1 | package org.greenstone.gsdl3.gs3build.indexers;
|
---|
2 |
|
---|
3 | import org.greenstone.gsdl3.gs3build.doctypes.DocumentID;
|
---|
4 | import org.greenstone.gsdl3.gs3build.doctypes.DocumentInterface;
|
---|
5 |
|
---|
6 | import java.io.InputStream;
|
---|
7 | import java.io.OutputStream;
|
---|
8 | import java.io.IOException;
|
---|
9 |
|
---|
10 | public class MGPPIndexer extends AbstractIndexer
|
---|
11 | {
|
---|
12 | int pass;
|
---|
13 | String name;
|
---|
14 | boolean firstDocument;
|
---|
15 | String outputDirectory;
|
---|
16 | String outputStem;
|
---|
17 | InputStream indexerFeedback;
|
---|
18 | InputStream indexerErrors;
|
---|
19 | OutputStream indexerTextfeed;
|
---|
20 | Process mgpp_passes;
|
---|
21 | static final String documentSeparator = "<Document>";
|
---|
22 |
|
---|
23 | public static final String MGPP_INDEX_TYPE = "mgpp";
|
---|
24 |
|
---|
25 | public MGPPIndexer(String name)
|
---|
26 | { this.name = name;
|
---|
27 | }
|
---|
28 |
|
---|
29 | public String getName()
|
---|
30 | { return this.name;
|
---|
31 | }
|
---|
32 |
|
---|
33 | /**
|
---|
34 | * The output directory should be (collection)/building/text/ for
|
---|
35 | * normal Greenstone builds
|
---|
36 | */
|
---|
37 | public boolean configure(String label, String value)
|
---|
38 | {
|
---|
39 | if (label.equals("outputDir")) {
|
---|
40 | this.outputDirectory = value;
|
---|
41 | this.outputStem = value + "/index";
|
---|
42 | }
|
---|
43 | this.pass = 0;
|
---|
44 | return true;
|
---|
45 | }
|
---|
46 |
|
---|
47 | public String getIndexType()
|
---|
48 | { return MGPP_INDEX_TYPE;
|
---|
49 | }
|
---|
50 |
|
---|
51 | public boolean addIndex(String name, String level, String field)
|
---|
52 | { return true;
|
---|
53 | }
|
---|
54 |
|
---|
55 | /**
|
---|
56 | * Index a single document; the document interface can be used to extract individual
|
---|
57 | * metadata items etc. as required or desired and index those instead or as well as
|
---|
58 | * the body text of the document.
|
---|
59 | */
|
---|
60 | public boolean indexDocument(DocumentID docID, DocumentInterface document)
|
---|
61 | {
|
---|
62 | if (!this.firstDocument)
|
---|
63 | { // Send a '<document>' before the document itself
|
---|
64 | try {
|
---|
65 | this.indexerTextfeed.write(documentSeparator.getBytes(), 0, documentSeparator.getBytes().length);
|
---|
66 | }
|
---|
67 | catch (IOException ex)
|
---|
68 | { System.out.println("Bad output on end of document" + ex);
|
---|
69 | ex.printStackTrace();
|
---|
70 | return false;
|
---|
71 | }
|
---|
72 | }
|
---|
73 | String docText = document.getDocumentText();
|
---|
74 |
|
---|
75 | byte [] bytes = docText.getBytes();
|
---|
76 | int pos = 0, end = bytes.length;
|
---|
77 |
|
---|
78 | try {
|
---|
79 | while (pos < end) {
|
---|
80 | this.indexerTextfeed.write(bytes, pos, (end - pos > 512 ? 512 : end - pos));
|
---|
81 | pos = pos + 512;
|
---|
82 |
|
---|
83 | try {
|
---|
84 | while (this.indexerFeedback.available() > 0)
|
---|
85 | { byte b[] = new byte[this.indexerFeedback.available()];
|
---|
86 | System.out.println("Feedback of " + this.indexerFeedback.available());
|
---|
87 | this.indexerFeedback.read(b);
|
---|
88 | System.out.println(b);
|
---|
89 | }
|
---|
90 | }
|
---|
91 | catch (IOException ex)
|
---|
92 | {
|
---|
93 | }
|
---|
94 |
|
---|
95 |
|
---|
96 | try {
|
---|
97 | while (this.indexerErrors.available() > 0)
|
---|
98 | { byte b[] = new byte[this.indexerErrors.available()];
|
---|
99 | System.out.println("Feedback of " + this.indexerErrors.available());
|
---|
100 | this.indexerErrors.read(b);
|
---|
101 | System.out.println(new String(b));
|
---|
102 | }
|
---|
103 | }
|
---|
104 | catch (IOException ex)
|
---|
105 | {
|
---|
106 | }
|
---|
107 | }
|
---|
108 | }
|
---|
109 | catch (IOException ex)
|
---|
110 | { System.out.println("Bad output during document write " + ex + " " + pos + " " + end);
|
---|
111 | ex.printStackTrace();
|
---|
112 | return false;
|
---|
113 | }
|
---|
114 | this.firstDocument = false;
|
---|
115 |
|
---|
116 | try {
|
---|
117 | while (this.indexerErrors.available() > 0)
|
---|
118 | { char c = (char) this.indexerErrors.read();
|
---|
119 | System.out.println(c);
|
---|
120 | }
|
---|
121 | while (this.indexerFeedback.available() > 0)
|
---|
122 | { byte b[] = new byte[this.indexerFeedback.available()];
|
---|
123 | System.out.println("Feedback of " + this.indexerFeedback.available());
|
---|
124 | this.indexerFeedback.read(b);
|
---|
125 | }
|
---|
126 | }
|
---|
127 | catch (IOException ex)
|
---|
128 | {
|
---|
129 | }
|
---|
130 | return true;
|
---|
131 | }
|
---|
132 |
|
---|
133 | /**
|
---|
134 | * Initialise the pass: open required files, check status
|
---|
135 | */
|
---|
136 | public boolean startPass(int passNumber)
|
---|
137 | { this.pass = passNumber;
|
---|
138 | this.firstDocument = true;
|
---|
139 |
|
---|
140 | try {
|
---|
141 | switch (this.pass) {
|
---|
142 | case 0:
|
---|
143 | mgpp_passes = Runtime.getRuntime().exec("mgpp_passes -f " + this.outputStem + " -T1");
|
---|
144 | break;
|
---|
145 |
|
---|
146 | case 1:
|
---|
147 | mgpp_passes = Runtime.getRuntime().exec("mgpp_passes -f " + this.outputStem +" -T2");
|
---|
148 | break;
|
---|
149 |
|
---|
150 | case 2:
|
---|
151 | mgpp_passes = Runtime.getRuntime().exec("mgpp_passes -f " + this.outputStem +" -I1");
|
---|
152 | break;
|
---|
153 |
|
---|
154 | case 3:
|
---|
155 | Process p = Runtime.getRuntime().exec("mgpp_perf_hash_build -f " + this.outputStem);
|
---|
156 | p.waitFor();
|
---|
157 |
|
---|
158 | mgpp_passes = Runtime.getRuntime().exec("mgpp_passes -f " + this.outputStem +" -I2");
|
---|
159 | break;
|
---|
160 | }
|
---|
161 |
|
---|
162 | this.indexerFeedback = mgpp_passes.getInputStream();
|
---|
163 | this.indexerErrors = mgpp_passes.getErrorStream();
|
---|
164 | this.indexerTextfeed = mgpp_passes.getOutputStream();
|
---|
165 | }
|
---|
166 | catch (IOException ex)
|
---|
167 | { System.out.println(ex);
|
---|
168 | ex.printStackTrace();
|
---|
169 | return false;
|
---|
170 | }
|
---|
171 | catch (InterruptedException ex)
|
---|
172 | { System.out.println(ex);
|
---|
173 | ex.printStackTrace();
|
---|
174 | return false;
|
---|
175 | }
|
---|
176 | System.out.println("Pass " + this.pass);
|
---|
177 | return true;
|
---|
178 | }
|
---|
179 |
|
---|
180 | /**
|
---|
181 | * Complete a pass - reset file counters, close files, etc.
|
---|
182 | */
|
---|
183 | public boolean endPass(int passNumber)
|
---|
184 | { // TODO: end pass
|
---|
185 | Process p;
|
---|
186 |
|
---|
187 | try {
|
---|
188 | this.indexerTextfeed.write((char) 2);
|
---|
189 | this.indexerTextfeed.write(4);
|
---|
190 | while (this.indexerErrors.available() > 0)
|
---|
191 | { char c = (char) this.indexerErrors.read();
|
---|
192 | System.out.print(c);
|
---|
193 | }
|
---|
194 | while (this.indexerFeedback.available() > 0)
|
---|
195 | { byte b[] = new byte[this.indexerFeedback.available()];
|
---|
196 | System.out.print("Feedback of " + this.indexerFeedback.available());
|
---|
197 | this.indexerFeedback.read(b);
|
---|
198 | }
|
---|
199 |
|
---|
200 | this.indexerTextfeed.close();
|
---|
201 | Thread.sleep(1000);
|
---|
202 | this.mgpp_passes.waitFor();
|
---|
203 | }
|
---|
204 | catch (IOException ex)
|
---|
205 | { System.out.println(ex);
|
---|
206 | }
|
---|
207 | catch (InterruptedException ex)
|
---|
208 | { System.out.println(ex);
|
---|
209 | }
|
---|
210 | System.out.println("Completed with " + this.mgpp_passes.exitValue());
|
---|
211 |
|
---|
212 | try {
|
---|
213 | switch (this.pass)
|
---|
214 | {
|
---|
215 | case 0:
|
---|
216 | System.out.println("Compressing dictionary");
|
---|
217 | p = Runtime.getRuntime().exec("mgpp_compression_dict -f " + this.outputStem + " -S -H -2 -k 5120");
|
---|
218 | p.waitFor();
|
---|
219 | System.out.println(p.exitValue());
|
---|
220 | break;
|
---|
221 |
|
---|
222 | case 3:
|
---|
223 | p = Runtime.getRuntime().exec("mgpp_weights_build -f " + this.outputStem);
|
---|
224 | p.waitFor();
|
---|
225 | System.out.println(p.exitValue());
|
---|
226 |
|
---|
227 | p = Runtime.getRuntime().exec("mgpp_invf_dict -f " + this.outputStem);
|
---|
228 | p.waitFor();
|
---|
229 | System.out.println(p.exitValue());
|
---|
230 |
|
---|
231 | p = Runtime.getRuntime().exec("mgpp_stem_idx -b 4096 -s1 -f " + this.outputStem + " -d " + this.outputDirectory);
|
---|
232 | p.waitFor();
|
---|
233 | System.out.println(p.exitValue());
|
---|
234 | p = Runtime.getRuntime().exec("mgpp_stem_idx -b 4096 -s2 -f " + this.outputStem + " -d " + this.outputDirectory);
|
---|
235 | p.waitFor();
|
---|
236 | System.out.println(p.exitValue());
|
---|
237 | p = Runtime.getRuntime().exec("mgpp_stem_idx -b 4096 -s3 -f " + this.outputStem + " -d " + this.outputDirectory);
|
---|
238 | p.waitFor();
|
---|
239 | System.out.println(p.exitValue());
|
---|
240 | break;
|
---|
241 | }
|
---|
242 | }
|
---|
243 | catch (IOException ex)
|
---|
244 | { System.out.println(ex);
|
---|
245 | ex.printStackTrace();
|
---|
246 | return false;
|
---|
247 | }
|
---|
248 | catch (InterruptedException ex)
|
---|
249 | { System.out.println(ex);
|
---|
250 | ex.printStackTrace();
|
---|
251 | return false;
|
---|
252 | }
|
---|
253 | return true;
|
---|
254 | }
|
---|
255 |
|
---|
256 | /**
|
---|
257 | * Do any tidying up
|
---|
258 | */
|
---|
259 | public void tidyup()
|
---|
260 | {
|
---|
261 | }
|
---|
262 |
|
---|
263 | /**
|
---|
264 | * Return the number of passes required for this index.
|
---|
265 | */
|
---|
266 | public int getNumberOfPasses()
|
---|
267 | { return 4;
|
---|
268 | }
|
---|
269 | }
|
---|