source: trunk/gsdl3/src/java/org/greenstone/gsdl3/gs3build/indexers/MGIndexer.java@ 9006

Last change on this file since 9006 was 9006, checked in by kjdon, 19 years ago

now outputs an indexStem element in buildCOnfig

  • Property svn:keywords set to Author Date Id Revision
File size: 21.4 KB
Line 
1package org.greenstone.gsdl3.gs3build.indexers;
2
3import java.util.List;
4import java.util.ArrayList;
5import java.util.Iterator;
6
7import java.io.File;
8import java.io.InputStream;
9import java.io.OutputStream;
10import java.io.IOException;
11import java.io.BufferedReader;
12import java.io.InputStreamReader;
13
14import org.w3c.dom.*;
15
16import org.greenstone.mg.*;
17
18import org.greenstone.gsdl3.gs3build.doctypes.DocumentID;
19import org.greenstone.gsdl3.gs3build.doctypes.DocumentInterface;
20import org.greenstone.gsdl3.gs3build.doctypes.HTMLDocument;
21import org.greenstone.gsdl3.gs3build.doctypes.METSDocument;
22import org.greenstone.gsdl3.gs3build.metadata.*;
23import org.greenstone.gsdl3.gs3build.xpointer.XPointer;
24import org.greenstone.gsdl3.util.GSXML;
25import org.greenstone.gsdl3.util.Misc;
26import org.greenstone.gsdl3.util.Processing;
27
28public class MGIndexer extends AbstractIndexer
29{
30 int pass;
31 int documentSeqNo;
32 int sectionSeqNo;
33 boolean firstDocument;
34 String outputDirectory;
35// InputStream indexerFeedback;
36// InputStream indexerErrors;
37 //OutputStream indexerTextfeed;
38 StringBuffer indexBuffer;
39 //Process mg_passes;
40 File textDirectory;
41 File indexDirectory;
42 String indexStem;
43 String textStem;
44 List indexes;
45 String overallName;
46
47 String currentIndexName;
48 String currentIndexLevel;
49 String currentIndexField;
50
51 MGPassesWrapper mgPasses;
52
53 static final char END_OF_DOCUMENT = (char) 2;
54 static final char END_OF_SECTION = (char) 3; // actually this is end of para for mg
55 static final char END_OF_STREAM = (char) 4;
56
57 public static final String MG_INDEX_TYPE = "mg";
58 public static final String INDEX_FILE_STEM = "index";
59
60 class MGIndex
61 { String name=null;
62 String level=null;
63 String field=null;
64 boolean error = false;// assume built until we get an error
65
66 public MGIndex(String name, String level, String field)
67 { this.name = name;
68 this.level = level;
69 this.field = field;
70 }
71
72 public MGIndex(String indexLabel)
73 { int colonAt = indexLabel.indexOf(':');
74
75 if (colonAt >= 0)
76 { this.field = indexLabel.substring(colonAt+1);
77 this.level = indexLabel.substring(0, colonAt);
78 createIndexName();
79 }
80 }
81
82 public String getLevel()
83 { return this.level;
84 }
85
86 public String getField()
87 { return this.field;
88 }
89
90 public String getName()
91 {
92 if (this.name==null || this.name.equals("")) {
93 createIndexName();
94 }
95 return this.name;
96 }
97
98 public boolean hasError() {
99 return this.error;
100 }
101 public void setError(boolean b) {
102 this.error = b;
103 }
104
105 private void createIndexName() {
106 StringBuffer new_name = new StringBuffer();
107 new_name.append(Character.toLowerCase((char) this.level.charAt(0)));
108
109 int c, w;
110 w = 0;
111 c = 0;
112 while (c < this.field.length() && w < 2) {
113 char ch = this.field.charAt(c);
114
115 ch = Character.toLowerCase(ch);
116 if (Character.isLetter(ch)) {
117 if (ch != 'a' && ch != 'e' && ch != 'i' &&
118 ch != 'o' && ch != 'u') {
119 new_name.append(ch);
120 w++;
121 }
122 }
123 c ++;
124 }
125 this.name = new_name.toString();
126
127 }
128 } // MGIndex
129
130 public MGIndexer(String name)
131 { this.indexes = new ArrayList();
132 this.overallName = name;
133
134 }
135
136 public String getIndexType()
137 { return MG_INDEX_TYPE;
138 }
139
140 public String getName()
141 { return this.overallName;
142 }
143
144// private String getIndexDirectory(String level, String field)
145// { StringBuffer directory = new StringBuffer();
146// directory.append(Character.toLowerCase((char) level.charAt(0)));
147
148// int c, w;
149// w = 0;
150// c = 0;
151// while (c < field.length() && w < 2) {
152// char ch = field.charAt(c);
153
154// ch = Character.toLowerCase(ch);
155// if (Character.isLetter(ch)) {
156// if (ch != 'a' && ch != 'e' && ch != 'i' &&
157// ch != 'o' && ch != 'u') {
158// directory.append(ch);
159// w++;
160// }
161// }
162// c ++;
163// }
164// return directory.toString();
165// }
166
167 /**
168 * The output directory should be (collection)/building/text/ for
169 * normal Greenstone builds.
170 *
171 * @param <code>String</code> the label to configure
172 * @param <code>String</code> the value...
173 */
174 public boolean configure(String label, String value)
175 {
176 if (label.equals(IndexerManager.outputDir)) {
177 this.outputDirectory = value;
178 this.pass = 0;
179
180 // attempt to ensure that the text subdirectory exists
181 this.textDirectory = new File(outputDirectory, "text");
182 if (!textDirectory.exists()) {
183 if (!textDirectory.mkdir()) {
184 return false;
185 }
186 }
187 else if (!textDirectory.isDirectory()) {
188 return false;
189 }
190 this.textStem = this.textDirectory.getPath() + File.separator + INDEX_FILE_STEM;
191
192 // Sign to the user which mg directory is being used...
193 System.out.println("Output MG directory is " + this.textStem);
194 }
195 else if (label.equals(IndexerInterface.GS2_INDEX_LABEL)) {
196 this.indexes.add(new MGIndex(value));
197 }
198
199 return true;
200 }
201
202 public boolean addIndex(String name, String level, String field)
203 {
204 MGIndex index = new MGIndex(name, level, field);
205 this.indexes.add(index);
206 return true;
207 }
208
209 private Node recurseDOM(DocumentInterface metsDoc, Node node,
210 AbstractStructure structure, StringBuffer textBuffer,
211 StringBuffer extraBuffer, String namespace)
212 //String name, String namespace, String field)
213 {
214 // send out the ctrl-c...if this is
215 if (structure.getStructureType().equals(METSDivision.DIVISION_TYPE)) {
216 // try doing this for all index types
217 if ((this.currentIndexName != null)) { // && this.level != null && this.level.equals(IndexerInterface.SECTION_LEVEL)) { //name.startsWith("s")) {
218 METSDivision division = (METSDivision) structure;
219
220 // get the division metadata block
221 METSDescriptive descriptive;
222 String metadataId = division.getDefaultMetadataReference();
223 if (metadataId == null) {
224 descriptive = metsDoc.getDocumentMetadata().createDescriptive(division.getLabel());
225 division.addMetadataReference(descriptive.getID());
226 }
227 else {
228 // Get the descriptive item...
229 descriptive = metsDoc.getDocumentMetadata().getDescriptiveById(metadataId);
230 }
231
232 descriptive.addMetadata("gsdl3", "mgseqno", this.overallName + "." + Integer.toString(this.sectionSeqNo));
233
234 metsDoc.setChanged(true);
235 //metsDoc.setModified(true);
236 // System.out.println("Assigning " + this.sectionSeqNo + " to " + metsDoc.getID() + " " + division.getLabel());
237 } // section level
238
239 // append an 'end of section' marker
240 //textBuffer.append(END_OF_SECTION);
241 this.sectionSeqNo ++;
242
243 // for document-level indexes, always append an 'end of document' tag at the
244 // end of the document for each section. Otherwise, each section is followed
245 // by an end of document character. This ensures that all indexes use the
246 // same document numbering...
247 if (this.currentIndexLevel == null ||
248 this.currentIndexLevel.equals(IndexerInterface.DOCUMENT_LEVEL)) {
249 extraBuffer.append(END_OF_DOCUMENT);
250 }
251 else {
252 textBuffer.append(END_OF_DOCUMENT);
253 this.documentSeqNo ++;
254 }
255
256 // produce the body here for metadata output of divisions - in the case of
257 // text output, that will happen below...
258 if (!this.currentIndexField.equals("text"))
259 { METSDescriptive descriptive;
260
261 METSDivision division = (METSDivision) structure;
262
263 String metadataId = division.getDefaultMetadataReference();
264
265 descriptive = metsDoc.getDocumentMetadata().getDescriptiveById(metadataId);
266 if (descriptive != null) {
267 List values = descriptive.getMetadata(namespace, this.currentIndexField);
268
269 if (values != null) {
270 Iterator valueIter = values.iterator();
271 while (valueIter.hasNext()) {
272 String value = valueIter.next().toString();
273
274 textBuffer.append(value);
275 if (valueIter.hasNext()) {
276 //textBuffer.append(END_OF_SECTION);
277 }
278 }
279 }
280 }
281 }
282 }
283
284 // go through our children as required...
285 Iterator children = structure.getChildIterator();
286 Node startNode;
287 while (children.hasNext()) {
288 AbstractStructure child = (AbstractStructure) children.next();
289
290 // get xpointer for child
291 // get start position node
292 if (metsDoc.getDocumentType() == "METS"){
293 startNode = ((METSDocument) metsDoc).getSectionStartNode((METSDivision) child);
294 } else {
295 startNode = ((HTMLDocument) metsDoc).getSectionStartNode((METSDivision) child);
296 }
297 //Node startNode = ((HTMLDocument) metsDoc).getSectionStartNode((METSDivision) child);
298
299 // while this node isn't the child's start node, produce the HTML node text, if
300 // in text field mode...
301 if (this.currentIndexField.equals("text")) {
302 while (node != startNode) {
303 XPointer.printNode(node, textBuffer, false);
304
305 // print buffer to node
306 node = XPointer.getNextNode(node, (this.currentIndexField.equals("text") ? textBuffer : null));
307 }
308 }
309
310 // recurse to child
311 node = this.recurseDOM(metsDoc, node, child, textBuffer, extraBuffer, namespace); // name, namespace, field);
312 } // while next child
313
314 // close a document - the actual closing \B will be done by the main
315 // loop, so only a required \C is printed here...
316 if (structure.getStructureType().equals(METSStructure.STRUCTURE_TYPE)) {
317 while (node != null) {
318 if (this.currentIndexField.equals("text")) {
319 XPointer.printNode(node, textBuffer, false);
320 }
321 node = XPointer.getNextNode(node, (this.currentIndexField.equals("text") ? textBuffer : null));
322 }
323
324 //textBuffer.append(END_OF_SECTION);
325 this.sectionSeqNo ++;
326
327 }
328 return node;
329 }
330
331 private String prepareDOM(DocumentInterface metsDoc, Document document, METSStructure structure, String namespace)
332 // String name, String namespace, String field)
333 { StringBuffer extraBuffer = new StringBuffer();
334 Node node = document.getDocumentElement();
335 StringBuffer textBuffer = new StringBuffer();
336
337 this.recurseDOM(metsDoc, node, structure, textBuffer, extraBuffer, namespace); //name, namespace, field);
338 textBuffer.append(extraBuffer.toString());
339 return textBuffer.toString();
340 }
341
342 /**
343 * Index a single document; the document interface can be used to extract individual
344 * metadata items etc. as required or desired and index those instead or as well as
345 * the body text of the document.
346 */
347 public boolean indexDocument(DocumentID docID, DocumentInterface document)
348 {
349 if (this.pass == 0) {
350 document.removeAllMetadata("gsdl3", "mgseqno");
351 }
352
353 if (!this.firstDocument) {
354 this.indexBuffer.append(END_OF_DOCUMENT);
355 mgPasses.processDocument(indexBuffer.toString());
356 this.indexBuffer.delete(0, this.indexBuffer.length());
357
358 }
359
360 String docText = null;
361
362 int startSeqNo = this.sectionSeqNo;
363 this.sectionSeqNo ++;
364
365 Document domDocument = document.getDOMDocument();
366 if (domDocument != null) {
367 System.err.println("dom doc is not null");
368 METSStructure sections = document.getDocumentStructure().getStructure("Section");
369 if (sections != null) {
370 System.err.println("sections are not null");
371 docText = this.prepareDOM(document, domDocument, sections, "gsdl3"); //this.name, "gsdl3", this.field);
372 // System.out.println(docText);
373 }
374 }
375 if (docText == null) {
376 System.err.println("dom doc or sections was null - asking for doc text");
377 if (this.currentIndexField.equals("text")) {
378 //docText = Character.toString(END_OF_DOCUMENT) + document.getDocumentText();
379 docText = document.getDocumentText();
380 }
381 else {
382 StringBuffer textBuffer = new StringBuffer();
383 //textBuffer.append(END_OF_DOCUMENT);
384 List values = document.getDocumentMetadataItem("gsdl3", this.currentIndexField);
385 if (values != null) {
386 Iterator valueIter = values.iterator();
387 while (valueIter.hasNext()) {
388 String value = valueIter.next().toString();
389
390 textBuffer.append(value);
391 if (valueIter.hasNext()) {
392 //textBuffer.append(END_OF_SECTION);
393 // sectionSeqNo ++;
394 }
395 }
396 }
397 else {
398 textBuffer.append("No data");
399 }
400 docText = textBuffer.toString();
401 }
402 sectionSeqNo ++;
403 }
404
405
406 this.indexBuffer.append(docText);
407 // remember that we're not on the first document,
408 this.firstDocument = false;
409 // assign the sequence number on the first pass only, and increment the sequence number.
410 if (this.pass == 0) {
411 document.addDocumentMetadata("gsdl3", "mgseqno", this.overallName+"."+Integer.toString(startSeqNo));
412 }
413 this.documentSeqNo += 1;
414
415 return true;
416 }
417
418 /**
419 * Initialise the pass: open required files, check status
420 */
421 public boolean startPass(int passNumber) {
422
423
424 this.pass = passNumber;
425 this.firstDocument = true;
426 this.documentSeqNo = 1;
427 this.sectionSeqNo = 1;
428
429 this.mgPasses = new MGPassesWrapper();
430 this.indexBuffer = new StringBuffer();
431 int indexNo = (this.pass - 2) / 2;
432 MGIndex index = null;
433 if (this.pass >= 2) {
434 index = (MGIndex) this.indexes.get(indexNo);
435 if (index.hasError()) {
436 // an error has already occurred for this index, don't continue
437 System.out.println("pass "+this.pass+": aborted due to errors in the previous pass");
438 return false;
439 }
440 // attempt to ensure that the text subdirectory exists
441 this.indexDirectory = new File(outputDirectory, index.getName());
442 if (!indexDirectory.exists()) {
443 if (!indexDirectory.mkdir()) {
444 return false;
445 }
446 }
447 else if (!indexDirectory.isDirectory()) {
448 return false;
449 }
450
451 this.currentIndexLevel = index.getLevel();
452 this.currentIndexField = index.getField();
453 this.currentIndexName = index.getName();
454
455 if (this.currentIndexLevel == null || this.currentIndexField == null ) {
456 System.out.println("invalid index - level or field was null");
457 return false;
458 }
459 this.indexStem = this.indexDirectory.getPath() + File.separatorChar + INDEX_FILE_STEM; // TODO: modify for index
460 if (this.pass % 2 == 1) {
461 this.currentIndexName = null; // why???
462 }
463 }
464 else {
465
466 this.currentIndexField = "text";
467 this.currentIndexLevel = "section";
468 this.currentIndexName = null;
469 }
470
471 // get the parameters for this execution of mg_passes
472 mgPasses.setFileName((this.pass < 2 ? this.textDirectory.toString() : this.indexDirectory.toString())+File.separator+ "index");
473 if (!Misc.isWindows()) {
474 mgPasses.setBasePath("/");
475 }
476 int mgPass = this.pass < 2 ? this.pass : ((this.pass % 2) + 2);
477
478 mgPasses.setBufferSize(100000);
479
480 switch (mgPass) {
481 case 0:
482 // -b 100000 -T1
483 mgPasses.addPass(MGPassesWrapper.TEXT_PASS_1);
484
485
486 break;
487
488 case 1:
489 // -b 100000 -T2
490 mgPasses.addPass(MGPassesWrapper.TEXT_PASS_2);
491 break;
492
493 case 2:
494 // -b 100000 -2 -m 32 -s 0 -G -t 10 -N1
495 mgPasses.addPass(MGPassesWrapper.INDEX_PASS_1);
496 mgPasses.setInvfLevel(MGPassesWrapper.INVF_LEVEL_2);
497 mgPasses.setStemOptions(MGPassesWrapper.STEMMER_ENGLISH, MGPassesWrapper.NO_STEM_OR_CASE);
498 mgPasses.setInversionMemLimit(32);
499 mgPasses.ignoreSGMLTags(true);
500 break;
501
502 case 3:
503 // -b 100000 -2 -c 3 -G -t 10 -N2
504 mgPasses.addPass(MGPassesWrapper.INDEX_PASS_2);
505 mgPasses.setInvfLevel(MGPassesWrapper.INVF_LEVEL_2);
506 mgPasses.ignoreSGMLTags(true);
507 break;
508 }
509
510 mgPasses.init();
511 System.out.println("Pass " + this.pass);
512 return true;
513 }
514
515 /**
516 * Complete a pass - reset file counters, close files, etc.
517 */
518 public boolean endPass(int passNumber) {
519 Process p;
520
521 int indexNo = (passNumber - 2) / 2;
522 MGIndex index = null;
523 if (passNumber >= 2) {
524 index = (MGIndex) this.indexes.get(indexNo);
525 }
526 try {
527 this.indexBuffer.append(END_OF_DOCUMENT);
528 mgPasses.processDocument(indexBuffer.toString());
529 this.indexBuffer.delete(0, this.indexBuffer.length());
530 Thread.sleep(1000); // what for??
531 }
532 catch (InterruptedException ex) {
533 System.out.println(ex);
534 }
535 mgPasses.finish();
536 try {
537 Thread.sleep(1000);
538 } catch (Exception e) {}
539
540 int exit_value = 0;
541 System.out.println("Pass " + this.pass + " completed with " + exit_value);
542 if (exit_value !=0) {
543 //assume something has gone wrong, don't continue
544 if (index != null) {
545 index.setError(true);
546 return false;
547 }
548 }
549 int mgPass = this.pass < 2 ? this.pass : ((this.pass % 2) + 2);
550 String osextra = "";
551 if (!Misc.isWindows()) {
552 osextra = " -d / ";
553 }
554
555 switch (mgPass) {
556
557 case 0:
558 System.out.println("Compressing dictionary");
559 exit_value = Processing.runProcess("mg_compression_dict -f " + this.textDirectory.toString()+File.separator+"index" + osextra + " -S -H -2 -k 5120");
560 if (exit_value == 0) {
561 System.out.println("Compressed dictionary successfully written");
562 } else {
563 System.err.println("Error from mg_compression_dict: " + exit_value);
564 index.setError(true);
565
566 return false;
567 }
568 break;
569
570 case 2:
571 System.out.println("Creating perfect hash");
572 exit_value = Processing.runProcess("mg_perf_hash_build -f " + this.indexDirectory.toString()+File.separator+ "index"+osextra);
573 if (exit_value ==0) {
574 System.out.println("Perfect hashes completed");
575 } else {
576 System.err.println("Unable to build the perfect hash");
577 index.setError(true);
578 return false;
579 }
580 break;
581
582 case 3:
583 System.out.println("Writing weights file");
584 exit_value = Processing.runProcess("mg_weights_build -f " + this.indexStem + " -t " + this.textStem + osextra);
585 if (exit_value ==0) {
586 System.out.println("Weights file successfully written");
587 } else {
588 System.err.println("Unable to create weights file");
589 index.setError(true);
590 return false;
591 }
592
593 System.out.println("Creating inverted dictionary");
594 exit_value = Processing.runProcess("mg_invf_dict -f " + this.indexDirectory.toString()+File.separator+"index" + osextra);
595 if (exit_value ==0) {
596 System.out.println("Inverted dictionary file successfully written");
597 } else {
598 System.out.println("Unable to create inverted dictionary file");
599 index.setError(true);
600 return false;
601 }
602
603 System.out.println("Creating Stem indexes");
604 exit_value = Processing.runProcess("mg_stem_idx -b 4096 -s1 -f " + this.indexDirectory.toString()+File.separator+"index"+osextra);
605 if (exit_value == 0) {
606 System.out.println("Stemmed index 1 successfully written");
607 } else {
608 System.out.println("Unable to create stemmed index 1");
609 index.setError(true);
610 return false;
611 }
612
613 exit_value = Processing.runProcess("mg_stem_idx -b 4096 -s2 -f " + this.indexDirectory.toString()+File.separator+"index"+osextra);
614 if (exit_value == 0) {
615 System.out.println("Stemmed index 2 successfully written");
616 } else {
617 System.out.println("Unable to create stemmed index 2");
618 index.setError(true);
619 return false;
620 }
621 exit_value = Processing.runProcess("mg_stem_idx -b 4096 -s3 -f " + this.indexDirectory.toString()+File.separator+"index"+osextra);
622 if (exit_value == 0) {
623 System.out.println("Stemmed index 3 successfully written");
624 } else {
625 System.out.println("Unable to create stemmed index 3");
626 index.setError(true);
627 return false;
628 }
629
630 break;
631 } // switch
632
633 mgPasses = null;
634 return true;
635 }
636
637 /**
638 * Do any tidying up
639 */
640 public void tidyup()
641 {
642 }
643
644 /**
645 * Return the number of passes required for this index.
646 */
647 public int getNumberOfPasses()
648 { return 2 + this.indexes.size() * 2;
649 }
650
651 public boolean addServiceDescriptions(org.w3c.dom.Element service_rack_list) {
652 Document doc = service_rack_list.getOwnerDocument();
653
654 // generate the list of indexes
655 Element index_list = doc.createElement(GSXML.INDEX_ELEM+GSXML.LIST_MODIFIER);
656 boolean found_index = false;
657 String def_index = ""; // the default index will just be the first one created for now.
658 for (int i=0; i<this.indexes.size(); i++) {
659 MGIndex index = (MGIndex)this.indexes.get(i);
660 if (!index.hasError()) {
661 Element e = doc.createElement(GSXML.INDEX_ELEM);
662 e.setAttribute(GSXML.NAME_ATT, index.getName());
663 index_list.appendChild(e);
664 if (found_index == false) {
665 // this is the first index
666 found_index = true;
667 def_index = index.getName();
668 }
669 }
670 }
671
672 if (!found_index) {
673 // no indexes were able to be created, so we can't use them or the text
674 return false;
675 }
676 Element default_index = doc.createElement("defaultIndex");
677 default_index.setAttribute(GSXML.NAME_ATT, def_index);
678 Element base_index_name = doc.createElement("baseIndexPrefix");
679 base_index_name.setAttribute(GSXML.NAME_ATT, overallName);
680 Element index_stem = doc.createElement("indexStem");
681 index_stem.setAttribute(GSXML.NAME_ATT, "index");
682
683 Element search_service_elem = doc.createElement(GSXML.SERVICE_CLASS_ELEM);
684 Element retrieve_service_elem = doc.createElement(GSXML.SERVICE_CLASS_ELEM);
685 service_rack_list.appendChild(search_service_elem);
686 service_rack_list.appendChild(retrieve_service_elem);
687
688 search_service_elem.setAttribute(GSXML.NAME_ATT, "GS3MGSearch");
689 search_service_elem.appendChild(index_list);
690 search_service_elem.appendChild(default_index);
691 search_service_elem.appendChild(base_index_name);
692 search_service_elem.appendChild(index_stem);
693
694 retrieve_service_elem.setAttribute(GSXML.NAME_ATT, "GS3MGRetrieve");
695 retrieve_service_elem.appendChild(default_index.cloneNode(true));
696 retrieve_service_elem.appendChild(base_index_name.cloneNode(true));
697 retrieve_service_elem.appendChild(index_stem.cloneNode(true));
698
699 return true;
700 }
701
702}
703
Note: See TracBrowser for help on using the repository browser.