1 package Text::Tradition::Stemma;
4 use Encode qw( decode_utf8 );
7 use Graph::Reader::Dot;
8 use IPC::Run qw/ run binary /;
9 use Text::Tradition::Error;
14 Text::Tradition::Stemma - a representation of a I<stemma codicum> for a Text::Tradition
19 my $t = Text::Tradition->new(
20 'name' => 'this is a text',
22 'file' => '/path/to/tei_parallel_seg_file.xml' );
24 my $s = $tradition->add_stemma( dotfile => '/path/to/stemma.dot' );
28 Text::Tradition is a library for representation and analysis of collated
29 texts, particularly medieval ones. The Stemma is a representation of the
30 copying relationships between the witnesses in a Tradition, modelled with
31 a connected rooted directed acyclic graph (CRDAG).
35 The easiest way to define a stemma is to use a special form of the 'dot'
38 Each stemma opens with the line
42 and continues with a list of all manuscript witnesses in the stemma, whether
43 extant witnesses or missing archetypes or hyparchetypes. Each of these is
44 listed by its sigil on its own line, e.g.:
46 alpha [ class=hypothetical ]
47 1 [ class=hypothetical,label=* ]
50 Extant witnesses are listed with class=extant; missing or postulated witnesses
51 are listed with class=hypothetical. Anonymous hyparchetypes must be given a
52 unique name or number, but can be represented as anonymous with the addition
53 of 'label=*' to their lines. Greek letters or other special characters may be
54 used as names, but they must always be wrapped in double quotes.
56 Links between manuscripts are then listed with arrow notation, as below. These
57 lines show the direction of copying, one step at a time, for the entire stemma.
62 The final line in the definition should be the closing brace:
66 Thus for a set of extant manuscripts A, B, and C, where A and B were copied
67 from the archetype O and C was copied from B, the definition would be:
70 O [ class=hypothetical]
83 The constructor. This should generally be called from Text::Tradition, but
84 if called directly it takes the following options:
88 =item * dot - A filehandle open to a DOT representation of the stemma graph.
90 =item * graph - If no DOT specification is given, you can pass a Graph object
91 instead. The vertices of the graph should have an attribute 'class' set to
92 either of the values 'extant' or 'hypothetical'.
94 =item * is_undirected - If the graph specification (or graph object) is for an
95 undirected graph (e.g. a phylogenetic tree), this should be set.
103 use_ok( 'Text::Tradition::Stemma' );
105 # Try to create a bad graph
107 local $TODO = "cannot use stdout redirection trick with FastCGI";
109 open( $baddotfh, 't/data/besoin_bad.dot' ) or die "Could not open test dotfile";
111 my $stemma = Text::Tradition::Stemma->new( dot => $baddotfh );
112 ok( 0, "Created broken stemma from dotfile with syntax error" );
113 } catch( Text::Tradition::Error $e ) {
114 like( $e->message, qr/^Error trying to parse/, "Syntax error in dot threw exception" );
118 # Create a good graph
120 open( $dotfh, 't/data/florilegium.dot' ) or die "Could not open test dotfile";
121 binmode( $dotfh, ':utf8' );
122 my $stemma = Text::Tradition::Stemma->new( dot => $dotfh );
123 is( ref( $stemma ), 'Text::Tradition::Stemma', "Created stemma from good dotfile" );
124 is( scalar $stemma->witnesses, 13, "Found correct number of extant witnesses" );
125 is( scalar $stemma->hypotheticals, 8, "Found correct number of extant hypotheticals" );
126 my $found_unicode_sigil;
127 foreach my $h ( $stemma->hypotheticals ) {
128 $found_unicode_sigil = 1 if $h eq "\x{3b1}";
130 ok( $found_unicode_sigil, "Found a correctly encoded Unicode sigil" );
132 # TODO Create stemma from graph, create stemma from undirected graph,
133 # create stemma from incompletely-specified graph
141 isa => 'Text::Tradition::Collation',
142 clearer => 'clear_collation', # interim measure to remove refs in DB
149 predicate => 'has_graph',
152 has is_undirected => (
156 writer => 'set_undirected',
160 my( $self, $args ) = @_;
161 # If we have been handed a dotfile, initialize it into a graph.
162 if( exists $args->{'dot'} ) {
163 $self->_graph_from_dot( $args->{'dot'} );
168 before 'graph' => sub {
171 # Make sure all unclassed graph nodes are marked extant.
173 throw( "Cannot set graph to a non-Graph object" )
174 unless ref( $g ) eq 'Graph';
175 foreach my $v ( $g->vertices ) {
176 unless( $g->has_vertex_attribute( $v, 'class' ) ) {
177 $g->set_vertex_attribute( $v, 'class', 'extant' );
180 $self->set_undirected( $g->is_undirected );
184 sub _graph_from_dot {
185 my( $self, $dotfh ) = @_;
186 my $reader = Graph::Reader::Dot->new();
187 # Redirect STDOUT in order to trap any error messages - syntax errors
188 # are evidently not fatal.
191 open $saved_stderr, ">&STDOUT";
193 open STDOUT, ">", \$reader_out;
194 my $graph = $reader->read_graph( $dotfh );
196 open STDOUT, ">", \$saved_stderr;
197 if( $reader_out && $reader_out =~ /error/s ) {
198 throw( "Error trying to parse dot: $reader_out" );
200 throw( "Failed to create graph from dot" );
202 $self->graph( $graph );
207 =head2 as_dot( \%options )
209 Returns a normal dot representation of the stemma layout, suitable for rendering
210 with GraphViz. Options include:
214 =item * graph - A hashref of global graph options.
216 =item * node - A hashref of global node options.
218 =item * edge - A hashref of global edge options.
222 See the GraphViz documentation for the list of available options.
227 my( $self, $opts ) = @_;
229 ## See if we are including any a.c. witnesses in this graph.
230 my $graph = $self->graph;
231 if( exists $opts->{'layerwits'} ) {
233 map { $extant->{$_} = 1 } $self->witnesses;
234 $graph = $self->situation_graph( $extant, $opts->{'layerwits'} );
237 # Get default and specified options
244 'fillcolor' => 'white',
246 'shape' => 'ellipse', # Shape for the extant nodes
249 'arrowhead' => 'none',
251 @graphopts{ keys %{$opts->{'graph'}} } = values %{$opts->{'graph'}}
253 @nodeopts{ keys %{$opts->{'node'}} } = values %{$opts->{'node'}}
255 @edgeopts{ keys %{$opts->{'edge'}} } = values %{$opts->{'edge'}}
258 my $gdecl = $graph->is_directed ? 'digraph' : 'graph';
260 push( @dotlines, "$gdecl stemma {" );
261 ## Print out the global attributes
262 push( @dotlines, _make_dotline( 'graph', %graphopts ) ) if keys %graphopts;
263 push( @dotlines, _make_dotline( 'edge', %edgeopts ) ) if keys %edgeopts;
264 push( @dotlines, _make_dotline( 'node', %nodeopts ) ) if keys %nodeopts;
266 # Add each of the nodes.
267 foreach my $n ( $graph->vertices ) {
268 if( $graph->has_vertex_attribute( $n, 'label' ) ) {
269 my $ltext = $graph->get_vertex_attribute( $n, 'label' );
270 push( @dotlines, _make_dotline( $n, 'label' => $ltext ) );
272 # Use the default display settings.
273 $n = _dotquote( $n );
274 push( @dotlines, " $n;" );
277 # Add each of our edges.
278 foreach my $e ( $graph->edges ) {
279 my( $from, $to ) = map { _dotquote( $_ ) } @$e;
280 my $connector = $graph->is_directed ? '->' : '--';
281 push( @dotlines, " $from $connector $to;" );
283 push( @dotlines, '}' );
285 return join( "\n", @dotlines );
288 =head2 alter_graph( $dotstring )
290 Alters the graph of this stemma according to the definition specified
296 my( $self, $dotstring ) = @_;
298 open $dotfh, '<', \$dotstring;
299 binmode $dotfh, ':utf8';
300 $self->_graph_from_dot( $dotfh );
303 =head2 editable( $opts )
305 =head2 editable_graph( $graph, $opts )
307 Returns a version of the graph rendered in our definition format. The
308 output separates statements with a newline; set $opts->{'linesep'} to the
309 empty string or to a space if the result is to be sent via JSON.
311 If a situational version of the stemma is required, the arguments for
312 situation_graph should be passed via $opts->{'extant'} and $opts->{'layerwits'}.
317 my( $self, $opts ) = @_;
318 my $graph = $self->graph;
319 ## See if we need an editable version of a situational graph.
320 if( exists $opts->{'layerwits'} || exists $opts->{'extant'} ) {
321 my $extant = delete $opts->{'extant'} || {};
322 my $layerwits = delete $opts->{'layerwits'} || [];
323 $graph = $self->situation_graph( $extant, $layerwits );
325 return editable_graph( $graph, $opts );
329 my( $graph, $opts ) = @_;
332 my $join = ( $opts && exists $opts->{'linesep'} ) ? $opts->{'linesep'} : "\n";
333 my $gdecl = $graph->is_undirected ? 'graph' : 'digraph';
335 push( @dotlines, "$gdecl stemma {" );
336 my @real; # A cheap sort
337 foreach my $n ( sort $graph->vertices ) {
338 my $c = $graph->get_vertex_attribute( $n, 'class' );
339 $c = 'extant' unless $c;
340 if( $c eq 'extant' ) {
343 push( @dotlines, _make_dotline( $n, 'class' => $c ) );
346 # Now do the real ones
347 foreach my $n ( @real ) {
348 push( @dotlines, _make_dotline( $n, 'class' => 'extant' ) );
350 foreach my $e ( sort _by_vertex $graph->edges ) {
351 my( $from, $to ) = map { _dotquote( $_ ) } @$e;
352 my $conn = $graph->is_undirected ? '--' : '->';
353 push( @dotlines, " $from $conn $to;" );
355 push( @dotlines, '}' );
356 return join( $join, @dotlines );
360 my( $obj, %attr ) = @_;
362 foreach my $k ( keys %attr ) {
363 my $v = _dotquote( $attr{$k} );
364 push( @pairs, "$k=$v" );
366 return sprintf( " %s [ %s ];", _dotquote( $obj ), join( ', ', @pairs ) );
371 return $str if $str =~ /^[A-Za-z0-9]+$/;
373 $str = '"' . $str . '"';
378 return $a->[0].$a->[1] cmp $b->[0].$b->[1];
381 =head2 situation_graph( $extant, $layered )
383 Returns a graph which is the original stemma graph with all witnesses not
384 in the %$extant hash marked as hypothetical, and witness layers added to
385 the graph according to the list in @$layered. A layered (a.c.) witness is
386 added as a parent of its main version, and additionally shares all other
387 parents and children with that version.
391 sub situation_graph {
392 my( $self, $extant, $layerwits, $layerlabel ) = @_;
394 my $graph = $self->graph->copy;
395 foreach my $vertex ( $graph->vertices ) {
396 # Set as extant any vertex that is extant in the stemma AND
397 # exists in the $extant hash.
398 my $class = 'hypothetical';
399 $class = 'extant' if exists $extant->{$vertex} && $extant->{$vertex} &&
400 $self->graph->get_vertex_attribute( $vertex, 'class' ) ne 'hypothetical';
401 $graph->set_vertex_attribute( $vertex, 'class', $class );
404 # For each 'layered' witness in the layerwits array, add it to the graph
405 # as an ancestor of the 'main' witness, and otherwise with the same parent/
406 # child links as its main analogue.
407 # TOOD Handle case where B is copied from A but corrected from C
408 $layerlabel = ' (a.c.)' unless $layerlabel;
409 foreach my $lw ( @$layerwits ) {
410 # Add the layered witness and set it with the same attributes as
411 # its 'main' analogue
412 throw( "Cannot add a layer to a hypothetical witness $lw" )
413 unless $graph->get_vertex_attribute( $lw, 'class' ) eq 'extant';
414 my $lwac = $lw . $layerlabel;
415 $graph->add_vertex( $lwac );
416 $graph->set_vertex_attributes( $lwac,
417 $graph->get_vertex_attributes( $lw ) );
419 # Set it as ancestor to the main witness
420 $graph->add_edge( $lwac, $lw );
422 # Give it the same ancestors and descendants as the main witness has,
423 # bearing in mind that those ancestors and descendants might also just
424 # have had a layered witness defined.
425 foreach my $v ( $graph->predecessors( $lw ) ) {
426 next if $v eq $lwac; # Don't add a loop
427 $graph->add_edge( $v, $lwac );
428 $graph->add_edge( $v.$layerlabel, $lwac )
429 if $graph->has_vertex( $v.$layerlabel );
431 foreach my $v ( $graph->successors( $lw ) ) {
432 next if $v eq $lwac; # but this shouldn't occur
433 $graph->add_edge( $lwac, $v );
434 $graph->add_edge( $lwac, $v.$layerlabel )
435 if $graph->has_vertex( $v.$layerlabel );
443 Returns an SVG representation of the graph, calling as_dot first.
448 my( $self, $opts ) = @_;
449 my $dot = $self->as_dot( $opts );
450 my @cmd = ( '-Tsvg' );
451 unshift( @cmd, $self->is_undirected ? 'neato' : 'dot' );
453 my $dotfile = File::Temp->new();
455 # $dotfile->unlink_on_destroy(0);
456 binmode $dotfile, ':utf8';
459 push( @cmd, $dotfile->filename );
460 run( \@cmd, ">", binary(), \$svg );
461 # HACK: Parse the SVG and change the dimensions.
462 # Get rid of width and height attributes to allow scaling.
463 if( $opts->{'size'} ) {
465 my $parser = XML::LibXML->new( load_ext_dtd => 0 );
468 $svgdoc = $parser->parse_string( decode_utf8( $svg ) );
470 throw( "Could not reparse SVG: $@" ) if $@;
471 my( $ew, $eh ) = @{$opts->{'size'}};
472 # If the graph is wider than it is tall, set width to ew and remove height.
473 # Otherwise set height to eh and remove width.
474 # TODO Also scale the viewbox
475 my $width = $svgdoc->documentElement->getAttribute('width');
476 my $height = $svgdoc->documentElement->getAttribute('height');
479 my( $remove, $keep, $val, $viewbox );
480 if( $width > $height ) {
484 my $vbheight = $width / $ew * $height;
485 $viewbox = "0.00 0.00 $width.00" . sprintf( "%.2f", $vbheight );
490 my $vbwidth = $height / $eh * $width;
491 $viewbox = "0.00 0.00 " . sprintf( "%.2f", $vbwidth ) . " $height.00";
493 $svgdoc->documentElement->removeAttribute( $remove );
494 $svgdoc->documentElement->setAttribute( $keep, $val );
495 $svgdoc->documentElement->removeAttribute( 'viewBox' );
496 $svgdoc->documentElement->setAttribute( 'viewBox', $viewbox );
497 $svg = $svgdoc->toString();
500 return decode_utf8( $svg );
505 Returns a list of the extant witnesses represented in the stemma.
511 my @wits = grep { $self->graph->get_vertex_attribute( $_, 'class' ) eq 'extant' }
512 $self->graph->vertices;
518 Returns a list of the hypothetical witnesses represented in the stemma.
525 { $self->graph->get_vertex_attribute( $_, 'class' ) eq 'hypothetical' }
526 $self->graph->vertices;
530 =head2 root( $root_vertex ) {
532 If the stemma graph is undirected, make it directed with $root_vertex at the root.
533 If it is directed, re-root it.
538 my( $self, $rootvertex ) = @_;
540 if( $self->is_undirected ) {
541 $graph = $self->graph;
543 # Make an undirected version of this graph.
544 $graph = $self->graph->undirected_copy();
546 my $rooted = Graph->new();
547 $rooted->add_vertex( $rootvertex );
548 my @next = ( $rootvertex );
551 foreach my $v ( @next ) {
552 # Place its not-placed neighbors (ergo children) in the tree
554 foreach my $n ( grep { !$rooted->has_vertex( $_ ) }
555 $graph->neighbors( $v ) ) {
556 $rooted->add_vertex( $n );
557 $rooted->add_edge( $v, $n );
558 push( @children, $n );
563 # Set the vertex classes
564 map { $rooted->set_vertex_attribute( $_, 'class', 'hypothetical' ) }
565 $self->graph->hypotheticals;
566 map { $rooted->set_vertex_class( $_, 'class', 'extant' ) }
567 $self->graph->witnesses;
573 Text::Tradition::Error->throw(
574 'ident' => 'Stemma error',
581 __PACKAGE__->meta->make_immutable;
587 This package is free software and is provided "as is" without express
588 or implied warranty. You can redistribute it and/or modify it under
589 the same terms as Perl itself.
593 Tara L Andrews E<lt>aurum@cpan.orgE<gt>