void SplitFasta(const char* seqDataFile, const char *outpath, const char *ext)/*{{{*/
{
char seqFile[MAX_PATH+1] = "";
//char command[MAX_COMMAND_LINE+1] = "";
FILE *fpSeqData;
FILE *fpSeq;
fpSeqData = fopen(seqDataFile,"r");
checkfilestream(fpSeqData, seqDataFile, "r");
char id[SIZE_ID+1] = "";
char str[SIZE_ID+10+1] = "";
char rtname[MAX_PATH+1] = "";
rootname(seqDataFile,rtname);
int linesize;
Array1D <char> line_1darray(maxline+1);
char *line = line_1darray.array1D;
char delim[] = " \t\r\n>";
char *pch = NULL;
fpos_t pos;
//int i;
int cnt = 0 ;
while((linesize = fgetline(fpSeqData,line,maxline)) != EOF)
{
if(linesize == 0 || IsBlankLine(line)) continue;
if(line[0] == '>')//beginning of a record
{
cnt ++ ;
my_strcpy(str, line, SIZE_ID+10);
pch = strtok(str,delim);
while(pch != NULL) //take the first word following '>' as id
{
my_strcpy(id, pch, SIZE_ID);
break;
}
if(strcmp(id,"") == 0)// if no word following '>', set the id as rootname(seq-file)_cnt
{
sprintf(id,"%s_%d", rtname, cnt);
}
printf("%d\t:%s exported\n",cnt,id);
sprintf(seqFile,"%s/%s.%s",outpath,id,ext);
fpSeq = fopen(seqFile,"w");
checkfilestream(fpSeq, seqFile, "r");
fprintf(fpSeq,"%s\n",line);
while(!feof(fpSeqData))
{
fgetpos(fpSeqData,&pos);
linesize = fgetline(fpSeqData,line,maxline);
if(linesize == 0 || IsBlankLine(line)) continue;
if(line[0] == '>')
{
fsetpos(fpSeqData,&pos);
break;
}
else
fprintf(fpSeq,"%s\n",line);
}
fclose(fpSeq);
}
}
fclose(fpSeqData);
printf("%d sequences exported!\n",cnt);
}
/** Read examples into memory
*/
void CSeqFeature::LoadFeatures()
{
unsigned int tmpFidx = 0;
Scalar tmpFval = 0;
unsigned int featureCnt = 0;
unsigned int seqNum = 0;
unsigned int phiNum = 0;
unsigned int posNum1 = 0, posNum2 = 0;
std::string line = "";
std::string token = "";
std::ifstream featureFp;
featureFp.open(featureFile.c_str());
if(!featureFp.good())
{
string msg = "Cannot open feature file <" + featureFile + ">!";
throw CBMRMException(msg, "CSeqFeature::ScanFeatureFile()");
}
// read header information
int headerInfoCnt = 3; // min duration, max duration, feature dimension
do {
getline(featureFp, line);
trim(line);
if(IsBlankLine(line)) continue; // blank line
if(line[0] == '#') continue; // comment line
if(sscanf(line.c_str(),"maxDuration:%d",&maxDuration)==1) headerInfoCnt--;
if(sscanf(line.c_str(),"minDuration:%d",&minDuration)==1) headerInfoCnt--;
if(sscanf(line.c_str(),"globalFeatureDim:%d",&featureDimension)==1) headerInfoCnt--;
} while(!featureFp.eof() && (headerInfoCnt != 0));
assert(maxDuration >= minDuration);
assert(featureDimension < (1<<30)); // featureDimension is normally less then 1 billion
if(featureFp.eof())
throw CBMRMException("Feature file does not contain valid examples","CSeqFeature::LoadFeatures()");
// read sequences
nnz = 0;
while(!featureFp.eof())
{
// read sequence number
do {
getline(featureFp, line);
trim(line);
if(IsBlankLine(line)) continue; // blank line
if(line[0] == '#') continue; // comment line
if(sscanf(line.c_str(),"sequence:%d",&seqNum)==1) break;
} while(!featureFp.eof());
if(featureFp.eof())
throw CBMRMException("Feature file does not contain valid phi:*","CSeqFeature::LoadFeatures()");
// read phi:1 tag
phiNum = 0;
do {
getline(featureFp, line);
trim(line);
if(IsBlankLine(line)) continue; // blank line
if(line[0] == '#') continue; // comment line
if(sscanf(line.c_str(),"phi:%d",&phiNum)==1) break;
} while(!featureFp.eof());
if(featureFp.eof() || (phiNum != 1))
throw CBMRMException("Feature file does not contain valid phi:1","CSeqFeature::LoadFeatures()");
// read phi:1 sparse vectors
do {
getline(featureFp, line);
trim(line);
if(IsBlankLine(line)) continue; // blank line
if(line[0] == '#') continue; // comment line
if(sscanf(line.c_str(),"phi:%d",&phiNum) == 1)
break;
istringstream iss(line);
iss >> token;
if((sscanf(token.c_str(),"pos:%d",&posNum1) != 1))
throw CBMRMException("Feature file does not contain valid pos tag in phi:1","CSeqFeature::LoadFeatures()");
TheMatrix svec(1,featureDimension,SML::SPARSE);
featureCnt = 0;
while(!iss.eof())
{
iss >> token;
if(sscanf(token.c_str(),svec_feature_index_and_value_format.c_str(),&tmpFidx, &tmpFval) != 2)
{
ostringstream msg;
msg << "Invalid #" << featureCnt + 1 << " sparse vector element in phi:"<< phiNum << " seq:" << seqNum << " pos:" << posNum1;
throw CBMRMException(msg.str(),"CSeqFeature::LoadFeatures()");
}
svec.Set(0,tmpFidx,tmpFval);
nnz++;
}
if(featureCnt == 0)
throw CBMRMException("Feature file does not contain valid phi:2 sparse vector","CSeqFeature::LoadFeatures()");
phi_1.push_back(svec);
} while(!featureFp.eof());
if(phi_1.size() < 1)
throw CBMRMException("Feature file does not contain valid phi:1","CSeqFeature::LoadFeatures()");
numOfSeq = phi_1.size();
if(featureFp.eof() || (phiNum != 2))
throw CBMRMException("Feature file does not contain valid phi:2","CSeqFeature::LoadFeatures()");
// read phi:2 sparse vectors
unsigned int prevPosNum1 = 0, prevPosNum2 = 0;
vector<TheMatrix> tmp_phi_2_svecs;
featureCnt = 0;
do {
getline(featureFp, line);
trim(line);
if(IsBlankLine(line)) continue; // blank line
if(line[0] == '#') continue; // comment line
if((sscanf(line.c_str(),"phi:%d",&phiNum) == 1))
break;
istringstream iss(line);
iss >> token;
if((sscanf(token.c_str(),"pos:%d,%d",&posNum1,&posNum2) != 2))
throw CBMRMException("Feature file does not containt valid pos tag in phi:2","CSeqFeature::LoadFeatures()");
if(prevPosNum2 >= posNum2)
{
ostringstream msg;
msg << "previous posNum2 must be > current posNum2 in phi:2 (phi:2 pos:" << posNum1 << "," << posNum2;
throw CBMRMException(msg.str(),"CSeqFeature::LoadFeatures()");
}
if(prevPosNum1 >= posNum1)
{
ostringstream msg;
msg << "previous posNum1 must be > current posNum1 in phi:2 (phi:2 pos:" << posNum1 << "," << posNum2;
throw CBMRMException(msg.str(),"CSeqFeature::LoadFeatures()");
}
if(posNum1 != prevPosNum1)
{
phi_2.push_back(tmp_phi_2_svecs);
tmp_phi_2_svecs.clear();
}
TheMatrix svec(1,featureDimension,SML::SPARSE);
featureCnt = 0;
while(!iss.eof())
{
iss >> token;
if(sscanf(token.c_str(),svec_feature_index_and_value_format.c_str(),&tmpFidx, &tmpFval) != 2)
{
ostringstream msg;
msg << "Invalid #" << featureCnt + 1 << " sparse vector element in phi:"<< phiNum << " seq:" << seqNum << " pos:" << posNum1;
throw CBMRMException(msg.str(),"CSeqFeature::LoadFeatures()");
}
svec.Set(0,tmpFidx,tmpFval);
nnz++;
}
if(featureCnt == 0)
throw CBMRMException("Feature file does not containt valid phi:2 sparse vector","CSeqFeature::LoadFeatures()");
tmp_phi_2_svecs.push_back(svec);
} while(!featureFp.eof());
if(phi_2.size() < 1)
throw CBMRMException("Feature file does not contain phi:2","CSeqFeature::LoadFeatures()");
}
// data matrix density
density = ((double)nnz/featureDimension)/numOfSeq;
featureFp.close();
}