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 com.hughes.android.dictionary.DictionaryInfo;
18 import com.hughes.android.dictionary.DictionaryInfo.IndexInfo;
19 import com.hughes.android.dictionary.engine.RowBase.RowKey;
20 import com.hughes.util.CachingList;
21 import com.hughes.util.StringUtil;
22 import com.hughes.util.TransformingList;
23 import com.hughes.util.raf.RAFList;
24 import com.hughes.util.raf.RAFSerializable;
25 import com.hughes.util.raf.RAFSerializer;
26 import com.hughes.util.raf.SerializableSerializer;
27 import com.hughes.util.raf.UniformRAFList;
28 import com.ibm.icu.text.Transliterator;
30 import java.io.DataInput;
31 import java.io.DataOutput;
32 import java.io.IOException;
33 import java.io.PrintStream;
34 import java.io.RandomAccessFile;
35 import java.nio.channels.FileChannel;
36 import java.util.AbstractList;
37 import java.util.ArrayList;
38 import java.util.Collection;
39 import java.util.Collections;
40 import java.util.Comparator;
41 import java.util.EnumMap;
42 import java.util.HashSet;
43 import java.util.HashMap;
44 import java.util.List;
47 import java.util.concurrent.atomic.AtomicBoolean;
48 import java.util.regex.Pattern;
50 public final class Index implements RAFSerializable<Index> {
52 private static final int CACHE_SIZE = 5000;
54 public final Dictionary dict;
56 public final String shortName; // Typically the ISO code for the language.
57 public final String longName;
59 // persisted: tells how the entries are sorted.
60 public final Language sortLanguage;
61 private final String normalizerRules;
63 // Built from the two above.
64 private Transliterator normalizer;
67 public final List<IndexEntry> sortedIndexEntries;
70 private final Set<String> stoplist;
75 public final List<RowBase> rows;
76 public final boolean swapPairEntries;
79 @SuppressWarnings("WeakerAccess")
80 public int mainTokenCount = -1;
82 // --------------------------------------------------------------------------
84 public Index(final Dictionary dict, final String shortName, final String longName,
85 final Language sortLanguage, final String normalizerRules,
86 final boolean swapPairEntries, final Set<String> stoplist) {
88 this.shortName = shortName;
89 this.longName = longName;
90 this.sortLanguage = sortLanguage;
91 this.normalizerRules = normalizerRules;
92 this.swapPairEntries = swapPairEntries;
93 sortedIndexEntries = new ArrayList<>();
94 this.stoplist = stoplist;
95 rows = new ArrayList<>();
101 * Deferred initialization because it can be slow.
103 @SuppressWarnings("WeakerAccess")
104 public synchronized Transliterator normalizer() {
105 if (normalizer == null) {
106 normalizer = TransliteratorManager.get(normalizerRules);
112 * Note that using this comparator probably involves doing too many text
115 @SuppressWarnings("WeakerAccess")
116 public NormalizeComparator getSortComparator() {
117 return new NormalizeComparator(normalizer(), sortLanguage.getCollator(), dict.dictFileVersion);
120 public Index(final Dictionary dict, final FileChannel inp, final DataInput raf) throws IOException {
122 shortName = raf.readUTF();
123 longName = raf.readUTF();
124 final String languageCode = raf.readUTF();
125 sortLanguage = Language.lookup(languageCode);
126 normalizerRules = raf.readUTF();
127 swapPairEntries = raf.readBoolean();
128 if (sortLanguage == null) {
129 throw new IOException("Unsupported language: " + languageCode);
131 if (dict.dictFileVersion >= 2) {
132 mainTokenCount = raf.readInt();
134 sortedIndexEntries = CachingList.create(
135 RAFList.create(inp, new IndexEntrySerializer(dict.dictFileVersion == 6 ? inp : null), inp.position(),
136 dict.dictFileVersion, dict.dictInfo + " idx " + languageCode + ": "), CACHE_SIZE, true);
137 if (dict.dictFileVersion >= 7) {
138 int count = StringUtil.readVarInt(raf);
139 stoplist = new HashSet<>(count);
140 for (int i = 0; i < count; ++i) {
141 stoplist.add(raf.readUTF());
143 } else if (dict.dictFileVersion >= 4) {
144 stoplist = new SerializableSerializer<Set<String>>().read(raf);
146 stoplist = Collections.emptySet();
148 rows = CachingList.create(
149 UniformRAFList.create(inp, new RowBase.Serializer(this), inp.position()),
154 public void write(final DataOutput out) throws IOException {
155 RandomAccessFile raf = (RandomAccessFile)out;
156 raf.writeUTF(shortName);
157 raf.writeUTF(longName);
158 raf.writeUTF(sortLanguage.getIsoCode());
159 raf.writeUTF(normalizerRules);
160 raf.writeBoolean(swapPairEntries);
161 raf.writeInt(mainTokenCount);
162 RAFList.write(raf, sortedIndexEntries, new IndexEntrySerializer(null), 32, true);
163 StringUtil.writeVarInt(raf, stoplist.size());
164 for (String i : stoplist) {
167 UniformRAFList.write(raf, rows, new RowBase.Serializer(this), 3 /* bytes per entry */);
170 public void print(final PrintStream out) {
171 for (final RowBase row : rows) {
176 private final class IndexEntrySerializer implements RAFSerializer<IndexEntry> {
177 private final FileChannel ch;
179 IndexEntrySerializer(FileChannel ch) {
184 public IndexEntry read(DataInput raf) throws IOException {
185 return new IndexEntry(Index.this, ch, raf);
189 public void write(DataOutput raf, IndexEntry t) throws IOException {
194 public static final class IndexEntry implements RAFSerializable<Index.IndexEntry> {
195 public final String token;
196 private final String normalizedToken;
197 public final int startRow;
198 final int numRows; // doesn't count the token row!
199 public List<HtmlEntry> htmlEntries;
201 public IndexEntry(final Index index, final String token, final String normalizedToken,
202 final int startRow, final int numRows) {
203 assert token.equals(token.trim());
204 assert token.length() > 0;
206 this.normalizedToken = normalizedToken;
207 this.startRow = startRow;
208 this.numRows = numRows;
209 this.htmlEntries = new ArrayList<>();
212 IndexEntry(final Index index, final FileChannel ch, final DataInput raf) throws IOException {
213 token = raf.readUTF();
214 if (index.dict.dictFileVersion >= 7) {
215 startRow = StringUtil.readVarInt(raf);
216 numRows = StringUtil.readVarInt(raf);
218 startRow = raf.readInt();
219 numRows = raf.readInt();
221 final boolean hasNormalizedForm = raf.readBoolean();
222 normalizedToken = hasNormalizedForm ? raf.readUTF() : token;
223 if (index.dict.dictFileVersion >= 7) {
224 int size = StringUtil.readVarInt(raf);
226 this.htmlEntries = Collections.emptyList();
228 final int[] htmlEntryIndices = new int[size];
229 for (int i = 0; i < size; ++i) {
230 htmlEntryIndices[i] = StringUtil.readVarInt(raf);
232 this.htmlEntries = new AbstractList<HtmlEntry>() {
234 public HtmlEntry get(int i) {
235 return index.dict.htmlEntries.get(htmlEntryIndices[i]);
239 return htmlEntryIndices.length;
243 } else if (index.dict.dictFileVersion >= 6) {
244 this.htmlEntries = CachingList.create(
245 RAFList.create(ch, index.dict.htmlEntryIndexSerializer,
246 ch.position(), index.dict.dictFileVersion,
247 index.dict.dictInfo + " htmlEntries: "), 1, false);
249 this.htmlEntries = Collections.emptyList();
253 public void write(DataOutput raf) throws IOException {
255 StringUtil.writeVarInt(raf, startRow);
256 StringUtil.writeVarInt(raf, numRows);
257 final boolean hasNormalizedForm = !token.equals(normalizedToken);
258 raf.writeBoolean(hasNormalizedForm);
259 if (hasNormalizedForm) {
260 raf.writeUTF(normalizedToken);
262 StringUtil.writeVarInt(raf, htmlEntries.size());
263 for (HtmlEntry e : htmlEntries)
264 StringUtil.writeVarInt(raf, e.index());
267 public String toString() {
268 return String.format("%s@%d(%d)", token, startRow, numRows);
271 String normalizedToken() {
272 return normalizedToken;
276 private static final TransformingList.Transformer<IndexEntry, String> INDEX_ENTRY_TO_TOKEN = new TransformingList.Transformer<IndexEntry, String>() {
278 public String transform(IndexEntry t1) {
283 public IndexEntry findExact(final String exactToken) {
284 final int result = Collections.binarySearch(
285 TransformingList.create(sortedIndexEntries, INDEX_ENTRY_TO_TOKEN), exactToken,
286 getSortComparator());
288 return sortedIndexEntries.get(result);
293 public IndexEntry findInsertionPoint(String token, final AtomicBoolean interrupted) {
294 final int index = findInsertionPointIndex(token, interrupted);
295 return index != -1 ? sortedIndexEntries.get(index) : null;
298 private int compareIdx(String token, final Comparator<Object> sortCollator, int idx) {
299 final IndexEntry entry = sortedIndexEntries.get(idx);
300 return NormalizeComparator.compareWithoutDash(token, entry.normalizedToken(), sortCollator, dict.dictFileVersion);
303 private int findMatchLen(final Comparator<Object> sortCollator, String a, String b) {
305 int end = Math.min(a.length(), b.length());
308 int mid = (start + end + 1) / 2;
309 if (sortCollator.compare(a.substring(0, mid), b.substring(0, mid)) == 0)
317 private int findInsertionPointIndex(String token, final AtomicBoolean interrupted) {
318 token = normalizeToken(token);
321 int end = sortedIndexEntries.size();
323 final Comparator<Object> sortCollator = sortLanguage.getCollator();
324 while (start < end) {
325 final int mid = (start + end) / 2;
326 if (interrupted.get()) {
329 final IndexEntry midEntry = sortedIndexEntries.get(mid);
331 int comp = NormalizeComparator.compareWithoutDash(token, midEntry.normalizedToken(), sortCollator, dict.dictFileVersion);
333 comp = sortCollator.compare(token, midEntry.normalizedToken());
335 return windBackCase(token, mid, interrupted);
336 } else if (comp < 0) {
337 // System.out.println("Upper bound: " + midEntry + ", norm=" +
338 // midEntry.normalizedToken() + ", mid=" + mid);
340 // Hack for robustness if sort order is broken
342 compareIdx(token, sortCollator, mid + 1) > 0 &&
343 compareIdx(token, sortCollator, mid + 2) > 0) {
349 // System.out.println("Lower bound: " + midEntry + ", norm=" +
350 // midEntry.normalizedToken() + ", mid=" + mid);
352 // Hack for robustness if sort order is broken
353 if (mid - 2 >= start &&
354 compareIdx(token, sortCollator, mid - 1) < 0 &&
355 compareIdx(token, sortCollator, mid - 2) < 0) {
363 // if the word before is the better match, move
365 if (start > 0 && start < sortedIndexEntries.size()) {
366 String prev = sortedIndexEntries.get(start - 1).normalizedToken();
367 String next = sortedIndexEntries.get(start).normalizedToken();
368 if (findMatchLen(sortCollator, token, prev) >= findMatchLen(sortCollator, token, next))
372 // If we search for a substring of a string that's in there, return
374 int result = Math.min(start, sortedIndexEntries.size() - 1);
375 result = windBackCase(sortedIndexEntries.get(result).normalizedToken(), result, interrupted);
379 private final int windBackCase(final String token, int result, final AtomicBoolean interrupted) {
380 while (result > 0 && sortedIndexEntries.get(result - 1).normalizedToken().equals(token)) {
382 if (interrupted.get()) {
389 public IndexInfo getIndexInfo() {
390 return new DictionaryInfo.IndexInfo(shortName, sortedIndexEntries.size(), mainTokenCount);
393 private static final int MAX_SEARCH_ROWS = 1000;
395 private final Map<String, Integer> prefixToNumRows = new HashMap<>();
397 private synchronized final int getUpperBoundOnRowsStartingWith(final String normalizedPrefix,
398 final int maxRows, final AtomicBoolean interrupted) {
399 final Integer numRows = prefixToNumRows.get(normalizedPrefix);
400 if (numRows != null) {
403 final int insertionPointIndex = findInsertionPointIndex(normalizedPrefix, interrupted);
406 for (int index = insertionPointIndex; index < sortedIndexEntries.size(); ++index) {
407 if (interrupted.get()) {
410 final IndexEntry indexEntry = sortedIndexEntries.get(index);
411 if (!indexEntry.normalizedToken.startsWith(normalizedPrefix) &&
412 !NormalizeComparator.withoutDash(indexEntry.normalizedToken).startsWith(normalizedPrefix)) {
415 rowCount += indexEntry.numRows + indexEntry.htmlEntries.size();
416 if (rowCount > maxRows) {
417 System.out.println("Giving up, too many words with prefix: " + normalizedPrefix);
421 prefixToNumRows.put(normalizedPrefix, rowCount);
425 public final List<RowBase> multiWordSearch(
426 final String searchText, final List<String> searchTokens,
427 final AtomicBoolean interrupted) {
428 final long startMills = System.currentTimeMillis();
429 final List<RowBase> result = new ArrayList<>();
431 final Set<String> normalizedNonStoplist = new HashSet<>();
433 String bestPrefix = null;
434 int leastRows = Integer.MAX_VALUE;
435 final StringBuilder searchTokensRegex = new StringBuilder();
436 for (int i = 0; i < searchTokens.size(); ++i) {
437 if (interrupted.get()) {
440 final String searchToken = searchTokens.get(i);
441 final String normalized = normalizeToken(searchTokens.get(i));
442 // Normalize them all.
443 searchTokens.set(i, normalized);
445 if (!stoplist.contains(searchToken)) {
446 if (normalizedNonStoplist.add(normalized)) {
447 final int numRows = getUpperBoundOnRowsStartingWith(normalized,
448 MAX_SEARCH_ROWS, interrupted);
449 if (numRows != -1 && numRows < leastRows) {
451 // We really are done here.
452 return Collections.emptyList();
455 bestPrefix = normalized;
460 if (searchTokensRegex.length() > 0) {
461 searchTokensRegex.append("[\\s]*");
463 searchTokensRegex.append(Pattern.quote(normalized));
465 final Pattern pattern = Pattern.compile(searchTokensRegex.toString());
467 if (bestPrefix == null) {
468 bestPrefix = searchTokens.get(0);
469 System.out.println("Everything was in the stoplist!");
471 System.out.println("Searching using prefix: " + bestPrefix + ", leastRows=" + leastRows
472 + ", searchTokens=" + searchTokens);
474 // Place to store the things that match.
475 final Map<RowMatchType, List<RowBase>> matches = new EnumMap<>(
477 for (final RowMatchType rowMatchType : RowMatchType.values()) {
478 if (rowMatchType != RowMatchType.NO_MATCH) {
479 matches.put(rowMatchType, new ArrayList<RowBase>());
485 final int exactMatchIndex = findInsertionPointIndex(searchText, interrupted);
486 if (exactMatchIndex != -1) {
487 final IndexEntry exactMatch = sortedIndexEntries.get(exactMatchIndex);
488 if (pattern.matcher(exactMatch.token).find()) {
489 matches.get(RowMatchType.TITLE_MATCH).add(rows.get(exactMatch.startRow));
493 final String searchToken = bestPrefix;
494 final int insertionPointIndex = findInsertionPointIndex(searchToken, interrupted);
495 final Set<RowKey> rowsAlreadySeen = new HashSet<>();
496 for (int index = insertionPointIndex; index < sortedIndexEntries.size()
497 && matchCount < MAX_SEARCH_ROWS; ++index) {
498 if (interrupted.get()) {
501 final IndexEntry indexEntry = sortedIndexEntries.get(index);
502 if (!indexEntry.normalizedToken.startsWith(searchToken) &&
503 !NormalizeComparator.withoutDash(indexEntry.normalizedToken).startsWith(searchToken)) {
507 // System.out.println("Searching indexEntry: " + indexEntry.token);
509 // Extra +1 to skip token row.
510 for (int rowIndex = indexEntry.startRow + 1; rowIndex < indexEntry.startRow + 1
512 && rowIndex < rows.size(); ++rowIndex) {
513 if (interrupted.get()) {
516 final RowBase row = rows.get(rowIndex);
517 final RowBase.RowKey rowKey = row.getRowKey();
518 if (rowsAlreadySeen.contains(rowKey)) {
521 rowsAlreadySeen.add(rowKey);
522 final RowMatchType matchType = row.matches(searchTokens, pattern, normalizer(),
524 if (matchType != RowMatchType.NO_MATCH) {
525 matches.get(matchType).add(row);
532 // Sort them into a reasonable order.
533 final RowBase.LengthComparator lengthComparator = new RowBase.LengthComparator(
535 for (final Collection<RowBase> rows : matches.values()) {
536 final List<RowBase> ordered = new ArrayList<>(rows);
537 Collections.sort(ordered, lengthComparator);
538 result.addAll(ordered);
541 System.out.println("searchDuration: " + (System.currentTimeMillis() - startMills));
545 private String normalizeToken(final String searchToken) {
546 if (TransliteratorManager.init(null, null)) {
547 final Transliterator normalizer = normalizer();
548 return normalizer.transliterate(searchToken);
550 // Do our best since the Transliterators aren't up yet.
551 return searchToken.toLowerCase();