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.
18 package com.hughes.android.dictionary.engine;
20 import java.io.IOException;
21 import java.io.PrintStream;
22 import java.io.RandomAccessFile;
23 import java.util.ArrayList;
24 import java.util.Collection;
25 import java.util.Collections;
26 import java.util.EnumMap;
27 import java.util.HashSet;
28 import java.util.LinkedHashMap;
29 import java.util.LinkedHashSet;
30 import java.util.List;
33 import java.util.concurrent.atomic.AtomicBoolean;
34 import java.util.regex.Pattern;
36 import com.hughes.android.dictionary.DictionaryInfo;
37 import com.hughes.android.dictionary.DictionaryInfo.IndexInfo;
38 import com.hughes.android.dictionary.engine.RowBase.RowKey;
39 import com.hughes.util.CachingList;
40 import com.hughes.util.TransformingList;
41 import com.hughes.util.raf.RAFList;
42 import com.hughes.util.raf.RAFSerializable;
43 import com.hughes.util.raf.RAFSerializer;
44 import com.hughes.util.raf.SerializableSerializer;
45 import com.hughes.util.raf.UniformRAFList;
46 import com.ibm.icu.text.Collator;
47 import com.ibm.icu.text.Transliterator;
49 public final class Index implements RAFSerializable<Index> {
51 static final int CACHE_SIZE = 5000;
53 final Dictionary dict;
55 public final String shortName; // Typically the ISO code for the language.
56 public final String longName;
58 // persisted: tells how the entries are sorted.
59 public final Language sortLanguage;
60 final String normalizerRules;
62 // Built from the two above.
63 private Transliterator normalizer;
66 public final List<IndexEntry> sortedIndexEntries;
69 public final Set<String> stoplist;
74 public final List<RowBase> rows;
75 public final boolean swapPairEntries;
78 int mainTokenCount = -1;
80 // --------------------------------------------------------------------------
82 public Index(final Dictionary dict, final String shortName, final String longName, final Language sortLanguage, final String normalizerRules, final boolean swapPairEntries, final Set<String> stoplist) {
84 this.shortName = shortName;
85 this.longName = longName;
86 this.sortLanguage = sortLanguage;
87 this.normalizerRules = normalizerRules;
88 this.swapPairEntries = swapPairEntries;
89 sortedIndexEntries = new ArrayList<IndexEntry>();
90 this.stoplist = stoplist;
91 rows = new ArrayList<RowBase>();
97 * Deferred initialization because it can be slow.
99 public synchronized Transliterator normalizer() {
100 if (normalizer == null) {
101 normalizer = Transliterator.createFromRules("", normalizerRules, Transliterator.FORWARD);
107 * Note that using this comparator probably involves doing too many text normalizations.
109 public NormalizeComparator getSortComparator() {
110 return new NormalizeComparator(normalizer(), sortLanguage.getCollator());
113 public Index(final Dictionary dict, final RandomAccessFile raf) throws IOException {
115 shortName = raf.readUTF();
116 longName = raf.readUTF();
117 final String languageCode = raf.readUTF();
118 sortLanguage = Language.lookup(languageCode);
119 normalizerRules = raf.readUTF();
120 swapPairEntries = raf.readBoolean();
121 if (sortLanguage == null) {
122 throw new IOException("Unsupported language: " + languageCode);
124 if (dict.dictFileVersion >= 2) {
125 mainTokenCount = raf.readInt();
127 sortedIndexEntries = CachingList.create(RAFList.create(raf, IndexEntry.SERIALIZER, raf.getFilePointer()), CACHE_SIZE);
128 if (dict.dictFileVersion >= 4) {
129 stoplist = new SerializableSerializer<Set<String>>().read(raf);
131 stoplist = Collections.emptySet();
133 rows = CachingList.create(UniformRAFList.create(raf, new RowBase.Serializer(this), raf.getFilePointer()), CACHE_SIZE);
137 public void write(final RandomAccessFile raf) throws IOException {
138 raf.writeUTF(shortName);
139 raf.writeUTF(longName);
140 raf.writeUTF(sortLanguage.getIsoCode());
141 raf.writeUTF(normalizerRules);
142 raf.writeBoolean(swapPairEntries);
143 if (dict.dictFileVersion >= 2) {
144 raf.writeInt(mainTokenCount);
146 RAFList.write(raf, sortedIndexEntries, IndexEntry.SERIALIZER);
147 new SerializableSerializer<Set<String>>().write(raf, stoplist);
148 UniformRAFList.write(raf, (Collection<RowBase>) rows, new RowBase.Serializer(this), 5 /* bytes per entry */);
151 public void print(final PrintStream out) {
152 for (final RowBase row : rows) {
157 public static final class IndexEntry implements RAFSerializable<Index.IndexEntry> {
158 public final String token;
159 private final String normalizedToken;
160 public final int startRow;
161 public final int numRows; // doesn't count the token row!
164 static final RAFSerializer<IndexEntry> SERIALIZER = new RAFSerializer<IndexEntry> () {
166 public IndexEntry read(RandomAccessFile raf) throws IOException {
167 return new IndexEntry(raf);
170 public void write(RandomAccessFile raf, IndexEntry t) throws IOException {
174 public IndexEntry(final String token, final String normalizedToken, final int startRow, final int numRows) {
175 assert token.equals(token.trim());
176 assert token.length() > 0;
178 this.normalizedToken = normalizedToken;
179 this.startRow = startRow;
180 this.numRows = numRows;
183 public IndexEntry(final RandomAccessFile raf) throws IOException {
184 token = raf.readUTF();
185 startRow = raf.readInt();
186 numRows = raf.readInt();
187 final boolean hasNormalizedForm = raf.readBoolean();
188 normalizedToken = hasNormalizedForm ? raf.readUTF() : token;
191 public void write(RandomAccessFile raf) throws IOException {
193 raf.writeInt(startRow);
194 raf.writeInt(numRows);
195 final boolean hasNormalizedForm = !token.equals(normalizedToken);
196 raf.writeBoolean(hasNormalizedForm);
197 if (hasNormalizedForm) {
198 raf.writeUTF(normalizedToken);
202 public String toString() {
203 return String.format("%s@%d(%d)", token, startRow, numRows);
206 public String normalizedToken() {
207 return normalizedToken;
211 static final TransformingList.Transformer<IndexEntry, String> INDEX_ENTRY_TO_TOKEN = new TransformingList.Transformer<IndexEntry, String>() {
213 public String transform(IndexEntry t1) {
218 public IndexEntry findExact(final String exactToken) {
219 final int result = Collections.binarySearch(TransformingList.create(sortedIndexEntries, INDEX_ENTRY_TO_TOKEN), exactToken, getSortComparator());
221 return sortedIndexEntries.get(result);
226 public IndexEntry findInsertionPoint(String token, final AtomicBoolean interrupted) {
227 final int index = findInsertionPointIndex(token, interrupted);
228 return index != -1 ? sortedIndexEntries.get(index) : null;
231 public int findInsertionPointIndex(String token, final AtomicBoolean interrupted) {
232 token = normalizeToken(token);
235 int end = sortedIndexEntries.size();
237 final Collator sortCollator = sortLanguage.getCollator();
238 while (start < end) {
239 final int mid = (start + end) / 2;
240 if (interrupted.get()) {
243 final IndexEntry midEntry = sortedIndexEntries.get(mid);
245 final int comp = sortCollator.compare(token, midEntry.normalizedToken());
247 final int result = windBackCase(token, mid, interrupted);
249 } else if (comp < 0) {
250 //System.out.println("Upper bound: " + midEntry + ", norm=" + midEntry.normalizedToken() + ", mid=" + mid);
253 //System.out.println("Lower bound: " + midEntry + ", norm=" + midEntry.normalizedToken() + ", mid=" + mid);
258 // If we search for a substring of a string that's in there, return that.
259 int result = Math.min(start, sortedIndexEntries.size() - 1);
260 result = windBackCase(sortedIndexEntries.get(result).normalizedToken(), result, interrupted);
264 private final int windBackCase(final String token, int result, final AtomicBoolean interrupted) {
265 while (result > 0 && sortedIndexEntries.get(result - 1).normalizedToken().equals(token)) {
267 if (interrupted.get()) {
274 public IndexInfo getIndexInfo() {
275 return new DictionaryInfo.IndexInfo(shortName, sortedIndexEntries.size(), mainTokenCount);
278 private static final int MAX_SEARCH_ROWS = 1000;
280 private final Map<String,Integer> prefixToNumRows = new LinkedHashMap<String, Integer>();
281 private synchronized final int getUpperBoundOnRowsStartingWith(final String normalizedPrefix, final int maxRows, final AtomicBoolean interrupted) {
282 final Integer numRows = prefixToNumRows.get(normalizedPrefix);
283 if (numRows != null) {
286 final int insertionPointIndex = findInsertionPointIndex(normalizedPrefix, interrupted);
289 for (int index = insertionPointIndex; index < sortedIndexEntries.size(); ++index) {
290 if (interrupted.get()) { return -1; }
291 final IndexEntry indexEntry = sortedIndexEntries.get(index);
292 if (!indexEntry.normalizedToken.startsWith(normalizedPrefix)) {
295 rowCount += indexEntry.numRows;
296 if (rowCount > maxRows) {
297 System.out.println("Giving up, too many words with prefix: " + normalizedPrefix);
301 prefixToNumRows.put(normalizedPrefix, numRows);
306 public final List<RowBase> multiWordSearch(final List<String> searchTokens, final AtomicBoolean interrupted) {
307 final long startMills = System.currentTimeMillis();
308 final List<RowBase> result = new ArrayList<RowBase>();
310 final Set<String> normalizedNonStoplist = new LinkedHashSet<String>();
312 String bestPrefix = null;
313 int leastRows = Integer.MAX_VALUE;
314 final StringBuilder regex = new StringBuilder();
315 for (int i = 0; i < searchTokens.size(); ++i) {
316 if (interrupted.get()) { return null; }
317 final String searchToken = searchTokens.get(i);
318 final String normalized = normalizeToken(searchTokens.get(i));
319 // Normalize them all.
320 searchTokens.set(i, normalized);
322 if (!stoplist.contains(searchToken)) {
323 if (normalizedNonStoplist.add(normalized)) {
324 final int numRows = getUpperBoundOnRowsStartingWith(normalized, MAX_SEARCH_ROWS, interrupted);
325 if (numRows != -1 && numRows < leastRows) {
327 // We really are done here.
328 return Collections.emptyList();
331 bestPrefix = normalized;
336 if (regex.length() > 0) {
337 regex.append("[\\s]*");
339 regex.append(Pattern.quote(normalized));
341 final Pattern pattern = Pattern.compile(regex.toString());
343 if (bestPrefix == null) {
344 bestPrefix = searchTokens.get(0);
345 System.out.println("Everything was in the stoplist!");
347 System.out.println("Searching using prefix: " + bestPrefix + ", leastRows=" + leastRows + ", searchTokens=" + searchTokens);
349 // Place to store the things that match.
350 final Map<RowMatchType,List<RowBase>> matches = new EnumMap<RowMatchType, List<RowBase>>(RowMatchType.class);
351 for (final RowMatchType rowMatchType : RowMatchType.values()) {
352 if (rowMatchType != RowMatchType.NO_MATCH) {
353 matches.put(rowMatchType, new ArrayList<RowBase>());
358 final Set<RowKey> cachedRowKeys = new HashSet<RowBase.RowKey>();
360 // for (final String searchToken : searchTokens) {
361 final String searchToken = bestPrefix;
363 final int insertionPointIndex = findInsertionPointIndex(searchToken, interrupted);
365 for (int index = insertionPointIndex; index < sortedIndexEntries.size() && matchCount < MAX_SEARCH_ROWS; ++index) {
366 if (interrupted.get()) { return null; }
367 final IndexEntry indexEntry = sortedIndexEntries.get(index);
368 if (!indexEntry.normalizedToken.startsWith(searchToken)) {
372 // System.out.println("Searching indexEntry: " + indexEntry.token);
374 // Extra +1 to skip token row.
375 for (int rowIndex = indexEntry.startRow + 1; rowIndex < indexEntry.startRow + 1 + indexEntry.numRows && rowIndex < rows.size(); ++rowIndex) {
376 if (interrupted.get()) { return null; }
377 final RowBase row = rows.get(rowIndex);
378 final RowBase.RowKey rowKey = row.getRowKey();
379 if (cachedRowKeys.contains(rowKey)) {
382 cachedRowKeys.add(rowKey);
383 final RowMatchType matchType = row.matches(searchTokens, pattern, normalizer(), swapPairEntries);
384 if (matchType != RowMatchType.NO_MATCH) {
385 matches.get(matchType).add(row);
392 // Sort them into a reasonable order.
393 final RowBase.LengthComparator lengthComparator = new RowBase.LengthComparator(swapPairEntries);
394 for (final Collection<RowBase> rows : matches.values()) {
395 final List<RowBase> ordered = new ArrayList<RowBase>(rows);
396 Collections.sort(ordered, lengthComparator);
397 result.addAll(ordered);
400 System.out.println("searchDuration: " + (System.currentTimeMillis() - startMills));
404 private String normalizeToken(final String searchToken) {
405 if (TransliteratorManager.init(null)) {
406 final Transliterator normalizer = normalizer();
407 return normalizer.transliterate(searchToken);
409 // Do our best since the Transliterators aren't up yet.
410 return searchToken.toLowerCase();