[scpubgit/stemmatology.git] / analysis / lib / Text / Tradition / StemmaUtil.pm

package Text::Tradition::StemmaUtil;

use strict;
use warnings;
use Exporter 'import';
use vars qw/ @EXPORT_OK /;
use Bio::Phylo::IO;
use Encode qw( decode_utf8 );
use File::chdir;
use File::Temp;
use File::Which;
use Graph;
use Graph::Reader::Dot;
use IPC::Run qw/ run binary /;
use Text::Tradition::Error;
@EXPORT_OK = qw/ read_graph display_graph editable_graph 
	character_input phylip_pars parse_newick newick_to_svg /;

=head1 NAME

Text::Tradition::StemmaUtil - standalone utilities for stemma graph display and
distance tree calculations

=head1 DESCRIPTION

This package contains a set of utilities for displaying arbitrary stemmata and 
running phylogenetic analysis on text collations.

=head1 SUBROUTINES

=head2 read_graph( $dotstr) {

Parses the graph specification on the filehandle in $dotstr and returns a Graph 
object. This subroutine works around some deficiencies in Graph::Reader::Dot.

=cut

sub read_graph {
	my $dotstr = shift;
	# Graph::Reader::Dot does not handle bare non-ASCII Unicode word characters.
	# We get around this by wrapping all words in double quotes, as long as they 
	# aren't already wrapped, and as long as they aren't the initial '(di)?graph'.
	# Also need to deal correctly with the graph title.
	if( $dotstr =~ /^\s*((di)?graph)\s+(.*?)\s*\{(.*)$/s ) {
		my( $decl, $ident, $rest ) = ( $1, $3, $4 );
		unless( substr( $ident, 0, 1 ) eq '"' ) {
			$ident = '"'.$ident.'"';
		}
		$rest =~ s/(?<!")\b(\w+)\b(?!")/"$1"/g;
		$dotstr = "$decl $ident { $rest";
	}
		
	# Now open a filehandle onto the string and pass it to Graph::Reader::Dot.
	my $dotfh;
	open $dotfh, '<', \$dotstr;
	binmode $dotfh, ':utf8';
 	my $reader = Graph::Reader::Dot->new();
 	# Redirect STDOUT in order to trap any error messages - syntax errors
 	# are evidently not fatal.
	my $graph;
	my $reader_out;
	my $reader_err;
	{
		local(*STDOUT);
		open( STDOUT, ">", \$reader_out );
		local(*STDERR);
		open( STDERR, ">", \$reader_err );
		$graph = $reader->read_graph( $dotfh );
		close STDOUT;
		close STDERR;
	}
	if( $reader_out && $reader_out =~ /error/s ) {
		throw( "Error trying to parse dot: $reader_out" );
	} elsif( !$graph ) {
		throw( "Failed to create graph from dot" );
	}
	# Wrench the graph identifier out of the graph
	## HORRIBLE HACK but there is no API access to the graph identifier!
	$graph->set_graph_attribute( 'name', $graph->[4]->{'name'} );

	# Correct for implicit graph -> digraph quirk of reader
	if( $reader_err && $reader_err =~ /graph will be treated as digraph/ ) {
		my $udgraph = $graph->undirected_copy;
		$udgraph->set_graph_attributes( $graph->get_graph_attributes );
		foreach my $v ( $graph->vertices ) {
			$udgraph->set_vertex_attributes( $v, $graph->get_vertex_attributes( $v ) );
		}
		$graph = $udgraph;
	}
	
	return $graph;
}

=head2 display_graph( $graph, $opts )

Returns a dot specification intended for display, according to the logical 
attributes of the witnesses.

=cut

sub display_graph {
    my( $graph, $opts ) = @_;
    
    # Get default and specified options
    my %graphopts = (
    	# 'ratio' => 1,
    	'bgcolor' => 'transparent',
    );
    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 $gdecl = $graph->is_directed ? 'digraph' : 'graph';
	my $gname = $opts->{'name'} ? '"' . $opts->{'name'} . '"'
		: 'stemma';
	my @dotlines;
	push( @dotlines, "$gdecl $gname {" );
	## 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 ) {
    	my %vattr = ( 'id' => $n );  # Set the SVG element ID to the sigil itself
        if( $graph->has_vertex_attribute( $n, 'label' ) ) {
        	$vattr{'label'} = $graph->get_vertex_attribute( $n, 'label' );
        }
		push( @dotlines, _make_dotline( $n, %vattr ) );
    }
    # Add each of our edges.
    foreach my $e ( $graph->edges ) {
    	my( $from, $to ) = map { _dotquote( $_ ) } @$e;
    	my $connector = $graph->is_directed ? '->' : '--';
    	push( @dotlines, "  $from $connector $to;" );
    }
    push( @dotlines, '}' );
    
    return join( "\n", @dotlines );
}


=head2 editable_graph( $graph, $opts )

Returns a dot specification of a stemma graph with logical witness features,
intended for editing the stemma definition.

=cut

sub editable_graph {
	my( $graph, $opts ) = @_;

	# Create the graph
	my $join = ( $opts && exists $opts->{'linesep'} ) ? $opts->{'linesep'} : "\n";
	my $gdecl = $graph->is_undirected ? 'graph' : 'digraph';
	my $gname = exists $opts->{'name'} ? '"' . $opts->{'name'} . '"'
		: 'stemma';
	my @dotlines;
	push( @dotlines, "$gdecl $gname {" );
	my @real; # A cheap sort
    foreach my $n ( sort $graph->vertices ) {
    	my $c = $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 ) );
		}
    }
	# Now do the real ones
	foreach my $n ( @real ) {
		push( @dotlines, _make_dotline( $n, 'class' => 'extant' ) );
	}
	foreach my $e ( sort _by_vertex $graph->edges ) {
		my( $from, $to ) = map { _dotquote( $_ ) } @$e;
		my $conn = $graph->is_undirected ? '--' : '->';
		push( @dotlines, "  $from $conn $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 character_input( $tradition, $opts )

Returns a character matrix string suitable for Phylip programs, which 
corresponds to the given alignment table.  See Text::Tradition::Collation 
for a description of the alignment table format. Options include:

=over

=item * exclude_layer - Exclude layered witnesses from the character input, 
using only the 'main' text of the witnesses in the tradition.

=item * collapse - A reference to an array of relationship names that should
be treated as equivalent for the purposes of generating the character matrix.

=back

=cut

sub character_input {
    my ( $tradition, $opts ) = @_;
    my $table = $tradition->collation->alignment_table;
    if( $opts->{exclude_layer} ) {
    	# Filter out all alignment table rows that do not correspond
    	# to a named witness - these are the layered witnesses.
    	my $newtable = { alignment => [], length => $table->{length} };
    	foreach my $row ( @{$table->{alignment}} ) {
    		if( $tradition->has_witness( $row->{witness} ) ) {
    			push( @{$newtable->{alignment}}, $row );
    		}
    	}
    	$table = $newtable;
    }
    my $character_matrix = _make_character_matrix( $table, $opts );
    my $input = '';
    my $rows = scalar @{$character_matrix};
    my $columns = scalar @{$character_matrix->[0]} - 1;
    $input .= "\t$rows\t$columns\n";
    foreach my $row ( @{$character_matrix} ) {
        $input .= join( '', @$row ) . "\n";
    }
    return $input;
}

sub _make_character_matrix {
    my( $table, $opts ) = @_;
    # Push the names of the witnesses to initialize the rows of the matrix.
    my @matrix = map { [ _normalize_witname( $_->{'witness'} ) ] } 
    				@{$table->{'alignment'}};
    foreach my $token_index ( 0 .. $table->{'length'} - 1) {
        my @pos_tokens = map { $_->{'tokens'}->[$token_index] }
        						@{$table->{'alignment'}};
        my @pos_readings = map { $_ ? $_->{'t'} : $_ } @pos_tokens;
        my @chars = _convert_characters( \@pos_readings, $opts );
        foreach my $idx ( 0 .. $#matrix ) {
            push( @{$matrix[$idx]}, $chars[$idx] );
        }
    }
    return \@matrix;
} 

# Helper function to make the witness name something legal for pars

sub _normalize_witname {
    my( $witname ) = @_;
    $witname =~ s/\s+/ /g;
    $witname =~ s/[\[\]\(\)\:;,]//g;
    $witname = substr( $witname, 0, 10 );
    return sprintf( "%-10s", $witname );
}

sub _convert_characters {
    my( $row, $opts ) = @_;
    # 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 %equivalent;
    my %count;
    my $ctr = 0;
    foreach my $rdg ( @$row ) {
    	next unless $rdg;
    	next if $rdg->is_lacuna;
		next if exists $unique{$rdg->text};
		if( ref( $opts->{'collapse'} ) eq 'ARRAY' ) {
			my @exclude_types = @{$opts->{'collapse'}};
			my @set = $rdg->related_readings( sub { my $rel = shift;
				$rel->colocated && grep { $rel->type eq $_ } @exclude_types } );
			push( @set, $rdg );
			my $char = chr( 65 + $ctr++ );
			map { $unique{$_->text} = $char } @set;
			$count{$rdg->text} += scalar @set;
		} else {
			$unique{$rdg->text} = chr( 65 + $ctr++ );
			$count{$rdg->text}++;			
		}
    }
    # Try to keep variants under 8 by lacunizing any singletons.
    if( scalar( keys %unique ) > 8 ) {
		foreach my $word ( keys %count ) {
			if( $count{$word} == 1 ) {
				$unique{$word} = '?';
			}
		}
    }
    my %u = reverse %unique;
    if( scalar( keys %u ) > 8 ) {
        warn "Have more than 8 variants on this location; phylip will break";
    }
    my @chars = map { $_ ? $unique{$_->text} : $unique{'__UNDEF__' } } @$row;
    return @chars;
}

=head2 phylip_pars( $character_matrix )

Runs Phylip Pars on the given character matrix.  Returns results in Newick format.

=cut

sub phylip_pars {
	my( $charmatrix ) = @_;
    # 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 $charmatrix;
    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.
    my $program = File::Which::which( 'pars' );
    unless( $program && -x $program ) {
		throw( "Phylip pars not found in path" );
    }

    {
        # 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 = <TREE>;
        close TREE;
    }
    return join( '', @outtree ) if @outtree;

	# If we got this far, we are about to throw an error.
    my @error;
    if( -f "$phylip_dir/outfile" ) {
        open( OUTPUT, "$phylip_dir/outfile" ) or die "Could not open output for read";
        @error = <OUTPUT>;
        close OUTPUT;
    } else {
        push( @error, "Neither outtree nor output file was produced!" );
    }
    throw( join( '', @error ) );
}

=head2 parse_newick( $newick_string )

Parses the given Newick tree(s) into one or more Stemma objects with
undirected graphs.

=cut

sub parse_newick {
    my $newick = shift;
    # Parse the result into a set of trees and return them.
    my $forest = Bio::Phylo::IO->parse( 
        -format => 'newick',
        -string => $newick,
        );
    return map { _graph_from_bio( $_ ) } @{$forest->get_entities};
}

sub _graph_from_bio {
    my $tree = shift;
    my $graph = Graph->new( 'undirected' => 1 );
    # Give all the intermediate anonymous nodes a name.
    my $i = 0;
    my $classes = {};
    foreach my $n ( @{$tree->get_terminals} ) {
    	# The terminal nodes are our named witnesses.
		$classes->{$n->get_name} = 'extant';
	}
	foreach my $n ( @{$tree->get_internals} ) {
    	unless( defined $n->get_name && $n->get_name ne '' ) {
    		# Get an integer, make sure it's a unique name
    		while( exists $classes->{$i} ) {
    			$i++;
    		}
    		$n->set_name( $i++ );
    	}
    	$classes->{$n->get_name} = 'hypothetical';
    }
    _add_tree_children( $graph, $classes, undef, [ $tree->get_root ]);
    return $graph;
}

sub _add_tree_children {
    my( $graph, $classes, $parent, $tree_children ) = @_;
    foreach my $c ( @$tree_children ) {
        my $child = $c->get_name;
        $graph->add_vertex( $child );
        $graph->set_vertex_attribute( $child, 'class', $classes->{$child} );
        $graph->add_path( $parent, $child ) if defined $parent;
        _add_tree_children( $graph, $classes, $child, $c->get_children() );
    }
}

=head2 newick_to_svg( $newick_string )

Uses the FigTree utility (if installed) to transform the given Newick tree(s)
into a graph visualization.

=cut

sub newick_to_svg {
	my $newick = shift;
    my $program = File::Which::which( 'figtree' );
    unless( -x $program ) {
		throw( "FigTree commandline utility not found in path" );
    }
    my $svg;
    my $nfile = File::Temp->new();
    print $nfile $newick;
    close $nfile;
	my @cmd = ( $program, '-graphic', 'SVG', $nfile );
    run( \@cmd, ">", binary(), \$svg );
    return decode_utf8( $svg );
}

sub throw {
	Text::Tradition::Error->throw( 
		'ident' => 'StemmaUtil error',
		'message' => $_[0],
		);
}

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 E<lt>aurum@cpan.orgE<gt>
Commit	Line	Data
68454b71	1	package Text::Tradition::StemmaUtil;
	2
	3	use strict;
	4	use warnings;
	5	use Exporter 'import';
	6	use vars qw/ @EXPORT_OK /;
b02332ca	7	use Bio::Phylo::IO;
27e2a8fe	8	use Encode qw( decode_utf8 );
b02332ca	9	use File::chdir;
	10	use File::Temp;
	11	use File::Which;
	12	use Graph;
11954259	13	use Graph::Reader::Dot;
b02332ca	14	use IPC::Run qw/ run binary /;
edac47cc	15	use Text::Tradition::Error;
11954259	16	@EXPORT_OK = qw/ read_graph display_graph editable_graph
5873cf38	17	character_input phylip_pars parse_newick newick_to_svg /;
68454b71	18
027d819c	19	=head1 NAME
027d819c	20
5873cf38	21	Text::Tradition::StemmaUtil - standalone utilities for stemma graph display and
5873cf38	22	distance tree calculations
027d819c	23
	24	=head1 DESCRIPTION
	25
5873cf38	26	This package contains a set of utilities for displaying arbitrary stemmata and
5873cf38	27	running phylogenetic analysis on text collations.
027d819c	28
	29	=head1 SUBROUTINES
	30
11954259	31	=head2 read_graph( $dotstr) {
	32
	33	Parses the graph specification on the filehandle in $dotstr and returns a Graph
	34	object. This subroutine works around some deficiencies in Graph::Reader::Dot.
	35
	36	=cut
	37
	38	sub read_graph {
	39	my $dotstr = shift;
	40	# Graph::Reader::Dot does not handle bare non-ASCII Unicode word characters.
	41	# We get around this by wrapping all words in double quotes, as long as they
	42	# aren't already wrapped, and as long as they aren't the initial '(di)?graph'.
	43	# Also need to deal correctly with the graph title.
	44	if( $dotstr =~ /^\s((di)?graph)\s+(.?)\s\{(.)$/s ) {
	45	my( $decl, $ident, $rest ) = ( $1, $3, $4 );
	46	unless( substr( $ident, 0, 1 ) eq '"' ) {
	47	$ident = '"'.$ident.'"';
	48	}
	49	$rest =~ s/(?<!")\b(\w+)\b(?!")/"$1"/g;
	50	$dotstr = "$decl $ident { $rest";
	51	}
	52
	53	# Now open a filehandle onto the string and pass it to Graph::Reader::Dot.
	54	my $dotfh;
	55	open $dotfh, '<', \$dotstr;
	56	binmode $dotfh, ':utf8';
	57	my $reader = Graph::Reader::Dot->new();
	58	# Redirect STDOUT in order to trap any error messages - syntax errors
	59	# are evidently not fatal.
	60	my $graph;
	61	my $reader_out;
	62	my $reader_err;
	63	{
	64	local(*STDOUT);
	65	open( STDOUT, ">", \$reader_out );
	66	local(*STDERR);
	67	open( STDERR, ">", \$reader_err );
	68	$graph = $reader->read_graph( $dotfh );
	69	close STDOUT;
	70	close STDERR;
	71	}
	72	if( $reader_out && $reader_out =~ /error/s ) {
	73	throw( "Error trying to parse dot: $reader_out" );
	74	} elsif( !$graph ) {
	75	throw( "Failed to create graph from dot" );
	76	}
	77	# Wrench the graph identifier out of the graph
	78	## HORRIBLE HACK but there is no API access to the graph identifier!
	79	$graph->set_graph_attribute( 'name', $graph->[4]->{'name'} );
	80
	81	# Correct for implicit graph -> digraph quirk of reader
	82	if( $reader_err && $reader_err =~ /graph will be treated as digraph/ ) {
	83	my $udgraph = $graph->undirected_copy;
	84	$udgraph->set_graph_attributes( $graph->get_graph_attributes );
	85	foreach my $v ( $graph->vertices ) {
	86	$udgraph->set_vertex_attributes( $v, $graph->get_vertex_attributes( $v ) );
	87	}
	88	$graph = $udgraph;
	89	}
	90
	91	return $graph;
	92	}
	93
5873cf38	94	=head2 display_graph( $graph, $opts )
	95
	96	Returns a dot specification intended for display, according to the logical
	97	attributes of the witnesses.
	98
	99	=cut
	100
	101	sub display_graph {
	102	my( $graph, $opts ) = @_;
	103
	104	# Get default and specified options
	105	my %graphopts = (
	106	# 'ratio' => 1,
	107	'bgcolor' => 'transparent',
	108	);
	109	my %nodeopts = (
	110	'fontsize' => 11,
	111	'style' => 'filled',
	112	'fillcolor' => 'white',
	113	'color' => 'white',
	114	'shape' => 'ellipse', # Shape for the extant nodes
	115	);
	116	my %edgeopts = (
	117	'arrowhead' => 'none',
	118	);
	119	@graphopts{ keys %{$opts->{'graph'}} } = values %{$opts->{'graph'}}
	120	if $opts->{'graph'};
	121	@nodeopts{ keys %{$opts->{'node'}} } = values %{$opts->{'node'}}
	122	if $opts->{'node'};
	123	@edgeopts{ keys %{$opts->{'edge'}} } = values %{$opts->{'edge'}}
	124	if $opts->{'edge'};
	125
	126	my $gdecl = $graph->is_directed ? 'digraph' : 'graph';
	127	my $gname = $opts->{'name'} ? '"' . $opts->{'name'} . '"'
	128	: 'stemma';
	129	my @dotlines;
	130	push( @dotlines, "$gdecl $gname {" );
	131	## Print out the global attributes
	132	push( @dotlines, _make_dotline( 'graph', %graphopts ) ) if keys %graphopts;
	133	push( @dotlines, _make_dotline( 'edge', %edgeopts ) ) if keys %edgeopts;
	134	push( @dotlines, _make_dotline( 'node', %nodeopts ) ) if keys %nodeopts;
	135
	136	# Add each of the nodes.
	137	foreach my $n ( $graph->vertices ) {
	138	my %vattr = ( 'id' => $n ); # Set the SVG element ID to the sigil itself
	139	if( $graph->has_vertex_attribute( $n, 'label' ) ) {
	140	$vattr{'label'} = $graph->get_vertex_attribute( $n, 'label' );
	141	}
	142	push( @dotlines, _make_dotline( $n, %vattr ) );
	143	}
	144	# Add each of our edges.
	145	foreach my $e ( $graph->edges ) {
	146	my( $from, $to ) = map { _dotquote( $_ ) } @$e;
	147	my $connector = $graph->is_directed ? '->' : '--';
	148	push( @dotlines, " $from $connector $to;" );
	149	}
	150	push( @dotlines, '}' );
	151
	152	return join( "\n", @dotlines );
	153	}
	154
	155
	156	=head2 editable_graph( $graph, $opts )
	157
158	Returns a dot specification of a stemma graph with logical witness features,
159	intended for editing the stemma definition.
160
161	=cut
162
163	sub editable_graph {
164	my( $graph, $opts ) = @_;
165
166	# Create the graph
167	my $join = ( $opts && exists $opts->{'linesep'} ) ? $opts->{'linesep'} : "\n";
5873cf38	168	my $gdecl = $graph->is_undirected ? 'graph' : 'digraph';
	169	my $gname = exists $opts->{'name'} ? '"' . $opts->{'name'} . '"'
	170	: 'stemma';
	171	my @dotlines;
	172	push( @dotlines, "$gdecl $gname {" );
	173	my @real; # A cheap sort
	174	foreach my $n ( sort $graph->vertices ) {
	175	my $c = $graph->get_vertex_attribute( $n, 'class' );
	176	$c = 'extant' unless $c;
	177	if( $c eq 'extant' ) {
	178	push( @real, $n );
	179	} else {
e07fdf07	180	push( @dotlines, _make_dotline( $n, 'class' => $c ) );
5873cf38	181	}
	182	}
	183	# Now do the real ones
	184	foreach my $n ( @real ) {
e07fdf07	185	push( @dotlines, _make_dotline( $n, 'class' => 'extant' ) );
5873cf38	186	}
	187	foreach my $e ( sort _by_vertex $graph->edges ) {
	188	my( $from, $to ) = map { _dotquote( $_ ) } @$e;
	189	my $conn = $graph->is_undirected ? '--' : '->';
	190	push( @dotlines, " $from $conn $to;" );
	191	}
	192	push( @dotlines, '}' );
	193	return join( $join, @dotlines );
	194	}
	195
	196	sub _make_dotline {
	197	my( $obj, %attr ) = @_;
	198	my @pairs;
5873cf38	199	foreach my $k ( keys %attr ) {
e07fdf07	200	my $v = _dotquote( $attr{$k} );
5873cf38	201	push( @pairs, "$k=$v" );
	202	}
	203	return sprintf( " %s [ %s ];", _dotquote( $obj ), join( ', ', @pairs ) );
	204	}
	205
	206	sub _dotquote {
e07fdf07	207	my( $str ) = @_;
e07fdf07	208	return $str if $str =~ /^[A-Za-z0-9_]+$/;
5873cf38	209	$str =~ s/\"/\\\"/g;
	210	$str = '"' . $str . '"';
	211	return $str;
	212	}
	213
	214	sub _by_vertex {
	215	return $a->[0].$a->[1] cmp $b->[0].$b->[1];
	216	}
	217
ea45d2a6	218	=head2 character_input( $tradition, $opts )
9457207b	219
	220	Returns a character matrix string suitable for Phylip programs, which
	221	corresponds to the given alignment table. See Text::Tradition::Collation
ea45d2a6	222	for a description of the alignment table format. Options include:
	223
	224	=over
	225
	226	=item * exclude_layer - Exclude layered witnesses from the character input,
	227	using only the 'main' text of the witnesses in the tradition.
	228
	229	=item * collapse - A reference to an array of relationship names that should
	230	be treated as equivalent for the purposes of generating the character matrix.
	231
	232	=back
9457207b	233
027d819c	234	=cut
027d819c	235
9457207b	236	sub character_input {
ea45d2a6	237	my ( $tradition, $opts ) = @_;
	238	my $table = $tradition->collation->alignment_table;
	239	if( $opts->{exclude_layer} ) {
	240	# Filter out all alignment table rows that do not correspond
	241	# to a named witness - these are the layered witnesses.
b39e7cb5	242	my $newtable = { alignment => [], length => $table->{length} };
b39e7cb5	243	foreach my $row ( @{$table->{alignment}} ) {
ea45d2a6	244	if( $tradition->has_witness( $row->{witness} ) ) {
	245	push( @{$newtable->{alignment}}, $row );
	246	}
	247	}
	248	$table = $newtable;
	249	}
	250	my $character_matrix = _make_character_matrix( $table, $opts );
9457207b	251	my $input = '';
	252	my $rows = scalar @{$character_matrix};
	253	my $columns = scalar @{$character_matrix->[0]} - 1;
	254	$input .= "\t$rows\t$columns\n";
	255	foreach my $row ( @{$character_matrix} ) {
	256	$input .= join( '', @$row ) . "\n";
	257	}
	258	return $input;
	259	}
	260
027d819c	261	sub _make_character_matrix {
ea45d2a6	262	my( $table, $opts ) = @_;
68454b71	263	# Push the names of the witnesses to initialize the rows of the matrix.
	264	my @matrix = map { [ _normalize_witname( $_->{'witness'} ) ] }
	265	@{$table->{'alignment'}};
	266	foreach my $token_index ( 0 .. $table->{'length'} - 1) {
ea45d2a6	267	my @pos_tokens = map { $_->{'tokens'}->[$token_index] }
68454b71	268	@{$table->{'alignment'}};
ea45d2a6	269	my @pos_readings = map { $_ ? $_->{'t'} : $_ } @pos_tokens;
ea45d2a6	270	my @chars = _convert_characters( \@pos_readings, $opts );
68454b71	271	foreach my $idx ( 0 .. $#matrix ) {
	272	push( @{$matrix[$idx]}, $chars[$idx] );
	273	}
	274	}
	275	return \@matrix;
	276	}
	277
	278	# Helper function to make the witness name something legal for pars
	279
	280	sub _normalize_witname {
	281	my( $witname ) = @_;
	282	$witname =~ s/\s+/ /g;
	283	$witname =~ s/[\[\]\(\)\:;,]//g;
	284	$witname = substr( $witname, 0, 10 );
	285	return sprintf( "%-10s", $witname );
	286	}
	287
027d819c	288	sub _convert_characters {
ea45d2a6	289	my( $row, $opts ) = @_;
68454b71	290	# This is a simple algorithm that treats every reading as different.
	291	# Eventually we will want to be able to specify how relationships
	292	# affect the character matrix.
	293	my %unique = ( '__UNDEF__' => 'X',
	294	'#LACUNA#' => '?',
	295	);
ea45d2a6	296	my %equivalent;
68454b71	297	my %count;
68454b71	298	my $ctr = 0;
ea45d2a6	299	foreach my $rdg ( @$row ) {
	300	next unless $rdg;
	301	next if $rdg->is_lacuna;
	302	next if exists $unique{$rdg->text};
	303	if( ref( $opts->{'collapse'} ) eq 'ARRAY' ) {
	304	my @exclude_types = @{$opts->{'collapse'}};
	305	my @set = $rdg->related_readings( sub { my $rel = shift;
	306	$rel->colocated && grep { $rel->type eq $_ } @exclude_types } );
	307	push( @set, $rdg );
	308	my $char = chr( 65 + $ctr++ );
	309	map { $unique{$_->text} = $char } @set;
	310	$count{$rdg->text} += scalar @set;
	311	} else {
	312	$unique{$rdg->text} = chr( 65 + $ctr++ );
	313	$count{$rdg->text}++;
	314	}
68454b71	315	}
	316	# Try to keep variants under 8 by lacunizing any singletons.
	317	if( scalar( keys %unique ) > 8 ) {
	318	foreach my $word ( keys %count ) {
	319	if( $count{$word} == 1 ) {
	320	$unique{$word} = '?';
	321	}
	322	}
	323	}
	324	my %u = reverse %unique;
	325	if( scalar( keys %u ) > 8 ) {
	326	warn "Have more than 8 variants on this location; phylip will break";
	327	}
ea45d2a6	328	my @chars = map { $_ ? $unique{$_->text} : $unique{'__UNDEF__' } } @$row;
68454b71	329	return @chars;
	330	}
	331
027d819c	332	=head2 phylip_pars( $character_matrix )
	333
	334	Runs Phylip Pars on the given character matrix. Returns results in Newick format.
	335
	336	=cut
	337
b02332ca	338	sub phylip_pars {
	339	my( $charmatrix ) = @_;
	340	# Set up a temporary directory for all the default Phylip files.
	341	my $phylip_dir = File::Temp->newdir();
	342	# $phylip_dir->unlink_on_destroy(0);
	343	# We need an infile, and we need a command input file.
	344	open( MATRIX, ">$phylip_dir/infile" ) or die "Could not write $phylip_dir/infile";
	345	print MATRIX $charmatrix;
	346	close MATRIX;
	347
	348	open( CMD, ">$phylip_dir/cmdfile" ) or die "Could not write $phylip_dir/cmdfile";
	349	## TODO any configuration parameters we want to set here
	350	# U Search for best tree? Yes
	351	# S Search option? More thorough search
	352	# V Number of trees to save? 100
	353	# J Randomize input order of species? No. Use input order
	354	# O Outgroup root? No, use as outgroup species 1
	355	# T Use Threshold parsimony? No, use ordinary parsimony
	356	# W Sites weighted? No
	357	# M Analyze multiple data sets? No
	358	# I Input species interleaved? Yes
	359	# 0 Terminal type (IBM PC, ANSI, none)? ANSI
	360	# 1 Print out the data at start of run No
	361	# 2 Print indications of progress of run Yes
	362	# 3 Print out tree Yes
	363	# 4 Print out steps in each site No
	364	# 5 Print character at all nodes of tree No
	365	# 6 Write out trees onto tree file? Yes
	366	print CMD "Y\n";
	367	close CMD;
	368
	369	# And then we run the program.
	370	my $program = File::Which::which( 'pars' );
b39e7cb5	371	unless( $program && -x $program ) {
edac47cc	372	throw( "Phylip pars not found in path" );
b02332ca	373	}
	374
	375	{
	376	# We need to run it in our temporary directory where we have created
	377	# all the expected files.
	378	local $CWD = $phylip_dir;
	379	my @cmd = ( $program );
	380	run \@cmd, '<', 'cmdfile', '>', '/dev/null';
	381	}
	382	# Now our output should be in 'outfile' and our tree in 'outtree',
	383	# both in the temp directory.
	384
	385	my @outtree;
	386	if( -f "$phylip_dir/outtree" ) {
	387	open( TREE, "$phylip_dir/outtree" ) or die "Could not open outtree for read";
	388	@outtree = <TREE>;
	389	close TREE;
	390	}
edac47cc	391	return join( '', @outtree ) if @outtree;
b02332ca	392
69403daa	393	# If we got this far, we are about to throw an error.
b02332ca	394	my @error;
	395	if( -f "$phylip_dir/outfile" ) {
	396	open( OUTPUT, "$phylip_dir/outfile" ) or die "Could not open output for read";
	397	@error = <OUTPUT>;
	398	close OUTPUT;
	399	} else {
	400	push( @error, "Neither outtree nor output file was produced!" );
	401	}
edac47cc	402	throw( join( '', @error ) );
b02332ca	403	}
b02332ca	404
027d819c	405	=head2 parse_newick( $newick_string )
027d819c	406
ea45d2a6	407	Parses the given Newick tree(s) into one or more Stemma objects with
ea45d2a6	408	undirected graphs.
027d819c	409
	410	=cut
	411
b02332ca	412	sub parse_newick {
b02332ca	413	my $newick = shift;
5873cf38	414	# Parse the result into a set of trees and return them.
b02332ca	415	my $forest = Bio::Phylo::IO->parse(
	416	-format => 'newick',
	417	-string => $newick,
	418	);
5873cf38	419	return map { _graph_from_bio( $_ ) } @{$forest->get_entities};
ea45d2a6	420	}
	421
	422	sub _graph_from_bio {
	423	my $tree = shift;
	424	my $graph = Graph->new( 'undirected' => 1 );
	425	# Give all the intermediate anonymous nodes a name.
	426	my $i = 0;
	427	my $classes = {};
	428	foreach my $n ( @{$tree->get_terminals} ) {
	429	# The terminal nodes are our named witnesses.
	430	$classes->{$n->get_name} = 'extant';
	431	}
	432	foreach my $n ( @{$tree->get_internals} ) {
	433	unless( defined $n->get_name && $n->get_name ne '' ) {
	434	# Get an integer, make sure it's a unique name
	435	while( exists $classes->{$i} ) {
	436	$i++;
	437	}
	438	$n->set_name( $i++ );
	439	}
	440	$classes->{$n->get_name} = 'hypothetical';
	441	}
	442	_add_tree_children( $graph, $classes, undef, [ $tree->get_root ]);
	443	return $graph;
	444	}
	445
	446	sub _add_tree_children {
	447	my( $graph, $classes, $parent, $tree_children ) = @_;
	448	foreach my $c ( @$tree_children ) {
	449	my $child = $c->get_name;
	450	$graph->add_vertex( $child );
	451	$graph->set_vertex_attribute( $child, 'class', $classes->{$child} );
	452	$graph->add_path( $parent, $child ) if defined $parent;
	453	_add_tree_children( $graph, $classes, $child, $c->get_children() );
b02332ca	454	}
b02332ca	455	}
b02332ca	456
027d819c	457	=head2 newick_to_svg( $newick_string )
	458
	459	Uses the FigTree utility (if installed) to transform the given Newick tree(s)
	460	into a graph visualization.
	461
	462	=cut
	463
27e2a8fe	464	sub newick_to_svg {
	465	my $newick = shift;
	466	my $program = File::Which::which( 'figtree' );
	467	unless( -x $program ) {
edac47cc	468	throw( "FigTree commandline utility not found in path" );
27e2a8fe	469	}
	470	my $svg;
	471	my $nfile = File::Temp->new();
	472	print $nfile $newick;
	473	close $nfile;
	474	my @cmd = ( $program, '-graphic', 'SVG', $nfile );
	475	run( \@cmd, ">", binary(), \$svg );
	476	return decode_utf8( $svg );
	477	}
	478
edac47cc	479	sub throw {
	480	Text::Tradition::Error->throw(
	481	'ident' => 'StemmaUtil error',
	482	'message' => $_[0],
	483	);
	484	}
	485
027d819c	486	1;
	487
	488	=head1 LICENSE
	489
	490	This package is free software and is provided "as is" without express
	491	or implied warranty. You can redistribute it and/or modify it under
	492	the same terms as Perl itself.
	493
	494	=head1 AUTHOR
	495
	496	Tara L Andrews E<lt>aurum@cpan.orgE<gt>