/**
* It reads the information in the header of the file. It reads relation's name, attributes'
* names, and inputs and outputs.
*
* @param parser is the parser of the data set
* @param isTrain is a boolean indicating if this is a train set (and so parameters information
* must be read) or a test set (parameters information has not to be read).
*/
public void parseHeader(InstanceParser parser, boolean isTrain) {
// 1. Declaration of variables
Vector inputAttrNames = new Vector();
Vector outputAttrNames = new Vector();
boolean inputsDef = false;
boolean outputsDef = false;
String line, aux;
header = "";
int attCount = 0, lineCount = 0;
attHeader = null;
while (!(line = parser.getLine().trim()).equalsIgnoreCase("@data")) {
line = line.trim();
// System.out.println (" > Line read: " + line +"." );
lineCount++;
if (line.toLowerCase().indexOf("@relation") != -1) {
if (isTrain) Attributes.setRelationName(line.replaceAll("@relation", ""));
}
if (line.toLowerCase().indexOf("@attribute") != -1) {
if (isTrain) insertAttribute(line);
attCount++;
}
if (line.toLowerCase().indexOf("@inputs") != -1) {
attHeader = header;
inputsDef = true;
aux = line.substring(8);
if (isTrain) insertInputOutput(aux, lineCount, inputAttrNames, "inputs", isTrain);
}
if (line.toLowerCase().indexOf("@outputs") != -1) {
if (attHeader == null) attHeader = header;
outputsDef = true;
// System.out.println ( " >>> Defining the output !!!");
aux = line.substring(8);
if (isTrain) insertInputOutput(aux, lineCount, outputAttrNames, "outputs", isTrain);
System.out.println(" >> Size of the output is: " + outputAttrNames.size());
}
header += line + "\n";
}
if (attHeader == null) attHeader = header;
processInputsAndOutputs(isTrain, inputsDef, outputsDef, outputAttrNames, inputAttrNames);
} // end headerParse
/**
* This method reads all the information in a DB and load it to memory.
*
* @param fileName is the database file name.
* @param isTrain is a flag that indicate if the database is for a train or for a test.
* @throws DatasetException if there is any semantical error in the input file.
* @throws HeaderFormatException if there is any lexical or sintactical error in the header of the
* input file
*/
public void readSet(String fileName, boolean isTrain)
throws DatasetException, HeaderFormatException {
String line;
System.out.println("Opening the file: " + fileName + ".");
// Parsing the header of the DB.
errorLogger = new FormatErrorKeeper();
// Declaring an instance parser
InstanceParser parser = new InstanceParser(fileName, isTrain);
// Reading information in the header, i.e., @relation, @attribute, @inputs and @outputs
parseHeader(parser, isTrain);
System.out.println(
" The number of output attributes is: " + Attributes.getOutputNumAttributes());
// The attributes statistics are init if we are in train mode.
if (isTrain && Attributes.getOutputNumAttributes() == 1) {
Attributes.initStatistics();
}
// A temporal vector is used to store the instances read.
System.out.println("\n\n > Reading the data ");
Vector tempSet = new Vector(1000, 100000);
while ((line = parser.getLine()) != null) {
// System.out.println (" > Data line: " + line );
tempSet.addElement(new Instance(line, isTrain, tempSet.size()));
}
// The vector of instances is converted to an array of instances.
int sizeInstance = tempSet.size();
System.out.println(" > Number of instances read: " + tempSet.size());
instanceSet = new Instance[sizeInstance];
for (int i = 0; i < sizeInstance; i++) {
instanceSet[i] = (Instance) tempSet.elementAt(i);
}
// System.out.println("After converting all instances");
// System.out.println("The error logger has any error: "+errorLogger.getNumErrors());
if (errorLogger.getNumErrors() > 0) {
System.out.println(
"There has been " + errorLogger.getAllErrors().size() + " errors in the Dataset format.");
for (int k = 0; k < errorLogger.getNumErrors(); k++) {
errorLogger.getError(k).print();
}
throw new DatasetException(
"There has been " + errorLogger.getAllErrors().size() + " errors in the Dataset format",
errorLogger.getAllErrors());
}
System.out.println(
"\n > Finishing the statistics: (isTrain)"
+ isTrain
+ ", (# out attributes)"
+ Attributes.getOutputNumAttributes());
// If being on a train dataset, the statistics are finished
if (isTrain && Attributes.getOutputNumAttributes() == 1) {
Attributes.finishStatistics();
}
System.out.println(" >> File LOADED CORRECTLY!!");
} // end of InstanceSet constructor.