diff options
Diffstat (limited to 'java/src/com/android/inputmethod/latin/makedict')
15 files changed, 1686 insertions, 563 deletions
diff --git a/java/src/com/android/inputmethod/latin/makedict/AbstractDictDecoder.java b/java/src/com/android/inputmethod/latin/makedict/AbstractDictDecoder.java index 9f7f502ea..f8fa68f45 100644 --- a/java/src/com/android/inputmethod/latin/makedict/AbstractDictDecoder.java +++ b/java/src/com/android/inputmethod/latin/makedict/AbstractDictDecoder.java @@ -32,35 +32,35 @@ import java.util.TreeMap; * A base class of the binary dictionary decoder. */ public abstract class AbstractDictDecoder implements DictDecoder { - protected FileHeader readHeader(final DictBuffer dictBuffer) + private static final int SUCCESS = 0; + private static final int ERROR_CANNOT_READ = 1; + private static final int ERROR_WRONG_FORMAT = 2; + + protected FileHeader readHeader(final DictBuffer headerBuffer) throws IOException, UnsupportedFormatException { - if (dictBuffer == null) { + if (headerBuffer == null) { openDictBuffer(); } - final int version = HeaderReader.readVersion(dictBuffer); + final int version = HeaderReader.readVersion(headerBuffer); if (version < FormatSpec.MINIMUM_SUPPORTED_VERSION || version > FormatSpec.MAXIMUM_SUPPORTED_VERSION) { throw new UnsupportedFormatException("Unsupported version : " + version); } // TODO: Remove this field. - final int optionsFlags = HeaderReader.readOptionFlags(dictBuffer); - - final int headerSize = HeaderReader.readHeaderSize(dictBuffer); - + final int optionsFlags = HeaderReader.readOptionFlags(headerBuffer); + final int headerSize = HeaderReader.readHeaderSize(headerBuffer); if (headerSize < 0) { throw new UnsupportedFormatException("header size can't be negative."); } - final HashMap<String, String> attributes = HeaderReader.readAttributes(dictBuffer, + final HashMap<String, String> attributes = HeaderReader.readAttributes(headerBuffer, headerSize); final FileHeader header = new FileHeader(headerSize, - new FusionDictionary.DictionaryOptions(attributes, - 0 != (optionsFlags & FormatSpec.GERMAN_UMLAUT_PROCESSING_FLAG), - 0 != (optionsFlags & FormatSpec.FRENCH_LIGATURE_PROCESSING_FLAG)), - new FormatOptions(version, - 0 != (optionsFlags & FormatSpec.SUPPORTS_DYNAMIC_UPDATE))); + new FusionDictionary.DictionaryOptions(attributes), + new FormatOptions(version, + 0 != (optionsFlags & FormatSpec.CONTAINS_TIMESTAMP_FLAG))); return header; } @@ -203,4 +203,25 @@ public abstract class AbstractDictDecoder implements DictDecoder { return readLength; } } + + /** + * Check whether the header contains the expected information. This is a no-error method, + * that will return an error code and never throw a checked exception. + * @return an error code, either ERROR_* or SUCCESS. + */ + private int checkHeader() { + try { + readHeader(); + } catch (IOException e) { + return ERROR_CANNOT_READ; + } catch (UnsupportedFormatException e) { + return ERROR_WRONG_FORMAT; + } + return SUCCESS; + } + + @Override + public boolean hasValidRawBinaryDictionary() { + return checkHeader() == SUCCESS; + } } diff --git a/java/src/com/android/inputmethod/latin/makedict/BinaryDictDecoderUtils.java b/java/src/com/android/inputmethod/latin/makedict/BinaryDictDecoderUtils.java index 216492b4d..7f0aa777f 100644 --- a/java/src/com/android/inputmethod/latin/makedict/BinaryDictDecoderUtils.java +++ b/java/src/com/android/inputmethod/latin/makedict/BinaryDictDecoderUtils.java @@ -24,12 +24,9 @@ import com.android.inputmethod.latin.makedict.FusionDictionary.PtNodeArray; import com.android.inputmethod.latin.makedict.FusionDictionary.WeightedString; import java.io.File; -import java.io.FileInputStream; -import java.io.FileNotFoundException; import java.io.IOException; import java.io.OutputStream; import java.nio.ByteBuffer; -import java.nio.channels.FileChannel; import java.util.ArrayList; import java.util.Map; import java.util.TreeMap; @@ -169,6 +166,14 @@ public final class BinaryDictDecoderUtils { return size; } + static int getCharArraySize(final int[] chars, final int start, final int end) { + int size = 0; + for (int i = start; i < end; ++i) { + size += getCharSize(chars[i]); + } + return size; + } + /** * Writes a char array to a byte buffer. * @@ -200,8 +205,7 @@ public final class BinaryDictDecoderUtils { * @param word the string to write. * @return the size written, in bytes. */ - static int writeString(final byte[] buffer, final int origin, - final String word) { + static int writeString(final byte[] buffer, final int origin, final String word) { final int length = word.length(); int index = origin; for (int i = 0; i < length; i = word.offsetByCodePoints(i, 1)) { @@ -223,22 +227,62 @@ public final class BinaryDictDecoderUtils { * * This will also write the terminator byte. * - * @param buffer the OutputStream to write to. + * @param stream the OutputStream to write to. * @param word the string to write. + * @return the size written, in bytes. */ - static void writeString(final OutputStream buffer, final String word) throws IOException { + static int writeString(final OutputStream stream, final String word) throws IOException { final int length = word.length(); + int written = 0; for (int i = 0; i < length; i = word.offsetByCodePoints(i, 1)) { final int codePoint = word.codePointAt(i); - if (1 == getCharSize(codePoint)) { - buffer.write((byte) codePoint); + final int charSize = getCharSize(codePoint); + if (1 == charSize) { + stream.write((byte) codePoint); } else { - buffer.write((byte) (0xFF & (codePoint >> 16))); - buffer.write((byte) (0xFF & (codePoint >> 8))); - buffer.write((byte) (0xFF & codePoint)); + stream.write((byte) (0xFF & (codePoint >> 16))); + stream.write((byte) (0xFF & (codePoint >> 8))); + stream.write((byte) (0xFF & codePoint)); } + written += charSize; } - buffer.write(FormatSpec.PTNODE_CHARACTERS_TERMINATOR); + stream.write(FormatSpec.PTNODE_CHARACTERS_TERMINATOR); + written += FormatSpec.PTNODE_TERMINATOR_SIZE; + return written; + } + + /** + * Writes an array of code points with our character format to an OutputStream. + * + * This will also write the terminator byte. + * + * @param stream the OutputStream to write to. + * @param codePoints the array of code points + * @return the size written, in bytes. + */ + // TODO: Merge this method with writeCharArray and rename the various write* methods to + // make the difference clear. + static int writeCodePoints(final OutputStream stream, final int[] codePoints, + final int startIndex, final int endIndex) + throws IOException { + int written = 0; + for (int i = startIndex; i < endIndex; ++i) { + final int codePoint = codePoints[i]; + final int charSize = getCharSize(codePoint); + if (1 == charSize) { + stream.write((byte) codePoint); + } else { + stream.write((byte) (0xFF & (codePoint >> 16))); + stream.write((byte) (0xFF & (codePoint >> 8))); + stream.write((byte) (0xFF & codePoint)); + } + written += charSize; + } + if (endIndex - startIndex > 1) { + stream.write(FormatSpec.PTNODE_CHARACTERS_TERMINATOR); + written += FormatSpec.PTNODE_TERMINATOR_SIZE; + } + return written; } /** @@ -286,7 +330,7 @@ public final class BinaryDictDecoderUtils { static int readChildrenAddress(final DictBuffer dictBuffer, final int optionFlags, final FormatOptions options) { - if (options.mSupportsDynamicUpdate) { + if (options.supportsDynamicUpdate()) { final int address = dictBuffer.readUnsignedInt24(); if (address == 0) return FormatSpec.NO_CHILDREN_ADDRESS; if ((address & FormatSpec.MSB24) != 0) { @@ -496,11 +540,11 @@ public final class BinaryDictDecoderUtils { } // reach the end of the array. - if (options.mSupportsDynamicUpdate) { + if (options.supportsDynamicUpdate()) { final boolean hasValidForwardLink = dictDecoder.readAndFollowForwardLink(); if (!hasValidForwardLink) break; } - } while (options.mSupportsDynamicUpdate && dictDecoder.hasNextPtNodeArray()); + } while (options.supportsDynamicUpdate() && dictDecoder.hasNextPtNodeArray()); final PtNodeArray nodeArray = new PtNodeArray(nodeArrayContents); nodeArray.mCachedAddressBeforeUpdate = nodeArrayOriginPos; @@ -556,7 +600,7 @@ public final class BinaryDictDecoderUtils { Map<Integer, PtNodeArray> reverseNodeArrayMapping = new TreeMap<Integer, PtNodeArray>(); Map<Integer, PtNode> reversePtNodeMapping = new TreeMap<Integer, PtNode>(); - final PtNodeArray root = readNodeArray(dictDecoder, fileHeader.mHeaderSize, + final PtNodeArray root = readNodeArray(dictDecoder, fileHeader.mBodyOffset, reverseNodeArrayMapping, reversePtNodeMapping, fileHeader.mFormatOptions); FusionDictionary newDict = new FusionDictionary(root, fileHeader.mDictionaryOptions); @@ -592,32 +636,10 @@ public final class BinaryDictDecoderUtils { /** * Basic test to find out whether the file is a binary dictionary or not. * - * Concretely this only tests the magic number. - * * @param file The file to test. * @return true if it's a binary dictionary, false otherwise */ public static boolean isBinaryDictionary(final File file) { - FileInputStream inStream = null; - try { - inStream = new FileInputStream(file); - final ByteBuffer buffer = inStream.getChannel().map( - FileChannel.MapMode.READ_ONLY, 0, file.length()); - final int version = getFormatVersion(new ByteBufferDictBuffer(buffer)); - return (version >= FormatSpec.MINIMUM_SUPPORTED_VERSION - && version <= FormatSpec.MAXIMUM_SUPPORTED_VERSION); - } catch (FileNotFoundException e) { - return false; - } catch (IOException e) { - return false; - } finally { - if (inStream != null) { - try { - inStream.close(); - } catch (IOException e) { - // do nothing - } - } - } + return FormatSpec.getDictDecoder(file).hasValidRawBinaryDictionary(); } } diff --git a/java/src/com/android/inputmethod/latin/makedict/BinaryDictEncoderUtils.java b/java/src/com/android/inputmethod/latin/makedict/BinaryDictEncoderUtils.java index f761829de..8ba0797de 100644 --- a/java/src/com/android/inputmethod/latin/makedict/BinaryDictEncoderUtils.java +++ b/java/src/com/android/inputmethod/latin/makedict/BinaryDictEncoderUtils.java @@ -17,9 +17,9 @@ package com.android.inputmethod.latin.makedict; import com.android.inputmethod.latin.makedict.BinaryDictDecoderUtils.CharEncoding; +import com.android.inputmethod.latin.makedict.BinaryDictDecoderUtils.DictBuffer; import com.android.inputmethod.latin.makedict.FormatSpec.FormatOptions; import com.android.inputmethod.latin.makedict.FusionDictionary.PtNode; -import com.android.inputmethod.latin.makedict.FusionDictionary.DictionaryOptions; import com.android.inputmethod.latin.makedict.FusionDictionary.PtNodeArray; import com.android.inputmethod.latin.makedict.FusionDictionary.WeightedString; @@ -160,7 +160,7 @@ public class BinaryDictEncoderUtils { node.mCachedSize = nodeSize; size += nodeSize; } - if (options.mSupportsDynamicUpdate) { + if (options.supportsDynamicUpdate()) { size += FormatSpec.FORWARD_LINK_ADDRESS_SIZE; } ptNodeArray.mCachedSize = size; @@ -245,6 +245,26 @@ public class BinaryDictEncoderUtils { } } + static void writeUIntToDictBuffer(final DictBuffer dictBuffer, final int value, + final int size) { + switch(size) { + case 4: + dictBuffer.put((byte) ((value >> 24) & 0xFF)); + /* fall through */ + case 3: + dictBuffer.put((byte) ((value >> 16) & 0xFF)); + /* fall through */ + case 2: + dictBuffer.put((byte) ((value >> 8) & 0xFF)); + /* fall through */ + case 1: + dictBuffer.put((byte) (value & 0xFF)); + break; + default: + /* nop */ + } + } + // End utility methods // This method is responsible for finding a nice ordering of the nodes that favors run-time @@ -377,7 +397,7 @@ public class BinaryDictEncoderUtils { nodeSize += FormatSpec.PTNODE_FREQUENCY_SIZE; } } - if (formatOptions.mSupportsDynamicUpdate) { + if (formatOptions.supportsDynamicUpdate()) { nodeSize += FormatSpec.SIGNED_CHILDREN_ADDRESS_SIZE; } else if (null != ptNode.mChildren) { nodeSize += getByteSize(getOffsetToTargetNodeArrayDuringUpdate(ptNodeArray, @@ -397,7 +417,7 @@ public class BinaryDictEncoderUtils { ptNode.mCachedSize = nodeSize; size += nodeSize; } - if (formatOptions.mSupportsDynamicUpdate) { + if (formatOptions.supportsDynamicUpdate()) { size += FormatSpec.FORWARD_LINK_ADDRESS_SIZE; } if (ptNodeArray.mCachedSize != size) { @@ -513,7 +533,7 @@ public class BinaryDictEncoderUtils { if (passes > MAX_PASSES) throw new RuntimeException("Too many passes - probably a bug"); } while (changesDone); - if (formatOptions.mSupportsDynamicUpdate) { + if (formatOptions.supportsDynamicUpdate()) { computeParentAddresses(flatNodes); } final PtNodeArray lastPtNodeArray = flatNodes.get(flatNodes.size() - 1); @@ -622,7 +642,7 @@ public class BinaryDictEncoderUtils { byte flags = 0; if (hasMultipleChars) flags |= FormatSpec.FLAG_HAS_MULTIPLE_CHARS; if (isTerminal) flags |= FormatSpec.FLAG_IS_TERMINAL; - if (formatOptions.mSupportsDynamicUpdate) { + if (formatOptions.supportsDynamicUpdate()) { flags |= FormatSpec.FLAG_IS_NOT_MOVED; } else if (true) { switch (childrenAddressSize) { @@ -690,6 +710,13 @@ public class BinaryDictEncoderUtils { + word + " is " + unigramFrequency); bigramFrequency = unigramFrequency; } + bigramFlags += getBigramFrequencyDiff(unigramFrequency, bigramFrequency) + & FormatSpec.FLAG_BIGRAM_SHORTCUT_ATTR_FREQUENCY; + return bigramFlags; + } + + public static int getBigramFrequencyDiff(final int unigramFrequency, + final int bigramFrequency) { // We compute the difference between 255 (which means probability = 1) and the // unigram score. We split this into a number of discrete steps. // Now, the steps are numbered 0~15; 0 represents an increase of 1 step while 15 @@ -723,22 +750,15 @@ public class BinaryDictEncoderUtils { // include this bigram in the dictionary. For now, register as 0, and live with the // small over-estimation that we get in this case. TODO: actually remove this bigram // if discretizedFrequency < 0. - final int finalBigramFrequency = discretizedFrequency > 0 ? discretizedFrequency : 0; - bigramFlags += finalBigramFrequency & FormatSpec.FLAG_BIGRAM_SHORTCUT_ATTR_FREQUENCY; - return bigramFlags; + return discretizedFrequency > 0 ? discretizedFrequency : 0; } /** - * Makes the 2-byte value for options flags. + * Makes the 2-byte value for options flags. Unused at the moment, and always 0. */ - private static final int makeOptionsValue(final FusionDictionary dictionary, - final FormatOptions formatOptions) { - final DictionaryOptions options = dictionary.mOptions; - final boolean hasBigrams = dictionary.hasBigrams(); - return (options.mFrenchLigatureProcessing ? FormatSpec.FRENCH_LIGATURE_PROCESSING_FLAG : 0) - + (options.mGermanUmlautProcessing ? FormatSpec.GERMAN_UMLAUT_PROCESSING_FLAG : 0) - + (hasBigrams ? FormatSpec.CONTAINS_BIGRAMS_FLAG : 0) - + (formatOptions.mSupportsDynamicUpdate ? FormatSpec.SUPPORTS_DYNAMIC_UPDATE : 0); + private static final int makeOptionsValue(final FormatOptions formatOptions) { + // TODO: why doesn't this handle CONTAINS_TIMESTAMP_FLAG? + return 0; } /** @@ -826,7 +846,7 @@ public class BinaryDictEncoderUtils { } dictEncoder.writePtNode(ptNode, parentPosition, formatOptions, dict); } - if (formatOptions.mSupportsDynamicUpdate) { + if (formatOptions.supportsDynamicUpdate()) { dictEncoder.writeForwardLinkAddress(FormatSpec.NO_FORWARD_LINK_ADDRESS); } if (dictEncoder.getPosition() != ptNodeArray.mCachedAddressAfterUpdate @@ -927,7 +947,7 @@ public class BinaryDictEncoderUtils { headerBuffer.write((byte) (0xFF & version)); // Options flags - final int options = makeOptionsValue(dict, formatOptions); + final int options = makeOptionsValue(formatOptions); headerBuffer.write((byte) (0xFF & (options >> 8))); headerBuffer.write((byte) (0xFF & options)); final int headerSizeOffset = headerBuffer.size(); diff --git a/java/src/com/android/inputmethod/latin/makedict/BinaryDictIOUtils.java b/java/src/com/android/inputmethod/latin/makedict/BinaryDictIOUtils.java index 0f7d2f6c9..640d778bb 100644 --- a/java/src/com/android/inputmethod/latin/makedict/BinaryDictIOUtils.java +++ b/java/src/com/android/inputmethod/latin/makedict/BinaryDictIOUtils.java @@ -62,7 +62,7 @@ public final class BinaryDictIOUtils { * Retrieves all node arrays without recursive call. */ private static void readUnigramsAndBigramsBinaryInner(final DictDecoder dictDecoder, - final int headerSize, final Map<Integer, String> words, + final int bodyOffset, final Map<Integer, String> words, final Map<Integer, Integer> frequencies, final Map<Integer, ArrayList<PendingAttribute>> bigrams, final FormatOptions formatOptions) { @@ -71,7 +71,7 @@ public final class BinaryDictIOUtils { Stack<Position> stack = new Stack<Position>(); int index = 0; - Position initPos = new Position(headerSize, 0); + Position initPos = new Position(bodyOffset, 0); stack.push(initPos); while (!stack.empty()) { @@ -112,7 +112,7 @@ public final class BinaryDictIOUtils { } if (p.mPosition == p.mNumOfPtNode) { - if (formatOptions.mSupportsDynamicUpdate) { + if (formatOptions.supportsDynamicUpdate()) { final boolean hasValidForwardLinkAddress = dictDecoder.readAndFollowForwardLink(); if (hasValidForwardLinkAddress && dictDecoder.hasNextPtNodeArray()) { @@ -154,7 +154,7 @@ public final class BinaryDictIOUtils { UnsupportedFormatException { // Read header final FileHeader header = dictDecoder.readHeader(); - readUnigramsAndBigramsBinaryInner(dictDecoder, header.mHeaderSize, words, + readUnigramsAndBigramsBinaryInner(dictDecoder, header.mBodyOffset, words, frequencies, bigrams, header.mFormatOptions); } @@ -228,7 +228,7 @@ public final class BinaryDictIOUtils { // a forward link address that we need to consult and possibly resume // search on the next node array in the linked list. if (foundNextPtNode) break; - if (!header.mFormatOptions.mSupportsDynamicUpdate) { + if (!header.mFormatOptions.supportsDynamicUpdate()) { return FormatSpec.NOT_VALID_WORD; } @@ -245,8 +245,7 @@ public final class BinaryDictIOUtils { /** * @return the size written, in bytes. Always 3 bytes. */ - static int writeSInt24ToBuffer(final DictBuffer dictBuffer, - final int value) { + static int writeSInt24ToBuffer(final DictBuffer dictBuffer, final int value) { final int absValue = Math.abs(value); dictBuffer.put((byte)(((value < 0 ? 0x80 : 0) | (absValue >> 16)) & 0xFF)); dictBuffer.put((byte)((absValue >> 8) & 0xFF)); @@ -301,35 +300,6 @@ public final class BinaryDictIOUtils { } /** - * Write a string to a stream. - * - * @param destination the stream to write. - * @param word the string to be written. - * @return the size written, in bytes. - * @throws IOException - */ - private static int writeString(final OutputStream destination, final String word) - throws IOException { - int size = 0; - final int length = word.length(); - for (int i = 0; i < length; i = word.offsetByCodePoints(i, 1)) { - final int codePoint = word.codePointAt(i); - if (CharEncoding.getCharSize(codePoint) == 1) { - destination.write((byte)codePoint); - size++; - } else { - destination.write((byte)(0xFF & (codePoint >> 16))); - destination.write((byte)(0xFF & (codePoint >> 8))); - destination.write((byte)(0xFF & codePoint)); - size += 3; - } - } - destination.write((byte)FormatSpec.PTNODE_CHARACTERS_TERMINATOR); - size += FormatSpec.PTNODE_TERMINATOR_SIZE; - return size; - } - - /** * Write a PtNode to an output stream from a PtNodeInfo. * A PtNode is an in-memory representation of a node in the patricia trie. * A PtNode info is a container for low-level information about how the @@ -387,7 +357,7 @@ public final class BinaryDictIOUtils { destination.write((byte)BinaryDictEncoderUtils.makeShortcutFlags( shortcutIterator.hasNext(), target.mFrequency)); size++; - size += writeString(destination, target.mWord); + size += CharEncoding.writeString(destination, target.mWord); } } @@ -445,6 +415,27 @@ public final class BinaryDictIOUtils { } /** + * Writes a PtNodeCount to the stream. + * + * @param destination the stream to write. + * @param ptNodeCount the count. + * @return the size written in bytes. + */ + static int writePtNodeCount(final OutputStream destination, final int ptNodeCount) + throws IOException { + final int countSize = BinaryDictIOUtils.getPtNodeCountSize(ptNodeCount); + // the count must fit on one byte or two bytes. + // Please see comments in FormatSpec. + if (countSize != 1 && countSize != 2) { + throw new RuntimeException("Strange size from getPtNodeCountSize : " + countSize); + } + final int encodedPtNodeCount = (countSize == 2) ? + (ptNodeCount | FormatSpec.LARGE_PTNODE_ARRAY_SIZE_FIELD_SIZE_FLAG) : ptNodeCount; + BinaryDictEncoderUtils.writeUIntToStream(destination, encodedPtNodeCount, countSize); + return countSize; + } + + /** * Write a node array to the stream. * * @param destination the stream to write. @@ -454,18 +445,7 @@ public final class BinaryDictIOUtils { */ static int writeNodes(final OutputStream destination, final PtNodeInfo[] infos) throws IOException { - int size = getPtNodeCountSize(infos.length); - switch (getPtNodeCountSize(infos.length)) { - case 1: - destination.write((byte)infos.length); - break; - case 2: - destination.write((byte)(infos.length >> 8)); - destination.write((byte)(infos.length & 0xFF)); - break; - default: - throw new RuntimeException("Invalid node count size."); - } + int size = writePtNodeCount(destination, infos.length); for (final PtNodeInfo info : infos) size += writePtNode(destination, info); writeSInt24ToStream(destination, FormatSpec.NO_FORWARD_LINK_ADDRESS); return size + FormatSpec.FORWARD_LINK_ADDRESS_SIZE; @@ -527,7 +507,7 @@ public final class BinaryDictIOUtils { * Helper method to check whether the node is moved. */ public static boolean isMovedPtNode(final int flags, final FormatOptions options) { - return options.mSupportsDynamicUpdate + return options.supportsDynamicUpdate() && ((flags & FormatSpec.MASK_CHILDREN_ADDRESS_TYPE) == FormatSpec.FLAG_IS_MOVED); } @@ -536,14 +516,14 @@ public final class BinaryDictIOUtils { */ public static boolean supportsDynamicUpdate(final FormatOptions options) { return options.mVersion >= FormatSpec.FIRST_VERSION_WITH_DYNAMIC_UPDATE - && options.mSupportsDynamicUpdate; + && options.supportsDynamicUpdate(); } /** * Helper method to check whether the node is deleted. */ public static boolean isDeletedPtNode(final int flags, final FormatOptions formatOptions) { - return formatOptions.mSupportsDynamicUpdate + return formatOptions.supportsDynamicUpdate() && ((flags & FormatSpec.MASK_CHILDREN_ADDRESS_TYPE) == FormatSpec.FLAG_IS_DELETED); } @@ -566,7 +546,7 @@ public final class BinaryDictIOUtils { static int getChildrenAddressSize(final int optionFlags, final FormatOptions formatOptions) { - if (formatOptions.mSupportsDynamicUpdate) return FormatSpec.SIGNED_CHILDREN_ADDRESS_SIZE; + if (formatOptions.supportsDynamicUpdate()) return FormatSpec.SIGNED_CHILDREN_ADDRESS_SIZE; switch (optionFlags & FormatSpec.MASK_CHILDREN_ADDRESS_TYPE) { case FormatSpec.FLAG_CHILDREN_ADDRESS_TYPE_ONEBYTE: return 1; diff --git a/java/src/com/android/inputmethod/latin/makedict/DictDecoder.java b/java/src/com/android/inputmethod/latin/makedict/DictDecoder.java index 3dbeee099..b4838f00f 100644 --- a/java/src/com/android/inputmethod/latin/makedict/DictDecoder.java +++ b/java/src/com/android/inputmethod/latin/makedict/DictDecoder.java @@ -35,6 +35,7 @@ import java.util.TreeMap; /** * An interface of binary dictionary decoders. */ +// TODO: Straighten out responsibility for the buffer's file pointer. public interface DictDecoder { /** @@ -43,7 +44,7 @@ public interface DictDecoder { public FileHeader readHeader() throws IOException, UnsupportedFormatException; /** - * Reads PtNode from nodeAddress. + * Reads PtNode from ptNodePos. * @param ptNodePos the position of PtNode. * @param formatOptions the format options. * @return PtNodeInfo. @@ -127,7 +128,8 @@ public interface DictDecoder { * Opens the dictionary file and makes DictBuffer. */ @UsedForTesting - public void openDictBuffer() throws FileNotFoundException, IOException; + public void openDictBuffer() throws FileNotFoundException, IOException, + UnsupportedFormatException; @UsedForTesting public boolean isDictBufferOpen(); @@ -228,4 +230,9 @@ public interface DictDecoder { } public void skipPtNode(final FormatOptions formatOptions); + + /** + * @return whether this decoder has a valid binary dictionary that it can decode. + */ + public boolean hasValidRawBinaryDictionary(); } diff --git a/java/src/com/android/inputmethod/latin/makedict/DynamicBinaryDictIOUtils.java b/java/src/com/android/inputmethod/latin/makedict/DynamicBinaryDictIOUtils.java index 336277196..ff03190a3 100644 --- a/java/src/com/android/inputmethod/latin/makedict/DynamicBinaryDictIOUtils.java +++ b/java/src/com/android/inputmethod/latin/makedict/DynamicBinaryDictIOUtils.java @@ -22,6 +22,7 @@ import com.android.inputmethod.latin.makedict.BinaryDictDecoderUtils.DictBuffer; import com.android.inputmethod.latin.makedict.FormatSpec.FileHeader; import com.android.inputmethod.latin.makedict.FormatSpec.FormatOptions; import com.android.inputmethod.latin.makedict.FusionDictionary.WeightedString; +import com.android.inputmethod.latin.utils.CollectionUtils; import java.io.IOException; import java.io.OutputStream; @@ -36,7 +37,7 @@ import java.util.Arrays; @UsedForTesting public final class DynamicBinaryDictIOUtils { private static final boolean DBG = false; - private static final int MAX_JUMPS = 10000; + static final int MAX_JUMPS = 10000; private DynamicBinaryDictIOUtils() { // This utility class is not publicly instantiable. @@ -60,7 +61,7 @@ public final class DynamicBinaryDictIOUtils { final DictBuffer dictBuffer = dictUpdater.getDictBuffer(); final int originalPosition = dictBuffer.position(); dictBuffer.position(ptNodeOriginAddress); - if (!formatOptions.mSupportsDynamicUpdate) { + if (!formatOptions.supportsDynamicUpdate()) { throw new RuntimeException("this file format does not support parent addresses"); } final int flags = dictBuffer.readUnsignedByte(); @@ -101,7 +102,7 @@ public final class DynamicBinaryDictIOUtils { } if (!dictUpdater.readAndFollowForwardLink()) break; if (dictUpdater.getPosition() == FormatSpec.NO_FORWARD_LINK_ADDRESS) break; - } while (formatOptions.mSupportsDynamicUpdate); + } while (formatOptions.supportsDynamicUpdate()); dictUpdater.setPosition(originalPosition); } @@ -217,6 +218,25 @@ public final class DynamicBinaryDictIOUtils { } /** + * Converts a list of WeightedString to a list of PendingAttribute. + */ + public static ArrayList<PendingAttribute> resolveBigramPositions(final DictUpdater dictUpdater, + final ArrayList<WeightedString> bigramStrings) + throws IOException, UnsupportedFormatException { + if (bigramStrings == null) return CollectionUtils.newArrayList(); + final ArrayList<PendingAttribute> bigrams = CollectionUtils.newArrayList(); + for (final WeightedString bigram : bigramStrings) { + final int pos = dictUpdater.getTerminalPosition(bigram.mWord); + if (pos == FormatSpec.NOT_VALID_WORD) { + // TODO: figure out what is the correct thing to do here. + } else { + bigrams.add(new PendingAttribute(bigram.mFrequency, pos)); + } + } + return bigrams; + } + + /** * Insert a word into a binary dictionary. * * @param dictUpdater the dict updater. @@ -238,18 +258,9 @@ public final class DynamicBinaryDictIOUtils { final ArrayList<WeightedString> shortcuts, final boolean isNotAWord, final boolean isBlackListEntry) throws IOException, UnsupportedFormatException { - final ArrayList<PendingAttribute> bigrams = new ArrayList<PendingAttribute>(); + final ArrayList<PendingAttribute> bigrams = resolveBigramPositions(dictUpdater, + bigramStrings); final DictBuffer dictBuffer = dictUpdater.getDictBuffer(); - if (bigramStrings != null) { - for (final WeightedString bigram : bigramStrings) { - int position = dictUpdater.getTerminalPosition(bigram.mWord); - if (position == FormatSpec.NOT_VALID_WORD) { - // TODO: figure out what is the correct thing to do here. - } else { - bigrams.add(new PendingAttribute(bigram.mFrequency, position)); - } - } - } final boolean isTerminal = true; final boolean hasBigrams = !bigrams.isEmpty(); diff --git a/java/src/com/android/inputmethod/latin/makedict/FormatSpec.java b/java/src/com/android/inputmethod/latin/makedict/FormatSpec.java index 5a5d7af6b..20ddba836 100644 --- a/java/src/com/android/inputmethod/latin/makedict/FormatSpec.java +++ b/java/src/com/android/inputmethod/latin/makedict/FormatSpec.java @@ -37,13 +37,11 @@ public final class FormatSpec { * sion * * o | - * p | not used 4 bits - * t | has bigrams ? 1 bit, 1 = yes, 0 = no : CONTAINS_BIGRAMS_FLAG - * i | FRENCH_LIGATURE_PROCESSING_FLAG - * o | supports dynamic updates ? 1 bit, 1 = yes, 0 = no : SUPPORTS_DYNAMIC_UPDATE - * n | GERMAN_UMLAUT_PROCESSING_FLAG - * f | - * lags + * p | not used 3 bits + * t | each unigram and bigram entry has a time stamp? + * i | 1 bit, 1 = yes, 0 = no : CONTAINS_TIMESTAMP_FLAG + * o | + * nflags * * h | * e | size of the file header, 4bytes @@ -80,45 +78,36 @@ public final class FormatSpec { * s * * f | - * o | IF SUPPORTS_DYNAMIC_UPDATE (defined in the file header) - * r | forward link address, 3byte - * w | 1 byte = bbbbbbbb match - * a | case 1xxxxxxx => -((xxxxxxx << 16) + (next byte << 8) + next byte) - * r | otherwise => (xxxxxxx << 16) + (next byte << 8) + next byte - * d | - * linkaddress + * o | forward link address, 3byte + * r | 1 byte = bbbbbbbb match + * w | case 1xxxxxxx => -((xxxxxxx << 16) + (next byte << 8) + next byte) + * a | otherwise => (xxxxxxx << 16) + (next byte << 8) + next byte + * r | + * dlinkaddress */ /* Node (FusionDictionary.PtNode) layout is as follows: - * | IF !SUPPORTS_DYNAMIC_UPDATE - * | addressType xx : mask with MASK_CHILDREN_ADDRESS_TYPE - * | 2 bits, 00 = no children : FLAG_CHILDREN_ADDRESS_TYPE_NOADDRESS - * f | 01 = 1 byte : FLAG_CHILDREN_ADDRESS_TYPE_ONEBYTE - * l | 10 = 2 bytes : FLAG_CHILDREN_ADDRESS_TYPE_TWOBYTES - * a | 11 = 3 bytes : FLAG_CHILDREN_ADDRESS_TYPE_THREEBYTES - * g | ELSE - * s | is moved ? 2 bits, 11 = no : FLAG_IS_NOT_MOVED - * | This must be the same as FLAG_CHILDREN_ADDRESS_TYPE_THREEBYTES - * | 01 = yes : FLAG_IS_MOVED - * | the new address is stored in the same place as the parent address - * | is deleted? 10 = yes : FLAG_IS_DELETED - * | has several chars ? 1 bit, 1 = yes, 0 = no : FLAG_HAS_MULTIPLE_CHARS - * | has a terminal ? 1 bit, 1 = yes, 0 = no : FLAG_IS_TERMINAL - * | has shortcut targets ? 1 bit, 1 = yes, 0 = no : FLAG_HAS_SHORTCUT_TARGETS + * | is moved ? 2 bits, 11 = no : FLAG_IS_NOT_MOVED + * | This must be the same as FLAG_CHILDREN_ADDRESS_TYPE_THREEBYTES + * | 01 = yes : FLAG_IS_MOVED + * f | the new address is stored in the same place as the parent address + * l | is deleted? 10 = yes : FLAG_IS_DELETED + * a | has several chars ? 1 bit, 1 = yes, 0 = no : FLAG_HAS_MULTIPLE_CHARS + * g | has a terminal ? 1 bit, 1 = yes, 0 = no : FLAG_IS_TERMINAL + * s | has shortcut targets ? 1 bit, 1 = yes, 0 = no : FLAG_HAS_SHORTCUT_TARGETS * | has bigrams ? 1 bit, 1 = yes, 0 = no : FLAG_HAS_BIGRAMS * | is not a word ? 1 bit, 1 = yes, 0 = no : FLAG_IS_NOT_A_WORD * | is blacklisted ? 1 bit, 1 = yes, 0 = no : FLAG_IS_BLACKLISTED * * p | - * a | IF SUPPORTS_DYNAMIC_UPDATE (defined in the file header) - * r | parent address, 3byte - * e | 1 byte = bbbbbbbb match - * n | case 1xxxxxxx => -((0xxxxxxx << 16) + (next byte << 8) + next byte) - * t | otherwise => (bbbbbbbb << 16) + (next byte << 8) + next byte - * a | This address is relative to the head of the PtNode. - * d | If the node doesn't have a parent, this field is set to 0. + * a | parent address, 3byte + * r | 1 byte = bbbbbbbb match + * e | case 1xxxxxxx => -((0xxxxxxx << 16) + (next byte << 8) + next byte) + * n | otherwise => (bbbbbbbb << 16) + (next byte << 8) + next byte + * t | This address is relative to the head of the PtNode. + * a | If the node doesn't have a parent, this field is set to 0. * d | - * ress + * dress * * c | IF FLAG_HAS_MULTIPLE_CHARS * h | char, char, char, char n * (1 or 3 bytes) : use PtNodeInfo for i/o helpers @@ -132,23 +121,16 @@ public final class FormatSpec { * e | frequency 1 byte * q | * - * c | IF SUPPORTS_DYNAMIC_UPDATE - * h | children address, 3 bytes - * i | 1 byte = bbbbbbbb match - * l | case 1xxxxxxx => -((0xxxxxxx << 16) + (next byte << 8) + next byte) - * d | otherwise => (bbbbbbbb<<16) + (next byte << 8) + next byte - * r | if this node doesn't have children, this field is set to 0. - * e | (see BinaryDictEncoderUtils#writeVariableSignedAddress) - * n | ELSIF 00 = FLAG_CHILDREN_ADDRESS_TYPE_NOADDRESS == addressType - * a | // nothing - * d | ELSIF 01 = FLAG_CHILDREN_ADDRESS_TYPE_ONEBYTE == addressType - * d | children address, 1 byte - * r | ELSIF 10 = FLAG_CHILDREN_ADDRESS_TYPE_TWOBYTES == addressType - * e | children address, 2 bytes - * s | ELSE // 11 = FLAG_CHILDREN_ADDRESS_TYPE_THREEBYTES = addressType - * s | children address, 3 bytes - * | END - * | This address is relative to the position of this field. + * c | + * h | children address, 3 bytes + * i | 1 byte = bbbbbbbb match + * l | case 1xxxxxxx => -((0xxxxxxx << 16) + (next byte << 8) + next byte) + * d | otherwise => (bbbbbbbb<<16) + (next byte << 8) + next byte + * r | if this node doesn't have children, this field is set to 0. + * e | (see BinaryDictEncoderUtils#writeVariableSignedAddress) + * n | This address is relative to the position of this field. + * a | + * ddress * * | IF FLAG_IS_TERMINAL && FLAG_HAS_SHORTCUT_TARGETS * | shortcut string list @@ -197,20 +179,23 @@ public final class FormatSpec { */ public static final int MAGIC_NUMBER = 0x9BC13AFE; - static final int MINIMUM_SUPPORTED_VERSION = 2; - static final int MAXIMUM_SUPPORTED_VERSION = 4; static final int NOT_A_VERSION_NUMBER = -1; static final int FIRST_VERSION_WITH_DYNAMIC_UPDATE = 3; static final int FIRST_VERSION_WITH_TERMINAL_ID = 4; - static final int VERSION3 = 3; - static final int VERSION4 = 4; + + // These MUST have the same values as the relevant constants in format_utils.h. + // From version 4 on, we use version * 100 + revision as a version number. That allows + // us to change the format during development while having testing devices remove + // older files with each upgrade, while still having a readable versioning scheme. + public static final int VERSION2 = 2; + public static final int VERSION3 = 3; + public static final int VERSION4 = 400; + static final int MINIMUM_SUPPORTED_VERSION = VERSION2; + static final int MAXIMUM_SUPPORTED_VERSION = VERSION4; // These options need to be the same numeric values as the one in the native reading code. - static final int GERMAN_UMLAUT_PROCESSING_FLAG = 0x1; // TODO: Make the native reading code read this variable. - static final int SUPPORTS_DYNAMIC_UPDATE = 0x2; - static final int FRENCH_LIGATURE_PROCESSING_FLAG = 0x4; - static final int CONTAINS_BIGRAMS_FLAG = 0x8; + static final int CONTAINS_TIMESTAMP_FLAG = 0x10; // TODO: Make this value adaptative to content data, store it in the header, and // use it in the reading code. @@ -260,9 +245,12 @@ public final class FormatSpec { static final int PTNODE_ATTRIBUTE_MAX_ADDRESS_SIZE = 3; static final int PTNODE_SHORTCUT_LIST_SIZE_SIZE = 2; - // These values are used only by version 4 or later. + // These values are used only by version 4 or later. They MUST match the definitions in + // ver4_dict_constants.cpp. static final String TRIE_FILE_EXTENSION = ".trie"; + public static final String HEADER_FILE_EXTENSION = ".header"; static final String FREQ_FILE_EXTENSION = ".freq"; + static final String UNIGRAM_TIMESTAMP_FILE_EXTENSION = ".timestamp"; // tat = Terminal Address Table static final String TERMINAL_ADDRESS_TABLE_FILE_EXTENSION = ".tat"; static final String BIGRAM_FILE_EXTENSION = ".bigram"; @@ -271,19 +259,25 @@ public final class FormatSpec { static final String CONTENT_TABLE_FILE_SUFFIX = "_index"; static final int FREQUENCY_AND_FLAGS_SIZE = 2; static final int TERMINAL_ADDRESS_TABLE_ADDRESS_SIZE = 3; + static final int UNIGRAM_TIMESTAMP_SIZE = 4; // With the English main dictionary as of October 2013, the size of bigram address table is - // is 584KB with the block size being 4. - // This is 91% of that of full address table. - static final int BIGRAM_ADDRESS_TABLE_BLOCK_SIZE = 4; - static final int BIGRAM_CONTENT_COUNT = 1; + // is 345KB with the block size being 16. + // This is 54% of that of full address table. + static final int BIGRAM_ADDRESS_TABLE_BLOCK_SIZE = 16; + static final int BIGRAM_CONTENT_COUNT = 2; static final int BIGRAM_FREQ_CONTENT_INDEX = 0; + static final int BIGRAM_TIMESTAMP_CONTENT_INDEX = 1; static final String BIGRAM_FREQ_CONTENT_ID = "_freq"; + static final String BIGRAM_TIMESTAMP_CONTENT_ID = "_timestamp"; + static final int BIGRAM_TIMESTAMP_SIZE = 4; + static final int BIGRAM_COUNTER_SIZE = 1; + static final int BIGRAM_LEVEL_SIZE = 1; static final int SHORTCUT_CONTENT_COUNT = 1; static final int SHORTCUT_CONTENT_INDEX = 0; // With the English main dictionary as of October 2013, the size of shortcut address table is - // 29KB with the block size being 64. + // 26KB with the block size being 64. // This is only 4.4% of that of full address table. static final int SHORTCUT_ADDRESS_TABLE_BLOCK_SIZE = 64; static final String SHORTCUT_CONTENT_ID = "_shortcut"; @@ -294,6 +288,8 @@ public final class FormatSpec { static final int INVALID_CHARACTER = -1; static final int MAX_PTNODES_FOR_ONE_BYTE_PTNODE_COUNT = 0x7F; // 127 + // Large PtNode array size field size is 2 bytes. + static final int LARGE_PTNODE_ARRAY_SIZE_FIELD_SIZE_FLAG = 0x8000; static final int MAX_PTNODES_IN_A_PT_NODE_ARRAY = 0x7FFF; // 32767 static final int MAX_BIGRAMS_IN_A_PTNODE = 10000; static final int MAX_SHORTCUT_LIST_SIZE_IN_A_PTNODE = 0xFFFF; @@ -319,22 +315,22 @@ public final class FormatSpec { */ public static final class FormatOptions { public final int mVersion; - public final boolean mSupportsDynamicUpdate; public final boolean mHasTerminalId; + public final boolean mHasTimestamp; + @UsedForTesting public FormatOptions(final int version) { - this(version, false); + this(version, false /* hasTimestamp */); } - @UsedForTesting - public FormatOptions(final int version, final boolean supportsDynamicUpdate) { + public FormatOptions(final int version, final boolean hasTimestamp) { mVersion = version; - if (version < FIRST_VERSION_WITH_DYNAMIC_UPDATE && supportsDynamicUpdate) { - throw new RuntimeException("Dynamic updates are only supported with versions " - + FIRST_VERSION_WITH_DYNAMIC_UPDATE + " and ulterior."); - } - mSupportsDynamicUpdate = supportsDynamicUpdate; mHasTerminalId = (version >= FIRST_VERSION_WITH_TERMINAL_ID); + mHasTimestamp = hasTimestamp; + } + + public boolean supportsDynamicUpdate() { + return mVersion >= FIRST_VERSION_WITH_DYNAMIC_UPDATE; } } @@ -342,13 +338,13 @@ public final class FormatSpec { * Class representing file header. */ public static final class FileHeader { - public final int mHeaderSize; + public final int mBodyOffset; public final DictionaryOptions mDictionaryOptions; public final FormatOptions mFormatOptions; // Note that these are corresponding definitions in native code in latinime::HeaderPolicy // and latinime::HeaderReadWriteUtils. - public static final String SUPPORTS_DYNAMIC_UPDATE_ATTRIBUTE = "SUPPORTS_DYNAMIC_UPDATE"; public static final String USES_FORGETTING_CURVE_ATTRIBUTE = "USES_FORGETTING_CURVE"; + public static final String HAS_HISTORICAL_INFO_ATTRIBUTE = "HAS_HISTORICAL_INFO"; public static final String ATTRIBUTE_VALUE_TRUE = "1"; public static final String DICTIONARY_VERSION_ATTRIBUTE = "version"; @@ -357,9 +353,18 @@ public final class FormatSpec { private static final String DICTIONARY_DESCRIPTION_ATTRIBUTE = "description"; public FileHeader(final int headerSize, final DictionaryOptions dictionaryOptions, final FormatOptions formatOptions) { - mHeaderSize = headerSize; mDictionaryOptions = dictionaryOptions; mFormatOptions = formatOptions; + mBodyOffset = formatOptions.mVersion < VERSION4 ? headerSize : 0; + if (null == getLocaleString()) { + throw new RuntimeException("Cannot create a FileHeader without a locale"); + } + if (null == getVersion()) { + throw new RuntimeException("Cannot create a FileHeader without a version"); + } + if (null == getId()) { + throw new RuntimeException("Cannot create a FileHeader without an ID"); + } } // Helper method to get the locale as a String diff --git a/java/src/com/android/inputmethod/latin/makedict/FusionDictionary.java b/java/src/com/android/inputmethod/latin/makedict/FusionDictionary.java index 3bb218bea..fdf2ae7b5 100644 --- a/java/src/com/android/inputmethod/latin/makedict/FusionDictionary.java +++ b/java/src/com/android/inputmethod/latin/makedict/FusionDictionary.java @@ -303,14 +303,9 @@ public final class FusionDictionary implements Iterable<Word> { * Options global to the dictionary. */ public static final class DictionaryOptions { - public final boolean mGermanUmlautProcessing; - public final boolean mFrenchLigatureProcessing; public final HashMap<String, String> mAttributes; - public DictionaryOptions(final HashMap<String, String> attributes, - final boolean germanUmlautProcessing, final boolean frenchLigatureProcessing) { + public DictionaryOptions(final HashMap<String, String> attributes) { mAttributes = attributes; - mGermanUmlautProcessing = germanUmlautProcessing; - mFrenchLigatureProcessing = frenchLigatureProcessing; } @Override public String toString() { // Convenience method @@ -339,14 +334,6 @@ public final class FusionDictionary implements Iterable<Word> { } s.append("\n"); } - if (mGermanUmlautProcessing) { - s.append(indent); - s.append("Needs German umlaut processing\n"); - } - if (mFrenchLigatureProcessing) { - s.append(indent); - s.append("Needs French ligature processing\n"); - } return s.toString(); } } @@ -701,138 +688,6 @@ public final class FusionDictionary implements Iterable<Word> { } /** - * Recursively count the number of nodes in a given branch of the trie. - * - * @param nodeArray the node array to count. - * @return the number of nodes in this branch. - */ - public static int countNodeArrays(final PtNodeArray nodeArray) { - int size = 1; - for (int i = nodeArray.mData.size() - 1; i >= 0; --i) { - PtNode ptNode = nodeArray.mData.get(i); - if (null != ptNode.mChildren) - size += countNodeArrays(ptNode.mChildren); - } - return size; - } - - // Recursively find out whether there are any bigrams. - // This can be pretty expensive especially if there aren't any (we return as soon - // as we find one, so it's much cheaper if there are bigrams) - private static boolean hasBigramsInternal(final PtNodeArray nodeArray) { - if (null == nodeArray) return false; - for (int i = nodeArray.mData.size() - 1; i >= 0; --i) { - PtNode ptNode = nodeArray.mData.get(i); - if (null != ptNode.mBigrams) return true; - if (hasBigramsInternal(ptNode.mChildren)) return true; - } - return false; - } - - /** - * Finds out whether there are any bigrams in this dictionary. - * - * @return true if there is any bigram, false otherwise. - */ - // TODO: this is expensive especially for large dictionaries without any bigram. - // The up side is, this is always accurate and correct and uses no memory. We should - // find a more efficient way of doing this, without compromising too much on memory - // and ease of use. - public boolean hasBigrams() { - return hasBigramsInternal(mRootNodeArray); - } - - // Historically, the tails of the words were going to be merged to save space. - // However, that would prevent the code to search for a specific address in log(n) - // time so this was abandoned. - // The code is still of interest as it does add some compression to any dictionary - // that has no need for attributes. Implementations that does not read attributes should be - // able to read a dictionary with merged tails. - // Also, the following code does support frequencies, as in, it will only merges - // tails that share the same frequency. Though it would result in the above loss of - // performance while searching by address, it is still technically possible to merge - // tails that contain attributes, but this code does not take that into account - it does - // not compare attributes and will merge terminals with different attributes regardless. - public void mergeTails() { - MakedictLog.i("Do not merge tails"); - return; - -// MakedictLog.i("Merging PtNodes. Number of PtNodes : " + countPtNodes(root)); -// MakedictLog.i("Number of PtNodes : " + countPtNodes(root)); -// -// final HashMap<String, ArrayList<PtNodeArray>> repository = -// new HashMap<String, ArrayList<PtNodeArray>>(); -// mergeTailsInner(repository, root); -// -// MakedictLog.i("Number of different pseudohashes : " + repository.size()); -// int size = 0; -// for (ArrayList<PtNodeArray> a : repository.values()) { -// size += a.size(); -// } -// MakedictLog.i("Number of nodes after merge : " + (1 + size)); -// MakedictLog.i("Recursively seen nodes : " + countNodes(root)); - } - - // The following methods are used by the deactivated mergeTails() -// private static boolean isEqual(PtNodeArray a, PtNodeArray b) { -// if (null == a && null == b) return true; -// if (null == a || null == b) return false; -// if (a.data.size() != b.data.size()) return false; -// final int size = a.data.size(); -// for (int i = size - 1; i >= 0; --i) { -// PtNode aPtNode = a.data.get(i); -// PtNode bPtNode = b.data.get(i); -// if (aPtNode.frequency != bPtNode.frequency) return false; -// if (aPtNode.alternates == null && bPtNode.alternates != null) return false; -// if (aPtNode.alternates != null && !aPtNode.equals(bPtNode.alternates)) return false; -// if (!Arrays.equals(aPtNode.chars, bPtNode.chars)) return false; -// if (!isEqual(aPtNode.children, bPtNode.children)) return false; -// } -// return true; -// } - -// static private HashMap<String, ArrayList<PtNodeArray>> mergeTailsInner( -// final HashMap<String, ArrayList<PtNodeArray>> map, final PtNodeArray nodeArray) { -// final ArrayList<PtNode> branches = nodeArray.data; -// final int nodeSize = branches.size(); -// for (int i = 0; i < nodeSize; ++i) { -// PtNode ptNode = branches.get(i); -// if (null != ptNode.children) { -// String pseudoHash = getPseudoHash(ptNode.children); -// ArrayList<PtNodeArray> similarList = map.get(pseudoHash); -// if (null == similarList) { -// similarList = new ArrayList<PtNodeArray>(); -// map.put(pseudoHash, similarList); -// } -// boolean merged = false; -// for (PtNodeArray similar : similarList) { -// if (isEqual(ptNode.children, similar)) { -// ptNode.children = similar; -// merged = true; -// break; -// } -// } -// if (!merged) { -// similarList.add(ptNode.children); -// } -// mergeTailsInner(map, ptNode.children); -// } -// } -// return map; -// } - -// private static String getPseudoHash(final PtNodeArray nodeArray) { -// StringBuilder s = new StringBuilder(); -// for (PtNode ptNode : nodeArray.data) { -// s.append(ptNode.frequency); -// for (int ch : ptNode.chars) { -// s.append(Character.toChars(ch)); -// } -// } -// return s.toString(); -// } - - /** * Iterator to walk through a dictionary. * * This is purely for convenience. diff --git a/java/src/com/android/inputmethod/latin/makedict/SparseTableContentReader.java b/java/src/com/android/inputmethod/latin/makedict/SparseTableContentReader.java new file mode 100644 index 000000000..06088b651 --- /dev/null +++ b/java/src/com/android/inputmethod/latin/makedict/SparseTableContentReader.java @@ -0,0 +1,120 @@ +/* + * Copyright (C) 2013 The Android Open Source Project + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package com.android.inputmethod.latin.makedict; + +import com.android.inputmethod.latin.makedict.BinaryDictDecoderUtils.DictBuffer; +import com.android.inputmethod.latin.makedict.DictDecoder.DictionaryBufferFactory; + +import java.io.File; +import java.io.FileNotFoundException; +import java.io.IOException; + +/** + * An auxiliary class for reading SparseTable and data written by SparseTableContentWriter. + */ +public class SparseTableContentReader { + + /** + * An interface of a function which is passed to SparseTableContentReader.read. + */ + public interface SparseTableContentReaderInterface { + /** + * Reads data. + * + * @param buffer the DictBuffer. The position of the buffer is set to the head of data. + */ + public void read(final DictBuffer buffer); + } + + protected final int mContentCount; + protected final int mBlockSize; + protected final File mBaseDir; + protected final File mLookupTableFile; + protected final File[] mAddressTableFiles; + protected final File[] mContentFiles; + protected DictBuffer mLookupTableBuffer; + protected final DictBuffer[] mAddressTableBuffers; + private final DictBuffer[] mContentBuffers; + protected final DictionaryBufferFactory mFactory; + + /** + * Sole constructor of SparseTableContentReader. + * + * @param name the name of SparseTable. + * @param blockSize the block size of the content table. + * @param baseDir the directory which contains the files of the content table. + * @param contentFilenames the file names of content files. + * @param contentIds the ids of contents. These ids are used for a suffix of a name of + * address files and content files. + * @param factory the DictionaryBufferFactory which is used for opening the files. + */ + public SparseTableContentReader(final String name, final int blockSize, final File baseDir, + final String[] contentFilenames, final String[] contentIds, + final DictionaryBufferFactory factory) { + if (contentFilenames.length != contentIds.length) { + throw new RuntimeException("The length of contentFilenames and the length of" + + " contentIds are different " + contentFilenames.length + ", " + + contentIds.length); + } + mBlockSize = blockSize; + mBaseDir = baseDir; + mFactory = factory; + mContentCount = contentFilenames.length; + mLookupTableFile = new File(baseDir, name + FormatSpec.LOOKUP_TABLE_FILE_SUFFIX); + mAddressTableFiles = new File[mContentCount]; + mContentFiles = new File[mContentCount]; + for (int i = 0; i < mContentCount; ++i) { + mAddressTableFiles[i] = new File(mBaseDir, + name + FormatSpec.CONTENT_TABLE_FILE_SUFFIX + contentIds[i]); + mContentFiles[i] = new File(mBaseDir, contentFilenames[i] + contentIds[i]); + } + mAddressTableBuffers = new DictBuffer[mContentCount]; + mContentBuffers = new DictBuffer[mContentCount]; + } + + public void openBuffers() throws FileNotFoundException, IOException { + mLookupTableBuffer = mFactory.getDictionaryBuffer(mLookupTableFile); + for (int i = 0; i < mContentCount; ++i) { + mAddressTableBuffers[i] = mFactory.getDictionaryBuffer(mAddressTableFiles[i]); + mContentBuffers[i] = mFactory.getDictionaryBuffer(mContentFiles[i]); + } + } + + protected void read(final int contentIndex, final int index, + final SparseTableContentReaderInterface reader) { + if (index < 0 || (index / mBlockSize) * SparseTable.SIZE_OF_INT_IN_BYTES + >= mLookupTableBuffer.limit()) { + return; + } + + mLookupTableBuffer.position((index / mBlockSize) * SparseTable.SIZE_OF_INT_IN_BYTES); + final int posInAddressTable = mLookupTableBuffer.readInt(); + if (posInAddressTable == SparseTable.NOT_EXIST) { + return; + } + + mAddressTableBuffers[contentIndex].position( + (posInAddressTable + index % mBlockSize) * SparseTable.SIZE_OF_INT_IN_BYTES); + final int address = mAddressTableBuffers[contentIndex].readInt(); + if (address == SparseTable.NOT_EXIST) { + return; + } + + mContentBuffers[contentIndex].position(address); + reader.read(mContentBuffers[contentIndex]); + } +}
\ No newline at end of file diff --git a/java/src/com/android/inputmethod/latin/makedict/SparseTableContentUpdater.java b/java/src/com/android/inputmethod/latin/makedict/SparseTableContentUpdater.java new file mode 100644 index 000000000..4518f21b9 --- /dev/null +++ b/java/src/com/android/inputmethod/latin/makedict/SparseTableContentUpdater.java @@ -0,0 +1,123 @@ +/* + * Copyright (C) 2013 The Android Open Source Project + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package com.android.inputmethod.latin.makedict; + +import com.android.inputmethod.latin.makedict.DictDecoder.DictionaryBufferFactory; + +import java.io.File; +import java.io.FileOutputStream; +import java.io.IOException; +import java.io.OutputStream; + +/** + * An auxiliary class for updating data associated with SparseTable. + */ +public class SparseTableContentUpdater extends SparseTableContentReader { + protected OutputStream mLookupTableOutStream; + protected OutputStream[] mAddressTableOutStreams; + protected OutputStream[] mContentOutStreams; + + public SparseTableContentUpdater(final String name, final int blockSize, + final File baseDir, final String[] contentFilenames, final String[] contentIds, + final DictionaryBufferFactory factory) { + super(name, blockSize, baseDir, contentFilenames, contentIds, factory); + mAddressTableOutStreams = new OutputStream[mContentCount]; + mContentOutStreams = new OutputStream[mContentCount]; + } + + protected void openStreamsAndBuffers() throws IOException { + openBuffers(); + mLookupTableOutStream = new FileOutputStream(mLookupTableFile, true /* append */); + for (int i = 0; i < mContentCount; ++i) { + mAddressTableOutStreams[i] = new FileOutputStream(mAddressTableFiles[i], + true /* append */); + mContentOutStreams[i] = new FileOutputStream(mContentFiles[i], true /* append */); + } + } + + /** + * Set the contentIndex-th elements of contentId-th table. + * + * @param contentId the id of the content table. + * @param contentIndex the index where to set the valie. + * @param value the value to set. + */ + protected void setContentValue(final int contentId, final int contentIndex, final int value) + throws IOException { + if ((contentIndex / mBlockSize) * SparseTable.SIZE_OF_INT_IN_BYTES + >= mLookupTableBuffer.limit()) { + // Need to extend the lookup table + final int currentSize = mLookupTableBuffer.limit() + / SparseTable.SIZE_OF_INT_IN_BYTES; + final int target = contentIndex / mBlockSize + 1; + for (int i = currentSize; i < target; ++i) { + BinaryDictEncoderUtils.writeUIntToStream(mLookupTableOutStream, + SparseTable.NOT_EXIST, SparseTable.SIZE_OF_INT_IN_BYTES); + } + // We need to reopen the byte buffer of the lookup table because a MappedByteBuffer in + // Java isn't expanded automatically when the underlying file is expanded. + reopenLookupTable(); + } + + mLookupTableBuffer.position((contentIndex / mBlockSize) * SparseTable.SIZE_OF_INT_IN_BYTES); + int posInAddressTable = mLookupTableBuffer.readInt(); + if (posInAddressTable == SparseTable.NOT_EXIST) { + // Need to extend the address table + mLookupTableBuffer.position(mLookupTableBuffer.position() + - SparseTable.SIZE_OF_INT_IN_BYTES); + posInAddressTable = mAddressTableBuffers[0].limit() / mBlockSize; + BinaryDictEncoderUtils.writeUIntToDictBuffer(mLookupTableBuffer, + posInAddressTable, SparseTable.SIZE_OF_INT_IN_BYTES); + for (int i = 0; i < mContentCount; ++i) { + for (int j = 0; j < mBlockSize; ++j) { + BinaryDictEncoderUtils.writeUIntToStream(mAddressTableOutStreams[i], + SparseTable.NOT_EXIST, SparseTable.SIZE_OF_INT_IN_BYTES); + } + } + // We need to reopen the byte buffers of the address tables because a MappedByteBuffer + // in Java isn't expanded automatically when the underlying file is expanded. + reopenAddressTables(); + } + posInAddressTable += (contentIndex % mBlockSize) * SparseTable.SIZE_OF_INT_IN_BYTES; + + mAddressTableBuffers[contentId].position(posInAddressTable); + BinaryDictEncoderUtils.writeUIntToDictBuffer(mAddressTableBuffers[contentId], + value, SparseTable.SIZE_OF_INT_IN_BYTES); + } + + private void reopenLookupTable() throws IOException { + mLookupTableOutStream.flush(); + mLookupTableBuffer = mFactory.getDictionaryBuffer(mLookupTableFile); + } + + private void reopenAddressTables() throws IOException { + for (int i = 0; i < mContentCount; ++i) { + mAddressTableOutStreams[i].flush(); + mAddressTableBuffers[i] = mFactory.getDictionaryBuffer(mAddressTableFiles[i]); + } + } + + protected void close() throws IOException { + mLookupTableOutStream.close(); + for (final OutputStream stream : mAddressTableOutStreams) { + stream.close(); + } + for (final OutputStream stream : mContentOutStreams) { + stream.close(); + } + } +} diff --git a/java/src/com/android/inputmethod/latin/makedict/SparseTableContentWriter.java b/java/src/com/android/inputmethod/latin/makedict/SparseTableContentWriter.java new file mode 100644 index 000000000..49f0fd624 --- /dev/null +++ b/java/src/com/android/inputmethod/latin/makedict/SparseTableContentWriter.java @@ -0,0 +1,93 @@ +/* + * Copyright (C) 2013 The Android Open Source Project + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License. + */ + +package com.android.inputmethod.latin.makedict; + +import java.io.File; +import java.io.FileNotFoundException; +import java.io.FileOutputStream; +import java.io.IOException; +import java.io.OutputStream; + +/** + * An auxiliary class for writing data associated with SparseTable to files. + */ +public class SparseTableContentWriter { + public interface SparseTableContentWriterInterface { + public void write(final OutputStream outStream) throws IOException; + } + + private final int mContentCount; + private final SparseTable mSparseTable; + private final File mLookupTableFile; + protected final File mBaseDir; + private final File[] mAddressTableFiles; + private final File[] mContentFiles; + protected final OutputStream[] mContentOutStreams; + + /** + * Sole constructor of SparseTableContentWriter. + * + * @param name the name of SparseTable. + * @param initialCapacity the initial capacity of SparseTable. + * @param blockSize the block size of the content table. + * @param baseDir the directory which contains the files of the content table. + * @param contentFilenames the file names of content files. + * @param contentIds the ids of contents. These ids are used for a suffix of a name of address + * files and content files. + */ + public SparseTableContentWriter(final String name, final int initialCapacity, + final int blockSize, final File baseDir, final String[] contentFilenames, + final String[] contentIds) { + if (contentFilenames.length != contentIds.length) { + throw new RuntimeException("The length of contentFilenames and the length of" + + " contentIds are different " + contentFilenames.length + ", " + + contentIds.length); + } + mContentCount = contentFilenames.length; + mSparseTable = new SparseTable(initialCapacity, blockSize, mContentCount); + mLookupTableFile = new File(baseDir, name + FormatSpec.LOOKUP_TABLE_FILE_SUFFIX); + mAddressTableFiles = new File[mContentCount]; + mContentFiles = new File[mContentCount]; + mBaseDir = baseDir; + for (int i = 0; i < mContentCount; ++i) { + mAddressTableFiles[i] = new File(mBaseDir, + name + FormatSpec.CONTENT_TABLE_FILE_SUFFIX + contentIds[i]); + mContentFiles[i] = new File(mBaseDir, contentFilenames[i] + contentIds[i]); + } + mContentOutStreams = new OutputStream[mContentCount]; + } + + public void openStreams() throws FileNotFoundException { + for (int i = 0; i < mContentCount; ++i) { + mContentOutStreams[i] = new FileOutputStream(mContentFiles[i]); + } + } + + protected void write(final int contentIndex, final int index, + final SparseTableContentWriterInterface writer) throws IOException { + mSparseTable.set(contentIndex, index, (int) mContentFiles[contentIndex].length()); + writer.write(mContentOutStreams[contentIndex]); + mContentOutStreams[contentIndex].flush(); + } + + public void closeStreams() throws IOException { + mSparseTable.writeToFiles(mLookupTableFile, mAddressTableFiles); + for (int i = 0; i < mContentCount; ++i) { + mContentOutStreams[i].close(); + } + } +}
\ No newline at end of file diff --git a/java/src/com/android/inputmethod/latin/makedict/Ver3DictEncoder.java b/java/src/com/android/inputmethod/latin/makedict/Ver3DictEncoder.java index d9e19899c..92eb861d6 100644 --- a/java/src/com/android/inputmethod/latin/makedict/Ver3DictEncoder.java +++ b/java/src/com/android/inputmethod/latin/makedict/Ver3DictEncoder.java @@ -129,7 +129,9 @@ public class Ver3DictEncoder implements DictEncoder { if (countSize != 1 && countSize != 2) { throw new RuntimeException("Strange size from getGroupCountSize : " + countSize); } - mPosition = BinaryDictEncoderUtils.writeUIntToBuffer(mBuffer, mPosition, ptNodeCount, + final int encodedPtNodeCount = (countSize == 2) ? + (ptNodeCount | FormatSpec.LARGE_PTNODE_ARRAY_SIZE_FIELD_SIZE_FLAG) : ptNodeCount; + mPosition = BinaryDictEncoderUtils.writeUIntToBuffer(mBuffer, mPosition, encodedPtNodeCount, countSize); } @@ -167,7 +169,7 @@ public class Ver3DictEncoder implements DictEncoder { private void writeChildrenPosition(final PtNode ptNode, final FormatOptions formatOptions) { final int childrenPos = BinaryDictEncoderUtils.getChildrenPosition(ptNode, formatOptions); - if (formatOptions.mSupportsDynamicUpdate) { + if (formatOptions.supportsDynamicUpdate()) { mPosition += BinaryDictEncoderUtils.writeSignedChildrenPosition(mBuffer, mPosition, childrenPos); } else { diff --git a/java/src/com/android/inputmethod/latin/makedict/Ver4DictDecoder.java b/java/src/com/android/inputmethod/latin/makedict/Ver4DictDecoder.java index 53729075f..3be62f066 100644 --- a/java/src/com/android/inputmethod/latin/makedict/Ver4DictDecoder.java +++ b/java/src/com/android/inputmethod/latin/makedict/Ver4DictDecoder.java @@ -40,26 +40,52 @@ import java.util.Arrays; public class Ver4DictDecoder extends AbstractDictDecoder { private static final String TAG = Ver4DictDecoder.class.getSimpleName(); - private static final int FILETYPE_TRIE = 1; - private static final int FILETYPE_FREQUENCY = 2; - private static final int FILETYPE_TERMINAL_ADDRESS_TABLE = 3; - private static final int FILETYPE_BIGRAM_FREQ = 4; - private static final int FILETYPE_SHORTCUT = 5; - - private final File mDictDirectory; - private final DictionaryBufferFactory mBufferFactory; + protected static final int FILETYPE_TRIE = 1; + protected static final int FILETYPE_FREQUENCY = 2; + protected static final int FILETYPE_TERMINAL_ADDRESS_TABLE = 3; + protected static final int FILETYPE_BIGRAM_FREQ = 4; + protected static final int FILETYPE_SHORTCUT = 5; + protected static final int FILETYPE_HEADER = 6; + + protected final File mDictDirectory; + protected final DictionaryBufferFactory mBufferFactory; protected DictBuffer mDictBuffer; - private DictBuffer mFrequencyBuffer; - private DictBuffer mTerminalAddressTableBuffer; - private DictBuffer mBigramBuffer; - private DictBuffer mShortcutBuffer; - private SparseTable mBigramAddressTable; - private SparseTable mShortcutAddressTable; + protected DictBuffer mHeaderBuffer; + protected DictBuffer mFrequencyBuffer; + protected DictBuffer mTerminalAddressTableBuffer; + private BigramContentReader mBigramReader; + private ShortcutContentReader mShortcutReader; + + /** + * Raw PtNode info straight out of a trie file in version 4 dictionary. + */ + protected static final class Ver4PtNodeInfo { + public final int mFlags; + public final int[] mCharacters; + public final int mTerminalId; + public final int mChildrenPos; + public final int mParentPos; + public final int mNodeSize; + public int mStartIndexOfCharacters; + public int mEndIndexOfCharacters; // exclusive + + public Ver4PtNodeInfo(final int flags, final int[] characters, final int terminalId, + final int childrenPos, final int parentPos, final int nodeSize) { + mFlags = flags; + mCharacters = characters; + mTerminalId = terminalId; + mChildrenPos = childrenPos; + mParentPos = parentPos; + mNodeSize = nodeSize; + mStartIndexOfCharacters = 0; + mEndIndexOfCharacters = characters.length; + } + } @UsedForTesting /* package */ Ver4DictDecoder(final File dictDirectory, final int factoryFlag) { mDictDirectory = dictDirectory; - mDictBuffer = mFrequencyBuffer = null; + mDictBuffer = mHeaderBuffer = mFrequencyBuffer = null; if ((factoryFlag & MASK_DICTBUFFER) == USE_READONLY_BYTEBUFFER) { mBufferFactory = new DictionaryBufferFromReadOnlyByteBufferFactory(); @@ -76,13 +102,16 @@ public class Ver4DictDecoder extends AbstractDictDecoder { /* package */ Ver4DictDecoder(final File dictDirectory, final DictionaryBufferFactory factory) { mDictDirectory = dictDirectory; mBufferFactory = factory; - mDictBuffer = mFrequencyBuffer = null; + mDictBuffer = mHeaderBuffer = mFrequencyBuffer = null; } - private File getFile(final int fileType) { + protected File getFile(final int fileType) throws UnsupportedFormatException { if (fileType == FILETYPE_TRIE) { return new File(mDictDirectory, mDictDirectory.getName() + FormatSpec.TRIE_FILE_EXTENSION); + } else if (fileType == FILETYPE_HEADER) { + return new File(mDictDirectory, + mDictDirectory.getName() + FormatSpec.HEADER_FILE_EXTENSION); } else if (fileType == FILETYPE_FREQUENCY) { return new File(mDictDirectory, mDictDirectory.getName() + FormatSpec.FREQ_FILE_EXTENSION); @@ -98,20 +127,27 @@ public class Ver4DictDecoder extends AbstractDictDecoder { mDictDirectory.getName() + FormatSpec.SHORTCUT_FILE_EXTENSION + FormatSpec.SHORTCUT_CONTENT_ID); } else { - throw new RuntimeException("Unsupported kind of file : " + fileType); + throw new UnsupportedFormatException("Unsupported kind of file : " + fileType); } } @Override - public void openDictBuffer() throws FileNotFoundException, IOException { + public void openDictBuffer() throws FileNotFoundException, IOException, + UnsupportedFormatException { + if (!mDictDirectory.isDirectory()) { + throw new UnsupportedFormatException("Format 4 dictionary needs a directory"); + } + mHeaderBuffer = mBufferFactory.getDictionaryBuffer(getFile(FILETYPE_HEADER)); mDictBuffer = mBufferFactory.getDictionaryBuffer(getFile(FILETYPE_TRIE)); mFrequencyBuffer = mBufferFactory.getDictionaryBuffer(getFile(FILETYPE_FREQUENCY)); mTerminalAddressTableBuffer = mBufferFactory.getDictionaryBuffer( getFile(FILETYPE_TERMINAL_ADDRESS_TABLE)); - mBigramBuffer = mBufferFactory.getDictionaryBuffer(getFile(FILETYPE_BIGRAM_FREQ)); - loadBigramAddressSparseTable(); - mShortcutBuffer = mBufferFactory.getDictionaryBuffer(getFile(FILETYPE_SHORTCUT)); - loadShortcutAddressSparseTable(); + mBigramReader = new BigramContentReader(mDictDirectory.getName(), + mDictDirectory, mBufferFactory, false); + mBigramReader.openBuffers(); + mShortcutReader = new ShortcutContentReader(mDictDirectory.getName(), mDictDirectory, + mBufferFactory); + mShortcutReader.openBuffers(); } @Override @@ -119,42 +155,134 @@ public class Ver4DictDecoder extends AbstractDictDecoder { return mDictBuffer != null; } + @UsedForTesting + /* package */ DictBuffer getHeaderBuffer() { + return mHeaderBuffer; + } + + @UsedForTesting /* package */ DictBuffer getDictBuffer() { return mDictBuffer; } @Override public FileHeader readHeader() throws IOException, UnsupportedFormatException { - if (mDictBuffer == null) { + if (mHeaderBuffer == null) { openDictBuffer(); } - final FileHeader header = super.readHeader(mDictBuffer); + mHeaderBuffer.position(0); + final FileHeader header = super.readHeader(mHeaderBuffer); final int version = header.mFormatOptions.mVersion; - if (version != 4) { + if (version != FormatSpec.VERSION4) { throw new UnsupportedFormatException("File header has a wrong version : " + version); } return header; } - private void loadBigramAddressSparseTable() throws IOException { - final File lookupIndexFile = new File(mDictDirectory, mDictDirectory.getName() - + FormatSpec.BIGRAM_FILE_EXTENSION + FormatSpec.LOOKUP_TABLE_FILE_SUFFIX); - final File freqsFile = new File(mDictDirectory, mDictDirectory.getName() - + FormatSpec.BIGRAM_FILE_EXTENSION + FormatSpec.CONTENT_TABLE_FILE_SUFFIX - + FormatSpec.BIGRAM_FREQ_CONTENT_ID); - mBigramAddressTable = SparseTable.readFromFiles(lookupIndexFile, new File[] { freqsFile }, - FormatSpec.BIGRAM_ADDRESS_TABLE_BLOCK_SIZE); + /** + * An auxiliary class for reading bigrams. + */ + protected static class BigramContentReader extends SparseTableContentReader { + public BigramContentReader(final String name, final File baseDir, + final DictionaryBufferFactory factory, final boolean hasTimestamp) { + super(name + FormatSpec.BIGRAM_FILE_EXTENSION, + FormatSpec.BIGRAM_ADDRESS_TABLE_BLOCK_SIZE, baseDir, + getContentFilenames(name, hasTimestamp), getContentIds(hasTimestamp), factory); + } + + // TODO: Consolidate this method and BigramContentWriter.getContentFilenames. + protected static String[] getContentFilenames(final String name, + final boolean hasTimestamp) { + final String[] contentFilenames; + if (hasTimestamp) { + contentFilenames = new String[] { name + FormatSpec.BIGRAM_FILE_EXTENSION, + name + FormatSpec.BIGRAM_FILE_EXTENSION }; + } else { + contentFilenames = new String[] { name + FormatSpec.BIGRAM_FILE_EXTENSION }; + } + return contentFilenames; + } + + // TODO: Consolidate this method and BigramContentWriter.getContentIds. + protected static String[] getContentIds(final boolean hasTimestamp) { + final String[] contentIds; + if (hasTimestamp) { + contentIds = new String[] { FormatSpec.BIGRAM_FREQ_CONTENT_ID, + FormatSpec.BIGRAM_TIMESTAMP_CONTENT_ID }; + } else { + contentIds = new String[] { FormatSpec.BIGRAM_FREQ_CONTENT_ID }; + } + return contentIds; + } + + public ArrayList<PendingAttribute> readTargetsAndFrequencies(final int terminalId, + final DictBuffer terminalAddressTableBuffer) { + final ArrayList<PendingAttribute> bigrams = CollectionUtils.newArrayList(); + read(FormatSpec.BIGRAM_FREQ_CONTENT_INDEX, terminalId, + new SparseTableContentReaderInterface() { + @Override + public void read(final DictBuffer buffer) { + while (bigrams.size() < FormatSpec.MAX_BIGRAMS_IN_A_PTNODE) { + // If bigrams.size() reaches FormatSpec.MAX_BIGRAMS_IN_A_PTNODE, + // remaining bigram entries are ignored. + final int bigramFlags = buffer.readUnsignedByte(); + final int targetTerminalId = buffer.readUnsignedInt24(); + terminalAddressTableBuffer.position(targetTerminalId + * FormatSpec.TERMINAL_ADDRESS_TABLE_ADDRESS_SIZE); + final int targetAddress = + terminalAddressTableBuffer.readUnsignedInt24(); + bigrams.add(new PendingAttribute(bigramFlags + & FormatSpec.FLAG_BIGRAM_SHORTCUT_ATTR_FREQUENCY, + targetAddress)); + if (0 == (bigramFlags + & FormatSpec.FLAG_BIGRAM_SHORTCUT_ATTR_HAS_NEXT)) { + break; + } + } + if (bigrams.size() >= FormatSpec.MAX_BIGRAMS_IN_A_PTNODE) { + throw new RuntimeException("Too many bigrams in a PtNode (" + + bigrams.size() + " but max is " + + FormatSpec.MAX_BIGRAMS_IN_A_PTNODE + ")"); + } + } + }); + if (bigrams.isEmpty()) return null; + return bigrams; + } } - // TODO: Let's have something like SparseTableContentsReader in this class. - private void loadShortcutAddressSparseTable() throws IOException { - final File lookupIndexFile = new File(mDictDirectory, mDictDirectory.getName() - + FormatSpec.SHORTCUT_FILE_EXTENSION + FormatSpec.LOOKUP_TABLE_FILE_SUFFIX); - final File contentFile = new File(mDictDirectory, mDictDirectory.getName() - + FormatSpec.SHORTCUT_FILE_EXTENSION + FormatSpec.CONTENT_TABLE_FILE_SUFFIX - + FormatSpec.SHORTCUT_CONTENT_ID); - mShortcutAddressTable = SparseTable.readFromFiles(lookupIndexFile, - new File[] { contentFile }, FormatSpec.SHORTCUT_ADDRESS_TABLE_BLOCK_SIZE); + /** + * An auxiliary class for reading shortcuts. + */ + protected static class ShortcutContentReader extends SparseTableContentReader { + public ShortcutContentReader(final String name, final File baseDir, + final DictionaryBufferFactory factory) { + super(name + FormatSpec.SHORTCUT_FILE_EXTENSION, + FormatSpec.SHORTCUT_ADDRESS_TABLE_BLOCK_SIZE, baseDir, + new String[] { name + FormatSpec.SHORTCUT_FILE_EXTENSION }, + new String[] { FormatSpec.SHORTCUT_CONTENT_ID }, factory); + } + + public ArrayList<WeightedString> readShortcuts(final int terminalId) { + final ArrayList<WeightedString> shortcuts = CollectionUtils.newArrayList(); + read(FormatSpec.SHORTCUT_CONTENT_INDEX, terminalId, + new SparseTableContentReaderInterface() { + @Override + public void read(final DictBuffer buffer) { + while (true) { + final int flags = buffer.readUnsignedByte(); + final String word = CharEncoding.readString(buffer); + shortcuts.add(new WeightedString(word, + flags & FormatSpec.FLAG_BIGRAM_SHORTCUT_ATTR_FREQUENCY)); + if (0 == (flags & FormatSpec.FLAG_BIGRAM_SHORTCUT_ATTR_HAS_NEXT)) { + break; + } + } + } + }); + if (shortcuts.isEmpty()) return null; + return shortcuts; + } } protected static class PtNodeReader extends AbstractDictDecoder.PtNodeReader { @@ -168,102 +296,82 @@ public class Ver4DictDecoder extends AbstractDictDecoder { } } - private ArrayList<WeightedString> readShortcuts(final int terminalId) { - if (mShortcutAddressTable.get(0, terminalId) == SparseTable.NOT_EXIST) return null; - - final ArrayList<WeightedString> ret = CollectionUtils.newArrayList(); - final int posOfShortcuts = mShortcutAddressTable.get(FormatSpec.SHORTCUT_CONTENT_INDEX, - terminalId); - mShortcutBuffer.position(posOfShortcuts); - while (true) { - final int flags = mShortcutBuffer.readUnsignedByte(); - final String word = CharEncoding.readString(mShortcutBuffer); - ret.add(new WeightedString(word, - flags & FormatSpec.FLAG_BIGRAM_SHORTCUT_ATTR_FREQUENCY)); - if (0 == (flags & FormatSpec.FLAG_BIGRAM_SHORTCUT_ATTR_HAS_NEXT)) break; - } - return ret; - } + private final int[] mCharacterBufferForReadingVer4PtNodeInfo + = new int[FormatSpec.MAX_WORD_LENGTH]; + /** + * Reads PtNode from ptNodePos in the trie file and returns Ver4PtNodeInfo. + * + * @param ptNodePos the position of PtNode. + * @param options the format options. + * @return Ver4PtNodeInfo. + */ // TODO: Make this buffer thread safe. // TODO: Support words longer than FormatSpec.MAX_WORD_LENGTH. - private final int[] mCharacterBuffer = new int[FormatSpec.MAX_WORD_LENGTH]; - @Override - public PtNodeInfo readPtNode(int ptNodePos, FormatOptions options) { - int addressPointer = ptNodePos; + protected Ver4PtNodeInfo readVer4PtNodeInfo(final int ptNodePos, final FormatOptions options) { + int readingPos = ptNodePos; final int flags = PtNodeReader.readPtNodeOptionFlags(mDictBuffer); - addressPointer += FormatSpec.PTNODE_FLAGS_SIZE; + readingPos += FormatSpec.PTNODE_FLAGS_SIZE; - final int parentAddress = PtNodeReader.readParentAddress(mDictBuffer, options); + final int parentPos = PtNodeReader.readParentAddress(mDictBuffer, options); if (BinaryDictIOUtils.supportsDynamicUpdate(options)) { - addressPointer += FormatSpec.PARENT_ADDRESS_SIZE; + readingPos += FormatSpec.PARENT_ADDRESS_SIZE; } final int characters[]; if (0 != (flags & FormatSpec.FLAG_HAS_MULTIPLE_CHARS)) { int index = 0; int character = CharEncoding.readChar(mDictBuffer); - addressPointer += CharEncoding.getCharSize(character); + readingPos += CharEncoding.getCharSize(character); while (FormatSpec.INVALID_CHARACTER != character && index < FormatSpec.MAX_WORD_LENGTH) { - mCharacterBuffer[index++] = character; + mCharacterBufferForReadingVer4PtNodeInfo[index++] = character; character = CharEncoding.readChar(mDictBuffer); - addressPointer += CharEncoding.getCharSize(character); + readingPos += CharEncoding.getCharSize(character); } - characters = Arrays.copyOfRange(mCharacterBuffer, 0, index); + characters = Arrays.copyOfRange(mCharacterBufferForReadingVer4PtNodeInfo, 0, index); } else { final int character = CharEncoding.readChar(mDictBuffer); - addressPointer += CharEncoding.getCharSize(character); + readingPos += CharEncoding.getCharSize(character); characters = new int[] { character }; } final int terminalId; if (0 != (FormatSpec.FLAG_IS_TERMINAL & flags)) { terminalId = PtNodeReader.readTerminalId(mDictBuffer); - addressPointer += FormatSpec.PTNODE_TERMINAL_ID_SIZE; + readingPos += FormatSpec.PTNODE_TERMINAL_ID_SIZE; } else { terminalId = PtNode.NOT_A_TERMINAL; } + int childrenPos = PtNodeReader.readChildrenAddress(mDictBuffer, flags, options); + if (childrenPos != FormatSpec.NO_CHILDREN_ADDRESS) { + childrenPos += readingPos; + } + readingPos += BinaryDictIOUtils.getChildrenAddressSize(flags, options); + + return new Ver4PtNodeInfo(flags, characters, terminalId, childrenPos, parentPos, + readingPos - ptNodePos); + } + + @Override + public PtNodeInfo readPtNode(int ptNodePos, FormatOptions options) { + final Ver4PtNodeInfo nodeInfo = readVer4PtNodeInfo(ptNodePos, options); + final int frequency; - if (0 != (FormatSpec.FLAG_IS_TERMINAL & flags)) { - frequency = PtNodeReader.readFrequency(mFrequencyBuffer, terminalId); + if (0 != (FormatSpec.FLAG_IS_TERMINAL & nodeInfo.mFlags)) { + frequency = PtNodeReader.readFrequency(mFrequencyBuffer, nodeInfo.mTerminalId); } else { frequency = PtNode.NOT_A_TERMINAL; } - int childrenAddress = PtNodeReader.readChildrenAddress(mDictBuffer, flags, options); - if (childrenAddress != FormatSpec.NO_CHILDREN_ADDRESS) { - childrenAddress += addressPointer; - } - addressPointer += BinaryDictIOUtils.getChildrenAddressSize(flags, options); - final ArrayList<WeightedString> shortcutTargets = readShortcuts(terminalId); - - final ArrayList<PendingAttribute> bigrams; - if (0 != (flags & FormatSpec.FLAG_HAS_BIGRAMS)) { - bigrams = new ArrayList<PendingAttribute>(); - final int posOfBigrams = mBigramAddressTable.get(0 /* contentTableIndex */, terminalId); - mBigramBuffer.position(posOfBigrams); - while (bigrams.size() < FormatSpec.MAX_BIGRAMS_IN_A_PTNODE) { - // If bigrams.size() reaches FormatSpec.MAX_BIGRAMS_IN_A_PTNODE, - // remaining bigram entries are ignored. - final int bigramFlags = mBigramBuffer.readUnsignedByte(); - final int targetTerminalId = mBigramBuffer.readUnsignedInt24(); - mTerminalAddressTableBuffer.position( - targetTerminalId * FormatSpec.TERMINAL_ADDRESS_TABLE_ADDRESS_SIZE); - final int targetAddress = mTerminalAddressTableBuffer.readUnsignedInt24(); - bigrams.add(new PendingAttribute( - bigramFlags & FormatSpec.FLAG_BIGRAM_SHORTCUT_ATTR_FREQUENCY, - targetAddress)); - if (0 == (bigramFlags & FormatSpec.FLAG_BIGRAM_SHORTCUT_ATTR_HAS_NEXT)) break; - } - if (bigrams.size() >= FormatSpec.MAX_BIGRAMS_IN_A_PTNODE) { - throw new RuntimeException("Too many bigrams in a PtNode (" + bigrams.size() - + " but max is " + FormatSpec.MAX_BIGRAMS_IN_A_PTNODE + ")"); - } - } else { - bigrams = null; - } - return new PtNodeInfo(ptNodePos, addressPointer, flags, characters, frequency, - parentAddress, childrenAddress, shortcutTargets, bigrams); + + final ArrayList<WeightedString> shortcutTargets = mShortcutReader.readShortcuts( + nodeInfo.mTerminalId); + final ArrayList<PendingAttribute> bigrams = mBigramReader.readTargetsAndFrequencies( + nodeInfo.mTerminalId, mTerminalAddressTableBuffer); + + return new PtNodeInfo(ptNodePos, ptNodePos + nodeInfo.mNodeSize, nodeInfo.mFlags, + nodeInfo.mCharacters, frequency, nodeInfo.mParentPos, nodeInfo.mChildrenPos, + shortcutTargets, bigrams); } private void deleteDictFiles() { @@ -314,10 +422,14 @@ public class Ver4DictDecoder extends AbstractDictDecoder { @Override public boolean readAndFollowForwardLink() { - final int nextAddress = mDictBuffer.readUnsignedInt24(); - if (nextAddress >= 0 && nextAddress < mDictBuffer.limit()) { - mDictBuffer.position(nextAddress); - return true; + final int forwardLinkPos = mDictBuffer.position(); + int nextRelativePos = BinaryDictDecoderUtils.readSInt24(mDictBuffer); + if (nextRelativePos != FormatSpec.NO_FORWARD_LINK_ADDRESS) { + final int nextPos = forwardLinkPos + nextRelativePos; + if (nextPos >= 0 && nextPos < mDictBuffer.limit()) { + mDictBuffer.position(nextPos); + return true; + } } return false; } diff --git a/java/src/com/android/inputmethod/latin/makedict/Ver4DictEncoder.java b/java/src/com/android/inputmethod/latin/makedict/Ver4DictEncoder.java index f9dcacf77..8b80ebe63 100644 --- a/java/src/com/android/inputmethod/latin/makedict/Ver4DictEncoder.java +++ b/java/src/com/android/inputmethod/latin/makedict/Ver4DictEncoder.java @@ -25,6 +25,8 @@ import com.android.inputmethod.latin.makedict.FusionDictionary.DictionaryOptions import com.android.inputmethod.latin.makedict.FusionDictionary.PtNode; import com.android.inputmethod.latin.makedict.FusionDictionary.PtNodeArray; import com.android.inputmethod.latin.makedict.FusionDictionary.WeightedString; +import com.android.inputmethod.latin.utils.CollectionUtils; +import com.android.inputmethod.latin.utils.FileUtils; import java.io.File; import java.io.FileNotFoundException; @@ -32,6 +34,8 @@ import java.io.FileOutputStream; import java.io.IOException; import java.io.OutputStream; import java.util.ArrayList; +import java.util.Collections; +import java.util.Comparator; import java.util.Iterator; /** @@ -42,9 +46,10 @@ public class Ver4DictEncoder implements DictEncoder { private final File mDictPlacedDir; private byte[] mTrieBuf; private int mTriePos; - private int mHeaderSize; private OutputStream mTrieOutStream; + private OutputStream mHeaderOutStream; private OutputStream mFreqOutStream; + private OutputStream mUnigramTimestampOutStream; private OutputStream mTerminalAddressTableOutStream; private File mDictDir; private String mBaseFilename; @@ -56,73 +61,41 @@ public class Ver4DictEncoder implements DictEncoder { mDictPlacedDir = dictPlacedDir; } - private interface SparseTableContentWriterInterface { - public void write(final OutputStream outStream) throws IOException; - } + private static class BigramContentWriter extends SparseTableContentWriter { + private final boolean mWriteTimestamp; - private static class SparseTableContentWriter { - private final int mContentCount; - private final SparseTable mSparseTable; - private final File mLookupTableFile; - protected final File mBaseDir; - private final File[] mAddressTableFiles; - private final File[] mContentFiles; - protected final OutputStream[] mContentOutStreams; - - public SparseTableContentWriter(final String name, final int contentCount, - final int initialCapacity, final int blockSize, final File baseDir, - final String[] contentFilenames, final String[] contentIds) { - if (contentFilenames.length != contentIds.length) { - throw new RuntimeException("The length of contentFilenames and the length of" - + " contentIds are different " + contentFilenames.length + ", " - + contentIds.length); - } - mContentCount = contentCount; - mSparseTable = new SparseTable(initialCapacity, blockSize, contentCount); - mLookupTableFile = new File(baseDir, name + FormatSpec.LOOKUP_TABLE_FILE_SUFFIX); - mAddressTableFiles = new File[mContentCount]; - mContentFiles = new File[mContentCount]; - mBaseDir = baseDir; - for (int i = 0; i < mContentCount; ++i) { - mAddressTableFiles[i] = new File(mBaseDir, - name + FormatSpec.CONTENT_TABLE_FILE_SUFFIX + contentIds[i]); - mContentFiles[i] = new File(mBaseDir, contentFilenames[i] + contentIds[i]); - } - mContentOutStreams = new OutputStream[mContentCount]; + public BigramContentWriter(final String name, final int initialCapacity, + final File baseDir, final boolean writeTimestamp) { + super(name + FormatSpec.BIGRAM_FILE_EXTENSION, initialCapacity, + FormatSpec.BIGRAM_ADDRESS_TABLE_BLOCK_SIZE, baseDir, + getContentFilenames(name, writeTimestamp), getContentIds(writeTimestamp)); + mWriteTimestamp = writeTimestamp; } - public void openStreams() throws FileNotFoundException { - for (int i = 0; i < mContentCount; ++i) { - mContentOutStreams[i] = new FileOutputStream(mContentFiles[i]); + private static String[] getContentFilenames(final String name, + final boolean writeTimestamp) { + final String[] contentFilenames; + if (writeTimestamp) { + contentFilenames = new String[] { name + FormatSpec.BIGRAM_FILE_EXTENSION, + name + FormatSpec.BIGRAM_FILE_EXTENSION }; + } else { + contentFilenames = new String[] { name + FormatSpec.BIGRAM_FILE_EXTENSION }; } + return contentFilenames; } - protected void write(final int contentIndex, final int index, - final SparseTableContentWriterInterface writer) throws IOException { - mSparseTable.set(contentIndex, index, (int) mContentFiles[contentIndex].length()); - writer.write(mContentOutStreams[contentIndex]); - mContentOutStreams[contentIndex].flush(); - } - - public void closeStreams() throws IOException { - mSparseTable.writeToFiles(mLookupTableFile, mAddressTableFiles); - for (int i = 0; i < mContentCount; ++i) { - mContentOutStreams[i].close(); + private static String[] getContentIds(final boolean writeTimestamp) { + final String[] contentIds; + if (writeTimestamp) { + contentIds = new String[] { FormatSpec.BIGRAM_FREQ_CONTENT_ID, + FormatSpec.BIGRAM_TIMESTAMP_CONTENT_ID }; + } else { + contentIds = new String[] { FormatSpec.BIGRAM_FREQ_CONTENT_ID }; } - } - } - - private static class BigramContentWriter extends SparseTableContentWriter { - - public BigramContentWriter(final String name, final int initialCapacity, - final File baseDir) { - super(name + FormatSpec.BIGRAM_FILE_EXTENSION, FormatSpec.BIGRAM_CONTENT_COUNT, - initialCapacity, FormatSpec.BIGRAM_ADDRESS_TABLE_BLOCK_SIZE, baseDir, - new String[] { name + FormatSpec.BIGRAM_FILE_EXTENSION }, - new String[] { FormatSpec.BIGRAM_FREQ_CONTENT_ID }); + return contentIds; } - public void writeBigramsForOneWord(final int terminalId, + public void writeBigramsForOneWord(final int terminalId, final int bigramCount, final Iterator<WeightedString> bigramIterator, final FusionDictionary dict) throws IOException { write(FormatSpec.BIGRAM_FREQ_CONTENT_INDEX, terminalId, @@ -130,8 +103,16 @@ public class Ver4DictEncoder implements DictEncoder { @Override public void write(final OutputStream outStream) throws IOException { writeBigramsForOneWordInternal(outStream, bigramIterator, dict); - } - }); + }}); + if (mWriteTimestamp) { + write(FormatSpec.BIGRAM_TIMESTAMP_CONTENT_INDEX, terminalId, + new SparseTableContentWriterInterface() { + @Override + public void write(final OutputStream outStream) throws IOException { + initBigramTimestampsCountersAndLevelsForOneWordInternal(outStream, + bigramCount); + }}); + } } private void writeBigramsForOneWordInternal(final OutputStream outStream, @@ -151,13 +132,26 @@ public class Ver4DictEncoder implements DictEncoder { FormatSpec.PTNODE_ATTRIBUTE_MAX_ADDRESS_SIZE); } } + + private void initBigramTimestampsCountersAndLevelsForOneWordInternal( + final OutputStream outStream, final int bigramCount) throws IOException { + for (int i = 0; i < bigramCount; ++i) { + // TODO: Figure out what initial values should be. + BinaryDictEncoderUtils.writeUIntToStream(outStream, 0 /* value */, + FormatSpec.BIGRAM_TIMESTAMP_SIZE); + BinaryDictEncoderUtils.writeUIntToStream(outStream, 0 /* value */, + FormatSpec.BIGRAM_COUNTER_SIZE); + BinaryDictEncoderUtils.writeUIntToStream(outStream, 0 /* value */, + FormatSpec.BIGRAM_LEVEL_SIZE); + } + } } private static class ShortcutContentWriter extends SparseTableContentWriter { public ShortcutContentWriter(final String name, final int initialCapacity, final File baseDir) { - super(name + FormatSpec.SHORTCUT_FILE_EXTENSION, FormatSpec.SHORTCUT_CONTENT_COUNT, - initialCapacity, FormatSpec.SHORTCUT_ADDRESS_TABLE_BLOCK_SIZE, baseDir, + super(name + FormatSpec.SHORTCUT_FILE_EXTENSION, initialCapacity, + FormatSpec.SHORTCUT_ADDRESS_TABLE_BLOCK_SIZE, baseDir, new String[] { name + FormatSpec.SHORTCUT_FILE_EXTENSION }, new String[] { FormatSpec.SHORTCUT_CONTENT_ID }); } @@ -192,19 +186,26 @@ public class Ver4DictEncoder implements DictEncoder { mBaseFilename = header.getId() + "." + header.getVersion(); mDictDir = new File(mDictPlacedDir, mBaseFilename); final File trieFile = new File(mDictDir, mBaseFilename + FormatSpec.TRIE_FILE_EXTENSION); + final File headerFile = new File(mDictDir, + mBaseFilename + FormatSpec.HEADER_FILE_EXTENSION); final File freqFile = new File(mDictDir, mBaseFilename + FormatSpec.FREQ_FILE_EXTENSION); + final File timestampFile = new File(mDictDir, + mBaseFilename + FormatSpec.UNIGRAM_TIMESTAMP_FILE_EXTENSION); final File terminalAddressTableFile = new File(mDictDir, mBaseFilename + FormatSpec.TERMINAL_ADDRESS_TABLE_FILE_EXTENSION); if (!mDictDir.isDirectory()) { - if (mDictDir.exists()) mDictDir.delete(); + if (mDictDir.exists()) { + FileUtils.deleteRecursively(mDictDir); + } mDictDir.mkdirs(); } - if (!trieFile.exists()) trieFile.createNewFile(); - if (!freqFile.exists()) freqFile.createNewFile(); - if (!terminalAddressTableFile.exists()) terminalAddressTableFile.createNewFile(); mTrieOutStream = new FileOutputStream(trieFile); + mHeaderOutStream = new FileOutputStream(headerFile); mFreqOutStream = new FileOutputStream(freqFile); mTerminalAddressTableOutStream = new FileOutputStream(terminalAddressTableFile); + if (formatOptions.mHasTimestamp) { + mUnigramTimestampOutStream = new FileOutputStream(timestampFile); + } } private void close() throws IOException { @@ -212,14 +213,21 @@ public class Ver4DictEncoder implements DictEncoder { if (mTrieOutStream != null) { mTrieOutStream.close(); } + if (mHeaderOutStream != null) { + mHeaderOutStream.close(); + } if (mFreqOutStream != null) { mFreqOutStream.close(); } if (mTerminalAddressTableOutStream != null) { mTerminalAddressTableOutStream.close(); } + if (mUnigramTimestampOutStream != null) { + mUnigramTimestampOutStream.close(); + } } finally { mTrieOutStream = null; + mHeaderOutStream = null; mFreqOutStream = null; mTerminalAddressTableOutStream = null; } @@ -240,16 +248,34 @@ public class Ver4DictEncoder implements DictEncoder { openStreams(formatOptions, dict.mOptions); } - mHeaderSize = BinaryDictEncoderUtils.writeDictionaryHeader(mTrieOutStream, dict, - formatOptions); + BinaryDictEncoderUtils.writeDictionaryHeader(mHeaderOutStream, dict, formatOptions); MakedictLog.i("Flattening the tree..."); ArrayList<PtNodeArray> flatNodes = BinaryDictEncoderUtils.flattenTree(dict.mRootNodeArray); int terminalCount = 0; + final ArrayList<PtNode> nodes = CollectionUtils.newArrayList(); for (final PtNodeArray array : flatNodes) { for (final PtNode node : array.mData) { - if (node.isTerminal()) node.mTerminalId = terminalCount++; + if (node.isTerminal()) { + nodes.add(node); + node.mTerminalId = terminalCount++; + } + } + } + Collections.sort(nodes, new Comparator<PtNode>() { + @Override + public int compare(final PtNode lhs, final PtNode rhs) { + if (lhs.mFrequency != rhs.mFrequency) { + return lhs.mFrequency < rhs.mFrequency ? -1 : 1; + } + if (lhs.mTerminalId < rhs.mTerminalId) return -1; + if (lhs.mTerminalId > rhs.mTerminalId) return 1; + return 0; } + }); + int count = 0; + for (final PtNode node : nodes) { + node.mTerminalId = count++; } MakedictLog.i("Computing addresses..."); @@ -257,7 +283,11 @@ public class Ver4DictEncoder implements DictEncoder { if (MakedictLog.DBG) BinaryDictEncoderUtils.checkFlatPtNodeArrayList(flatNodes); writeTerminalData(flatNodes, terminalCount); - mBigramWriter = new BigramContentWriter(mBaseFilename, terminalCount, mDictDir); + if (formatOptions.mHasTimestamp) { + initUnigramTimestamps(terminalCount); + } + mBigramWriter = new BigramContentWriter(mBaseFilename, terminalCount, mDictDir, + formatOptions.mHasTimestamp); writeBigrams(flatNodes, dict); mShortcutWriter = new ShortcutContentWriter(mBaseFilename, terminalCount, mDictDir); writeShortcuts(flatNodes); @@ -282,7 +312,7 @@ public class Ver4DictEncoder implements DictEncoder { @Override public void setPosition(int position) { - if (mTrieBuf == null || position < 0 || position >- mTrieBuf.length) return; + if (mTrieBuf == null || position < 0 || position > mTrieBuf.length) return; mTriePos = position; } @@ -299,7 +329,9 @@ public class Ver4DictEncoder implements DictEncoder { if (countSize != 1 && countSize != 2) { throw new RuntimeException("Strange size from getPtNodeCountSize : " + countSize); } - mTriePos = BinaryDictEncoderUtils.writeUIntToBuffer(mTrieBuf, mTriePos, ptNodeCount, + final int encodedPtNodeCount = (countSize == 2) ? + (ptNodeCount | FormatSpec.LARGE_PTNODE_ARRAY_SIZE_FIELD_SIZE_FLAG) : ptNodeCount; + mTriePos = BinaryDictEncoderUtils.writeUIntToBuffer(mTrieBuf, mTriePos, encodedPtNodeCount, countSize); } @@ -333,7 +365,7 @@ public class Ver4DictEncoder implements DictEncoder { private void writeChildrenPosition(PtNode ptNode, FormatOptions formatOptions) { final int childrenPos = BinaryDictEncoderUtils.getChildrenPosition(ptNode, formatOptions); - if (formatOptions.mSupportsDynamicUpdate) { + if (formatOptions.supportsDynamicUpdate()) { mTriePos += BinaryDictEncoderUtils.writeSignedChildrenPosition(mTrieBuf, mTriePos, childrenPos); } else { @@ -348,7 +380,7 @@ public class Ver4DictEncoder implements DictEncoder { for (final PtNodeArray nodeArray : flatNodes) { for (final PtNode ptNode : nodeArray.mData) { if (ptNode.mBigrams != null) { - mBigramWriter.writeBigramsForOneWord(ptNode.mTerminalId, + mBigramWriter.writeBigramsForOneWord(ptNode.mTerminalId, ptNode.mBigrams.size(), ptNode.mBigrams.iterator(), dict); } } @@ -400,7 +432,7 @@ public class Ver4DictEncoder implements DictEncoder { ptNode.mFrequency, FormatSpec.FREQUENCY_AND_FLAGS_SIZE); BinaryDictEncoderUtils.writeUIntToBuffer(terminalAddressTableBuf, ptNode.mTerminalId * FormatSpec.TERMINAL_ADDRESS_TABLE_ADDRESS_SIZE, - ptNode.mCachedAddressAfterUpdate + mHeaderSize, + ptNode.mCachedAddressAfterUpdate, FormatSpec.TERMINAL_ADDRESS_TABLE_ADDRESS_SIZE); } } @@ -408,4 +440,11 @@ public class Ver4DictEncoder implements DictEncoder { mFreqOutStream.write(freqBuf); mTerminalAddressTableOutStream.write(terminalAddressTableBuf); } + + private void initUnigramTimestamps(final int terminalCount) throws IOException { + // Initial value of time stamps for each word is 0. + final byte[] unigramTimestampBuf = + new byte[terminalCount * FormatSpec.UNIGRAM_TIMESTAMP_SIZE]; + mUnigramTimestampOutStream.write(unigramTimestampBuf); + } } diff --git a/java/src/com/android/inputmethod/latin/makedict/Ver4DictUpdater.java b/java/src/com/android/inputmethod/latin/makedict/Ver4DictUpdater.java index 3d8f186ba..c46bc36bb 100644 --- a/java/src/com/android/inputmethod/latin/makedict/Ver4DictUpdater.java +++ b/java/src/com/android/inputmethod/latin/makedict/Ver4DictUpdater.java @@ -17,29 +17,130 @@ package com.android.inputmethod.latin.makedict; import com.android.inputmethod.annotations.UsedForTesting; +import com.android.inputmethod.latin.makedict.BinaryDictDecoderUtils.CharEncoding; +import com.android.inputmethod.latin.makedict.FormatSpec.FileHeader; +import com.android.inputmethod.latin.makedict.FormatSpec.FormatOptions; +import com.android.inputmethod.latin.makedict.FusionDictionary.PtNode; import com.android.inputmethod.latin.makedict.FusionDictionary.WeightedString; +import com.android.inputmethod.latin.utils.CollectionUtils; + +import android.util.Log; import java.io.File; +import java.io.FileOutputStream; import java.io.IOException; +import java.io.OutputStream; import java.util.ArrayList; +import java.util.Arrays; +import java.util.Iterator; /** * An implementation of DictUpdater for version 4 binary dictionary. */ @UsedForTesting public class Ver4DictUpdater extends Ver4DictDecoder implements DictUpdater { + private static final String TAG = Ver4DictUpdater.class.getSimpleName(); + + private OutputStream mDictStream; + private final File mFrequencyFile; @UsedForTesting - public Ver4DictUpdater(final File dictDirectory, final int factoryType) { + public Ver4DictUpdater(final File dictDirectory, final int factoryType) + throws UnsupportedFormatException { // DictUpdater must have an updatable DictBuffer. super(dictDirectory, ((factoryType & MASK_DICTBUFFER) == USE_BYTEARRAY) ? USE_BYTEARRAY : USE_WRITABLE_BYTEBUFFER); + mFrequencyFile = getFile(FILETYPE_FREQUENCY); + } + + private static class BigramContentUpdater extends SparseTableContentUpdater { + public BigramContentUpdater(final String name, final File baseDir, + final boolean hasTimestamp) { + super(name + FormatSpec.BIGRAM_FILE_EXTENSION, + FormatSpec.BIGRAM_ADDRESS_TABLE_BLOCK_SIZE, baseDir, + BigramContentReader.getContentFilenames(name, hasTimestamp), + BigramContentReader.getContentIds(hasTimestamp), + new DictionaryBufferFromWritableByteBufferFactory()); + } + + public void insertBigramEntries(final int terminalId, final int frequency, + final ArrayList<PendingAttribute> entries) throws IOException { + if (terminalId < 0) { + throw new RuntimeException("Invalid terminal id : " + terminalId); + } + openStreamsAndBuffers(); + + if (entries == null || entries.isEmpty()) { + setContentValue(FormatSpec.BIGRAM_FREQ_CONTENT_INDEX, terminalId, + SparseTable.NOT_EXIST); + return; + } + final int positionOfEntries = + (int) mContentFiles[FormatSpec.BIGRAM_FREQ_CONTENT_INDEX].length(); + setContentValue(FormatSpec.BIGRAM_FREQ_CONTENT_INDEX, terminalId, positionOfEntries); + + final Iterator<PendingAttribute> bigramIterator = entries.iterator(); + while (bigramIterator.hasNext()) { + final PendingAttribute entry = bigramIterator.next(); + final int flags = BinaryDictEncoderUtils.makeBigramFlags(bigramIterator.hasNext(), + 0 /* offset */, entry.mFrequency, frequency, "" /* word */); + BinaryDictEncoderUtils.writeUIntToStream( + mContentOutStreams[FormatSpec.BIGRAM_FREQ_CONTENT_INDEX], flags, + FormatSpec.PTNODE_ATTRIBUTE_FLAGS_SIZE); + BinaryDictEncoderUtils.writeUIntToStream( + mContentOutStreams[FormatSpec.BIGRAM_FREQ_CONTENT_INDEX], entry.mAddress, + FormatSpec.PTNODE_ATTRIBUTE_MAX_ADDRESS_SIZE); + } + close(); + } + } + + private static class ShortcutContentUpdater extends SparseTableContentUpdater { + public ShortcutContentUpdater(final String name, final File baseDir) { + super(name + FormatSpec.SHORTCUT_FILE_EXTENSION, + FormatSpec.SHORTCUT_ADDRESS_TABLE_BLOCK_SIZE, baseDir, + new String[] { name + FormatSpec.SHORTCUT_FILE_EXTENSION }, + new String[] { FormatSpec.SHORTCUT_CONTENT_ID }, + new DictionaryBufferFromWritableByteBufferFactory()); + } + + public void insertShortcuts(final int terminalId, + final ArrayList<WeightedString> shortcuts) throws IOException { + if (terminalId < 0) { + throw new RuntimeException("Invalid terminal id : " + terminalId); + } + openStreamsAndBuffers(); + if (shortcuts == null || shortcuts.isEmpty()) { + setContentValue(FormatSpec.SHORTCUT_CONTENT_INDEX, terminalId, + SparseTable.NOT_EXIST); + return; + } + + final int positionOfShortcuts = + (int) mContentFiles[FormatSpec.SHORTCUT_CONTENT_INDEX].length(); + setContentValue(FormatSpec.SHORTCUT_CONTENT_INDEX, terminalId, positionOfShortcuts); + + final Iterator<WeightedString> shortcutIterator = shortcuts.iterator(); + while (shortcutIterator.hasNext()) { + final WeightedString target = shortcutIterator.next(); + final int shortcutFlags = BinaryDictEncoderUtils.makeShortcutFlags( + shortcutIterator.hasNext(), target.mFrequency); + BinaryDictEncoderUtils.writeUIntToStream( + mContentOutStreams[FormatSpec.SHORTCUT_CONTENT_INDEX], shortcutFlags, + FormatSpec.PTNODE_ATTRIBUTE_FLAGS_SIZE); + CharEncoding.writeString(mContentOutStreams[FormatSpec.SHORTCUT_CONTENT_INDEX], + target.mWord); + } + close(); + } } @Override public void deleteWord(final String word) throws IOException, UnsupportedFormatException { - if (mDictBuffer == null) openDictBuffer(); - readHeader(); + if (mDictBuffer == null) { + openDictBuffer(); + readHeader(); + } final int wordPos = getTerminalPosition(word); if (wordPos != FormatSpec.NOT_VALID_WORD) { mDictBuffer.position(wordPos); @@ -49,11 +150,623 @@ public class Ver4DictUpdater extends Ver4DictDecoder implements DictUpdater { } } - @Override + private int getNewTerminalId() { + // The size of frequency file is FormatSpec.FREQUENCY_AND_FLAGS_SIZE * number of terminals + // because each terminal always has a frequency. + // So we can get a fresh terminal id by this logic. + // CAVEAT: we are reading the file size from the disk each time: beware of race conditions, + // even on one thread. + return (int) (mFrequencyFile.length() / FormatSpec.FREQUENCY_AND_FLAGS_SIZE); + } + + private void updateParentPosIfNotMoved(final int nodePos, final int newParentPos, + final FormatOptions formatOptions) { + final int originalPos = getPosition(); + setPosition(nodePos); + final int flags = PtNodeReader.readPtNodeOptionFlags(mDictBuffer); + if (!BinaryDictIOUtils.isMovedPtNode(flags, formatOptions)) { + final int parentOffset = newParentPos - nodePos; + BinaryDictIOUtils.writeSInt24ToBuffer(mDictBuffer, parentOffset); + } + setPosition(originalPos); + } + + private void updateParentPositions(final int nodeArrayPos, final int newParentPos, + final FormatOptions formatOptions) { + final int originalPos = mDictBuffer.position(); + mDictBuffer.position(nodeArrayPos); + int jumpCount = 0; + do { + final int count = readPtNodeCount(); + for (int i = 0; i < count; ++i) { + updateParentPosIfNotMoved(getPosition(), newParentPos, formatOptions); + skipPtNode(formatOptions); + } + if (!readAndFollowForwardLink()) break; + } while (jumpCount++ < DynamicBinaryDictIOUtils.MAX_JUMPS); + setPosition(originalPos); + } + + private void updateChildrenPos(final int nodePos, final int newChildrenPos, + final FormatOptions options) { + final int originalPos = getPosition(); + setPosition(nodePos); + final int flags = PtNodeReader.readPtNodeOptionFlags(mDictBuffer); + PtNodeReader.readParentAddress(mDictBuffer, options); + BinaryDictIOUtils.skipString(mDictBuffer, + (flags & FormatSpec.FLAG_HAS_MULTIPLE_CHARS) != 0); + if ((flags & FormatSpec.FLAG_IS_TERMINAL) != 0) PtNodeReader.readTerminalId(mDictBuffer); + final int basePos = getPosition(); + BinaryDictIOUtils.writeSInt24ToBuffer(mDictBuffer, newChildrenPos - basePos); + setPosition(originalPos); + } + + private void updateTerminalPosition(final int terminalId, final int position) { + if (terminalId == PtNode.NOT_A_TERMINAL + || terminalId * FormatSpec.TERMINAL_ADDRESS_TABLE_ADDRESS_SIZE + >= mTerminalAddressTableBuffer.limit()) return; + mTerminalAddressTableBuffer.position(terminalId + * FormatSpec.TERMINAL_ADDRESS_TABLE_ADDRESS_SIZE); + BinaryDictEncoderUtils.writeUIntToDictBuffer(mTerminalAddressTableBuffer, position, + FormatSpec.TERMINAL_ADDRESS_TABLE_ADDRESS_SIZE); + } + + private void updateForwardLink(final int nodeArrayPos, final int newForwardLink, + final FormatOptions formatOptions) { + final int originalPos = getPosition(); + setPosition(nodeArrayPos); + int jumpCount = 0; + while (jumpCount++ < DynamicBinaryDictIOUtils.MAX_JUMPS) { + final int ptNodeCount = readPtNodeCount(); + for (int i = 0; i < ptNodeCount; ++i) { + skipPtNode(formatOptions); + } + final int forwardLinkPos = getPosition(); + if (!readAndFollowForwardLink()) { + setPosition(forwardLinkPos); + BinaryDictIOUtils.writeSInt24ToBuffer(mDictBuffer, newForwardLink - forwardLinkPos); + break; + } + } + setPosition(originalPos); + } + + private void markPtNodeAsMoved(final int nodePos, final int newNodePos, + final FormatOptions options) { + final int originalPos = getPosition(); + updateParentPosIfNotMoved(nodePos, newNodePos, options); + setPosition(nodePos); + final int currentFlags = PtNodeReader.readPtNodeOptionFlags(mDictBuffer); + setPosition(nodePos); + mDictBuffer.put((byte) (FormatSpec.FLAG_IS_MOVED + | (currentFlags & (~FormatSpec.MASK_MOVE_AND_DELETE_FLAG)))); + final int offset = newNodePos - nodePos; + BinaryDictIOUtils.writeSInt24ToBuffer(mDictBuffer, offset); + setPosition(originalPos); + } + + /** + * Writes a PtNode to an output stream from a Ver4PtNodeInfo. + * + * @param nodePos the position of the head of the PtNode. + * @param info the PtNode info to be written. + * @return the size written, in bytes. + */ + private int writePtNode(final int nodePos, final Ver4PtNodeInfo info) throws IOException { + int written = 0; + + // Write flags. + mDictStream.write((byte) (info.mFlags & 0xFF)); + written += FormatSpec.PTNODE_FLAGS_SIZE; + + // Write the parent position. + final int parentOffset = info.mParentPos == FormatSpec.NO_PARENT_ADDRESS ? + FormatSpec.NO_PARENT_ADDRESS : info.mParentPos - nodePos; + BinaryDictIOUtils.writeSInt24ToStream(mDictStream, parentOffset); + written += FormatSpec.PARENT_ADDRESS_SIZE; + + // Write a string. + if (((info.mFlags & FormatSpec.FLAG_HAS_MULTIPLE_CHARS) != 0) + != (info.mEndIndexOfCharacters - info.mStartIndexOfCharacters > 1)) { + throw new RuntimeException("Inconsistent flags : hasMultipleChars = " + + ((info.mFlags & FormatSpec.FLAG_HAS_MULTIPLE_CHARS) != 0) + ", length = " + + (info.mEndIndexOfCharacters - info.mStartIndexOfCharacters)); + } + written += CharEncoding.writeCodePoints(mDictStream, info.mCharacters, + info.mStartIndexOfCharacters, info.mEndIndexOfCharacters); + + // Write the terminal id. + if ((info.mFlags & FormatSpec.FLAG_IS_TERMINAL) != 0) { + BinaryDictEncoderUtils.writeUIntToStream(mDictStream, info.mTerminalId, + FormatSpec.PTNODE_TERMINAL_ID_SIZE); + written += FormatSpec.PTNODE_TERMINAL_ID_SIZE; + } + + // Write the children position. + final int childrenOffset = info.mChildrenPos == FormatSpec.NO_CHILDREN_ADDRESS + ? 0 : info.mChildrenPos - (nodePos + written); + BinaryDictIOUtils.writeSInt24ToStream(mDictStream, childrenOffset); + written += FormatSpec.SIGNED_CHILDREN_ADDRESS_SIZE; + + return written; + } + + /** + * Helper method to split and move PtNode. + * + * @param ptNodeArrayPos the position of PtNodeArray which contains the split and moved PtNode. + * @param splittedPtNodeToMovePos the position of the split and moved PtNode. + * @param newParent the parent PtNode after splitting. + * @param newChildren the children PtNodes after splitting. + * @param newParentStartPos where to write the new parent. + * @param formatOptions the format options. + */ + private void writeSplittedPtNodes(final int ptNodeArrayPos, final int splittedPtNodeToMovePos, + final Ver4PtNodeInfo newParent, final Ver4PtNodeInfo[] newChildren, + final int newParentStartPos, + final FormatOptions formatOptions) throws IOException { + updateTerminalPosition(newParent.mTerminalId, + newParentStartPos + 1 /* size of PtNodeCount */); + int written = writePtNodeArray(newParentStartPos, new Ver4PtNodeInfo[] { newParent }, + FormatSpec.NO_FORWARD_LINK_ADDRESS); + final int childrenStartPos = newParentStartPos + written; + writePtNodeArray(childrenStartPos, newChildren, FormatSpec.NO_FORWARD_LINK_ADDRESS); + int childrenNodePos = childrenStartPos + 1 /* size of PtNodeCount */; + for (final Ver4PtNodeInfo info : newChildren) { + updateTerminalPosition(info.mTerminalId, childrenNodePos); + childrenNodePos += computePtNodeSize(info.mCharacters, info.mStartIndexOfCharacters, + info.mEndIndexOfCharacters, + (info.mFlags & FormatSpec.FLAG_IS_TERMINAL) != 0); + } + + // Mark as moved. + markPtNodeAsMoved(splittedPtNodeToMovePos, newParentStartPos + 1 /* size of PtNodeCount */, + formatOptions); + updateForwardLink(ptNodeArrayPos, newParentStartPos, formatOptions); + } + + /** + * Writes a node array to the stream. + * + * @param nodeArrayPos the position of the head of the node array. + * @param infos an array of Ver4PtNodeInfo to be written. + * @return the written length in bytes. + */ + private int writePtNodeArray(final int nodeArrayPos, final Ver4PtNodeInfo[] infos, + final int forwardLink) throws IOException { + int written = BinaryDictIOUtils.writePtNodeCount(mDictStream, infos.length); + for (int i = 0; i < infos.length; ++i) { + written += writePtNode(nodeArrayPos + written, infos[i]); + } + BinaryDictIOUtils.writeSInt24ToStream(mDictStream, forwardLink); + written += FormatSpec.FORWARD_LINK_ADDRESS_SIZE; + return written; + } + + private int computePtNodeSize(final int[] codePoints, final int startIndex, final int endIndex, + final boolean isTerminal) { + return FormatSpec.PTNODE_FLAGS_SIZE + FormatSpec.PARENT_ADDRESS_SIZE + + CharEncoding.getCharArraySize(codePoints, startIndex, endIndex) + + (endIndex - startIndex > 1 ? FormatSpec.PTNODE_TERMINATOR_SIZE : 0) + + (isTerminal ? FormatSpec.PTNODE_TERMINAL_ID_SIZE : 0) + + FormatSpec.SIGNED_CHILDREN_ADDRESS_SIZE; + } + + private void writeNewSinglePtNodeWithAttributes(final int[] codePoints, + final boolean hasShortcuts, final int terminalId, final boolean hasBigrams, + final boolean isNotAWord, final boolean isBlackListEntry, final int parentPos, + final FormatOptions formatOptions) throws IOException { + final int newNodeArrayPos = mDictBuffer.limit(); + final int newNodeFlags = BinaryDictEncoderUtils.makePtNodeFlags(codePoints.length > 1, + terminalId != PtNode.NOT_A_TERMINAL, FormatSpec.FLAG_IS_NOT_MOVED, hasShortcuts, + hasBigrams, isNotAWord, isBlackListEntry, formatOptions); + final Ver4PtNodeInfo info = new Ver4PtNodeInfo(newNodeFlags, codePoints, terminalId, + FormatSpec.NO_CHILDREN_ADDRESS, parentPos, 0 /* nodeSize */); + writePtNodeArray(newNodeArrayPos, new Ver4PtNodeInfo[] { info }, + FormatSpec.NO_FORWARD_LINK_ADDRESS); + } + + private int setMultipleCharsInFlags(final int currentFlags, final boolean hasMultipleChars) { + final int flags; + if (hasMultipleChars) { + flags = currentFlags | FormatSpec.FLAG_HAS_MULTIPLE_CHARS; + } else { + flags = currentFlags & (~FormatSpec.FLAG_HAS_MULTIPLE_CHARS); + } + return flags; + } + + private int setIsNotAWordInFlags(final int currentFlags, final boolean isNotAWord) { + final int flags; + if (isNotAWord) { + flags = currentFlags | FormatSpec.FLAG_IS_NOT_A_WORD; + } else { + flags = currentFlags & (~FormatSpec.FLAG_IS_NOT_A_WORD); + } + return flags; + } + + private int setIsBlackListEntryInFlags(final int currentFlags, final boolean isBlackListEntry) { + final int flags; + if (isBlackListEntry) { + flags = currentFlags | FormatSpec.FLAG_IS_BLACKLISTED; + } else { + flags = currentFlags & (~FormatSpec.FLAG_IS_BLACKLISTED); + } + return flags; + } + + /** + * Splits a PtNode. + * + * abcd - ef + * + * -> inserting "abc" + * + * abc - d - ef + * + * @param nodeArrayToSplitPos the position of PtNodeArray which contains the PtNode to split. + * @param nodeToSplitPos the position of the PtNode to split. + * @param nodeToSplitInfo the information of the PtNode to split. + * @param indexToSplit the index where to split in the code points array. + * @param parentOfNodeToSplitPos the absolute position of a parent of the node to split. + * @param newTerminalId the terminal id of the inserted node (corresponds to "d"). + * @param hasShortcuts whether the inserted word should have shortcuts. + * @param hasBigrams whether the inserted word should have bigrams. + * @param isNotAWord whether the inserted word should be not a word. + * @param isBlackListEntry whether the inserted word should be a black list entry. + * @param formatOptions the format options. + */ + private void splitOnly(final int nodeArrayToSplitPos, final int nodeToSplitPos, + final Ver4PtNodeInfo nodeToSplitInfo, final int indexToSplit, + final int parentOfNodeToSplitPos, final int newTerminalId, final boolean hasShortcuts, + final boolean hasBigrams, final boolean isNotAWord, final boolean isBlackListEntry, + final FormatOptions formatOptions) throws IOException { + final int parentNodeArrayStartPos = mDictBuffer.limit(); + final int parentNodeStartPos = parentNodeArrayStartPos + 1 /* size of PtNodeCount */; + final int parentFlags = BinaryDictEncoderUtils.makePtNodeFlags(indexToSplit > 1, + true /* isTerminal */, FormatSpec.FLAG_IS_NOT_MOVED, hasShortcuts, hasBigrams, + isNotAWord, isBlackListEntry, formatOptions); + final Ver4PtNodeInfo parentInfo = new Ver4PtNodeInfo(parentFlags, + nodeToSplitInfo.mCharacters, newTerminalId, parentNodeStartPos + + computePtNodeSize(nodeToSplitInfo.mCharacters, 0, indexToSplit, true) + + FormatSpec.FORWARD_LINK_ADDRESS_SIZE, + parentOfNodeToSplitPos, 0 /* nodeSize */); + parentInfo.mStartIndexOfCharacters = 0; + parentInfo.mEndIndexOfCharacters = indexToSplit; + + // Write the child. + final int childrenFlags = setMultipleCharsInFlags(nodeToSplitInfo.mFlags, + nodeToSplitInfo.mCharacters.length - indexToSplit > 1); + final Ver4PtNodeInfo childrenInfo = new Ver4PtNodeInfo(childrenFlags, + nodeToSplitInfo.mCharacters, nodeToSplitInfo.mTerminalId, + nodeToSplitInfo.mChildrenPos, parentNodeStartPos, 0 /* nodeSize */); + childrenInfo.mStartIndexOfCharacters = indexToSplit; + childrenInfo.mEndIndexOfCharacters = nodeToSplitInfo.mCharacters.length; + if (nodeToSplitInfo.mChildrenPos != FormatSpec.NO_CHILDREN_ADDRESS) { + updateParentPositions(nodeToSplitInfo.mChildrenPos, + parentInfo.mChildrenPos + 1 /* size of PtNodeCount */, formatOptions); + } + + writeSplittedPtNodes(nodeArrayToSplitPos, nodeToSplitPos, parentInfo, + new Ver4PtNodeInfo[] { childrenInfo }, parentNodeArrayStartPos, formatOptions); + } + + /** + * Split and branch a PtNode. + * + * ab - cd + * + * -> inserting "ac" + * + * a - b - cd + * | + * - c + * + * @param nodeArrayToSplitPos the position of PtNodeArray which contains the PtNode to split. + * @param nodeToSplitPos the position of the PtNode to split. + * @param nodeToSplitInfo the information of the PtNode to split. + * @param indexToSplit the index where to split in the code points array. + * @param parentOfNodeToSplitPos the absolute position of parent of the node to split. + * @param newWordSuffixCodePoints the suffix of the newly inserted word (corresponds to "c"). + * @param startIndexOfNewWordSuffixCodePoints the start index in newWordSuffixCodePoints where + * the suffix starts. + * @param newTerminalId the terminal id of the inserted node (correspond to "c"). + * @param hasShortcuts whether the inserted word should have shortcuts. + * @param hasBigrams whether the inserted word should have bigrams. + * @param isNotAWord whether the inserted word should be not a word. + * @param isBlackListEntry whether the inserted word should be a black list entry. + * @param formatOptions the format options. + */ + private void splitAndBranch(final int nodeArrayToSplitPos, final int nodeToSplitPos, + final Ver4PtNodeInfo nodeToSplitInfo, final int indexToSplit, + final int parentOfNodeToSplitPos, final int[] newWordSuffixCodePoints, + final int startIndexOfNewWordSuffixCodePoints, + final int newTerminalId, + final boolean hasShortcuts, final boolean hasBigrams, final boolean isNotAWord, + final boolean isBlackListEntry, final FormatOptions formatOptions) throws IOException { + final int parentNodeArrayStartPos = mDictBuffer.limit(); + final int parentNodeStartPos = parentNodeArrayStartPos + 1 /* size of PtNodeCount */; + final int parentFlags = BinaryDictEncoderUtils.makePtNodeFlags( + indexToSplit > 1, + false /* isTerminal */, FormatSpec.FLAG_IS_NOT_MOVED, + false /* hasShortcut */, false /* hasBigrams */, + false /* isNotAWord */, false /* isBlackListEntry */, formatOptions); + final Ver4PtNodeInfo parentInfo = new Ver4PtNodeInfo(parentFlags, + nodeToSplitInfo.mCharacters, PtNode.NOT_A_TERMINAL, + parentNodeStartPos + + computePtNodeSize(nodeToSplitInfo.mCharacters, 0, indexToSplit, false) + + FormatSpec.FORWARD_LINK_ADDRESS_SIZE, + parentOfNodeToSplitPos, 0 /* nodeSize */); + parentInfo.mStartIndexOfCharacters = 0; + parentInfo.mEndIndexOfCharacters = indexToSplit; + + final int childrenNodeArrayStartPos = parentNodeStartPos + + computePtNodeSize(nodeToSplitInfo.mCharacters, 0, indexToSplit, false) + + FormatSpec.FORWARD_LINK_ADDRESS_SIZE; + final int firstChildrenFlags = BinaryDictEncoderUtils.makePtNodeFlags( + newWordSuffixCodePoints.length - startIndexOfNewWordSuffixCodePoints > 1, + true /* isTerminal */, FormatSpec.FLAG_IS_NOT_MOVED, hasShortcuts, hasBigrams, + isNotAWord, isBlackListEntry, formatOptions); + final Ver4PtNodeInfo firstChildrenInfo = new Ver4PtNodeInfo(firstChildrenFlags, + newWordSuffixCodePoints, newTerminalId, + FormatSpec.NO_CHILDREN_ADDRESS, parentNodeStartPos, + 0 /* nodeSize */); + firstChildrenInfo.mStartIndexOfCharacters = startIndexOfNewWordSuffixCodePoints; + firstChildrenInfo.mEndIndexOfCharacters = newWordSuffixCodePoints.length; + + final int secondChildrenStartPos = childrenNodeArrayStartPos + 1 /* size of ptNodeCount */ + + computePtNodeSize(newWordSuffixCodePoints, startIndexOfNewWordSuffixCodePoints, + newWordSuffixCodePoints.length, true /* isTerminal */); + final int secondChildrenFlags = setMultipleCharsInFlags(nodeToSplitInfo.mFlags, + nodeToSplitInfo.mCharacters.length - indexToSplit > 1); + final Ver4PtNodeInfo secondChildrenInfo = new Ver4PtNodeInfo(secondChildrenFlags, + nodeToSplitInfo.mCharacters, nodeToSplitInfo.mTerminalId, + nodeToSplitInfo.mChildrenPos, parentNodeStartPos, 0 /* nodeSize */); + secondChildrenInfo.mStartIndexOfCharacters = indexToSplit; + secondChildrenInfo.mEndIndexOfCharacters = nodeToSplitInfo.mCharacters.length; + if (nodeToSplitInfo.mChildrenPos != FormatSpec.NO_CHILDREN_ADDRESS) { + updateParentPositions(nodeToSplitInfo.mChildrenPos, secondChildrenStartPos, + formatOptions); + } + + writeSplittedPtNodes(nodeArrayToSplitPos, nodeToSplitPos, parentInfo, + new Ver4PtNodeInfo[] { firstChildrenInfo, secondChildrenInfo }, + parentNodeArrayStartPos, formatOptions); + } + + /** + * Inserts a word into the trie file and returns the position of inserted terminal node. + * If the insertion is failed, returns FormatSpec.NOT_VALID_WORD. + */ + @UsedForTesting + private int insertWordToTrie(final String word, final int newTerminalId, + final boolean isNotAWord, final boolean isBlackListEntry, final boolean hasBigrams, + final boolean hasShortcuts) throws IOException, UnsupportedFormatException { + setPosition(0); + final FileHeader header = readHeader(); + + final int[] codePoints = FusionDictionary.getCodePoints(word); + final int wordLen = codePoints.length; + + int wordPos = 0; + for (int depth = 0; depth < FormatSpec.MAX_WORD_LENGTH; /* nop */) { + final int nodeArrayPos = getPosition(); + final int ptNodeCount = readPtNodeCount(); + boolean goToChildren = false; + int parentPos = FormatSpec.NO_PARENT_ADDRESS; + for (int i = 0; i < ptNodeCount; ++i) { + final int nodePos = getPosition(); + final Ver4PtNodeInfo nodeInfo = readVer4PtNodeInfo(nodePos, header.mFormatOptions); + if (BinaryDictIOUtils.isMovedPtNode(nodeInfo.mFlags, header.mFormatOptions)) { + continue; + } + if (nodeInfo.mParentPos != FormatSpec.NO_PARENT_ADDRESS) { + parentPos = nodePos + nodeInfo.mParentPos; + } + + final boolean firstCharacterMatched = + codePoints[wordPos] == nodeInfo.mCharacters[0]; + boolean allCharactersMatched = true; + int firstDifferentCharacterIndex = -1; + for (int p = 0; p < nodeInfo.mCharacters.length; ++p) { + if (wordPos + p >= codePoints.length) break; + if (codePoints[wordPos + p] != nodeInfo.mCharacters[p]) { + if (firstDifferentCharacterIndex == -1) { + firstDifferentCharacterIndex = p; + } + allCharactersMatched = false; + } + } + + if (!firstCharacterMatched) { + // Go to the next sibling node. + continue; + } + + if (!allCharactersMatched) { + final int parentNodeArrayStartPos = mDictBuffer.limit(); + splitAndBranch(nodeArrayPos, nodePos, nodeInfo, firstDifferentCharacterIndex, + parentPos, codePoints, wordPos + firstDifferentCharacterIndex, + newTerminalId, hasShortcuts, hasBigrams, isNotAWord, + isBlackListEntry, header.mFormatOptions); + + return parentNodeArrayStartPos + computePtNodeSize(codePoints, wordPos, + wordPos + firstDifferentCharacterIndex, false) + + FormatSpec.FORWARD_LINK_ADDRESS_SIZE + 1 /* size of PtNodeCount */; + } + + if (wordLen - wordPos < nodeInfo.mCharacters.length) { + final int parentNodeArrayStartPos = mDictBuffer.limit(); + splitOnly(nodeArrayPos, nodePos, nodeInfo, wordLen - wordPos, parentPos, + newTerminalId, hasShortcuts, hasBigrams, isNotAWord, isBlackListEntry, + header.mFormatOptions); + + // Return the position of the inserted word. + return parentNodeArrayStartPos + 1 /* size of PtNodeCount */; + } + + wordPos += nodeInfo.mCharacters.length; + if (wordPos == wordLen) { + // This dictionary already contains the word. + Log.e(TAG, "Something went wrong. If the word is already contained, " + + " there is no need to insert new PtNode."); + return FormatSpec.NOT_VALID_WORD; + } + if (nodeInfo.mChildrenPos == FormatSpec.NO_CHILDREN_ADDRESS) { + // There are no children. + // We need to add a new node as a child of this node. + final int newNodeArrayPos = mDictBuffer.limit(); + final int[] newNodeCodePoints = Arrays.copyOfRange(codePoints, wordPos, + codePoints.length); + writeNewSinglePtNodeWithAttributes(newNodeCodePoints, hasShortcuts, + newTerminalId, hasBigrams, isNotAWord, isBlackListEntry, nodePos, + header.mFormatOptions); + updateChildrenPos(nodePos, newNodeArrayPos, header.mFormatOptions); + return newNodeArrayPos + 1 /* size of PtNodeCount */; + } else { + // Found the matched node. + // Go to the children of this node. + setPosition(nodeInfo.mChildrenPos); + goToChildren = true; + depth++; + break; + } + } + + if (goToChildren) continue; + if (!readAndFollowForwardLink()) { + // Add a new node that contains [wordPos, word.length()-1]. + // and update the forward link. + final int newNodeArrayPos = mDictBuffer.limit(); + final int[] newCodePoints = Arrays.copyOfRange(codePoints, wordPos, + codePoints.length); + writeNewSinglePtNodeWithAttributes(newCodePoints, hasShortcuts, newTerminalId, + hasBigrams, isNotAWord, isBlackListEntry, parentPos, header.mFormatOptions); + updateForwardLink(nodeArrayPos, newNodeArrayPos, header.mFormatOptions); + return newNodeArrayPos + 1 /* size of PtNodeCount */; + } + } + return FormatSpec.NOT_VALID_WORD; + } + + private void updateFrequency(final int terminalId, final int frequency) { + mFrequencyBuffer.position(terminalId * FormatSpec.FREQUENCY_AND_FLAGS_SIZE); + BinaryDictEncoderUtils.writeUIntToDictBuffer(mFrequencyBuffer, frequency, + FormatSpec.FREQUENCY_AND_FLAGS_SIZE); + } + + private void insertFrequency(final int frequency) throws IOException { + final OutputStream frequencyStream = new FileOutputStream(mFrequencyFile, + true /* append */); + BinaryDictEncoderUtils.writeUIntToStream(frequencyStream, frequency, + FormatSpec.FREQUENCY_AND_FLAGS_SIZE); + frequencyStream.close(); + } + + private void insertTerminalPosition(final int posOfTerminal) throws IOException, + UnsupportedFormatException { + final OutputStream terminalPosStream = new FileOutputStream( + getFile(FILETYPE_TERMINAL_ADDRESS_TABLE), true /* append */); + BinaryDictEncoderUtils.writeUIntToStream(terminalPosStream, posOfTerminal, + FormatSpec.TERMINAL_ADDRESS_TABLE_ADDRESS_SIZE); + terminalPosStream.close(); + } + + private void insertBigrams(final int terminalId, final int frequency, + final ArrayList<PendingAttribute> bigramAddresses) + throws IOException, UnsupportedFormatException { + openDictBuffer(); + final BigramContentUpdater updater = new BigramContentUpdater(mDictDirectory.getName(), + mDictDirectory, false); + + // Convert addresses to terminal ids. + final ArrayList<PendingAttribute> bigrams = CollectionUtils.newArrayList(); + mDictBuffer.position(0); + final FileHeader header = readHeader(); + for (PendingAttribute attr : bigramAddresses) { + mDictBuffer.position(attr.mAddress); + final Ver4PtNodeInfo info = readVer4PtNodeInfo(attr.mAddress, header.mFormatOptions); + if (info.mTerminalId == PtNode.NOT_A_TERMINAL) { + throw new RuntimeException("We can't have a bigram target that's not a terminal."); + } + bigrams.add(new PendingAttribute(frequency, info.mTerminalId)); + } + updater.insertBigramEntries(terminalId, frequency, bigrams); + close(); + } + + private void insertShortcuts(final int terminalId, final ArrayList<WeightedString> shortcuts) + throws IOException { + final ShortcutContentUpdater updater = new ShortcutContentUpdater(mDictDirectory.getName(), + mDictDirectory); + updater.insertShortcuts(terminalId, shortcuts); + } + + private void openBuffersAndStream() throws IOException, UnsupportedFormatException { + openDictBuffer(); + mDictStream = new FileOutputStream(getFile(FILETYPE_TRIE), true /* append */); + } + + private void close() throws IOException { + if (mDictStream != null) { + mDictStream.close(); + mDictStream = null; + } + mDictBuffer = null; + mFrequencyBuffer = null; + mTerminalAddressTableBuffer = null; + } + + private void updateAttributes(final int posOfWord, final int frequency, + final ArrayList<WeightedString> bigramStrings, + final ArrayList<WeightedString> shortcuts, final boolean isNotAWord, + final boolean isBlackListEntry) throws IOException, UnsupportedFormatException { + mDictBuffer.position(0); + final FileHeader header = readHeader(); + mDictBuffer.position(posOfWord); + final Ver4PtNodeInfo info = readVer4PtNodeInfo(posOfWord, header.mFormatOptions); + final int terminalId = info.mTerminalId; + + // Update the flags. + final int newFlags = setIsNotAWordInFlags( + setIsBlackListEntryInFlags(info.mFlags, isBlackListEntry), isNotAWord); + mDictBuffer.position(posOfWord); + mDictBuffer.put((byte) newFlags); + + updateFrequency(terminalId, frequency); + insertBigrams(terminalId, frequency, + DynamicBinaryDictIOUtils.resolveBigramPositions(this, bigramStrings)); + insertShortcuts(terminalId, shortcuts); + } + + @Override @UsedForTesting public void insertWord(final String word, final int frequency, final ArrayList<WeightedString> bigramStrings, final ArrayList<WeightedString> shortcuts, final boolean isNotAWord, final boolean isBlackListEntry) throws IOException, UnsupportedFormatException { - // TODO: Implement this method. + final int newTerminalId = getNewTerminalId(); + + openBuffersAndStream(); + final int posOfWord = getTerminalPosition(word); + if (posOfWord != FormatSpec.NOT_VALID_WORD) { + // The word is already contained in the dictionary. + updateAttributes(posOfWord, frequency, bigramStrings, shortcuts, isNotAWord, + isBlackListEntry); + close(); + return; + } + + // Insert new PtNode into trie. + final int posOfTerminal = insertWordToTrie(word, newTerminalId, isNotAWord, + isBlackListEntry, bigramStrings != null && !bigramStrings.isEmpty(), + shortcuts != null && !shortcuts.isEmpty()); + insertFrequency(frequency); + insertTerminalPosition(posOfTerminal); + close(); + + insertBigrams(newTerminalId, frequency, + DynamicBinaryDictIOUtils.resolveBigramPositions(this, bigramStrings)); + insertShortcuts(newTerminalId, shortcuts); } } |