X-Git-Url: http://git.shadowcat.co.uk/gitweb/gitweb.cgi?a=blobdiff_plain;f=lib%2FText%2FTradition%2FCollation.pm;h=68b2adf5997badcdbbd2e3c48b421913c43dd150;hb=679f17e1a60a81370df8cbb49b94a2b5d19e3a98;hp=ed780d77052e74c843734cb7f6dcd444f882f985;hpb=a344ffcf67e33abd15e20253f9c24baebbe0aa04;p=scpubgit%2Fstemmatology.git diff --git a/lib/Text/Tradition/Collation.pm b/lib/Text/Tradition/Collation.pm index ed780d7..68b2adf 100644 --- a/lib/Text/Tradition/Collation.pm +++ b/lib/Text/Tradition/Collation.pm @@ -2,6 +2,7 @@ package Text::Tradition::Collation; use Encode qw( decode_utf8 ); use File::Temp; +use File::Which; use Graph; use IPC::Run qw( run binary ); use Text::CSV_XS; @@ -27,6 +28,8 @@ has 'relations' => ( handles => { relationships => 'relationships', related_readings => 'related_readings', + get_relationship => 'get_relationship', + del_relationship => 'del_relationship', }, writer => '_set_relations', ); @@ -74,6 +77,12 @@ has 'ac_label' => ( default => ' (a.c.)', ); +has 'wordsep' => ( + is => 'rw', + isa => 'Str', + default => ' ', + ); + has 'start' => ( is => 'ro', isa => 'Text::Tradition::Collation::Reading', @@ -87,6 +96,26 @@ has 'end' => ( writer => '_set_end', weak_ref => 1, ); + +has 'cached_svg' => ( + is => 'rw', + isa => 'Str', + predicate => 'has_cached_svg', + clearer => 'wipe_svg', + ); + +has 'cached_table' => ( + is => 'rw', + isa => 'HashRef', + predicate => 'has_cached_table', + clearer => 'wipe_table', + ); + +has '_graphcalc_done' => ( + is => 'rw', + isa => 'Bool', + default => undef, + ); =head1 NAME @@ -140,6 +169,9 @@ representing another layer of path for the given witness - that is, when a text has more than one possible reading due to scribal corrections or the like. Defaults to ' (a.c.)'. +=item * wordsep - The string used to separate words in the original text. +Defaults to ' '. + =back =head1 ACCESSORS @@ -154,6 +186,8 @@ the like. Defaults to ' (a.c.)'. =head2 ac_label +=head2 wordsep + Simple accessors for collation attributes. =head2 start @@ -182,10 +216,14 @@ See L for the available arguments. Removes the given reading from the collation, implicitly removing its paths and relationships. -=head2 merge_readings( $main, $second ) +=head2 merge_readings( $main, $second, $concatenate, $with_str ) + +Merges the $second reading into the $main one. If $concatenate is true, then +the merged node will carry the text of both readings, concatenated with either +$with_str (if specified) or a sensible default (the empty string if the +appropriate 'join_*' flag is set on either reading, or else $self->wordsep.) -Merges the $second reading into the $main one. -The arguments may be either readings or reading IDs. +The first two arguments may be either readings or reading IDs. =head2 has_reading( $id ) @@ -250,6 +288,7 @@ sub add_reading { if( $self->reading( $reading->id ) ) { throw( "Collation already has a reading with id " . $reading->id ); } + $self->_graphcalc_done(0); $self->_add_reading( $reading->id => $reading ); # Once the reading has been added, put it in both graphs. $self->sequence->add_vertex( $reading->id ); @@ -266,6 +305,8 @@ around del_reading => sub { $arg = $arg->id; } # Remove the reading from the graphs. + $self->_graphcalc_done(0); + $self->_clear_cache; # Explicitly clear caches to GC the reading $self->sequence->delete_vertex( $arg ); $self->relations->delete_reading( $arg ); @@ -273,14 +314,58 @@ around del_reading => sub { $self->$orig( $arg ); }; -# merge_readings( $main, $to_be_deleted ); +=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 $rno = scalar $c->readings; +# Split n21 for testing purposes +my $new_r = $c->add_reading( { 'id' => 'n21p0', 'text' => 'un', 'join_next' => 1 } ); +my $old_r = $c->reading( 'n21' ); +$old_r->alter_text( 'to' ); +$c->del_path( 'n20', 'n21', 'A' ); +$c->add_path( 'n20', 'n21p0', 'A' ); +$c->add_path( 'n21p0', 'n21', 'A' ); +$c->flatten_ranks(); +ok( $c->reading( 'n21p0' ), "New reading exists" ); +is( scalar $c->readings, $rno, "Reading add offset by flatten_ranks" ); + +# Combine n3 and n4 ( with his ) +$c->merge_readings( 'n3', 'n4', 1 ); +ok( !$c->reading('n4'), "Reading n4 is gone" ); +is( $c->reading('n3')->text, 'with his', "Reading n3 has both words" ); + +# Collapse n9 and n10 ( rood / root ) +$c->merge_readings( 'n9', 'n10' ); +ok( !$c->reading('n10'), "Reading n10 is gone" ); +is( $c->reading('n9')->text, 'rood', "Reading n9 has an unchanged word" ); + +# Combine n21 and n21p0 +my $remaining = $c->reading('n21'); +$remaining ||= $c->reading('n22'); # one of these should still exist +$c->merge_readings( 'n21p0', $remaining, 1 ); +ok( !$c->reading('n21'), "Reading $remaining is gone" ); +is( $c->reading('n21p0')->text, 'unto', "Reading n21p0 merged correctly" ); + +=end testing + +=cut sub merge_readings { my $self = shift; # We only need the IDs for adding paths to the graph, not the reading # objects themselves. - my( $kept, $deleted, $combine_char ) = $self->_stringify_args( @_ ); + my( $kept, $deleted, $combine, $combine_char ) = $self->_stringify_args( @_ ); + $self->_graphcalc_done(0); # The kept reading should inherit the paths and the relationships # of the deleted reading. @@ -298,11 +383,15 @@ sub merge_readings { $self->relations->merge_readings( $kept, $deleted, $combine_char ); # Do the deletion deed. - if( $combine_char ) { + if( $combine ) { my $kept_obj = $self->reading( $kept ); - my $new_text = join( $combine_char, $kept_obj->text, - $self->reading( $deleted )->text ); - $kept_obj->alter_text( $new_text ); + my $del_obj = $self->reading( $deleted ); + my $joinstr = $combine_char; + unless( defined $joinstr ) { + $joinstr = '' if $kept_obj->join_next || $del_obj->join_prior; + $joinstr = $self->wordsep unless defined $joinstr; + } + $kept_obj->alter_text( join( $joinstr, $kept_obj->text, $del_obj->text ) ); } $self->del_reading( $deleted ); } @@ -310,12 +399,12 @@ sub merge_readings { # Helper function for manipulating the graph. sub _stringify_args { - my( $self, $first, $second, $arg ) = @_; + my( $self, $first, $second, @args ) = @_; $first = $first->id if ref( $first ) eq 'Text::Tradition::Collation::Reading'; $second = $second->id if ref( $second ) eq 'Text::Tradition::Collation::Reading'; - return( $first, $second, $arg ); + return( $first, $second, @args ); } # Helper function for manipulating the graph. @@ -336,6 +425,7 @@ sub add_path { # objects themselves. my( $source, $target, $wit ) = $self->_stringify_args( @_ ); + $self->_graphcalc_done(0); # Connect the readings $self->sequence->add_edge( $source, $target ); # Note the witness in question @@ -356,6 +446,7 @@ sub del_path { # objects themselves. my( $source, $target, $wit ) = $self->_stringify_args( @args ); + $self->_graphcalc_done(0); if( $self->sequence->has_edge_attribute( $source, $target, $wit ) ) { $self->sequence->delete_edge_attribute( $source, $target, $wit ); } @@ -384,6 +475,7 @@ be called via $tradition->del_witness. sub clear_witness { my( $self, @sigils ) = @_; + $self->_graphcalc_done(0); # Clear the witness(es) out of the paths foreach my $e ( $self->paths ) { foreach my $sig ( @sigils ) { @@ -404,11 +496,21 @@ sub add_relationship { my( $source, $target, $opts ) = $self->_stringify_args( @_ ); 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 $self->end->has_rank; + $self->_graphcalc_done(0); return @vectors; } +around qw/ get_relationship del_relationship / => sub { + my $orig = shift; + my $self = shift; + my @args = @_; + if( @args == 1 && ref( $args[0] ) eq 'ARRAY' ) { + @args = @{$_[0]}; + } + my( $source, $target ) = $self->_stringify_args( @args ); + $self->$orig( $source, $target ); +}; + =head2 reading_witnesses( $reading ) Return a list of sigils corresponding to the witnesses in which the reading appears. @@ -427,81 +529,73 @@ sub reading_witnesses { my $wits = $self->sequence->get_edge_attributes( @$e ); @all_witnesses{ keys %$wits } = 1; } + my $acstr = $self->ac_label; + foreach my $acwit ( grep { $_ =~ s/^(.*)\Q$acstr\E$/$1/ } keys %all_witnesses ) { + delete $all_witnesses{$acwit.$acstr} if exists $all_witnesses{$acwit}; + } return keys %all_witnesses; } =head1 OUTPUT METHODS -=head2 as_svg +=head2 as_svg( \%options ) Returns an SVG string that represents the graph, via as_dot and graphviz. +See as_dot for a list of options. Must have GraphViz (dot) installed to run. =cut sub as_svg { - my( $self ) = @_; - - my @cmd = qw/dot -Tsvg/; - my( $svg, $err ); - my $dotfile = File::Temp->new(); - ## USE FOR DEBUGGING - # $dotfile->unlink_on_destroy(0); - binmode $dotfile, ':utf8'; - print $dotfile $self->as_dot(); - push( @cmd, $dotfile->filename ); - run( \@cmd, ">", binary(), \$svg ); - # HACK part 3 - remove silent node+edge - my $parser = XML::LibXML->new(); - my $svgdom = $parser->parse_string( $svg ); - my $xpc = XML::LibXML::XPathContext->new( $svgdom->documentElement ); - $xpc->registerNs( 'svg', 'http://www.w3.org/2000/svg' ); - my @hacknodes = $xpc->findnodes( '//svg:g[contains(child::svg:title, "#SILENT#")]' ); - foreach my $h ( @hacknodes ) { - $h->parentNode->removeChild( $h ); - } - return decode_utf8( $svgdom->toString() ); + my( $self, $opts ) = @_; + throw( "Need GraphViz installed to output SVG" ) + unless File::Which::which( 'dot' ); + my $want_subgraph = exists $opts->{'from'} || exists $opts->{'to'}; + $self->calculate_ranks() unless $self->_graphcalc_done; + if( !$self->has_cached_svg || $opts->{'recalc'} || $want_subgraph ) { + my @cmd = qw/dot -Tsvg/; + my( $svg, $err ); + my $dotfile = File::Temp->new(); + ## USE FOR DEBUGGING + # $dotfile->unlink_on_destroy(0); + binmode $dotfile, ':utf8'; + print $dotfile $self->as_dot( $opts ); + push( @cmd, $dotfile->filename ); + run( \@cmd, ">", binary(), \$svg ); + $svg = decode_utf8( $svg ); + $self->cached_svg( $svg ) unless $want_subgraph; + return $svg; + } else { + return $self->cached_svg; + } } -=head2 svg_subgraph( $from, $to ) -Returns an SVG string that represents the portion of the graph given by the -specified range. The $from and $to variables refer to ranks within the graph. +=head2 as_dot( \%options ) -=cut +Returns a string that is the collation graph expressed in dot +(i.e. GraphViz) format. Options include: -sub svg_subgraph { - my( $self, $from, $to ) = @_; - - my $dot = $self->as_dot( $from, $to ); - unless( $dot ) { - throw( "Could not output a graph with range $from - $to" ); - } - - my @cmd = qw/dot -Tsvg/; - my( $svg, $err ); - 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 ); - $svg = decode_utf8( $svg ); - return $svg; -} +=over 4 +=item * from -=head2 as_dot( $from, $to ) +=item * to -Returns a string that is the collation graph expressed in dot -(i.e. GraphViz) format. If $from or $to is passed, as_dot creates -a subgraph rather than the entire graph. +=item * color_common + +=back =cut sub as_dot { - my( $self, $startrank, $endrank ) = @_; - + my( $self, $opts ) = @_; + my $startrank = $opts->{'from'} if $opts; + my $endrank = $opts->{'to'} if $opts; + my $color_common = $opts->{'color_common'} if $opts; + my $STRAIGHTENHACK = !$startrank && !$endrank && $self->end->rank + && $self->end->rank > 100; + $STRAIGHTENHACK = 1 if $opts->{'straight'}; # even for subgraphs or small graphs + # Check the arguments if( $startrank ) { return if $endrank && $startrank > $endrank; @@ -512,7 +606,6 @@ sub as_dot { $endrank = undef if $endrank == $self->end->rank; } - # TODO consider making some of these things configurable my $graph_name = $self->tradition->name; $graph_name =~ s/[^\w\s]//g; $graph_name = join( '_', split( /\s+/, $graph_name ) ); @@ -522,7 +615,7 @@ sub as_dot { 'bgcolor' => 'none', ); my %node_attrs = ( - 'fontsize' => 11, + 'fontsize' => 14, 'fillcolor' => 'white', 'style' => 'filled', 'shape' => 'ellipse' @@ -544,13 +637,19 @@ sub as_dot { if( $endrank ) { $dot .= "\t\"#SUBEND#\" [ label=\"...\" ];\n"; } - if( !$startrank && !$endrank ) { + if( $STRAIGHTENHACK ) { ## HACK part 1 - $dot .= "\tsubgraph { rank=same \"#START#\" \"#SILENT#\" }\n"; - $dot .= "\t\"#SILENT#\" [ color=white,penwidth=0,label=\"\" ];" + my $startlabel = $startrank ? 'SUBSTART' : 'START'; + $dot .= "\tsubgraph { rank=same \"#$startlabel#\" \"#SILENT#\" }\n"; + $dot .= "\t\"#SILENT#\" [ shape=diamond,color=white,penwidth=0,label=\"\" ];" } my %used; # Keep track of the readings that actually appear in the graph - foreach my $reading ( $self->readings ) { + # Sort the readings by rank if we have ranks; this speeds layout. + my @all_readings = $self->end->has_rank + ? sort { $a->rank <=> $b->rank } $self->readings + : $self->readings; + # TODO Refrain from outputting lacuna nodes - just grey out the edges. + foreach my $reading ( @all_readings ) { # Only output readings within our rank range. next if $startrank && $reading->rank < $startrank; next if $endrank && $reading->rank > $endrank; @@ -559,29 +658,41 @@ sub as_dot { next if $reading->id eq $reading->text; my $rattrs; my $label = $reading->text; + $label .= '-' if $reading->join_next; + $label = "-$label" if $reading->join_prior; $label =~ s/\"/\\\"/g; $rattrs->{'label'} = $label; - # TODO make this an option? - # $rattrs->{'fillcolor'} = 'green' if $reading->is_common; + $rattrs->{'fillcolor'} = '#b3f36d' if $reading->is_common && $color_common; $dot .= sprintf( "\t\"%s\" %s;\n", $reading->id, _dot_attr_string( $rattrs ) ); } - # Add the real edges + # Add the real edges. Need to weight one edge per rank jump, in a + # continuous line. + # my $weighted = $self->_add_edge_weights; 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 $label = $self->path_display_label( $self->path_witnesses( $edge ) ); + my $label = $self->_path_display_label( $self->path_witnesses( $edge ) ); my $variables = { %edge_attrs, 'label' => $label }; + # Account for the rank gap if necessary - 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; + my $rank0 = $self->reading( $edge->[0] )->rank + if $self->reading( $edge->[0] )->has_rank; + my $rank1 = $self->reading( $edge->[1] )->rank + if $self->reading( $edge->[1] )->has_rank; + if( defined $rank0 && defined $rank1 && $rank1 - $rank0 > 1 ) { + $variables->{'minlen'} = $rank1 - $rank0; } + + # Add the calculated edge weights + # if( exists $weighted->{$edge->[0]} + # && $weighted->{$edge->[0]} eq $edge->[1] ) { + # # $variables->{'color'} = 'red'; + # $variables->{'weight'} = 3.0; + # } + # 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 @@ -597,22 +708,23 @@ sub as_dot { } # 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 $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 $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#\" $varopts;"; } # HACK part 2 - if( !$startrank && !$endrank ) { - $dot .= "\t\"#END#\" -> \"#SILENT#\" [ color=white,penwidth=0 ];\n"; - } - + if( $STRAIGHTENHACK ) { + my $endlabel = $endrank ? 'SUBEND' : 'END'; + $dot .= "\t\"#$endlabel#\" -> \"#SILENT#\" [ color=white,penwidth=0 ];\n"; + } + $dot .= "}\n"; return $dot; } @@ -627,6 +739,39 @@ sub _dot_attr_string { return( '[ ' . join( ', ', @attrs ) . ' ]' ); } +sub _add_edge_weights { + my $self = shift; + # Walk the graph from START to END, choosing the successor node with + # the largest number of witness paths each time. + my $weighted = {}; + my $curr = $self->start->id; + my $ranked = $self->end->has_rank; + while( $curr ne $self->end->id ) { + my $rank = $ranked ? $self->reading( $curr )->rank : 0; + my @succ = sort { $self->path_witnesses( $curr, $a ) + <=> $self->path_witnesses( $curr, $b ) } + $self->sequence->successors( $curr ); + my $next = pop @succ; + my $nextrank = $ranked ? $self->reading( $next )->rank : 0; + # Try to avoid lacunae in the weighted path. + while( @succ && + ( $self->reading( $next )->is_lacuna || + $nextrank - $rank > 1 ) ){ + $next = pop @succ; + } + $weighted->{$curr} = $next; + $curr = $next; + } + return $weighted; +} + +=head2 path_witnesses( $edge ) + +Returns the list of sigils whose witnesses are associated with the given edge. +The edge can be passed as either an array or an arrayref of ( $source, $target ). + +=cut + sub path_witnesses { my( $self, @edge ) = @_; # If edge is an arrayref, cope. @@ -638,7 +783,7 @@ sub path_witnesses { return @wits; } -sub path_display_label { +sub _path_display_label { my $self = shift; my @wits = sort @_; my $maj = scalar( $self->tradition->witnesses ) * 0.6; @@ -649,7 +794,26 @@ sub path_display_label { return join( ', ', @wits ); } } - + +=head2 readings_at_rank( $rank ) + +Returns a list of readings at a given rank, taken from the alignment table. + +=cut + +sub readings_at_rank { + my( $self, $rank ) = @_; + my $table = $self->alignment_table; + # Table rank is real rank - 1. + my @elements = map { $_->{'tokens'}->[$rank-1] } @{$table->{'alignment'}}; + my %readings; + foreach my $e ( @elements ) { + next unless ref( $e ) eq 'HASH'; + next unless exists $e->{'t'}; + $readings{$e->{'t'}->id} = $e->{'t'}; + } + return values %readings; +} =head2 as_graphml @@ -699,7 +863,8 @@ is( scalar $st->collation->relationships, 3, "Reparsed collation has new relatio sub as_graphml { my( $self ) = @_; - + $self->calculate_ranks unless $self->_graphcalc_done; + # Some namespaces my $graphml_ns = 'http://graphml.graphdrawing.org/xmlns'; my $xsi_ns = 'http://www.w3.org/2001/XMLSchema-instance'; @@ -712,57 +877,90 @@ sub as_graphml { $graphml->setDocumentElement( $root ); $root->setNamespace( $xsi_ns, 'xsi', 0 ); $root->setAttributeNS( $xsi_ns, 'schemaLocation', $graphml_schema ); + + # List of attribute types to save on our objects and their corresponding + # GraphML types + my %save_types = ( + 'Str' => 'string', + 'Int' => 'int', + 'Bool' => 'boolean', + 'RelationshipType' => 'string', + 'RelationshipScope' => 'string', + ); + + # List of attribute names *not* to save on our objects. + # We will also not save any attribute beginning with _. + my %skipsave; + map { $skipsave{$_} = 1 } qw/ cached_svg /; - # Add the data keys for the graph + # Add the data keys for the graph. Include an extra key 'version' for the + # GraphML output version. my %graph_data_keys; my $gdi = 0; - my @graph_attributes = qw/ version wit_list_separator baselabel linear ac_label /; - foreach my $datum ( @graph_attributes ) { + my %graph_attributes = ( 'version' => 'string' ); + # Graph attributes include those of Tradition and those of Collation. + my %gattr_from; + my $tmeta = $self->tradition->meta; + my $cmeta = $self->meta; + map { $gattr_from{$_->name} = 'Tradition' } $tmeta->get_all_attributes; + map { $gattr_from{$_->name} = 'Collation' } $cmeta->get_all_attributes; + foreach my $attr ( ( $tmeta->get_all_attributes, $cmeta->get_all_attributes ) ) { + next if $attr->name =~ /^_/; + next if $skipsave{$attr->name}; + next unless $save_types{$attr->type_constraint->name}; + $graph_attributes{$attr->name} = $save_types{$attr->type_constraint->name}; + } + + foreach my $datum ( sort keys %graph_attributes ) { $graph_data_keys{$datum} = 'dg'.$gdi++; my $key = $root->addNewChild( $graphml_ns, 'key' ); $key->setAttribute( 'attr.name', $datum ); - $key->setAttribute( 'attr.type', $key eq 'linear' ? 'boolean' : 'string' ); + $key->setAttribute( 'attr.type', $graph_attributes{$datum} ); $key->setAttribute( 'for', 'graph' ); $key->setAttribute( 'id', $graph_data_keys{$datum} ); } - # Add the data keys for nodes + # Add the data keys for reading nodes + my %reading_attributes; + my $rmeta = Text::Tradition::Collation::Reading->meta; + foreach my $attr( $rmeta->get_all_attributes ) { + next if $attr->name =~ /^_/; + next if $skipsave{$attr->name}; + next unless $save_types{$attr->type_constraint->name}; + $reading_attributes{$attr->name} = $save_types{$attr->type_constraint->name}; + } my %node_data_keys; my $ndi = 0; - my %node_data = ( - id => 'string', - text => 'string', - rank => 'string', - is_start => 'boolean', - is_end => 'boolean', - is_lacuna => 'boolean', - ); - foreach my $datum ( keys %node_data ) { + foreach my $datum ( sort keys %reading_attributes ) { $node_data_keys{$datum} = 'dn'.$ndi++; my $key = $root->addNewChild( $graphml_ns, 'key' ); $key->setAttribute( 'attr.name', $datum ); - $key->setAttribute( 'attr.type', $node_data{$datum} ); + $key->setAttribute( 'attr.type', $reading_attributes{$datum} ); $key->setAttribute( 'for', 'node' ); $key->setAttribute( 'id', $node_data_keys{$datum} ); } - # Add the data keys for edges, i.e. witnesses + # Add the data keys for edges, that is, paths and relationships. Path + # data does not come from a Moose class so is here manually. my $edi = 0; my %edge_data_keys; - my %edge_data = ( - class => 'string', # Class, deprecated soon + my %edge_attributes = ( witness => 'string', # ID/label for a path - relationship => 'string', # ID/label for a relationship extra => 'boolean', # Path key - scope => 'string', # Relationship key - non_correctable => 'boolean', # Relationship key - non_independent => 'boolean', # Relationship key ); - foreach my $datum ( keys %edge_data ) { + my @path_attributes = keys %edge_attributes; # track our manual additions + my $pmeta = Text::Tradition::Collation::Relationship->meta; + foreach my $attr( $pmeta->get_all_attributes ) { + next if $attr->name =~ /^_/; + next if $skipsave{$attr->name}; + next unless $save_types{$attr->type_constraint->name}; + $edge_attributes{$attr->name} = $save_types{$attr->type_constraint->name}; + } + foreach my $datum ( sort keys %edge_attributes ) { $edge_data_keys{$datum} = 'de'.$edi++; my $key = $root->addNewChild( $graphml_ns, 'key' ); $key->setAttribute( 'attr.name', $datum ); - $key->setAttribute( 'attr.type', $edge_data{$datum} ); + $key->setAttribute( 'attr.type', $edge_attributes{$datum} ); $key->setAttribute( 'for', 'edge' ); $key->setAttribute( 'id', $edge_data_keys{$datum} ); } @@ -778,8 +976,15 @@ sub as_graphml { $sgraph->setAttribute( 'parse.order', 'nodesfirst' ); # Collation attribute data - foreach my $datum ( @graph_attributes ) { - my $value = $datum eq 'version' ? '3.0' : $self->$datum; + foreach my $datum ( keys %graph_attributes ) { + my $value; + if( $datum eq 'version' ) { + $value = '3.1'; + } elsif( $gattr_from{$datum} eq 'Tradition' ) { + $value = $self->tradition->$datum; + } else { + $value = $self->$datum; + } _add_graphml_data( $sgraph, $graph_data_keys{$datum}, $value ); } @@ -792,7 +997,7 @@ sub as_graphml { my $node_xmlid = 'n' . $node_ctr++; $node_hash{ $n->id } = $node_xmlid; $node_el->setAttribute( 'id', $node_xmlid ); - foreach my $d ( keys %node_data ) { + foreach my $d ( keys %reading_attributes ) { my $nval = $n->$d; _add_graphml_data( $node_el, $node_data_keys{$d}, $nval ) if defined $nval; @@ -824,12 +1029,12 @@ sub as_graphml { _add_graphml_data( $edge_el, $edge_data_keys{'extra'}, $aclabel ); } _add_graphml_data( $edge_el, $edge_data_keys{'witness'}, $base ); - _add_graphml_data( $edge_el, $edge_data_keys{'class'}, 'path' ); } } # Add the relationship graph to the XML - $self->relations->as_graphml( $graphml_ns, $root, \%node_hash, + map { delete $edge_data_keys{$_} } @path_attributes; + $self->relations->_as_graphml( $graphml_ns, $root, \%node_hash, $node_data_keys{'id'}, \%edge_data_keys ); # Save and return the thing @@ -854,7 +1059,7 @@ row per witness (or witness uncorrected.) sub as_csv { my( $self ) = @_; - my $table = $self->make_alignment_table; + my $table = $self->alignment_table; my $csv = Text::CSV_XS->new( { binary => 1, quote_null => 0 } ); my @result; # Make the header row @@ -863,14 +1068,14 @@ sub as_csv { # Make the rest of the rows foreach my $idx ( 0 .. $table->{'length'} - 1 ) { my @rowobjs = map { $_->{'tokens'}->[$idx] } @{$table->{'alignment'}}; - my @row = map { $_ ? $_->{'t'} : $_ } @rowobjs; + my @row = map { $_ ? $_->{'t'}->text : $_ } @rowobjs; $csv->combine( @row ); push( @result, decode_utf8( $csv->string ) ); } return join( "\n", @result ); } -=head2 make_alignment_table( $use_refs, $include_witnesses ) +=head2 alignment_table( $use_refs, $include_witnesses ) Return a reference to an alignment table, in a slightly enhanced CollateX format which looks like this: @@ -890,35 +1095,38 @@ keys have a true hash value will be included. =cut -sub make_alignment_table { - my( $self, $noderefs, $include ) = @_; - unless( $self->linear ) { - throw( "Need a linear graph in order to make an alignment table" ); - } +sub alignment_table { + my( $self ) = @_; + $self->calculate_ranks() unless $self->_graphcalc_done; + return $self->cached_table if $self->has_cached_table; + + # Make sure we can do this + throw( "Need a linear graph in order to make an alignment table" ) + unless $self->linear; + $self->calculate_ranks unless $self->end->has_rank; + my $table = { 'alignment' => [], 'length' => $self->end->rank - 1 }; my @all_pos = ( 1 .. $self->end->rank - 1 ); foreach my $wit ( sort { $a->sigil cmp $b->sigil } $self->tradition->witnesses ) { - if( $include ) { - next unless $include->{$wit->sigil}; - } # print STDERR "Making witness row(s) for " . $wit->sigil . "\n"; my @wit_path = $self->reading_sequence( $self->start, $self->end, $wit->sigil ); - my @row = _make_witness_row( \@wit_path, \@all_pos, $noderefs ); + my @row = _make_witness_row( \@wit_path, \@all_pos ); push( @{$table->{'alignment'}}, { '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 ); - my @ac_row = _make_witness_row( \@wit_ac_path, \@all_pos, $noderefs ); + my @ac_row = _make_witness_row( \@wit_ac_path, \@all_pos ); push( @{$table->{'alignment'}}, { 'witness' => $wit->sigil.$self->ac_label, 'tokens' => \@ac_row } ); } } - return $table; + $self->cached_table( $table ); + return $table; } sub _make_witness_row { - my( $path, $positions, $noderefs ) = @_; + my( $path, $positions ) = @_; my %char_hash; map { $char_hash{$_} = undef } @$positions; my $debug = 0; @@ -927,8 +1135,7 @@ sub _make_witness_row { $rtext = '#LACUNA#' if $rdg->is_lacuna; print STDERR "rank " . $rdg->rank . "\n" if $debug; # print STDERR "No rank for " . $rdg->id . "\n" unless defined $rdg->rank; - $char_hash{$rdg->rank} = $noderefs ? { 't' => $rdg } - : { 't' => $rtext }; + $char_hash{$rdg->rank} = { 't' => $rdg }; } my @row = map { $char_hash{$_} } @$positions; # Fill in lacuna markers for undef spots in the row @@ -937,8 +1144,8 @@ sub _make_witness_row { foreach my $el ( @row ) { # If we are using node reference, make the lacuna node appear many times # in the table. If not, use the lacuna tag. - if( $last_el && _el_is_lacuna( $last_el ) && !defined $el ) { - $el = $noderefs ? $last_el : { 't' => '#LACUNA#' }; + if( $last_el && $last_el->{'t'}->is_lacuna && !defined $el ) { + $el = $last_el; } push( @filled_row, $el ); $last_el = $el; @@ -946,30 +1153,6 @@ sub _make_witness_row { return @filled_row; } -# Tiny utility function to say if a table element is a lacuna -sub _el_is_lacuna { - my $el = shift; - return 1 if $el->{'t'} eq '#LACUNA#'; - return 1 if ref( $el->{'t'} ) eq 'Text::Tradition::Collation::Reading' - && $el->{'t'}->is_lacuna; - return 0; -} - -# Helper to turn the witnesses along columns rather than rows. Assumes -# equal-sized rows. -sub _turn_table { - my( $table ) = @_; - my $result = []; - return $result unless scalar @$table; - my $nrows = scalar @{$table->[0]}; - foreach my $idx ( 0 .. $nrows - 1 ) { - foreach my $wit ( 0 .. $#{$table} ) { - $result->[$idx]->[$wit] = $table->[$wit]->[$idx]; - } - } - return $result; -} - =head1 NAVIGATION METHODS =head2 reading_sequence( $first, $last, $sigil, $backup ) @@ -1138,7 +1321,17 @@ sub path_text { $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 ); + my $pathtext = ''; + my $last; + foreach my $r ( @path ) { + if( $r->join_prior || !$last || $last->join_next ) { + $pathtext .= $r->text; + } else { + $pathtext .= ' ' . $r->text; + } + $last = $r; + } + return $pathtext; } =head1 INITIALIZATION METHODS @@ -1196,10 +1389,35 @@ sub make_witness_path { Calculate the reading ranks (that is, their aligned positions relative to each other) for the graph. This can only be called on linear collations. +=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; + +# Make an svg +my $table = $c->alignment_table; +ok( $c->has_cached_table, "Alignment table was cached" ); +is( $c->alignment_table, $table, "Cached table returned upon second call" ); +$c->calculate_ranks; +is( $c->alignment_table, $table, "Cached table retained with no rank change" ); +$c->add_relationship( 'n24', 'n23', { 'type' => 'spelling' } ); +isnt( $c->alignment_table, $table, "Alignment table changed after relationship add" ); + +=end testing + =cut sub calculate_ranks { my $self = shift; + # Save the existing ranks, in case we need to invalidate the cached SVG. + my %existing_ranks; # Walk a version of the graph where every node linked by a relationship # edge is fundamentally the same node, and do a topological ranking on # the nodes in this graph. @@ -1227,6 +1445,7 @@ sub calculate_ranks { # Add the edges. foreach my $r ( $self->readings ) { + $existing_ranks{$r} = $r->rank; foreach my $n ( $self->sequence->successors( $r->id ) ) { my( $tfrom, $tto ) = ( $rel_containers{$r->id}, $rel_containers{$n} ); @@ -1256,6 +1475,20 @@ sub calculate_ranks { throw( "Ranks not calculated after $last - do you have a cycle in the graph?" ); } } + # Do we need to invalidate the cached data? + if( $self->has_cached_svg || $self->has_cached_table ) { + foreach my $r ( $self->readings ) { + next if defined( $existing_ranks{$r} ) + && $existing_ranks{$r} == $r->rank; + # Something has changed, so clear the cache + $self->_clear_cache; + # ...and recalculate the common readings. + $self->calculate_common_readings(); + last; + } + } + # The graph calculation information is now up to date. + $self->_graphcalc_done(1); } sub _assign_rank { @@ -1293,6 +1526,13 @@ sub _assign_rank { return @next_nodes; } +sub _clear_cache { + my $self = shift; + $self->wipe_svg if $self->has_cached_svg; + $self->wipe_table if $self->has_cached_table; +} + + =head2 flatten_ranks A convenience method for parsing collation data. Searches the graph for readings @@ -1303,18 +1543,25 @@ with the same text at the same rank, and merges any that are found. sub flatten_ranks { my $self = shift; my %unique_rank_rdg; + my $changed; foreach my $rdg ( $self->readings ) { next unless $rdg->has_rank; my $key = $rdg->rank . "||" . $rdg->text; if( exists $unique_rank_rdg{$key} ) { # Combine! # print STDERR "Combining readings at same rank: $key\n"; + $changed = 1; $self->merge_readings( $unique_rank_rdg{$key}, $rdg ); + # TODO see if this now makes a common point. } else { $unique_rank_rdg{$key} = $rdg; } } + # If we merged readings, the ranks are still fine but the alignment + # table is wrong. Wipe it. + $self->wipe_table() if $changed; } + =head2 calculate_common_readings @@ -1338,7 +1585,7 @@ 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 /; +my @expected = qw/ n1 n11 n16 n19 n20 n5 n6 n7 /; is_deeply( \@marked, \@expected, "Found correct list of common readings" ); =end testing @@ -1348,9 +1595,13 @@ is_deeply( \@marked, \@expected, "Found correct list of common readings" ); sub calculate_common_readings { my $self = shift; my @common; - my $table = $self->make_alignment_table( 1 ); + map { $_->is_common( 0 ) } $self->readings; + # Implicitly calls calculate_ranks + my $table = $self->alignment_table; foreach my $idx ( 0 .. $table->{'length'} - 1 ) { - my @row = map { $_->{'tokens'}->[$idx]->{'t'} } @{$table->{'alignment'}}; + my @row = map { $_->{'tokens'}->[$idx] + ? $_->{'tokens'}->[$idx]->{'t'} : '' } + @{$table->{'alignment'}}; my %hash; foreach my $r ( @row ) { if( $r ) { @@ -1379,14 +1630,22 @@ original texts. 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 ) ); + my @readings = $self->reading_sequence( $self->start, $self->end, $wit->sigil ); + my @text; + foreach my $r ( @readings ) { + next if $r->is_meta; + push( @text, $r->text ); + } $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 ); + my @ucrdgs = $self->reading_sequence( $self->start, $self->end, + $wit->sigil.$self->ac_label ); + my @uctext; + foreach my $r ( @ucrdgs ) { + next if $r->is_meta; + push( @uctext, $r->text ); + } + $wit->layertext( \@uctext ); } } } @@ -1413,14 +1672,14 @@ my $t = Text::Tradition->new( ); my $c = $t->collation; -is( $c->common_predecessor( 'n9', 'n23' )->id, +is( $c->common_predecessor( 'n24', 'n23' )->id, 'n20', "Found correct common predecessor" ); -is( $c->common_successor( 'n9', 'n23' )->id, +is( $c->common_successor( 'n24', 'n23' )->id, '#END#', "Found correct common successor" ); is( $c->common_predecessor( 'n19', 'n17' )->id, 'n16', "Found correct common predecessor for readings on same path" ); -is( $c->common_successor( 'n21', 'n26' )->id, +is( $c->common_successor( 'n21', 'n10' )->id, '#END#', "Found correct common successor for readings on same path" ); =end testing @@ -1431,16 +1690,16 @@ is( $c->common_successor( 'n21', 'n26' )->id, sub common_predecessor { my $self = shift; my( $r1, $r2 ) = $self->_objectify_args( @_ ); - return $self->common_in_path( $r1, $r2, 'predecessors' ); + return $self->_common_in_path( $r1, $r2, 'predecessors' ); } sub common_successor { my $self = shift; my( $r1, $r2 ) = $self->_objectify_args( @_ ); - return $self->common_in_path( $r1, $r2, 'successors' ); + return $self->_common_in_path( $r1, $r2, 'successors' ); } -sub common_in_path { +sub _common_in_path { my( $self, $r1, $r2, $dir ) = @_; my $iter = $r1->rank > $r2->rank ? $r1->rank : $r2->rank; $iter = $self->end->rank - $iter if $dir eq 'successors'; @@ -1475,10 +1734,12 @@ sub throw { no Moose; __PACKAGE__->meta->make_immutable; -=head1 BUGS / TODO +=head1 LICENSE -=over +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. -=item * Get rid of $backup in reading_sequence +=head1 AUTHOR -=back +Tara L Andrews Eaurum@cpan.orgE