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=e04811c4b7cb0a1f963276b6cacfe8c2e1d596e9;hpb=e79c23c720dcef073275b5847371b832cac75dfe;p=scpubgit%2Fstemmatology.git diff --git a/lib/Text/Tradition/Stemma.pm b/lib/Text/Tradition/Stemma.pm index e04811c..ec17145 100644 --- a/lib/Text/Tradition/Stemma.pm +++ b/lib/Text/Tradition/Stemma.pm @@ -2,305 +2,399 @@ package Text::Tradition::Stemma; use Bio::Phylo::IO; use Encode qw( decode_utf8 ); -use File::chdir; use File::Temp; use Graph; -use Graph::Convert; 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::Balanced qw/ extract_bracketed /; + +=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', ); - -has apsp => ( - is => 'rw', - isa => 'Graph', - ); - -has distance_trees => ( - is => 'ro', - isa => 'ArrayRef[Graph]', - writer => '_save_distance_trees', - predicate => 'has_distance_trees', - ); sub BUILD { my( $self, $args ) = @_; # If we have been handed a dotfile, initialize it into a graph. if( exists $args->{'dot'} ) { - # Open the file, assume UTF-8 - open( my $dot, $args->{'dot'} ) or warn "Failed to read dot file"; - # TODO don't bother if we haven't opened - binmode $dot, ":utf8"; - my $reader = Graph::Reader::Dot->new(); - my $graph = $reader->read_graph( $dot ); - $graph - ? $self->graph( $graph ) - : warn "Failed to parse dot file " . $args->{'dot'}; + $self->_graph_from_dot( $args->{'dot'} ); } +} + +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 { + throw( "Failed to parse dot in $dotfh" ); + } +} + +=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 ) = @_; - # If we have a graph, calculate all the shortest paths between nodes, - # disregarding direction. - if( $self->has_graph ) { - my $undirected; - if( $self->graph->is_directed ) { - # Make an undirected version. - $undirected = Graph->new( 'undirected' => 1 ); - foreach my $v ( $self->graph->vertices ) { - $undirected->add_vertex( $v ); - } - foreach my $e ( $self->graph->edges ) { - $undirected->add_edge( @$e ); - } + ## 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; + + # 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 { - $undirected = $self->graph; + # Use the default display settings. + $n = _dotquote( $n ); + push( @dotlines, " $n;" ); } - $self->apsp( $undirected->APSP_Floyd_Warshall() ); } + # 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 ); } -# Render the stemma as SVG. -sub as_svg { - my $self = shift; - # TODO add options for display, someday - my $dgraph = Graph::Convert->as_graph_easy( $self->graph ); - # Set some class display attributes for 'hypothetical' and 'extant' nodes - $dgraph->set_attribute( 'flow', 'south' ); - foreach my $n ( $dgraph->nodes ) { - if( $n->attribute( 'class' ) eq 'hypothetical' ) { - $n->set_attribute( 'shape', 'point' ); - $n->set_attribute( 'pointshape', 'diamond' ); - } else { - $n->set_attribute( 'shape', 'ellipse' ); - } +=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'} ); + } + + # 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 ) ); + } } - - # Render to svg via graphviz + # 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; +} + +=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, $err ); + my $svg; my $dotfile = File::Temp->new(); ## TODO REMOVE # $dotfile->unlink_on_destroy(0); binmode $dotfile, ':utf8'; - print $dotfile $dgraph->as_graphviz(); + print $dotfile $dot; push( @cmd, $dotfile->filename ); run( \@cmd, ">", binary(), \$svg ); - $svg = decode_utf8( $svg ); - return $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 ); } -#### Methods for calculating phylogenetic trees #### +=head2 witnesses -before 'distance_trees' => sub { - my $self = shift; - my %args = @_; - # TODO allow specification of method for calculating distance tree - if( $args{'recalc'} || !$self->has_distance_trees ) { - # We need to make a tree before we can return it. - my( $ok, $result ) = $self->run_phylip_pars(); - if( $ok ) { - $self->_save_distance_trees( _parse_newick( $result ) ); - } else { - warn "Failed to calculate distance tree: $result"; - } - } -}; - -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 $table = $self->collation->make_alignment_table; - # Push the names of the witnesses to initialize the rows of the matrix. - my @matrix = map { [ $self->_normalize_ac( $_ ) ] } @{$table->[0]}; - $DB::single = 1; - foreach my $token_index ( 1 .. $#{$table} ) { - # First implementation: make dumb alignment table, caring about - # nothing except which reading is in which position. - my @chars = convert_characters( $table->[$token_index] ); - foreach my $idx ( 0 .. $#matrix ) { - push( @{$matrix[$idx]}, $chars[$idx] ); - } - } - $self->_save_character_matrix( \@matrix ); -} - -sub _normalize_ac { - my( $self, $witname ) = @_; - my $ac = $self->collation->ac_label; - if( $witname =~ /(.*)\Q$ac\E$/ ) { - $witname = $1 . '_ac'; - } - return sprintf( "%-10s", $witname ); -} +Returns a list of the extant witnesses represented in the stemma. -sub convert_characters { - my $row = shift; - # 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 %unique = ( '__UNDEF__' => 'X', - '#LACUNA#' => '?', - ); - my $ctr = 0; - foreach my $word ( @$row ) { - if( $word && !exists $unique{$word} ) { - $unique{$word} = chr( 65 + $ctr ); - $ctr++; - } - } - if( scalar( keys %unique ) > 8 ) { - warn "Have more than 8 variants on this location; phylip will break"; - } - my @chars = map { $_ ? $unique{$_} : $unique{'__UNDEF__' } } @$row; - return @chars; -} +=cut -sub phylip_pars_input { +sub witnesses { my $self = shift; - $self->make_character_matrix unless $self->has_character_matrix; - my $matrix = ''; - my $rows = scalar @{$self->character_matrix}; - my $columns = scalar @{$self->character_matrix->[0]} - 1; - $matrix .= "\t$rows\t$columns\n"; - foreach my $row ( @{$self->character_matrix} ) { - $matrix .= join( '', @$row ) . "\n"; - } - return $matrix; + my @wits = grep { $self->graph->get_vertex_attribute( $_, 'class' ) eq 'extant' } + $self->graph->vertices; + return @wits; } -sub run_phylip_pars { +sub hypotheticals { my $self = shift; - - # Set up a temporary directory for all the default Phylip files. - my $phylip_dir = File::Temp->newdir(); - # $phylip_dir->unlink_on_destroy(0); - # We need an infile, and we need a command input file. - open( MATRIX, ">$phylip_dir/infile" ) or die "Could not write $phylip_dir/infile"; - print MATRIX $self->phylip_pars_input(); - close MATRIX; - - open( CMD, ">$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"; - } - - { - # 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/outfile" ) { - open( OUTPUT, "$phylip_dir/outfile" ) 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 ) ); + my @wits = grep + { $self->graph->get_vertex_attribute( $_, 'class' ) eq 'hypothetical' } + $self->graph->vertices; + return @wits; } -sub _parse_newick { - my $newick = shift; - my @trees; - # Parse the result into a tree - my $forest = Bio::Phylo::IO->parse( - -format => 'newick', - -string => $newick, - ); - # Turn the tree into a graph, starting with the root node - foreach my $tree ( @{$forest->get_entities} ) { - push( @trees, _graph_from_bio( $tree ) ); - } - return \@trees; +sub throw { + Text::Tradition::Error->throw( + 'ident' => 'Stemma error', + 'message' => $_[0], + ); } -sub _graph_from_bio { - my $tree = shift; - my $graph = Graph->new( 'undirected' => 1 ); - # Give all the intermediate anonymous nodes a name. - my $i = 0; - foreach my $n ( @{$tree->get_entities} ) { - next if $n->get_name; - $n->set_name( $i++ ); - } - my $root = $tree->get_root->get_name; - $graph->add_vertex( $root ); - _add_tree_children( $graph, $root, $tree->get_root->get_children() ); - return $graph; -} - -sub _add_tree_children { - my( $graph, $parent, $tree_children ) = @_; - foreach my $c ( @$tree_children ) { - my $child = $c->get_name; - $graph->add_vertex( $child ); - $graph->add_path( $parent, $child ); - _add_tree_children( $graph, $child, $c->get_children() ); - } -} 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