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 ); |
732152b1 |
12 | $self->run_analysis( $args->{'file'}, $args->{'stemmadot'} ); |
d71100ed |
13 | return $self; |
14 | } |
15 | |
16 | sub run_analysis { |
732152b1 |
17 | my( $self, $file, $stemmadot ) = @_; |
d71100ed |
18 | # What we will return |
19 | my $svg; |
20 | my $variants = []; |
21 | |
3d79e248 |
22 | # Read in the file and stemma |
d71100ed |
23 | my $tradition = Text::Tradition->new( |
3d79e248 |
24 | 'input' => 'Self', |
25 | 'file' => $file, |
d71100ed |
26 | 'linear' => 1, |
27 | ); |
732152b1 |
28 | $self->{'title'} = $tradition->name; |
29 | |
d71100ed |
30 | my $stemma = Text::Tradition::Stemma->new( |
31 | 'collation' => $tradition->collation, |
32 | 'dot' => $stemmadot, |
33 | ); |
34 | # We will return the stemma picture |
35 | $svg = $stemma->as_svg; |
36 | ### DIRTY HACK |
37 | $svg =~ s/transform=\"scale\(1 1\)/transform=\"scale\(0.7 0.7\)/; |
732152b1 |
38 | $self->{'svg'} = $svg; |
d71100ed |
39 | |
40 | # We have the collation, so get the alignment table with witnesses in rows. |
41 | # Also return the reading objects in the table, rather than just the words. |
42 | |
43 | my $all_wits_table = $tradition->collation->make_alignment_table( 'refs' ); |
44 | |
45 | # For each column in the alignment table, we want to see if the existing |
46 | # groupings of witnesses match our stemma hypothesis. We also want, at the |
47 | # end, to produce an HTML table with all the variants. |
48 | my $html_columns = 0; |
732152b1 |
49 | my ( $total, $genealogical, $conflicts ) = ( 0, 0, 0 ); |
d71100ed |
50 | |
51 | # Strip the list of sigla and save it for correlation to the readings. |
52 | my $col_wits = shift @$all_wits_table; |
53 | |
54 | # We will return a data structure, an array for each row that looks like: |
55 | # { id = X, genealogical = Y, readings = [ text = X, group = Y], empty = N } |
56 | foreach my $i ( 0 .. $#$all_wits_table ) { |
57 | # For each column in the table, group the readings by witness. |
58 | my $rdg_wits = {}; |
59 | my $col_rdgs = shift @$all_wits_table; |
60 | my $rank; |
c4e11e3f |
61 | my $lacunose = []; |
d71100ed |
62 | foreach my $j ( 0 .. $#{$col_rdgs} ) { |
63 | my $rdg = $col_rdgs->[$j]; |
d71100ed |
64 | my $rdg_text = '(omitted)'; # Initialize in case of empty reading |
65 | if( $rdg ) { |
c4e11e3f |
66 | if( $rdg->is_lacuna ) { |
67 | $rdg_text = undef; # Don't count lacunae |
68 | push( @$lacunose, $col_wits->[$j] ); |
69 | } else { |
70 | $rdg_text = $rdg->text; |
71 | # Get the rank from any real reading; they should be identical. |
72 | $rank = $rdg->rank; |
73 | } |
d71100ed |
74 | } |
75 | if( defined $rdg_text ) { |
76 | # Initialize the witness array if we haven't got one yet |
77 | $rdg_wits->{$rdg_text} = [] unless $rdg_wits->{$rdg_text}; |
78 | # Add the relevant witness, subject to a.c. logic |
79 | add_variant_wit( $rdg_wits->{$rdg_text}, $col_wits->[$j], |
80 | $tradition->collation->ac_label ); |
81 | } |
82 | } |
83 | |
84 | # See if this column has any potentially genealogical variants. |
85 | # If not, skip to the next. |
86 | $total++ unless scalar keys %$rdg_wits == 1; |
87 | my( $groups, $readings ) = useful_variant( $rdg_wits ); |
88 | next unless $groups && $readings; |
89 | |
d71100ed |
90 | # Keep track of our widest row |
91 | $html_columns = scalar @$groups if scalar @$groups > $html_columns; |
92 | |
93 | # We can already look up witnesses for a reading; we also want to look |
94 | # up readings for a given witness. |
95 | my $group_readings = {}; |
96 | foreach my $x ( 0 .. $#$groups ) { |
97 | $group_readings->{wit_stringify( $groups->[$x] )} = $readings->[$x]; |
98 | } |
99 | |
100 | # For all the groups with more than one member, collect the list of all |
101 | # contiguous vertices needed to connect them. |
102 | # TODO: deal with a.c. reading logic |
732152b1 |
103 | my $variant_row = analyze_variant_location( $group_readings, $groups, |
104 | $stemma->apsp, $lacunose ); |
105 | $variant_row->{'id'} = $rank; |
106 | $genealogical++ if $variant_row->{'genealogical'}; |
107 | $conflicts += grep { $_->{'conflict'} } @{$variant_row->{'readings'}}; |
108 | |
d71100ed |
109 | # Now run the same analysis given the calculated distance tree(s). |
732152b1 |
110 | # my @trees = @{$stemma->distance_trees}; |
111 | # if( @trees ) { |
112 | # foreach my $tree ( 0 .. $#trees ) { |
113 | # my $dc = analyze_variant_location( $group_readings, $groups, |
114 | # $stemma->distance_apsps->[$tree] ); |
115 | # foreach my $rdg ( keys %$dc ) { |
116 | # my $var = $dc->{$rdg}; |
117 | # # TODO Do something with this |
118 | # } |
119 | # } |
120 | # } |
121 | |
d71100ed |
122 | # Record that we used this variant in an analysis |
123 | push( @$variants, $variant_row ); |
124 | } |
125 | |
732152b1 |
126 | # Go through our variant rows, after we have seen all of them once, |
127 | # and add the number of empty columns needed by each. |
d71100ed |
128 | foreach my $row ( @$variants ) { |
129 | my $empty = $html_columns - scalar @{$row->{'readings'}}; |
130 | $row->{'empty'} = $empty; |
131 | } |
132 | |
732152b1 |
133 | # Populate self with our analysis data. |
134 | $self->{'variants'} = $variants; |
135 | $self->{'variant_count'} = $total; |
136 | $self->{'conflict_count'} = $conflicts; |
137 | $self->{'genealogical_count'} = $genealogical; |
d71100ed |
138 | } |
139 | |
732152b1 |
140 | # variant_row -> genealogical |
141 | # -> readings [ { text, group, conflict, missing } ] |
142 | |
d71100ed |
143 | sub analyze_variant_location { |
732152b1 |
144 | my( $group_readings, $groups, $apsp, $lacunose ) = @_; |
d71100ed |
145 | my %contig; |
146 | my $conflict = {}; |
732152b1 |
147 | my %missing; |
148 | map { $missing{$_} = 1 } @$lacunose; |
149 | my $variant_row = { 'readings' => [] }; |
d71100ed |
150 | foreach my $g ( sort { scalar @$b <=> scalar @$a } @$groups ) { |
151 | my @members = @$g; |
732152b1 |
152 | my $gst = wit_stringify( $g ); # $gst is now the name of this group. |
153 | map { $contig{$_} = $gst } @members; # All members are in this group. |
154 | while( @members ) { |
155 | # Gather the list of vertices that are needed to join all members. |
156 | my $curr = pop @members; |
157 | foreach my $m ( @members ) { |
158 | foreach my $v ( $apsp->path_vertices( $curr, $m ) ) { |
159 | $contig{$v} = $gst unless exists $contig{$v}; |
160 | next if $contig{$v} eq $gst; |
161 | # Record what is conflicting. TODO do we use this? |
162 | $conflict->{$group_readings->{$gst}} = $group_readings->{$contig{$v}}; |
163 | } |
d71100ed |
164 | } |
165 | } |
732152b1 |
166 | # Write the reading. |
167 | my $reading = { 'text' => $group_readings->{$gst}, |
168 | 'missing' => wit_stringify( $lacunose ), |
169 | 'conflict' => exists( $conflict->{$group_readings->{$gst}} ) }; |
170 | if( $reading->{'conflict'} ) { |
171 | $reading->{'group'} = $gst; |
172 | } else { |
173 | my @all_vertices = grep { $contig{$_} eq $gst && !$missing{$_} } keys %contig; |
174 | $reading->{'group'} = wit_stringify( \@all_vertices ); |
175 | } |
176 | push( @{$variant_row->{'readings'}}, $reading ); |
d71100ed |
177 | } |
732152b1 |
178 | $variant_row->{'genealogical'} = keys %$conflict ? undef : 1; |
179 | return $variant_row; |
d71100ed |
180 | } |
181 | |
182 | # Add the variant, subject to a.c. representation logic. |
183 | # This assumes that we will see the 'main' version before the a.c. version. |
184 | sub add_variant_wit { |
185 | my( $arr, $wit, $acstr ) = @_; |
186 | my $skip; |
187 | if( $wit =~ /^(.*)\Q$acstr\E$/ ) { |
188 | my $real = $1; |
189 | $skip = grep { $_ =~ /^\Q$real\E$/ } @$arr; |
190 | } |
191 | push( @$arr, $wit ) unless $skip; |
192 | } |
193 | |
194 | # Return an answer if the variant is useful, i.e. if there are at least 2 variants |
195 | # with at least 2 witnesses each. |
196 | sub useful_variant { |
197 | my( $readings ) = @_; |
198 | my $total = keys %$readings; |
199 | foreach my $var ( keys %$readings ) { |
200 | $total-- if @{$readings->{$var}} == 1; |
201 | } |
202 | return( undef, undef ) if $total <= 1; |
203 | my( $groups, $text ); |
204 | foreach my $var ( keys %$readings ) { |
205 | push( @$groups, $readings->{$var} ); |
206 | push( @$text, $var ); |
207 | } |
208 | return( $groups, $text ); |
209 | } |
210 | |
211 | # Take an array of witness groupings and produce a string like |
212 | # ['A','B'] / ['C','D','E'] / ['F'] |
213 | |
214 | sub wit_stringify { |
215 | my $groups = shift; |
216 | my @gst; |
217 | # If we were passed an array of witnesses instead of an array of |
218 | # groupings, then "group" the witnesses first. |
219 | unless( ref( $groups->[0] ) ) { |
220 | my $mkgrp = [ $groups ]; |
221 | $groups = $mkgrp; |
222 | } |
223 | foreach my $g ( @$groups ) { |
224 | push( @gst, '[' . join( ',', map { "'$_'" } @$g ) . ']' ); |
225 | } |
226 | return join( ' / ', @gst ); |
227 | } |
228 | |
229 | 1; |