text::Tradition::Collation::RelationshipStore;
+use Text::Tradition::Error;
use XML::LibXML;
use Moose;
default => 1,
);
-has 'collapse_punctuation' => (
- is => 'rw',
- isa => 'Bool',
- default => 1,
- );
-
has 'ac_label' => (
is => 'rw',
isa => 'Str',
transposed readings should be treated as two linked readings rather than one,
and therefore whether the collation graph is acyclic. Defaults to true.
-=item * collapse_punctuation - TODO
-
=item * baselabel - The default label for the path taken by a base text
(if any). Defaults to 'base text'.
=head2 linear
-=head2 collapse_punctuation
-
=head2 wit_list_separator
=head2 baselabel
}
# First check to see if a reading with this ID exists.
if( $self->reading( $reading->id ) ) {
- warn "Collation already has a reading with id " . $reading->id;
- return undef;
+ throw( "Collation already has a reading with id " . $reading->id );
}
$self->_add_reading( $reading->id => $reading );
# Once the reading has been added, put it in both graphs.
sub add_relationship {
my $self = shift;
my( $source, $target, $opts ) = $self->_stringify_args( @_ );
- my( $ret, @vectors ) = $self->relations->add_relationship( $source,
+ my( @vectors ) = $self->relations->add_relationship( $source,
$self->reading( $source ), $target, $self->reading( $target ), $opts );
# Force a full rank recalculation every time. Yuck.
- $self->calculate_ranks() if $ret && $self->end->has_rank;
- return( $ret, @vectors );
+ $self->calculate_ranks() if $self->end->has_rank;
+ return @vectors;
}
=head2 reading_witnesses( $reading )
my $dot = $self->as_dot( $from, $to );
unless( $dot ) {
- warn "Could not output a graph with range $from - $to";
- return;
+ throw( "Could not output a graph with range $from - $to" );
}
my @cmd = qw/dot -Tsvg/;
my $graph_name = $self->tradition->name;
$graph_name =~ s/[^\w\s]//g;
$graph_name = join( '_', split( /\s+/, $graph_name ) );
+
+ my %graph_attrs = (
+ 'rankdir' => 'LR',
+ 'bgcolor' => 'none',
+ );
+ my %node_attrs = (
+ 'fontsize' => 11,
+ 'fillcolor' => 'white',
+ 'style' => 'filled',
+ 'shape' => 'ellipse'
+ );
+ my %edge_attrs = (
+ 'arrowhead' => 'open',
+ 'color' => '#000000',
+ 'fontcolor' => '#000000',
+ );
+
my $dot = sprintf( "digraph %s {\n", $graph_name );
- $dot .= "\tedge [ arrowhead=open ];\n";
- $dot .= "\tgraph [ rankdir=LR,bgcolor=none ];\n";
- $dot .= sprintf( "\tnode [ fontsize=%d, fillcolor=%s, style=%s, shape=%s ];\n",
- 11, "white", "filled", "ellipse" );
+ $dot .= "\tgraph " . _dot_attr_string( \%graph_attrs ) . ";\n";
+ $dot .= "\tnode " . _dot_attr_string( \%node_attrs ) . ";\n";
# Output substitute start/end readings if necessary
if( $startrank ) {
$dot .= "\t\"#SUBEND#\" [ label=\"...\" ];\n";
}
my %used; # Keep track of the readings that actually appear in the graph
- my %subedges;
- my %subend;
foreach my $reading ( $self->readings ) {
# Only output readings within our rank range.
next if $startrank && $reading->rank < $startrank;
next if $endrank && $reading->rank > $endrank;
$used{$reading->id} = 1;
- $subedges{$reading->id} = '#SUBSTART#'
- if $startrank && $startrank == $reading->rank;
- $subedges{$reading->id} = '#SUBEND#'
- if $endrank && $endrank == $reading->rank;
# Need not output nodes without separate labels
next if $reading->id eq $reading->text;
- my $label = $reading->punctuated_form;
+ my $rattrs;
+ my $label = $reading->text;
$label =~ s/\"/\\\"/g;
- $dot .= sprintf( "\t\"%s\" [ label=\"%s\" ];\n", $reading->id, $label );
+ $rattrs->{'label'} = $label;
+ # TODO make this an option?
+ # $rattrs->{'fillcolor'} = 'green' if $reading->is_common;
+ $dot .= sprintf( "\t\"%s\" %s;\n", $reading->id, _dot_attr_string( $rattrs ) );
}
- # Add substitute start and end edges if necessary
- foreach my $node ( keys %subedges ) {
- my @vector = ( $subedges{$node}, $node );
- @vector = reverse( @vector ) if $vector[0] =~ /END/;
- my $witstr = join( ', ', sort $self->reading_witnesses( $self->reading( $node ) ) );
- my %variables = ( 'color' => '#000000',
- 'fontcolor' => '#000000',
- 'label' => $witstr,
- );
- my $varopts = join( ', ', map { $_.'="'.$variables{$_}.'"' } sort keys %variables );
- $dot .= sprintf( "\t\"%s\" -> \"%s\" [ %s ];\n", @vector, $varopts );
- }
-
# Add the real edges
my @edges = $self->paths;
+ my( %substart, %subend );
foreach my $edge ( @edges ) {
# Do we need to output this edge?
- if( $used{$edge->[0]} && $used{$edge->[1]} ) {;
- my %variables = ( 'color' => '#000000',
- 'fontcolor' => '#000000',
- 'label' => join( ', ', $self->path_display_label( $edge ) ),
- );
- my $varopts = join( ', ', map { $_.'="'.$variables{$_}.'"' } sort keys %variables );
+ if( $used{$edge->[0]} && $used{$edge->[1]} ) {
+ my $label = $self->path_display_label( $self->path_witnesses( $edge ) );
+ my $variables = { %edge_attrs, 'label' => $label };
# Account for the rank gap if necessary
- my $rankgap = $self->reading( $edge->[1] )->rank
+ if( $self->reading( $edge->[1] )->has_rank
+ && $self->reading( $edge->[0] )->has_rank
+ && $self->reading( $edge->[1] )->rank
+ - $self->reading( $edge->[0] )->rank > 1 ) {
+ $variables->{'minlen'} = $self->reading( $edge->[1] )->rank
- $self->reading( $edge->[0] )->rank;
- $varopts .= ", minlen=$rankgap" if $rankgap > 1;
- $dot .= sprintf( "\t\"%s\" -> \"%s\" [ %s ];\n",
- $edge->[0], $edge->[1], $varopts );
+ }
+ # EXPERIMENTAL: make edge width reflect no. of witnesses
+ my $extrawidth = scalar( $self->path_witnesses( $edge ) ) * 0.2;
+ $variables->{'penwidth'} = $extrawidth + 0.8; # gives 1 for a single wit
+
+ my $varopts = _dot_attr_string( $variables );
+ $dot .= sprintf( "\t\"%s\" -> \"%s\" %s;\n",
+ $edge->[0], $edge->[1], $varopts );
+ } elsif( $used{$edge->[0]} ) {
+ $subend{$edge->[0]} = 1;
+ } elsif( $used{$edge->[1]} ) {
+ $substart{$edge->[1]} = 1;
}
}
-
+ # Add substitute start and end edges if necessary
+ foreach my $node ( keys %substart ) {
+ my $witstr = $self->path_display_label ( $self->reading_witnesses( $self->reading( $node ) ) );
+ my $variables = { %edge_attrs, 'label' => $witstr };
+ my $varopts = _dot_attr_string( $variables );
+ $dot .= "\t\"#SUBSTART#\" -> \"$node\" $varopts;";
+ }
+ foreach my $node ( keys %subend ) {
+ my $witstr = $self->path_display_label ( $self->reading_witnesses( $self->reading( $node ) ) );
+ my $variables = { %edge_attrs, 'label' => $witstr };
+ my $varopts = _dot_attr_string( $variables );
+ $dot .= "\t\"$node\" -> \"#SUBEND#\" $varopts;";
+ }
+
$dot .= "}\n";
return $dot;
}
+sub _dot_attr_string {
+ my( $hash ) = @_;
+ my @attrs;
+ foreach my $k ( sort keys %$hash ) {
+ my $v = $hash->{$k};
+ push( @attrs, $k.'="'.$v.'"' );
+ }
+ return( '[ ' . join( ', ', @attrs ) . ' ]' );
+}
+
sub path_witnesses {
my( $self, @edge ) = @_;
# If edge is an arrayref, cope.
@edge = @$e;
}
my @wits = keys %{$self->sequence->get_edge_attributes( @edge )};
- return sort @wits;
+ return @wits;
}
sub path_display_label {
- my( $self, $edge ) = @_;
- my @wits = $self->path_witnesses( $edge );
+ my $self = shift;
+ my @wits = sort @_;
my $maj = scalar( $self->tradition->witnesses ) * 0.6;
if( scalar @wits > $maj ) {
+ # TODO break out a.c. wits
return 'majority';
} else {
return join( ', ', @wits );
$node_el->setAttribute( 'id', $node_xmlid );
foreach my $d ( keys %node_data ) {
my $nval = $n->$d;
- $nval = $n->punctuated_form if $d eq 'text';
_add_graphml_data( $node_el, $node_data_keys{$d}, $nval )
if defined $nval;
}
my $edge_ctr = 0;
foreach my $e ( sort { $a->[0] cmp $b->[0] } $self->sequence->edges() ) {
# We add an edge in the graphml for every witness in $e.
- foreach my $wit ( $self->path_witnesses( $e ) ) {
+ foreach my $wit ( sort $self->path_witnesses( $e ) ) {
my( $id, $from, $to ) = ( 'e'.$edge_ctr++,
$node_hash{ $e->[0] },
$node_hash{ $e->[1] } );
sub make_alignment_table {
my( $self, $noderefs, $include ) = @_;
unless( $self->linear ) {
- warn "Need a linear graph in order to make an alignment table";
- return;
+ throw( "Need a linear graph in order to make an alignment table" );
}
my $table = { 'alignment' => [], 'length' => $self->end->rank - 1 };
my @all_pos = ( 1 .. $self->end->rank - 1 );
{ 'witness' => $wit->sigil, 'tokens' => \@row } );
if( $wit->is_layered ) {
my @wit_ac_path = $self->reading_sequence( $self->start, $self->end,
- $wit->sigil.$self->ac_label, $wit->sigil );
+ $wit->sigil.$self->ac_label );
my @ac_row = _make_witness_row( \@wit_ac_path, \@all_pos, $noderefs );
push( @{$table->{'alignment'}},
{ 'witness' => $wit->sigil.$self->ac_label, 'tokens' => \@ac_row } );
=cut
# TODO Think about returning some lazy-eval iterator.
+# TODO Get rid of backup; we should know from what witness is whether we need it.
sub reading_sequence {
- my( $self, $start, $end, $witness, $backup ) = @_;
+ my( $self, $start, $end, $witness ) = @_;
$witness = $self->baselabel unless $witness;
my @readings = ( $start );
my $n = $start;
while( $n && $n->id ne $end->id ) {
if( exists( $seen{$n->id} ) ) {
- warn "Detected loop at " . $n->id;
- last;
+ throw( "Detected loop for $witness at " . $n->id );
}
$seen{$n->id} = 1;
- my $next = $self->next_reading( $n, $witness, $backup );
+ my $next = $self->next_reading( $n, $witness );
unless( $next ) {
- warn "Did not find any path for $witness from reading " . $n->id;
- last;
+ throw( "Did not find any path for $witness from reading " . $n->id );
}
push( @readings, $next );
$n = $next;
}
# Check that the last reading is our end reading.
my $last = $readings[$#readings];
- warn "Last reading found from " . $start->text .
- " for witness $witness is not the end!"
+ throw( "Last reading found from " . $start->text .
+ " for witness $witness is not the end!" ) # TODO do we get this far?
unless $last->id eq $end->id;
return @readings;
}
sub _find_linked_reading {
- my( $self, $direction, $node, $path, $alt_path ) = @_;
+ my( $self, $direction, $node, $path ) = @_;
+
+ # Get a backup if we are dealing with a layered witness
+ my $alt_path;
+ my $aclabel = $self->ac_label;
+ if( $path && $path =~ /^(.*)\Q$aclabel\E$/ ) {
+ $alt_path = $1;
+ }
+
my @linked_paths = $direction eq 'next'
? $self->sequence->edges_from( $node )
: $self->sequence->edges_to( $node );
if( $self->sequence->has_edge_attribute( @$le, $self->baselabel ) ) {
$base_le = $le;
}
- my @le_wits = $self->path_witnesses( $le );
+ my @le_wits = sort $self->path_witnesses( $le );
if( _is_within( \@path_wits, \@le_wits ) ) {
# This is the right path.
return $direction eq 'next' ? $le->[1] : $le->[0];
my $regex = $self->wit_list_separator;
my @answer = split( /\Q$regex\E/, $label );
return @answer;
-}
+}
+
+=head2 common_readings
+
+Returns the list of common readings in the graph (i.e. those readings that are
+shared by all non-lacunose witnesses.)
+
+=cut
+
+sub common_readings {
+ my $self = shift;
+ my @common = grep { $_->is_common } $self->readings;
+ return @common;
+}
+
+=head2 path_text( $sigil, $mainsigil [, $start, $end ] )
+
+Returns the text of a witness (plus its backup, if we are using a layer)
+as stored in the collation. The text is returned as a string, where the
+individual readings are joined with spaces and the meta-readings (e.g.
+lacunae) are omitted. Optional specification of $start and $end allows
+the generation of a subset of the witness text.
+
+=cut
+sub path_text {
+ my( $self, $wit, $start, $end ) = @_;
+ $start = $self->start unless $start;
+ $end = $self->end unless $end;
+ my @path = grep { !$_->is_meta } $self->reading_sequence( $start, $end, $wit );
+ return join( ' ', map { $_->text } @path );
+}
=head1 INITIALIZATION METHODS
if( defined $node_ranks->{$rel_containers{$r->id}} ) {
$r->rank( $node_ranks->{$rel_containers{$r->id}} );
} else {
- die "No rank calculated for node " . $r->id
- . " - do you have a cycle in the graph?";
+ # Die. Find the last rank we calculated.
+ my @all_defined = sort { $node_ranks->{$rel_containers{$a->id}}
+ <=> $node_ranks->{$rel_containers{$b->id}} }
+ $self->readings;
+ my $last = pop @all_defined;
+ throw( "Ranks not calculated after $last - do you have a cycle in the graph?" );
}
}
}
}
}
+=head2 calculate_common_readings
+
+Goes through the graph identifying the readings that appear in every witness
+(apart from those with lacunae at that spot.) Marks them as common and returns
+the list.
+
+=begin testing
+
+use Text::Tradition;
+
+my $cxfile = 't/data/Collatex-16.xml';
+my $t = Text::Tradition->new(
+ 'name' => 'inline',
+ 'input' => 'CollateX',
+ 'file' => $cxfile,
+ );
+my $c = $t->collation;
+
+my @common = $c->calculate_common_readings();
+is( scalar @common, 8, "Found correct number of common readings" );
+my @marked = sort $c->common_readings();
+is( scalar @common, 8, "All common readings got marked as such" );
+my @expected = qw/ n1 n12 n16 n19 n20 n5 n6 n7 /;
+is_deeply( \@marked, \@expected, "Found correct list of common readings" );
+
+=end testing
+
+=cut
+
+sub calculate_common_readings {
+ my $self = shift;
+ my @common;
+ my $table = $self->make_alignment_table( 1 );
+ foreach my $idx ( 0 .. $table->{'length'} - 1 ) {
+ my @row = map { $_->{'tokens'}->[$idx]->{'t'} } @{$table->{'alignment'}};
+ my %hash;
+ foreach my $r ( @row ) {
+ if( $r ) {
+ $hash{$r->id} = $r unless $r->is_meta;
+ } else {
+ $hash{'UNDEF'} = $r;
+ }
+ }
+ if( keys %hash == 1 && !exists $hash{'UNDEF'} ) {
+ my( $r ) = values %hash;
+ $r->is_common( 1 );
+ push( @common, $r );
+ }
+ }
+ return @common;
+}
+
+=head2 text_from_paths
+
+Calculate the text array for all witnesses from the path, for later consistency
+checking. Only to be used if there is no non-graph-based way to know the
+original texts.
+
+=cut
+
+sub text_from_paths {
+ my $self = shift;
+ foreach my $wit ( $self->tradition->witnesses ) {
+ my @text = split( /\s+/,
+ $self->reading_sequence( $self->start, $self->end, $wit->sigil ) );
+ $wit->text( \@text );
+ if( $wit->is_layered ) {
+ my @uctext = split( /\s+/,
+ $self->reading_sequence( $self->start, $self->end,
+ $wit->sigil.$self->ac_label ) );
+ $wit->text( \@uctext );
+ }
+ }
+}
=head1 UTILITY FUNCTIONS
return $dir eq 'predecessors' ? pop( @answer ) : shift ( @answer );
}
+sub throw {
+ Text::Tradition::Error->throw(
+ 'ident' => 'Collation error',
+ 'message' => $_[0],
+ );
+}
+
no Moose;
__PACKAGE__->meta->make_immutable;