From fd46e87da2e91583e0b0ab0dd7ffa5e329012861 Mon Sep 17 00:00:00 2001 From: Yuichiro Hanada Date: Thu, 19 Sep 2013 21:09:34 +0900 Subject: [PATCH] Separate bigram entries from the trie file. Bug: 10920165 Change-Id: I42eb7f228994a56c2f31a1c83904c8c360b6d088 --- .../makedict/BinaryDictEncoderUtils.java | 14 ++-- .../latin/makedict/FormatSpec.java | 4 + .../latin/makedict/SparseTable.java | 44 +++++++++++ .../latin/makedict/Ver4DictDecoder.java | 34 ++++++++- .../latin/makedict/Ver4DictEncoder.java | 76 +++++++++++++------ 5 files changed, 142 insertions(+), 30 deletions(-) diff --git a/java/src/com/android/inputmethod/latin/makedict/BinaryDictEncoderUtils.java b/java/src/com/android/inputmethod/latin/makedict/BinaryDictEncoderUtils.java index 6cc0bfb76..af61f2979 100644 --- a/java/src/com/android/inputmethod/latin/makedict/BinaryDictEncoderUtils.java +++ b/java/src/com/android/inputmethod/latin/makedict/BinaryDictEncoderUtils.java @@ -385,12 +385,14 @@ public class BinaryDictEncoderUtils { nodeSize + size, ptNode.mChildren)); } nodeSize += getShortcutListSize(ptNode.mShortcutTargets); - if (null != ptNode.mBigrams) { - for (WeightedString bigram : ptNode.mBigrams) { - final int offset = getOffsetToTargetPtNodeDuringUpdate(ptNodeArray, - nodeSize + size + FormatSpec.PTNODE_ATTRIBUTE_FLAGS_SIZE, - FusionDictionary.findWordInTree(dict.mRootNodeArray, bigram.mWord)); - nodeSize += getByteSize(offset) + FormatSpec.PTNODE_ATTRIBUTE_FLAGS_SIZE; + if (formatOptions.mVersion < FormatSpec.FIRST_VERSION_WITH_TERMINAL_ID) { + if (null != ptNode.mBigrams) { + for (WeightedString bigram : ptNode.mBigrams) { + final int offset = getOffsetToTargetPtNodeDuringUpdate(ptNodeArray, + nodeSize + size + FormatSpec.PTNODE_ATTRIBUTE_FLAGS_SIZE, + FusionDictionary.findWordInTree(dict.mRootNodeArray, bigram.mWord)); + nodeSize += getByteSize(offset) + FormatSpec.PTNODE_ATTRIBUTE_FLAGS_SIZE; + } } } ptNode.mCachedSize = nodeSize; diff --git a/java/src/com/android/inputmethod/latin/makedict/FormatSpec.java b/java/src/com/android/inputmethod/latin/makedict/FormatSpec.java index 2765222af..9481a8c14 100644 --- a/java/src/com/android/inputmethod/latin/makedict/FormatSpec.java +++ b/java/src/com/android/inputmethod/latin/makedict/FormatSpec.java @@ -265,8 +265,12 @@ public final class FormatSpec { static final String FREQ_FILE_EXTENSION = ".freq"; // tat = Terminal Address Table static final String TERMINAL_ADDRESS_TABLE_FILE_EXTENSION = ".tat"; + static final String BIGRAM_FILE_EXTENSION = ".bigram"; + static final String BIGRAM_LOOKUP_TABLE_FILE_EXTENSION = ".bigram_lookup"; + static final String BIGRAM_ADDRESS_TABLE_FILE_EXTENSION = ".bigram_index"; static final int FREQUENCY_AND_FLAGS_SIZE = 2; static final int TERMINAL_ADDRESS_TABLE_ADDRESS_SIZE = 3; + static final int BIGRAM_ADDRESS_TABLE_BLOCK_SIZE = 4; static final int NO_CHILDREN_ADDRESS = Integer.MIN_VALUE; static final int NO_PARENT_ADDRESS = 0; diff --git a/java/src/com/android/inputmethod/latin/makedict/SparseTable.java b/java/src/com/android/inputmethod/latin/makedict/SparseTable.java index 0b9cf91d2..96d057a44 100644 --- a/java/src/com/android/inputmethod/latin/makedict/SparseTable.java +++ b/java/src/com/android/inputmethod/latin/makedict/SparseTable.java @@ -18,6 +18,9 @@ package com.android.inputmethod.latin.makedict; import com.android.inputmethod.annotations.UsedForTesting; +import java.io.File; +import java.io.FileInputStream; +import java.io.FileOutputStream; import java.io.IOException; import java.io.OutputStream; import java.util.ArrayList; @@ -147,4 +150,45 @@ public class SparseTable { BinaryDictEncoderUtils.writeUIntToStream(contentOutStream, index, 4); } } + + @UsedForTesting + public void writeToFiles(final File lookupTableFile, final File contentFile) + throws IOException { + FileOutputStream lookupTableOutStream = null; + FileOutputStream contentOutStream = null; + try { + lookupTableOutStream = new FileOutputStream(lookupTableFile); + contentOutStream = new FileOutputStream(contentFile); + write(lookupTableOutStream, contentOutStream); + } finally { + if (lookupTableOutStream != null) { + lookupTableOutStream.close(); + } + if (contentOutStream != null) { + contentOutStream.close(); + } + } + } + + private static byte[] readFileToByteArray(final File file) throws IOException { + final byte[] contents = new byte[(int) file.length()]; + FileInputStream inStream = null; + try { + inStream = new FileInputStream(file); + inStream.read(contents); + } finally { + if (inStream != null) { + inStream.close(); + } + } + return contents; + } + + @UsedForTesting + public static SparseTable readFromFiles(final File lookupTableFile, final File contentFile, + final int blockSize) throws IOException { + final byte[] lookupTable = readFileToByteArray(lookupTableFile); + final byte[] content = readFileToByteArray(contentFile); + return new SparseTable(lookupTable, content, blockSize); + } } diff --git a/java/src/com/android/inputmethod/latin/makedict/Ver4DictDecoder.java b/java/src/com/android/inputmethod/latin/makedict/Ver4DictDecoder.java index 4c8ff8ea4..0aa431966 100644 --- a/java/src/com/android/inputmethod/latin/makedict/Ver4DictDecoder.java +++ b/java/src/com/android/inputmethod/latin/makedict/Ver4DictDecoder.java @@ -42,12 +42,15 @@ public class Ver4DictDecoder extends DictDecoder { 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 = 4; private final File mDictDirectory; private final DictionaryBufferFactory mBufferFactory; private DictBuffer mDictBuffer; private DictBuffer mFrequencyBuffer; private DictBuffer mTerminalAddressTableBuffer; + private DictBuffer mBigramBuffer; + private SparseTable mBigramAddressTable; @UsedForTesting /* package */ Ver4DictDecoder(final File dictDirectory, final int factoryFlag) { @@ -82,6 +85,9 @@ public class Ver4DictDecoder extends DictDecoder { } else if (fileType == FILETYPE_TERMINAL_ADDRESS_TABLE) { return new File(mDictDirectory, mDictDirectory.getName() + FormatSpec.TERMINAL_ADDRESS_TABLE_FILE_EXTENSION); + } else if (fileType == FILETYPE_BIGRAM) { + return new File(mDictDirectory, + mDictDirectory.getName() + FormatSpec.BIGRAM_FILE_EXTENSION); } else { throw new RuntimeException("Unsupported kind of file : " + fileType); } @@ -94,6 +100,8 @@ public class Ver4DictDecoder extends DictDecoder { mFrequencyBuffer = mBufferFactory.getDictionaryBuffer(getFile(FILETYPE_FREQUENCY)); mTerminalAddressTableBuffer = mBufferFactory.getDictionaryBuffer( getFile(FILETYPE_TERMINAL_ADDRESS_TABLE)); + mBigramBuffer = mBufferFactory.getDictionaryBuffer(getFile(FILETYPE_BIGRAM)); + loadBigramAddressSparseTable(); } @Override @@ -118,6 +126,15 @@ public class Ver4DictDecoder extends DictDecoder { return header; } + private void loadBigramAddressSparseTable() throws IOException { + final File lookupIndexFile = new File(mDictDirectory, + mDictDirectory.getName() + FormatSpec.BIGRAM_LOOKUP_TABLE_FILE_EXTENSION); + final File contentFile = new File(mDictDirectory, + mDictDirectory.getName() + FormatSpec.BIGRAM_ADDRESS_TABLE_FILE_EXTENSION); + mBigramAddressTable = SparseTable.readFromFiles(lookupIndexFile, contentFile, + FormatSpec.BIGRAM_ADDRESS_TABLE_BLOCK_SIZE); + } + protected static class PtNodeReader extends DictDecoder.PtNodeReader { protected static int readFrequency(final DictBuffer frequencyBuffer, final int terminalId) { frequencyBuffer.position(terminalId * FormatSpec.FREQUENCY_AND_FLAGS_SIZE + 1); @@ -191,8 +208,21 @@ public class Ver4DictDecoder extends DictDecoder { final ArrayList bigrams; if (0 != (flags & FormatSpec.FLAG_HAS_BIGRAMS)) { bigrams = new ArrayList(); - addressPointer += PtNodeReader.readBigramAddresses(mDictBuffer, bigrams, - addressPointer); + final int posOfBigrams = mBigramAddressTable.get(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) { MakedictLog.d("too many bigrams in a node."); } diff --git a/java/src/com/android/inputmethod/latin/makedict/Ver4DictEncoder.java b/java/src/com/android/inputmethod/latin/makedict/Ver4DictEncoder.java index 4fb89671f..4c25faf88 100644 --- a/java/src/com/android/inputmethod/latin/makedict/Ver4DictEncoder.java +++ b/java/src/com/android/inputmethod/latin/makedict/Ver4DictEncoder.java @@ -26,6 +26,7 @@ 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 java.io.ByteArrayOutputStream; import java.io.File; import java.io.FileNotFoundException; import java.io.FileOutputStream; @@ -43,9 +44,13 @@ public class Ver4DictEncoder implements DictEncoder { private byte[] mTrieBuf; private int mTriePos; private int mHeaderSize; + private SparseTable mBigramAddressTable; private OutputStream mTrieOutStream; private OutputStream mFreqOutStream; private OutputStream mTerminalAddressTableOutStream; + private OutputStream mBigramOutStream; + private File mDictDir; + private String mBaseFilename; @UsedForTesting public Ver4DictEncoder(final File dictPlacedDir) { @@ -55,12 +60,14 @@ public class Ver4DictEncoder implements DictEncoder { private void openStreams(final FormatOptions formatOptions, final DictionaryOptions dictOptions) throws FileNotFoundException, IOException { final FileHeader header = new FileHeader(0, dictOptions, formatOptions); - final String filename = header.getId() + "." + header.getVersion(); - final File mDictDir = new File(mDictPlacedDir, filename); - final File trieFile = new File(mDictDir, filename + FormatSpec.TRIE_FILE_EXTENSION); - final File freqFile = new File(mDictDir, filename + FormatSpec.FREQ_FILE_EXTENSION); + mBaseFilename = header.getId() + "." + header.getVersion(); + mDictDir = new File(mDictPlacedDir, mBaseFilename); + final File trieFile = new File(mDictDir, mBaseFilename + FormatSpec.TRIE_FILE_EXTENSION); + final File freqFile = new File(mDictDir, mBaseFilename + FormatSpec.FREQ_FILE_EXTENSION); final File terminalAddressTableFile = new File(mDictDir, - filename + FormatSpec.TERMINAL_ADDRESS_TABLE_FILE_EXTENSION); + mBaseFilename + FormatSpec.TERMINAL_ADDRESS_TABLE_FILE_EXTENSION); + final File bigramFile = new File(mDictDir, + mBaseFilename + FormatSpec.BIGRAM_FILE_EXTENSION); if (!mDictDir.isDirectory()) { if (mDictDir.exists()) mDictDir.delete(); mDictDir.mkdirs(); @@ -71,6 +78,7 @@ public class Ver4DictEncoder implements DictEncoder { mTrieOutStream = new FileOutputStream(trieFile); mFreqOutStream = new FileOutputStream(freqFile); mTerminalAddressTableOutStream = new FileOutputStream(terminalAddressTableFile); + mBigramOutStream = new FileOutputStream(bigramFile); } private void close() throws IOException { @@ -84,10 +92,14 @@ public class Ver4DictEncoder implements DictEncoder { if (mTerminalAddressTableOutStream != null) { mTerminalAddressTableOutStream.close(); } + if (mBigramOutStream != null) { + mBigramOutStream.close(); + } } finally { mTrieOutStream = null; mFreqOutStream = null; mTerminalAddressTableOutStream = null; + mBigramOutStream = null; } } @@ -123,6 +135,10 @@ public class Ver4DictEncoder implements DictEncoder { if (MakedictLog.DBG) BinaryDictEncoderUtils.checkFlatPtNodeArrayList(flatNodes); writeTerminalData(flatNodes, terminalCount); + mBigramAddressTable = new SparseTable(terminalCount, + FormatSpec.BIGRAM_ADDRESS_TABLE_BLOCK_SIZE); + writeBigrams(flatNodes, dict); + writeBigramAddressSparseTable(); final PtNodeArray lastNodeArray = flatNodes.get(flatNodes.size() - 1); final int bufferSize = lastNodeArray.mCachedAddressAfterUpdate + lastNodeArray.mCachedSize; @@ -230,24 +246,41 @@ public class Ver4DictEncoder implements DictEncoder { shortcutByteSize, FormatSpec.PTNODE_SHORTCUT_LIST_SIZE_SIZE); } - private void writeBigrams(ArrayList bigrams, FusionDictionary dict) { - if (bigrams == null) return; + private void writeBigrams(final ArrayList flatNodes, final FusionDictionary dict) + throws IOException { + final ByteArrayOutputStream bigramBuffer = new ByteArrayOutputStream(); - final Iterator bigramIterator = bigrams.iterator(); - while (bigramIterator.hasNext()) { - final WeightedString bigram = bigramIterator.next(); - final PtNode target = - FusionDictionary.findWordInTree(dict.mRootNodeArray, bigram.mWord); - final int addressOfBigram = target.mCachedAddressAfterUpdate; - final int unigramFrequencyForThisWord = target.mFrequency; - final int offset = addressOfBigram - - (mTriePos + FormatSpec.PTNODE_ATTRIBUTE_FLAGS_SIZE); - int bigramFlags = BinaryDictEncoderUtils.makeBigramFlags(bigramIterator.hasNext(), - offset, bigram.mFrequency, unigramFrequencyForThisWord, bigram.mWord); - mTrieBuf[mTriePos++] = (byte) bigramFlags; - mTriePos += BinaryDictEncoderUtils.writeChildrenPosition(mTrieBuf, - mTriePos, Math.abs(offset)); + for (final PtNodeArray nodeArray : flatNodes) { + for (final PtNode ptNode : nodeArray.mData) { + if (ptNode.mBigrams != null) { + final int startPos = bigramBuffer.size(); + mBigramAddressTable.set(ptNode.mTerminalId, startPos); + final Iterator bigramIterator = ptNode.mBigrams.iterator(); + while (bigramIterator.hasNext()) { + final WeightedString bigram = bigramIterator.next(); + final PtNode target = + FusionDictionary.findWordInTree(dict.mRootNodeArray, bigram.mWord); + final int unigramFrequencyForThisWord = target.mFrequency; + final int bigramFlags = BinaryDictEncoderUtils.makeBigramFlags( + bigramIterator.hasNext(), 0, bigram.mFrequency, + unigramFrequencyForThisWord, bigram.mWord); + BinaryDictEncoderUtils.writeUIntToStream(bigramBuffer, bigramFlags, + FormatSpec.PTNODE_ATTRIBUTE_FLAGS_SIZE); + BinaryDictEncoderUtils.writeUIntToStream(bigramBuffer, target.mTerminalId, + FormatSpec.PTNODE_ATTRIBUTE_MAX_ADDRESS_SIZE); + } + } + } } + bigramBuffer.writeTo(mBigramOutStream); + } + + private void writeBigramAddressSparseTable() throws IOException { + final File lookupIndexFile = + new File(mDictDir, mBaseFilename + FormatSpec.BIGRAM_LOOKUP_TABLE_FILE_EXTENSION); + final File contentFile = + new File(mDictDir, mBaseFilename + FormatSpec.BIGRAM_ADDRESS_TABLE_FILE_EXTENSION); + mBigramAddressTable.writeToFiles(lookupIndexFile, contentFile); } @Override @@ -267,7 +300,6 @@ public class Ver4DictEncoder implements DictEncoder { } writeChildrenPosition(ptNode, formatOptions); writeShortcuts(ptNode.mShortcutTargets); - writeBigrams(ptNode.mBigrams, dict); } private void writeTerminalData(final ArrayList flatNodes,