X-Git-Url: http://gitweb.fperrin.net/?a=blobdiff_plain;f=src%2Fcom%2Fhughes%2Fandroid%2Fdictionary%2Fengine%2FIndex.java;h=fd69bb857e4bd53c08152211c91bccf4115e3d10;hb=60b71b3a5c6fbcd12fbff3d814e8199aec7fba92;hp=e52ac62f74aaaf65a722b2c2f930ff02fbc2096b;hpb=1ba421ef7c195d84bdb140af232b922014c8d812;p=Dictionary.git diff --git a/src/com/hughes/android/dictionary/engine/Index.java b/src/com/hughes/android/dictionary/engine/Index.java index e52ac62..fd69bb8 100644 --- a/src/com/hughes/android/dictionary/engine/Index.java +++ b/src/com/hughes/android/dictionary/engine/Index.java @@ -12,26 +12,8 @@ // See the License for the specific language governing permissions and // limitations under the License. -/** - * - */ - package com.hughes.android.dictionary.engine; -import com.hughes.android.dictionary.DictionaryInfo; -import com.hughes.android.dictionary.DictionaryInfo.IndexInfo; -import com.hughes.android.dictionary.engine.RowBase.RowKey; -import com.hughes.util.CachingList; -import com.hughes.util.StringUtil; -import com.hughes.util.TransformingList; -import com.hughes.util.raf.RAFList; -import com.hughes.util.raf.RAFSerializable; -import com.hughes.util.raf.RAFSerializer; -import com.hughes.util.raf.SerializableSerializer; -import com.hughes.util.raf.UniformRAFList; -import java.text.Collator; -import com.ibm.icu.text.Transliterator; - import java.io.DataInput; import java.io.DataOutput; import java.io.IOException; @@ -43,17 +25,29 @@ import java.util.Collection; import java.util.Collections; import java.util.Comparator; import java.util.EnumMap; -import java.util.HashSet; import java.util.HashMap; +import java.util.HashSet; import java.util.List; import java.util.Map; import java.util.Set; import java.util.concurrent.atomic.AtomicBoolean; import java.util.regex.Pattern; -public final class Index implements RAFSerializable { +import com.hughes.android.dictionary.DictionaryInfo; +import com.hughes.android.dictionary.DictionaryInfo.IndexInfo; +import com.hughes.android.dictionary.engine.RowBase.RowKey; +import com.hughes.util.CachingList; +import com.hughes.util.DataInputBuffer; +import com.hughes.util.StringUtil; +import com.hughes.util.TransformingList; +import com.hughes.util.raf.RAFList; +import com.hughes.util.raf.RAFSerializer; +import com.hughes.util.raf.UniformRAFList; +import com.ibm.icu.text.Transliterator; + +public final class Index { - static final int CACHE_SIZE = 5000; + private static final int CACHE_SIZE = 5000; public final Dictionary dict; @@ -62,7 +56,7 @@ public final class Index implements RAFSerializable { // persisted: tells how the entries are sorted. public final Language sortLanguage; - final String normalizerRules; + public final String normalizerRules; // Built from the two above. private Transliterator normalizer; @@ -80,22 +74,23 @@ public final class Index implements RAFSerializable { public final boolean swapPairEntries; // Version 2: - int mainTokenCount = -1; + @SuppressWarnings("WeakerAccess") + public int mainTokenCount = -1; // -------------------------------------------------------------------------- public Index(final Dictionary dict, final String shortName, final String longName, - final Language sortLanguage, final String normalizerRules, - final boolean swapPairEntries, final Set stoplist) { + final Language sortLanguage, final String normalizerRules, + final boolean swapPairEntries, final Set stoplist) { this.dict = dict; this.shortName = shortName; this.longName = longName; this.sortLanguage = sortLanguage; this.normalizerRules = normalizerRules; this.swapPairEntries = swapPairEntries; - sortedIndexEntries = new ArrayList(); + sortedIndexEntries = new ArrayList<>(); this.stoplist = stoplist; - rows = new ArrayList(); + rows = new ArrayList<>(); normalizer = null; } @@ -103,6 +98,7 @@ public final class Index implements RAFSerializable { /** * Deferred initialization because it can be slow. */ + @SuppressWarnings("WeakerAccess") public synchronized Transliterator normalizer() { if (normalizer == null) { normalizer = TransliteratorManager.get(normalizerRules); @@ -114,45 +110,48 @@ public final class Index implements RAFSerializable { * Note that using this comparator probably involves doing too many text * normalizations. */ + @SuppressWarnings("WeakerAccess") public NormalizeComparator getSortComparator() { return new NormalizeComparator(normalizer(), sortLanguage.getCollator(), dict.dictFileVersion); } - public Index(final Dictionary dict, final DataInput inp) throws IOException { + public Index(final Dictionary dict, final DataInputBuffer raf) throws IOException { this.dict = dict; - RandomAccessFile raf = (RandomAccessFile)inp; shortName = raf.readUTF(); longName = raf.readUTF(); final String languageCode = raf.readUTF(); sortLanguage = Language.lookup(languageCode); normalizerRules = raf.readUTF(); swapPairEntries = raf.readBoolean(); - if (sortLanguage == null) { - throw new IOException("Unsupported language: " + languageCode); - } if (dict.dictFileVersion >= 2) { mainTokenCount = raf.readInt(); } sortedIndexEntries = CachingList.create( - RAFList.create(raf, indexEntrySerializer, raf.getFilePointer(), - dict.dictFileVersion, dict.dictInfo + " idx " + languageCode + ": "), CACHE_SIZE); + RAFList.create(raf, new IndexEntrySerializer(), + dict.dictFileVersion, dict.dictInfo + " idx " + languageCode + ": "), CACHE_SIZE, true); if (dict.dictFileVersion >= 7) { int count = StringUtil.readVarInt(raf); - stoplist = new HashSet(count); + stoplist = new HashSet<>(count); for (int i = 0; i < count; ++i) { stoplist.add(raf.readUTF()); } - } else if (dict.dictFileVersion >= 4) { - stoplist = new SerializableSerializer>().read(raf); + } else if (dict.dictFileVersion >= 4) { + raf.readInt(); // length + raf.skipBytes(52); + stoplist = new HashSet<>(); + byte b; + while ((b = raf.readByte()) == 0x74) { + stoplist.add(raf.readUTF()); + } + if (b != 0x78) throw new IOException("Invalid data in dictionary stoplist!"); } else { stoplist = Collections.emptySet(); } rows = CachingList.create( - UniformRAFList.create(raf, new RowBase.Serializer(this), raf.getFilePointer()), - CACHE_SIZE); + UniformRAFList.create(raf, new RowBase.Serializer(this)), + CACHE_SIZE, true); } - @Override public void write(final DataOutput out) throws IOException { RandomAccessFile raf = (RandomAccessFile)out; raf.writeUTF(shortName); @@ -160,10 +159,8 @@ public final class Index implements RAFSerializable { raf.writeUTF(sortLanguage.getIsoCode()); raf.writeUTF(normalizerRules); raf.writeBoolean(swapPairEntries); - if (dict.dictFileVersion >= 2) { - raf.writeInt(mainTokenCount); - } - RAFList.write(raf, sortedIndexEntries, indexEntrySerializer, 32, true); + raf.writeInt(mainTokenCount); + RAFList.write(raf, sortedIndexEntries, new IndexEntrySerializer(), 32, true); StringUtil.writeVarInt(raf, stoplist.size()); for (String i : stoplist) { raf.writeUTF(i); @@ -177,7 +174,7 @@ public final class Index implements RAFSerializable { } } - private final RAFSerializer indexEntrySerializer = new RAFSerializer() { + private final class IndexEntrySerializer implements RAFSerializer { @Override public IndexEntry read(DataInput raf) throws IOException { return new IndexEntry(Index.this, raf); @@ -187,27 +184,27 @@ public final class Index implements RAFSerializable { public void write(DataOutput raf, IndexEntry t) throws IOException { t.write(raf); } - }; + } - public static final class IndexEntry implements RAFSerializable { + public static final class IndexEntry { public final String token; private final String normalizedToken; public final int startRow; - public final int numRows; // doesn't count the token row! + final int numRows; // doesn't count the token row! public List htmlEntries; public IndexEntry(final Index index, final String token, final String normalizedToken, - final int startRow, final int numRows) { + final int startRow, final int numRows) { assert token.equals(token.trim()); assert token.length() > 0; this.token = token; this.normalizedToken = normalizedToken; this.startRow = startRow; this.numRows = numRows; - this.htmlEntries = new ArrayList(); + this.htmlEntries = new ArrayList<>(); } - public IndexEntry(final Index index, final DataInput raf) throws IOException { + IndexEntry(final Index index, final DataInput raf) throws IOException { token = raf.readUTF(); if (index.dict.dictFileVersion >= 7) { startRow = StringUtil.readVarInt(raf); @@ -227,22 +224,23 @@ public final class Index implements RAFSerializable { for (int i = 0; i < size; ++i) { htmlEntryIndices[i] = StringUtil.readVarInt(raf); } - this.htmlEntries = CachingList.create(new AbstractList() { + this.htmlEntries = new AbstractList() { @Override public HtmlEntry get(int i) { return index.dict.htmlEntries.get(htmlEntryIndices[i]); } + @Override public int size() { return htmlEntryIndices.length; } - }, 1); + }; } } else if (index.dict.dictFileVersion >= 6) { this.htmlEntries = CachingList.create( - RAFList.create((RandomAccessFile)raf, index.dict.htmlEntryIndexSerializer, - ((RandomAccessFile)raf).getFilePointer(), index.dict.dictFileVersion, - index.dict.dictInfo + " htmlEntries: "), 1); + RAFList.create((DataInputBuffer)raf, index.dict.htmlEntryIndexSerializer, + index.dict.dictFileVersion, + index.dict.dictInfo + " htmlEntries: "), 1, false); } else { this.htmlEntries = Collections.emptyList(); } @@ -266,12 +264,12 @@ public final class Index implements RAFSerializable { return String.format("%s@%d(%d)", token, startRow, numRows); } - public String normalizedToken() { + String normalizedToken() { return normalizedToken; } } - static final TransformingList.Transformer INDEX_ENTRY_TO_TOKEN = new TransformingList.Transformer() { + private static final TransformingList.Transformer INDEX_ENTRY_TO_TOKEN = new TransformingList.Transformer() { @Override public String transform(IndexEntry t1) { return t1.token; @@ -280,8 +278,8 @@ public final class Index implements RAFSerializable { public IndexEntry findExact(final String exactToken) { final int result = Collections.binarySearch( - TransformingList.create(sortedIndexEntries, INDEX_ENTRY_TO_TOKEN), exactToken, - getSortComparator()); + TransformingList.create(sortedIndexEntries, INDEX_ENTRY_TO_TOKEN), exactToken, + getSortComparator()); if (result >= 0) { return sortedIndexEntries.get(result); } @@ -293,13 +291,32 @@ public final class Index implements RAFSerializable { return index != -1 ? sortedIndexEntries.get(index) : null; } - public int findInsertionPointIndex(String token, final AtomicBoolean interrupted) { + private int compareIdx(String token, final Comparator sortCollator, int idx) { + final IndexEntry entry = sortedIndexEntries.get(idx); + return NormalizeComparator.compareWithoutDash(token, entry.normalizedToken(), sortCollator, dict.dictFileVersion); + } + + private int findMatchLen(final Comparator sortCollator, String a, String b) { + int start = 0; + int end = Math.min(a.length(), b.length()); + while (start < end) + { + int mid = (start + end + 1) / 2; + if (sortCollator.compare(a.substring(0, mid), b.substring(0, mid)) == 0) + start = mid; + else + end = mid - 1; + } + return start; + } + + private int findInsertionPointIndex(String token, final AtomicBoolean interrupted) { token = normalizeToken(token); int start = 0; int end = sortedIndexEntries.size(); - final Comparator sortCollator = sortLanguage.getCollator(); + final Comparator sortCollator = sortLanguage.getCollator(); while (start < end) { final int mid = (start + end) / 2; if (interrupted.get()) { @@ -311,19 +328,43 @@ public final class Index implements RAFSerializable { if (comp == 0) comp = sortCollator.compare(token, midEntry.normalizedToken()); if (comp == 0) { - final int result = windBackCase(token, mid, interrupted); - return result; + return windBackCase(token, mid, interrupted); } else if (comp < 0) { // System.out.println("Upper bound: " + midEntry + ", norm=" + // midEntry.normalizedToken() + ", mid=" + mid); - end = mid; + + // Hack for robustness if sort order is broken + if (mid + 2 < end && + compareIdx(token, sortCollator, mid + 1) > 0 && + compareIdx(token, sortCollator, mid + 2) > 0) { + start = mid; + } else { + end = mid; + } } else { // System.out.println("Lower bound: " + midEntry + ", norm=" + // midEntry.normalizedToken() + ", mid=" + mid); - start = mid + 1; + + // Hack for robustness if sort order is broken + if (mid - 2 >= start && + compareIdx(token, sortCollator, mid - 1) < 0 && + compareIdx(token, sortCollator, mid - 2) < 0) { + end = mid + 1; + } else { + start = mid + 1; + } } } + // if the word before is the better match, move + // our result to it + if (start > 0 && start < sortedIndexEntries.size()) { + String prev = sortedIndexEntries.get(start - 1).normalizedToken(); + String next = sortedIndexEntries.get(start).normalizedToken(); + if (findMatchLen(sortCollator, token, prev) >= findMatchLen(sortCollator, token, next)) + start--; + } + // If we search for a substring of a string that's in there, return // that. int result = Math.min(start, sortedIndexEntries.size() - 1); @@ -331,7 +372,7 @@ public final class Index implements RAFSerializable { return result; } - private final int windBackCase(final String token, int result, final AtomicBoolean interrupted) { + private int windBackCase(final String token, int result, final AtomicBoolean interrupted) { while (result > 0 && sortedIndexEntries.get(result - 1).normalizedToken().equals(token)) { --result; if (interrupted.get()) { @@ -347,10 +388,10 @@ public final class Index implements RAFSerializable { private static final int MAX_SEARCH_ROWS = 1000; - private final Map prefixToNumRows = new HashMap(); + private final Map prefixToNumRows = new HashMap<>(); - private synchronized final int getUpperBoundOnRowsStartingWith(final String normalizedPrefix, - final int maxRows, final AtomicBoolean interrupted) { + private synchronized int getUpperBoundOnRowsStartingWith(final String normalizedPrefix, + final int maxRows, final AtomicBoolean interrupted) { final Integer numRows = prefixToNumRows.get(normalizedPrefix); if (numRows != null) { return numRows; @@ -363,7 +404,8 @@ public final class Index implements RAFSerializable { return -1; } final IndexEntry indexEntry = sortedIndexEntries.get(index); - if (!indexEntry.normalizedToken.startsWith(normalizedPrefix)) { + if (!indexEntry.normalizedToken.startsWith(normalizedPrefix) && + !NormalizeComparator.withoutDash(indexEntry.normalizedToken).startsWith(normalizedPrefix)) { break; } rowCount += indexEntry.numRows + indexEntry.htmlEntries.size(); @@ -372,17 +414,17 @@ public final class Index implements RAFSerializable { break; } } - prefixToNumRows.put(normalizedPrefix, numRows); + prefixToNumRows.put(normalizedPrefix, rowCount); return rowCount; } public final List multiWordSearch( - final String searchText, final List searchTokens, - final AtomicBoolean interrupted) { + final String searchText, final List searchTokens, + final AtomicBoolean interrupted) { final long startMills = System.currentTimeMillis(); - final List result = new ArrayList(); + final List result = new ArrayList<>(); - final Set normalizedNonStoplist = new HashSet(); + final Set normalizedNonStoplist = new HashSet<>(); String bestPrefix = null; int leastRows = Integer.MAX_VALUE; @@ -399,7 +441,7 @@ public final class Index implements RAFSerializable { if (!stoplist.contains(searchToken)) { if (normalizedNonStoplist.add(normalized)) { final int numRows = getUpperBoundOnRowsStartingWith(normalized, - MAX_SEARCH_ROWS, interrupted); + MAX_SEARCH_ROWS, interrupted); if (numRows != -1 && numRows < leastRows) { if (numRows == 0) { // We really are done here. @@ -423,10 +465,10 @@ public final class Index implements RAFSerializable { System.out.println("Everything was in the stoplist!"); } System.out.println("Searching using prefix: " + bestPrefix + ", leastRows=" + leastRows - + ", searchTokens=" + searchTokens); + + ", searchTokens=" + searchTokens); // Place to store the things that match. - final Map> matches = new EnumMap>( + final Map> matches = new EnumMap<>( RowMatchType.class); for (final RowMatchType rowMatchType : RowMatchType.values()) { if (rowMatchType != RowMatchType.NO_MATCH) { @@ -446,14 +488,15 @@ public final class Index implements RAFSerializable { final String searchToken = bestPrefix; final int insertionPointIndex = findInsertionPointIndex(searchToken, interrupted); - final Set rowsAlreadySeen = new HashSet(); + final Set rowsAlreadySeen = new HashSet<>(); for (int index = insertionPointIndex; index < sortedIndexEntries.size() && matchCount < MAX_SEARCH_ROWS; ++index) { if (interrupted.get()) { return null; } final IndexEntry indexEntry = sortedIndexEntries.get(index); - if (!indexEntry.normalizedToken.startsWith(searchToken)) { + if (!indexEntry.normalizedToken.startsWith(searchToken) && + !NormalizeComparator.withoutDash(indexEntry.normalizedToken).startsWith(searchToken)) { break; } @@ -473,7 +516,7 @@ public final class Index implements RAFSerializable { } rowsAlreadySeen.add(rowKey); final RowMatchType matchType = row.matches(searchTokens, pattern, normalizer(), - swapPairEntries); + swapPairEntries); if (matchType != RowMatchType.NO_MATCH) { matches.get(matchType).add(row); ++matchCount; @@ -484,9 +527,9 @@ public final class Index implements RAFSerializable { // Sort them into a reasonable order. final RowBase.LengthComparator lengthComparator = new RowBase.LengthComparator( - swapPairEntries); + swapPairEntries); for (final Collection rows : matches.values()) { - final List ordered = new ArrayList(rows); + final List ordered = new ArrayList<>(rows); Collections.sort(ordered, lengthComparator); result.addAll(ordered); } @@ -496,7 +539,7 @@ public final class Index implements RAFSerializable { } private String normalizeToken(final String searchToken) { - if (TransliteratorManager.init(null)) { + if (TransliteratorManager.init(null, null)) { final Transliterator normalizer = normalizer(); return normalizer.transliterate(searchToken); } else {