Commit | Line | Data |
d71100ed |
1 | package Text::Tradition::Analysis; |
2 | |
3 | use strict; |
4 | use warnings; |
e4386ba9 |
5 | use Benchmark; |
88a6bac5 |
6 | use Encode qw/ encode_utf8 /; |
d1348d38 |
7 | use Exporter 'import'; |
b4cb2d60 |
8 | use Graph; |
88a6bac5 |
9 | use JSON qw/ encode_json decode_json /; |
10 | use LWP::UserAgent; |
d71100ed |
11 | use Text::Tradition; |
12 | use Text::Tradition::Stemma; |
13 | |
d1348d38 |
14 | use vars qw/ @EXPORT_OK /; |
a2cf85dd |
15 | @EXPORT_OK = qw/ run_analysis group_variants analyze_variant_location wit_stringify /; |
d1348d38 |
16 | |
7f52eac8 |
17 | =head1 NAME |
18 | |
19 | Text::Tradition::Analysis - functions for stemma analysis of a tradition |
20 | |
21 | =head1 SYNOPSIS |
22 | |
23 | use Text::Tradition; |
24 | use Text::Tradition::Analysis qw/ run_analysis analyze_variant_location /; |
25 | my $t = Text::Tradition->new( |
26 | 'name' => 'this is a text', |
27 | 'input' => 'TEI', |
28 | 'file' => '/path/to/tei_parallel_seg_file.xml' ); |
29 | $t->add_stemma( 'dotfile' => $stemmafile ); |
30 | |
31 | my $variant_data = run_analysis( $tradition ); |
32 | # Recalculate rank $n treating all orthographic variants as equivalent |
33 | my $reanalyze = analyze_variant_location( $tradition, $n, 0, 'orthographic' ); |
34 | |
35 | =head1 DESCRIPTION |
36 | |
37 | Text::Tradition is a library for representation and analysis of collated |
38 | texts, particularly medieval ones. The Collation is the central feature of |
39 | a Tradition, where the text, its sequence of readings, and its relationships |
40 | between readings are actually kept. |
41 | |
42 | =head1 SUBROUTINES |
43 | |
88a6bac5 |
44 | =head2 run_analysis( $tradition, %opts ) |
7f52eac8 |
45 | |
88a6bac5 |
46 | Runs the analysis described in analyze_variant_location on every location in the |
47 | collation of the given tradition, with the given options. These include: |
7f52eac8 |
48 | |
88a6bac5 |
49 | =over 4 |
50 | |
51 | =item * stemma_id - Specify which of the tradition's stemmata to use. Default |
52 | is 0 (i.e. the first). |
53 | |
54 | =item * ranks - Specify a list of location ranks to analyze; exclude the rest. |
55 | |
56 | =item * merge_types - Specify a list of relationship types, where related readings |
57 | should be treated as identical for the purposes of analysis. |
58 | |
59 | =back |
7f52eac8 |
60 | |
61 | =begin testing |
62 | |
63 | use Text::Tradition; |
64 | use Text::Tradition::Analysis qw/ run_analysis analyze_variant_location /; |
65 | |
66 | my $datafile = 't/data/florilegium_tei_ps.xml'; |
67 | my $tradition = Text::Tradition->new( 'input' => 'TEI', |
68 | 'name' => 'test0', |
69 | 'file' => $datafile ); |
70 | my $s = $tradition->add_stemma( 'dotfile' => 't/data/florilegium.dot' ); |
71 | is( ref( $s ), 'Text::Tradition::Stemma', "Added stemma to tradition" ); |
72 | |
f00cefe8 |
73 | my %expected_genealogical = ( |
a44aaf2a |
74 | 1 => 0, |
f00cefe8 |
75 | 2 => 1, |
a44aaf2a |
76 | 3 => 0, |
77 | 5 => 0, |
78 | 7 => 0, |
79 | 8 => 0, |
80 | 10 => 0, |
f00cefe8 |
81 | 13 => 1, |
a44aaf2a |
82 | 33 => 0, |
83 | 34 => 0, |
84 | 37 => 0, |
85 | 60 => 0, |
f00cefe8 |
86 | 81 => 1, |
a44aaf2a |
87 | 84 => 0, |
88 | 87 => 0, |
89 | 101 => 0, |
90 | 102 => 0, |
f00cefe8 |
91 | 122 => 1, |
a44aaf2a |
92 | 157 => 0, |
f00cefe8 |
93 | 166 => 1, |
94 | 169 => 1, |
a44aaf2a |
95 | 200 => 0, |
f00cefe8 |
96 | 216 => 1, |
97 | 217 => 1, |
98 | 219 => 1, |
99 | 241 => 1, |
100 | 242 => 1, |
101 | 243 => 1, |
102 | ); |
103 | |
7f52eac8 |
104 | my $data = run_analysis( $tradition ); |
f00cefe8 |
105 | foreach my $row ( @{$data->{'variants'}} ) { |
a44aaf2a |
106 | # Account for rows that used to be "not useful" |
107 | unless( exists $expected_genealogical{$row->{'id'}} ) { |
108 | $expected_genealogical{$row->{'id'}} = 1; |
109 | } |
adc08836 |
110 | my $gen_bool = $row->{'genealogical'} ? 1 : 0; |
111 | is( $gen_bool, $expected_genealogical{$row->{'id'}}, |
f00cefe8 |
112 | "Got correct genealogical flag for row " . $row->{'id'} ); |
113 | } |
a44aaf2a |
114 | is( $data->{'variant_count'}, 58, "Got right total variant number" ); |
b4cb2d60 |
115 | # TODO Make something meaningful of conflict count, maybe test other bits |
7f52eac8 |
116 | |
117 | =end testing |
118 | |
119 | =cut |
120 | |
d71100ed |
121 | sub run_analysis { |
88a6bac5 |
122 | my( $tradition, %opts ) = @_; |
f00cefe8 |
123 | my $c = $tradition->collation; |
88a6bac5 |
124 | |
125 | my $stemma_id = $opts{'stemma_id'} || 0; |
1d73ecad |
126 | my @ranks = ref( $opts{'ranks'} ) eq 'ARRAY' ? @{$opts{'ranks'}} : (); |
127 | my @collapse = ref( $opts{'merge_types'} ) eq 'ARRAY' ? @{$opts{'merge_types'}} : (); |
88a6bac5 |
128 | |
129 | # Get the stemma |
130 | my $stemma = $tradition->stemma( $stemma_id ); |
b4cb2d60 |
131 | |
88a6bac5 |
132 | # Figure out which witnesses we are working with |
133 | my @lacunose = $stemma->hypotheticals; |
fae07016 |
134 | my @tradition_wits = map { $_->sigil } $tradition->witnesses; |
b4cb2d60 |
135 | map { push( @tradition_wits, $_->sigil.$c->ac_label ) if $_->is_layered } |
fae07016 |
136 | $tradition->witnesses; |
137 | push( @lacunose, _symmdiff( [ $stemma->witnesses ], \@tradition_wits ) ); |
88a6bac5 |
138 | |
139 | # Find and mark 'common' ranks for exclusion, unless they were |
140 | # explicitly specified. |
141 | unless( @ranks ) { |
142 | my %common_rank; |
a44aaf2a |
143 | foreach my $rdg ( $c->common_readings ) { |
88a6bac5 |
144 | $common_rank{$rdg->rank} = 1; |
145 | } |
146 | @ranks = grep { !$common_rank{$_} } ( 1 .. $c->end->rank-1 ); |
d71100ed |
147 | } |
7f52eac8 |
148 | |
88a6bac5 |
149 | # Group the variants to send to the solver |
150 | my @groups; |
a44aaf2a |
151 | my %lacunae; |
88a6bac5 |
152 | foreach my $rank ( @ranks ) { |
a44aaf2a |
153 | my $missing = [ @lacunose ]; |
154 | push( @groups, group_variants( $tradition, $rank, $missing, \@collapse ) ); |
155 | $lacunae{$rank} = $missing; |
d71100ed |
156 | } |
b4cb2d60 |
157 | $DB::single = 1; |
88a6bac5 |
158 | # Parse the answer |
e59b8faa |
159 | my $answer = solve_variants( $stemma, @groups ); |
fae07016 |
160 | |
88a6bac5 |
161 | # Do further analysis on the answer |
a44aaf2a |
162 | my $conflict_count = 0; |
88a6bac5 |
163 | foreach my $idx ( 0 .. $#ranks ) { |
164 | my $location = $answer->{'variants'}->[$idx]; |
165 | # Add the rank back in |
166 | $location->{'id'} = $ranks[$idx]; |
a44aaf2a |
167 | # Add the lacunae back in |
168 | $location->{'missing'} = $lacunae{$ranks[$idx]}; |
88a6bac5 |
169 | # Run the extra analysis we need. |
88a6bac5 |
170 | analyze_location( $tradition, $stemma->graph, $location ); |
a44aaf2a |
171 | # Add the reading text back in |
172 | foreach my $rdghash ( @{$location->{'readings'}} ) { |
173 | $conflict_count++ |
174 | if exists $rdghash->{'conflict'} && $rdghash->{'conflict'}; |
175 | my $rdg = $c->reading( $rdghash->{'readingid'} ); |
176 | $rdghash->{'text'} = $rdg ? $rdg->text : $rdghash->{'readingid'}; |
177 | } |
88a6bac5 |
178 | } |
a44aaf2a |
179 | $answer->{'conflict_count'} = $conflict_count; |
f00cefe8 |
180 | |
88a6bac5 |
181 | return $answer; |
d71100ed |
182 | } |
183 | |
7f52eac8 |
184 | =head2 group_variants( $tradition, $rank, $lacunose, @merge_relationship_types ) |
185 | |
186 | Groups the variants at the given $rank of the collation, treating any |
187 | relationships in @merge_relationship_types as equivalent. $lacunose should |
188 | be a reference to an array, to which the sigla of lacunose witnesses at this |
189 | rank will be appended. |
190 | |
191 | Returns two ordered lists $readings, $groups, where $readings->[$n] is attested |
192 | by the witnesses listed in $groups->[$n]. |
193 | |
194 | =cut |
195 | |
196 | # Return group_readings, groups, lacunose |
d1348d38 |
197 | sub group_variants { |
7f52eac8 |
198 | my( $tradition, $rank, $lacunose, $collapse ) = @_; |
199 | my $c = $tradition->collation; |
b4cb2d60 |
200 | my $aclabel = $c->ac_label; |
7f52eac8 |
201 | # Get the alignment table readings |
202 | my %readings_at_rank; |
203 | my @gap_wits; |
1d73ecad |
204 | foreach my $tablewit ( @{$c->alignment_table->{'alignment'}} ) { |
7f52eac8 |
205 | my $rdg = $tablewit->{'tokens'}->[$rank-1]; |
fae07016 |
206 | my $wit = $tablewit->{'witness'}; |
7f52eac8 |
207 | if( $rdg && $rdg->{'t'}->is_lacuna ) { |
b4cb2d60 |
208 | _add_to_witlist( $wit, $lacunose, $aclabel ); |
7f52eac8 |
209 | } elsif( $rdg ) { |
210 | $readings_at_rank{$rdg->{'t'}->text} = $rdg->{'t'}; |
211 | } else { |
b4cb2d60 |
212 | _add_to_witlist( $wit, \@gap_wits, $aclabel ); |
7f52eac8 |
213 | } |
214 | } |
d1348d38 |
215 | |
7f52eac8 |
216 | # Group the readings, collapsing groups by relationship if needed |
217 | my %grouped_readings; |
b4cb2d60 |
218 | foreach my $rdg ( sort { $b->witnesses <=> $a->witnesses } |
219 | values %readings_at_rank ) { |
7f52eac8 |
220 | # Skip readings that have been collapsed into others. |
f00cefe8 |
221 | next if exists $grouped_readings{$rdg->id} && !$grouped_readings{$rdg->id}; |
7f52eac8 |
222 | my @wits = $rdg->witnesses; |
223 | if( $collapse ) { |
224 | my $filter = sub { my $r = $_[0]; grep { $_ eq $r->type } @$collapse; }; |
225 | foreach my $other ( $rdg->related_readings( $filter ) ) { |
fae07016 |
226 | my @otherwits = $other->witnesses; |
fae07016 |
227 | push( @wits, @otherwits ); |
f00cefe8 |
228 | $grouped_readings{$other->id} = 0; |
d1348d38 |
229 | } |
230 | } |
f00cefe8 |
231 | $grouped_readings{$rdg->id} = \@wits; |
7f52eac8 |
232 | } |
233 | $grouped_readings{'(omitted)'} = \@gap_wits if @gap_wits; |
234 | # Get rid of our collapsed readings |
235 | map { delete $grouped_readings{$_} unless $grouped_readings{$_} } |
236 | keys %grouped_readings |
237 | if $collapse; |
238 | |
5be0cdeb |
239 | return \%grouped_readings; |
d1348d38 |
240 | } |
241 | |
88a6bac5 |
242 | =head2 solve_variants( $graph, @groups ) |
243 | |
244 | Sends the set of groups to the external graph solver service and returns |
245 | a cleaned-up answer, adding the rank IDs back where they belong. |
246 | |
247 | The JSON has the form |
248 | { "graph": [ stemmagraph DOT string without newlines ], |
249 | "groupings": [ array of arrays of groups, one per rank ] } |
250 | |
251 | The answer has the form |
252 | { "variants" => [ array of variant location structures ], |
253 | "variant_count" => total, |
254 | "conflict_count" => number of conflicts detected, |
255 | "genealogical_count" => number of solutions found } |
256 | |
257 | =cut |
258 | |
259 | sub solve_variants { |
e59b8faa |
260 | my( $stemma, @groups ) = @_; |
88a6bac5 |
261 | |
262 | # Make the json with stemma + groups |
b4cb2d60 |
263 | my $groupings = []; |
88a6bac5 |
264 | foreach my $ghash ( @groups ) { |
265 | my @grouping; |
266 | foreach my $k ( sort keys %$ghash ) { |
267 | push( @grouping, $ghash->{$k} ); |
268 | } |
b4cb2d60 |
269 | push( @$groupings, \@grouping ); |
88a6bac5 |
270 | } |
b4cb2d60 |
271 | ## Witness map is a HACK to get around limitations in node names from IDP |
272 | my $witness_map = {}; |
273 | my $json = encode_json( _safe_wit_strings( $stemma, $groupings, $witness_map ) ); |
88a6bac5 |
274 | |
275 | # Send it off and get the result |
276 | my $solver_url = 'http://byzantini.st/cgi-bin/graphcalc.cgi'; |
277 | my $ua = LWP::UserAgent->new(); |
278 | my $resp = $ua->post( $solver_url, 'Content-Type' => 'application/json', |
279 | 'Content' => $json ); |
280 | |
281 | my $answer; |
282 | if( $resp->is_success ) { |
b4cb2d60 |
283 | $answer = _desanitize_names( decode_json( $resp->content ), $witness_map ); |
88a6bac5 |
284 | } else { |
fae07016 |
285 | # Fall back to the old method. |
286 | warn "IDP solver returned " . $resp->status_line . " / " . $resp->content |
287 | . "; falling back to perl method"; |
b4cb2d60 |
288 | $answer = perl_solver( $stemma, @$groupings ); |
88a6bac5 |
289 | } |
290 | |
291 | # Fold the result back into what we know about the groups. |
292 | my $variants = []; |
293 | my $genealogical = 0; |
294 | foreach my $idx ( 0 .. $#groups ) { |
295 | my( $calc_groups, $result ) = @{$answer->[$idx]}; |
296 | $genealogical++ if $result; |
297 | my $input_group = $groups[$idx]; |
298 | foreach my $k ( sort keys %$input_group ) { |
299 | my $cg = shift @$calc_groups; |
300 | $input_group->{$k} = $cg; |
301 | } |
302 | my $vstruct = { |
303 | 'genealogical' => $result, |
304 | 'readings' => [], |
fae07016 |
305 | }; |
88a6bac5 |
306 | foreach my $k ( keys %$input_group ) { |
307 | push( @{$vstruct->{'readings'}}, |
fae07016 |
308 | { 'readingid' => $k, 'group' => $input_group->{$k}} ); |
88a6bac5 |
309 | } |
310 | push( @$variants, $vstruct ); |
311 | } |
312 | |
313 | return { 'variants' => $variants, |
314 | 'variant_count' => scalar @$variants, |
315 | 'genealogical_count' => $genealogical }; |
316 | } |
317 | |
b4cb2d60 |
318 | #### HACKERY to cope with IDP's limited idea of what a node name looks like ### |
319 | |
320 | sub _safe_wit_strings { |
321 | my( $stemma, $groupings, $witness_map ) = @_; |
322 | my $safegraph = Graph->new(); |
323 | # Convert the graph to a safe representation and store the conversion. |
324 | foreach my $n ( $stemma->graph->vertices ) { |
325 | my $sn = _safe_witstr( $n ); |
326 | warn "Ambiguous stringification $sn for $n and " . $witness_map->{$sn} |
327 | if exists $witness_map->{$sn}; |
328 | $witness_map->{$sn} = $n; |
329 | $safegraph->add_vertex( $sn ); |
330 | $safegraph->set_vertex_attributes( $sn, |
331 | $stemma->graph->get_vertex_attributes( $n ) ); |
332 | } |
333 | foreach my $e ( $stemma->graph->edges ) { |
334 | my @safe_e = ( _safe_witstr( $e->[0] ), _safe_witstr( $e->[1] ) ); |
335 | $safegraph->add_edge( @safe_e ); |
336 | } |
337 | my $safe_stemma = Text::Tradition::Stemma->new( |
338 | 'collation' => $stemma->collation, 'graph' => $safegraph ); |
339 | |
340 | # Now convert the witness groupings to a safe representation. |
341 | my $safe_groupings = []; |
342 | foreach my $grouping ( @$groupings ) { |
343 | my $safe_grouping = []; |
344 | foreach my $group ( @$grouping ) { |
345 | my $safe_group = []; |
346 | foreach my $n ( @$group ) { |
347 | my $sn = _safe_witstr( $n ); |
348 | warn "Ambiguous stringification $sn for $n and " . $witness_map->{$sn} |
349 | if exists $witness_map->{$sn} && $witness_map->{$sn} ne $n; |
350 | $witness_map->{$sn} = $n; |
351 | push( @$safe_group, $sn ); |
352 | } |
353 | push( @$safe_grouping, $safe_group ); |
354 | } |
355 | push( @$safe_groupings, $safe_grouping ); |
356 | } |
357 | |
358 | # Return it all in the struct we expect. We have stored the reductions |
359 | # in the $witness_map that we were passed. |
360 | return { 'graph' => $safe_stemma->editable( ' ' ), 'groupings' => $safe_groupings }; |
361 | } |
362 | |
363 | sub _safe_witstr { |
364 | my $witstr = shift; |
365 | $witstr =~ s/\s+/_/g; |
366 | $witstr =~ s/[^\w\d-]//g; |
367 | return $witstr; |
368 | } |
369 | |
370 | sub _desanitize_names { |
371 | my( $jsonstruct, $witness_map ) = @_; |
372 | my $result = []; |
373 | foreach my $grouping ( @$jsonstruct ) { |
374 | my $real_grouping = []; |
375 | foreach my $element ( @$grouping ) { |
376 | if( ref( $element ) eq 'ARRAY' ) { |
377 | # it's the groupset. |
378 | my $real_groupset = []; |
379 | foreach my $group ( @$element ) { |
380 | my $real_group = []; |
381 | foreach my $n ( @$group ) { |
382 | my $rn = $witness_map->{$n}; |
383 | push( @$real_group, $rn ); |
384 | } |
385 | push( @$real_groupset, $real_group ); |
386 | } |
387 | push( @$real_grouping, $real_groupset ); |
388 | } else { |
389 | # It is the boolean, not actually a group. |
390 | push( @$real_grouping, $element ); |
391 | } |
392 | } |
393 | push( @$result, $real_grouping ); |
394 | } |
395 | return $result; |
396 | } |
397 | |
398 | ### END HACKERY ### |
399 | |
fae07016 |
400 | =head2 analyze_location ( $tradition, $graph, $location_hash ) |
7f52eac8 |
401 | |
fae07016 |
402 | Given the tradition, its stemma graph, and the solution from the graph solver, |
403 | work out the rest of the information we want. For each reading we need missing, |
404 | conflict, reading_parents, independent_occurrence, followed, not_followed, and follow_unknown. Alters the location_hash in place. |
7f52eac8 |
405 | |
406 | =cut |
732152b1 |
407 | |
fae07016 |
408 | sub analyze_location { |
409 | my ( $tradition, $graph, $variant_row ) = @_; |
410 | |
411 | # Make a hash of all known node memberships, and make the subgraphs. |
412 | my $contig = {}; |
413 | my $reading_roots = {}; |
414 | my $subgraph = {}; |
415 | foreach my $rdghash ( @{$variant_row->{'readings'}} ) { |
416 | my $rid = $rdghash->{'readingid'}; |
417 | map { $contig->{$_} = $rid } @{$rdghash->{'group'}}; |
c4a4fb1b |
418 | |
fae07016 |
419 | # Make the subgraph. |
420 | my $part = $graph->copy; |
421 | my %these_vertices; |
422 | map { $these_vertices{$_} = 1 } @{$rdghash->{'group'}}; |
423 | $part->delete_vertices( grep { !$these_vertices{$_} } $part->vertices ); |
424 | $subgraph->{$rid} = $part; |
425 | # Get the reading roots. |
426 | map { $reading_roots->{$_} = $rid } $part->predecessorless_vertices; |
bebec0e9 |
427 | } |
428 | |
fae07016 |
429 | # Now that we have all the node group memberships, calculate followed/ |
bebec0e9 |
430 | # non-followed/unknown values for each reading. Also figure out the |
431 | # reading's evident parent(s). |
432 | foreach my $rdghash ( @{$variant_row->{'readings'}} ) { |
fae07016 |
433 | # Group string key - TODO do we need this? |
434 | my $gst = wit_stringify( $rdghash->{'group'} ); |
435 | my $rid = $rdghash->{'readingid'}; |
436 | # Get the subgraph |
437 | my $part = $subgraph->{$rid}; |
438 | |
439 | # Start figuring things out. |
bebec0e9 |
440 | my @roots = $part->predecessorless_vertices; |
441 | $rdghash->{'independent_occurrence'} = scalar @roots; |
442 | $rdghash->{'followed'} = scalar( $part->vertices ) - scalar( @roots ); |
443 | # Find the parent readings, if any, of this reading. |
f00cefe8 |
444 | my %rdgparents; |
bebec0e9 |
445 | foreach my $wit ( @roots ) { |
f00cefe8 |
446 | # Look in the main stemma to find this witness's extant or known-reading |
447 | # immediate ancestor(s), and look up the reading that each ancestor olds. |
448 | my @check = $graph->predecessors( $wit ); |
449 | while( @check ) { |
450 | my @next; |
451 | foreach my $wparent( @check ) { |
fae07016 |
452 | my $preading = $contig->{$wparent}; |
453 | if( $preading ) { |
454 | $rdgparents{$preading} = 1; |
f00cefe8 |
455 | } else { |
456 | push( @next, $graph->predecessors( $wparent ) ); |
457 | } |
458 | } |
459 | @check = @next; |
460 | } |
bebec0e9 |
461 | } |
f00cefe8 |
462 | $rdghash->{'reading_parents'} = [ keys %rdgparents ]; |
bebec0e9 |
463 | |
464 | # Find the number of times this reading was altered, and the number of |
465 | # times we're not sure. |
466 | my( %nofollow, %unknownfollow ); |
467 | foreach my $wit ( $part->vertices ) { |
468 | foreach my $wchild ( $graph->successors( $wit ) ) { |
469 | next if $part->has_vertex( $wchild ); |
fae07016 |
470 | if( $reading_roots->{$wchild} && $contig->{$wchild} ) { |
bebec0e9 |
471 | # It definitely changed here. |
472 | $nofollow{$wchild} = 1; |
473 | } elsif( !($contig->{$wchild}) ) { |
474 | # The child is a hypothetical node not definitely in |
475 | # any group. Answer is unknown. |
476 | $unknownfollow{$wchild} = 1; |
477 | } # else it's a non-root node in a known group, and therefore |
478 | # is presumed to have its reading from its group, not this link. |
479 | } |
480 | } |
481 | $rdghash->{'not_followed'} = keys %nofollow; |
482 | $rdghash->{'follow_unknown'} = keys %unknownfollow; |
fae07016 |
483 | |
484 | # Now say whether this reading represents a conflict. |
485 | unless( $variant_row->{'genealogical'} ) { |
486 | $rdghash->{'conflict'} = @roots != 1; |
487 | } |
c4a4fb1b |
488 | } |
d71100ed |
489 | } |
490 | |
fae07016 |
491 | |
492 | =head2 perl_solver( $tradition, $rank, $stemma_id, @merge_relationship_types ) |
493 | |
494 | ** NOTE ** This method should hopefully not be called - it is not guaranteed |
495 | to be correct. Serves as a backup for the real solver. |
496 | |
497 | Runs an analysis of the given tradition, at the location given in $rank, |
498 | against the graph of the stemma specified in $stemma_id. The argument |
499 | @merge_relationship_types is an optional list of relationship types for |
500 | which readings so related should be treated as equivalent. |
501 | |
502 | Returns a nested array data structure as follows: |
503 | |
504 | [ [ group_list, is_genealogical ], [ group_list, is_genealogical ] ... ] |
505 | |
506 | where the group list is the array of arrays passed in for each element of @groups, |
507 | possibly with the addition of hypothetical readings. |
508 | |
509 | |
510 | =cut |
511 | |
512 | sub perl_solver { |
e59b8faa |
513 | my( $stemma, @groups ) = @_; |
514 | my $graph = $stemma->graph; |
515 | my @answer; |
516 | foreach my $g ( @groups ) { |
517 | push( @answer, _solve_variant_location( $graph, $g ) ); |
518 | } |
519 | return \@answer; |
fae07016 |
520 | } |
521 | |
e59b8faa |
522 | sub _solve_variant_location { |
523 | my( $graph, $groups ) = @_; |
fae07016 |
524 | # Now do the work. |
e59b8faa |
525 | my $contig = {}; |
526 | my $subgraph = {}; |
527 | my $is_conflicted; |
528 | my $conflict = {}; |
529 | |
530 | # Mark each ms as in its own group, first. |
531 | foreach my $g ( @$groups ) { |
532 | my $gst = wit_stringify( $g ); |
533 | map { $contig->{$_} = $gst } @$g; |
534 | } |
535 | |
536 | # Now for each unmarked node in the graph, initialize an array |
537 | # for possible group memberships. We will use this later to |
538 | # resolve potential conflicts. |
539 | map { $contig->{$_} = [] unless $contig->{$_} } $graph->vertices; |
540 | foreach my $g ( sort { scalar @$b <=> scalar @$a } @$groups ) { |
541 | my $gst = wit_stringify( $g ); # This is the group name |
542 | # Copy the graph, and delete all non-members from the new graph. |
543 | my $part = $graph->copy; |
544 | my @group_roots; |
545 | $part->delete_vertices( |
546 | grep { !ref( $contig->{$_} ) && $contig->{$_} ne $gst } $graph->vertices ); |
547 | |
548 | # Now look to see if our group is connected. |
549 | if( @$g > 1 ) { |
550 | # We have to take directionality into account. |
551 | # How many root nodes do we have? |
552 | my @roots = grep { ref( $contig->{$_} ) || $contig->{$_} eq $gst } |
553 | $part->predecessorless_vertices; |
554 | # Assuming that @$g > 1, find the first root node that has at |
555 | # least one successor belonging to our group. If this reading |
556 | # is genealogical, there should be only one, but we will check |
557 | # that implicitly later. |
558 | foreach my $root ( @roots ) { |
559 | # Prune the tree to get rid of extraneous hypotheticals. |
560 | $root = _prune_subtree( $part, $root, $contig ); |
561 | next unless $root; |
562 | # Save this root for our group. |
563 | push( @group_roots, $root ); |
564 | # Get all the successor nodes of our root. |
565 | } |
566 | } else { |
567 | # Dispense with the trivial case of one reading. |
568 | my $wit = $g->[0]; |
569 | @group_roots = ( $wit ); |
570 | foreach my $v ( $part->vertices ) { |
571 | $part->delete_vertex( $v ) unless $v eq $wit; |
572 | } |
573 | } |
574 | |
575 | if( @group_roots > 1 ) { |
576 | $conflict->{$gst} = 1; |
577 | $is_conflicted = 1; |
578 | } |
579 | # Paint the 'hypotheticals' with our group. |
580 | foreach my $wit ( $part->vertices ) { |
581 | if( ref( $contig->{$wit} ) ) { |
582 | push( @{$contig->{$wit}}, $gst ); |
583 | } elsif( $contig->{$wit} ne $gst ) { |
584 | warn "How did we get here?"; |
585 | } |
586 | } |
587 | |
588 | |
589 | # Save the relevant subgraph. |
590 | $subgraph->{$gst} = $part; |
591 | } |
592 | |
593 | # For each of our hypothetical readings, flatten its 'contig' array if |
594 | # the array contains zero or one group. If we have any unflattened arrays, |
595 | # we may need to run the resolution process. If the reading is already known |
596 | # to have a conflict, flatten the 'contig' array to nothing; we won't resolve |
597 | # it. |
598 | my @resolve; |
599 | foreach my $wit ( keys %$contig ) { |
600 | next unless ref( $contig->{$wit} ); |
601 | if( @{$contig->{$wit}} > 1 ) { |
602 | if( $is_conflicted ) { |
603 | $contig->{$wit} = ''; # We aren't going to decide. |
604 | } else { |
605 | push( @resolve, $wit ); |
606 | } |
607 | } else { |
608 | my $gst = pop @{$contig->{$wit}}; |
609 | $contig->{$wit} = $gst || ''; |
610 | } |
611 | } |
612 | |
613 | if( @resolve ) { |
614 | my $still_contig = {}; |
615 | foreach my $h ( @resolve ) { |
616 | # For each of the hypothetical readings with more than one possibility, |
617 | # try deleting it from each of its member subgraphs in turn, and see |
618 | # if that breaks the contiguous grouping. |
619 | # TODO This can still break in a corner case where group A can use |
620 | # either vertex 1 or 2, and group B can use either vertex 2 or 1. |
621 | # Revisit this if necessary; it could get brute-force nasty. |
622 | foreach my $gst ( @{$contig->{$h}} ) { |
623 | my $gpart = $subgraph->{$gst}->copy(); |
624 | # If we have come this far, there is only one root and everything |
625 | # is reachable from it. |
626 | my( $root ) = $gpart->predecessorless_vertices; |
627 | my $reachable = {}; |
628 | map { $reachable->{$_} = 1 } $gpart->vertices; |
629 | |
630 | # Try deleting the hypothetical node. |
631 | $gpart->delete_vertex( $h ); |
632 | if( $h eq $root ) { |
633 | # See if we still have a single root. |
634 | my @roots = $gpart->predecessorless_vertices; |
635 | warn "This shouldn't have happened" unless @roots; |
636 | if( @roots > 1 ) { |
637 | # $h is needed by this group. |
638 | if( exists( $still_contig->{$h} ) ) { |
639 | # Conflict! |
640 | $conflict->{$gst} = 1; |
641 | $still_contig->{$h} = ''; |
642 | } else { |
643 | $still_contig->{$h} = $gst; |
644 | } |
645 | } |
646 | } else { |
647 | # $h is somewhere in the middle. See if everything |
648 | # else can still be reached from the root. |
649 | my %still_reachable = ( $root => 1 ); |
650 | map { $still_reachable{$_} = 1 } |
651 | $gpart->all_successors( $root ); |
652 | foreach my $v ( keys %$reachable ) { |
653 | next if $v eq $h; |
654 | if( !$still_reachable{$v} |
655 | && ( $contig->{$v} eq $gst |
656 | || ( exists $still_contig->{$v} |
657 | && $still_contig->{$v} eq $gst ) ) ) { |
658 | # We need $h. |
659 | if( exists $still_contig->{$h} ) { |
660 | # Conflict! |
661 | $conflict->{$gst} = 1; |
662 | $still_contig->{$h} = ''; |
663 | } else { |
664 | $still_contig->{$h} = $gst; |
665 | } |
666 | last; |
667 | } # else we don't need $h in this group. |
668 | } # end foreach $v |
669 | } # endif $h eq $root |
670 | } # end foreach $gst |
671 | } # end foreach $h |
672 | |
673 | # Now we have some hypothetical vertices in $still_contig that are the |
674 | # "real" group memberships. Replace these in $contig. |
675 | foreach my $v ( keys %$contig ) { |
676 | next unless ref $contig->{$v}; |
677 | $contig->{$v} = $still_contig->{$v}; |
678 | } |
679 | } # end if @resolve |
680 | |
681 | my $is_genealogical = keys %$conflict ? JSON::false : JSON::true; |
682 | my $variant_row = [ [], $is_genealogical ]; |
683 | # Fill in the groupings from $contig. |
684 | foreach my $g ( @$groups ) { |
685 | my $gst = wit_stringify( $g ); |
686 | my @realgroup = grep { $contig->{$_} eq $gst } keys %$contig; |
687 | push( @{$variant_row->[0]}, \@realgroup ); |
688 | } |
689 | return $variant_row; |
690 | } |
fae07016 |
691 | |
7f52eac8 |
692 | sub _prune_subtree { |
231d71fc |
693 | my( $tree, $root, $contighash ) = @_; |
694 | # First, delete hypothetical leaves / orphans until there are none left. |
695 | my @orphan_hypotheticals = grep { ref( $contighash->{$_} ) } |
696 | $tree->successorless_vertices; |
697 | while( @orphan_hypotheticals ) { |
698 | $tree->delete_vertices( @orphan_hypotheticals ); |
699 | @orphan_hypotheticals = grep { ref( $contighash->{$_} ) } |
700 | $tree->successorless_vertices; |
701 | } |
702 | # Then delete a hypothetical root with only one successor, moving the |
bebec0e9 |
703 | # root to the first child that has no other predecessors. |
231d71fc |
704 | while( $tree->successors( $root ) == 1 && ref $contighash->{$root} ) { |
705 | my @nextroot = $tree->successors( $root ); |
706 | $tree->delete_vertex( $root ); |
bebec0e9 |
707 | ( $root ) = grep { $tree->is_predecessorless_vertex( $_ ) } @nextroot; |
231d71fc |
708 | } |
709 | # The tree has been modified in place, but we need to know the new root. |
bebec0e9 |
710 | $root = undef unless $root && $tree->has_vertex( $root ); |
231d71fc |
711 | return $root; |
712 | } |
d71100ed |
713 | # Add the variant, subject to a.c. representation logic. |
714 | # This assumes that we will see the 'main' version before the a.c. version. |
715 | sub add_variant_wit { |
716 | my( $arr, $wit, $acstr ) = @_; |
717 | my $skip; |
718 | if( $wit =~ /^(.*)\Q$acstr\E$/ ) { |
719 | my $real = $1; |
720 | $skip = grep { $_ =~ /^\Q$real\E$/ } @$arr; |
721 | } |
722 | push( @$arr, $wit ) unless $skip; |
723 | } |
724 | |
5be0cdeb |
725 | sub _useful_variant { |
726 | my( $group_readings, $graph, $acstr ) = @_; |
727 | |
728 | # TODO Decide what to do with AC witnesses |
729 | |
730 | # Sort by group size and return |
731 | my $is_useful = 0; |
732 | my( @readings, @groups ); # The sorted groups for our answer. |
733 | foreach my $rdg ( sort { @{$group_readings->{$b}} <=> @{$group_readings->{$a}} } |
734 | keys %$group_readings ) { |
735 | push( @readings, $rdg ); |
736 | push( @groups, $group_readings->{$rdg} ); |
737 | if( @{$group_readings->{$rdg}} > 1 ) { |
738 | $is_useful++; |
739 | } else { |
740 | my( $wit ) = @{$group_readings->{$rdg}}; |
741 | $wit =~ s/^(.*)\Q$acstr\E$/$1/; |
742 | $is_useful++ unless( $graph->is_sink_vertex( $wit ) ); |
743 | } |
744 | } |
745 | if( $is_useful > 1 ) { |
746 | return( \@readings, \@groups ); |
747 | } else { |
748 | return( [], [] ); |
749 | } |
750 | } |
751 | |
7f52eac8 |
752 | =head2 wit_stringify( $groups ) |
753 | |
754 | Takes an array of witness groupings and produces a string like |
755 | ['A','B'] / ['C','D','E'] / ['F'] |
d71100ed |
756 | |
7f52eac8 |
757 | =cut |
d71100ed |
758 | |
759 | sub wit_stringify { |
760 | my $groups = shift; |
761 | my @gst; |
762 | # If we were passed an array of witnesses instead of an array of |
763 | # groupings, then "group" the witnesses first. |
764 | unless( ref( $groups->[0] ) ) { |
765 | my $mkgrp = [ $groups ]; |
766 | $groups = $mkgrp; |
767 | } |
768 | foreach my $g ( @$groups ) { |
769 | push( @gst, '[' . join( ',', map { "'$_'" } @$g ) . ']' ); |
770 | } |
771 | return join( ' / ', @gst ); |
772 | } |
7f52eac8 |
773 | |
5be0cdeb |
774 | # Helper function to ensure that X and X a.c. never appear in the same list. |
775 | sub _add_to_witlist { |
776 | my( $wit, $list, $acstr ) = @_; |
777 | my %inlist; |
778 | my $idx = 0; |
779 | map { $inlist{$_} = $idx++ } @$list; |
780 | if( $wit =~ /^(.*)\Q$acstr\E$/ ) { |
781 | my $acwit = $1; |
782 | unless( exists $inlist{$acwit} ) { |
783 | push( @$list, $acwit.$acstr ); |
784 | } |
785 | } else { |
786 | if( exists( $inlist{$wit.$acstr} ) ) { |
787 | # Replace the a.c. version with the main witness |
788 | my $i = $inlist{$wit.$acstr}; |
789 | $list->[$i] = $wit; |
790 | } else { |
791 | push( @$list, $wit ); |
792 | } |
793 | } |
794 | } |
795 | |
bebec0e9 |
796 | sub _symmdiff { |
797 | my( $lista, $listb ) = @_; |
7f52eac8 |
798 | my %union; |
799 | my %scalars; |
800 | map { $union{$_} = 1; $scalars{$_} = $_ } @$lista; |
801 | map { $union{$_} += 1; $scalars{$_} = $_ } @$listb; |
bebec0e9 |
802 | my @set = grep { $union{$_} == 1 } keys %union; |
7f52eac8 |
803 | return map { $scalars{$_} } @set; |
804 | } |
805 | |
806 | 1; |
807 | |
808 | =head1 LICENSE |
809 | |
810 | This package is free software and is provided "as is" without express |
811 | or implied warranty. You can redistribute it and/or modify it under |
812 | the same terms as Perl itself. |
813 | |
814 | =head1 AUTHOR |
815 | |
816 | Tara L Andrews E<lt>aurum@cpan.orgE<gt> |