Commit | Line | Data |
d71100ed |
1 | package Text::Tradition::Analysis; |
2 | |
3 | use strict; |
4 | use warnings; |
428bcf0b |
5 | use Algorithm::Diff; # for word similarity measure |
03c0a7d7 |
6 | use Encode qw/ decode_utf8 encode_utf8 /; |
d1348d38 |
7 | use Exporter 'import'; |
b4cb2d60 |
8 | use Graph; |
a745c3d9 |
9 | use JSON qw/ to_json decode_json /; |
03c0a7d7 |
10 | use LWP::UserAgent; |
7e17346f |
11 | use Set::Scalar; |
7e17346f |
12 | use Text::Tradition::Analysis::Result; |
13 | use Text::Tradition::Directory; |
d71100ed |
14 | use Text::Tradition::Stemma; |
d120c995 |
15 | use TryCatch; |
d71100ed |
16 | |
d1348d38 |
17 | use vars qw/ @EXPORT_OK /; |
a2cf85dd |
18 | @EXPORT_OK = qw/ run_analysis group_variants analyze_variant_location wit_stringify /; |
d1348d38 |
19 | |
03c0a7d7 |
20 | my $SOLVER_URL = 'http://byzantini.st/cgi-bin/graphcalc.cgi'; |
7e17346f |
21 | my $unsolved_problems = {}; |
5c44c598 |
22 | |
7f52eac8 |
23 | =head1 NAME |
24 | |
25 | Text::Tradition::Analysis - functions for stemma analysis of a tradition |
26 | |
27 | =head1 SYNOPSIS |
28 | |
29 | use Text::Tradition; |
30 | use Text::Tradition::Analysis qw/ run_analysis analyze_variant_location /; |
31 | my $t = Text::Tradition->new( |
32 | 'name' => 'this is a text', |
33 | 'input' => 'TEI', |
34 | 'file' => '/path/to/tei_parallel_seg_file.xml' ); |
35 | $t->add_stemma( 'dotfile' => $stemmafile ); |
36 | |
37 | my $variant_data = run_analysis( $tradition ); |
38 | # Recalculate rank $n treating all orthographic variants as equivalent |
39 | my $reanalyze = analyze_variant_location( $tradition, $n, 0, 'orthographic' ); |
40 | |
41 | =head1 DESCRIPTION |
42 | |
43 | Text::Tradition is a library for representation and analysis of collated |
44 | texts, particularly medieval ones. The Collation is the central feature of |
45 | a Tradition, where the text, its sequence of readings, and its relationships |
46 | between readings are actually kept. |
47 | |
48 | =head1 SUBROUTINES |
49 | |
88a6bac5 |
50 | =head2 run_analysis( $tradition, %opts ) |
7f52eac8 |
51 | |
88a6bac5 |
52 | Runs the analysis described in analyze_variant_location on every location in the |
53 | collation of the given tradition, with the given options. These include: |
7f52eac8 |
54 | |
88a6bac5 |
55 | =over 4 |
56 | |
57 | =item * stemma_id - Specify which of the tradition's stemmata to use. Default |
58 | is 0 (i.e. the first). |
59 | |
60 | =item * ranks - Specify a list of location ranks to analyze; exclude the rest. |
61 | |
62 | =item * merge_types - Specify a list of relationship types, where related readings |
63 | should be treated as identical for the purposes of analysis. |
64 | |
ffa22d6f |
65 | =item * exclude_type1 - Exclude those ranks whose groupings have only type-1 variants. |
66 | |
88a6bac5 |
67 | =back |
7f52eac8 |
68 | |
69 | =begin testing |
70 | |
71 | use Text::Tradition; |
72 | use Text::Tradition::Analysis qw/ run_analysis analyze_variant_location /; |
73 | |
74 | my $datafile = 't/data/florilegium_tei_ps.xml'; |
75 | my $tradition = Text::Tradition->new( 'input' => 'TEI', |
76 | 'name' => 'test0', |
77 | 'file' => $datafile ); |
78 | my $s = $tradition->add_stemma( 'dotfile' => 't/data/florilegium.dot' ); |
79 | is( ref( $s ), 'Text::Tradition::Stemma', "Added stemma to tradition" ); |
80 | |
f00cefe8 |
81 | my %expected_genealogical = ( |
a44aaf2a |
82 | 1 => 0, |
f00cefe8 |
83 | 2 => 1, |
a44aaf2a |
84 | 3 => 0, |
85 | 5 => 0, |
86 | 7 => 0, |
87 | 8 => 0, |
88 | 10 => 0, |
f00cefe8 |
89 | 13 => 1, |
a44aaf2a |
90 | 33 => 0, |
91 | 34 => 0, |
92 | 37 => 0, |
93 | 60 => 0, |
f00cefe8 |
94 | 81 => 1, |
a44aaf2a |
95 | 84 => 0, |
96 | 87 => 0, |
97 | 101 => 0, |
98 | 102 => 0, |
f00cefe8 |
99 | 122 => 1, |
a44aaf2a |
100 | 157 => 0, |
f00cefe8 |
101 | 166 => 1, |
102 | 169 => 1, |
a44aaf2a |
103 | 200 => 0, |
f00cefe8 |
104 | 216 => 1, |
105 | 217 => 1, |
106 | 219 => 1, |
107 | 241 => 1, |
108 | 242 => 1, |
109 | 243 => 1, |
110 | ); |
111 | |
7e17346f |
112 | my $data = run_analysis( $tradition, calcdsn => 'dbi:SQLite:dbname=t/data/analysis.db' ); |
7234b01d |
113 | my $c = $tradition->collation; |
f00cefe8 |
114 | foreach my $row ( @{$data->{'variants'}} ) { |
a44aaf2a |
115 | # Account for rows that used to be "not useful" |
116 | unless( exists $expected_genealogical{$row->{'id'}} ) { |
117 | $expected_genealogical{$row->{'id'}} = 1; |
118 | } |
18f48b82 |
119 | my $gen_bool = $row->{'genealogical'} ? 1 : 0; |
120 | is( $gen_bool, $expected_genealogical{$row->{'id'}}, |
f00cefe8 |
121 | "Got correct genealogical flag for row " . $row->{'id'} ); |
7234b01d |
122 | # Check that we have the right row with the right groups |
123 | my $rank = $row->{'id'}; |
124 | foreach my $rdghash ( @{$row->{'readings'}} ) { |
125 | # Skip 'readings' that aren't really |
126 | next unless $c->reading( $rdghash->{'readingid'} ); |
127 | # Check the rank |
128 | is( $c->reading( $rdghash->{'readingid'} )->rank, $rank, |
129 | "Got correct reading rank" ); |
130 | # Check the witnesses |
131 | my @realwits = sort $c->reading_witnesses( $rdghash->{'readingid'} ); |
132 | my @sgrp = sort @{$rdghash->{'group'}}; |
133 | is_deeply( \@sgrp, \@realwits, "Reading analyzed with correct groups" ); |
134 | } |
f00cefe8 |
135 | } |
a44aaf2a |
136 | is( $data->{'variant_count'}, 58, "Got right total variant number" ); |
b4cb2d60 |
137 | # TODO Make something meaningful of conflict count, maybe test other bits |
7f52eac8 |
138 | |
139 | =end testing |
140 | |
141 | =cut |
142 | |
d71100ed |
143 | sub run_analysis { |
88a6bac5 |
144 | my( $tradition, %opts ) = @_; |
f00cefe8 |
145 | my $c = $tradition->collation; |
7e17346f |
146 | my $aclabel = $c->ac_label; |
88a6bac5 |
147 | |
148 | my $stemma_id = $opts{'stemma_id'} || 0; |
1d73ecad |
149 | my @ranks = ref( $opts{'ranks'} ) eq 'ARRAY' ? @{$opts{'ranks'}} : (); |
e23225e7 |
150 | my $collapse = Set::Scalar->new(); |
151 | if( $opts{'merge_types'} && ref( $opts{'merge_types'} ) eq 'ARRAY' ) { |
152 | $collapse->insert( @{$opts{'merge_types'}} ); |
153 | } elsif( $opts{'merge_types'} ) { |
154 | $collapse->insert( $opts{'merge_types'} ); |
155 | } |
7e17346f |
156 | |
157 | # Make sure we have a lookup DB for graph calculation results; this will die |
158 | # if calcdir or calcdsn isn't passed. |
03c0a7d7 |
159 | my $dir; |
160 | if( exists $opts{'calcdir'} ) { |
161 | $dir = delete $opts{'calcdir'} |
162 | } elsif ( exists $opts{'calcdsn'} ) { |
163 | $dir = Text::Tradition::Directory->new( dsn => $opts{'calcdsn'} ); |
164 | } |
88a6bac5 |
165 | |
166 | # Get the stemma |
167 | my $stemma = $tradition->stemma( $stemma_id ); |
b4cb2d60 |
168 | |
4ce27d42 |
169 | # Figure out which witnesses we are working with - that is, the ones that |
170 | # appear both in the stemma and in the tradition. All others are 'lacunose' |
171 | # for our purposes. |
7e17346f |
172 | my $lacunose = Set::Scalar->new( $stemma->hypotheticals ); |
173 | my $stemma_wits = Set::Scalar->new( $stemma->witnesses ); |
174 | my $tradition_wits = Set::Scalar->new( map { $_->sigil } $tradition->witnesses ); |
175 | $lacunose->insert( $stemma_wits->symmetric_difference( $tradition_wits )->members ); |
88a6bac5 |
176 | |
177 | # Find and mark 'common' ranks for exclusion, unless they were |
178 | # explicitly specified. |
179 | unless( @ranks ) { |
180 | my %common_rank; |
a44aaf2a |
181 | foreach my $rdg ( $c->common_readings ) { |
88a6bac5 |
182 | $common_rank{$rdg->rank} = 1; |
183 | } |
184 | @ranks = grep { !$common_rank{$_} } ( 1 .. $c->end->rank-1 ); |
d71100ed |
185 | } |
7f52eac8 |
186 | |
88a6bac5 |
187 | # Group the variants to send to the solver |
188 | my @groups; |
f629cb3b |
189 | my @use_ranks; |
a44aaf2a |
190 | my %lacunae; |
94654e27 |
191 | my $moved = {}; |
88a6bac5 |
192 | foreach my $rank ( @ranks ) { |
7e17346f |
193 | my $missing = $lacunose->clone(); |
e23225e7 |
194 | my $rankgroup = group_variants( $tradition, $rank, $missing, $moved, $collapse ); |
94654e27 |
195 | # Filter out any empty rankgroups |
196 | # (e.g. from the later rank for a transposition) |
197 | next unless keys %$rankgroup; |
7e17346f |
198 | # Get the graph for this rankgroup |
199 | my $rankgraph = _graph_for_grouping( $stemma, $rankgroup, $missing, $aclabel ); |
ffa22d6f |
200 | if( $opts{'exclude_type1'} ) { |
201 | # Check to see whether this is a "useful" group. |
7e17346f |
202 | next unless _useful_variant( $rankgroup, $rankgraph, $aclabel ); |
ffa22d6f |
203 | } |
f629cb3b |
204 | push( @use_ranks, $rank ); |
7e17346f |
205 | push( @groups, { grouping => $rankgroup, graph => $rankgraph } ); |
a44aaf2a |
206 | $lacunae{$rank} = $missing; |
d71100ed |
207 | } |
4ce27d42 |
208 | # Run the solver |
b7bd7aa5 |
209 | my $answer; |
210 | try { |
211 | $answer = solve_variants( $dir, @groups ); |
212 | } catch ( Text::Tradition::Error $e ) { |
213 | if( $e->message =~ /IDP/ ) { |
214 | # Something is wrong with the solver; make the variants table anyway |
215 | $answer->{'variants'} = []; |
216 | map { push( @{$answer->{'variants'}}, _init_unsolved( $_, 'IDP error' ) ) } |
217 | @groups; |
218 | } else { |
219 | # Something else is wrong; error out. |
220 | $e->throw; |
221 | } |
222 | } |
fae07016 |
223 | |
88a6bac5 |
224 | # Do further analysis on the answer |
a44aaf2a |
225 | my $conflict_count = 0; |
7e17346f |
226 | my $reversion_count = 0; |
f629cb3b |
227 | foreach my $idx ( 0 .. $#use_ranks ) { |
88a6bac5 |
228 | my $location = $answer->{'variants'}->[$idx]; |
229 | # Add the rank back in |
94654e27 |
230 | my $rank = $use_ranks[$idx]; |
231 | $location->{'id'} = $rank; |
7234b01d |
232 | # Note what our lacunae are |
f629cb3b |
233 | my %lmiss; |
7234b01d |
234 | map { $lmiss{$_} = 1 } @{$lacunae{$use_ranks[$idx]}}; |
7234b01d |
235 | $location->{'missing'} = [ keys %lmiss ]; |
236 | |
88a6bac5 |
237 | # Run the extra analysis we need. |
428bcf0b |
238 | ## TODO We run through all the variants in this call, so |
239 | ## why not add the reading data there instead of here below? |
7e17346f |
240 | my $graph = $groups[$idx]->{graph}; |
241 | analyze_location( $tradition, $graph, $location, \%lmiss ); |
7234b01d |
242 | |
638e2a95 |
243 | my @layerwits; |
7234b01d |
244 | # Do the final post-analysis tidying up of the data. |
a44aaf2a |
245 | foreach my $rdghash ( @{$location->{'readings'}} ) { |
e23225e7 |
246 | $conflict_count++ if $rdghash->{'is_conflict'}; |
247 | $reversion_count++ if $rdghash->{'is_reverted'}; |
94654e27 |
248 | # Add the reading text back in, setting display value as needed |
a44aaf2a |
249 | my $rdg = $c->reading( $rdghash->{'readingid'} ); |
94654e27 |
250 | if( $rdg ) { |
251 | $rdghash->{'text'} = $rdg->text . |
252 | ( $rdg->rank == $rank ? '' : ' [' . $rdg->rank . ']' ); |
7b7abf10 |
253 | $rdghash->{'is_ungrammatical'} = $rdg->grammar_invalid; |
254 | $rdghash->{'is_nonsense'} = $rdg->is_nonsense; |
94654e27 |
255 | } |
f629cb3b |
256 | # Remove lacunose witnesses from this reading's list now that the |
7234b01d |
257 | # analysis is done |
f629cb3b |
258 | my @realgroup; |
7234b01d |
259 | map { push( @realgroup, $_ ) unless $lmiss{$_} } @{$rdghash->{'group'}}; |
f629cb3b |
260 | $rdghash->{'group'} = \@realgroup; |
638e2a95 |
261 | # Note any layered witnesses that appear in this group |
262 | foreach( @realgroup ) { |
263 | if( $_ =~ /^(.*)\Q$aclabel\E$/ ) { |
264 | push( @layerwits, $1 ); |
265 | } |
266 | } |
a44aaf2a |
267 | } |
638e2a95 |
268 | $location->{'layerwits'} = \@layerwits if @layerwits; |
88a6bac5 |
269 | } |
a44aaf2a |
270 | $answer->{'conflict_count'} = $conflict_count; |
7e17346f |
271 | $answer->{'reversion_count'} = $reversion_count; |
f00cefe8 |
272 | |
88a6bac5 |
273 | return $answer; |
d71100ed |
274 | } |
275 | |
e23225e7 |
276 | =head2 group_variants( $tradition, $rank, $lacunose, $transposed, $merge_relationship_types ) |
7f52eac8 |
277 | |
278 | Groups the variants at the given $rank of the collation, treating any |
e23225e7 |
279 | relationships in the set $merge_relationship_types as equivalent. |
280 | $lacunose should be a reference to an array, to which the sigla of lacunose |
281 | witnesses at this rank will be appended; $transposed should be a reference |
282 | to a hash, wherein the identities of transposed readings and their |
283 | relatives will be stored. |
7f52eac8 |
284 | |
ffa22d6f |
285 | Returns a hash $group_readings where $rdg is attested by the witnesses listed |
286 | in $group_readings->{$rdg}. |
7f52eac8 |
287 | |
288 | =cut |
289 | |
290 | # Return group_readings, groups, lacunose |
d1348d38 |
291 | sub group_variants { |
94654e27 |
292 | my( $tradition, $rank, $lacunose, $transposed, $collapse ) = @_; |
7f52eac8 |
293 | my $c = $tradition->collation; |
335a62ef |
294 | my $aclabel = $c->ac_label; |
d120c995 |
295 | my $table = $c->alignment_table; |
7f52eac8 |
296 | # Get the alignment table readings |
297 | my %readings_at_rank; |
7e17346f |
298 | my $check_for_gaps = Set::Scalar->new(); |
94654e27 |
299 | my %moved_wits; |
d120c995 |
300 | my $has_transposition; |
301 | foreach my $tablewit ( @{$table->{'alignment'}} ) { |
7f52eac8 |
302 | my $rdg = $tablewit->{'tokens'}->[$rank-1]; |
fae07016 |
303 | my $wit = $tablewit->{'witness'}; |
ffa22d6f |
304 | # Exclude the witness if it is "lacunose" which if we got here |
305 | # means "not in the stemma". |
7e17346f |
306 | next if _is_lacunose( $wit, $lacunose, $aclabel ); |
94654e27 |
307 | # Note if the witness is actually in a lacuna |
7f52eac8 |
308 | if( $rdg && $rdg->{'t'}->is_lacuna ) { |
335a62ef |
309 | _add_to_witlist( $wit, $lacunose, $aclabel ); |
94654e27 |
310 | # Otherwise the witness either has a positive reading... |
7f52eac8 |
311 | } elsif( $rdg ) { |
94654e27 |
312 | # If the reading has been counted elsewhere as a transposition, ignore it. |
313 | if( $transposed->{$rdg->{'t'}->id} ) { |
d120c995 |
314 | # TODO Does this cope with three-way transpositions? |
94654e27 |
315 | map { $moved_wits{$_} = 1 } @{$transposed->{$rdg->{'t'}->id}}; |
316 | next; |
317 | } |
318 | # Otherwise, record it... |
319 | $readings_at_rank{$rdg->{'t'}->id} = $rdg->{'t'}; |
320 | # ...and grab any transpositions, and their relations. |
321 | my @transp = grep { $_->rank != $rank } $rdg->{'t'}->related_readings(); |
322 | foreach my $trdg ( @transp ) { |
d120c995 |
323 | next if exists $readings_at_rank{$trdg->id}; |
324 | $has_transposition = 1; |
325 | my @affected_wits = _table_witnesses( |
7e17346f |
326 | $table, $trdg, $lacunose, $aclabel ); |
d120c995 |
327 | next unless @affected_wits; |
328 | map { $moved_wits{$_} = 1 } @affected_wits; |
329 | $transposed->{$trdg->id} = |
7e17346f |
330 | [ _table_witnesses( $table, $rdg->{'t'}, $lacunose, $aclabel ) ]; |
94654e27 |
331 | $readings_at_rank{$trdg->id} = $trdg; |
332 | } |
333 | # ...or it is empty, ergo a gap. |
7f52eac8 |
334 | } else { |
7e17346f |
335 | _add_to_witlist( $wit, $check_for_gaps, $aclabel ); |
7f52eac8 |
336 | } |
337 | } |
7e17346f |
338 | my $gap_wits = Set::Scalar->new(); |
339 | map { _add_to_witlist( $_, $gap_wits, $aclabel ) |
340 | unless $moved_wits{$_} } $check_for_gaps->members; |
341 | |
342 | # Group the readings, collapsing groups by relationship if needed. |
d120c995 |
343 | my $grouped_readings = {}; |
4ce27d42 |
344 | foreach my $rdg ( values %readings_at_rank ) { |
7f52eac8 |
345 | # Skip readings that have been collapsed into others. |
d120c995 |
346 | next if exists $grouped_readings->{$rdg->id} |
347 | && $grouped_readings->{$rdg->id} eq 'COLLAPSE'; |
4ce27d42 |
348 | # Get the witness list, including from readings collapsed into this one. |
7e17346f |
349 | my @wits = _table_witnesses( $table, $rdg, $lacunose, $aclabel ); |
e23225e7 |
350 | if( $collapse && $collapse->size ) { |
351 | my $filter = sub { $collapse->has( $_[0]->type ) }; |
7f52eac8 |
352 | foreach my $other ( $rdg->related_readings( $filter ) ) { |
7e17346f |
353 | my @otherwits = _table_witnesses( $table, $other, $lacunose, $aclabel ); |
fae07016 |
354 | push( @wits, @otherwits ); |
d120c995 |
355 | $grouped_readings->{$other->id} = 'COLLAPSE'; |
d1348d38 |
356 | } |
357 | } |
7e17346f |
358 | $grouped_readings->{$rdg->id} = Set::Scalar->new( @wits ); |
7f52eac8 |
359 | } |
7e17346f |
360 | if( $gap_wits->members ) { |
361 | $grouped_readings->{'(omitted)'} = $gap_wits; |
362 | } |
363 | |
7f52eac8 |
364 | # Get rid of our collapsed readings |
7e17346f |
365 | map { delete $grouped_readings->{$_} if( |
366 | $grouped_readings->{$_} eq 'COLLAPSE' |
367 | || $grouped_readings->{$_}->is_empty ) } |
368 | keys %$grouped_readings; |
d120c995 |
369 | |
370 | # If something was transposed, check the groups for doubled-up readings |
371 | if( $has_transposition ) { |
62a39b8f |
372 | # print STDERR "Group for rank $rank:\n"; |
373 | # map { print STDERR "\t$_: " . join( ' ' , @{$grouped_readings->{$_}} ) . "\n" } |
374 | # keys %$grouped_readings; |
d120c995 |
375 | _check_transposed_consistency( $c, $rank, $transposed, $grouped_readings ); |
376 | } |
7f52eac8 |
377 | |
4ce27d42 |
378 | # Return the result |
d120c995 |
379 | return $grouped_readings; |
380 | } |
381 | |
382 | # Helper function to query the alignment table for all witnesses (a.c. included) |
383 | # that have a given reading at its rank. |
384 | sub _table_witnesses { |
385 | my( $table, $trdg, $lacunose, $aclabel ) = @_; |
386 | my $tableidx = $trdg->rank - 1; |
7e17346f |
387 | my $has_reading = Set::Scalar->new(); |
d120c995 |
388 | foreach my $row ( @{$table->{'alignment'}} ) { |
389 | my $wit = $row->{'witness'}; |
7e17346f |
390 | next if _is_lacunose( $wit, $lacunose, $aclabel ); |
d120c995 |
391 | my $rdg = $row->{'tokens'}->[$tableidx]; |
392 | next unless exists $rdg->{'t'} && defined $rdg->{'t'}; |
7e17346f |
393 | _add_to_witlist( $wit, $has_reading, $aclabel ) |
d120c995 |
394 | if $rdg->{'t'}->id eq $trdg->id; |
395 | } |
7e17346f |
396 | return $has_reading->members; |
397 | } |
398 | |
399 | # Helper function to see if a witness is lacunose even if we are asking about |
400 | # the a.c. version |
401 | sub _is_lacunose { |
402 | my ( $wit, $lac, $acstr ) = @_; |
403 | if( $wit =~ /^(.*)\Q$acstr\E$/ ) { |
404 | $wit = $1; |
405 | } |
406 | return $lac->has( $wit ); |
d1348d38 |
407 | } |
408 | |
335a62ef |
409 | # Helper function to ensure that X and X a.c. never appear in the same list. |
410 | sub _add_to_witlist { |
411 | my( $wit, $list, $acstr ) = @_; |
335a62ef |
412 | if( $wit =~ /^(.*)\Q$acstr\E$/ ) { |
7e17346f |
413 | # Don't add X a.c. if we already have X |
414 | return if $list->has( $1 ); |
335a62ef |
415 | } else { |
7e17346f |
416 | # Delete X a.c. if we are about to add X |
417 | $list->delete( $wit.$acstr ); |
335a62ef |
418 | } |
7e17346f |
419 | $list->insert( $wit ); |
335a62ef |
420 | } |
421 | |
d120c995 |
422 | sub _check_transposed_consistency { |
423 | my( $c, $rank, $transposed, $groupings ) = @_; |
424 | my %seen_wits; |
425 | my %thisrank; |
426 | # Note which readings are actually at this rank, and which witnesses |
427 | # belong to which reading. |
428 | foreach my $rdg ( keys %$groupings ) { |
429 | my $rdgobj = $c->reading( $rdg ); |
430 | # Count '(omitted)' as a reading at this rank |
431 | $thisrank{$rdg} = 1 if !$rdgobj || $rdgobj->rank == $rank; |
432 | map { push( @{$seen_wits{$_}}, $rdg ) } @{$groupings->{$rdg}}; |
433 | } |
434 | # Our work is done if we have no witness belonging to more than one |
435 | # reading. |
436 | my @doubled = grep { scalar @{$seen_wits{$_}} > 1 } keys %seen_wits; |
437 | return unless @doubled; |
438 | # If we have a symmetric related transposition, drop the non-rank readings. |
439 | if( @doubled == scalar keys %seen_wits ) { |
440 | foreach my $rdg ( keys %$groupings ) { |
441 | if( !$thisrank{$rdg} ) { |
442 | my $groupstr = wit_stringify( $groupings->{$rdg} ); |
443 | my ( $matched ) = grep { $groupstr eq wit_stringify( $groupings->{$_} ) } |
444 | keys %thisrank; |
445 | delete $groupings->{$rdg}; |
446 | # If we found a group match, assume there is a symmetry happening. |
447 | # TODO think more about this |
62a39b8f |
448 | # print STDERR "*** Deleting symmetric reading $rdg\n"; |
d120c995 |
449 | unless( $matched ) { |
450 | delete $transposed->{$rdg}; |
451 | warn "Found problem in evident symmetry with reading $rdg"; |
452 | } |
453 | } |
454 | } |
455 | # Otherwise 'unhook' the transposed reading(s) that have duplicates. |
456 | } else { |
457 | foreach my $dup ( @doubled ) { |
458 | foreach my $rdg ( @{$seen_wits{$dup}} ) { |
459 | next if $thisrank{$rdg}; |
460 | next unless exists $groupings->{$rdg}; |
62a39b8f |
461 | # print STDERR "*** Deleting asymmetric doubled-up reading $rdg\n"; |
d120c995 |
462 | delete $groupings->{$rdg}; |
463 | delete $transposed->{$rdg}; |
464 | } |
465 | } |
466 | # and put any now-orphaned readings into an 'omitted' reading. |
467 | foreach my $wit ( keys %seen_wits ) { |
468 | unless( grep { exists $groupings->{$_} } @{$seen_wits{$wit}} ) { |
7e17346f |
469 | $groupings->{'(omitted)'} = Set::Scalar->new() |
470 | unless exists $groupings->{'(omitted)'}; |
d120c995 |
471 | _add_to_witlist( $wit, $groupings->{'(omitted)'}, $c->ac_label ); |
472 | } |
473 | } |
474 | } |
475 | } |
476 | |
7e17346f |
477 | # For the given grouping, return its situation graph based on the stemma. |
478 | sub _graph_for_grouping { |
479 | my( $stemma, $grouping, $lacunose, $aclabel ) = @_; |
480 | my $acwits = []; |
481 | my $extant = {}; |
482 | foreach my $gs ( values %$grouping ) { |
483 | map { |
484 | if( $_ =~ /^(.*)\Q$aclabel\E$/ ) { |
485 | push( @$acwits, $1 ) unless $lacunose->has( $1 ); |
486 | } else { |
487 | $extant->{$_} = 1 unless $lacunose->has( $_ ); |
488 | } |
489 | } $gs->members; |
490 | } |
491 | my $graph; |
492 | try { |
493 | # contig contains all extant wits and all hypothetical wits |
494 | # needed to make up the groups. |
495 | $graph = $stemma->situation_graph( $extant, $acwits ); |
496 | } catch ( Text::Tradition::Error $e ) { |
03c0a7d7 |
497 | throw( "Could not extend graph with given extant and a.c. witnesses: " |
498 | . $e->message ); |
7e17346f |
499 | } catch { |
03c0a7d7 |
500 | throw( "Could not extend graph with a.c. witnesses @$acwits" ); |
7e17346f |
501 | } |
502 | return $graph; |
503 | } |
88a6bac5 |
504 | |
7e17346f |
505 | =head2 solve_variants( $calcdir, @groups ) |
506 | |
507 | Looks up the set of groups in the answers provided by the external graph solver |
508 | service and returns a cleaned-up answer, adding the rank IDs back where they belong. |
88a6bac5 |
509 | |
88a6bac5 |
510 | The answer has the form |
511 | { "variants" => [ array of variant location structures ], |
512 | "variant_count" => total, |
513 | "conflict_count" => number of conflicts detected, |
514 | "genealogical_count" => number of solutions found } |
515 | |
516 | =cut |
517 | |
518 | sub solve_variants { |
03c0a7d7 |
519 | my( @groups ) = @_; |
520 | |
a745c3d9 |
521 | # Are we using a local result directory, or did we pass an empty value |
522 | # for the directory? |
03c0a7d7 |
523 | my $dir; |
a745c3d9 |
524 | unless( ref( $groups[0] ) eq 'HASH' ) { |
03c0a7d7 |
525 | $dir = shift @groups; |
526 | } |
335a62ef |
527 | |
03c0a7d7 |
528 | ## For each graph/group combo, make a Text::Tradition::Analysis::Result |
529 | ## object so that we can send it off for IDP lookup. |
7e17346f |
530 | my $variants = []; |
531 | my $genealogical = 0; # counter |
03c0a7d7 |
532 | # TODO Optimize for unique graph problems |
554e2e7d |
533 | my %problems; |
7e17346f |
534 | foreach my $graphproblem ( @groups ) { |
03c0a7d7 |
535 | # Construct the calc result key and look up its answer |
536 | my $problem = Text::Tradition::Analysis::Result->new( |
537 | graph => $graphproblem->{'graph'}, |
538 | setlist => [ values %{$graphproblem->{'grouping'}} ] ); |
554e2e7d |
539 | if( exists $problems{$problem->object_key} ) { |
540 | $problem = $problems{$problem->object_key}; |
541 | } else { |
542 | $problems{$problem->object_key} = $problem; |
543 | } |
544 | $graphproblem->{'object'} = $problem; |
03c0a7d7 |
545 | } |
546 | |
554e2e7d |
547 | my %results; |
03c0a7d7 |
548 | if( $dir ) { |
549 | my $scope = $dir->new_scope; |
554e2e7d |
550 | map { $results{$_} = $dir->lookup( $_ ) || $problems{$_} } keys %problems; |
551 | } else { |
552 | my $json = JSON->new->allow_blessed->convert_blessed->utf8->encode( |
553 | [ values %problems ] ); |
03c0a7d7 |
554 | # Send it off and get the result |
555 | # print STDERR "Sending request: " . decode_utf8( $json ) . "\n"; |
556 | my $ua = LWP::UserAgent->new(); |
557 | my $resp = $ua->post( $SOLVER_URL, 'Content-Type' => 'application/json', |
558 | 'Content' => $json ); |
a745c3d9 |
559 | my $answer; |
03c0a7d7 |
560 | if( $resp->is_success ) { |
a745c3d9 |
561 | $answer = decode_json( $resp->content ); |
03c0a7d7 |
562 | throw( "Unexpected answer from IDP: $answer" ) unless ref( $answer ) eq 'ARRAY'; |
563 | } else { |
564 | throw( "IDP solver returned " . $resp->status_line . " / " . $resp->content |
565 | . "; cannot run graph analysis" ); |
566 | } |
567 | # One more sanity check |
568 | throw( "Something went wrong with answer symmetricity" ) |
554e2e7d |
569 | unless keys( %problems ) == @$answer; |
03c0a7d7 |
570 | # Convert the results |
554e2e7d |
571 | foreach my $a ( @$answer ) { |
572 | my $r = Text::Tradition::Analysis::Result->new( $a ); |
573 | $results{$r->object_key} = $r; |
574 | } |
03c0a7d7 |
575 | } |
576 | |
577 | # We now have a single JSON-encoded Result object per problem sent. Fold its |
578 | # answers into our variant info structure. |
554e2e7d |
579 | foreach my $graphproblem ( @groups ) { |
580 | my $result = $results{$graphproblem->{'object'}->object_key} |
581 | || $graphproblem->{'object'}; |
03c0a7d7 |
582 | |
7e17346f |
583 | # Initialize the result structure for this graph problem |
b7bd7aa5 |
584 | my $vstruct; |
585 | if( $result->status eq 'OK' ) { |
586 | $vstruct = { readings => [] }; |
587 | push( @$variants, $vstruct ); |
588 | } else { |
589 | push( @$variants, _init_unsolved( $graphproblem, $result->status ) ); |
7e17346f |
590 | next; |
5c44c598 |
591 | } |
b7bd7aa5 |
592 | |
7e17346f |
593 | # 1. Did the group evaluate as genealogical? |
03c0a7d7 |
594 | $vstruct->{genealogical} = $result->is_genealogical; |
595 | $genealogical++ if $result->is_genealogical; |
7e17346f |
596 | |
597 | # 2. What are the calculated minimum groupings for each variant loc? |
598 | foreach my $rid ( keys %{$graphproblem->{grouping}} ) { |
599 | my $inputset = $graphproblem->{grouping}->{$rid}; |
03c0a7d7 |
600 | my $minset = $result->minimum_grouping_for( $inputset ); |
7e17346f |
601 | push( @{$vstruct->{readings}}, { readingid => $rid, group => $minset } ); |
6d25a3a0 |
602 | } |
7e17346f |
603 | |
604 | # 3. What are the sources and classes calculated for each witness? |
03c0a7d7 |
605 | $vstruct->{witcopy_types} = { $result->classes }; |
7e17346f |
606 | $vstruct->{reading_roots} = {}; |
03c0a7d7 |
607 | map { $vstruct->{reading_roots}->{$_} = 1 } $result->sources; |
7e17346f |
608 | |
88a6bac5 |
609 | } |
610 | |
611 | return { 'variants' => $variants, |
612 | 'variant_count' => scalar @$variants, |
613 | 'genealogical_count' => $genealogical }; |
614 | } |
615 | |
b7bd7aa5 |
616 | sub _init_unsolved { |
617 | my( $graphproblem, $status ) = @_; |
618 | my $vstruct = { 'readings' => [] }; |
619 | $vstruct->{'unsolved'} = $status; |
620 | foreach my $rid ( keys %{$graphproblem->{grouping}} ) { |
621 | push( @{$vstruct->{readings}}, { readingid => $rid, |
622 | group => [ $graphproblem->{grouping}->{$rid}->members ] } ); |
623 | } |
624 | return $vstruct; |
625 | } |
626 | |
fae07016 |
627 | =head2 analyze_location ( $tradition, $graph, $location_hash ) |
7f52eac8 |
628 | |
fae07016 |
629 | Given the tradition, its stemma graph, and the solution from the graph solver, |
630 | work out the rest of the information we want. For each reading we need missing, |
5c44c598 |
631 | conflict, reading_parents, independent_occurrence, followed, not_followed, |
632 | and follow_unknown. Alters the location_hash in place. |
7f52eac8 |
633 | |
634 | =cut |
732152b1 |
635 | |
fae07016 |
636 | sub analyze_location { |
7e17346f |
637 | my ( $tradition, $graph, $variant_row, $lacunose ) = @_; |
94654e27 |
638 | my $c = $tradition->collation; |
fae07016 |
639 | |
03c0a7d7 |
640 | if( exists $variant_row->{'unsolved'} ) { |
641 | return; |
642 | } |
643 | my $reading_roots = delete $variant_row->{'reading_roots'}; |
644 | my $classinfo = delete $variant_row->{'witcopy_types'}; |
645 | |
fae07016 |
646 | # Make a hash of all known node memberships, and make the subgraphs. |
647 | my $contig = {}; |
fae07016 |
648 | my $subgraph = {}; |
638e2a95 |
649 | my $acstr = $c->ac_label; |
650 | my @acwits; |
5c44c598 |
651 | |
5c44c598 |
652 | # Note which witnesses positively belong to which group. This information |
653 | # comes ultimately from the IDP solver. |
654 | # Also make a note of the reading's roots. |
fae07016 |
655 | foreach my $rdghash ( @{$variant_row->{'readings'}} ) { |
656 | my $rid = $rdghash->{'readingid'}; |
5c44c598 |
657 | my @roots; |
638e2a95 |
658 | foreach my $wit ( @{$rdghash->{'group'}} ) { |
659 | $contig->{$wit} = $rid; |
660 | if( $wit =~ /^(.*)\Q$acstr\E$/ ) { |
661 | push( @acwits, $1 ); |
662 | } |
5c44c598 |
663 | if( exists $reading_roots->{$wit} && $reading_roots->{$wit} ) { |
664 | push( @roots, $wit ); |
665 | } |
638e2a95 |
666 | } |
5c44c598 |
667 | $rdghash->{'independent_occurrence'} = \@roots; |
94654e27 |
668 | } |
7e17346f |
669 | |
fae07016 |
670 | # Now that we have all the node group memberships, calculate followed/ |
bebec0e9 |
671 | # non-followed/unknown values for each reading. Also figure out the |
672 | # reading's evident parent(s). |
673 | foreach my $rdghash ( @{$variant_row->{'readings'}} ) { |
fae07016 |
674 | my $rid = $rdghash->{'readingid'}; |
428bcf0b |
675 | my $rdg = $c->reading( $rid ); |
5c44c598 |
676 | my @roots = @{$rdghash->{'independent_occurrence'}}; |
7e17346f |
677 | my @reversions; |
678 | if( $classinfo ) { |
679 | @reversions = grep { $classinfo->{$_} eq 'revert' } |
680 | $rdghash->{'group'}->members; |
be590045 |
681 | $rdghash->{'reversions'} = \@reversions; |
7e17346f |
682 | } |
5c44c598 |
683 | my @group = @{$rdghash->{'group'}}; |
fae07016 |
684 | |
685 | # Start figuring things out. |
7e17346f |
686 | $rdghash->{'followed'} = scalar( @group ) |
687 | - ( scalar( @roots ) + scalar( @reversions ) ); |
bebec0e9 |
688 | # Find the parent readings, if any, of this reading. |
7e17346f |
689 | my $sourceparents = _find_reading_parents( $rid, $graph, $contig, @roots ); |
7e17346f |
690 | # Work out relationships between readings and their non-followed parent. |
691 | _resolve_parent_relationships( $c, $rid, $rdg, $sourceparents ); |
be590045 |
692 | $rdghash->{'source_parents'} = $sourceparents; |
e23225e7 |
693 | |
694 | if( @reversions ) { |
695 | my $revparents = _find_reading_parents( $rid, $graph, $contig, @reversions ); |
696 | _resolve_parent_relationships( $c, $rid, $rdg, $revparents ); |
697 | $rdghash->{'reversion_parents'} = $revparents; |
698 | } |
bebec0e9 |
699 | |
700 | # Find the number of times this reading was altered, and the number of |
701 | # times we're not sure. |
702 | my( %nofollow, %unknownfollow ); |
5c44c598 |
703 | foreach my $wit ( @{$rdghash->{'group'}} ) { |
bebec0e9 |
704 | foreach my $wchild ( $graph->successors( $wit ) ) { |
5c44c598 |
705 | if( $reading_roots->{$wchild} && $contig->{$wchild} |
706 | && $contig->{$wchild} ne $rid ) { |
bebec0e9 |
707 | # It definitely changed here. |
708 | $nofollow{$wchild} = 1; |
709 | } elsif( !($contig->{$wchild}) ) { |
710 | # The child is a hypothetical node not definitely in |
711 | # any group. Answer is unknown. |
712 | $unknownfollow{$wchild} = 1; |
5c44c598 |
713 | } # else it is either in our group, or it is a non-root node in a |
714 | # known group and therefore is presumed to have its reading from |
715 | # its group, not this link. |
bebec0e9 |
716 | } |
717 | } |
718 | $rdghash->{'not_followed'} = keys %nofollow; |
719 | $rdghash->{'follow_unknown'} = keys %unknownfollow; |
fae07016 |
720 | |
721 | # Now say whether this reading represents a conflict. |
722 | unless( $variant_row->{'genealogical'} ) { |
7e17346f |
723 | $rdghash->{'is_conflict'} = @roots != 1; |
e23225e7 |
724 | $rdghash->{'is_reverted'} = scalar @reversions; |
7e17346f |
725 | } |
726 | } |
727 | } |
728 | |
729 | sub _find_reading_parents { |
730 | my( $rid, $graph, $contig, @list ) = @_; |
731 | my $parenthash = {}; |
732 | foreach my $wit ( @list ) { |
733 | # Look in the stemma graph to find this witness's extant or known-reading |
734 | # immediate ancestor(s), and look up the reading that each ancestor holds. |
735 | my @check = $graph->predecessors( $wit ); |
736 | while( @check ) { |
737 | my @next; |
738 | foreach my $wparent( @check ) { |
739 | my $preading = $contig->{$wparent}; |
740 | if( $preading && $preading ne $rid ) { |
741 | $parenthash->{$preading} = 1; |
742 | } else { |
743 | push( @next, $graph->predecessors( $wparent ) ); |
744 | } |
745 | } |
746 | @check = @next; |
747 | } |
748 | } |
749 | return $parenthash; |
750 | } |
751 | |
752 | sub _resolve_parent_relationships { |
753 | my( $c, $rid, $rdg, $rdgparents ) = @_; |
754 | foreach my $p ( keys %$rdgparents ) { |
755 | # Resolve the relationship of the parent to the reading, and |
756 | # save it in our hash. |
757 | my $pobj = $c->reading( $p ); |
758 | my $prep = $pobj ? $pobj->id . ' (' . $pobj->text . ')' : $p; |
759 | my $phash = { 'label' => $prep }; |
760 | if( $pobj ) { |
761 | my $rel = $c->get_relationship( $p, $rid ); |
762 | if( $rel ) { |
763 | _add_to_hash( $rel, $phash ); |
764 | } elsif( $rdg ) { |
765 | # First check for a transposed relationship |
766 | if( $rdg->rank != $pobj->rank ) { |
767 | foreach my $ti ( $rdg->related_readings( 'transposition' ) ) { |
768 | next unless $ti->text eq $rdg->text; |
769 | $rel = $c->get_relationship( $ti, $pobj ); |
770 | if( $rel ) { |
771 | _add_to_hash( $rel, $phash, 1 ); |
772 | last; |
773 | } |
774 | } |
775 | unless( $rel ) { |
776 | foreach my $ti ( $pobj->related_readings( 'transposition' ) ) { |
777 | next unless $ti->text eq $pobj->text; |
778 | $rel = $c->get_relationship( $ti, $rdg ); |
779 | if( $rel ) { |
780 | _add_to_hash( $rel, $phash, 1 ); |
781 | last; |
782 | } |
783 | } |
5c44c598 |
784 | } |
785 | } |
7e17346f |
786 | unless( $rel ) { |
787 | # and then check for sheer word similarity. |
788 | my $rtext = $rdg->text; |
789 | my $ptext = $pobj->text; |
790 | if( similar( $rtext, $ptext ) ) { |
791 | # say STDERR "Words $rtext and $ptext judged similar"; |
792 | $phash->{relation} = { type => 'wordsimilar' }; |
793 | } |
794 | } |
5c44c598 |
795 | } else { |
7e17346f |
796 | $phash->{relation} = { type => 'deletion' }; |
5c44c598 |
797 | } |
7e17346f |
798 | # Get the attributes of the parent object while we are here |
799 | $phash->{'text'} = $pobj->text if $pobj; |
800 | $phash->{'is_nonsense'} = $pobj->is_nonsense; |
801 | $phash->{'is_ungrammatical'} = $pobj->grammar_invalid; |
802 | } elsif( $p eq '(omitted)' ) { |
803 | $phash->{relation} = { type => 'addition' }; |
804 | } |
805 | # Save it |
806 | $rdgparents->{$p} = $phash; |
807 | } |
d71100ed |
808 | } |
809 | |
428bcf0b |
810 | sub _add_to_hash { |
811 | my( $rel, $phash, $is_transposed ) = @_; |
812 | $phash->{relation} = { type => $rel->type }; |
813 | $phash->{relation}->{transposed} = 1 if $is_transposed; |
814 | $phash->{relation}->{annotation} = $rel->annotation |
815 | if $rel->has_annotation; |
816 | } |
817 | |
818 | =head2 similar( $word1, $word2 ) |
819 | |
820 | Use Algorithm::Diff to get a sense of how close the words are to each other. |
821 | This will hopefully handle substitutions a bit more nicely than Levenshtein. |
822 | |
823 | =cut |
824 | |
825 | #!/usr/bin/env perl |
826 | |
827 | sub similar { |
828 | my( $word1, $word2 ) = sort { length($a) <=> length($b) } @_; |
829 | my @let1 = split( '', lc( $word1 ) ); |
830 | my @let2 = split( '', lc( $word2 ) ); |
831 | my $diff = Algorithm::Diff->new( \@let1, \@let2 ); |
832 | my $mag = 0; |
833 | while( $diff->Next ) { |
834 | if( $diff->Same ) { |
835 | # Take off points for longer strings |
836 | my $cs = $diff->Range(1) - 2; |
837 | $cs = 0 if $cs < 0; |
838 | $mag -= $cs; |
839 | } elsif( !$diff->Items(1) ) { |
840 | $mag += $diff->Range(2); |
841 | } elsif( !$diff->Items(2) ) { |
842 | $mag += $diff->Range(1); |
843 | } else { |
844 | # Split the difference for substitutions |
845 | my $c1 = $diff->Range(1) || 1; |
846 | my $c2 = $diff->Range(2) || 1; |
847 | my $cd = ( $c1 + $c2 ) / 2; |
848 | $mag += $cd; |
849 | } |
850 | } |
851 | return ( $mag <= length( $word1 ) / 2 ); |
852 | } |
853 | |
5be0cdeb |
854 | sub _useful_variant { |
7e17346f |
855 | my( $rankgroup, $rankgraph, $acstr ) = @_; |
5be0cdeb |
856 | |
857 | # Sort by group size and return |
858 | my $is_useful = 0; |
7e17346f |
859 | foreach my $rdg ( keys %$rankgroup ) { |
860 | my @wits = $rankgroup->{$rdg}->members; |
861 | if( @wits > 1 ) { |
5be0cdeb |
862 | $is_useful++; |
863 | } else { |
7e17346f |
864 | $is_useful++ unless( $rankgraph->is_sink_vertex( $wits[0] ) |
865 | || $wits[0] =~ /\Q$acstr\E$/ ); |
5be0cdeb |
866 | } |
867 | } |
7e17346f |
868 | return $is_useful > 1; |
5be0cdeb |
869 | } |
870 | |
7f52eac8 |
871 | =head2 wit_stringify( $groups ) |
872 | |
873 | Takes an array of witness groupings and produces a string like |
874 | ['A','B'] / ['C','D','E'] / ['F'] |
d71100ed |
875 | |
7f52eac8 |
876 | =cut |
d71100ed |
877 | |
878 | sub wit_stringify { |
879 | my $groups = shift; |
880 | my @gst; |
881 | # If we were passed an array of witnesses instead of an array of |
882 | # groupings, then "group" the witnesses first. |
883 | unless( ref( $groups->[0] ) ) { |
884 | my $mkgrp = [ $groups ]; |
885 | $groups = $mkgrp; |
886 | } |
887 | foreach my $g ( @$groups ) { |
888 | push( @gst, '[' . join( ',', map { "'$_'" } @$g ) . ']' ); |
889 | } |
890 | return join( ' / ', @gst ); |
891 | } |
7f52eac8 |
892 | |
7f52eac8 |
893 | 1; |
894 | |
a745c3d9 |
895 | sub throw { |
896 | Text::Tradition::Error->throw( |
897 | 'ident' => 'Analysis error', |
898 | 'message' => $_[0], |
899 | ); |
900 | } |
901 | |
7f52eac8 |
902 | =head1 LICENSE |
903 | |
904 | This package is free software and is provided "as is" without express |
905 | or implied warranty. You can redistribute it and/or modify it under |
906 | the same terms as Perl itself. |
907 | |
908 | =head1 AUTHOR |
909 | |
910 | Tara L Andrews E<lt>aurum@cpan.orgE<gt> |