void mfl_reroot_subtree(node *n, node *atip, node *subtr, node *base, node *up,
node *dn, tree *swapingon, tree **savedtrees, int ntax,
int nchar, int numnodes, mfl_searchrec *searchrec,
int diff)
{
/* Traverses the subtree and re-roots it at each branch in preorder and then
* calls regrafting traversal (same as used in SPR).*/
if (searchrec->success) {
return;
}
mfl_join_nodes(base->next->next, n->edge);
mfl_join_nodes(base->next, n);
// Reoptimize the subtree base
mfl_reopt_subtr_root(base, nchar);
if (!base->next->edge->origbase && !base->next->next->edge->origbase) {
up->visited = 0;
dn->visited = 0;
}
mfl_regrafting_traversal(up, subtr, swapingon, savedtrees, ntax, nchar,
numnodes, searchrec, diff);
if (searchrec->success) {
up->visited = 1;
dn->visited = 1;
base->skip = true;
return;
}
mfl_regrafting_traversal(dn, subtr, swapingon, savedtrees, ntax, nchar,
numnodes, searchrec, diff);
up->visited = 1;
dn->visited = 1;
base->skip = true;
if (searchrec->success) {
return;
}
// Remove the base
mfl_join_nodes(base->next->edge, base->next->next->edge);
if (n->tip) {
return;
}
mfl_reroot_subtree(n->next->edge, atip, subtr, base, up, dn, swapingon,
savedtrees, ntax, nchar, numnodes, searchrec, diff);
if (searchrec->success) {
return;
}
mfl_reroot_subtree(n->next->next->edge, atip, subtr, base, up, dn,
swapingon, savedtrees, ntax, nchar, numnodes, searchrec,
diff);
}
void mfl_insert_branch(node *br, node *target, int ntax)
{
// Inserts a branch with a ring base into another branch
node *br1, *br2, *tdesc;
tdesc = target->edge;
if (br->tip) {
br1 = br->edge->next;
br2 = br1->next;
}
else {
br1 = mfl_seek_ringnode(br, ntax);
br2 = mfl_seek_ringnode(br1->next, ntax);
}
if (br1->edge || br2->edge) {
dbg_printf("Error in branch insertion\n");
return;
}
mfl_join_nodes(br1, target);
mfl_join_nodes(br2, tdesc);
}
struct node * cpyfromNWK(char *nwktr, int nwklen, int ntax, int numnodes,
int *pos, nodearray nds, bool isRooted)
{
int i, tipnum;
char tipbuf[10];
node *n, *nlst, *p;
n = mfl_seek_internal(ntax - 1, numnodes, nds);
n->initialized = 1;
nlst = n;
do {
*pos = *pos + 1;
if (nwktr[*pos] == ')')
{
nlst->next = n;
return n;
}
if (isdigit(nwktr[*pos]))
{
for (i = 0; nwktr[*pos] != ',' && nwktr[*pos] != ')'; ++i, ++*pos)
{
tipbuf[i] = nwktr[*pos];
}
tipbuf[i] = '\0';
tipnum = atoi(tipbuf);
nlst->next = mfl_allocnode();
nlst = nlst->next;
mfl_join_nodes(nds[tipnum - 1], nlst);
if (nwktr[*pos] == ')')
{
nlst->next = n;
mfl_set_ring_to_n(n);
return n;
}
}
if (nwktr[*pos] == '(')
{
nlst->next = mfl_allocnode();
nlst = nlst->next;
p = cpyfromNWK(nwktr, nwklen, ntax, numnodes, pos, nds, isRooted);
mfl_join_nodes(p, nlst);
}
} while (nwktr[*pos]);
nlst->next = n;
return n;
}
tree *mfl_addseq_randasis(int ntax, int nchar, int numnodes,
charstate *tipdata, bool addRandom,
tree **savedtrees)
{
int i, nbeslen = 0;
int *bestlen = &nbeslen;
int *taxarray;
node *p, *bestpos;
tree *newtree = mfl_alloc_noring(ntax, numnodes);
taxarray = (int*)malloc(ntax * sizeof(int));
memset(taxarray, 0, ntax * sizeof(int)); // This is to see if I can fix the problem
mfl_init_taxarray(taxarray, ntax);
if (addRandom) {
dbg_printf("Joining taxa by random addition sequence\n");
mfl_shuffle(taxarray, ntax);
}
else {
dbg_printf("Joining taxa according to order in matrix\n");
}
p = newtree->trnodes[ntax + 1];
mfl_newring(p, ntax);
i = 0;
do {
mfl_join_nodes(newtree->trnodes[taxarray[i] - 1], p);
p = p->next;
++i;
} while (p != newtree->trnodes[ntax + 1]);
mfl_temproot(newtree, taxarray[0] - 1, ntax);
newtree->length = *bestlen;
bestpos = newtree->trnodes[taxarray[0] - 1];
for (i = 3; i < ntax; ++i) {
mfl_newring(newtree->trnodes[ntax + i - 1], ntax);
mfl_join_nodes(newtree->trnodes[taxarray[i] - 1], newtree->trnodes[ntax + i - 1]->next);
mfl_insert_branch(newtree->trnodes[taxarray[i] - 1], newtree->trnodes[taxarray[0] - 1], ntax);
*bestlen = mfl_get_sttreelen(newtree, tipdata, ntax, nchar, bestlen);
//dbg_printf("Preliminary length: %i\n", *bestlen);
mfl_remove_branch(newtree->trnodes[taxarray[i] - 1]);
bestpos = mfl_tryall(newtree->root, newtree->trnodes[taxarray[i] - 1], bestpos, ntax, nchar,
numnodes, bestlen, newtree, savedtrees, tipdata);
//Join the nodes//
mfl_insert_branch(newtree->trnodes[taxarray[i] - 1], bestpos, ntax);
*bestlen = 0;
}
mfl_undo_temproot(ntax, newtree);
//newtree->bipartitions = mfl_tree_biparts(newtree, ntax, numnodes);
free(taxarray);
return newtree;
}
struct tree *randunrooted(int ntax, int numnodes)
{
/* Returns a random unrooted tree*/
int i;
int *taxarray;
node *p, *q;
tree *randtree;
taxarray =(int*) malloc(ntax * sizeof(int));
mfl_init_taxarray(taxarray, ntax);
mfl_shuffle(taxarray, ntax);
randtree = mfl_alloctree(ntax, numnodes);
randtree->trnodes[0]->start = true;
randtree->trnodes[0]->edge = randtree->trnodes[taxarray[0]];
// Join all the internal nodes (except the root) together
for (i = 1; i <= (ntax - 3); ++i) {
p = randtree->trnodes[ntax + i]->next->next;
q = randtree->trnodes[ntax + i + 1];
mfl_join_nodes(p, q);
}
// Add all the tips to the appropriate internal nodes
mfl_join_nodes(randtree->trnodes[ntax + 1], randtree->trnodes[taxarray[0] - 1]);
for (i = 1; i < ntax - 1; ++i) {
randtree->trnodes[taxarray[i] - 1]->edge = randtree->trnodes[ntax + i]->next;
randtree->trnodes[ntax + i]->next->edge = randtree->trnodes[taxarray[i] - 1];
}
randtree->trnodes[2 * ntax - 2]->next->next->edge = randtree->trnodes[taxarray[i] - 1];
randtree->trnodes[taxarray[i] - 1]->edge = randtree->trnodes[2 * ntax - 2]->next->next;
free(taxarray);
/*printNewick(randtree->trnodes[0]);*/
dbg_printf(";\n");
return (randtree);
}
void mfl_remove_branch(node *n)
{
node *p, *q, *nb;
nb = n->edge;
p = nb->next->edge;
q = nb->next->next->edge;
nb->next->edge = NULL;
nb->next->next->edge = NULL;
mfl_join_nodes(p, q);
}
void mfl_bisection_traversal(node *n, tree *swapingon, tree **savedtrees,
int ntax, int nchar, int numnodes,
mfl_searchrec *searchrec)
{
/* Traverses a binary tree clipping out a subtree in postorder and passing
* a pointer to the subtree to mfl_reroot_subtree. */
int i, diff, *srcchanging, *tgtchanging;
node *p, *q, *src, *tgt, *s_dn, *s_up, *t_dn, *t_up, *t_dn_N, *t_up_N;
node *s_dn_N, *s_up_N, *atip;
nodearray nds = swapingon->trnodes;
mfl_undone_tree(swapingon->trnodes, numnodes);
for (i = ntax + 1; i < numnodes; ++i) {
p = nds[i];
q = nds[i];
do {
src = p->edge;
tgt = p;
src->done = true;
if (!tgt->done) {
if (!src->skip) {
src->skip = true;
if (src->tip) {
memcpy(src->apomorphies, src->tempapos,
nchar * sizeof(charstate));
}
else {
mfl_set_updown(src, &s_up, &s_dn);
s_up_N = s_up->edge;
s_dn_N = s_dn->edge;
if (src->tocalcroot) {
srcchanging = mfl_get_subtr_changing(src, NULL, NULL,
nchar);
mfl_reopt_subtr(src, swapingon, nchar, numnodes,
srcchanging);
free(srcchanging);
}
else {
if (src->next->edge->tip && src->next->next->edge->tip) {
mfl_set_rootstates(src, nchar, NULL);
}
else {
srcchanging = mfl_get_tgt_changing(s_dn->edge, s_up,
s_dn, nchar);
mfl_join_nodes(s_dn, s_up);
mfl_partial_downpass(s_dn, swapingon, numnodes, ntax,
nchar, srcchanging);
mfl_join_nodes(s_up_N, s_up);
mfl_join_nodes(s_dn_N, s_dn);
mfl_reopt_subtr_root(src, nchar);
free(srcchanging);
}
}
}
mfl_set_updown(tgt, &t_up, &t_dn);
t_up_N = t_up->edge;
t_dn_N = t_dn->edge;
if (!tgt->tocalcroot) {
tgtchanging = mfl_get_tgt_changing(t_dn->edge, t_up, t_dn,
nchar);
// Pop out redundant node:
mfl_join_nodes(t_up, t_dn);
mfl_partial_downpass(t_dn, swapingon, numnodes, ntax, nchar,
tgtchanging);
free(tgtchanging);
}
else {
tgtchanging = mfl_get_subtr_changing(tgt, NULL, NULL,
nchar);
tgt->isroot = true;
mfl_reopt_subtr(tgt, swapingon, nchar, numnodes,
tgtchanging);
tgt->isroot = false;
free(tgtchanging);
// Pop out the redundant node:
mfl_join_nodes(t_up, t_dn);
}
diff = mfl_subtr_reinsertion(src, t_up, t_dn, nchar);
t_up->visited = true;
t_dn->visited = true;
if (mfl_subtr_isrerootable(src)) {
atip = mfl_find_atip(src);
mfl_join_nodes(s_up, s_dn);
s_up->origbase = true;
s_dn->origbase = true;
mfl_reroot_subtree(atip->edge, atip, src->edge->next->next,
src, t_up, t_dn, swapingon, savedtrees,
ntax, nchar, numnodes, searchrec, diff);
s_up->origbase = false;
s_dn->origbase = false;
if (searchrec->success) {
t_up->visited = false;
t_dn->visited = false;
src->skip = false;
return;
}
mfl_join_nodes(s_up_N, s_up);
mfl_join_nodes(s_dn_N, s_dn);
}
else {
mfl_regrafting_traversal(t_up, src->edge->next->next,
swapingon, savedtrees, ntax, nchar,
numnodes, searchrec, diff);
mfl_regrafting_traversal(t_dn, src->edge->next->next,
swapingon, savedtrees, ntax, nchar,
numnodes, searchrec, diff);
}
t_up->visited = false;
t_dn->visited = false;
src->skip = false;
if (searchrec->success) {
return;
}
mfl_join_nodes(t_up, t_up_N);
mfl_join_nodes(t_dn, t_dn_N);
mfl_restore_origstates(swapingon, ntax, numnodes, nchar);
src->skip = false;
}
else {
src->skip = false;
}
}
assert(q == nds[i]);
p = p->next;
} while (p != nds[i]);
}
}
void mfl_subtree_pruning(node *n, tree *swapingon, tree **savedtrees, int ntax,
int nchar, int numnodes, mfl_searchrec *searchrec)
{
/* Indexes over the node array of a tree, pruning subtrees and calling
* the regrafting function */
int i, diff = 0, *srcchanging, *tgtchanging;
node *p, *q, *src, *tgt, *s_dn, *s_up, *t_dn, *t_up, *t_dn_N, *t_up_N;
node *s_dn_N, *s_up_N;
nodearray nds = swapingon->trnodes;
mfl_undone_tree(swapingon->trnodes, numnodes);
for (i = ntax + 1; i < numnodes; ++i) {
p = nds[i];
q = nds[i];
do {
if (!p->edge->skip) {
p->edge->skip = true;
src = p->edge;
tgt = p;
if (!tgt->tocalcroot) {
mfl_set_updown(tgt, &t_up, &t_dn);
tgtchanging = mfl_get_tgt_changing(t_dn->edge, t_up, t_dn,
nchar);
// Pop out redundant node:
t_up_N = t_up->edge;
t_dn_N = t_dn->edge;
mfl_join_nodes(t_up, t_dn);
mfl_partial_downpass(t_dn, swapingon, numnodes, ntax, nchar,
tgtchanging);
free(tgtchanging);
}
else {
mfl_set_updown(tgt, &t_up, &t_dn);
if (t_up->tip && t_dn->tip) {
p = p->next;
continue;
}
else {
//mfl_set_updown(tgt, &t_up, &t_dn);
tgtchanging = mfl_get_subtr_changing(tgt, NULL, NULL,
nchar);
tgt->isroot = true;
mfl_reopt_subtr(tgt, swapingon, nchar, numnodes,
tgtchanging);
tgt->isroot = false;
free(tgtchanging);
}
// Pop out the redundant node:
t_up_N = t_up->edge;
t_dn_N = t_dn->edge;
mfl_join_nodes(t_up, t_dn);
}
if (src->tip) {
memcpy(src->apomorphies, src->tempapos,
nchar * sizeof(charstate));
}
else if (src->tocalcroot) {
mfl_set_rootstates(src, nchar, NULL);
/*if (src->next->edge->tip && src->next->next->edge->tip) {
printNewick(src);
dbg_printf("\n");
}*/
}
else {
if (src->next->edge->tip && src->next->next->edge->tip) {
mfl_set_rootstates(src, nchar, NULL);
}
else {
mfl_set_updown(src, &s_up, &s_dn);
srcchanging = mfl_get_tgt_changing(s_dn->edge, s_up,
s_dn, nchar);
s_up_N = s_up->edge;
s_dn_N = s_dn->edge;
mfl_join_nodes(s_dn, s_up);
mfl_partial_downpass(s_dn, swapingon, numnodes, ntax,
nchar, srcchanging);
mfl_join_nodes(s_up_N, s_up);
mfl_join_nodes(s_dn_N, s_dn);
mfl_reopt_subtr_root(src, nchar);
free(srcchanging);
}
}
diff = mfl_subtr_reinsertion(src, t_up, t_dn, nchar);
// Perform all reinsertions of SRC on TGT
/* OPTIMIZATION: This program can be greatly speeded up by
* taking advantage of the fact that the two subtrees are
* already reoptimized at this stage. Once one set of
* reinsertions has been completed and did not result in a
* better tree, this function can reverse the order and begin
* reinserting the old target tree on the source tree. In this
* diff may need to be recalculated. */
t_up->visited = true;
t_dn->visited = true;
mfl_regrafting_traversal(t_up, src->edge->next->next, swapingon,
savedtrees, ntax, nchar, numnodes,
searchrec, diff);
mfl_regrafting_traversal(t_dn, src->edge->next->next, swapingon,
savedtrees, ntax, nchar, numnodes,
searchrec, diff);
t_up->visited = false;
t_dn->visited = false;
p->edge->skip = false;
if (searchrec->success) {
return;
}
mfl_join_nodes(t_up, t_up_N);
mfl_join_nodes(t_dn, t_dn_N);
mfl_restore_origstates(swapingon, ntax, numnodes, nchar);
}
assert(q == nds[i]);
p->edge->skip = false;
p = p->next;
} while (p != nds[i]);
}
}
void mfl_regrafting_traversal(node *n, node *subtr, tree *swapingon,
tree **savedtrees, int ntax, int nchar, int numnodes, mfl_searchrec *searchrec, int diff)
{
/* Called from within any subtree pruning algorithm used in either SPR or
* TBR branch swapping. Traverses a binary tree in preorder, inserting the
* subtree at each node and (hopefully) skipping a reinsertion at the original
* site of pruning (flagged by the "visited" boolean values in those nodes) */
if (searchrec->success) {
return;
}
int trlength = 0;
int al = 0;
node *up;
node *down, *top;
if (!(n->visited) && !(n->edge->visited)) {
up = n->edge;
al = mfl_locreopt_cost(subtr->next->edge, n, up, nchar, diff);
trlength = searchrec->bestinrep - diff + al;
assert(trlength >= 0);
searchrec->niter_total = searchrec->niter_total + 1;
#ifdef MFY_DEBUG
/*** BEGIN COMMENT OUT BEFORE COMMIT ***
int trulen = 0;
if (subtr->tocalcroot) {
down = subtr;
top = subtr->next->next;
}
else {
down = subtr->next->next;
top = subtr;
}
mfl_join_nodes(top, up);
mfl_join_nodes(down, n);
mfl_temproot(swapingon, 0, ntax);
mfl_count_postorder(swapingon->root, &trulen, nchar, NULL);
trulen = 0;
mfl_fitch_preorder(swapingon->root, nchar, &trulen);
mfl_undo_temproot(ntax, swapingon);
if (trulen != trlength) {
if (subtr->next->edge->tocalcroot) {
printNewick(subtr->next->edge);
dbg_printf("\n");
printNewick(swapingon->trnodes[0]);
dbg_printf("\n");
}
dbg_printf("estimated: %i\n", trlength);
dbg_printf("true: %i\n", trulen);
dbg_printf("diff: %i\n", diff);
dbg_printf("al: %i\n\n", al);
//trlength = trulen;
//dbg_printf("report mismatch\n");
}
else {
dbg_printf("MATCH:\n");
dbg_printf("estimated: %i\n", trlength);
dbg_printf("true: %i\n", trulen);
dbg_printf("diff: %i\n", diff);
dbg_printf("al: %i\n\n", al);
}
trlength = trulen;
mfl_join_nodes(n, up);
*** END COMMENT OUT BEFORE COMMIT ***/
#endif
if (trlength < searchrec->bestinrep) {
dbg_printf("length: %i\n", trlength);
mfl_join_nodes(subtr->next->next, up);
mfl_join_nodes(subtr, n);
searchrec->foundbettertr = true;
searchrec->success = true;
swapingon->length = trlength;
searchrec->bestinrep = trlength;
if (searchrec->bestinrep < searchrec->bestlength) {
searchrec->bestlength = searchrec->bestinrep;
mfl_reinit_treebuffer(savedtrees, swapingon, &searchrec->nextinbuffer, numnodes);
}
else {
mfl_reinit_tbinrange(savedtrees, swapingon, searchrec->trbufstart, &searchrec->nextinbuffer, numnodes);
}
free(swapingon->compressedtr);
swapingon->compressedtr = mfl_compress_tree(swapingon, ntax, numnodes);
searchrec->nextinbuffer = searchrec->nextinbuffer + 1;
return;
}
searchrec->foundbettertr = false;
searchrec->success = false;
if (trlength == searchrec->bestinrep)
{
if (subtr->tocalcroot) {
down = subtr;
top = subtr->next->next;
}
else {
down = subtr->next->next;
top = subtr;
}
mfl_join_nodes(top, up);
mfl_join_nodes(down, n);
if (!mfl_compare_alltrees(swapingon, savedtrees, ntax, numnodes, &searchrec->trbufstart, searchrec->nextinbuffer))
{
if (searchrec->currentreplicate > 0) {
if (trlength == searchrec->bestlength) {
long int bstart = 0;
if (mfl_compare_alltrees(swapingon, savedtrees, ntax, numnodes, &bstart, searchrec->trbufstart)) {
mfl_reinit_tbinrange(savedtrees, swapingon, searchrec->trbufstart, &searchrec->nextinbuffer, numnodes);
searchrec->success = true;
return;
}
}
}
savedtrees[searchrec->nextinbuffer] = mfl_copytree(swapingon, ntax, numnodes);
savedtrees[searchrec->nextinbuffer]->index = searchrec->nextinbuffer;
savedtrees[searchrec->nextinbuffer]->length = trlength;
savedtrees[searchrec->nextinbuffer]->swapped = false;
searchrec->nextinbuffer = searchrec->nextinbuffer + 1;
}
trlength = 0;
}
mfl_join_nodes(n, up);
}
if (n->tip) {
return;
}
mfl_regrafting_traversal(n->next->edge, subtr, swapingon, savedtrees, ntax,
nchar, numnodes, searchrec, diff);
if (searchrec->success) {
return;
}
mfl_regrafting_traversal(n->next->next->edge, subtr, swapingon, savedtrees, ntax,
nchar, numnodes, searchrec, diff);
}