weka.knowledgeflow.steps.Sorter Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of weka-stable Show documentation
Show all versions of weka-stable Show documentation
The Waikato Environment for Knowledge Analysis (WEKA), a machine
learning workbench. This is the stable version. Apart from bugfixes, this version
does not receive any other updates.
/*
* This program is free software: you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation, either version 3 of the License, or
* (at your option) any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program. If not, see m_incrementalBuffer;
/** List of sorted temp files for incremental operation */
protected transient List m_bufferFiles;
/** Size of the in-memory buffer */
protected String m_bufferSize = "10000";
/** Size of the in-memory buffer after resolving any environment vars */
protected int m_bufferSizeI = 10000;
/** Holds indexes of string attributes, keyed by attribute name */
protected Map m_stringAttIndexes;
/** Holds the internal textual description of the sort definitions */
protected String m_sortDetails;
/**
* The directory to hold the temp files - if not set the system tmp directory
* is used
*/
protected File m_tempDirectory = new File("");
/** format of instances for current incoming connection (if any) */
protected Instances m_connectedFormat;
/** True if we've been reset */
protected boolean m_isReset;
/** True if processing streaming data */
protected boolean m_streaming;
/** To (re)use when streaming */
protected Data m_streamingData;
/**
* Get the size of the in-memory buffer
*
* @return the size of the in-memory buffer
*/
public String getBufferSize() {
return m_bufferSize;
}
/**
* Set the size of the in-memory buffer
*
* @param buffSize the size of the in-memory buffer
*/
@OptionMetadata(displayName = "Size of in-mem streaming buffer",
description = "Number of instances to sort in memory before writing to a "
+ "temp file (instance connections only)", displayOrder = 1)
public void setBufferSize(String buffSize) {
m_bufferSize = buffSize;
}
/**
* Set the directory to use for temporary files during incremental operation
*
* @param tempDir the temp dir to use
*/
@FilePropertyMetadata(fileChooserDialogType = KFGUIConsts.OPEN_DIALOG,
directoriesOnly = true)
@OptionMetadata(displayName = "Directory for temp files",
description = "Where to store temporary files when spilling to disk",
displayOrder = 2)
public void setTempDirectory(File tempDir) {
m_tempDirectory = tempDir;
}
/**
* Get the directory to use for temporary files during incremental operation
*
* @return the temp dir to use
*/
public File getTempDirectory() {
return m_tempDirectory;
}
/**
* Set the sort rules to use
*
* @param sortDetails the sort rules in internal string representation
*/
@ProgrammaticProperty
public void setSortDetails(String sortDetails) {
m_sortDetails = sortDetails;
}
/**
* Get the sort rules to use
*
* @return the sort rules in internal string representation
*/
public String getSortDetails() {
return m_sortDetails;
}
/**
* Initialize the step.
*
* @throws WekaException if a problem occurs during initialization
*/
@Override
public void stepInit() throws WekaException {
m_isReset = true;
m_streaming = false;
m_stringAttIndexes = new HashMap();
m_bufferFiles = new ArrayList();
m_streamingData = new Data(StepManager.CON_INSTANCE);
}
/**
* Get a list of incoming connection types that this step can accept. Ideally
* (and if appropriate), this should take into account the state of the step
* and any existing incoming connections. E.g. a step might be able to accept
* one (and only one) incoming batch data connection.
*
* @return a list of incoming connections that this step can accept given its
* current state
*/
@Override
public List getIncomingConnectionTypes() {
if (getStepManager().numIncomingConnections() == 0) {
return Arrays.asList(StepManager.CON_INSTANCE, StepManager.CON_DATASET,
StepManager.CON_TRAININGSET, StepManager.CON_TESTSET);
}
return null;
}
/**
* Get a list of outgoing connection types that this step can produce. Ideally
* (and if appropriate), this should take into account the state of the step
* and the incoming connections. E.g. depending on what incoming connection is
* present, a step might be able to produce a trainingSet output, a testSet
* output or neither, but not both.
*
* @return a list of outgoing connections that this step can produce
*/
@Override
public List getOutgoingConnectionTypes() {
List result = new ArrayList();
if (getStepManager().numIncomingConnectionsOfType(StepManager.CON_INSTANCE) > 0) {
result.add(StepManager.CON_INSTANCE);
}
if (getStepManager().numIncomingConnectionsOfType(StepManager.CON_DATASET) > 0) {
result.add(StepManager.CON_DATASET);
}
if (getStepManager().numIncomingConnectionsOfType(
StepManager.CON_TRAININGSET) > 0) {
result.add(StepManager.CON_TRAININGSET);
}
if (getStepManager().numIncomingConnectionsOfType(StepManager.CON_TESTSET) > 0) {
result.add(StepManager.CON_TESTSET);
}
return result;
}
/**
* Initialize given the supplied instances structure
*
* @param structure the structure to initialize with
*/
protected void init(Instances structure) {
m_connectedFormat = structure;
List sortRules = new ArrayList();
if (m_sortDetails != null && m_sortDetails.length() > 0) {
String[] sortParts = m_sortDetails.split("@@sort-rule@@");
for (String s : sortParts) {
SortRule r = new SortRule(s.trim());
r.init(getStepManager().getExecutionEnvironment()
.getEnvironmentVariables(), structure);
sortRules.add(r);
}
m_sortComparator = new SortComparator(sortRules);
}
// check for string attributes
m_stringAttIndexes = new HashMap();
for (int i = 0; i < structure.numAttributes(); i++) {
if (structure.attribute(i).isString()) {
m_stringAttIndexes.put(structure.attribute(i).name(), new Integer(i));
}
}
if (m_stringAttIndexes.size() == 0) {
m_stringAttIndexes = null;
}
if (m_streaming) {
String buffSize = environmentSubstitute(m_bufferSize);
m_bufferSizeI = Integer.parseInt(buffSize);
m_incrementalBuffer = new ArrayList(m_bufferSizeI);
}
}
/**
* Process an incoming data payload (if the step accepts incoming connections)
*
* @param data the data to process
* @throws WekaException if a problem occurs
*/
@Override
public void processIncoming(Data data) throws WekaException {
if (m_isReset) {
Instances structure;
if (data.getConnectionName().equals(StepManager.CON_INSTANCE)) {
Instance inst = data.getPrimaryPayload();
structure = new Instances(inst.dataset(), 0);
m_streaming = true;
getStepManager().logBasic(
"Starting streaming sort. Using streaming " + "buffer size: "
+ m_bufferSizeI);
m_isReset = false;
} else {
structure = data.getPrimaryPayload();
structure = new Instances(structure, 0);
}
init(structure);
}
if (m_streaming) {
processIncremental(data);
} else {
processBatch(data);
}
if (isStopRequested()) {
getStepManager().interrupted();
} else if (!m_streaming) {
getStepManager().finished();
}
}
/**
* Process batch data
*
* @param data the data to process
* @throws WekaException if a problem occurs
*/
protected void processBatch(Data data) throws WekaException {
getStepManager().processing();
Instances insts = data.getPrimaryPayload();
getStepManager().logBasic("Sorting " + insts.relationName());
List instances = new ArrayList();
for (int i = 0; i < insts.numInstances(); i++) {
InstanceHolder h = new InstanceHolder();
h.m_instance = insts.instance(i);
instances.add(h);
}
Collections.sort(instances, m_sortComparator);
Instances output = new Instances(insts, 0);
for (int i = 0; i < instances.size(); i++) {
output.add(instances.get(i).m_instance);
}
Data outputD = new Data(data.getConnectionName(), output);
outputD.setPayloadElement(StepManager.CON_AUX_DATA_SET_NUM,
data.getPayloadElement(StepManager.CON_AUX_DATA_SET_NUM));
outputD.setPayloadElement(StepManager.CON_AUX_DATA_MAX_SET_NUM,
data.getPayloadElement(StepManager.CON_AUX_DATA_MAX_SET_NUM));
getStepManager().outputData(outputD);
}
/**
* Process incremental data
*
* @param data the data to process
* @throws WekaException if a problem occurs
*/
protected void processIncremental(Data data) throws WekaException {
if (isStopRequested()) {
return;
}
if (getStepManager().isStreamFinished(data)) {
emitBufferedInstances();
} else {
getStepManager().throughputUpdateStart();
InstanceHolder tempH = new InstanceHolder();
tempH.m_instance = data.getPrimaryPayload();
tempH.m_fileNumber = -1; // unused here
if (m_stringAttIndexes != null) {
copyStringAttVals(tempH);
}
m_incrementalBuffer.add(tempH);
if (m_incrementalBuffer.size() == m_bufferSizeI) {
// time to sort and write this to a temp file
try {
sortBuffer(true);
} catch (Exception ex) {
throw new WekaException(ex);
}
}
getStepManager().throughputUpdateEnd();
}
}
/**
* Output any buffered instances
*
* @throws WekaException if a problem occurs
*/
protected void emitBufferedInstances() throws WekaException {
if (isStopRequested()) {
return;
}
if (m_incrementalBuffer.size() > 0) {
try {
getStepManager().throughputUpdateStart();
sortBuffer(false);
getStepManager().throughputUpdateEnd();
} catch (Exception ex) {
throw new WekaException(ex);
}
if (m_bufferFiles.size() == 0) {
// we only have the in memory buffer
getStepManager().logDetailed("Emitting in memory buffer");
Instances newHeader =
new Instances(m_incrementalBuffer.get(0).m_instance.dataset(), 0);
for (int i = 0; i < m_incrementalBuffer.size(); i++) {
getStepManager().throughputUpdateStart();
InstanceHolder currentH = m_incrementalBuffer.get(i);
currentH.m_instance.setDataset(newHeader);
if (m_stringAttIndexes != null) {
for (String attName : m_stringAttIndexes.keySet()) {
boolean setValToZero =
newHeader.attribute(attName).numValues() > 0;
newHeader.attribute(attName).setStringValue(
currentH.m_stringVals.get(attName));
if (setValToZero) {
currentH.m_instance.setValue(newHeader.attribute(attName), 0);
}
}
}
if (isStopRequested()) {
return;
}
m_streamingData.setPayloadElement(StepManager.CON_INSTANCE,
currentH.m_instance);
getStepManager().throughputUpdateEnd();
getStepManager().outputData(m_streamingData);
if (i == m_incrementalBuffer.size() - 1) {
// end of stream
m_streamingData.clearPayload();
getStepManager().throughputFinished(m_streamingData);
}
}
return;
}
}
List inputStreams = new ArrayList();
// for the interleaving part of the merge sort
List merger = new ArrayList();
Instances tempHeader = new Instances(m_connectedFormat, 0);
// add an instance from the in-memory buffer first
if (m_incrementalBuffer.size() > 0) {
InstanceHolder tempH = m_incrementalBuffer.remove(0);
merger.add(tempH);
}
if (isStopRequested()) {
return;
}
if (m_bufferFiles.size() > 0) {
getStepManager().logDetailed("Merging temp files");
}
// open all temp buffer files and read one instance from each
for (int i = 0; i < m_bufferFiles.size(); i++) {
ObjectInputStream ois = null;
try {
FileInputStream fis = new FileInputStream(m_bufferFiles.get(i));
BufferedInputStream bis = new BufferedInputStream(fis, 50000);
ois = new ObjectInputStream(bis);
InstanceHolder tempH = (InstanceHolder) ois.readObject();
if (tempH != null) {
inputStreams.add(ois);
tempH.m_fileNumber = i;
merger.add(tempH);
} else {
// no instances?!??
ois.close();
}
} catch (Exception ex) {
if (ois != null) {
try {
ois.close();
} catch (Exception e) {
throw new WekaException(e);
}
}
throw new WekaException(ex);
}
}
Collections.sort(merger, m_sortComparator);
int mergeCount = 0;
do {
if (isStopRequested()) {
return;
}
InstanceHolder holder = merger.remove(0);
holder.m_instance.setDataset(tempHeader);
if (m_stringAttIndexes != null) {
for (String attName : m_stringAttIndexes.keySet()) {
boolean setValToZero =
(tempHeader.attribute(attName).numValues() > 1);
tempHeader.attribute(attName).setStringValue(
holder.m_stringVals.get(attName));
if (setValToZero) {
holder.m_instance.setValue(tempHeader.attribute(attName), 0);
}
}
}
m_streamingData.setPayloadElement(StepManager.CON_INSTANCE,
holder.m_instance);
mergeCount++;
getStepManager().outputData(m_streamingData);
getStepManager().throughputUpdateStart();
if (mergeCount % m_bufferSizeI == 0) {
getStepManager().logDetailed("Merged " + mergeCount + " instances");
}
int smallest = holder.m_fileNumber;
// now get another instance from the source of "smallest"
InstanceHolder nextH = null;
if (smallest == -1) {
if (m_incrementalBuffer.size() > 0) {
nextH = m_incrementalBuffer.remove(0);
nextH.m_fileNumber = -1;
}
} else {
ObjectInputStream tis = inputStreams.get(smallest);
try {
InstanceHolder tempH = (InstanceHolder) tis.readObject();
if (tempH != null) {
nextH = tempH;
nextH.m_fileNumber = smallest;
} else {
throw new Exception("end of buffer");
}
} catch (Exception ex) {
// EOF
try {
getStepManager().logDetailed("Closing temp file");
tis.close();
} catch (Exception e) {
throw new WekaException(ex);
}
File file = m_bufferFiles.remove(smallest);
// file.delete();
inputStreams.remove(smallest);
// update file numbers
for (InstanceHolder h : merger) {
if (h.m_fileNumber != -1 && h.m_fileNumber > smallest) {
h.m_fileNumber--;
}
}
}
}
if (nextH != null) {
// find the correct position (i.e. interleave) for this new Instance
int index = Collections.binarySearch(merger, nextH, m_sortComparator);
if (index < 0) {
merger.add(index * -1 - 1, nextH);
} else {
merger.add(index, nextH);
}
nextH = null;
}
getStepManager().throughputUpdateEnd();
} while (merger.size() > 0 && !isStopRequested());
if (!isStopRequested()) {
// signal end of stream
m_streamingData.clearPayload();
getStepManager().throughputFinished(m_streamingData);
} else {
// try an close any input streams still open...
for (ObjectInputStream is : inputStreams) {
try {
is.close();
} catch (Exception ex) {
// ignore
}
}
}
}
/**
* Sort the buffer
*
* @param write true if the buffer sould be written to a tmp file
* @throws Exception if a problem occurs
*/
private void sortBuffer(boolean write) throws Exception {
getStepManager().logBasic("Sorting in memory buffer");
Collections.sort(m_incrementalBuffer, m_sortComparator);
if (!write) {
return;
}
if (isStopRequested()) {
return;
}
String tmpDir = m_tempDirectory.toString();
File tempFile = File.createTempFile("Sorter", ".tmp");
if (tmpDir != null && tmpDir.length() > 0) {
tmpDir = environmentSubstitute(tmpDir);
File tempDir = new File(tmpDir);
if (tempDir.exists() && tempDir.canWrite()) {
String filename = tempFile.getName();
tempFile = new File(tmpDir + File.separator + filename);
tempFile.deleteOnExit();
}
}
getStepManager().logDebug("Temp file: " + tempFile.toString());
m_bufferFiles.add(tempFile);
FileOutputStream fos = new FileOutputStream(tempFile);
BufferedOutputStream bos = new BufferedOutputStream(fos, 50000);
ObjectOutputStream oos = new ObjectOutputStream(bos);
getStepManager().logDetailed(
"Writing buffer to temp file " + m_bufferFiles.size()
+ ". Buffer contains " + m_incrementalBuffer.size() + " instances");
for (int i = 0; i < m_incrementalBuffer.size(); i++) {
InstanceHolder temp = m_incrementalBuffer.get(i);
temp.m_instance.setDataset(null);
oos.writeObject(temp);
if (i % (m_bufferSizeI / 10) == 0) {
oos.reset();
}
}
bos.flush();
oos.close();
m_incrementalBuffer.clear();
}
private void copyStringAttVals(InstanceHolder holder) {
for (String attName : m_stringAttIndexes.keySet()) {
Attribute att = holder.m_instance.dataset().attribute(attName);
String val = holder.m_instance.stringValue(att);
if (holder.m_stringVals == null) {
holder.m_stringVals = new HashMap();
}
holder.m_stringVals.put(attName, val);
}
}
/**
* Inner class that holds instances and the index of the temp file that holds
* them (if operating in incremental mode)
*/
protected static class InstanceHolder implements Serializable {
/** For serialization */
private static final long serialVersionUID = -3985730394250172995L;
/** The instance */
protected Instance m_instance;
/** index into the list of files on disk */
protected int m_fileNumber;
/**
* for incremental operation, if string attributes are present then we need
* to store them with each instance - since incremental streaming in the
* knowledge flow only maintains one string value in memory (and hence in
* the header) at any one time
*/
protected Map m_stringVals;
}
/**
* Comparator that applies the sort rules to {@code InstanceHolder}s
*/
protected static class SortComparator implements Comparator {
/** The rules to apply */
protected List m_sortRules;
/**
* Constructor
*
* @param sortRules the rules to apply
*/
public SortComparator(List sortRules) {
m_sortRules = sortRules;
}
/**
* @param o1 the first {@code InstanceHolder} to compare
* @param o2 the second {@code InstanceHolder} to compare
* @return the result of the comparison - the first rule that returns a
* non-zero comparison value
*/
@Override
public int compare(InstanceHolder o1, InstanceHolder o2) {
int cmp = 0;
for (SortRule sr : m_sortRules) {
cmp = sr.compare(o1, o2);
if (cmp != 0) {
return cmp;
}
}
return 0;
}
}
/**
* Implements a sorting rule based on a single attribute
*/
public static class SortRule implements Comparator {
/** Name or index of the attribute to compare on */
protected String m_attributeNameOrIndex;
/** The actual attribute to compare on */
protected Attribute m_attribute;
/** True for descending instead of ascending order */
protected boolean m_descending;
/**
* Constructor
*
* @param att the name or index of the attribute to compare on
* @param descending true if order should be descending
*/
public SortRule(String att, boolean descending) {
m_attributeNameOrIndex = att;
m_descending = descending;
}
/**
* Constructor
*/
public SortRule() {
}
/**
* Constructor
*
* @param setup the definition of a sort rule
*/
public SortRule(String setup) {
parseFromInternal(setup);
}
protected void parseFromInternal(String setup) {
String[] parts = setup.split("@@SR@@");
if (parts.length != 2) {
throw new IllegalArgumentException("Malformed sort rule: " + setup);
}
m_attributeNameOrIndex = parts[0].trim();
m_descending = parts[1].equalsIgnoreCase("Y");
}
/**
* Gets the rule in internal format
*
* @return the rule in internal format
*/
public String toStringInternal() {
return m_attributeNameOrIndex + "@@SR@@" + (m_descending ? "Y" : "N");
}
/**
* Prints the rule in human readable format
*
* @return a human readable formatted rule
*/
@Override
public String toString() {
StringBuffer res = new StringBuffer();
res.append("Attribute: " + m_attributeNameOrIndex + " - sort "
+ (m_descending ? "descending" : "ascending"));
return res.toString();
}
/**
* Set the name or index of the attribute to sort on
*
* @param att the name or index of tha attribute to sort on
*/
public void setAttribute(String att) {
m_attributeNameOrIndex = att;
}
/**
* Get the name or index of the attribute to sort on
*
* @return the name or index of the attribute to sort on
*/
public String getAttribute() {
return m_attributeNameOrIndex;
}
/**
* Set whether the sort should be descending rather than ascending
*
* @param d true for a descending sort
*/
public void setDescending(boolean d) {
m_descending = d;
}
/**
* Return true if the sort is descending
*
* @return true if the sort is descending
*/
public boolean getDescending() {
return m_descending;
}
/**
* Initialize the rule
*
* @param env the environment variables to use
* @param structure the structure of the instances that the rule will
* opperate on
*/
public void init(Environment env, Instances structure) {
String attNameI = m_attributeNameOrIndex;
try {
attNameI = env.substitute(attNameI);
} catch (Exception ex) {
}
if (attNameI.equalsIgnoreCase("/first")) {
m_attribute = structure.attribute(0);
} else if (attNameI.equalsIgnoreCase("/last")) {
m_attribute = structure.attribute(structure.numAttributes() - 1);
} else {
// try actual attribute name
m_attribute = structure.attribute(attNameI);
if (m_attribute == null) {
// try as an index
try {
int index = Integer.parseInt(attNameI);
m_attribute = structure.attribute(index);
} catch (NumberFormatException n) {
throw new IllegalArgumentException("Unable to locate attribute "
+ attNameI + " as either a named attribute or as a valid "
+ "attribute index");
}
}
}
}
/**
* Compare two instances according to the rule
*
* @param o1 the first instance
* @param o2 the second instance
* @return the result of the comparison
*/
@Override
public int compare(InstanceHolder o1, InstanceHolder o2) {
// both missing is equal
if (o1.m_instance.isMissing(m_attribute)
&& o2.m_instance.isMissing(m_attribute)) {
return 0;
}
// one missing - missing instances should all be at the end
// regardless of whether order is ascending or descending
if (o1.m_instance.isMissing(m_attribute)) {
return 1;
}
if (o2.m_instance.isMissing(m_attribute)) {
return -1;
}
int cmp = 0;
if (!m_attribute.isString() && !m_attribute.isRelationValued()) {
double val1 = o1.m_instance.value(m_attribute);
double val2 = o2.m_instance.value(m_attribute);
cmp = Double.compare(val1, val2);
} else if (m_attribute.isString()) {
String val1 = o1.m_stringVals.get(m_attribute.name());
String val2 = o2.m_stringVals.get(m_attribute.name());
/*
* String val1 = o1.stringValue(m_attribute); String val2 =
* o2.stringValue(m_attribute);
*/
// TODO case insensitive?
cmp = val1.compareTo(val2);
} else {
throw new IllegalArgumentException("Can't sort according to "
+ "relation-valued attribute values!");
}
if (m_descending) {
return -cmp;
}
return cmp;
}
}
/**
* Return the fully qualified name of a custom editor component (JComponent)
* to use for editing the properties of the step. This method can return null,
* in which case the system will dynamically generate an editor using the
* GenericObjectEditor
*
* @return the fully qualified name of a step editor component
*/
@Override
public String getCustomEditorForStep() {
return "weka.gui.knowledgeflow.steps.SorterStepEditorDialog";
}
}