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 if (dict.dictFileVersion >= 2) {
162 raf.writeInt(mainTokenCount);
164 RAFList.write(raf, sortedIndexEntries, new IndexEntrySerializer(null), 32, true);
165 StringUtil.writeVarInt(raf, stoplist.size());
166 for (String i : stoplist) {
169 UniformRAFList.write(raf, rows, new RowBase.Serializer(this), 3 /* bytes per entry */);
172 public void print(final PrintStream out) {
173 for (final RowBase row : rows) {
178 private final class IndexEntrySerializer implements RAFSerializer<IndexEntry> {
179 private final FileChannel ch;
181 IndexEntrySerializer(FileChannel ch) {
186 public IndexEntry read(DataInput raf) throws IOException {
187 return new IndexEntry(Index.this, ch, raf);
191 public void write(DataOutput raf, IndexEntry t) throws IOException {
196 public static final class IndexEntry implements RAFSerializable<Index.IndexEntry> {
197 public final String token;
198 private final String normalizedToken;
199 public final int startRow;
200 final int numRows; // doesn't count the token row!
201 public List<HtmlEntry> htmlEntries;
203 public IndexEntry(final Index index, final String token, final String normalizedToken,
204 final int startRow, final int numRows) {
205 assert token.equals(token.trim());
206 assert token.length() > 0;
208 this.normalizedToken = normalizedToken;
209 this.startRow = startRow;
210 this.numRows = numRows;
211 this.htmlEntries = new ArrayList<>();
214 IndexEntry(final Index index, final FileChannel ch, final DataInput raf) throws IOException {
215 token = raf.readUTF();
216 if (index.dict.dictFileVersion >= 7) {
217 startRow = StringUtil.readVarInt(raf);
218 numRows = StringUtil.readVarInt(raf);
220 startRow = raf.readInt();
221 numRows = raf.readInt();
223 final boolean hasNormalizedForm = raf.readBoolean();
224 normalizedToken = hasNormalizedForm ? raf.readUTF() : token;
225 if (index.dict.dictFileVersion >= 7) {
226 int size = StringUtil.readVarInt(raf);
228 this.htmlEntries = Collections.emptyList();
230 final int[] htmlEntryIndices = new int[size];
231 for (int i = 0; i < size; ++i) {
232 htmlEntryIndices[i] = StringUtil.readVarInt(raf);
234 this.htmlEntries = new AbstractList<HtmlEntry>() {
236 public HtmlEntry get(int i) {
237 return index.dict.htmlEntries.get(htmlEntryIndices[i]);
241 return htmlEntryIndices.length;
245 } else if (index.dict.dictFileVersion >= 6) {
246 this.htmlEntries = CachingList.create(
247 RAFList.create(ch, index.dict.htmlEntryIndexSerializer,
248 ch.position(), index.dict.dictFileVersion,
249 index.dict.dictInfo + " htmlEntries: "), 1, false);
251 this.htmlEntries = Collections.emptyList();
255 public void write(DataOutput raf) throws IOException {
257 StringUtil.writeVarInt(raf, startRow);
258 StringUtil.writeVarInt(raf, numRows);
259 final boolean hasNormalizedForm = !token.equals(normalizedToken);
260 raf.writeBoolean(hasNormalizedForm);
261 if (hasNormalizedForm) {
262 raf.writeUTF(normalizedToken);
264 StringUtil.writeVarInt(raf, htmlEntries.size());
265 for (HtmlEntry e : htmlEntries)
266 StringUtil.writeVarInt(raf, e.index());
269 public String toString() {
270 return String.format("%s@%d(%d)", token, startRow, numRows);
273 String normalizedToken() {
274 return normalizedToken;
278 private static final TransformingList.Transformer<IndexEntry, String> INDEX_ENTRY_TO_TOKEN = new TransformingList.Transformer<IndexEntry, String>() {
280 public String transform(IndexEntry t1) {
285 public IndexEntry findExact(final String exactToken) {
286 final int result = Collections.binarySearch(
287 TransformingList.create(sortedIndexEntries, INDEX_ENTRY_TO_TOKEN), exactToken,
288 getSortComparator());
290 return sortedIndexEntries.get(result);
295 public IndexEntry findInsertionPoint(String token, final AtomicBoolean interrupted) {
296 final int index = findInsertionPointIndex(token, interrupted);
297 return index != -1 ? sortedIndexEntries.get(index) : null;
300 private int compareIdx(String token, final Comparator<Object> sortCollator, int idx) {
301 final IndexEntry entry = sortedIndexEntries.get(idx);
302 return NormalizeComparator.compareWithoutDash(token, entry.normalizedToken(), sortCollator, dict.dictFileVersion);
305 private int findMatchLen(final Comparator<Object> sortCollator, String a, String b) {
307 int end = Math.min(a.length(), b.length());
310 int mid = (start + end + 1) / 2;
311 if (sortCollator.compare(a.substring(0, mid), b.substring(0, mid)) == 0)
319 private int findInsertionPointIndex(String token, final AtomicBoolean interrupted) {
320 token = normalizeToken(token);
323 int end = sortedIndexEntries.size();
325 final Comparator<Object> sortCollator = sortLanguage.getCollator();
326 while (start < end) {
327 final int mid = (start + end) / 2;
328 if (interrupted.get()) {
331 final IndexEntry midEntry = sortedIndexEntries.get(mid);
333 int comp = NormalizeComparator.compareWithoutDash(token, midEntry.normalizedToken(), sortCollator, dict.dictFileVersion);
335 comp = sortCollator.compare(token, midEntry.normalizedToken());
337 return windBackCase(token, mid, interrupted);
338 } else if (comp < 0) {
339 // System.out.println("Upper bound: " + midEntry + ", norm=" +
340 // midEntry.normalizedToken() + ", mid=" + mid);
342 // Hack for robustness if sort order is broken
344 compareIdx(token, sortCollator, mid + 1) > 0 &&
345 compareIdx(token, sortCollator, mid + 2) > 0) {
351 // System.out.println("Lower bound: " + midEntry + ", norm=" +
352 // midEntry.normalizedToken() + ", mid=" + mid);
354 // Hack for robustness if sort order is broken
355 if (mid - 2 >= start &&
356 compareIdx(token, sortCollator, mid - 1) < 0 &&
357 compareIdx(token, sortCollator, mid - 2) < 0) {
365 // if the word before is the better match, move
367 if (start > 0 && start < sortedIndexEntries.size()) {
368 String prev = sortedIndexEntries.get(start - 1).normalizedToken();
369 String next = sortedIndexEntries.get(start).normalizedToken();
370 if (findMatchLen(sortCollator, token, prev) >= findMatchLen(sortCollator, token, next))
374 // If we search for a substring of a string that's in there, return
376 int result = Math.min(start, sortedIndexEntries.size() - 1);
377 result = windBackCase(sortedIndexEntries.get(result).normalizedToken(), result, interrupted);
381 private final int windBackCase(final String token, int result, final AtomicBoolean interrupted) {
382 while (result > 0 && sortedIndexEntries.get(result - 1).normalizedToken().equals(token)) {
384 if (interrupted.get()) {
391 public IndexInfo getIndexInfo() {
392 return new DictionaryInfo.IndexInfo(shortName, sortedIndexEntries.size(), mainTokenCount);
395 private static final int MAX_SEARCH_ROWS = 1000;
397 private final Map<String, Integer> prefixToNumRows = new HashMap<>();
399 private synchronized final int getUpperBoundOnRowsStartingWith(final String normalizedPrefix,
400 final int maxRows, final AtomicBoolean interrupted) {
401 final Integer numRows = prefixToNumRows.get(normalizedPrefix);
402 if (numRows != null) {
405 final int insertionPointIndex = findInsertionPointIndex(normalizedPrefix, interrupted);
408 for (int index = insertionPointIndex; index < sortedIndexEntries.size(); ++index) {
409 if (interrupted.get()) {
412 final IndexEntry indexEntry = sortedIndexEntries.get(index);
413 if (!indexEntry.normalizedToken.startsWith(normalizedPrefix) &&
414 !NormalizeComparator.withoutDash(indexEntry.normalizedToken).startsWith(normalizedPrefix)) {
417 rowCount += indexEntry.numRows + indexEntry.htmlEntries.size();
418 if (rowCount > maxRows) {
419 System.out.println("Giving up, too many words with prefix: " + normalizedPrefix);
423 prefixToNumRows.put(normalizedPrefix, rowCount);
427 public final List<RowBase> multiWordSearch(
428 final String searchText, final List<String> searchTokens,
429 final AtomicBoolean interrupted) {
430 final long startMills = System.currentTimeMillis();
431 final List<RowBase> result = new ArrayList<>();
433 final Set<String> normalizedNonStoplist = new HashSet<>();
435 String bestPrefix = null;
436 int leastRows = Integer.MAX_VALUE;
437 final StringBuilder searchTokensRegex = new StringBuilder();
438 for (int i = 0; i < searchTokens.size(); ++i) {
439 if (interrupted.get()) {
442 final String searchToken = searchTokens.get(i);
443 final String normalized = normalizeToken(searchTokens.get(i));
444 // Normalize them all.
445 searchTokens.set(i, normalized);
447 if (!stoplist.contains(searchToken)) {
448 if (normalizedNonStoplist.add(normalized)) {
449 final int numRows = getUpperBoundOnRowsStartingWith(normalized,
450 MAX_SEARCH_ROWS, interrupted);
451 if (numRows != -1 && numRows < leastRows) {
453 // We really are done here.
454 return Collections.emptyList();
457 bestPrefix = normalized;
462 if (searchTokensRegex.length() > 0) {
463 searchTokensRegex.append("[\\s]*");
465 searchTokensRegex.append(Pattern.quote(normalized));
467 final Pattern pattern = Pattern.compile(searchTokensRegex.toString());
469 if (bestPrefix == null) {
470 bestPrefix = searchTokens.get(0);
471 System.out.println("Everything was in the stoplist!");
473 System.out.println("Searching using prefix: " + bestPrefix + ", leastRows=" + leastRows
474 + ", searchTokens=" + searchTokens);
476 // Place to store the things that match.
477 final Map<RowMatchType, List<RowBase>> matches = new EnumMap<>(
479 for (final RowMatchType rowMatchType : RowMatchType.values()) {
480 if (rowMatchType != RowMatchType.NO_MATCH) {
481 matches.put(rowMatchType, new ArrayList<RowBase>());
487 final int exactMatchIndex = findInsertionPointIndex(searchText, interrupted);
488 if (exactMatchIndex != -1) {
489 final IndexEntry exactMatch = sortedIndexEntries.get(exactMatchIndex);
490 if (pattern.matcher(exactMatch.token).find()) {
491 matches.get(RowMatchType.TITLE_MATCH).add(rows.get(exactMatch.startRow));
495 final String searchToken = bestPrefix;
496 final int insertionPointIndex = findInsertionPointIndex(searchToken, interrupted);
497 final Set<RowKey> rowsAlreadySeen = new HashSet<>();
498 for (int index = insertionPointIndex; index < sortedIndexEntries.size()
499 && matchCount < MAX_SEARCH_ROWS; ++index) {
500 if (interrupted.get()) {
503 final IndexEntry indexEntry = sortedIndexEntries.get(index);
504 if (!indexEntry.normalizedToken.startsWith(searchToken) &&
505 !NormalizeComparator.withoutDash(indexEntry.normalizedToken).startsWith(searchToken)) {
509 // System.out.println("Searching indexEntry: " + indexEntry.token);
511 // Extra +1 to skip token row.
512 for (int rowIndex = indexEntry.startRow + 1; rowIndex < indexEntry.startRow + 1
514 && rowIndex < rows.size(); ++rowIndex) {
515 if (interrupted.get()) {
518 final RowBase row = rows.get(rowIndex);
519 final RowBase.RowKey rowKey = row.getRowKey();
520 if (rowsAlreadySeen.contains(rowKey)) {
523 rowsAlreadySeen.add(rowKey);
524 final RowMatchType matchType = row.matches(searchTokens, pattern, normalizer(),
526 if (matchType != RowMatchType.NO_MATCH) {
527 matches.get(matchType).add(row);
534 // Sort them into a reasonable order.
535 final RowBase.LengthComparator lengthComparator = new RowBase.LengthComparator(
537 for (final Collection<RowBase> rows : matches.values()) {
538 final List<RowBase> ordered = new ArrayList<>(rows);
539 Collections.sort(ordered, lengthComparator);
540 result.addAll(ordered);
543 System.out.println("searchDuration: " + (System.currentTimeMillis() - startMills));
547 private String normalizeToken(final String searchToken) {
548 if (TransliteratorManager.init(null, null)) {
549 final Transliterator normalizer = normalizer();
550 return normalizer.transliterate(searchToken);
552 // Do our best since the Transliterators aren't up yet.
553 return searchToken.toLowerCase();