1 | package org.greenstone.gsdl3.gs3build.indexers;
|
---|
2 |
|
---|
3 | import java.util.List;
|
---|
4 | import java.util.ArrayList;
|
---|
5 | import java.util.Iterator;
|
---|
6 |
|
---|
7 | import java.io.File;
|
---|
8 | import java.io.InputStream;
|
---|
9 | import java.io.OutputStream;
|
---|
10 | import java.io.IOException;
|
---|
11 |
|
---|
12 | import org.w3c.dom.*;
|
---|
13 |
|
---|
14 | import org.greenstone.gsdl3.gs3build.doctypes.DocumentID;
|
---|
15 | import org.greenstone.gsdl3.gs3build.doctypes.DocumentInterface;
|
---|
16 | import org.greenstone.gsdl3.gs3build.doctypes.HTMLDocument;
|
---|
17 | import org.greenstone.gsdl3.gs3build.metadata.*;
|
---|
18 | import org.greenstone.gsdl3.gs3build.xpointer.XPointer;
|
---|
19 |
|
---|
20 | public class MGIndexer implements IndexerInterface
|
---|
21 | {
|
---|
22 | int pass;
|
---|
23 | int documentSeqNo;
|
---|
24 | int sectionSeqNo;
|
---|
25 | boolean firstDocument;
|
---|
26 | String outputDirectory;
|
---|
27 | InputStream indexerFeedback;
|
---|
28 | InputStream indexerErrors;
|
---|
29 | OutputStream indexerTextfeed;
|
---|
30 | Process mg_passes;
|
---|
31 | File textDirectory;
|
---|
32 | File indexDirectory;
|
---|
33 | String indexStem;
|
---|
34 | String textStem;
|
---|
35 | List indexes;
|
---|
36 | String level;
|
---|
37 |
|
---|
38 | public static final String MG_INDEX = "Index";
|
---|
39 |
|
---|
40 | class MGIndex
|
---|
41 | { String level;
|
---|
42 | String field;
|
---|
43 |
|
---|
44 | public MGIndex(String indexLabel)
|
---|
45 | { int colonAt = indexLabel.indexOf(':');
|
---|
46 | if (colonAt >= 0)
|
---|
47 | { field = indexLabel.substring(colonAt+1);
|
---|
48 | level = indexLabel.substring(0, colonAt);
|
---|
49 | }
|
---|
50 | else
|
---|
51 | {
|
---|
52 | }
|
---|
53 | }
|
---|
54 |
|
---|
55 | public String getLevel()
|
---|
56 | { return this.level;
|
---|
57 | }
|
---|
58 |
|
---|
59 | public String getField()
|
---|
60 | { return this.field;
|
---|
61 | }
|
---|
62 | }
|
---|
63 |
|
---|
64 | public MGIndexer()
|
---|
65 | { this.indexes = new ArrayList();
|
---|
66 | }
|
---|
67 |
|
---|
68 | private String getIndexDirectory(String level, String field)
|
---|
69 | { StringBuffer directory = new StringBuffer();
|
---|
70 | directory.append(Character.toLowerCase((char) level.charAt(0)));
|
---|
71 |
|
---|
72 | int c, w;
|
---|
73 | w = 0;
|
---|
74 | c = 0;
|
---|
75 | while (c < field.length() && w < 2) {
|
---|
76 | char ch = field.charAt(c);
|
---|
77 |
|
---|
78 | ch = Character.toLowerCase(ch);
|
---|
79 | if (Character.isLetter(ch)) {
|
---|
80 | if (ch != 'a' && ch != 'e' && ch != 'i' &&
|
---|
81 | ch != 'o' && ch != 'u') {
|
---|
82 | directory.append(ch);
|
---|
83 | w++;
|
---|
84 | }
|
---|
85 | }
|
---|
86 | c ++;
|
---|
87 | }
|
---|
88 | return directory.toString();
|
---|
89 | }
|
---|
90 |
|
---|
91 | /**
|
---|
92 | * The output directory should be (collection)/building/text/ for
|
---|
93 | * normal Greenstone builds
|
---|
94 | */
|
---|
95 | public boolean configure(String label, String value)
|
---|
96 | {
|
---|
97 | if (label.equals(IndexerManager.outputDir)) {
|
---|
98 | this.outputDirectory = value;
|
---|
99 | this.textStem = value + "/text/index";
|
---|
100 | this.pass = 0;
|
---|
101 |
|
---|
102 | // attempt to ensure that the text subdirectory exists
|
---|
103 | this.textDirectory = new File(outputDirectory, "text");
|
---|
104 | if (!textDirectory.exists()) {
|
---|
105 | if (!textDirectory.mkdir()) {
|
---|
106 | return false;
|
---|
107 | }
|
---|
108 | }
|
---|
109 | else if (!textDirectory.isDirectory()) {
|
---|
110 | return false;
|
---|
111 | }
|
---|
112 |
|
---|
113 | // Sign to the user which mg directory is being used...
|
---|
114 | System.out.println("Output MG directory is " + this.textStem);
|
---|
115 | }
|
---|
116 | else if (label.equals(MG_INDEX)) {
|
---|
117 | this.indexes.add(new MGIndex(value));
|
---|
118 | }
|
---|
119 |
|
---|
120 | return true;
|
---|
121 | }
|
---|
122 |
|
---|
123 | private Node recurseDOM(DocumentInterface metsDoc, Node node,
|
---|
124 | AbstractStructure structure, StringBuffer buffer)
|
---|
125 | {
|
---|
126 | // send out the ctrl-c...if this is
|
---|
127 | if (structure.getStructureType().equals(METSDivision.DIVISION_TYPE)) {
|
---|
128 | if (this.pass == 0) {
|
---|
129 | METSDivision division = (METSDivision) structure;
|
---|
130 |
|
---|
131 | // get the division metadata block
|
---|
132 | METSDescriptive descriptive;
|
---|
133 | String metadataId = division.getDefaultMetadataReference();
|
---|
134 | if (metadataId == null) {
|
---|
135 | descriptive = metsDoc.getDocumentMetadata().createDescriptive(division.getLabel());
|
---|
136 | division.addMetadataReference(descriptive.getID());
|
---|
137 | }
|
---|
138 | else {
|
---|
139 | // Get the descriptive item...
|
---|
140 | descriptive = metsDoc.getDocumentMetadata().getDescriptiveById(metadataId);
|
---|
141 | }
|
---|
142 |
|
---|
143 | descriptive.addMetadata("gsdl3", "mgseqno", Integer.toString(this.sectionSeqNo));
|
---|
144 | }
|
---|
145 |
|
---|
146 | buffer.append((char) 3);
|
---|
147 | if (this.level != null &&
|
---|
148 | this.level.equals("section")) {
|
---|
149 | buffer.append((char) 2);
|
---|
150 | }
|
---|
151 | this.sectionSeqNo ++;
|
---|
152 | }
|
---|
153 |
|
---|
154 | // go through our children as required...
|
---|
155 | Iterator children = structure.getChildIterator();
|
---|
156 | while (children.hasNext()) {
|
---|
157 | AbstractStructure child = (AbstractStructure) children.next();
|
---|
158 |
|
---|
159 | // get xpointer for child
|
---|
160 | // get start position node
|
---|
161 | Node startNode = ((HTMLDocument) metsDoc).getSectionStartNode((METSDivision) child);
|
---|
162 |
|
---|
163 | // while this node isn't the child's start node, produce the node
|
---|
164 | while (node != startNode) {
|
---|
165 | XPointer.printNode(node, buffer, false);
|
---|
166 | // print buffer to node
|
---|
167 | node = XPointer.getNextNode(node, buffer);
|
---|
168 | }
|
---|
169 |
|
---|
170 | // recurse to child
|
---|
171 | this.recurseDOM(metsDoc, node, child, buffer);
|
---|
172 | }
|
---|
173 |
|
---|
174 | // close a document - the actual closing \B will be done by the main
|
---|
175 | // loop, so only a required \C is printed here...
|
---|
176 | if (structure.getStructureType().equals(METSStructure.STRUCTURE_TYPE)) {
|
---|
177 | while (node != null) {
|
---|
178 | XPointer.printNode(node, buffer, false);
|
---|
179 | node = XPointer.getNextNode(node, buffer);
|
---|
180 | }
|
---|
181 | buffer.append((char) 3);
|
---|
182 | this.sectionSeqNo ++;
|
---|
183 | }
|
---|
184 | return node;
|
---|
185 | }
|
---|
186 |
|
---|
187 | private String prepareDOM(DocumentInterface metsDoc, Document document, METSStructure structure)
|
---|
188 | { Node node = document.getDocumentElement();
|
---|
189 | StringBuffer textBuffer = new StringBuffer();
|
---|
190 |
|
---|
191 | this.recurseDOM(metsDoc, node, structure, textBuffer);
|
---|
192 | return textBuffer.toString();
|
---|
193 | }
|
---|
194 |
|
---|
195 | /**
|
---|
196 | * Index a single document; the document interface can be used to extract individual
|
---|
197 | * metadata items etc. as required or desired and index those instead or as well as
|
---|
198 | * the body text of the document.
|
---|
199 | */
|
---|
200 | public boolean indexDocument(DocumentID docID, DocumentInterface document)
|
---|
201 | {
|
---|
202 | if (!this.firstDocument)
|
---|
203 | { // Send a 'CTRL-B' before the document itself
|
---|
204 | try {
|
---|
205 | this.indexerTextfeed.write(2);
|
---|
206 | }
|
---|
207 | catch (IOException ex)
|
---|
208 | { System.out.println("Bad output on end of document" + ex);
|
---|
209 | ex.printStackTrace();
|
---|
210 | return false;
|
---|
211 | }
|
---|
212 | }
|
---|
213 | String docText = null;
|
---|
214 |
|
---|
215 | Document domDocument = document.getDOMDocument();
|
---|
216 | if (domDocument != null) {
|
---|
217 | METSStructure sections = document.getDocumentStructure().getStructure("Section");
|
---|
218 | if (sections != null) {
|
---|
219 | docText = this.prepareDOM(document, domDocument, sections);
|
---|
220 | // System.out.println(docText);
|
---|
221 | }
|
---|
222 | }
|
---|
223 | if (docText == null) {
|
---|
224 | docText = document.getDocumentText();
|
---|
225 | }
|
---|
226 |
|
---|
227 | byte [] bytes = docText.getBytes();
|
---|
228 | int pos = 0, end = bytes.length;
|
---|
229 |
|
---|
230 | try {
|
---|
231 | while (pos < end) {
|
---|
232 | this.indexerTextfeed.write(bytes, pos, (end - pos > 512 ? 512 : end - pos));
|
---|
233 | pos = pos + 512;
|
---|
234 |
|
---|
235 | try {
|
---|
236 | while (this.indexerFeedback.available() > 0)
|
---|
237 | { byte b[] = new byte[this.indexerFeedback.available()];
|
---|
238 | System.out.println("Feedback of " + this.indexerFeedback.available());
|
---|
239 | this.indexerFeedback.read(b);
|
---|
240 | System.out.println(b);
|
---|
241 | }
|
---|
242 | }
|
---|
243 | catch (IOException ex)
|
---|
244 | {
|
---|
245 | }
|
---|
246 |
|
---|
247 |
|
---|
248 | try {
|
---|
249 | while (this.indexerErrors.available() > 0)
|
---|
250 | { byte b[] = new byte[this.indexerErrors.available()];
|
---|
251 | System.out.println("Feedback of " + this.indexerErrors.available());
|
---|
252 | this.indexerErrors.read(b);
|
---|
253 | System.out.println(new String(b));
|
---|
254 | }
|
---|
255 | }
|
---|
256 | catch (IOException ex)
|
---|
257 | {
|
---|
258 | }
|
---|
259 | }
|
---|
260 | }
|
---|
261 | catch (IOException ex)
|
---|
262 | { System.out.println("Bad output during document write " + ex + " " + pos + " " + end);
|
---|
263 | ex.printStackTrace();
|
---|
264 | return false;
|
---|
265 | }
|
---|
266 |
|
---|
267 | // remember that we're not on the first document, assign the sequence number
|
---|
268 | // on the first pass only, and increment the sequence number.
|
---|
269 | this.firstDocument = false;
|
---|
270 | if (this.pass == 0) {
|
---|
271 | document.addDocumentMetadata("gsdl3", "mgseqno", Integer.toString(this.documentSeqNo));
|
---|
272 | }
|
---|
273 | this.documentSeqNo += 1;
|
---|
274 |
|
---|
275 | try {
|
---|
276 | while (this.indexerErrors.available() > 0)
|
---|
277 | { char c = (char) this.indexerErrors.read();
|
---|
278 | System.out.println(c);
|
---|
279 | }
|
---|
280 | while (this.indexerFeedback.available() > 0)
|
---|
281 | { byte b[] = new byte[this.indexerFeedback.available()];
|
---|
282 | System.out.println("Feedback of " + this.indexerFeedback.available());
|
---|
283 | this.indexerFeedback.read(b);
|
---|
284 | }
|
---|
285 | }
|
---|
286 | catch (IOException ex)
|
---|
287 | {
|
---|
288 | }
|
---|
289 | return true;
|
---|
290 | }
|
---|
291 |
|
---|
292 | /**
|
---|
293 | * Initialise the pass: open required files, check status
|
---|
294 | */
|
---|
295 | public boolean startPass(int passNumber)
|
---|
296 | {
|
---|
297 | this.pass = passNumber;
|
---|
298 | this.firstDocument = true;
|
---|
299 | this.documentSeqNo = 1;
|
---|
300 | this.sectionSeqNo = 1;
|
---|
301 |
|
---|
302 | int indexNo = (this.pass - 2) / 2;
|
---|
303 | if (indexNo >= 0) {
|
---|
304 | MGIndex index = (MGIndex) this.indexes.get(indexNo);
|
---|
305 |
|
---|
306 | // attempt to ensure that the text subdirectory exists
|
---|
307 | this.indexDirectory = new File(outputDirectory, this.getIndexDirectory(index.getLevel(), index.getField()));
|
---|
308 | if (!indexDirectory.exists()) {
|
---|
309 | if (!indexDirectory.mkdir()) {
|
---|
310 | return false;
|
---|
311 | }
|
---|
312 | }
|
---|
313 | else if (!indexDirectory.isDirectory()) {
|
---|
314 | return false;
|
---|
315 | }
|
---|
316 |
|
---|
317 | this.indexStem = this.outputDirectory + File.separatorChar +
|
---|
318 | this.getIndexDirectory("document", "text") +
|
---|
319 | File.separatorChar + "index"; // TODO: modify for index
|
---|
320 | this.level = index.getLevel();
|
---|
321 | }
|
---|
322 |
|
---|
323 | // get the parameters for this execution of mg_passes
|
---|
324 | String pathParams = "-f index -d " + (this.pass < 2 ? this.textDirectory.toString() : this.indexDirectory.toString());
|
---|
325 |
|
---|
326 | int mgPass = this.pass < 2 ? this.pass : ((this.pass % 2) + 2);
|
---|
327 |
|
---|
328 | try {
|
---|
329 | switch (mgPass) {
|
---|
330 | case 0:
|
---|
331 | mg_passes = Runtime.getRuntime().exec("mg_passes " + pathParams + " -b 100000 -T1");
|
---|
332 | break;
|
---|
333 |
|
---|
334 | case 1:
|
---|
335 | mg_passes = Runtime.getRuntime().exec("mg_passes " + pathParams +" -b 100000 -T2");
|
---|
336 | break;
|
---|
337 |
|
---|
338 | case 2:
|
---|
339 | mg_passes = Runtime.getRuntime().exec("mg_passes " + pathParams + " -b 100000 -2 -m 32 -s 0 -G -t 10 -N1");
|
---|
340 | break;
|
---|
341 |
|
---|
342 | case 3:
|
---|
343 | Process p = Runtime.getRuntime().exec("mg_perf_hash_build -f index -d " + this.indexDirectory.toString());
|
---|
344 | p.waitFor();
|
---|
345 | System.out.println(p.exitValue());
|
---|
346 |
|
---|
347 | mg_passes = Runtime.getRuntime().exec("mg_passes " + pathParams +" -b 100000 -2 -c 3 -G -t 10 -N2");
|
---|
348 | break;
|
---|
349 | }
|
---|
350 |
|
---|
351 | this.indexerFeedback = mg_passes.getInputStream();
|
---|
352 | this.indexerErrors = mg_passes.getErrorStream();
|
---|
353 | this.indexerTextfeed = mg_passes.getOutputStream();
|
---|
354 | }
|
---|
355 | catch (IOException ex)
|
---|
356 | { System.out.println(ex);
|
---|
357 | ex.printStackTrace();
|
---|
358 | return false;
|
---|
359 | }
|
---|
360 | catch (InterruptedException ex)
|
---|
361 | { System.out.println(ex);
|
---|
362 | ex.printStackTrace();
|
---|
363 | return false;
|
---|
364 | }
|
---|
365 | System.out.println("Pass " + this.pass);
|
---|
366 | return true;
|
---|
367 | }
|
---|
368 |
|
---|
369 | /**
|
---|
370 | * Complete a pass - reset file counters, close files, etc.
|
---|
371 | */
|
---|
372 | public boolean endPass(int passNumber)
|
---|
373 | { Process p;
|
---|
374 |
|
---|
375 | try {
|
---|
376 | this.indexerTextfeed.write((char) 2);
|
---|
377 | this.indexerTextfeed.write(4);
|
---|
378 | while (this.indexerErrors.available() > 0)
|
---|
379 | { char c = (char) this.indexerErrors.read();
|
---|
380 | System.out.print(c);
|
---|
381 | }
|
---|
382 | while (this.indexerFeedback.available() > 0)
|
---|
383 | { byte b[] = new byte[this.indexerFeedback.available()];
|
---|
384 | System.out.print("Feedback of " + this.indexerFeedback.available());
|
---|
385 | this.indexerFeedback.read(b);
|
---|
386 | }
|
---|
387 |
|
---|
388 | this.indexerTextfeed.close();
|
---|
389 | Thread.sleep(1000);
|
---|
390 | this.mg_passes.waitFor();
|
---|
391 | }
|
---|
392 | catch (IOException ex)
|
---|
393 | { System.out.println(ex);
|
---|
394 | }
|
---|
395 | catch (InterruptedException ex)
|
---|
396 | { System.out.println(ex);
|
---|
397 | }
|
---|
398 | System.out.println("Completed with " + this.mg_passes.exitValue());
|
---|
399 |
|
---|
400 | int mgPass = this.pass < 2 ? this.pass : ((this.pass % 2) + 2);
|
---|
401 |
|
---|
402 | try {
|
---|
403 | switch (mgPass)
|
---|
404 | {
|
---|
405 | case 0:
|
---|
406 | System.out.println("Compressing dictionary");
|
---|
407 | p = Runtime.getRuntime().exec("mg_compression_dict -f index -d " + this.textDirectory.toString() + " -S -H -2 -k 5120");
|
---|
408 | p.waitFor();
|
---|
409 | if (p.exitValue() != 0) {
|
---|
410 | System.out.println("Error from mg_compression_dict: " + p.exitValue());
|
---|
411 | }
|
---|
412 | break;
|
---|
413 |
|
---|
414 | case 3:
|
---|
415 | p = Runtime.getRuntime().exec("mg_weights_build -f " + this.indexStem + " -t " + this.textStem + " -d /");
|
---|
416 | p.waitFor();
|
---|
417 | System.out.println(p.exitValue());
|
---|
418 |
|
---|
419 | p = Runtime.getRuntime().exec("mg_invf_dict -f index -d " + this.indexDirectory.toString());
|
---|
420 | p.waitFor();
|
---|
421 | System.out.println(p.exitValue());
|
---|
422 |
|
---|
423 | p = Runtime.getRuntime().exec("mg_stem_idx -b 4096 -s1 -f index -d " + this.indexDirectory.toString());
|
---|
424 | p.waitFor();
|
---|
425 | System.out.println(p.exitValue());
|
---|
426 | p = Runtime.getRuntime().exec("mg_stem_idx -b 4096 -s2 -f index -d " + this.indexDirectory.toString());
|
---|
427 | p.waitFor();
|
---|
428 | System.out.println(p.exitValue());
|
---|
429 | p = Runtime.getRuntime().exec("mg_stem_idx -b 4096 -s3 -f index -d " + this.indexDirectory.toString());
|
---|
430 | p.waitFor();
|
---|
431 | System.out.println(p.exitValue());
|
---|
432 | break;
|
---|
433 | }
|
---|
434 | }
|
---|
435 | catch (IOException ex)
|
---|
436 | { System.out.println(ex);
|
---|
437 | ex.printStackTrace();
|
---|
438 | return false;
|
---|
439 | }
|
---|
440 | catch (InterruptedException ex)
|
---|
441 | { System.out.println(ex);
|
---|
442 | ex.printStackTrace();
|
---|
443 | return false;
|
---|
444 | }
|
---|
445 | return true;
|
---|
446 | }
|
---|
447 |
|
---|
448 | /**
|
---|
449 | * Do any tidying up
|
---|
450 | */
|
---|
451 | public void tidyup()
|
---|
452 | {
|
---|
453 | }
|
---|
454 |
|
---|
455 | /**
|
---|
456 | * Return the number of passes required for this index.
|
---|
457 | */
|
---|
458 | public int getNumberOfPasses()
|
---|
459 | { return 2 + this.indexes.size() * 2;
|
---|
460 | }
|
---|
461 | }
|
---|