1 package com.hughes.android.dictionary;
3 import java.io.BufferedOutputStream;
4 import java.io.DataOutputStream;
6 import java.io.FileOutputStream;
7 import java.io.IOException;
8 import java.io.RandomAccessFile;
9 import java.util.ArrayList;
10 import java.util.Arrays;
11 import java.util.List;
12 import java.util.regex.Matcher;
13 import java.util.regex.Pattern;
15 import javax.xml.parsers.ParserConfigurationException;
16 import javax.xml.parsers.SAXParser;
17 import javax.xml.parsers.SAXParserFactory;
19 import org.xml.sax.Attributes;
20 import org.xml.sax.SAXException;
22 public class WiktionarySplitter extends org.xml.sax.helpers.DefaultHandler {
24 static class Section implements java.io.Serializable {
25 private static final long serialVersionUID = -7676549898325856822L;
31 public Section(final String title, final String heading, final String text) {
33 this.heading = heading;
36 //System.out.printf("TITLE:%s\nHEADING:%s\nTEXT:%s\n\n\n\n\n\n", title, heading, text);
40 static class Selector {
44 public Selector(final String filename, final String pattern) throws IOException {
45 this.out = new DataOutputStream(new BufferedOutputStream(new FileOutputStream(filename)));
46 this.pattern = Pattern.compile(pattern);
50 final List<Selector> selectors = new ArrayList<Selector>();
51 StringBuilder titleBuilder;
52 StringBuilder textBuilder;
53 StringBuilder currentBuilder = null;
55 public static void main(final String[] args) throws SAXException, IOException, ParserConfigurationException {
56 final SAXParser parser = SAXParserFactory.newInstance().newSAXParser();
57 final WiktionarySplitter wiktionarySplitter = new WiktionarySplitter();
60 final File file = new File(args[0]);
61 final List<Selector> selectors = wiktionarySplitter.selectors;
62 for (int i = 1; i < args.length; i += 2) {
63 final Selector selector = new Selector(args[i], args[i+1]);
64 selectors.add(selector);
67 if (selectors.isEmpty()) {
68 selectors.addAll(Arrays.asList(
69 new Selector("wikiSplit/arabic.data", ".*[Ar]rabic.*"),
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/hindi.data", ".*[Hh]indi.*"),
77 new Selector("wikiSplit/italian.data", ".*[Ii]talian.*"),
78 new Selector("wikiSplit/japanese.data", ".*[Jj]apanese.*"),
79 new Selector("wikiSplit/korean.data", ".*[Kk]orean.*"),
80 new Selector("wikiSplit/persian.data", ".*[Pp]ersian.*"),
81 new Selector("wikiSplit/romanian.data", ".*[Rr]omanian.*"),
82 new Selector("wikiSplit/russian.data", ".*[Rr]ussian.*"),
83 new Selector("wikiSplit/spanish.data", ".*[Ss]panish.*"),
84 new Selector("wikiSplit/swedish.data", ".*[Ss]wedish.*"),
85 new Selector("wikiSplit/thai.data", ".*[Tt]hai.*"),
86 new Selector("wikiSplit/vietnamese.data", ".*[Vv]ietnamese.*")
91 parser.parse(file, wiktionarySplitter);
94 for (final Selector selector : selectors) {
99 static final Pattern headingStart = Pattern.compile("^(=+)[^=]+=+", Pattern.MULTILINE);
101 private void endPage() {
102 final String title = titleBuilder.toString();
103 System.out.println("endPage: " + title);
105 String text = textBuilder.toString();
107 while (text.length() > 0) {
109 final Matcher startMatcher = headingStart.matcher(text);
110 if (!startMatcher.find()) {
113 text = text.substring(startMatcher.end());
115 final String heading = startMatcher.group();
116 for (final Selector selector : selectors) {
117 if (selector.pattern.matcher(heading).find()) {
120 final int depth = startMatcher.group(1).length();
121 final Pattern endPattern = Pattern.compile(String.format("^={1,%d}[^=]+=+", depth), Pattern.MULTILINE);
123 final Matcher endMatcher = endPattern.matcher(text);
125 if (endMatcher.find()) {
126 end = endMatcher.start();
131 final String sectionText = text.substring(0, end);
132 final Section section = new Section(title, heading, sectionText);
135 selector.out.writeUTF(section.title);
136 selector.out.writeUTF(section.heading);
137 final byte[] bytes = section.text.getBytes("UTF8");
138 selector.out.writeInt(bytes.length);
139 selector.out.write(bytes);
140 } catch (IOException e) {
141 throw new RuntimeException(e);
144 text = text.substring(end);
151 // -----------------------------------------------------------------------
154 public void startElement(String uri, String localName, String qName,
155 Attributes attributes) {
156 currentBuilder = null;
157 if ("page".equals(qName)) {
158 titleBuilder = new StringBuilder();
160 // Start with "\n" to better match certain strings.
161 textBuilder = new StringBuilder("\n");
162 } else if ("title".equals(qName)) {
163 currentBuilder = titleBuilder;
164 } else if ("text".equals(qName)) {
165 currentBuilder = textBuilder;
170 public void characters(char[] ch, int start, int length) throws SAXException {
171 if (currentBuilder != null) {
172 currentBuilder.append(ch, start, length);
177 public void endElement(String uri, String localName, String qName)
178 throws SAXException {
179 currentBuilder = null;
180 if ("page".equals(qName)) {
186 public void parse(final File file) throws ParserConfigurationException,
187 SAXException, IOException {
188 final SAXParser parser = SAXParserFactory.newInstance().newSAXParser();
189 parser.parse(file, this);