1 | package org.nzdl.gsdl.LuceneWrap;
|
---|
2 |
|
---|
3 |
|
---|
4 | import org.xml.sax.Attributes;
|
---|
5 | import org.xml.sax.helpers.DefaultHandler;
|
---|
6 | import org.xml.sax.InputSource;
|
---|
7 | import org.xml.sax.SAXException;
|
---|
8 | import org.xml.sax.XMLReader;
|
---|
9 |
|
---|
10 | import javax.xml.parsers.SAXParser;
|
---|
11 | import javax.xml.parsers.SAXParserFactory;
|
---|
12 |
|
---|
13 | import org.apache.lucene.document.Document;
|
---|
14 | import org.apache.lucene.document.Field;
|
---|
15 | import org.apache.lucene.index.IndexWriter;
|
---|
16 | import org.apache.lucene.analysis.standard.StandardAnalyzer;
|
---|
17 |
|
---|
18 | import java.util.Stack;
|
---|
19 | import java.io.FileInputStream;
|
---|
20 | import java.io.File;
|
---|
21 | import java.io.StringReader;
|
---|
22 | import java.net.URL;
|
---|
23 |
|
---|
24 |
|
---|
25 |
|
---|
26 | public class Indexer extends DefaultHandler
|
---|
27 | {
|
---|
28 | IndexWriter writer_ = null;
|
---|
29 | SAXParser sax_parser_ = null;
|
---|
30 | String doc_tag_level_ = null;
|
---|
31 |
|
---|
32 | Stack stack_ = null;
|
---|
33 | String path_ = "";
|
---|
34 |
|
---|
35 | Document current_doc_ = null;
|
---|
36 | String current_node_ = "";
|
---|
37 | String indexable_current_node_ = "";
|
---|
38 | String current_contents_ = "";
|
---|
39 |
|
---|
40 | protected String file_id_ = null;
|
---|
41 |
|
---|
42 | /** pass in true if want to create a new index, false if want to use the existing one */
|
---|
43 | public Indexer (String doc_tag_level, File index_dir, boolean create)
|
---|
44 | {
|
---|
45 | doc_tag_level_ = doc_tag_level;
|
---|
46 |
|
---|
47 | try {
|
---|
48 | stack_ = new Stack();
|
---|
49 | SAXParserFactory sax_factory = SAXParserFactory.newInstance();
|
---|
50 | sax_parser_ = sax_factory.newSAXParser();
|
---|
51 |
|
---|
52 | XMLReader reader = sax_parser_.getXMLReader();
|
---|
53 | reader.setFeature("http://xml.org/sax/features/validation", false);
|
---|
54 |
|
---|
55 | writer_ = new IndexWriter(index_dir.getPath(), new StandardAnalyzer(), create);
|
---|
56 | // by default, will only index 10,000 words per document
|
---|
57 | // Can throw out_of_memory errors
|
---|
58 | writer_.setMaxFieldLength(Integer.MAX_VALUE);
|
---|
59 | if (create) {
|
---|
60 | writer_.optimize();
|
---|
61 | }
|
---|
62 |
|
---|
63 | } catch (Exception e) {
|
---|
64 | // do nothing!
|
---|
65 | }
|
---|
66 | }
|
---|
67 |
|
---|
68 | /** index one document */
|
---|
69 | public void index (String file_id, File file)
|
---|
70 | {
|
---|
71 | file_id_ = file_id;
|
---|
72 | path_ = "";
|
---|
73 | String base_path = file.getPath();
|
---|
74 | base_path = base_path.substring(0, base_path.lastIndexOf(File.separatorChar));
|
---|
75 |
|
---|
76 | try {
|
---|
77 | sax_parser_.parse(new InputSource(new FileInputStream(file)), this);
|
---|
78 | }
|
---|
79 | catch (Exception e) {
|
---|
80 | println("parse error:");
|
---|
81 | e.printStackTrace();
|
---|
82 | }
|
---|
83 | }
|
---|
84 |
|
---|
85 | /** index one document stored as string*/
|
---|
86 | public void index (String xml_text)
|
---|
87 | {
|
---|
88 | file_id_ = "<xml doc on stdin>";
|
---|
89 | path_ = "";
|
---|
90 |
|
---|
91 | try {
|
---|
92 | sax_parser_.parse(new InputSource(new StringReader(xml_text)), this);
|
---|
93 | }
|
---|
94 | catch (Exception e) {
|
---|
95 | println("parse error:");
|
---|
96 | e.printStackTrace();
|
---|
97 | }
|
---|
98 | }
|
---|
99 |
|
---|
100 | public void finish()
|
---|
101 | {
|
---|
102 | /** optimise the index */
|
---|
103 | try {
|
---|
104 | writer_.optimize();
|
---|
105 | writer_.close();
|
---|
106 | }
|
---|
107 | catch (Exception e) {
|
---|
108 | }
|
---|
109 | }
|
---|
110 |
|
---|
111 | protected void print(String s)
|
---|
112 | {
|
---|
113 | System.out.print(s);
|
---|
114 | }
|
---|
115 |
|
---|
116 | protected void println(String s)
|
---|
117 | {
|
---|
118 | System.out.println(s);
|
---|
119 | }
|
---|
120 |
|
---|
121 | public void startDocument() throws SAXException
|
---|
122 | {
|
---|
123 | println("Starting to index " + file_id_);
|
---|
124 | print("[");
|
---|
125 | }
|
---|
126 |
|
---|
127 | public void endDocument() throws SAXException
|
---|
128 | {
|
---|
129 | println("]");
|
---|
130 | println("... indexing finished.");
|
---|
131 | }
|
---|
132 |
|
---|
133 | public void startElement(String uri, String localName, String qName, Attributes atts)
|
---|
134 | throws SAXException
|
---|
135 | {
|
---|
136 | path_ = appendPathLink(path_, qName, atts);
|
---|
137 |
|
---|
138 | if (qName.equals(doc_tag_level_)) {
|
---|
139 | pushOnStack(); // start new doc
|
---|
140 | current_node_ = qName;
|
---|
141 | String node_id = atts.getValue("gs2:id");
|
---|
142 |
|
---|
143 | print(" " + qName + ": " + node_id );
|
---|
144 | current_doc_.add(new Field("nodeID", node_id, Field.Store.YES, Field.Index.NO));
|
---|
145 | }
|
---|
146 |
|
---|
147 | if (XMLTagInfo.isIndexable(atts)) {
|
---|
148 | indexable_current_node_ = qName;
|
---|
149 | }
|
---|
150 | else {
|
---|
151 | indexable_current_node_ = "";
|
---|
152 | }
|
---|
153 |
|
---|
154 | }
|
---|
155 | public void endElement(String uri, String localName, String qName) throws SAXException
|
---|
156 | {
|
---|
157 | if (qName.equals(indexable_current_node_)) {
|
---|
158 | current_doc_.add(new Field(qName, current_contents_, Field.Store.NO, Field.Index.TOKENIZED));
|
---|
159 | current_contents_ = "";
|
---|
160 | }
|
---|
161 |
|
---|
162 | if (qName.equals(doc_tag_level_)) {
|
---|
163 | try {
|
---|
164 | writer_.addDocument(current_doc_);
|
---|
165 | }
|
---|
166 | catch (java.io.IOException e) {
|
---|
167 | e.printStackTrace();
|
---|
168 | }
|
---|
169 | popOffStack(); // end document
|
---|
170 | }
|
---|
171 |
|
---|
172 | path_ = removePathLink(path_);
|
---|
173 | }
|
---|
174 |
|
---|
175 | public void characters(char ch[], int start, int length) throws SAXException
|
---|
176 | {
|
---|
177 | String data = new String(ch, start, length).trim();
|
---|
178 | if (data.length() > 0 ) {
|
---|
179 | current_contents_ += data;
|
---|
180 | }
|
---|
181 | }
|
---|
182 |
|
---|
183 | protected String appendPathLink(String path, String qName, Attributes atts)
|
---|
184 | {
|
---|
185 |
|
---|
186 | path = path + "/"+qName;
|
---|
187 | if (atts.getLength()>0) {
|
---|
188 | String id = atts.getValue("gs2:id");
|
---|
189 | if (id != null) {
|
---|
190 | path += "[@gs2:id='"+id+"']";
|
---|
191 | }
|
---|
192 | else {
|
---|
193 | id = atts.getValue("gs3:id");
|
---|
194 | if (id != null) {
|
---|
195 | path += "[@gs3:id='"+id+"']";
|
---|
196 | }
|
---|
197 | }
|
---|
198 | }
|
---|
199 | return path;
|
---|
200 | }
|
---|
201 | protected String removePathLink(String path)
|
---|
202 | {
|
---|
203 |
|
---|
204 | int i=path.lastIndexOf('/');
|
---|
205 | if (i==-1) {
|
---|
206 | path="";
|
---|
207 | } else {
|
---|
208 | path = path.substring(0, i);
|
---|
209 | }
|
---|
210 | return path;
|
---|
211 | }
|
---|
212 | /** these are what we save on the stack */
|
---|
213 | private class MyDocument
|
---|
214 | {
|
---|
215 | public Document doc = null;
|
---|
216 | public String contents = null;
|
---|
217 | public String tagname = "";
|
---|
218 |
|
---|
219 | }
|
---|
220 |
|
---|
221 | protected void pushOnStack()
|
---|
222 | {
|
---|
223 | if (current_doc_ != null) {
|
---|
224 | MyDocument save = new MyDocument();
|
---|
225 | save.doc = current_doc_;
|
---|
226 | save.contents = current_contents_;
|
---|
227 | save.tagname = current_node_;
|
---|
228 | stack_.push(save);
|
---|
229 | }
|
---|
230 | current_doc_ = new Document();
|
---|
231 | current_contents_ = "";
|
---|
232 | current_node_ = "";
|
---|
233 | }
|
---|
234 |
|
---|
235 | protected void popOffStack()
|
---|
236 | {
|
---|
237 | if (!stack_.empty()) {
|
---|
238 | MyDocument saved = (MyDocument)stack_.pop();
|
---|
239 | current_doc_ = saved.doc;
|
---|
240 | current_contents_ = saved.contents;
|
---|
241 | current_node_ = saved.tagname;
|
---|
242 | } else {
|
---|
243 | current_doc_ = new Document();
|
---|
244 | current_contents_ = "";
|
---|
245 | current_node_ = "";
|
---|
246 | }
|
---|
247 | }
|
---|
248 |
|
---|
249 |
|
---|
250 | }
|
---|
251 |
|
---|
252 |
|
---|