1 // Copyright 2011 Google Inc. All Rights Reserved.
3 // Licensed under the Apache License, Version 2.0 (the "License");
4 // you may not use this file except in compliance with the License.
5 // You may obtain a copy of the License at
7 // http://www.apache.org/licenses/LICENSE-2.0
9 // Unless required by applicable law or agreed to in writing, software
10 // distributed under the License is distributed on an "AS IS" BASIS,
11 // WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
12 // See the License for the specific language governing permissions and
13 // limitations under the License.
15 package com.hughes.android.dictionary.engine;
17 import java.io.BufferedOutputStream;
18 import java.io.DataOutputStream;
20 import java.io.FileOutputStream;
21 import java.io.IOException;
22 import java.util.ArrayList;
23 import java.util.Arrays;
24 import java.util.List;
25 import java.util.regex.Matcher;
26 import java.util.regex.Pattern;
28 import javax.xml.parsers.ParserConfigurationException;
29 import javax.xml.parsers.SAXParser;
30 import javax.xml.parsers.SAXParserFactory;
32 import org.xml.sax.Attributes;
33 import org.xml.sax.SAXException;
35 public class WiktionarySplitter extends org.xml.sax.helpers.DefaultHandler {
37 private static final String FILE_TO_SPLIT = "data/inputs/enwiktionary-20111224-pages-articles.xml";
39 static class Section implements java.io.Serializable {
40 private static final long serialVersionUID = -7676549898325856822L;
46 public Section(final String title, final String heading, final String text) {
48 this.heading = heading;
51 //System.out.printf("TITLE:%s\nHEADING:%s\nTEXT:%s\n\n\n\n\n\n", title, heading, text);
55 static class Selector {
59 public Selector(final String filename, final String pattern) throws IOException {
60 this.out = new DataOutputStream(new BufferedOutputStream(new FileOutputStream(filename)));
61 this.pattern = Pattern.compile(pattern, Pattern.CASE_INSENSITIVE);
65 final List<Selector> selectors = new ArrayList<Selector>();
66 StringBuilder titleBuilder;
67 StringBuilder textBuilder;
68 StringBuilder currentBuilder = null;
70 public static void main(final String[] args) throws SAXException, IOException, ParserConfigurationException {
71 final SAXParser parser = SAXParserFactory.newInstance().newSAXParser();
72 final WiktionarySplitter wiktionarySplitter = new WiktionarySplitter();
76 final List<Selector> selectors = wiktionarySplitter.selectors;
77 for (int i = 1; i < args.length; i += 2) {
78 final Selector selector = new Selector(args[i], args[i+1]);
79 selectors.add(selector);
82 if (selectors.isEmpty()) {
83 selectors.addAll(Arrays.asList(
84 new Selector("data/inputs/enWikiSplit/AF.data", ".*[Aa]frikaans.*"),
85 new Selector("data/inputs/enWikiSplit/AR.data", ".*[Aa]rabic.*"),
86 new Selector("data/inputs/enWikiSplit/HY.data", ".*[Aa]rmenian.*"),
87 new Selector("data/inputs/enWikiSplit/HR.data", ".*[Cc]roatian.*"),
88 new Selector("data/inputs/enWikiSplit/CS.data", ".*[Cc]zech.*"),
89 new Selector("data/inputs/enWikiSplit/ZH.data", ".*[Cc]hinese.*|.*[Mm]andarin.*|.*Cantonese.*"),
90 new Selector("data/inputs/enWikiSplit/DA.data", ".*[Dd]anish.*"),
91 new Selector("data/inputs/enWikiSplit/NL.data", ".*[Dd]utch.*"),
92 new Selector("data/inputs/enWikiSplit/EN.data", ".*[Ee]nglish.*"),
93 new Selector("data/inputs/enWikiSplit/FI.data", ".*[Ff]innish.*"),
94 new Selector("data/inputs/enWikiSplit/FR.data", ".*[Ff]rench.*"),
95 new Selector("data/inputs/enWikiSplit/DE.data", ".*[Gg]erman.*"),
96 new Selector("data/inputs/enWikiSplit/EL.data", ".*[Gg]reek.*"),
97 new Selector("data/inputs/enWikiSplit/haw.data", ".*[Hh]awaiian.*"),
98 new Selector("data/inputs/enWikiSplit/HE.data", ".*[Hh]ebrew.*"),
99 new Selector("data/inputs/enWikiSplit/HI.data", ".*[Hh]indi.*"),
100 new Selector("data/inputs/enWikiSplit/IS.data", ".*[Ii]celandic.*"),
101 new Selector("data/inputs/enWikiSplit/GA.data", ".*[Ii]rish.*"),
102 new Selector("data/inputs/enWikiSplit/IT.data", ".*[Ii]talian.*"),
103 new Selector("data/inputs/enWikiSplit/LT.data", ".*[Ll]ithuanian.*"),
104 new Selector("data/inputs/enWikiSplit/JA.data", ".*[Jj]apanese.*"),
105 new Selector("data/inputs/enWikiSplit/KO.data", ".*[Kk]orean.*"),
106 new Selector("data/inputs/enWikiSplit/KU.data", ".*[Kk]urdish.*"),
107 new Selector("data/inputs/enWikiSplit/MS.data", ".*[Mm]alay.*"),
108 new Selector("data/inputs/enWikiSplit/MI.data", ".*[Mm]aori.*"),
109 new Selector("data/inputs/enWikiSplit/MN.data", ".*[Mm]ongolian.*"),
110 new Selector("data/inputs/enWikiSplit/NO.data", ".*[Nn]orwegian.*"),
111 new Selector("data/inputs/enWikiSplit/FA.data", ".*[Pp]ersian.*"),
112 new Selector("data/inputs/enWikiSplit/PT.data", ".*[Pp]ortuguese.*"),
113 new Selector("data/inputs/enWikiSplit/PL.data", ".*[Pp]olish.*"),
114 new Selector("data/inputs/enWikiSplit/RO.data", ".*[Rr]omanian.*"),
115 new Selector("data/inputs/enWikiSplit/RU.data", ".*[Rr]ussian.*"),
116 new Selector("data/inputs/enWikiSplit/SA.data", ".*[Ss]anskrit.*"),
117 new Selector("data/inputs/enWikiSplit/SR.data", ".*[Ss]erbian.*"),
118 new Selector("data/inputs/enWikiSplit/SO.data", ".*[Ss]omali.*"),
119 new Selector("data/inputs/enWikiSplit/ES.data", ".*[Ss]panish.*"),
120 new Selector("data/inputs/enWikiSplit/SV.data", ".*[Ss]wedish.*"),
121 new Selector("data/inputs/enWikiSplit/TG.data", ".*[Tt]ajik.*"),
122 new Selector("data/inputs/enWikiSplit/TH.data", ".*[Tt]hai.*"),
123 new Selector("data/inputs/enWikiSplit/BO.data", ".*[Tt]ibetan.*"),
124 new Selector("data/inputs/enWikiSplit/TR.data", ".*[Tt]urkish.*"),
125 new Selector("data/inputs/enWikiSplit/UK.data", ".*[Uu]krainian.*"),
126 new Selector("data/inputs/enWikiSplit/VI.data", ".*[Vv]ietnamese.*"),
127 new Selector("data/inputs/enWikiSplit/CI.data", ".*[Ww]elsh.*"),
128 new Selector("data/inputs/enWikiSplit/YI.data", ".*[Yy]iddish.*"),
129 new Selector("data/inputs/enWikiSplit/ZU.data", ".*[Zz]ulu.*")
134 parser.parse(new File(FILE_TO_SPLIT), wiktionarySplitter);
137 for (final Selector selector : selectors) {
138 selector.out.close();
142 static final Pattern headingStart = Pattern.compile("^(=+)[^=]+=+", Pattern.MULTILINE);
145 private void endPage() {
146 final String title = titleBuilder.toString();
147 if (++pageCount % 1000 == 0) {
148 System.out.println("endPage: " + title + ", count=" + pageCount);
151 String text = textBuilder.toString();
153 while (text.length() > 0) {
155 final Matcher startMatcher = headingStart.matcher(text);
156 if (!startMatcher.find()) {
159 text = text.substring(startMatcher.end());
161 final String heading = startMatcher.group();
162 for (final Selector selector : selectors) {
163 if (selector.pattern.matcher(heading).find()) {
166 final int depth = startMatcher.group(1).length();
167 final Pattern endPattern = Pattern.compile(String.format("^={1,%d}[^=]+=+", depth), Pattern.MULTILINE);
169 final Matcher endMatcher = endPattern.matcher(text);
171 if (endMatcher.find()) {
172 end = endMatcher.start();
177 final String sectionText = text.substring(0, end);
178 final Section section = new Section(title, heading, sectionText);
181 selector.out.writeUTF(section.title);
182 selector.out.writeUTF(section.heading);
183 final byte[] bytes = section.text.getBytes("UTF8");
184 selector.out.writeInt(bytes.length);
185 selector.out.write(bytes);
186 } catch (IOException e) {
187 throw new RuntimeException(e);
190 text = text.substring(end);
197 // -----------------------------------------------------------------------
200 public void startElement(String uri, String localName, String qName,
201 Attributes attributes) {
202 currentBuilder = null;
203 if ("page".equals(qName)) {
204 titleBuilder = new StringBuilder();
206 // Start with "\n" to better match certain strings.
207 textBuilder = new StringBuilder("\n");
208 } else if ("title".equals(qName)) {
209 currentBuilder = titleBuilder;
210 } else if ("text".equals(qName)) {
211 currentBuilder = textBuilder;
216 public void characters(char[] ch, int start, int length) throws SAXException {
217 if (currentBuilder != null) {
218 currentBuilder.append(ch, start, length);
223 public void endElement(String uri, String localName, String qName)
224 throws SAXException {
225 currentBuilder = null;
226 if ("page".equals(qName)) {
232 public void parse(final File file) throws ParserConfigurationException,
233 SAXException, IOException {
234 final SAXParser parser = SAXParserFactory.newInstance().newSAXParser();
235 parser.parse(file, this);