Commit | Line | Data |
d71100ed |
1 | package Text::Tradition::Analysis; |
2 | |
3 | use strict; |
4 | use warnings; |
5 | use Text::Tradition; |
6 | use Text::Tradition::Stemma; |
7 | |
8 | sub new { |
9 | my( $class, $args ) = @_; |
10 | my $self = {}; |
d71100ed |
11 | bless( $self, $class ); |
e367f5c0 |
12 | $self->{'data'} = []; |
13 | foreach my $t ( @{$args->{'traditions'}} ) { |
14 | $self->run_analysis( $t->{'file'}, $t->{'stemmadot'} ); |
15 | } |
d71100ed |
16 | return $self; |
17 | } |
18 | |
19 | sub run_analysis { |
732152b1 |
20 | my( $self, $file, $stemmadot ) = @_; |
d71100ed |
21 | # What we will return |
22 | my $svg; |
23 | my $variants = []; |
e367f5c0 |
24 | my $data = {}; |
d71100ed |
25 | |
3d79e248 |
26 | # Read in the file and stemma |
d71100ed |
27 | my $tradition = Text::Tradition->new( |
3d79e248 |
28 | 'input' => 'Self', |
29 | 'file' => $file, |
d71100ed |
30 | 'linear' => 1, |
31 | ); |
e367f5c0 |
32 | $data->{'title'} = $tradition->name; |
732152b1 |
33 | |
d71100ed |
34 | my $stemma = Text::Tradition::Stemma->new( |
35 | 'collation' => $tradition->collation, |
36 | 'dot' => $stemmadot, |
37 | ); |
38 | # We will return the stemma picture |
e367f5c0 |
39 | $svg = $stemma->as_svg( { size => "8,7.5" } );; |
40 | $data->{'svg'} = $svg; |
d71100ed |
41 | |
42 | # We have the collation, so get the alignment table with witnesses in rows. |
43 | # Also return the reading objects in the table, rather than just the words. |
44 | |
45 | my $all_wits_table = $tradition->collation->make_alignment_table( 'refs' ); |
46 | |
47 | # For each column in the alignment table, we want to see if the existing |
48 | # groupings of witnesses match our stemma hypothesis. We also want, at the |
49 | # end, to produce an HTML table with all the variants. |
50 | my $html_columns = 0; |
732152b1 |
51 | my ( $total, $genealogical, $conflicts ) = ( 0, 0, 0 ); |
d71100ed |
52 | |
53 | # Strip the list of sigla and save it for correlation to the readings. |
54 | my $col_wits = shift @$all_wits_table; |
55 | |
56 | # We will return a data structure, an array for each row that looks like: |
57 | # { id = X, genealogical = Y, readings = [ text = X, group = Y], empty = N } |
58 | foreach my $i ( 0 .. $#$all_wits_table ) { |
59 | # For each column in the table, group the readings by witness. |
60 | my $rdg_wits = {}; |
61 | my $col_rdgs = shift @$all_wits_table; |
62 | my $rank; |
c4e11e3f |
63 | my $lacunose = []; |
d71100ed |
64 | foreach my $j ( 0 .. $#{$col_rdgs} ) { |
65 | my $rdg = $col_rdgs->[$j]; |
d71100ed |
66 | my $rdg_text = '(omitted)'; # Initialize in case of empty reading |
67 | if( $rdg ) { |
c4e11e3f |
68 | if( $rdg->is_lacuna ) { |
69 | $rdg_text = undef; # Don't count lacunae |
70 | push( @$lacunose, $col_wits->[$j] ); |
71 | } else { |
72 | $rdg_text = $rdg->text; |
73 | # Get the rank from any real reading; they should be identical. |
74 | $rank = $rdg->rank; |
75 | } |
d71100ed |
76 | } |
77 | if( defined $rdg_text ) { |
78 | # Initialize the witness array if we haven't got one yet |
79 | $rdg_wits->{$rdg_text} = [] unless $rdg_wits->{$rdg_text}; |
80 | # Add the relevant witness, subject to a.c. logic |
81 | add_variant_wit( $rdg_wits->{$rdg_text}, $col_wits->[$j], |
82 | $tradition->collation->ac_label ); |
83 | } |
84 | } |
85 | |
86 | # See if this column has any potentially genealogical variants. |
87 | # If not, skip to the next. |
88 | $total++ unless scalar keys %$rdg_wits == 1; |
89 | my( $groups, $readings ) = useful_variant( $rdg_wits ); |
90 | next unless $groups && $readings; |
91 | |
d71100ed |
92 | # Keep track of our widest row |
93 | $html_columns = scalar @$groups if scalar @$groups > $html_columns; |
94 | |
95 | # We can already look up witnesses for a reading; we also want to look |
96 | # up readings for a given witness. |
97 | my $group_readings = {}; |
98 | foreach my $x ( 0 .. $#$groups ) { |
99 | $group_readings->{wit_stringify( $groups->[$x] )} = $readings->[$x]; |
100 | } |
101 | |
102 | # For all the groups with more than one member, collect the list of all |
103 | # contiguous vertices needed to connect them. |
104 | # TODO: deal with a.c. reading logic |
94a077d6 |
105 | $DB::single = 1 if $rank == 25; |
732152b1 |
106 | my $variant_row = analyze_variant_location( $group_readings, $groups, |
107 | $stemma->apsp, $lacunose ); |
108 | $variant_row->{'id'} = $rank; |
109 | $genealogical++ if $variant_row->{'genealogical'}; |
110 | $conflicts += grep { $_->{'conflict'} } @{$variant_row->{'readings'}}; |
111 | |
d71100ed |
112 | # Now run the same analysis given the calculated distance tree(s). |
732152b1 |
113 | # my @trees = @{$stemma->distance_trees}; |
114 | # if( @trees ) { |
115 | # foreach my $tree ( 0 .. $#trees ) { |
116 | # my $dc = analyze_variant_location( $group_readings, $groups, |
117 | # $stemma->distance_apsps->[$tree] ); |
118 | # foreach my $rdg ( keys %$dc ) { |
119 | # my $var = $dc->{$rdg}; |
120 | # # TODO Do something with this |
121 | # } |
122 | # } |
123 | # } |
124 | |
d71100ed |
125 | # Record that we used this variant in an analysis |
126 | push( @$variants, $variant_row ); |
127 | } |
128 | |
732152b1 |
129 | # Go through our variant rows, after we have seen all of them once, |
130 | # and add the number of empty columns needed by each. |
d71100ed |
131 | foreach my $row ( @$variants ) { |
132 | my $empty = $html_columns - scalar @{$row->{'readings'}}; |
133 | $row->{'empty'} = $empty; |
134 | } |
135 | |
732152b1 |
136 | # Populate self with our analysis data. |
e367f5c0 |
137 | $data->{'variants'} = $variants; |
138 | $data->{'variant_count'} = $total; |
139 | $data->{'conflict_count'} = $conflicts; |
140 | $data->{'genealogical_count'} = $genealogical; |
141 | push( @{$self->{'data'}}, $data ); |
d71100ed |
142 | } |
143 | |
732152b1 |
144 | # variant_row -> genealogical |
145 | # -> readings [ { text, group, conflict, missing } ] |
146 | |
d71100ed |
147 | sub analyze_variant_location { |
732152b1 |
148 | my( $group_readings, $groups, $apsp, $lacunose ) = @_; |
d71100ed |
149 | my %contig; |
150 | my $conflict = {}; |
732152b1 |
151 | my %missing; |
152 | map { $missing{$_} = 1 } @$lacunose; |
153 | my $variant_row = { 'readings' => [] }; |
94a077d6 |
154 | # Mark each ms as in its own group, first. |
155 | foreach my $g ( @$groups ) { |
156 | my $gst = wit_stringify( $g ); |
157 | map { $contig{$_} = $gst } @$g; |
158 | } |
d71100ed |
159 | foreach my $g ( sort { scalar @$b <=> scalar @$a } @$groups ) { |
160 | my @members = @$g; |
732152b1 |
161 | my $gst = wit_stringify( $g ); # $gst is now the name of this group. |
732152b1 |
162 | while( @members ) { |
163 | # Gather the list of vertices that are needed to join all members. |
164 | my $curr = pop @members; |
165 | foreach my $m ( @members ) { |
166 | foreach my $v ( $apsp->path_vertices( $curr, $m ) ) { |
167 | $contig{$v} = $gst unless exists $contig{$v}; |
168 | next if $contig{$v} eq $gst; |
169 | # Record what is conflicting. TODO do we use this? |
170 | $conflict->{$group_readings->{$gst}} = $group_readings->{$contig{$v}}; |
171 | } |
d71100ed |
172 | } |
173 | } |
732152b1 |
174 | # Write the reading. |
175 | my $reading = { 'text' => $group_readings->{$gst}, |
176 | 'missing' => wit_stringify( $lacunose ), |
177 | 'conflict' => exists( $conflict->{$group_readings->{$gst}} ) }; |
178 | if( $reading->{'conflict'} ) { |
179 | $reading->{'group'} = $gst; |
180 | } else { |
181 | my @all_vertices = grep { $contig{$_} eq $gst && !$missing{$_} } keys %contig; |
182 | $reading->{'group'} = wit_stringify( \@all_vertices ); |
183 | } |
184 | push( @{$variant_row->{'readings'}}, $reading ); |
d71100ed |
185 | } |
732152b1 |
186 | $variant_row->{'genealogical'} = keys %$conflict ? undef : 1; |
187 | return $variant_row; |
d71100ed |
188 | } |
189 | |
190 | # Add the variant, subject to a.c. representation logic. |
191 | # This assumes that we will see the 'main' version before the a.c. version. |
192 | sub add_variant_wit { |
193 | my( $arr, $wit, $acstr ) = @_; |
194 | my $skip; |
195 | if( $wit =~ /^(.*)\Q$acstr\E$/ ) { |
196 | my $real = $1; |
197 | $skip = grep { $_ =~ /^\Q$real\E$/ } @$arr; |
198 | } |
199 | push( @$arr, $wit ) unless $skip; |
200 | } |
201 | |
202 | # Return an answer if the variant is useful, i.e. if there are at least 2 variants |
203 | # with at least 2 witnesses each. |
204 | sub useful_variant { |
205 | my( $readings ) = @_; |
206 | my $total = keys %$readings; |
207 | foreach my $var ( keys %$readings ) { |
208 | $total-- if @{$readings->{$var}} == 1; |
209 | } |
210 | return( undef, undef ) if $total <= 1; |
211 | my( $groups, $text ); |
212 | foreach my $var ( keys %$readings ) { |
213 | push( @$groups, $readings->{$var} ); |
214 | push( @$text, $var ); |
215 | } |
216 | return( $groups, $text ); |
217 | } |
218 | |
219 | # Take an array of witness groupings and produce a string like |
220 | # ['A','B'] / ['C','D','E'] / ['F'] |
221 | |
222 | sub wit_stringify { |
223 | my $groups = shift; |
224 | my @gst; |
225 | # If we were passed an array of witnesses instead of an array of |
226 | # groupings, then "group" the witnesses first. |
227 | unless( ref( $groups->[0] ) ) { |
228 | my $mkgrp = [ $groups ]; |
229 | $groups = $mkgrp; |
230 | } |
231 | foreach my $g ( @$groups ) { |
232 | push( @gst, '[' . join( ',', map { "'$_'" } @$g ) . ']' ); |
233 | } |
234 | return join( ' / ', @gst ); |
235 | } |
236 | |
237 | 1; |