src/com/hughes/android/dictionary/engine/Index.java

   1 // Copyright 2011 Google Inc. All Rights Reserved.
   2 //
   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
   6 //
   7 //     http://www.apache.org/licenses/LICENSE-2.0
   8 //
   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.
  14
  15 /**
  16  *
  17  */
  18
  19 package com.hughes.android.dictionary.engine;
  20
  21 import com.hughes.android.dictionary.DictionaryInfo;
  22 import com.hughes.android.dictionary.DictionaryInfo.IndexInfo;
  23 import com.hughes.android.dictionary.engine.RowBase.RowKey;
  24 import com.hughes.util.CachingList;
  25 import com.hughes.util.StringUtil;
  26 import com.hughes.util.TransformingList;
  27 import com.hughes.util.raf.RAFList;
  28 import com.hughes.util.raf.RAFSerializable;
  29 import com.hughes.util.raf.RAFSerializer;
  30 import com.hughes.util.raf.SerializableSerializer;
  31 import com.hughes.util.raf.UniformRAFList;
  32 import java.text.Collator;
  33 import com.ibm.icu.text.Transliterator;
  34
  35 import java.io.DataInput;
  36 import java.io.DataOutput;
  37 import java.io.IOException;
  38 import java.io.PrintStream;
  39 import java.io.RandomAccessFile;
  40 import java.util.AbstractList;
  41 import java.util.ArrayList;
  42 import java.util.Collection;
  43 import java.util.Collections;
  44 import java.util.Comparator;
  45 import java.util.EnumMap;
  46 import java.util.HashSet;
  47 import java.util.HashMap;
  48 import java.util.List;
  49 import java.util.Map;
  50 import java.util.Set;
  51 import java.util.concurrent.atomic.AtomicBoolean;
  52 import java.util.regex.Pattern;
  53
  54 public final class Index implements RAFSerializable<Index> {
  55
  56     static final int CACHE_SIZE = 5000;
  57
  58     public final Dictionary dict;
  59
  60     public final String shortName; // Typically the ISO code for the language.
  61     public final String longName;
  62
  63     // persisted: tells how the entries are sorted.
  64     public final Language sortLanguage;
  65     final String normalizerRules;
  66
  67     // Built from the two above.
  68     private Transliterator normalizer;
  69
  70     // persisted
  71     public final List<IndexEntry> sortedIndexEntries;
  72
  73     // persisted.
  74     public final Set<String> stoplist;
  75
  76     // One big list!
  77     // Various sub-types.
  78     // persisted
  79     public final List<RowBase> rows;
  80     public final boolean swapPairEntries;
  81
  82     // Version 2:
  83     int mainTokenCount = -1;
  84
  85     // --------------------------------------------------------------------------
  86
  87     public Index(final Dictionary dict, final String shortName, final String longName,
  88                  final Language sortLanguage, final String normalizerRules,
  89                  final boolean swapPairEntries, final Set<String> stoplist) {
  90         this.dict = dict;
  91         this.shortName = shortName;
  92         this.longName = longName;
  93         this.sortLanguage = sortLanguage;
  94         this.normalizerRules = normalizerRules;
  95         this.swapPairEntries = swapPairEntries;
  96         sortedIndexEntries = new ArrayList<IndexEntry>();
  97         this.stoplist = stoplist;
  98         rows = new ArrayList<RowBase>();
  99
 100         normalizer = null;
 101     }
 102
 103     /**
 104      * Deferred initialization because it can be slow.
 105      */
 106     public synchronized Transliterator normalizer() {
 107         if (normalizer == null) {
 108             normalizer = TransliteratorManager.get(normalizerRules);
 109         }
 110         return normalizer;
 111     }
 112
 113     /**
 114      * Note that using this comparator probably involves doing too many text
 115      * normalizations.
 116      */
 117     public NormalizeComparator getSortComparator() {
 118         return new NormalizeComparator(normalizer(), sortLanguage.getCollator(), dict.dictFileVersion);
 119     }
 120
 121     public Index(final Dictionary dict, final DataInput inp) throws IOException {
 122         this.dict = dict;
 123         RandomAccessFile raf = (RandomAccessFile)inp;
 124         shortName = raf.readUTF();
 125         longName = raf.readUTF();
 126         final String languageCode = raf.readUTF();
 127         sortLanguage = Language.lookup(languageCode);
 128         normalizerRules = raf.readUTF();
 129         swapPairEntries = raf.readBoolean();
 130         if (sortLanguage == null) {
 131             throw new IOException("Unsupported language: " + languageCode);
 132         }
 133         if (dict.dictFileVersion >= 2) {
 134             mainTokenCount = raf.readInt();
 135         }
 136         sortedIndexEntries = CachingList.create(
 137                                  RAFList.create(raf, indexEntrySerializer, raf.getFilePointer(),
 138                                                 dict.dictFileVersion, dict.dictInfo + " idx " + languageCode + ": "), CACHE_SIZE);
 139         if (dict.dictFileVersion >= 7) {
 140             int count = StringUtil.readVarInt(raf);
 141             stoplist = new HashSet<String>(count);
 142             for (int i = 0; i < count; ++i) {
 143                 stoplist.add(raf.readUTF());
 144             }
 145         } else if (dict.dictFileVersion >= 4) {
 146             stoplist = new SerializableSerializer<Set<String>>().read(raf);
 147         } else {
 148             stoplist = Collections.emptySet();
 149         }
 150         rows = CachingList.create(
 151                    UniformRAFList.create(raf, new RowBase.Serializer(this), raf.getFilePointer()),
 152                    CACHE_SIZE);
 153     }
 154
 155     @Override
 156     public void write(final DataOutput out) throws IOException {
 157         RandomAccessFile raf = (RandomAccessFile)out;
 158         raf.writeUTF(shortName);
 159         raf.writeUTF(longName);
 160         raf.writeUTF(sortLanguage.getIsoCode());
 161         raf.writeUTF(normalizerRules);
 162         raf.writeBoolean(swapPairEntries);
 163         if (dict.dictFileVersion >= 2) {
 164             raf.writeInt(mainTokenCount);
 165         }
 166         RAFList.write(raf, sortedIndexEntries, indexEntrySerializer, 32, true);
 167         StringUtil.writeVarInt(raf, stoplist.size());
 168         for (String i : stoplist) {
 169             raf.writeUTF(i);
 170         }
 171         UniformRAFList.write(raf, rows, new RowBase.Serializer(this), 3 /* bytes per entry */);
 172     }
 173
 174     public void print(final PrintStream out) {
 175         for (final RowBase row : rows) {
 176             row.print(out);
 177         }
 178     }
 179
 180     private final RAFSerializer<IndexEntry> indexEntrySerializer = new RAFSerializer<IndexEntry>() {
 181         @Override
 182         public IndexEntry read(DataInput raf) throws IOException {
 183             return new IndexEntry(Index.this, raf);
 184         }
 185
 186         @Override
 187         public void write(DataOutput raf, IndexEntry t) throws IOException {
 188             t.write(raf);
 189         }
 190     };
 191
 192     public static final class IndexEntry implements RAFSerializable<Index.IndexEntry> {
 193         public final String token;
 194         private final String normalizedToken;
 195         public final int startRow;
 196         public final int numRows; // doesn't count the token row!
 197         public List<HtmlEntry> htmlEntries;
 198
 199         public IndexEntry(final Index index, final String token, final String normalizedToken,
 200                           final int startRow, final int numRows) {
 201             assert token.equals(token.trim());
 202             assert token.length() > 0;
 203             this.token = token;
 204             this.normalizedToken = normalizedToken;
 205             this.startRow = startRow;
 206             this.numRows = numRows;
 207             this.htmlEntries = new ArrayList<HtmlEntry>();
 208         }
 209
 210         public IndexEntry(final Index index, final DataInput raf) throws IOException {
 211             token = raf.readUTF();
 212             if (index.dict.dictFileVersion >= 7) {
 213                 startRow = StringUtil.readVarInt(raf);
 214                 numRows = StringUtil.readVarInt(raf);
 215             } else {
 216                 startRow = raf.readInt();
 217                 numRows = raf.readInt();
 218             }
 219             final boolean hasNormalizedForm = raf.readBoolean();
 220             normalizedToken = hasNormalizedForm ? raf.readUTF() : token;
 221             if (index.dict.dictFileVersion >= 7) {
 222                 int size = StringUtil.readVarInt(raf);
 223                 if (size == 0) {
 224                     this.htmlEntries = Collections.emptyList();
 225                 } else {
 226                     final int[] htmlEntryIndices = new int[size];
 227                     for (int i = 0; i < size; ++i) {
 228                         htmlEntryIndices[i] = StringUtil.readVarInt(raf);
 229                     }
 230                     this.htmlEntries = new AbstractList<HtmlEntry>() {
 231                         @Override
 232                         public HtmlEntry get(int i) {
 233                             return index.dict.htmlEntries.get(htmlEntryIndices[i]);
 234                         }
 235                         @Override
 236                         public int size() {
 237                             return htmlEntryIndices.length;
 238                         }
 239                     };
 240                 }
 241             } else if (index.dict.dictFileVersion >= 6) {
 242                 this.htmlEntries = CachingList.create(
 243                                        RAFList.create((RandomAccessFile)raf, index.dict.htmlEntryIndexSerializer,
 244                                                       ((RandomAccessFile)raf).getFilePointer(), index.dict.dictFileVersion,
 245                                                       index.dict.dictInfo + " htmlEntries: "), 1);
 246             } else {
 247                 this.htmlEntries = Collections.emptyList();
 248             }
 249         }
 250
 251         public void write(DataOutput raf) throws IOException {
 252             raf.writeUTF(token);
 253             StringUtil.writeVarInt(raf, startRow);
 254             StringUtil.writeVarInt(raf, numRows);
 255             final boolean hasNormalizedForm = !token.equals(normalizedToken);
 256             raf.writeBoolean(hasNormalizedForm);
 257             if (hasNormalizedForm) {
 258                 raf.writeUTF(normalizedToken);
 259             }
 260             StringUtil.writeVarInt(raf, htmlEntries.size());
 261             for (HtmlEntry e : htmlEntries)
 262                 StringUtil.writeVarInt(raf, e.index());
 263         }
 264
 265         public String toString() {
 266             return String.format("%s@%d(%d)", token, startRow, numRows);
 267         }
 268
 269         public String normalizedToken() {
 270             return normalizedToken;
 271         }
 272     }
 273
 274     static final TransformingList.Transformer<IndexEntry, String> INDEX_ENTRY_TO_TOKEN = new TransformingList.Transformer<IndexEntry, String>() {
 275         @Override
 276         public String transform(IndexEntry t1) {
 277             return t1.token;
 278         }
 279     };
 280
 281     public IndexEntry findExact(final String exactToken) {
 282         final int result = Collections.binarySearch(
 283                                TransformingList.create(sortedIndexEntries, INDEX_ENTRY_TO_TOKEN), exactToken,
 284                                getSortComparator());
 285         if (result >= 0) {
 286             return sortedIndexEntries.get(result);
 287         }
 288         return null;
 289     }
 290
 291     public IndexEntry findInsertionPoint(String token, final AtomicBoolean interrupted) {
 292         final int index = findInsertionPointIndex(token, interrupted);
 293         return index != -1 ? sortedIndexEntries.get(index) : null;
 294     }
 295
 296     public int findInsertionPointIndex(String token, final AtomicBoolean interrupted) {
 297         token = normalizeToken(token);
 298
 299         int start = 0;
 300         int end = sortedIndexEntries.size();
 301
 302         final Comparator sortCollator = sortLanguage.getCollator();
 303         while (start < end) {
 304             final int mid = (start + end) / 2;
 305             if (interrupted.get()) {
 306                 return -1;
 307             }
 308             final IndexEntry midEntry = sortedIndexEntries.get(mid);
 309
 310             int comp = NormalizeComparator.compareWithoutDash(token, midEntry.normalizedToken(), sortCollator, dict.dictFileVersion);
 311             if (comp == 0)
 312                 comp = sortCollator.compare(token, midEntry.normalizedToken());
 313             if (comp == 0) {
 314                 final int result = windBackCase(token, mid, interrupted);
 315                 return result;
 316             } else if (comp < 0) {
 317                 // System.out.println("Upper bound: " + midEntry + ", norm=" +
 318                 // midEntry.normalizedToken() + ", mid=" + mid);
 319                 end = mid;
 320             } else {
 321                 // System.out.println("Lower bound: " + midEntry + ", norm=" +
 322                 // midEntry.normalizedToken() + ", mid=" + mid);
 323                 start = mid + 1;
 324             }
 325         }
 326
 327         // If we search for a substring of a string that's in there, return
 328         // that.
 329         int result = Math.min(start, sortedIndexEntries.size() - 1);
 330         result = windBackCase(sortedIndexEntries.get(result).normalizedToken(), result, interrupted);
 331         return result;
 332     }
 333
 334     private final int windBackCase(final String token, int result, final AtomicBoolean interrupted) {
 335         while (result > 0 && sortedIndexEntries.get(result - 1).normalizedToken().equals(token)) {
 336             --result;
 337             if (interrupted.get()) {
 338                 return result;
 339             }
 340         }
 341         return result;
 342     }
 343
 344     public IndexInfo getIndexInfo() {
 345         return new DictionaryInfo.IndexInfo(shortName, sortedIndexEntries.size(), mainTokenCount);
 346     }
 347
 348     private static final int MAX_SEARCH_ROWS = 1000;
 349
 350     private final Map<String, Integer> prefixToNumRows = new HashMap<String, Integer>();
 351
 352     private synchronized final int getUpperBoundOnRowsStartingWith(final String normalizedPrefix,
 353             final int maxRows, final AtomicBoolean interrupted) {
 354         final Integer numRows = prefixToNumRows.get(normalizedPrefix);
 355         if (numRows != null) {
 356             return numRows;
 357         }
 358         final int insertionPointIndex = findInsertionPointIndex(normalizedPrefix, interrupted);
 359
 360         int rowCount = 0;
 361         for (int index = insertionPointIndex; index < sortedIndexEntries.size(); ++index) {
 362             if (interrupted.get()) {
 363                 return -1;
 364             }
 365             final IndexEntry indexEntry = sortedIndexEntries.get(index);
 366             if (!indexEntry.normalizedToken.startsWith(normalizedPrefix)) {
 367                 break;
 368             }
 369             rowCount += indexEntry.numRows + indexEntry.htmlEntries.size();
 370             if (rowCount > maxRows) {
 371                 System.out.println("Giving up, too many words with prefix: " + normalizedPrefix);
 372                 break;
 373             }
 374         }
 375         prefixToNumRows.put(normalizedPrefix, numRows);
 376         return rowCount;
 377     }
 378
 379     public final List<RowBase> multiWordSearch(
 380         final String searchText, final List<String> searchTokens,
 381         final AtomicBoolean interrupted) {
 382         final long startMills = System.currentTimeMillis();
 383         final List<RowBase> result = new ArrayList<RowBase>();
 384
 385         final Set<String> normalizedNonStoplist = new HashSet<String>();
 386
 387         String bestPrefix = null;
 388         int leastRows = Integer.MAX_VALUE;
 389         final StringBuilder searchTokensRegex = new StringBuilder();
 390         for (int i = 0; i < searchTokens.size(); ++i) {
 391             if (interrupted.get()) {
 392                 return null;
 393             }
 394             final String searchToken = searchTokens.get(i);
 395             final String normalized = normalizeToken(searchTokens.get(i));
 396             // Normalize them all.
 397             searchTokens.set(i, normalized);
 398
 399             if (!stoplist.contains(searchToken)) {
 400                 if (normalizedNonStoplist.add(normalized)) {
 401                     final int numRows = getUpperBoundOnRowsStartingWith(normalized,
 402                                         MAX_SEARCH_ROWS, interrupted);
 403                     if (numRows != -1 && numRows < leastRows) {
 404                         if (numRows == 0) {
 405                             // We really are done here.
 406                             return Collections.emptyList();
 407                         }
 408                         leastRows = numRows;
 409                         bestPrefix = normalized;
 410                     }
 411                 }
 412             }
 413
 414             if (searchTokensRegex.length() > 0) {
 415                 searchTokensRegex.append("[\\s]*");
 416             }
 417             searchTokensRegex.append(Pattern.quote(normalized));
 418         }
 419         final Pattern pattern = Pattern.compile(searchTokensRegex.toString());
 420
 421         if (bestPrefix == null) {
 422             bestPrefix = searchTokens.get(0);
 423             System.out.println("Everything was in the stoplist!");
 424         }
 425         System.out.println("Searching using prefix: " + bestPrefix + ", leastRows=" + leastRows
 426                            + ", searchTokens=" + searchTokens);
 427
 428         // Place to store the things that match.
 429         final Map<RowMatchType, List<RowBase>> matches = new EnumMap<RowMatchType, List<RowBase>>(
 430             RowMatchType.class);
 431         for (final RowMatchType rowMatchType : RowMatchType.values()) {
 432             if (rowMatchType != RowMatchType.NO_MATCH) {
 433                 matches.put(rowMatchType, new ArrayList<RowBase>());
 434             }
 435         }
 436
 437         int matchCount = 0;
 438
 439         final int exactMatchIndex = findInsertionPointIndex(searchText, interrupted);
 440         if (exactMatchIndex != -1) {
 441             final IndexEntry exactMatch = sortedIndexEntries.get(exactMatchIndex);
 442             if (pattern.matcher(exactMatch.token).find()) {
 443                 matches.get(RowMatchType.TITLE_MATCH).add(rows.get(exactMatch.startRow));
 444             }
 445         }
 446
 447         final String searchToken = bestPrefix;
 448         final int insertionPointIndex = findInsertionPointIndex(searchToken, interrupted);
 449         final Set<RowKey> rowsAlreadySeen = new HashSet<RowBase.RowKey>();
 450         for (int index = insertionPointIndex; index < sortedIndexEntries.size()
 451                 && matchCount < MAX_SEARCH_ROWS; ++index) {
 452             if (interrupted.get()) {
 453                 return null;
 454             }
 455             final IndexEntry indexEntry = sortedIndexEntries.get(index);
 456             if (!indexEntry.normalizedToken.startsWith(searchToken)) {
 457                 break;
 458             }
 459
 460             // System.out.println("Searching indexEntry: " + indexEntry.token);
 461
 462             // Extra +1 to skip token row.
 463             for (int rowIndex = indexEntry.startRow + 1; rowIndex < indexEntry.startRow + 1
 464                     + indexEntry.numRows
 465                     && rowIndex < rows.size(); ++rowIndex) {
 466                 if (interrupted.get()) {
 467                     return null;
 468                 }
 469                 final RowBase row = rows.get(rowIndex);
 470                 final RowBase.RowKey rowKey = row.getRowKey();
 471                 if (rowsAlreadySeen.contains(rowKey)) {
 472                     continue;
 473                 }
 474                 rowsAlreadySeen.add(rowKey);
 475                 final RowMatchType matchType = row.matches(searchTokens, pattern, normalizer(),
 476                                                swapPairEntries);
 477                 if (matchType != RowMatchType.NO_MATCH) {
 478                     matches.get(matchType).add(row);
 479                     ++matchCount;
 480                 }
 481             }
 482         }
 483         // } // searchTokens
 484
 485         // Sort them into a reasonable order.
 486         final RowBase.LengthComparator lengthComparator = new RowBase.LengthComparator(
 487             swapPairEntries);
 488         for (final Collection<RowBase> rows : matches.values()) {
 489             final List<RowBase> ordered = new ArrayList<RowBase>(rows);
 490             Collections.sort(ordered, lengthComparator);
 491             result.addAll(ordered);
 492         }
 493
 494         System.out.println("searchDuration: " + (System.currentTimeMillis() - startMills));
 495         return result;
 496     }
 497
 498     private String normalizeToken(final String searchToken) {
 499         if (TransliteratorManager.init(null)) {
 500             final Transliterator normalizer = normalizer();
 501             return normalizer.transliterate(searchToken);
 502         } else {
 503             // Do our best since the Transliterators aren't up yet.
 504             return searchToken.toLowerCase();
 505         }
 506     }
 507
 508 }