void CSVImporter::initColumn(Column &column, const string &name, const vector<string> &cells)
{
// we treat single spaces as missing values, because SPSS saves missing values as a single space in CSV files
column.setName(name);
// try to make the column nominal
bool success = true;
set<int> uniqueValues;
Column::Ints::iterator intInputItr = column.AsInts.begin();
Labels &labels = column.labels();
labels.clear();
BOOST_FOREACH(const string &value, cells)
{
if (value != "NaN" && value != "nan" && value != "" && value != " ")
{
try
{
int v = lexical_cast<int>(value);
uniqueValues.insert(v);
*intInputItr = v;
}
catch (...)
{
// column can't be made nominal numeric
success = false;
break;
}
}
else
{
*intInputItr = INT_MIN;
}
intInputItr++;
}
if (success && uniqueValues.size() <= 24)
{
labels.clear();
BOOST_FOREACH(int value, uniqueValues)
{
(void)uniqueValues;
labels.add(value);
}
void SPSSImporter::setColumnLabeledData(Column &column, size_t numCases, const SPSSColumn &spssCol)
{
column.labels().clear();
// Add lables from the SPSS file first.
map<double, string> lbs;
for (SPSSColumn::LabelByValueDict::const_iterator it = spssCol.spssLables.begin();
it != spssCol.spssLables.end(); ++it)
lbs.insert( pair<double, string>(it->first.dbl, it->second) );
// Add labels for numeric values (if not already present)..
vector<bool> isMissing(numCases);
for (size_t i = 0; i < numCases; ++i)
{
if (spssCol.missingChecker().isMissingValue(_floatInfo, spssCol.numerics[i]) == false)
{
isMissing[i] = false;
if (lbs.find(spssCol.numerics[i]) == lbs.end())
lbs.insert( pair<double, string>( spssCol.numerics[i], spssCol.format(spssCol.numerics[i], _floatInfo)) );
}
else
isMissing[i] = true;
}
// Extract the data were are going to use.
vector<int> dataToInsert;
map<int, string> labels;
// We cannot insert doubles as data valuesm and get labels
// for them to work (JASP limitation).
if (spssCol.containsFraction())
{
// Generate an index value for each data point.
for (size_t i = 0; i < numCases; ++i)
{
// Find insert the index as a data point, if not missing value.
if (isMissing[i])
dataToInsert.push_back(INT_MIN);
else
{
map<double, string>::iterator fltLabeI = lbs.find(spssCol.numerics[i]);
dataToInsert.push_back( distance(lbs.begin(), fltLabeI) );
// Pair the inserted value with a lable string.
labels.insert(pair<int, string>(dataToInsert.back(), fltLabeI->second));
}
}
}
else
{
// Use the raw data as the index to labels.
for (size_t i = 0; i < numCases; ++i)
{
// insert the (rounded) value as the data point.
if (isMissing[i])
dataToInsert.push_back(INT_MIN);
else
{
dataToInsert.push_back( static_cast<int>(spssCol.numerics[i]) );
map<double, string>::iterator fltLabeI = lbs.find(spssCol.numerics[i]);
// pair the inserted value with a lable string.
labels.insert(pair<int, string>( static_cast<int>(fltLabeI->first), fltLabeI->second));
}
}
}
// Insert the labels into the JASP data set.
for (map<int, string>::const_iterator it = labels.begin(); it != labels.end(); ++it)
column.labels().add(it->first, it->second);
// Insert the data into the data set.
Column::Ints::iterator intInputItr = column.AsInts.begin();
for (size_t i = 0; i < dataToInsert.size(); ++i, ++intInputItr)
*intInputItr = dataToInsert[i];
}