package org.greenstone.gsdl3.gs3build.indexers;
import java.util.List;
import java.util.ArrayList;
import java.util.Iterator;
import java.io.File;
import java.io.InputStream;
import java.io.OutputStream;
import java.io.IOException;
import java.io.BufferedReader;
import java.io.InputStreamReader;
import org.w3c.dom.*;
import org.greenstone.mg.*;
import org.greenstone.gsdl3.gs3build.doctypes.DocumentID;
import org.greenstone.gsdl3.gs3build.doctypes.DocumentInterface;
import org.greenstone.gsdl3.gs3build.doctypes.HTMLDocument;
import org.greenstone.gsdl3.gs3build.doctypes.METSDocument;
import org.greenstone.gsdl3.gs3build.metadata.*;
import org.greenstone.gsdl3.gs3build.xpointer.XPointer;
import org.greenstone.gsdl3.util.GSXML;
import org.greenstone.gsdl3.util.Misc;
import org.greenstone.gsdl3.util.Processing;
public class MGIndexer extends AbstractIndexer
{
int pass;
int documentSeqNo;
int sectionSeqNo;
boolean firstDocument;
String outputDirectory;
// InputStream indexerFeedback;
// InputStream indexerErrors;
//OutputStream indexerTextfeed;
StringBuffer indexBuffer;
//Process mg_passes;
File textDirectory;
File indexDirectory;
String indexStem;
String textStem;
List indexes;
String overallName;
String currentIndexName;
String currentIndexLevel;
String currentIndexField;
MGPassesWrapper mgPasses;
static final char END_OF_DOCUMENT = (char) 2;
static final char END_OF_SECTION = (char) 3; // actually this is end of para for mg
static final char END_OF_STREAM = (char) 4;
public static final String MG_INDEX_TYPE = "mg";
public static final String INDEX_FILE_STEM = "index";
class MGIndex
{
String name=null;
String level=null;
String field=null;
boolean error = false;// assume built until we get an error
public MGIndex(String name, String level, String field)
{
this.name = name;
this.level = level;
this.field = field;
}
public MGIndex(String indexLabel)
{
int colonAt = indexLabel.indexOf(':');
if (colonAt >= 0) {
this.field = indexLabel.substring(colonAt+1);
this.level = indexLabel.substring(0, colonAt);
createIndexName();
}
}
public String getLevel()
{
return this.level;
}
public String getField()
{
return this.field;
}
public String getName()
{
if (this.name==null || this.name.equals("")) {
createIndexName();
}
return this.name;
}
public boolean hasError() {
return this.error;
}
public void setError(boolean b) {
this.error = b;
}
private void createIndexName() {
StringBuffer new_name = new StringBuffer();
new_name.append(Character.toLowerCase((char) this.level.charAt(0)));
int c, w;
w = 0;
c = 0;
while (c < this.field.length() && w < 2) {
char ch = this.field.charAt(c);
ch = Character.toLowerCase(ch);
if (Character.isLetter(ch)) {
if (ch != 'a' && ch != 'e' && ch != 'i' &&
ch != 'o' && ch != 'u') {
new_name.append(ch);
w++;
}
}
c ++;
}
this.name = new_name.toString();
}
} // MGIndex
public MGIndexer(String name)
{
this.indexes = new ArrayList();
this.overallName = name;
}
public String getIndexType()
{
return MG_INDEX_TYPE;
}
public String getName()
{
return this.overallName;
}
// private String getIndexDirectory(String level, String field)
// { StringBuffer directory = new StringBuffer();
// directory.append(Character.toLowerCase((char) level.charAt(0)));
// int c, w;
// w = 0;
// c = 0;
// while (c < field.length() && w < 2) {
// char ch = field.charAt(c);
// ch = Character.toLowerCase(ch);
// if (Character.isLetter(ch)) {
// if (ch != 'a' && ch != 'e' && ch != 'i' &&
// ch != 'o' && ch != 'u') {
// directory.append(ch);
// w++;
// }
// }
// c ++;
// }
// return directory.toString();
// }
/**
* The output directory should be (collection)/building/text/ for
* normal Greenstone builds.
*
* @param String
the label to configure
* @param String
the value...
*/
public boolean configure(String label, String value)
{
if (label.equals(IndexerManager.outputDir)) {
this.outputDirectory = value;
this.pass = 0;
// attempt to ensure that the text subdirectory exists
this.textDirectory = new File(outputDirectory, "text");
if (!textDirectory.exists()) {
if (!textDirectory.mkdir()) {
return false;
}
}
else if (!textDirectory.isDirectory()) {
return false;
}
this.textStem = this.textDirectory.getPath() + File.separator + INDEX_FILE_STEM;
// Sign to the user which mg directory is being used...
System.out.println("Output MG directory is " + this.textStem);
}
else if (label.equals(IndexerInterface.GS2_INDEX_LABEL)) {
this.indexes.add(new MGIndex(value));
}
return true;
}
public boolean addIndex(String name, String level, String field)
{
MGIndex index = new MGIndex(name, level, field);
this.indexes.add(index);
return true;
}
private Node recurseDOM(DocumentInterface metsDoc, Node node,
AbstractStructure structure, StringBuffer textBuffer,
StringBuffer extraBuffer, String namespace)
//String name, String namespace, String field)
{
// send out the ctrl-c...if this is
if (structure.getStructureType().equals(METSDivision.DIVISION_TYPE)) {
// try doing this for all index types
if ((this.currentIndexName != null)) { // && this.level != null && this.level.equals(IndexerInterface.SECTION_LEVEL)) { //name.startsWith("s")) {
METSDivision division = (METSDivision) structure;
// get the division metadata block
METSDescriptive descriptive;
String metadataId = division.getDefaultMetadataReference();
if (metadataId == null) {
descriptive = metsDoc.getDocumentMetadata().createDescriptive(division.getLabel());
division.addMetadataReference(descriptive.getID());
}
else {
// Get the descriptive item...
descriptive = metsDoc.getDocumentMetadata().getDescriptiveById(metadataId);
}
descriptive.addMetadata("gsdl3", "mgseqno", this.overallName + "." + Integer.toString(this.sectionSeqNo));
metsDoc.setChanged(true);
//metsDoc.setModified(true);
// System.out.println("Assigning " + this.sectionSeqNo + " to " + metsDoc.getID() + " " + division.getLabel());
} // section level
// append an 'end of section' marker
//textBuffer.append(END_OF_SECTION);
this.sectionSeqNo ++;
// for document-level indexes, always append an 'end of document' tag at the
// end of the document for each section. Otherwise, each section is followed
// by an end of document character. This ensures that all indexes use the
// same document numbering...
if (this.currentIndexLevel == null ||
this.currentIndexLevel.equals(IndexerInterface.DOCUMENT_LEVEL)) {
extraBuffer.append(END_OF_DOCUMENT);
}
else {
textBuffer.append(END_OF_DOCUMENT);
this.documentSeqNo ++;
}
// produce the body here for metadata output of divisions - in the case of
// text output, that will happen below...
if (!this.currentIndexField.equals("text")) {
METSDescriptive descriptive;
METSDivision division = (METSDivision) structure;
String metadataId = division.getDefaultMetadataReference();
descriptive = metsDoc.getDocumentMetadata().getDescriptiveById(metadataId);
if (descriptive != null) {
List values = descriptive.getMetadata(namespace, this.currentIndexField);
if (values != null) {
Iterator valueIter = values.iterator();
while (valueIter.hasNext()) {
String value = valueIter.next().toString();
textBuffer.append(value);
if (valueIter.hasNext()) {
//textBuffer.append(END_OF_SECTION);
}
}
}
}
}
}
// go through our children as required...
Iterator children = structure.getChildIterator();
Node startNode;
while (children.hasNext()) {
AbstractStructure child = (AbstractStructure) children.next();
// get xpointer for child
// get start position node
if (metsDoc.getDocumentType() == "METS"){
startNode = ((METSDocument) metsDoc).getSectionStartNode((METSDivision) child);
} else {
startNode = ((HTMLDocument) metsDoc).getSectionStartNode((METSDivision) child);
}
//Node startNode = ((HTMLDocument) metsDoc).getSectionStartNode((METSDivision) child);
// while this node isn't the child's start node, produce the HTML node text, if
// in text field mode...
if (this.currentIndexField.equals("text")) {
while (node != startNode) {
XPointer.printNode(node, textBuffer, false);
// print buffer to node
node = XPointer.getNextNode(node, (this.currentIndexField.equals("text") ? textBuffer : null));
}
}
// recurse to child
node = this.recurseDOM(metsDoc, node, child, textBuffer, extraBuffer, namespace); // name, namespace, field);
} // while next child
// close a document - the actual closing \B will be done by the main
// loop, so only a required \C is printed here...
if (structure.getStructureType().equals(METSStructure.STRUCTURE_TYPE)) {
while (node != null) {
if (this.currentIndexField.equals("text")) {
XPointer.printNode(node, textBuffer, false);
}
node = XPointer.getNextNode(node, (this.currentIndexField.equals("text") ? textBuffer : null));
}
//textBuffer.append(END_OF_SECTION);
this.sectionSeqNo ++;
}
return node;
}
private String prepareDOM(DocumentInterface metsDoc, Document document, METSStructure structure, String namespace)
{
// String name, String namespace, String field)
StringBuffer extraBuffer = new StringBuffer();
Node node = document.getDocumentElement();
StringBuffer textBuffer = new StringBuffer();
this.recurseDOM(metsDoc, node, structure, textBuffer, extraBuffer, namespace); //name, namespace, field);
textBuffer.append(extraBuffer.toString());
return textBuffer.toString();
}
/**
* Index a single document; the document interface can be used to extract individual
* metadata items etc. as required or desired and index those instead or as well as
* the body text of the document.
*/
public boolean indexDocument(DocumentID docID, DocumentInterface document)
{
if (this.pass == 0) {
document.removeAllMetadata("gsdl3", "mgseqno");
}
if (!this.firstDocument) {
this.indexBuffer.append(END_OF_DOCUMENT);
mgPasses.processDocument(indexBuffer.toString());
this.indexBuffer.delete(0, this.indexBuffer.length());
}
String docText = null;
int startSeqNo = this.sectionSeqNo;
this.sectionSeqNo ++;
Document domDocument = document.getDOMDocument();
if (domDocument != null) {
System.err.println("dom doc is not null");
METSStructure sections = document.getDocumentStructure().getStructure("Section");
if (sections != null) {
System.err.println("sections are not null");
docText = this.prepareDOM(document, domDocument, sections, "gsdl3"); //this.name, "gsdl3", this.field);
// System.out.println(docText);
}
}
if (docText == null) {
System.err.println("dom doc or sections was null - asking for doc text");
if (this.currentIndexField.equals("text")) {
//docText = Character.toString(END_OF_DOCUMENT) + document.getDocumentText();
docText = document.getDocumentText();
}
else {
StringBuffer textBuffer = new StringBuffer();
//textBuffer.append(END_OF_DOCUMENT);
List values = document.getDocumentMetadataItem("gsdl3", this.currentIndexField);
if (values != null) {
Iterator valueIter = values.iterator();
while (valueIter.hasNext()) {
String value = valueIter.next().toString();
textBuffer.append(value);
if (valueIter.hasNext()) {
//textBuffer.append(END_OF_SECTION);
// sectionSeqNo ++;
}
}
}
else {
textBuffer.append("No data");
}
docText = textBuffer.toString();
}
sectionSeqNo ++;
}
this.indexBuffer.append(docText);
// remember that we're not on the first document,
this.firstDocument = false;
// assign the sequence number on the first pass only, and increment the sequence number.
if (this.pass == 0) {
document.addDocumentMetadata("gsdl3", "mgseqno", this.overallName+"."+Integer.toString(startSeqNo));
}
this.documentSeqNo += 1;
return true;
}
/**
* Initialise the pass: open required files, check status
*/
public boolean startPass(int passNumber) {
this.pass = passNumber;
this.firstDocument = true;
this.documentSeqNo = 1;
this.sectionSeqNo = 1;
this.mgPasses = new MGPassesWrapper();
this.indexBuffer = new StringBuffer();
int indexNo = (this.pass - 2) / 2;
MGIndex index = null;
if (this.pass >= 2) {
index = (MGIndex) this.indexes.get(indexNo);
if (index.hasError()) {
// an error has already occurred for this index, don't continue
System.out.println("pass "+this.pass+": aborted due to errors in the previous pass");
return false;
}
// attempt to ensure that the text subdirectory exists
this.indexDirectory = new File(outputDirectory, index.getName());
if (!indexDirectory.exists()) {
if (!indexDirectory.mkdir()) {
return false;
}
}
else if (!indexDirectory.isDirectory()) {
return false;
}
this.currentIndexLevel = index.getLevel();
this.currentIndexField = index.getField();
this.currentIndexName = index.getName();
if (this.currentIndexLevel == null || this.currentIndexField == null ) {
System.out.println("invalid index - level or field was null");
return false;
}
this.indexStem = this.indexDirectory.getPath() + File.separatorChar + INDEX_FILE_STEM; // TODO: modify for index
if (this.pass % 2 == 1) {
this.currentIndexName = null; // why???
}
}
else {
this.currentIndexField = "text";
this.currentIndexLevel = "section";
this.currentIndexName = null;
}
// get the parameters for this execution of mg_passes
mgPasses.setFileName((this.pass < 2 ? this.textDirectory.toString() : this.indexDirectory.toString())+File.separator+ "index");
if (!Misc.isWindows()) {
mgPasses.setBasePath("/");
}
int mgPass = this.pass < 2 ? this.pass : ((this.pass % 2) + 2);
mgPasses.setBufferSize(100000);
switch (mgPass) {
case 0:
// -b 100000 -T1
mgPasses.addPass(MGPassesWrapper.TEXT_PASS_1);
break;
case 1:
// -b 100000 -T2
mgPasses.addPass(MGPassesWrapper.TEXT_PASS_2);
break;
case 2:
// -b 100000 -2 -m 32 -s 0 -G -t 10 -N1
mgPasses.addPass(MGPassesWrapper.INDEX_PASS_1);
mgPasses.setInvfLevel(MGPassesWrapper.INVF_LEVEL_2);
mgPasses.setStemOptions(MGPassesWrapper.STEMMER_ENGLISH, MGPassesWrapper.NO_STEM_OR_CASE);
mgPasses.setInversionMemLimit(32);
mgPasses.ignoreSGMLTags(true);
break;
case 3:
// -b 100000 -2 -c 3 -G -t 10 -N2
mgPasses.addPass(MGPassesWrapper.INDEX_PASS_2);
mgPasses.setInvfLevel(MGPassesWrapper.INVF_LEVEL_2);
mgPasses.ignoreSGMLTags(true);
break;
}
mgPasses.init();
System.out.println("Pass " + this.pass);
return true;
}
/**
* Complete a pass - reset file counters, close files, etc.
*/
public boolean endPass(int passNumber) {
Process p;
int indexNo = (passNumber - 2) / 2;
MGIndex index = null;
if (passNumber >= 2) {
index = (MGIndex) this.indexes.get(indexNo);
}
try {
this.indexBuffer.append(END_OF_DOCUMENT);
mgPasses.processDocument(indexBuffer.toString());
this.indexBuffer.delete(0, this.indexBuffer.length());
Thread.sleep(1000); // what for??
}
catch (InterruptedException ex) {
System.out.println(ex);
}
mgPasses.finish();
try {
Thread.sleep(1000);
} catch (Exception e) {}
int exit_value = 0;
System.out.println("Pass " + this.pass + " completed with " + exit_value);
if (exit_value !=0) {
//assume something has gone wrong, don't continue
if (index != null) {
index.setError(true);
return false;
}
}
int mgPass = this.pass < 2 ? this.pass : ((this.pass % 2) + 2);
String osextra = "";
if (!Misc.isWindows()) {
osextra = " -d / ";
}
switch (mgPass) {
case 0:
System.out.println("Compressing dictionary");
exit_value = Processing.runProcess("mg_compression_dict -f " + this.textDirectory.toString()+File.separator+"index" + osextra + " -S -H -2 -k 5120");
if (exit_value == 0) {
System.out.println("Compressed dictionary successfully written");
} else {
System.err.println("Error from mg_compression_dict: " + exit_value);
index.setError(true);
return false;
}
break;
case 2:
System.out.println("Creating perfect hash");
exit_value = Processing.runProcess("mg_perf_hash_build -f " + this.indexDirectory.toString()+File.separator+ "index"+osextra);
if (exit_value ==0) {
System.out.println("Perfect hashes completed");
} else {
System.err.println("Unable to build the perfect hash");
index.setError(true);
return false;
}
break;
case 3:
System.out.println("Writing weights file");
exit_value = Processing.runProcess("mg_weights_build -f " + this.indexStem + " -t " + this.textStem + osextra);
if (exit_value ==0) {
System.out.println("Weights file successfully written");
} else {
System.err.println("Unable to create weights file");
index.setError(true);
return false;
}
System.out.println("Creating inverted dictionary");
exit_value = Processing.runProcess("mg_invf_dict -f " + this.indexDirectory.toString()+File.separator+"index" + osextra);
if (exit_value ==0) {
System.out.println("Inverted dictionary file successfully written");
} else {
System.out.println("Unable to create inverted dictionary file");
index.setError(true);
return false;
}
System.out.println("Creating Stem indexes");
exit_value = Processing.runProcess("mg_stem_idx -b 4096 -s1 -f " + this.indexDirectory.toString()+File.separator+"index"+osextra);
if (exit_value == 0) {
System.out.println("Stemmed index 1 successfully written");
} else {
System.out.println("Unable to create stemmed index 1");
index.setError(true);
return false;
}
exit_value = Processing.runProcess("mg_stem_idx -b 4096 -s2 -f " + this.indexDirectory.toString()+File.separator+"index"+osextra);
if (exit_value == 0) {
System.out.println("Stemmed index 2 successfully written");
} else {
System.out.println("Unable to create stemmed index 2");
index.setError(true);
return false;
}
exit_value = Processing.runProcess("mg_stem_idx -b 4096 -s3 -f " + this.indexDirectory.toString()+File.separator+"index"+osextra);
if (exit_value == 0) {
System.out.println("Stemmed index 3 successfully written");
} else {
System.out.println("Unable to create stemmed index 3");
index.setError(true);
return false;
}
break;
} // switch
mgPasses = null;
return true;
}
/**
* Do any tidying up
*/
public void tidyup()
{
}
/**
* Return the number of passes required for this index.
*/
public int getNumberOfPasses()
{
return 2 + this.indexes.size() * 2;
}
public boolean addServiceDescriptions(org.w3c.dom.Element service_rack_list)
{
Document doc = service_rack_list.getOwnerDocument();
// generate the list of indexes
Element index_list = doc.createElement(GSXML.INDEX_ELEM+GSXML.LIST_MODIFIER);
boolean found_index = false;
String def_index = ""; // the default index will just be the first one created for now.
for (int i=0; i