1 package com.hughes.android.dictionary.engine;
3 import java.io.BufferedOutputStream;
4 import java.io.DataOutputStream;
6 import java.io.FileOutputStream;
7 import java.io.IOException;
8 import java.util.ArrayList;
9 import java.util.Arrays;
10 import java.util.List;
11 import java.util.regex.Matcher;
12 import java.util.regex.Pattern;
14 import javax.xml.parsers.ParserConfigurationException;
15 import javax.xml.parsers.SAXParser;
16 import javax.xml.parsers.SAXParserFactory;
18 import org.xml.sax.Attributes;
19 import org.xml.sax.SAXException;
21 public class WiktionarySplitter extends org.xml.sax.helpers.DefaultHandler {
23 static class Section implements java.io.Serializable {
24 private static final long serialVersionUID = -7676549898325856822L;
30 public Section(final String title, final String heading, final String text) {
32 this.heading = heading;
35 //System.out.printf("TITLE:%s\nHEADING:%s\nTEXT:%s\n\n\n\n\n\n", title, heading, text);
39 static class Selector {
43 public Selector(final String filename, final String pattern) throws IOException {
44 this.out = new DataOutputStream(new BufferedOutputStream(new FileOutputStream(filename)));
45 this.pattern = Pattern.compile(pattern);
49 final List<Selector> selectors = new ArrayList<Selector>();
50 StringBuilder titleBuilder;
51 StringBuilder textBuilder;
52 StringBuilder currentBuilder = null;
54 public static void main(final String[] args) throws SAXException, IOException, ParserConfigurationException {
55 final SAXParser parser = SAXParserFactory.newInstance().newSAXParser();
56 final WiktionarySplitter wiktionarySplitter = new WiktionarySplitter();
59 final File file = new File(args[0]);
60 final List<Selector> selectors = wiktionarySplitter.selectors;
61 for (int i = 1; i < args.length; i += 2) {
62 final Selector selector = new Selector(args[i], args[i+1]);
63 selectors.add(selector);
66 if (selectors.isEmpty()) {
67 selectors.addAll(Arrays.asList(
68 new Selector("wikiSplit/arabic.data", ".*[Ar]rabic.*"),
69 new Selector("wikiSplit/croation.data", ".*[Cc]roation.*"),
70 new Selector("wikiSplit/czech.data", ".*[Cc]zech.*"),
71 new Selector("wikiSplit/mandarin.data", ".*[Mm]andarin|[Cc]hinese.*"),
72 new Selector("wikiSplit/dutch.data", ".*[Du]utch.*"),
73 new Selector("wikiSplit/english.data", ".*[Ee]nglish.*"),
74 new Selector("wikiSplit/french.data", ".*[Ff]rench.*"),
75 new Selector("wikiSplit/german.data", ".*[Gg]erman.*"),
76 new Selector("wikiSplit/greek.data", ".*[Gg]reek.*"),
77 new Selector("wikiSplit/hindi.data", ".*[Hh]indi.*"),
78 new Selector("wikiSplit/italian.data", ".*[Ii]talian.*"),
79 new Selector("wikiSplit/japanese.data", ".*[Jj]apanese.*"),
80 new Selector("wikiSplit/korean.data", ".*[Kk]orean.*"),
81 new Selector("wikiSplit/persian.data", ".*[Pp]ersian.*"),
82 new Selector("wikiSplit/portuguese.data", ".*[Pp]ortuguese.*"),
83 new Selector("wikiSplit/romanian.data", ".*[Rr]omanian.*"),
84 new Selector("wikiSplit/russian.data", ".*[Rr]ussian.*"),
85 new Selector("wikiSplit/spanish.data", ".*[Ss]panish.*"),
86 new Selector("wikiSplit/swedish.data", ".*[Ss]wedish.*"),
87 new Selector("wikiSplit/thai.data", ".*[Tt]hai.*"),
88 new Selector("wikiSplit/vietnamese.data", ".*[Vv]ietnamese.*")
93 parser.parse(file, wiktionarySplitter);
96 for (final Selector selector : selectors) {
101 static final Pattern headingStart = Pattern.compile("^(=+)[^=]+=+", Pattern.MULTILINE);
103 private void endPage() {
104 final String title = titleBuilder.toString();
105 System.out.println("endPage: " + title);
107 String text = textBuilder.toString();
109 while (text.length() > 0) {
111 final Matcher startMatcher = headingStart.matcher(text);
112 if (!startMatcher.find()) {
115 text = text.substring(startMatcher.end());
117 final String heading = startMatcher.group();
118 for (final Selector selector : selectors) {
119 if (selector.pattern.matcher(heading).find()) {
122 final int depth = startMatcher.group(1).length();
123 final Pattern endPattern = Pattern.compile(String.format("^={1,%d}[^=]+=+", depth), Pattern.MULTILINE);
125 final Matcher endMatcher = endPattern.matcher(text);
127 if (endMatcher.find()) {
128 end = endMatcher.start();
133 final String sectionText = text.substring(0, end);
134 final Section section = new Section(title, heading, sectionText);
137 selector.out.writeUTF(section.title);
138 selector.out.writeUTF(section.heading);
139 final byte[] bytes = section.text.getBytes("UTF8");
140 selector.out.writeInt(bytes.length);
141 selector.out.write(bytes);
142 } catch (IOException e) {
143 throw new RuntimeException(e);
146 text = text.substring(end);
153 // -----------------------------------------------------------------------
156 public void startElement(String uri, String localName, String qName,
157 Attributes attributes) {
158 currentBuilder = null;
159 if ("page".equals(qName)) {
160 titleBuilder = new StringBuilder();
162 // Start with "\n" to better match certain strings.
163 textBuilder = new StringBuilder("\n");
164 } else if ("title".equals(qName)) {
165 currentBuilder = titleBuilder;
166 } else if ("text".equals(qName)) {
167 currentBuilder = textBuilder;
172 public void characters(char[] ch, int start, int length) throws SAXException {
173 if (currentBuilder != null) {
174 currentBuilder.append(ch, start, length);
179 public void endElement(String uri, String localName, String qName)
180 throws SAXException {
181 currentBuilder = null;
182 if ("page".equals(qName)) {
188 public void parse(final File file) throws ParserConfigurationException,
189 SAXException, IOException {
190 final SAXParser parser = SAXParserFactory.newInstance().newSAXParser();
191 parser.parse(file, this);