pedigree.h 56.2 KB
Newer Older
1
2
3
4
5
6
7
8
9
10
#ifndef _SPELL_BAYES_CSV_H_
#define _SPELL_BAYES_CSV_H_

#include <iostream>
#include <fstream>
#include <string>
#include <sstream>
#include <stdexcept>
#include <vector>
#include <utility>
11
#include <unordered_set>
12
13


Damien Leroux's avatar
Damien Leroux committed
14
15
/*#include "permutation.h"*/
/*#include "symmetry.h"*/
16
17
#include "geno_matrix.h"
#include "linear_combination.h"
Damien Leroux's avatar
Damien Leroux committed
18
#include "pedigree_tree.h"
19
#include "bayes/output.h"
20
21


22
23
24
25
26
27
28
29
30
31
32
struct bn_label_type {
    char first;
    char second;
    char first_allele;
    char second_allele;
    bn_label_type() : first(0), second(0), first_allele(0), second_allele(0) {}
    bn_label_type(int) : first(0), second(0), first_allele(0), second_allele(0) {}
    bn_label_type(char f, char s, char fa, char sa)
        : first(f), second(s), first_allele(fa), second_allele(sa)
    {}

33
34
35
    int compact() const { return *(int*) this; }
    int& compact() { return *(int*) this; }

36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
    friend std::ostream& operator << (std::ostream& os, const bn_label_type& bl)
    {
        if (bl.second != GAMETE_EMPTY) {
            return os << bl.first << ((int) bl.first_allele) << bl.second << ((int) bl.second_allele);
        } else {
            return os << bl.first << ((int) bl.first_allele);
        }
    }

    bool operator < (const bn_label_type& other) const
    {
        /*return first < other.first || (first == other.first*/
            /*&& (second < other.second || (second == other.second*/
            /*&& (first_allele < other.first_allele || (first_allele == other.first_allele*/
            /*&& second_allele < other.second_allele)))));*/
        return (*(int*)this) < (*(int*) &other);
    }

    bool operator == (const bn_label_type& other) const
    {
        return (*(int*)this) == (*(int*) &other);
    }
58
59
60
61
62

    bool operator != (const bn_label_type& other) const
    {
        return (*(int*)this) != (*(int*) &other);
    }
63
64
65
};

typedef combination_type<size_t, bn_label_type> genotype_comb_type;
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223


template <typename Arg>
int read_field(std::stringstream& s, char sep, Arg& arg)
{
    std::string field;
    std::getline(s, field, sep);
    /*MSG_DEBUG("CSV FIELD |" << field << "|");*/
    std::stringstream ss(field);
    ss >> arg;
    return 0;
}


#define do_with_arg_pack(_expr) do { using _ = int[]; (void)_{0, ((_expr), void(), 0)...}; } while(0)

template <typename... Args>
void read_csv_line(std::istream& is, char sep, Args&... args)
{
    std::string line;
    std::getline(is, line);
    /*MSG_DEBUG("CSV LINE |" << line << "|");*/
    std::stringstream ss(line);
    do_with_arg_pack(read_field(ss, sep, args));
}





struct pedigree_item {
    std::string gen_name;
    int id;
    int p1;
    int p2;

    pedigree_item(const char* gn, int i, int a, int b)
        : gen_name(gn), id(i), p1(a), p2(b)
    {}

    pedigree_item(std::istream& is, char field_sep=';')
        : gen_name()
    {
        id = p1 = p2 = 0;
        read_csv_line(is, field_sep, gen_name, id, p1, p2);
        if (id == p1 && id == p2 && id == 0) {
            return;
        }
        if (id <= p1 || id <= p2) {
            throw std::runtime_error("Bad ID! ID must be greater than p1 AND p2");
            /*MSG_DEBUG("BAD ID!! " << id << " must be greater than " << p1 << " AND " << p2);*/
        }
    }

    bool is_ancestor() const { return p1 == 0 && p2 == 0; }
    bool is_self() const { return p1 > 0 && p1 == p2; }
    bool is_cross() const { return p1 > 0 && p2 > 0 && p1 != p2; }
    bool is_dh() const { return p1 != p2 && p1 >= 0 && p2 >= 0 && (p1 * p2) == 0; }
    bool is_bullshit() const { return !(is_ancestor() || is_self() || is_cross() || is_dh()); }
};


std::vector<pedigree_item>
read_csv(const std::string& pedigree_file, char field_sep=';');



typedef std::map<size_t, size_t> ancestor_list_type;


ancestor_list_type reentrants(const ancestor_list_type& a)
{
    ancestor_list_type ret;
    for (const auto& kv: a) {
        if (kv.second > 1) {
            ret.emplace(kv);
        }
    }
    return ret;
}


ancestor_list_type operator + (const ancestor_list_type& a1, const ancestor_list_type& a2)
{
    ancestor_list_type ret(a1);
    for (const auto& kv: a2) {
        ret[kv.first] += kv.second;
    }
    return ret;
}


ancestor_list_type operator / (const ancestor_list_type& a, const ancestor_list_type& restr)
{
    ancestor_list_type ret;
    for (const auto& kv: a) {
        auto i = restr.find(kv.first);
        if (i != restr.end()) {
            ret.emplace(kv.first, std::min(kv.second, i->second));
        }
    }
    return ret;
}


ancestor_list_type operator % (const ancestor_list_type& a, const ancestor_list_type& restr)
{
    ancestor_list_type ret;
    for (const auto& kv: a) {
        if (restr.find(kv.first) != restr.end()) {
            ret.emplace(kv);
        }
    }
    return ret;
}


ancestor_list_type operator - (const ancestor_list_type& a, const ancestor_list_type& restr)
{
    ancestor_list_type ret;
    for (const auto& kv: a) {
        auto it = restr.find(kv.first);
        if (it == restr.end()) {
            ret.emplace(kv);
        } else if (kv.second > it->second) {
            ret.emplace(kv.first, kv.second - it->second);
        }
    }
    return ret;
}


ancestor_list_type operator * (const ancestor_list_type& a, size_t weight)
{
    ancestor_list_type ret;
    for (const auto& kv: a) {
        ret.emplace(kv.first, kv.second * weight);
    }
    return ret;
}


std::ostream& operator << (std::ostream& os, const ancestor_list_type& a)
{
    auto i = a.begin();
    auto j = a.end();
    if (i != j) {
        os << i->first << ':' << i->second;
        for (++i; i != j; ++i) {
            os << ' ' << i->first << ':' << i->second;
        }
    } else {
        os << "empty";
    }
    return os;
}


224
#if 0
225
226
227
228
229
230
231
232
233
234
235
236
label_type operator * (label_type a, label_type b)
{
    label_type ret;
    if (a.second == GAMETE_EMPTY) {
        ret = {a.first, b.first};
    } else {
        if (b.second == GAMETE_EMPTY) {
            ret = {SELECT(a, b.first), GAMETE_EMPTY};
        } else {
            ret = {SELECT(a, b.first), SELECT(a, b.second)};
        }
    }
237
238
239
    /*MSG_DEBUG("" << a << " * " << b << " = " << ret);*/
    return ret;
}
240
#endif
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257


#define SELECT_A(__p, __b) ((__b) == GAMETE_R ? (__p).second_allele : (__p).first_allele)

bn_label_type operator * (bn_label_type a, bn_label_type b)
{
    bn_label_type ret;
    if (a.second == GAMETE_EMPTY) {
        ret = {a.first, b.first, a.first_allele, b.first_allele};
    } else {
        if (b.second == GAMETE_EMPTY) {
            ret = {SELECT(a, b.first), GAMETE_EMPTY, SELECT_A(a, b.first), 0};
        } else {
            ret = {SELECT(a, b.first), SELECT(a, b.second), SELECT_A(a, b.first), SELECT_A(a, b.second)};
        }
    }
    /*MSG_DEBUG("" << a << " * " << b << " = " << ret);*/
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
    return ret;
}


template <typename F> struct make_one;
template <> struct make_one<MatrixXd> {
    static MatrixXd& _(bool der) {
        static MatrixXd one = MatrixXd::Ones(1, 1);
        static MatrixXd zero = MatrixXd::Zero(1, 1);
        return der ? zero : one;
    }
};
template <> struct make_one<VectorXd> {
    static VectorXd& _(bool der) {
        static VectorXd one = VectorXd::Ones(1);
        static VectorXd zero = VectorXd::Zero(1);
        return der ? zero : one;
    }
};


279
280
281
282
283
284
285
286
287
288
template <class DERIVED>
struct rw_any {
    bool fourcc(std::ifstream& ifs, const char* cc)
    {
        if (check_fourcc(ifs, cc)) {
            MSG_ERROR("File is not valid or has been corrupted", "");
            return true;
        }
        return false;
    }
289

290
291
292
293
294
    bool fourcc(std::ofstream& ofs, const char* cc)
    {
        write_fourcc(ofs, cc);
        return false;
    }
295

296
    virtual ~rw_any() {}
297

298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
    DERIVED& ref() { return *dynamic_cast<DERIVED*>(this); }

    void operator () (std::ifstream& ifs, std::string& s) { s = read_str(ifs); }
    void operator () (std::ofstream& ofs, const std::string& s) { write_str(ofs, s); }

    void operator () (std::ifstream& ifs, char& i) { i = read_char(ifs); }
    void operator () (std::ofstream& ofs, char i) { write_char(ofs, i); }

    void operator () (std::ifstream& ifs, int& i) { i = read_int(ifs); }
    void operator () (std::ofstream& ofs, int i) { write_int(ofs, i); }

    void operator () (std::ifstream& ifs, bool& i) { i = !!read_char(ifs); }
    void operator () (std::ofstream& ofs, bool i) { write_char(ofs, i); }

    void operator () (std::ifstream& ifs, double& i) { i = read_double(ifs); }
    void operator () (std::ofstream& ofs, double i) { write_double(ofs, i); }

    void operator () (std::ifstream& ifs, size_t& i) { i = read_size(ifs); }
    void operator () (std::ofstream& ofs, size_t i) { write_size(ofs, i); }

    void operator () (std::ifstream& ifs, label_type& l) { l = {read_char(ifs), read_char(ifs)}; }
    void operator () (std::ofstream& ofs, const label_type& l) { write_char(ofs, l.first()); write_char(ofs, l.second()); }

    void operator () (std::ifstream& ifs, bn_label_type& l) { l.compact() = read_int(ifs); }
    void operator () (std::ofstream& ofs, const bn_label_type& l) { write_int(ofs, l.compact()); }

    template <typename V>
        void operator () (std::ifstream& ifs, std::set<V>& vec)
326
        {
327
328
329
330
331
332
333
334
            if (fourcc(ifs, "OSET")) { return; }
            size_t sz = read_size(ifs);
            vec.clear();
            for (size_t i = 0; i < sz; ++i) {
                V tmp;
                ref() (ifs, tmp);
                vec.emplace(tmp);
            }
335
336
        }

337
338
    template <typename V>
        void operator () (std::ofstream& ofs, const std::set<V>& vec)
339
        {
340
341
342
343
            if (fourcc(ofs, "OSET")) { return; }
            write_size(ofs, vec.size());
            for (const auto& e: vec) {
                ref() (ofs, e);
344
345
346
347
            }
        }


348
349
350
351
352
353
354
355
356
357
358
359
    template <typename V>
        void operator () (std::ifstream& ifs, std::unordered_set<V>& vec)
        {
            if (fourcc(ifs, "USET")) { return; }
            size_t sz = read_size(ifs);
            vec.clear();
            for (size_t i = 0; i < sz; ++i) {
                V tmp;
                ref() (ifs, tmp);
                vec.emplace(tmp);
            }
        }
360

361
362
363
364
365
366
367
368
369
    template <typename V>
        void operator () (std::ofstream& ofs, const std::unordered_set<V>& vec)
        {
            if (fourcc(ofs, "USET")) { return; }
            write_size(ofs, vec.size());
            for (const auto& e: vec) {
                ref() (ofs, e);
            }
        }
370

371
    void operator () (std::ifstream& ifs, std::vector<bool>::reference i) { i = !!read_char(ifs); }
372

373
374
375
376
377
378
379
380
381
382
    template <typename V, typename A>
        void operator () (std::ifstream& ifs, std::vector<V, A>& vec)
        {
            if (fourcc(ifs, "VECT")) { return; }
            size_t sz = read_size(ifs);
            vec.clear();
            vec.reserve(sz);
            for (size_t i = 0; i < sz; ++i) {
                vec.emplace_back();
                ref() (ifs, vec.back());
383
            }
384
        }
385

386
387
388
389
390
391
392
    template <typename V, typename A>
        void operator () (std::ofstream& ofs, const std::vector<V, A>& vec)
        {
            if (fourcc(ofs, "VECT")) { return; }
            write_size(ofs, vec.size());
            for (const auto& e: vec) {
                ref() (ofs, e);
393
            }
394
        }
395

396
397
398
399
400
401
402
403
404
405
406
407
    template <typename K, typename V, typename A, typename C>
        void operator () (std::ifstream& ifs, std::map<K, V, A, C>& map)
        {
            if (fourcc(ifs, "MAP ")) { return; }
            size_t count = read_size(ifs);
            map.clear();
            for (size_t i = 0; i < count; ++i) {
                K key;
                V value;
                ref() (ifs, key);
                ref() (ifs, value);
                map.emplace(std::move(key), std::move(value));
408
            }
409
        }
410

411
412
413
414
415
416
417
418
    template <typename K, typename V, typename A, typename C>
        void operator () (std::ofstream& ofs, const std::map<K, V, A, C>& map)
        {
            if (fourcc(ofs, "MAP ")) { return; }
            write_size(ofs, map.size());
            for (const auto& kv: map) {
                ref() (ofs, kv.first);
                ref() (ofs, kv.second);
419
            }
420
        }
421

422
423
    template <typename SCALAR, int ROW, int COL, int C, int D, int E>
        void operator () (std::ifstream& ifs, Eigen::Matrix<SCALAR, ROW, COL, C, D, E>& mat) { read_matrix(ifs, mat); }
424

425
426
    template <typename SCALAR, int ROW, int COL, int C, int D, int E>
        void operator () (std::ofstream& ofs, const Eigen::Matrix<SCALAR, ROW, COL, C, D, E>& mat) { write_matrix(ofs, mat); }
427

428
429
    void operator () (std::ifstream& ifs, geno_matrix& mat) { read_geno_matrix(ifs, mat); }
    void operator () (std::ofstream& ofs, geno_matrix& mat) { write_geno_matrix(ofs, mat); }
430

431
432
433
434
435
436
437
438
439
    void operator () (std::ifstream& ifs, std::shared_ptr<geno_matrix>& ptr)
    {
        ptr.reset();
        ptr = std::make_shared<geno_matrix>();
        ref() (ifs, *ptr);
        if (!ptr->size()) {
            ptr.reset();
        }
    }
440

441
442
443
444
445
446
447
448
449
450
    void operator () (std::ofstream& ofs, const std::shared_ptr<geno_matrix>& ptr)
    {
        if (ptr) {
            ref() (ofs, *ptr.get());
        } else {
            geno_matrix _;
            ref() (ofs, _);
        }
    }
};
451
452


453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
struct rw_base : public rw_any<rw_base> {
    virtual ~rw_base() {}
    using rw_any<rw_base>::fourcc;
    using rw_any<rw_base>::ref;
    using rw_any<rw_base>::operator ();
};

template <typename PARENT_TYPE, typename STATE_TYPE>
struct rw_comb : public rw_any<rw_comb<PARENT_TYPE, STATE_TYPE>> {
    typedef combination_type<PARENT_TYPE, STATE_TYPE> comb_type;

    virtual ~rw_comb() {}

    using rw_any<rw_comb<PARENT_TYPE, STATE_TYPE>>::fourcc;
    using rw_any<rw_comb<PARENT_TYPE, STATE_TYPE>>::ref;
    using rw_any<rw_comb<PARENT_TYPE, STATE_TYPE>>::operator ();

    void operator () (std::ifstream& fs, typename comb_type::key_type& key) { ref() (fs, key.parent); ref() (fs, key.state); }

    void operator () (std::ofstream& fs, const typename comb_type::key_type& key) { ref() (fs, key.parent); ref() (fs, key.state); }

    void operator () (std::ifstream& fs, typename comb_type::key_list& keys) { ref() (fs, keys.keys); }

    void operator () (std::ofstream& fs, const typename comb_type::key_list& keys) { ref() (fs, keys.keys); }

    void operator () (std::ifstream& fs, typename comb_type::element_type& elem) { ref() (fs, elem.keys); ref() (fs, elem.coef); }

    void operator () (std::ofstream& fs, const typename comb_type::element_type& elem) { ref() (fs, elem.keys); ref() (fs, elem.coef); }

    void operator () (std::ifstream& fs, comb_type& comb)
    {
        if (fourcc(fs, "COMB")) { return; }
        ref() (fs, comb.m_combination);
    }

    void operator () (std::ofstream& fs, const comb_type& comb)
    {
        if (fourcc(fs, "COMB")) { return; }
        ref() (fs, comb.m_combination);
    }
};


struct rw_tree : public rw_any<rw_tree> {
    virtual ~rw_tree() {}

    using rw_any<rw_tree>::fourcc;
    using rw_any<rw_tree>::ref;
    using rw_any<rw_tree>::operator ();

    void operator () (std::ifstream& fs, pedigree_tree_type::node_type& node) { ref() (fs, node.p1); ref() (fs, node.p2); }
    void operator () (std::ofstream& fs, const pedigree_tree_type::node_type& node) { ref() (fs, node.p1); ref() (fs, node.p2); }

    void operator () (std::ifstream& fs, pedigree_tree_type& tree)
    {
        ref() (fs, tree.m_leaves);
        ref() (fs, tree.m_roots);
        ref() (fs, tree.m_nodes);
        ref() (fs, tree.m_must_recompute);
        ref() (fs, tree.m_node_number_to_ind_number);
        ref() (fs, tree.m_original_ordering);
    }

    void operator () (std::ofstream& fs, const pedigree_tree_type& tree)
    {
        ref() (fs, tree.m_leaves);
        ref() (fs, tree.m_roots);
        ref() (fs, tree.m_nodes);
        ref() (fs, tree.m_must_recompute);
        ref() (fs, tree.m_node_number_to_ind_number);
        ref() (fs, tree.m_original_ordering);
    }
525
526
527
};


528
529
530
531
532
533
534
/* TODO extraire l'arbre du pedigree
 * TODO opérations sur l'arbre :
 * TODO - insérer un nouveau noeud étant donné {P1, P2} (Pi étant soit néant soit un noeud existant)
 * TODO - extraire sous-arbre étant donné {RACINE, {FEUILLES}}
 * TODO - comparer deux arbres
 * TODO - pour deux arbres comparables, déterminer la rotation du second pour matcher le premier
 */
535
536
537
538

/*
 * pedigree_type: implements all facilities to compute proper geno_matrices for any pedigree, including reentrant individuals.
 */
539
struct pedigree_type {
540
541
542
    /*
     * pedigree tree implementation
     */
Damien Leroux's avatar
Damien Leroux committed
543
    pedigree_tree_type tree;
544
545
546
547
548

    /*
     * pedigree tree metadata
     */
    typedef size_t geno_matrix_index_type;
Damien Leroux's avatar
Damien Leroux committed
549
    typedef int individual_index_type;
550
551
552
553
    std::vector<std::shared_ptr<geno_matrix>> generations;
    std::vector<geno_matrix_index_type> node_generations;
    std::map<individual_index_type, char> ancestor_letters;
    std::map<geno_matrix_index_type, std::string> generation_names;
554
    /*std::vector<VectorLC> LC;*/
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569

    /*
     * geno_matrix cache to avoid recomputing identical generations
     */
    std::map<geno_matrix_index_type, geno_matrix_index_type> cache_gamete;
    std::map<std::pair<geno_matrix_index_type, geno_matrix_index_type>, geno_matrix_index_type> cache_geno;

    /*
     * geno_matrix database
     */

    std::map<std::string, std::set<geno_matrix_index_type>> geno_matrix_by_generation_name;
    std::map<std::string, std::vector<individual_index_type>> individuals_by_generation_name;
    std::map<individual_index_type, const std::string*> generation_name_by_individual;

570
571
572
573
574
575
576
577
578
579
580
    /*
     * overlump control
     */

    size_t max_states;

    /*
     * BN metadata
     */
    size_t n_alleles;

581
582
583
584
585
586
587
588
589
590
591
592
593
    /*
     * Metadata for XML output and recreating command line
     */
    std::string filename;

    /*
     * Actual output: LC and factors for bayesian network
     */
    std::vector<std::vector<gencomb_type>> LC;
    std::vector<std::vector<std::map<bn_label_type, genotype_comb_type>>> factor_messages;
    std::vector<std::vector<size_t>> individuals_in_factors;
    /* i-th element means the i-th variable receives a message through this factor from variables in genotype_comb_type:keys */

594
595
596
597
    /*
     * default ctor
     */
    pedigree_type()
Damien Leroux's avatar
Damien Leroux committed
598
599
600
        : tree(), node_generations(), ancestor_letters(), generation_names(),
          cache_gamete(), cache_geno(),
          max_states(NONE),
601
602
603
604
          n_alleles(1),
          filename("<none>"),
          LC(),
          factor_messages()
605
606
607
608
609
610
611
612
613
    {
        __init();
    }

    void __init()
    {
        generations.emplace_back();
    }

Damien Leroux's avatar
Damien Leroux committed
614
#if 0
615
    /*
616
     * prealloc ctor
617
618
619
     */
    pedigree_type(size_t n_ind)
    {
620
        n_alleles = 1;
Damien Leroux's avatar
Damien Leroux committed
621
        max_states = NONE;
622
623
624
625
626
627
        nodes.reserve(3 * n_ind);
        /*ind_generations.reserve(n_ind);*/
        ind_number_to_node_number.reserve(n_ind);
        __init();
    }

Damien Leroux's avatar
Damien Leroux committed
628
629
    size_t last_node_index() const { return tree.size() - 1; }
#endif
630

631
    individual_index_type spawn_gamete(const std::string&, int parent_node)
632
    {
Damien Leroux's avatar
Damien Leroux committed
633
        int n = tree.add_node(parent_node);
634
        node_generations.emplace_back(node_generations[parent_node]);
Damien Leroux's avatar
Damien Leroux committed
635
636
637
638
639
        /*MSG_DEBUG_INDENT_EXPR("[compute " << gamete_name << " gamete] ");*/
        /*compute_generation(n);*/
        /*compute_LC(n);*/
        /*MSG_DEBUG_DEDENT;*/
        return n;
640
641
    }

642
    individual_index_type spawn(const std::string& generation_name, std::initializer_list<individual_index_type> parents)
643
    {
Damien Leroux's avatar
Damien Leroux committed
644
        individual_index_type ind = tree.next_ind_idx();
645
646
647
648
649
        switch (parents.size()) {
            case 0: /* ancestor */
                {
                    MSG_DEBUG_INDENT_EXPR("[compute gen #" << ind << "] ");
                    MSG_DEBUG("ANCESTOR");
Damien Leroux's avatar
Damien Leroux committed
650
651
                    int n = tree.add_node();
                    MSG_DEBUG("node=" << n << " ind=" << ind);
652
                    compute_generation(generation_name, n);
Damien Leroux's avatar
Damien Leroux committed
653
                    /*compute_LC(n);*/
654
655
656
657
658
659
660
                    MSG_DEBUG_DEDENT;
                }
                break;
            case 1: /* doubling */
                {
                    MSG_DEBUG_INDENT_EXPR("[compute gen #" << ind << "] ");
                    MSG_DEBUG("DOUBLING");
Damien Leroux's avatar
Damien Leroux committed
661
662
663
664
                    individual_index_type p1 = *parents.begin();
                    int g1 = spawn_gamete("M", tree.ind2node(p1));
                    int n = tree.add_node(g1, g1);
                    MSG_DEBUG("node=" << n << " ind=" << ind);
Damien Leroux's avatar
Damien Leroux committed
665
666
                    /*compute_generation(generation_name, n);*/
                    /*compute_data_for_bn(n);*/
667
668
669
670
671
672
673
674
675
                    MSG_DEBUG_DEDENT;
                }
                break;
            case 2: /* crossing & selfing */
                {
                    MSG_DEBUG_INDENT_EXPR("[compute gen #" << ind << "] ");
                    MSG_DEBUG("CROSSING/SELFING");
                    auto i = parents.begin();
                    individual_index_type p1 = *i++;
Damien Leroux's avatar
Damien Leroux committed
676
                    int n1 = tree.ind2node(p1);
677
                    individual_index_type p2 = *i;
Damien Leroux's avatar
Damien Leroux committed
678
                    int n2 = tree.ind2node(p2);
679
                    /*MSG_DEBUG("p1=" << p1 << " p2=" << p2 << " n1=" << n1 << " n2=" << n2);*/
Damien Leroux's avatar
Damien Leroux committed
680
681
682
683
                    int g1 = spawn_gamete("M", n1);
                    int g2 = spawn_gamete("F", n2);
                    int n = tree.add_node(g1, g2);
                    MSG_DEBUG("node=" << n << " ind=" << ind);
Damien Leroux's avatar
Damien Leroux committed
684
685
                    /*compute_generation(generation_name, n);*/
                    /*compute_data_for_bn(n);*/
686
687
688
689
690
691
692
693
                    MSG_DEBUG_DEDENT;
                }
                break;
            default:;
        };
        return ind;
    }

694
695
696
697
    /*individual_index_type crossing(std::string& generation_name, individual_index_type p1, individual_index_type p2) { return spawn(generation_name, {p1, p2}); }*/
    /*individual_index_type selfing(std::string& generation_name, individual_index_type p1) { return spawn(generation_name, {p1, p1}); }*/
    /*individual_index_type dh(std::string& generation_name, individual_index_type p1) { return spawn(generation_name, {p1}); }*/
    /*individual_index_type ancestor(std::string& generation_name) { return spawn(generation_name, {}); }*/
698
699
700
701
702
703
704
705
706
707

    individual_index_type fill_db(const std::string& name, individual_index_type ind)
    {
        geno_matrix_by_generation_name[name].insert(get_gen_index(ind)).first;
        individuals_by_generation_name[name].push_back(ind);
        auto it = individuals_by_generation_name.find(name);
        generation_name_by_individual[ind] = &it->first;
        return ind;
    }

708
709
710
711
    individual_index_type crossing(const std::string& name, individual_index_type p1, individual_index_type p2) { return fill_db(name, spawn(name, {p1, p2})); }
    individual_index_type selfing(const std::string& name, individual_index_type p1) { return fill_db(name, spawn(name, {p1, p1})); }
    individual_index_type dh(const std::string& name, individual_index_type p1) { return fill_db(name, spawn(name, {p1})); }
    individual_index_type ancestor(const std::string& name) { return fill_db(name, spawn(name, {})); }
712

Damien Leroux's avatar
Damien Leroux committed
713
    void propagate_symmetries(int n, geno_matrix& gen)
714
    {
715
        MSG_DEBUG_INDENT_EXPR("[propagate symmetries #" << n << "] ");
Damien Leroux's avatar
Damien Leroux committed
716
717
718
719
720
721
722
        std::vector<int> in, out;
        auto expr = tree.extract_expression(n, in, out);
        std::vector<pedigree_tree_type> input_trees;
        input_trees.reserve(in.size());
        for (int t: in) {
            input_trees.emplace_back(tree.extract_subtree(t));
        }
723
724
725
726
727
728
729
730
731
732
733
        /*auto recompute = tree.get_deep_recompute_vec(n);*/
        /*MSG_DEBUG("RECOMPUTE: " << recompute);*/
        auto get_lumper
            = [&, this] (int node) -> MatrixXb
            {
                /*if (tree[node].is_gamete() || recompute[node]) {*/
                    /*MSG_DEBUG("NIL lumper for node #" << node << " because" << (recompute[node] && tree[node].is_gamete() ? " recompute flag is set and it is a gamete" : tree[node].is_gamete() ? " it is a gamete" : " recompute flag is set"));*/
                    /*return {};*/
                /*}*/
                return get_node_gen(node)->collect.cast<bool>();
            };
Damien Leroux's avatar
Damien Leroux committed
734
        symmetry_propagator sp(expr);
735
        MSG_DEBUG_INDENT_EXPR("[SYMMETRIES] ");
Damien Leroux's avatar
Damien Leroux committed
736
737
738
739
740
        gen.symmetries = sp.compute_propagated_symmetries(
                [&] (int i) -> const pedigree_tree_type& { MSG_DEBUG("request subtree for input #" << i); return input_trees[i]; },
                [&] (int node) -> const symmetry_group_type& { MSG_DEBUG("request symmetries for node #" << node); return get_node_gen(node)->symmetries; },
                [&] (int node) -> const symmetry_group_type& { MSG_DEBUG("request latent symmetries for node #" << node); return get_node_gen(node)->latent_symmetries; },
                gen.labels, gen.inf_mat,
741
                get_lumper
Damien Leroux's avatar
Damien Leroux committed
742
                );
743
744
        MSG_DEBUG_DEDENT;
        MSG_DEBUG_INDENT_EXPR("[LATENT SYMMETRIES] ");
Damien Leroux's avatar
Damien Leroux committed
745
746
747
748
749
        auto temp = sp.compute_propagated_latent_symmetries(
                [&] (int i) -> const pedigree_tree_type& { MSG_DEBUG("request subtree for input #" << i); return input_trees[i]; },
                [&] (int node) -> const symmetry_group_type& { MSG_DEBUG("request symmetries for node #" << node); return get_node_gen(node)->symmetries; },
                [&] (int node) -> const symmetry_group_type& { MSG_DEBUG("request latent symmetries for node #" << node); return get_node_gen(node)->latent_symmetries; },
                gen.labels, gen.inf_mat,
750
                get_lumper
Damien Leroux's avatar
Damien Leroux committed
751
                );
752
        MSG_DEBUG_DEDENT;
Damien Leroux's avatar
Damien Leroux committed
753
754
        MSG_DEBUG(temp);
        gen.latent_symmetries = temp - gen.symmetries;
755
        MSG_DEBUG_INDENT_EXPR("[AFTER SYMMETRY PROPAGATION] ");
Damien Leroux's avatar
Damien Leroux committed
756
        MSG_DEBUG(gen);
757
        MSG_DEBUG_DEDENT;
758
        MSG_DEBUG_DEDENT;
759
760
    }

761
    void compute_generation(const std::string& generation_name, int n)
762
    {
Damien Leroux's avatar
Damien Leroux committed
763
        MSG_DEBUG("Computing generation for node " << tree.make_node_label(n));
764
        /*MSG_DEBUG(render_tree());*/
765

Damien Leroux's avatar
Damien Leroux committed
766
767
        int np1 = tree.get_p1(n);
        int np2 = tree.get_p2(n);
768
769
770
        node_generations.emplace_back(generations.size());
        geno_matrix new_gen;
        geno_matrix_index_type* cached_gen = NULL;
Damien Leroux's avatar
Damien Leroux committed
771
        if (np1 == NONE && np2 == NONE) {
772
773
774
            /* ancestor */
            char l = 'a' + ancestor_letters.size();
            ancestor_letters[n] = l;
775
            new_gen = ancestor_matrix(generation_name, l);
Damien Leroux's avatar
Damien Leroux committed
776
        } else if (np2 == NONE) {
777
778
            /* gamete */
            auto gp = node_generations[np1];
Damien Leroux's avatar
Damien Leroux committed
779
            auto& cache = cache_gamete;
780
781
782
783
784
785
            cached_gen = &cache[gp];
            if (*cached_gen) {
                MSG_DEBUG("GENERATION HAS ALREADY BEEN COMPUTED");
                node_generations[n] = *cached_gen;
                return;
            }
Damien Leroux's avatar
Damien Leroux committed
786
            new_gen = kronecker(*generations[gp], gamete);
787
            new_gen.name = generation_name;
788
789
            /*MSG_DEBUG("TMP GAMETE GEN");*/
            /*MSG_DEBUG(new_gen);*/
790
        } else {
791
792
            /*auto ngp1 = node_generations[np1];*/
            /*auto ngp2 = node_generations[np2];*/
Damien Leroux's avatar
Damien Leroux committed
793
794
795
            /*auto gp1 = generations[ngp1];*/
            /*auto gp2 = generations[ngp2];*/
            MSG_DEBUG("Child of " << tree.node2ind(tree.get_p1(np1)) << " and " << tree.node2ind(tree.get_p1(np2)));
796

797
798
799
800
            /* use node's grandparents generations, not parents (no gamete generation) */
            size_t g1 = node_generations[tree.get_p1(np1)];
            size_t g2 = node_generations[tree.get_p1(np2)];
            cached_gen = &cache_geno[{g1, g2}];
801
802
803
804
805
806
            if (*cached_gen) {
                MSG_DEBUG("GENERATION HAS ALREADY BEEN COMPUTED");
                node_generations[n] = *cached_gen;
                return;
            }

Damien Leroux's avatar
Damien Leroux committed
807
808
            const auto& recompute = tree.get_recompute_vec(n);

809
810
811
            std::vector<bool> visited_clear(recompute.size(), false);
            std::vector<bool> visited;

812
813
            new_gen.name = generation_name;

814
815
816
817
818
819
            visited = visited_clear;
            MSG_DEBUG("COMPUTING INF_MAT");
            new_gen.inf_mat = eval(n, &geno_matrix::inf_mat, &pedigree_type::kron_d, recompute, visited);
            MSG_DEBUG(MATRIX_SIZE(new_gen.inf_mat));
            visited = visited_clear;
            MSG_DEBUG("COMPUTING DIAG");
820
            new_gen.diag = eval(n, &geno_matrix::diag, &pedigree_type::kron_d_diag, recompute, visited);
821
822
823
824
825
826
827
828
829
830
831
            visited = visited_clear;
            MSG_DEBUG("COMPUTING STAT_DIST");
            new_gen.stat_dist = eval(n, &geno_matrix::stat_dist, &pedigree_type::kron, recompute, visited);
            visited = visited_clear;
            MSG_DEBUG("COMPUTING P");
            new_gen.p = eval(n, &geno_matrix::p, &pedigree_type::kron, recompute, visited);
            visited = visited_clear;
            MSG_DEBUG("COMPUTING P_INV");
            new_gen.p_inv = eval(n, &geno_matrix::p_inv, &pedigree_type::kron_d, recompute, visited);
            /*new_gen.labels = eval_labels(n, recompute, visited_clear);*/
            new_gen.labels = eval_vector(n, recompute, &pedigree_type::get_labels, reentrant_label);
Damien Leroux's avatar
Damien Leroux committed
832
            new_gen.variant = (tree[n].is_genotype()
833
                               ? Geno
Damien Leroux's avatar
Damien Leroux committed
834
                               : Gamete);
835
836
837
838
839
840
841
842
            new_gen.dispatch = new_gen.collect = MatrixXd::Identity(new_gen.diag.size(), new_gen.diag.size());
            MSG_DEBUG(MATRIX_SIZE(new_gen.inf_mat));
            MSG_DEBUG(MATRIX_SIZE(new_gen.p));
            MSG_DEBUG(MATRIX_SIZE(new_gen.p_inv));
            MSG_DEBUG(MATRIX_SIZE(new_gen.diag));
            MSG_DEBUG("new_gen.labels.size()=" << new_gen.labels.size());
            MSG_QUEUE_FLUSH();

Damien Leroux's avatar
Damien Leroux committed
843
            /*if (!(ind_number_to_node_number.size() == 9 && tree.size() == 23)) {*/
844
                /*MSG_DEBUG("PROPAGATING SYMMETRIES");*/
845
                /*propagate_symmetries(new_gen, recompute, n);*/
846
847
                /*study_expression_symmetries(new_gen);*/
                /*complete_symmetries(new_gen);*/
848
                /*MSG_DEBUG("COMPUTING LATENT SYMMETRY");*/
849
            /*}*/
850
851
852
        }
        node_generations[n] = generations.size();
        generations.emplace_back(new geno_matrix());
853
854
        /*MSG_DEBUG("BEFORE LUMPING");*/
        /*MSG_DEBUG(new_gen);*/
855
        *generations.back() = lump(new_gen, max_states);
856
857
858
        /*if (tree[n].is_crossing()) {*/
            /*propagate_symmetries(n, *generations.back());*/
        /*} else if (tree[n].is_ancestor()) {*/
Damien Leroux's avatar
Damien Leroux committed
859
            generations.back()->symmetries = symmetry_group_type(generations.back()->labels);
860
        /*}*/
861
862
863
864
865
        /**node_generations[n] = lump(new_gen);*/
        if (cached_gen) {
            *cached_gen = node_generations[n];
        }
        MSG_DEBUG("DONE COMPUTING GENERATION FOR NODE #" << n);
866
        MSG_DEBUG_INDENT_EXPR("[RESULT " << tree.make_node_label(n) << " gen#" << node_generations[n] << "] ");
867
868
        MSG_DEBUG((*generations.back()));
        MSG_DEBUG_DEDENT;
869
        /*MSG_DEBUG((*generations[node_generations[n]]));*/
870
871
872
        MSG_DEBUG("=========================================================================");
    }

873
    std::map<genotype_comb_type::key_list, double> GLC_norm_factors(const std::vector<genotype_comb_type>& expanded)
874
    {
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
        std::map<genotype_comb_type::key_list, double> ret;
        for (const auto& e: expanded) {
            const auto& elem = e.m_combination.front();  /* all LC are singletons in the expanded vector */
            for (const auto& k: elem.keys) {
                auto sub_k = elem.keys - k;
                ret[sub_k] += elem.coef;
            }
        }
        return ret;
    }

    void compute_data_for_bn(int n)
    {
        compute_LC(n);

        factor_messages.resize(n + 1);
        for (auto& dest_f: compute_factors(n, true)) {
            factor_messages[dest_f.first].emplace_back(std::move(dest_f.second));
893
        }
894
895
896
897
898
899
900
901
902
903
904
905
906
        for (auto& dest_f: compute_factors(n, false)) {
            factor_messages[dest_f.first].emplace_back(std::move(dest_f.second));
        }
        individuals_in_factors.resize(n + 1);
        std::vector<size_t>& iif = individuals_in_factors.back();
        std::vector<int> in;
        std::vector<int> out;
        auto expr = tree.extract_expression(n, in, out);
        iif.reserve(expr.m_nodes.size());
        for (size_t i = 0; i < expr.m_nodes.size(); ++i) {
            if (expr.m_nodes[i].is_genotype()) {
                iif.push_back(expr.original_node_number(i));
            }
907
        }
908
909
        MSG_DEBUG("INDIVIDUALS IN FACTOR: " << iif);
    }
910

911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
    void
        compute_LC(int n)
        {
            int p1 = tree.get_p1(n);
            std::vector<gencomb_type> lc;
            const geno_matrix& m = *generations[node_generations[n]];
            if (p1 == NONE) {
                lc.emplace_back(1.);
            } else {
                lc = eval_vector(tree.size() - 1, tree.get_recompute_vec(n), &pedigree_type::get_LC, reentrant_LC);
            }
            VectorLC tmp(lc.size()), lumped;
            for (size_t i = 0; i < lc.size(); ++i) {
                tmp(i) = lc[i];
            }
            lumped = m.collect.cast<gencomb_type>() * tmp;
            LC.resize(n + 1);
            LC.back().assign(lumped.data(), lumped.data() + lumped.size());
            MSG_DEBUG("Computed new LC:");
            MSG_DEBUG("" << LC.back());
931
932
        }

933
934
935
936
937
    std::map<size_t, std::map<bn_label_type, genotype_comb_type>>
        compute_factors(int n, bool up)
        {
            std::vector<genotype_comb_type> glc;
            std::vector<bn_label_type> bn_labels;
938

939
940
941
            if (tree.get_p1(n) == NONE) {
                return {};
            }
942

943
944
945
946
947
948
949
950
951
952
            if (up) {
                std::vector<bool> recompute(n + 1, false);
                recompute[n] = true;
                recompute[tree.get_p1(n)] = true;
                recompute[tree.get_p2(n)] = true;
                glc = eval_vector(tree.size() - 1, recompute, &pedigree_type::get_GLC, reentrant_GLC);
                bn_labels = eval_vector(tree.size() - 1, recompute, &pedigree_type::get_bn_labels, reentrant_bn_label);
            } else {
                glc = eval_vector(tree.size() - 1, tree.get_recompute_vec(n), &pedigree_type::get_GLC, reentrant_GLC);
                bn_labels = eval_vector(tree.size() - 1, tree.get_recompute_vec(n), &pedigree_type::get_bn_labels, reentrant_bn_label);
953
954
            }

955
956
957
958
959
960
961
            MSG_DEBUG("glc.size = " << glc.size());
            MSG_DEBUG("bn_labels.size = " << bn_labels.size());

            std::map<bn_label_type, genotype_comb_type> glc_map;
            for (size_t i = 0; i < bn_labels.size(); ++i) {
                glc_map[bn_labels[i]] += glc[i];
            }
962

963
964
965
966
967
968
969
970
971
972
            MSG_DEBUG("Corresponding GLC:");
            std::vector<genotype_comb_type> expanded_glc;
            for (const auto& kv: glc_map) {
                MSG_DEBUG("" << kv.first << " = " << kv.second);
                for (const auto& elem: kv.second.m_combination) {
                    expanded_glc.emplace_back();
                    expanded_glc.back().m_combination.emplace_back(elem);
                    auto& keys = expanded_glc.back().m_combination.back().keys.keys;
                    keys.emplace_back(tree.size() - 1, kv.first);
                    std::sort(keys.begin(), keys.end());
973
                }
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
            }

            MSG_DEBUG("Sparse GLC coefs:");
            for (const auto& g: expanded_glc) {
                MSG_DEBUG("" << g);
            }

            auto glc_norm_factors = GLC_norm_factors(expanded_glc);
            for (const auto& kn: glc_norm_factors) {
                MSG_DEBUG('|' << kn.first << "| = " << kn.second);
            }

            std::map<size_t, std::map<bn_label_type, genotype_comb_type>> messages;
            size_t n_nodes = expanded_glc.front().m_combination.front().keys.keys.size();
            size_t first, last;
            if (up) {
                first = 0;
                last = n_nodes - 2;
            } else {
                first = 0;
                last = n_nodes - 1;
            }
            for (const auto& lc: expanded_glc) {
                const auto& elems = lc.m_combination.front();
                const auto& keys = elems.keys.keys;
                for (size_t ni = first; ni <= last; ++ni) {
                    genotype_comb_type tmp;
For faster browsing, not all history is shown. View entire blame