zzh
/
FastBQSR
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

修改多维数组实现方式，更有利于连续访存，修改reclatable的索引次序

This commit is contained in:
zzh 2025-12-30 01:21:13 +08:00
parent b526306e87
commit 3815a67618
7 changed files with 164 additions and 188 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
src/bqsr/bqsr_entry.cpp
View File

@ -85,9 +85,7 @@ void collapseQualityScoreTableToReadGroupTable(Array2D<RecalDatum> &byReadGroupT
// 遍历quality table // 遍历quality table
_Foreach3DK(byQualTable, qualDatum, { _Foreach3DK(byQualTable, qualDatum, {
if (qualDatum.numObservations > 0) { if (qualDatum.numObservations > 0) {
int rgKey = k1; byReadGroupTable(k1, k2).combine(qualDatum);
int eventIndex = k3;
byReadGroupTable[rgKey][eventIndex].combine(qualDatum);
} }
}); });
} }

24
src/bqsr/covariate.cpp
View File

@ -25,7 +25,7 @@ int CycleCovariate::MAXIMUM_CYCLE_VALUE;
// 对一条read计算协变量（该协变量被上一个read用过） // 对一条read计算协变量（该协变量被上一个read用过）
void CovariateUtils::ComputeCovariates(SamData& sd, sam_hdr_t* header, PerReadCovariateMatrix& values, void CovariateUtils::ComputeCovariates(SamData& sd, sam_hdr_t* header, PerReadCovariateMatrix& values,
bool recordIndelValues) { bool recordIndelValues) {
// ReadGroupCovariate::RecordValues(sd, header, values, recordIndelValues); ReadGroupCovariate::RecordValues(sd, header, values, recordIndelValues);
BaseQualityCovariate::RecordValues(sd, header, values, recordIndelValues); BaseQualityCovariate::RecordValues(sd, header, values, recordIndelValues);
ContextCovariate::RecordValues(sd, header, values, recordIndelValues); ContextCovariate::RecordValues(sd, header, values, recordIndelValues);
CycleCovariate::RecordValues(sd, header, values, recordIndelValues); CycleCovariate::RecordValues(sd, header, values, recordIndelValues);
@ -43,7 +43,7 @@ void ReadGroupCovariate::RecordValues(SamData& sd, sam_hdr_t* header, PerReadCov
key = RgToId[rgVal]; key = RgToId[rgVal];
} }
for (int i = 0; i < sd.read_len; ++i) { for (int i = 0; i < sd.read_len; ++i) {
CovariateUtils::SetCovariate(key, key, key, i, ReadGroupCovariate::index, values); CovariateUtils::SetReadGroup(key, key, key, i, values);
} }
} }
@ -51,11 +51,11 @@ void ReadGroupCovariate::RecordValues(SamData& sd, sam_hdr_t* header, PerReadCov
void BaseQualityCovariate::RecordValues(SamData& sd, sam_hdr_t* header, PerReadCovariateMatrix& values, void BaseQualityCovariate::RecordValues(SamData& sd, sam_hdr_t* header, PerReadCovariateMatrix& values,
bool recordIndelValues) { bool recordIndelValues) {
// 在前面的处理过后，quals应该和base长度一致了 // 在前面的处理过后，quals应该和base长度一致了
#define __bq_set_cov(ins, del) \ #define __bq_set_cov(ins, del) \
do { \ do { \
for (int i = 0; i < sd.read_len; ++i) { \ for (int i = 0; i < sd.read_len; ++i) { \
CovariateUtils::SetCovariate(quals[i], (ins), (del), i, BaseQualityCovariate::index, values); \ CovariateUtils::SetBaseQual(quals[i], (ins), (del), i, values); \
} \ } \
} while (0) } while (0)
const int INDEL_QUAL = 45; const int INDEL_QUAL = 45;
@ -261,7 +261,7 @@ void ContextCovariate::RecordValues(SamData& sd, sam_hdr_t* header, PerReadCovar
// don't bother zeroing out if we are going to overwrite the whole array // don't bother zeroing out if we are going to overwrite the whole array
for (int i = 0; i < originalReadLength; i++) { for (int i = 0; i < originalReadLength; i++) {
// this base has been clipped off, so zero out the covariate values here // this base has been clipped off, so zero out the covariate values here
CovariateUtils::SetCovariate(0, 0, 0, i, ContextCovariate::index, values); CovariateUtils::SetContext(0, 0, 0, i, values);
} }
} }
@ -273,12 +273,12 @@ void ContextCovariate::RecordValues(SamData& sd, sam_hdr_t* header, PerReadCovar
for (int i = 0; i < readLengthAfterClipping; i++) { for (int i = 0; i < readLengthAfterClipping; i++) {
const int readOffset = GetStrandedOffset(negativeStrand, i, readLengthAfterClipping); const int readOffset = GetStrandedOffset(negativeStrand, i, readLengthAfterClipping);
const int indelKey = indelKeys[i]; const int indelKey = indelKeys[i];
CovariateUtils::SetCovariate(nBasePairContextAtEachCycle[i], indelKey, indelKey, readOffset, ContextCovariate::index, values); CovariateUtils::SetContext(nBasePairContextAtEachCycle[i], indelKey, indelKey, readOffset, values);
} }
} else { } else {
for (int i = 0; i < readLengthAfterClipping; i++) { for (int i = 0; i < readLengthAfterClipping; i++) {
const int readOffset = GetStrandedOffset(negativeStrand, i, readLengthAfterClipping); const int readOffset = GetStrandedOffset(negativeStrand, i, readLengthAfterClipping);
CovariateUtils::SetCovariate(nBasePairContextAtEachCycle[i], 0, 0, readOffset, ContextCovariate::index, values); CovariateUtils::SetContext(nBasePairContextAtEachCycle[i], 0, 0, readOffset, values);
} }
} }
} }
@ -330,12 +330,12 @@ void CycleCovariate::RecordValues(SamData& sd, sam_hdr_t* header, PerReadCovaria
for (int i = 0; i < readLength; i++) { for (int i = 0; i < readLength; i++) {
const int substitutionKey = CycleKey(sd, i, false, MAXIMUM_CYCLE_VALUE); const int substitutionKey = CycleKey(sd, i, false, MAXIMUM_CYCLE_VALUE);
const int indelKey = CycleKey(sd, i, true, MAXIMUM_CYCLE_VALUE); const int indelKey = CycleKey(sd, i, true, MAXIMUM_CYCLE_VALUE);
CovariateUtils::SetCovariate(substitutionKey, indelKey, indelKey, i, CycleCovariate::index, values); CovariateUtils::SetCycle(substitutionKey, indelKey, indelKey, i, values);
} }
} else { } else {
for (int i = 0; i < readLength; i++) { for (int i = 0; i < readLength; i++) {
const int substitutionKey = CycleKey(sd, i, false, MAXIMUM_CYCLE_VALUE); const int substitutionKey = CycleKey(sd, i, false, MAXIMUM_CYCLE_VALUE);
CovariateUtils::SetCovariate(substitutionKey, 0, 0, i, CycleCovariate::index, values); CovariateUtils::SetCycle(substitutionKey, 0, 0, i, values);
} }
} }
} }

88
src/bqsr/covariate.h
View File

@ -25,12 +25,27 @@ using std::map;
using std::string; using std::string;
using std::vector; using std::vector;
// 协变量的值, 4个协变量
struct CovariateValues {
int readGroup = 0;
int baseQuality = 0;
int context = -1;
int cycle = -1;
void clear() {
readGroup = 0;
baseQuality = 0;
context = -1;
cycle = -1;
}
};
/** /**
* This is where we store the per-read covariates, also indexed by (event type) and (read position). * This is where we store the per-read covariates, also indexed by (event type) and (read position).
* Thus the array has shape { event type } x { read position (aka cycle) } x { covariate }. * Thus the array has shape { event type } x { read position (aka cycle) } x { covariate }.
* For instance, { covariate } is by default 4-dimensional (read group, base quality, context, cycle). * For instance, { covariate } is by default 4-dimensional (read group, base quality, context, cycle).
*/ */
typedef vector<vector<vector<int>>> PerReadCovariateMatrix; // 三维数组，第一维是event type，第二维是read position，第三维是协变量数组(结构体)(base quality, context, cycle)
typedef vector<vector<CovariateValues>> PerReadCovariateMatrix;
// 变异类型（snp, insert, deletion） // 变异类型（snp, insert, deletion）
struct EventTypeValue { struct EventTypeValue {
@ -48,33 +63,6 @@ struct EventType {
static vector<EventTypeValue> EVENTS; static vector<EventTypeValue> EVENTS;
}; };
// 协变量相关的工具类
struct CovariateUtils {
static constexpr int MAX_READ_LENGTH = 300; // 最大read长度
static constexpr int NUM_COVARIATES = 4;
// 初始化PerReadCovariateMatrix
static void InitPerReadCovMat(PerReadCovariateMatrix& matrix) {
matrix.resize(EventType::EVENT_SIZE);
for (int event_type = 0; event_type < EventType::EVENT_SIZE; ++event_type) {
matrix[event_type].resize(MAX_READ_LENGTH);
for (int pos = 0; pos < MAX_READ_LENGTH; ++pos) {
matrix[event_type][pos].resize(NUM_COVARIATES, 0);
}
}
}
// 设置协变量
static void SetCovariate(int mismatch, int insertion, int deletion, int readOffset, int covIndex, PerReadCovariateMatrix& matrix) {
matrix[EventType::BASE_SUBSTITUTION.index][readOffset][covIndex] = mismatch;
matrix[EventType::BASE_INSERTION.index][readOffset][covIndex] = insertion;
matrix[EventType::BASE_DELETION.index][readOffset][covIndex] = deletion;
}
// 对一条read计算协变量（该协变量被上一个read用过）
static void ComputeCovariates(SamData& ad, sam_hdr_t* header, PerReadCovariateMatrix& values, bool recordIndelValues);
};
// Read group协变量 // Read group协变量
struct ReadGroupCovariate { struct ReadGroupCovariate {
static constexpr int index = 0; // 在协变量数组中的索引位置 static constexpr int index = 0; // 在协变量数组中的索引位置
@ -270,8 +258,44 @@ struct CycleCovariate {
static void RecordValues(SamData& ad, sam_hdr_t* header, PerReadCovariateMatrix& values, bool recordIndelValues); static void RecordValues(SamData& ad, sam_hdr_t* header, PerReadCovariateMatrix& values, bool recordIndelValues);
}; };
// 好像不需要 // 协变量相关的工具类
struct StandardCovariateList { struct CovariateUtils {
ReadGroupCovariate readGroupCovariate; static constexpr int MAX_READ_LENGTH = 300; // 最大read长度
BaseQualityCovariate qualityScoreCovariate; static constexpr int NUM_COVARIATES = 4;
// 初始化PerReadCovariateMatrix
static void InitPerReadCovMat(PerReadCovariateMatrix& matrix) {
matrix.resize(EventType::EVENT_SIZE);
for (int event_type = 0; event_type < EventType::EVENT_SIZE; ++event_type) {
matrix[event_type].resize(MAX_READ_LENGTH);
}
}
// 设置协变量
static inline void SetReadGroup(int mismatch, int insertion, int deletion, int readOffset, PerReadCovariateMatrix& matrix) {
matrix[EventType::BASE_SUBSTITUTION.index][readOffset].readGroup = mismatch;
matrix[EventType::BASE_INSERTION.index][readOffset].readGroup = insertion;
matrix[EventType::BASE_DELETION.index][readOffset].readGroup = deletion;
}
static inline void SetBaseQual(int mismatch, int insertion, int deletion, int readOffset, PerReadCovariateMatrix& matrix) {
matrix[EventType::BASE_SUBSTITUTION.index][readOffset].baseQuality = mismatch;
matrix[EventType::BASE_INSERTION.index][readOffset].baseQuality = insertion;
matrix[EventType::BASE_DELETION.index][readOffset].baseQuality = deletion;
}
static inline void SetContext(int mismatch, int insertion, int deletion, int readOffset, PerReadCovariateMatrix& matrix) {
matrix[EventType::BASE_SUBSTITUTION.index][readOffset].context = mismatch;
matrix[EventType::BASE_INSERTION.index][readOffset].context = insertion;
matrix[EventType::BASE_DELETION.index][readOffset].context = deletion;
}
static inline void SetCycle(int mismatch, int insertion, int deletion, int readOffset, PerReadCovariateMatrix& matrix) {
matrix[EventType::BASE_SUBSTITUTION.index][readOffset].cycle = mismatch;
matrix[EventType::BASE_INSERTION.index][readOffset].cycle = insertion;
matrix[EventType::BASE_DELETION.index][readOffset].cycle = deletion;
}
// 对一条read计算协变量（该协变量被上一个read用过）
static void ComputeCovariates(SamData& ad, sam_hdr_t* header, PerReadCovariateMatrix& values, bool recordIndelValues);
}; };

157
src/bqsr/nested_array.h
View File

@ -16,129 +16,102 @@ using std::vector;
template <class T> template <class T>
struct Array2D { struct Array2D {
vector<vector<T>> data; vector<T> data;
Array2D() { } int d1 = 0, d2 = 0;
int s1 = 0, cap = 0; // strides for each dimension
Array2D() {}
Array2D(int dim1, int dim2) { init(dim1, dim2); } Array2D(int dim1, int dim2) { init(dim1, dim2); }
void init(int dim1, int dim2) { data.resize(dim1); for (auto& v : data) v.resize(dim2); } void init(int dim1, int dim2) { d1 = dim1; d2 = dim2; s1 = dim2; cap = d1 * d2; data.resize(cap); }
inline T& get(int k1, int k2) { return data[k1][k2]; } inline T& operator()(int k1, int k2) { return data[k1 * s1 + k2]; }
// 根据关键字，在对应位置插入数据
inline void put(const T& value, int k1, int k2) { data[k1][k2] = value; }
inline vector<T>& operator[](size_t idx) { return data[idx]; }
inline const vector<T>& operator[](size_t idx) const { return data[idx]; }
#define _Foreach2D(array, valName, codes) \ #define _Foreach2D(array, valName, codes) \
for (auto& arr1 : array.data) { \ for (auto& valName : array.data) { \
for (auto& valName : arr1) { \ codes; \
codes; \
} \
} }
#define _Foreach2DK(array, valName, codes) \ #define _Foreach2DK(array, valName, codes) \
do { \ do { \
int k1 = 0; \ int k1 = 0, k2 = 0, kg = 0; \
for (auto& arr1 : array.data) { \ for (auto& valName : array.data) { \
int k2 = 0; \ codes; \
for (auto& valName : arr1) { \ ++kg; \
codes; \ k1 = kg / (array.s1); \
++k2; \ k2 = kg % (array.s1); \
} \
++k1; \
} \ } \
} while (0) } while (0)
}; };
// 不能是bool
template <class T> template <class T>
struct Array3D { struct Array3D {
vector<vector<vector<T>>> data; vector<T> data;
int d1 = 0, d2 = 0, d3 = 0;
int s1 = 0, s2 = 0, cap = 0; // strides for each dimension
Array3D() {} Array3D() {}
Array3D(int dim1, int dim2, int dim3) { init(dim1, dim2, dim3); } Array3D(int dim1, int dim2, int dim3) { init(dim1, dim2, dim3); }
void init(int dim1, int dim2, int dim3) { void init(int dim1, int dim2, int dim3) {
data.resize(dim1); d1 = dim1; d2 = dim2; d3 = dim3;
for (auto& v : data) v.resize(dim2); s1 = dim2 * dim3; s2 = dim3; cap = d1 * d2 * d3;
for (auto& v1 : data) data.resize(cap);
for (auto& v2 : v1) v2.resize(dim3);
} }
inline T& get(int k1, int k2, int k3) { return data[k1][k2][k3]; } inline T& operator()(int k1, int k2, int k3) {
// 根据关键字，在对应位置插入数据 return data[k1 * s1 + k2 * s2 + k3];
inline void put(const T& value, int k1, int k2, int k3) { data[k1][k2][k3] = value; }
inline vector<vector<T>>& operator[](size_t idx) { return data[idx]; }
#define _Foreach3D(array, valName, codes) \
for (auto& arr1 : array.data) { \
for (auto& arr2 : arr1) { \
for (auto& valName : arr2) { \
codes; \
} \
} \
} }
#define _Foreach3DK(array, valName, codes) \ #define _Foreach3D(array, valName, codes) \
do { \ for (auto& valName : array.data) { \
int k1 = 0; \ codes; \
for (auto& arr1 : array.data) { \ }
int k2 = 0; \
for (auto& arr2 : arr1) { \ #define _Foreach3DK(array, valName, codes) \
int k3 = 0; \ do { \
for (auto& valName : arr2) { \ int k1 = 0, k2 = 0, k3 = 0, kg = 0; \
codes; \ for (auto& valName : array.data) { \
++k3; \ codes; \
} \ ++kg; \
++k2; \ const int mod1 = kg % (array.s1); \
} \ k1 = kg / (array.s1); \
++k1; \ k2 = mod1 / (array.s2); \
} \ k3 = mod1 % (array.s2); \
} \
} while (0) } while (0)
}; };
template <class T> template <class T>
struct Array4D { struct Array4D {
vector<vector<vector<vector<T>>>> data; vector<T> data;
int d1 = 0, d2 = 0, d3 = 0, d4;
int s1 = 0, s2 = 0, s3 = 0, cap = 0; // strides for each dimension
Array4D() {} Array4D() {}
Array4D(int dim1, int dim2, int dim3, int dim4) { init(dim1, dim2, dim3, dim4); } Array4D(int dim1, int dim2, int dim3, int dim4) { init(dim1, dim2, dim3, dim4); }
void init(int dim1, int dim2, int dim3, int dim4) { void init(int dim1, int dim2, int dim3, int dim4) {
data.resize(dim1); d1 = dim1; d2 = dim2; d3 = dim3; d4 = dim4;
for (auto& v : data) v.resize(dim2); s1 = d2 * d3 * d4; s2 = d3 * d4; s3 = d4; cap = d1 * d2 * d3 * d4;
for (auto& v1 : data) data.resize(cap);
for (auto& v2 : v1) v2.resize(dim3);
for (auto& v1 : data)
for (auto& v2 : v1)
for (auto& v3 : v2) v3.resize(dim4);
} }
inline T& get(int k1, int k2, int k3, int k4) { return data[k1][k2][k3][k4]; } inline T& operator()(int k1, int k2, int k3, int k4) { return data[k1 * s1 + k2 * s2 + k3 * s3 + k4]; }
// 根据关键字，在对应位置插入数据
inline void put(const T& value, int k1, int k2, int k3, int k4) { data[k1][k2][k3][k4] = value; } #define _Foreach4D(array, valName, codes) \
inline vector<vector<vector<T>>>& operator[](size_t idx) { return data[idx]; } for (auto& valName : array.data) { \
#define _Foreach4D(array, valName, codes) \ codes; \
for (auto& arr1 : array.data) { \
for (auto& arr2 : arr1) { \
for (auto& arr3 : arr2) { \
for (auto& valName : arr3) { \
codes; \
} \
} \
} \
} }
#define _Foreach4DK(array, valName, codes) \ #define _Foreach4DK(array, valName, codes) \
do { \ do { \
int k1 = 0; \ int k1 = 0, k2 = 0, k3 = 0, k4 = 0, kg = 0; \
for (auto& arr1 : array.data) { \ for (auto& valName : array.data) { \
int k2 = 0; \ codes; \
for (auto& arr2 : arr1) { \ ++kg; \
int k3 = 0; \ const int mod1 = kg % (array.s1); \
for (auto& arr3 : arr2) { \ const int mod2 = mod1 % (array.s2); \
int k4 = 0; \ k1 = kg / (array.s1); \
for (auto& valName : arr3) { \ k2 = mod1 / (array.s2); \
codes; \ k3 = mod2 / (array.s3); \
++k4; \ k4 = mod2 % (array.s3); \
} \ } \
++k3; \
} \
++k2; \
} \
++k1; \
} \
} while (0) } while (0)
}; };
// 类似一个tensor // 类似一个tensor，性能太低，不用了
template <class T> template <class T>
struct NestedArray { struct NestedArray {
vector<T> data; vector<T> data;

4
src/bqsr/recal_datum.h
View File

@ -97,13 +97,13 @@ struct RecalDatum {
empiricalQuality = UNINITIALIZED_EMPIRICAL_QUALITY; empiricalQuality = UNINITIALIZED_EMPIRICAL_QUALITY;
} }
void increment(const uint64_t incObservations, const double incMismatches) { inline void increment(const uint64_t incObservations, const double incMismatches) {
numObservations += incObservations; numObservations += incObservations;
numMismatches += (incMismatches * MULTIPLIER); // the multiplier used to avoid underflow, or something like that. numMismatches += (incMismatches * MULTIPLIER); // the multiplier used to avoid underflow, or something like that.
empiricalQuality = UNINITIALIZED_EMPIRICAL_QUALITY; empiricalQuality = UNINITIALIZED_EMPIRICAL_QUALITY;
} }
void increment(const uint64_t incObservations, const double incMismatches, int baseQuality) { inline void increment(const uint64_t incObservations, const double incMismatches, int baseQuality) {
numObservations += incObservations; numObservations += incObservations;
numMismatches += (incMismatches * MULTIPLIER); // the multiplier used to avoid underflow, or something like that. numMismatches += (incMismatches * MULTIPLIER); // the multiplier used to avoid underflow, or something like that.
reportedQuality = baseQuality; reportedQuality = baseQuality;

8
src/bqsr/recal_tables.h
View File

@ -34,11 +34,11 @@ struct RecalTables {
void init(int _numReadGroups) { void init(int _numReadGroups) {
numReadGroups = _numReadGroups; numReadGroups = _numReadGroups;
// 初始化readgroup和quality两个table // 初始化readgroup和quality两个table
readGroupTable.init(numReadGroups, eventDimension); readGroupTable.init(eventDimension, numReadGroups);
qualityScoreTable.init(numReadGroups, qualDimension, eventDimension); qualityScoreTable.init(eventDimension, numReadGroups, qualDimension);
// 初始化context和cycle两个table // 初始化context和cycle两个table
contextTable.init(numReadGroups, qualDimension, ContextCovariate::MaximumKeyValue() + 1, eventDimension); contextTable.init(eventDimension, numReadGroups, qualDimension, ContextCovariate::MaximumKeyValue() + 1);
cycleTable.init(numReadGroups, qualDimension, CycleCovariate::MaximumKeyValue() + 1, eventDimension); cycleTable.init(eventDimension, numReadGroups, qualDimension, CycleCovariate::MaximumKeyValue() + 1);
} }
}; };

67
src/bqsr/recal_utils.h
View File

@ -55,44 +55,25 @@ struct RecalUtils {
Array4D<RecalDatum>& cycleTable = recalTables.cycleTable; Array4D<RecalDatum>& cycleTable = recalTables.cycleTable;
int readLength = read.read_len; int readLength = read.read_len;
for (int offset = 0; offset < readLength; ++offset) { for (int idx = 0; idx < validEventTypes.size(); ++idx) {
// if (read.rid == 46114) { for (int offset = 0; offset < readLength; ++offset) {
// fprintf(gf[3], "%d %d\n", offset, info.skips[offset] ? 1 : 0); if (!info.skips[offset]) {
// }
if (!info.skips[offset]) {
// if (true){ // 不跳过当前位置
for (int idx = 0; idx < validEventTypes.size(); ++idx) {
// 获取四个值，readgroup / qualityscore / context / cycle // 获取四个值，readgroup / qualityscore / context / cycle
EventTypeValue& event = validEventTypes[idx]; EventTypeValue& event = validEventTypes[idx];
vector<int>& covariatesAtOffset = readCovars[event.index][offset]; CovariateValues& cv = readCovars[event.index][offset];
uint8_t qual = info.getQual(event, offset); uint8_t qual = info.getQual(event, offset);
double isError = info.getErrorFraction(event, offset); double isError = info.getErrorFraction(event, offset);
int readGroup = covariatesAtOffset[ReadGroupCovariate::index]; // 处理quality score covariate
int baseQuality = covariatesAtOffset[BaseQualityCovariate::index]; qualityScoreTable(event.index, cv.readGroup, cv.baseQuality).increment(1, isError, cv.baseQuality);
// 处理base quality score协变量
// RecalUtils::IncrementDatum3keys(qualityScoreTable, qual, isError, readGroup, baseQuality, event.index);
//if (read.rid == 46114) {
// fprintf(gf[3], "%d %d %f\n", offset, baseQuality, isError);
//}
qualityScoreTable[readGroup][baseQuality][event.index].increment(1, isError, baseQuality);
auto& d = qualityScoreTable[readGroup][baseQuality][event.index];
// spdlog::info("isError {} : {}, mis {}, obs {}", isError, info.snp_errs[offset], d.numMismatches, d.numObservations);
// 处理context covariate // 处理context covariate
int contextCovariate = covariatesAtOffset[ContextCovariate::index]; if (cv.context >= 0)
if (contextCovariate >= 0) contextTable(event.index, cv.readGroup, cv.baseQuality, cv.context).increment(1, isError, cv.baseQuality);
contextTable[readGroup][baseQuality][contextCovariate][event.index].increment(1, isError, baseQuality);
// RecalUtils::IncrementDatum4keys(nsgv::gRecalTables.contextTable, qual, isError, readGroup, baseQuality, contextCovariate,
// event.index);
// 处理cycle covariate // 处理cycle covariate
int cycleCovariate = covariatesAtOffset[CycleCovariate::index]; if (cv.cycle >= 0)
if (cycleCovariate >= 0) cycleTable(event.index, cv.readGroup, cv.baseQuality, cv.cycle).increment(1, isError, cv.baseQuality);
cycleTable[readGroup][baseQuality][cycleCovariate][event.index].increment(1, isError, baseQuality);
// RecalUtils::IncrementDatum4keys(nsgv::gRecalTables.cycleTable, qual, isError, readGroup, baseQuality, cycleCovariate,
// event.index);
} }
} }
} }
@ -207,8 +188,8 @@ struct RecalUtils {
RecalDatum &dat = const_cast<RecalDatum&>(datum); RecalDatum &dat = const_cast<RecalDatum&>(datum);
if (dat.getNumObservations() > 0) { if (dat.getNumObservations() > 0) {
table.addRowData({ table.addRowData({
ReadGroupCovariate::IdToRg[k1], ReadGroupCovariate::IdToRg[k2],
ReportUtil::ToString(EventType::EVENTS[k2].representation), ReportUtil::ToString(EventType::EVENTS[k1].representation),
ReportUtil::ToString(dat.getEmpiricalQuality(), 4), ReportUtil::ToString(dat.getEmpiricalQuality(), 4),
ReportUtil::ToString(dat.getReportedQuality(), 4), ReportUtil::ToString(dat.getReportedQuality(), 4),
ReportUtil::ToString(dat.getNumObservations()), ReportUtil::ToString(dat.getNumObservations()),
@ -233,9 +214,9 @@ struct RecalUtils {
RecalDatum &dat = const_cast<RecalDatum&>(datum); RecalDatum &dat = const_cast<RecalDatum&>(datum);
if (dat.getNumObservations() > 0) { if (dat.getNumObservations() > 0) {
table.addRowData({ table.addRowData({
ReadGroupCovariate::IdToRg[k1], ReadGroupCovariate::IdToRg[k2],
ReportUtil::ToString(k2), ReportUtil::ToString(k3),
ReportUtil::ToString(EventType::EVENTS[k3].representation), ReportUtil::ToString(EventType::EVENTS[k1].representation),
ReportUtil::ToString(dat.getEmpiricalQuality(), 4), ReportUtil::ToString(dat.getEmpiricalQuality(), 4),
ReportUtil::ToString(dat.getNumObservations()), ReportUtil::ToString(dat.getNumObservations()),
ReportUtil::ToString(dat.getNumMismatches(), 2) ReportUtil::ToString(dat.getNumMismatches(), 2)
@ -260,11 +241,11 @@ struct RecalUtils {
RecalDatum &dat = const_cast<RecalDatum&>(datum); RecalDatum &dat = const_cast<RecalDatum&>(datum);
if (dat.getNumObservations() > 0) { if (dat.getNumObservations() > 0) {
table.addRowData({ table.addRowData({
ReadGroupCovariate::IdToRg[k1], ReadGroupCovariate::IdToRg[k2],
ReportUtil::ToString(k2), ReportUtil::ToString(k3),
ReportUtil::ToString(ContextCovariate::ContextFromKey(k3)), ReportUtil::ToString(ContextCovariate::ContextFromKey(k4)),
"Context", "Context",
ReportUtil::ToString(EventType::EVENTS[k4].representation), ReportUtil::ToString(EventType::EVENTS[k1].representation),
ReportUtil::ToString(dat.getEmpiricalQuality(), 4), ReportUtil::ToString(dat.getEmpiricalQuality(), 4),
ReportUtil::ToString(dat.getNumObservations()), ReportUtil::ToString(dat.getNumObservations()),
ReportUtil::ToString(dat.getNumMismatches(), 2) ReportUtil::ToString(dat.getNumMismatches(), 2)
@ -276,11 +257,11 @@ struct RecalUtils {
RecalDatum &dat = const_cast<RecalDatum&>(datum); RecalDatum &dat = const_cast<RecalDatum&>(datum);
if (dat.getNumObservations() > 0) { if (dat.getNumObservations() > 0) {
table.addRowData({ table.addRowData({
ReadGroupCovariate::IdToRg[k1], ReadGroupCovariate::IdToRg[k2],
ReportUtil::ToString(k2), ReportUtil::ToString(k3),
ReportUtil::ToString(CycleCovariate::CycleFromKey(k3)), ReportUtil::ToString(CycleCovariate::CycleFromKey(k4)),
"Cycle", "Cycle",
ReportUtil::ToString(EventType::EVENTS[k4].representation), ReportUtil::ToString(EventType::EVENTS[k1].representation),
ReportUtil::ToString(dat.getEmpiricalQuality(), 4), ReportUtil::ToString(dat.getEmpiricalQuality(), 4),
ReportUtil::ToString(dat.getNumObservations()), ReportUtil::ToString(dat.getNumObservations()),
ReportUtil::ToString(dat.getNumMismatches(), 2) ReportUtil::ToString(dat.getNumMismatches(), 2)