SEXP extract_pair_data(SEXP bam, SEXP index, SEXP chr, SEXP start, SEXP end, SEXP mapq, SEXP dedup, SEXP diagnostics) try {
// Checking input values.
if (!isInteger(mapq) || LENGTH(mapq)!=1) {
throw std::runtime_error("mapping quality should be an integer scalar");
}
const int minqual=asInteger(mapq);
if (!isLogical(dedup) || LENGTH(dedup)!=1) {
throw std::runtime_error("duplicate removal should be a logical scalar");
}
const bool rmdup=asLogical(dedup);
if (!isLogical(diagnostics) || LENGTH(diagnostics)!=1) {
throw std::runtime_error("diagnostics specification should be a logical scalar");
}
const bool getnames=asLogical(diagnostics);
// Initializing odds and ends.
BamFile bf(bam, index);
BamRead br;
BamIterator biter(bf, chr, start, end);
OutputContainer oc(getnames);
typedef std::map<std::pair<int, std::string>, AlignData> Holder;
std::deque<Holder> all_holders(4); // four holders, one for each strand/first combination; cut down searches.
std::pair<int, std::string> current;
Holder::iterator ith;
int curpos, mate_pos;
AlignData algn_data;
bool am_mapped, is_first;
bool mate_is_in;
std::set<std::string> identical_pos;
std::set<std::string>::iterator itip;
int last_identipos=-1;
while (bam_itr_next(bf.in, biter.iter, br.read) >= 0){
++oc.totals;
curpos = (br.read->core).pos + 1; // Getting 1-indexed position.
br.extract_data(algn_data);
am_mapped=br.is_well_mapped(minqual, rmdup);
/* Reasons to not add a read: */
// // If we can see that it is obviously unmapped (IMPOSSIBLE for a sorted file).
// if (((br.read -> core).flag & BAM_FUNMAP)!=0) {
// // We don't filter by additional mapping criteria, as we need to search 'holder' to pop out the partner and to store diagnostics.
// continue;
// }
// If it's a singleton.
if (((br.read -> core).flag & BAM_FPAIRED)==0) {
if (am_mapped) { oc.add_single(curpos, algn_data); }
continue;
}
// Or, if we can see that its partner is obviously unmapped.
if (((br.read -> core).flag & BAM_FMUNMAP)!=0) {
if (am_mapped) { oc.add_onemapped(curpos, algn_data); }
continue;
}
// Or if it's inter-chromosomal.
is_first=(((br.read->core).flag & BAM_FREAD1)!=0);
if (is_first==(((br.read->core).flag & BAM_FREAD2)!=0)) {
std::stringstream err;
err << "read '" << bam_get_qname(br.read) << "' must be either first or second in the pair";
throw std::runtime_error(err.str());
}
if ((br.read -> core).mtid!=(br.read -> core).tid) {
if (am_mapped) { oc.add_interchr(curpos, algn_data, bam_get_qname(br.read), is_first); }
continue;
}
/* Checking the map and adding it if it doesn't exist. */
current.second.assign(bam_get_qname(br.read));
mate_pos = (br.read -> core).mpos + 1; // 1-indexed position, again.
mate_is_in=false;
if (mate_pos < curpos) {
mate_is_in=true;
} else if (mate_pos == curpos) {
// Identical mpos to curpos needs careful handling to figure out whether we've already seen it.
if (curpos!=last_identipos) {
identical_pos.clear();
last_identipos=curpos;
}
itip=identical_pos.lower_bound(current.second);
if (itip!=identical_pos.end() && !(identical_pos.key_comp()(current.second, *itip))) {
mate_is_in=true;
identical_pos.erase(itip);
} else {
identical_pos.insert(itip, current.second);
}
}
if (mate_is_in) {
current.first = mate_pos;
Holder& holder=all_holders[int(!is_first) + 2*int(bam_is_mrev(br.read))];
ith=holder.find(current);
if (ith != holder.end()) {
if (!am_mapped) {
// Searching to pop out the mate, to reduce the size of 'holder' for the remaining searches (and to store diagnostics).
oc.add_onemapped((ith->first).first, ith->second);
holder.erase(ith);
continue;
}
oc.add_genuine(curpos, algn_data, (ith->first).first, ith->second, is_first);
holder.erase(ith);
} else if (am_mapped) {
// Only possible if the mate didn't get added because 'am_mapped' was false.
oc.add_onemapped(curpos, algn_data);
}
} else if (am_mapped) {
current.first = curpos;
Holder& holder=all_holders[int(is_first) + 2*int(algn_data.is_reverse)];
holder[current] = algn_data;
}
}
// Leftovers treated as one_unmapped; marked as paired, but the mate is not in file.
for (size_t h=0; h<all_holders.size(); ++h) {
Holder& holder=all_holders[h];
for (ith=holder.begin(); ith!=holder.end(); ++ith) {
oc.add_onemapped((ith->first).first, ith->second);
}
holder.clear();
}
// Storing all output.
SEXP output=PROTECT(allocVector(VECSXP, getnames ? 9 : 2));
try {
SET_VECTOR_ELT(output, 0, allocVector(VECSXP, 2));
SEXP left=VECTOR_ELT(output, 0);
store_int_output(left, 0, oc.forward_pos_out);
store_int_output(left, 1, oc.forward_len_out);
SET_VECTOR_ELT(output, 1, allocVector(VECSXP, 2));
SEXP right=VECTOR_ELT(output, 1);
store_int_output(right, 0, oc.reverse_pos_out);
store_int_output(right, 1, oc.reverse_len_out);
if (getnames) {
SET_VECTOR_ELT(output, 2, ScalarInteger(oc.totals));
SET_VECTOR_ELT(output, 3, allocVector(VECSXP, 2));
SEXP singles=VECTOR_ELT(output, 3);
store_int_output(singles, 0, oc.single_pos);
store_int_output(singles, 1, oc.single_len);
SET_VECTOR_ELT(output, 4, allocVector(VECSXP, 2));
SEXP first=VECTOR_ELT(output, 4);
store_int_output(first, 0, oc.ufirst_pos);
store_int_output(first, 1, oc.ufirst_len);
SET_VECTOR_ELT(output, 5, allocVector(VECSXP, 2));
SEXP second=VECTOR_ELT(output, 5);
store_int_output(second, 0, oc.usecond_pos);
store_int_output(second, 1, oc.usecond_len);
SET_VECTOR_ELT(output, 6, allocVector(VECSXP, 2));
SEXP onemap=VECTOR_ELT(output, 6);
store_int_output(onemap, 0, oc.onemap_pos);
store_int_output(onemap, 1, oc.onemap_len);
SET_VECTOR_ELT(output, 7, allocVector(VECSXP, 3));
SEXP interchr1=VECTOR_ELT(output, 7);
store_int_output(interchr1, 0, oc.ifirst_pos);
store_int_output(interchr1, 1, oc.ifirst_len);
store_names(interchr1, 2, oc.interchr_names_1);
SET_VECTOR_ELT(output, 8, allocVector(VECSXP, 3));
SEXP interchr2=VECTOR_ELT(output, 8);
store_int_output(interchr2, 0, oc.isecond_pos);
store_int_output(interchr2, 1, oc.isecond_len);
store_names(interchr2, 2, oc.interchr_names_2);
}
} catch (std::exception &e) {
UNPROTECT(1);
throw;
}
UNPROTECT(1);
return output;
} catch (std::exception &e) {
return mkString(e.what());
}
static int make_pair_key(key_data_t *key, bam1_t *bam) {
int32_t this_ref, this_coord, this_end;
int32_t other_ref, other_coord, other_end;
int32_t orientation, leftmost;
uint8_t *data;
char *cig;
this_ref = bam->core.tid + 1; // avoid a 0 being put into the hash
other_ref = bam->core.mtid + 1;
this_coord = unclipped_start(bam);
this_end = unclipped_end(bam);
if ((data = bam_aux_get(bam, "MC"))) {
cig = bam_aux2Z(data);
other_end = unclipped_other_end(bam->core.mpos, cig);
other_coord = unclipped_other_start(bam->core.mpos, cig);
} else {
fprintf(stderr, "[markdup] error: no MC tag.\n");
return 1;
}
// work out orientations
if (this_ref != other_ref) {
leftmost = this_ref < other_ref;
} else {
if (bam_is_rev(bam) == bam_is_mrev(bam)) {
if (!bam_is_rev(bam)) {
leftmost = this_coord <= other_coord;
} else {
leftmost = this_end <= other_end;
}
} else {
if (bam_is_rev(bam)) {
leftmost = this_end <= other_coord;
} else {
leftmost = this_coord <= other_end;
}
}
}
// pair orientation
if (leftmost) {
if (bam_is_rev(bam) == bam_is_mrev(bam)) {
other_coord = other_end;
if (!bam_is_rev(bam)) {
if (bam->core.flag & BAM_FREAD1) {
orientation = O_FF;
} else {
orientation = O_RR;
}
} else {
if (bam->core.flag & BAM_FREAD1) {
orientation = O_RR;
} else {
orientation = O_FF;
}
}
} else {
if (!bam_is_rev(bam)) {
orientation = O_FR;
other_coord = other_end;
} else {
orientation = O_RF;
this_coord = this_end;
}
}
} else {
if (bam_is_rev(bam) == bam_is_mrev(bam)) {
this_coord = this_end;
if (!bam_is_rev(bam)) {
if (bam->core.flag & BAM_FREAD1) {
orientation = O_RR;
} else {
orientation = O_FF;
}
} else {
if (bam->core.flag & BAM_FREAD1) {
orientation = O_FF;
} else {
orientation = O_RR;
}
}
} else {
if (!bam_is_rev(bam)) {
orientation = O_RF;
other_coord = other_end;
} else {
orientation = O_FR;
this_coord = this_end;
}
}
}
if (!leftmost)
leftmost = 13;
else
leftmost = 11;
key->single = 0;
key->this_ref = this_ref;
key->this_coord = this_coord;
key->other_ref = other_ref;
key->other_coord = other_coord;
key->leftmost = leftmost;
key->orientation = orientation;
return 0;
}
