X-Git-Url: http://git.shadowcat.co.uk/gitweb/gitweb.cgi?a=blobdiff_plain;f=lib%2FText%2FTradition%2FStemma.pm;h=ec1714520261157338e0f6a98e0a282b3d6cb02e;hb=335a62ef9e58093170ab3e2da0945fc545e9f982;hp=448b711c05affd4099d32c519446229d414931d5;hpb=9463b0bff2afe6185d9bdfda49ce9c9cdc176049;p=scpubgit%2Fstemmatology.git diff --git a/lib/Text/Tradition/Stemma.pm b/lib/Text/Tradition/Stemma.pm index 448b711..ec17145 100644 --- a/lib/Text/Tradition/Stemma.pm +++ b/lib/Text/Tradition/Stemma.pm @@ -1,167 +1,400 @@ package Text::Tradition::Stemma; -use File::chdir; +use Bio::Phylo::IO; +use Encode qw( decode_utf8 ); use File::Temp; -use IPC::Run qw/ run /; +use Graph; +use Graph::Reader::Dot; +use IPC::Run qw/ run binary /; +use Text::Tradition::Error; +use Text::Tradition::StemmaUtil qw/ character_input phylip_pars parse_newick /; +use XML::LibXML; use Moose; -use Text::Tradition::Collation::Position; + +=head1 NAME + +Text::Tradition::Stemma - a representation of a I for a Text::Tradition + +=head1 SYNOPSIS + + use Text::Tradition; + my $t = Text::Tradition->new( + 'name' => 'this is a text', + 'input' => 'TEI', + 'file' => '/path/to/tei_parallel_seg_file.xml' ); + + my $s = $tradition->add_stemma( dotfile => '/path/to/stemma.dot' ); + +=head1 DESCRIPTION + +Text::Tradition is a library for representation and analysis of collated +texts, particularly medieval ones. The Stemma is a representation of the +copying relationships between the witnesses in a Tradition, modelled with +a connected rooted directed acyclic graph (CRDAG). + +=head1 DOT SYNTAX + +The easiest way to define a stemma is to use a special form of the 'dot' +syntax of GraphViz. + +Each stemma opens with the line + + digraph Stemma { + +and continues with a list of all manuscript witnesses in the stemma, whether +extant witnesses or missing archetypes or hyparchetypes. Each of these is +listed by its sigil on its own line, e.g.: + + alpha [ class=hypothetical ] + 1 [ class=hypothetical,label=* ] + Ms4 [ class=extant ] + +Extant witnesses are listed with class=extant; missing or postulated witnesses +are listed with class=hypothetical. Anonymous hyparchetypes must be given a +unique name or number, but can be represented as anonymous with the addition +of 'label=*' to their lines. Greek letters or other special characters may be +used as names, but they must always be wrapped in double quotes. + +Links between manuscripts are then listed with arrow notation, as below. These +lines show the direction of copying, one step at a time, for the entire stemma. + + alpha -> 1 + 1 -> Ms4 + +The final line in the definition should be the closing brace: + + } + +Thus for a set of extant manuscripts A, B, and C, where A and B were copied +from the archetype O and C was copied from B, the definition would be: + + digraph Stemma { + O [ class=hypothetical] + A [ class=extant ] + B [ class=extant ] + C [ class=extant ] + O -> A + O -> B + B -> C + } + +=head1 CONSTRUCTOR + +=head2 new + +The constructor. This should generally be called from Text::Tradition, but +if called directly it takes the following options: + +=over + +=item * collation - The collation with which the stemma is associated. + +=item * dot - A filehandle open to a DOT representation of the stemma graph. + +=back + +=cut has collation => ( is => 'ro', isa => 'Text::Tradition::Collation', required => 1, + weak_ref => 1, ); -has character_matrix => ( - is => 'ro', - isa => 'ArrayRef[ArrayRef[Str]]', - writer => '_save_character_matrix', - predicate => 'has_character_matrix', +has graph => ( + is => 'rw', + isa => 'Graph', + predicate => 'has_graph', ); - -sub make_character_matrix { - my $self = shift; - unless( $self->collation->linear ) { - warn "Need a linear graph in order to make an alignment table"; - return; - } - my @all_pos = sort { Text::Tradition::Collation::Position::str_cmp( $a, $b ) } - $self->collation->possible_positions; - my $table = []; - my $characters = {}; - map { $characters->{$_} = {} } @all_pos; - foreach my $wit ( @{$self->collation->tradition->witnesses} ) { - # First implementation: make dumb alignment table, caring about - # nothing except which reading is in which position. - push( @$table, [ $wit->sigil, make_witness_row( $characters, $wit->path, - \@all_pos ) ] ); - if( $wit->has_ante_corr ) { - push( @$table, [ $wit->sigil . "_ac", - make_witness_row( $characters, $wit->uncorrected_path, - \@all_pos ) ] ); - } + +sub BUILD { + my( $self, $args ) = @_; + # If we have been handed a dotfile, initialize it into a graph. + if( exists $args->{'dot'} ) { + $self->_graph_from_dot( $args->{'dot'} ); } - $self->_save_character_matrix( $table ); } -sub make_witness_row { - my( $characters, $path, $positions ) = @_; - my @row; - my $pathdrift = 0; - foreach my $i( 0 .. $#{$positions} ) { - if( $path->[$i-$pathdrift]->position->minref eq $positions->[$i] ) { - push( @row, get_character( $path->[$i-$pathdrift], $characters ) ); +sub _graph_from_dot { + my( $self, $dotfh ) = @_; + my $reader = Graph::Reader::Dot->new(); + my $graph = $reader->read_graph( $dotfh ); + if( $graph ) { + $self->graph( $graph ); + # Go through the nodes and set any non-hypothetical node to extant. + foreach my $v ( $self->graph->vertices ) { + $self->graph->set_vertex_attribute( $v, 'class', 'extant' ) + unless $self->graph->has_vertex_attribute( $v, 'class' ); + } } else { - push( @row, 'X' ); - $pathdrift++; + throw( "Failed to parse dot in $dotfh" ); } - $i++; - } - return @row; } + +=head1 METHODS + +=head2 as_dot( \%options ) + +Returns a normal dot representation of the stemma layout, suitable for rendering +with GraphViz. Options include: + +=over + +=item * graph - A hashref of global graph options. + +=item * node - A hashref of global node options. + +=item * edge - A hashref of global edge options. + +=back + +See the GraphViz documentation for the list of available options. + +=cut + +sub as_dot { + my( $self, $opts ) = @_; + ## See if we are including any a.c. witnesses in this graph. + my $graph = $self->graph; + if( exists $opts->{'layerwits'} ) { + $graph = $self->extend_graph( $opts->{'layerwits'} ); + } + + # Get default and specified options + my %graphopts = ( + # 'ratio' => 1, + ); + my %nodeopts = ( + 'fontsize' => 11, + 'style' => 'filled', + 'fillcolor' => 'white', + 'color' => 'white', + 'shape' => 'ellipse', # Shape for the extant nodes + ); + my %edgeopts = ( + 'arrowhead' => 'none', + ); + @graphopts{ keys %{$opts->{'graph'}} } = values %{$opts->{'graph'}} + if $opts->{'graph'}; + @nodeopts{ keys %{$opts->{'node'}} } = values %{$opts->{'node'}} + if $opts->{'node'}; + @edgeopts{ keys %{$opts->{'edge'}} } = values %{$opts->{'edge'}} + if $opts->{'edge'}; + + my @dotlines; + push( @dotlines, 'digraph stemma {' ); + ## Print out the global attributes + push( @dotlines, _make_dotline( 'graph', %graphopts ) ) if keys %graphopts; + push( @dotlines, _make_dotline( 'edge', %edgeopts ) ) if keys %edgeopts; + push( @dotlines, _make_dotline( 'node', %nodeopts ) ) if keys %nodeopts; -sub get_character { - my( $reading, $characters ) = @_; - my $this_pos = $characters->{$reading->position->minref}; - # This is a simple algorithm that treats every reading as different. - # Eventually we will want to be able to specify how relationships - # affect the character matrix. - my $text = $reading->text; - unless( exists $this_pos->{$text} ) { - # We need to find what the next character is here, and record it. - my @all_chr = sort { $a <=> $b } values( %$this_pos ); - if( @all_chr == 8 ) { - warn "Already have eight variants at position " - . $reading->position->minref . "; not adding " . $reading->text; - return '?'; + # Add each of the nodes. + foreach my $n ( $graph->vertices ) { + if( $graph->has_vertex_attribute( $n, 'label' ) ) { + my $ltext = $graph->get_vertex_attribute( $n, 'label' ); + push( @dotlines, _make_dotline( $n, 'label' => $ltext ) ); + } else { + # Use the default display settings. + $n = _dotquote( $n ); + push( @dotlines, " $n;" ); + } + } + # Add each of our edges. + foreach my $e ( $graph->edges ) { + my( $from, $to ) = map { _dotquote( $_ ) } @$e; + push( @dotlines, " $from -> $to;" ); + } + push( @dotlines, '}' ); + + return join( "\n", @dotlines ); +} + +=head2 editable( $opts ) + +Returns a version of the graph rendered in our definition format. The +output separates statements with a newline; set $opts->{'linesep'} to the +empty string or to a space if the result is to be sent via JSON. + +Any layer witnesses to be included should be passed via $opts->{'layerwits'}. + +=cut + +sub editable { + my( $self, $opts ) = @_; + + ## See if we are including any a.c. witnesses in this graph. + my $graph = $self->graph; + if( exists $opts->{'layerwits'} ) { + $graph = $self->extend_graph( $opts->{'layerwits'} ); } - $this_pos->{$text} = scalar @all_chr; + + # Create the graph + my $join = ( $opts && exists $opts->{'linesep'} ) ? $opts->{'linesep'} : "\n"; + my @dotlines; + push( @dotlines, 'digraph stemma {' ); + my @real; # A cheap sort + foreach my $n ( sort $self->graph->vertices ) { + my $c = $self->graph->get_vertex_attribute( $n, 'class' ); + $c = 'extant' unless $c; + if( $c eq 'extant' ) { + push( @real, $n ); + } else { + push( @dotlines, _make_dotline( $n, 'class' => $c ) ); + } } - return $this_pos->{$text}; + # Now do the real ones + foreach my $n ( @real ) { + push( @dotlines, _make_dotline( $n, 'class' => 'extant' ) ); + } + foreach my $e ( sort _by_vertex $self->graph->edges ) { + my( $from, $to ) = map { _dotquote( $_ ) } @$e; + push( @dotlines, " $from -> $to;" ); + } + push( @dotlines, '}' ); + return join( $join, @dotlines ); +} + +sub _make_dotline { + my( $obj, %attr ) = @_; + my @pairs; + foreach my $k ( keys %attr ) { + my $v = _dotquote( $attr{$k} ); + push( @pairs, "$k=$v" ); + } + return sprintf( " %s [ %s ];", _dotquote( $obj ), join( ', ', @pairs ) ); +} + +sub _dotquote { + my( $str ) = @_; + return $str if $str =~ /^[A-Za-z0-9]+$/; + $str =~ s/\"/\\\"/g; + $str = '"' . $str . '"'; + return $str; +} + +sub _by_vertex { + return $a->[0].$a->[1] cmp $b->[0].$b->[1]; +} + +=head2 extend_graph( $layered_witnesses ) + +Returns a graph which is the original stemma with witness layers added for the +list in @$layered_witnesses. A layered (a.c.) witness is added as a parent +of its main version, and additionally shares all other parents and children with +that version. + +=cut + +sub extend_graph { + my( $self, $layerwits ) = @_; + # For each 'layered' witness in the layerwits array, add it to the graph + # as an ancestor of the 'main' witness, and otherwise with the same parent/ + # child links as its main analogue. + # TOOD Handle case where B is copied from A but corrected from C + + # Iterate through, adding a.c. witnesses + my $actag = $self->collation->ac_label; + my $graph = $self->graph->copy; + foreach my $lw ( @$layerwits ) { + # Add the layered witness and set it with the same attributes as + # its 'main' analogue + my $lwac = $lw . $self->collation->ac_label; + $graph->add_vertex( $lwac ); + $graph->set_vertex_attributes( $lwac, + $graph->get_vertex_attributes( $lw ) ); + + # Set it as ancestor to the main witness + $graph->add_edge( $lwac, $lw ); + + # Give it the same ancestors and descendants as the main witness has, + # bearing in mind that those ancestors and descendants might also just + # have had a layered witness defined. + foreach my $v ( $graph->predecessors( $lw ) ) { + next if $v eq $lwac; # Don't add a loop + $graph->add_edge( $v, $lwac ); + $graph->add_edge( $v.$self->collation->ac_label, $lwac ) + if $graph->has_vertex( $v.$self->collation->ac_label ); + } + foreach my $v ( $graph->successors( $lw ) ) { + next if $v eq $lwac; # but this shouldn't occur + $graph->add_edge( $lwac, $v ); + $graph->add_edge( $lwac, $v.$self->collation->ac_label ) + if $graph->has_vertex( $v.$self->collation->ac_label ); + } + } + return $graph; } -sub run_pars { +=head2 as_svg + +Returns an SVG representation of the graph, calling as_dot first. + +=cut + +sub as_svg { + my( $self, $opts ) = @_; + my $dot = $self->as_dot( $opts ); + my @cmd = qw/dot -Tsvg/; + my $svg; + my $dotfile = File::Temp->new(); + ## TODO REMOVE + # $dotfile->unlink_on_destroy(0); + binmode $dotfile, ':utf8'; + print $dotfile $dot; + push( @cmd, $dotfile->filename ); + run( \@cmd, ">", binary(), \$svg ); + # HACK: Parse the SVG and change the dimensions. + # Get rid of width and height attributes to allow scaling. + my $parser = XML::LibXML->new(); + my $svgdoc = $parser->parse_string( decode_utf8( $svg ) ); + $svgdoc->documentElement->removeAttribute('width'); + $svgdoc->documentElement->removeAttribute('height'); + # Return the result + return decode_utf8( $svgdoc->toString ); +} + +=head2 witnesses + +Returns a list of the extant witnesses represented in the stemma. + +=cut + +sub witnesses { my $self = shift; - $self->make_character_matrix unless $self->has_character_matrix; - - # Set up a temporary directory for all the default Phylip files. - my $phylip_dir = File::Temp->newdir(); - - # We need an infile, and we need a command input file. - $DB::single = 1; - open( MATRIX, ">$phylip_dir/infile" ) or die "Could not write $phylip_dir/infile"; - my $rows = scalar @{$self->character_matrix}; - my $columns = scalar @{$self->character_matrix->[0]} - 1; - print MATRIX "\t$rows\t$columns\n"; - foreach my $row ( @{$self->character_matrix} ) { - my $wit = shift @$row; - my $chars = join( '', @$row ); - print MATRIX sprintf( "%-10s%s\n", $wit, $chars ); - } - close MATRIX; - - open( CMD, ">$phylip_dir/cmdfile" ) or die "Could not write $phylip_dir/cmdfile"; - ## TODO any configuration parameters we want to set here -# U Search for best tree? Yes -# S Search option? More thorough search -# V Number of trees to save? 100 -# J Randomize input order of species? No. Use input order -# O Outgroup root? No, use as outgroup species 1 -# T Use Threshold parsimony? No, use ordinary parsimony -# W Sites weighted? No -# M Analyze multiple data sets? No -# I Input species interleaved? Yes -# 0 Terminal type (IBM PC, ANSI, none)? ANSI -# 1 Print out the data at start of run No -# 2 Print indications of progress of run Yes -# 3 Print out tree Yes -# 4 Print out steps in each site No -# 5 Print character at all nodes of tree No -# 6 Write out trees onto tree file? Yes - print CMD "Y\n"; - close CMD; - - # And then we run the program. - ### HACKY HACKY - my $PHYLIP_PATH = '/Users/tla/Projects/phylip-3.69/exe'; - my $program = "pars"; - if( $^O eq 'darwin' ) { - $program = "$PHYLIP_PATH/$program.app/Contents/MacOS/$program"; - } else { - $program = "$PHYLIP_PATH/$program"; - } + my @wits = grep { $self->graph->get_vertex_attribute( $_, 'class' ) eq 'extant' } + $self->graph->vertices; + return @wits; +} - { - # We need to run it in our temporary directory where we have created - # all the expected files. - local $CWD = $phylip_dir; - my @cmd = ( $program ); - run \@cmd, '<', 'cmdfile', '>', '/dev/null'; - } - # Now our output should be in 'outfile' and our tree in 'outtree', - # both in the temp directory. - - my @outtree; - if( -f "$phylip_dir/outtree" ) { - open( TREE, "$phylip_dir/outtree" ) or die "Could not open outtree for read"; - @outtree = ; - close TREE; - } - return( 1, join( '', @outtree ) ) if @outtree; - - my @error; - if( -f "$phylip_dir/output" ) { - open( OUTPUT, "$phylip_dir/output" ) or die "Could not open output for read"; - @error = ; - close OUTPUT; - } else { - push( @error, "Neither outtree nor output file was produced!" ); - } - return( undef, join( '', @error ) ); +sub hypotheticals { + my $self = shift; + my @wits = grep + { $self->graph->get_vertex_attribute( $_, 'class' ) eq 'hypothetical' } + $self->graph->vertices; + return @wits; +} + +sub throw { + Text::Tradition::Error->throw( + 'ident' => 'Stemma error', + 'message' => $_[0], + ); } + no Moose; __PACKAGE__->meta->make_immutable; 1; + +=head1 LICENSE + +This package is free software and is provided "as is" without express +or implied warranty. You can redistribute it and/or modify it under +the same terms as Perl itself. + +=head1 AUTHOR + +Tara L Andrews Eaurum@cpan.orgE