am 70e81af7: Hide physical structure of the dictionary from BinaryDictEncoderUtils.

* commit '70e81af72678d95c2a38412c478e06837a51c7cb':
  Hide physical structure of the dictionary from BinaryDictEncoderUtils.
main
Yuichiro Hanada 2013-09-08 23:20:51 -07:00 committed by Android Git Automerger
commit 45bede32b9
3 changed files with 230 additions and 150 deletions

View File

@ -27,7 +27,6 @@ import java.io.ByteArrayOutputStream;
import java.io.IOException; import java.io.IOException;
import java.io.OutputStream; import java.io.OutputStream;
import java.util.ArrayList; import java.util.ArrayList;
import java.util.Iterator;
/** /**
* Encodes binary files for a FusionDictionary. * Encodes binary files for a FusionDictionary.
@ -432,7 +431,7 @@ public class BinaryDictEncoderUtils {
* @param formatOptions file format options. * @param formatOptions file format options.
* @return the same array it was passed. The nodes have been updated for address and size. * @return the same array it was passed. The nodes have been updated for address and size.
*/ */
private static ArrayList<PtNodeArray> computeAddresses(final FusionDictionary dict, /* package */ static ArrayList<PtNodeArray> computeAddresses(final FusionDictionary dict,
final ArrayList<PtNodeArray> flatNodes, final FormatOptions formatOptions) { final ArrayList<PtNodeArray> flatNodes, final FormatOptions formatOptions) {
// First get the worst possible sizes and offsets // First get the worst possible sizes and offsets
for (final PtNodeArray n : flatNodes) calculatePtNodeArrayMaximumSize(n, formatOptions); for (final PtNodeArray n : flatNodes) calculatePtNodeArrayMaximumSize(n, formatOptions);
@ -484,7 +483,7 @@ public class BinaryDictEncoderUtils {
* *
* @param arrays the list of node arrays to check * @param arrays the list of node arrays to check
*/ */
private static void checkFlatPtNodeArrayList(final ArrayList<PtNodeArray> arrays) { /* package */ static void checkFlatPtNodeArrayList(final ArrayList<PtNodeArray> arrays) {
int offset = 0; int offset = 0;
int index = 0; int index = 0;
for (final PtNodeArray ptNodeArray : arrays) { for (final PtNodeArray ptNodeArray : arrays) {
@ -508,7 +507,8 @@ public class BinaryDictEncoderUtils {
* @param position the position to write. * @param position the position to write.
* @return the size in bytes the address actually took. * @return the size in bytes the address actually took.
*/ */
private static int writeChildrenPosition(final byte[] buffer, int index, final int position) { /* package */ static int writeChildrenPosition(final byte[] buffer, int index,
final int position) {
switch (getByteSize(position)) { switch (getByteSize(position)) {
case 1: case 1:
buffer[index++] = (byte)position; buffer[index++] = (byte)position;
@ -537,7 +537,7 @@ public class BinaryDictEncoderUtils {
* @param position the position to write. * @param position the position to write.
* @return the size in bytes the address actually took. * @return the size in bytes the address actually took.
*/ */
private static int writeSignedChildrenPosition(final byte[] buffer, int index, /* package */ static int writeSignedChildrenPosition(final byte[] buffer, int index,
final int position) { final int position) {
if (!BinaryDictIOUtils.hasChildrenAddress(position)) { if (!BinaryDictIOUtils.hasChildrenAddress(position)) {
buffer[index] = buffer[index + 1] = buffer[index + 2] = 0; buffer[index] = buffer[index + 1] = buffer[index + 2] = 0;
@ -598,7 +598,7 @@ public class BinaryDictEncoderUtils {
return flags; return flags;
} }
private static byte makePtNodeFlags(final PtNode node, final int ptNodeAddress, /* package */ static byte makePtNodeFlags(final PtNode node, final int ptNodeAddress,
final int childrenOffset, final FormatOptions formatOptions) { final int childrenOffset, final FormatOptions formatOptions) {
return (byte) makePtNodeFlags(node.mChars.length > 1, node.mFrequency >= 0, return (byte) makePtNodeFlags(node.mChars.length > 1, node.mFrequency >= 0,
getByteSize(childrenOffset), getByteSize(childrenOffset),
@ -616,7 +616,7 @@ public class BinaryDictEncoderUtils {
* @param word the second bigram, for debugging purposes * @param word the second bigram, for debugging purposes
* @return the flags * @return the flags
*/ */
private static final int makeBigramFlags(final boolean more, final int offset, /* package */ static final int makeBigramFlags(final boolean more, final int offset,
int bigramFrequency, final int unigramFrequency, final String word) { int bigramFrequency, final int unigramFrequency, final String word) {
int bigramFlags = (more ? FormatSpec.FLAG_BIGRAM_SHORTCUT_ATTR_HAS_NEXT : 0) int bigramFlags = (more ? FormatSpec.FLAG_BIGRAM_SHORTCUT_ATTR_HAS_NEXT : 0)
+ (offset < 0 ? FormatSpec.FLAG_BIGRAM_ATTR_OFFSET_NEGATIVE : 0); + (offset < 0 ? FormatSpec.FLAG_BIGRAM_ATTR_OFFSET_NEGATIVE : 0);
@ -702,7 +702,7 @@ public class BinaryDictEncoderUtils {
+ (frequency & FormatSpec.FLAG_BIGRAM_SHORTCUT_ATTR_FREQUENCY); + (frequency & FormatSpec.FLAG_BIGRAM_SHORTCUT_ATTR_FREQUENCY);
} }
private static final int writeParentAddress(final byte[] buffer, final int index, /* package */ static final int writeParentAddress(final byte[] buffer, final int index,
final int address, final FormatOptions formatOptions) { final int address, final FormatOptions formatOptions) {
if (BinaryDictIOUtils.supportsDynamicUpdate(formatOptions)) { if (BinaryDictIOUtils.supportsDynamicUpdate(formatOptions)) {
if (address == FormatSpec.NO_PARENT_ADDRESS) { if (address == FormatSpec.NO_PARENT_ADDRESS) {
@ -721,7 +721,7 @@ public class BinaryDictEncoderUtils {
} }
} }
private static final int getChildrenPosition(final PtNode ptNode, /* package */ static final int getChildrenPosition(final PtNode ptNode,
final FormatOptions formatOptions) { final FormatOptions formatOptions) {
int positionOfChildrenPosField = ptNode.mCachedAddressAfterUpdate int positionOfChildrenPosField = ptNode.mCachedAddressAfterUpdate
+ getNodeHeaderSize(ptNode, formatOptions); + getNodeHeaderSize(ptNode, formatOptions);
@ -736,35 +736,29 @@ public class BinaryDictEncoderUtils {
* Write a PtNodeArray to memory. The PtNodeArray is expected to have its final position cached. * Write a PtNodeArray to memory. The PtNodeArray is expected to have its final position cached.
* *
* @param dict the dictionary the node array is a part of (for relative offsets). * @param dict the dictionary the node array is a part of (for relative offsets).
* @param buffer the memory buffer to write to. * @param dictEncoder the dictionary encoder.
* @param ptNodeArray the node array to write. * @param ptNodeArray the node array to write.
* @param formatOptions file format options. * @param formatOptions file format options.
* @return the address of the END of the node.
*/ */
@SuppressWarnings("unused") @SuppressWarnings("unused")
private static int writePlacedNode(final FusionDictionary dict, byte[] buffer, /* package */ static void writePlacedNode(final FusionDictionary dict,
final PtNodeArray ptNodeArray, final FormatOptions formatOptions) { final DictEncoder dictEncoder, final PtNodeArray ptNodeArray,
final FormatOptions formatOptions) {
// TODO: Make the code in common with BinaryDictIOUtils#writePtNode // TODO: Make the code in common with BinaryDictIOUtils#writePtNode
int index = ptNodeArray.mCachedAddressAfterUpdate; dictEncoder.setPosition(ptNodeArray.mCachedAddressAfterUpdate);
final int ptNodeCount = ptNodeArray.mData.size(); final int ptNodeCount = ptNodeArray.mData.size();
final int countSize = getPtNodeCountSize(ptNodeArray); dictEncoder.writePtNodeCount(ptNodeCount);
final int parentAddress = ptNodeArray.mCachedParentAddress; final int parentPosition =
if (1 == countSize) { (ptNodeArray.mCachedParentAddress == FormatSpec.NO_PARENT_ADDRESS)
buffer[index++] = (byte)ptNodeCount; ? FormatSpec.NO_PARENT_ADDRESS
} else if (2 == countSize) { : ptNodeArray.mCachedParentAddress + ptNodeArray.mCachedAddressAfterUpdate;
// We need to signal 2-byte size by setting the top bit of the MSB to 1, so
// we | 0x80 to do this.
buffer[index++] = (byte)((ptNodeCount >> 8) | 0x80);
buffer[index++] = (byte)(ptNodeCount & 0xFF);
} else {
throw new RuntimeException("Strange size from getGroupCountSize : " + countSize);
}
for (int i = 0; i < ptNodeCount; ++i) { for (int i = 0; i < ptNodeCount; ++i) {
final PtNode ptNode = ptNodeArray.mData.get(i); final PtNode ptNode = ptNodeArray.mData.get(i);
if (index != ptNode.mCachedAddressAfterUpdate) { if (dictEncoder.getPosition() != ptNode.mCachedAddressAfterUpdate) {
throw new RuntimeException("Bug: write index is not the same as the cached address " throw new RuntimeException("Bug: write index is not the same as the cached address "
+ "of the node : " + index + " <> " + ptNode.mCachedAddressAfterUpdate); + "of the node : " + dictEncoder.getPosition() + " <> "
+ ptNode.mCachedAddressAfterUpdate);
} }
// Sanity checks. // Sanity checks.
if (DBG && ptNode.mFrequency > FormatSpec.MAX_TERMINAL_FREQUENCY) { if (DBG && ptNode.mFrequency > FormatSpec.MAX_TERMINAL_FREQUENCY) {
@ -773,85 +767,23 @@ public class BinaryDictEncoderUtils {
+ " : " + ptNode.mFrequency); + " : " + ptNode.mFrequency);
} }
final int childrenPosition = getChildrenPosition(ptNode, formatOptions); dictEncoder.writePtNodeFlags(ptNode, parentPosition, formatOptions);
buffer[index++] = makePtNodeFlags(ptNode, index, childrenPosition, dictEncoder.writeParentPosition(parentPosition, ptNode, formatOptions);
formatOptions); dictEncoder.writeCharacters(ptNode.mChars, ptNode.hasSeveralChars());
dictEncoder.writeFrequency(ptNode.mFrequency);
if (parentAddress == FormatSpec.NO_PARENT_ADDRESS) { dictEncoder.writeChildrenPosition(ptNode, formatOptions);
index = writeParentAddress(buffer, index, parentAddress, formatOptions); dictEncoder.writeShortcuts(ptNode.mShortcutTargets);
} else { dictEncoder.writeBigrams(ptNode.mBigrams, dict);
index = writeParentAddress(buffer, index, parentAddress
+ (ptNodeArray.mCachedAddressAfterUpdate
- ptNode.mCachedAddressAfterUpdate),
formatOptions);
}
index = CharEncoding.writeCharArray(ptNode.mChars, buffer, index);
if (ptNode.hasSeveralChars()) {
buffer[index++] = FormatSpec.PTNODE_CHARACTERS_TERMINATOR;
}
if (ptNode.mFrequency >= 0) {
buffer[index++] = (byte) ptNode.mFrequency;
}
if (formatOptions.mSupportsDynamicUpdate) {
index += writeSignedChildrenPosition(buffer, index, childrenPosition);
} else {
index += writeChildrenPosition(buffer, index, childrenPosition);
}
// Write shortcuts
if (null != ptNode.mShortcutTargets && !ptNode.mShortcutTargets.isEmpty()) {
final int indexOfShortcutByteSize = index;
index += FormatSpec.PTNODE_SHORTCUT_LIST_SIZE_SIZE;
final Iterator<WeightedString> shortcutIterator =
ptNode.mShortcutTargets.iterator();
while (shortcutIterator.hasNext()) {
final WeightedString target = shortcutIterator.next();
int shortcutFlags = makeShortcutFlags(shortcutIterator.hasNext(),
target.mFrequency);
buffer[index++] = (byte)shortcutFlags;
final int shortcutShift = CharEncoding.writeString(buffer, index, target.mWord);
index += shortcutShift;
}
final int shortcutByteSize = index - indexOfShortcutByteSize;
if (shortcutByteSize > 0xFFFF) {
throw new RuntimeException("Shortcut list too large");
}
buffer[indexOfShortcutByteSize] = (byte)(shortcutByteSize >> 8);
buffer[indexOfShortcutByteSize + 1] = (byte)(shortcutByteSize & 0xFF);
}
// Write bigrams
if (null != ptNode.mBigrams) {
final Iterator<WeightedString> bigramIterator = ptNode.mBigrams.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
- (index + FormatSpec.PTNODE_ATTRIBUTE_FLAGS_SIZE);
int bigramFlags = makeBigramFlags(bigramIterator.hasNext(), offset,
bigram.mFrequency, unigramFrequencyForThisWord, bigram.mWord);
buffer[index++] = (byte)bigramFlags;
final int bigramShift = writeChildrenPosition(buffer, index, Math.abs(offset));
index += bigramShift;
}
}
} }
if (formatOptions.mSupportsDynamicUpdate) { if (formatOptions.mSupportsDynamicUpdate) {
buffer[index] = buffer[index + 1] = buffer[index + 2] dictEncoder.writeForwardLinkAddress(FormatSpec.NO_FORWARD_LINK_ADDRESS);
= FormatSpec.NO_FORWARD_LINK_ADDRESS;
index += FormatSpec.FORWARD_LINK_ADDRESS_SIZE;
} }
if (index != ptNodeArray.mCachedAddressAfterUpdate + ptNodeArray.mCachedSize) { if (dictEncoder.getPosition() != ptNodeArray.mCachedAddressAfterUpdate
throw new RuntimeException( + ptNodeArray.mCachedSize) {
"Not the same size : written " + (index - ptNodeArray.mCachedAddressAfterUpdate) throw new RuntimeException("Not the same size : written "
+ (dictEncoder.getPosition() - ptNodeArray.mCachedAddressAfterUpdate)
+ " bytes from a node that should have " + ptNodeArray.mCachedSize + " bytes"); + " bytes from a node that should have " + ptNodeArray.mCachedSize + " bytes");
} }
return index;
} }
/** /**
@ -862,7 +794,7 @@ public class BinaryDictEncoderUtils {
* *
* @param ptNodeArrays the list of PtNode arrays. * @param ptNodeArrays the list of PtNode arrays.
*/ */
private static void showStatistics(ArrayList<PtNodeArray> ptNodeArrays) { /* package */ static void showStatistics(ArrayList<PtNodeArray> ptNodeArrays) {
int firstTerminalAddress = Integer.MAX_VALUE; int firstTerminalAddress = Integer.MAX_VALUE;
int lastTerminalAddress = Integer.MIN_VALUE; int lastTerminalAddress = Integer.MIN_VALUE;
int size = 0; int size = 0;
@ -968,49 +900,4 @@ public class BinaryDictEncoderUtils {
headerBuffer.close(); headerBuffer.close();
} }
/**
* Dumps a FusionDictionary to a file.
*
* @param destination the stream to write the dictionary body to.
* @param dict the dictionary to write.
* @param formatOptions file format options.
*/
/* package */ static void writeDictionaryBody(final OutputStream destination,
final FusionDictionary dict, final FormatOptions formatOptions) throws IOException {
// Addresses are limited to 3 bytes, but since addresses can be relative to each node
// array, the structure itself is not limited to 16MB. However, if it is over 16MB deciding
// the order of the PtNode arrays becomes a quite complicated problem, because though the
// dictionary itself does not have a size limit, each node array must still be within 16MB
// of all its children and parents. As long as this is ensured, the dictionary file may
// grow to any size.
// Leave the choice of the optimal node order to the flattenTree function.
MakedictLog.i("Flattening the tree...");
ArrayList<PtNodeArray> flatNodes = flattenTree(dict.mRootNodeArray);
MakedictLog.i("Computing addresses...");
computeAddresses(dict, flatNodes, formatOptions);
MakedictLog.i("Checking PtNode array...");
if (DBG) checkFlatPtNodeArrayList(flatNodes);
// Create a buffer that matches the final dictionary size.
final PtNodeArray lastNodeArray = flatNodes.get(flatNodes.size() - 1);
final int bufferSize = lastNodeArray.mCachedAddressAfterUpdate + lastNodeArray.mCachedSize;
final byte[] buffer = new byte[bufferSize];
MakedictLog.i("Writing file...");
int dataEndOffset = 0;
for (PtNodeArray nodeArray : flatNodes) {
dataEndOffset = writePlacedNode(dict, buffer, nodeArray, formatOptions);
}
if (DBG) showStatistics(flatNodes);
destination.write(buffer, 0, dataEndOffset);
destination.close();
MakedictLog.i("Done");
}
} }

View File

@ -17,8 +17,11 @@
package com.android.inputmethod.latin.makedict; package com.android.inputmethod.latin.makedict;
import com.android.inputmethod.latin.makedict.FormatSpec.FormatOptions; 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 java.io.IOException; import java.io.IOException;
import java.util.ArrayList;
/** /**
* An interface of binary dictionary encoder. * An interface of binary dictionary encoder.
@ -26,4 +29,32 @@ import java.io.IOException;
public interface DictEncoder { public interface DictEncoder {
public void writeDictionary(final FusionDictionary dict, final FormatOptions formatOptions) public void writeDictionary(final FusionDictionary dict, final FormatOptions formatOptions)
throws IOException, UnsupportedFormatException; throws IOException, UnsupportedFormatException;
public void setPosition(final int position);
public int getPosition();
public void writePtNodeCount(final int ptNodeCount);
public void writePtNodeFlags(final PtNode ptNode, final int parentAddress,
final FormatOptions formatOptions);
public void writeParentPosition(final int parentPosition, final PtNode ptNode,
final FormatOptions formatOptions);
public void writeCharacters(final int[] characters, final boolean hasSeveralChars);
public void writeFrequency(final int frequency);
public void writeChildrenPosition(final PtNode ptNode, final FormatOptions formatOptions);
/**
* Write a shortcut attributes list to memory.
*
* @param shortcuts the shortcut attributes list.
*/
public void writeShortcuts(final ArrayList<WeightedString> shortcuts);
/**
* Write a bigram attributes list to memory.
*
* @param bigrams the bigram attributes list.
* @param dict the dictionary the node array is a part of (for relative offsets).
*/
public void writeBigrams(final ArrayList<WeightedString> bigrams, final FusionDictionary dict);
public void writeForwardLinkAddress(final int forwardLinkAddress);
} }

View File

@ -16,13 +16,19 @@
package com.android.inputmethod.latin.makedict; package com.android.inputmethod.latin.makedict;
import com.android.inputmethod.latin.makedict.BinaryDictDecoderUtils.CharEncoding;
import com.android.inputmethod.latin.makedict.FormatSpec.FormatOptions; import com.android.inputmethod.latin.makedict.FormatSpec.FormatOptions;
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.File; import java.io.File;
import java.io.FileNotFoundException; import java.io.FileNotFoundException;
import java.io.FileOutputStream; import java.io.FileOutputStream;
import java.io.IOException; import java.io.IOException;
import java.io.OutputStream; import java.io.OutputStream;
import java.util.ArrayList;
import java.util.Iterator;
/** /**
* An implementation of DictEncoder for version 3 binary dictionary. * An implementation of DictEncoder for version 3 binary dictionary.
@ -31,10 +37,13 @@ public class Ver3DictEncoder implements DictEncoder {
private final File mDictFile; private final File mDictFile;
private OutputStream mOutStream; private OutputStream mOutStream;
private byte[] mBuffer;
private int mPosition;
public Ver3DictEncoder(final File dictFile) { public Ver3DictEncoder(final File dictFile) {
mDictFile = dictFile; mDictFile = dictFile;
mOutStream = null; mOutStream = null;
mBuffer = null;
} }
// This constructor is used only by BinaryDictOffdeviceUtilsTests. // This constructor is used only by BinaryDictOffdeviceUtilsTests.
@ -57,13 +66,166 @@ public class Ver3DictEncoder implements DictEncoder {
} }
@Override @Override
public void writeDictionary(FusionDictionary dict, FormatOptions formatOptions) public void writeDictionary(final FusionDictionary dict, final FormatOptions formatOptions)
throws IOException, UnsupportedFormatException { throws IOException, UnsupportedFormatException {
if (mOutStream == null) { if (mOutStream == null) {
openStream(); openStream();
} }
BinaryDictEncoderUtils.writeDictionaryHeader(mOutStream, dict, formatOptions); BinaryDictEncoderUtils.writeDictionaryHeader(mOutStream, dict, formatOptions);
BinaryDictEncoderUtils.writeDictionaryBody(mOutStream, dict, formatOptions);
// Addresses are limited to 3 bytes, but since addresses can be relative to each node
// array, the structure itself is not limited to 16MB. However, if it is over 16MB deciding
// the order of the PtNode arrays becomes a quite complicated problem, because though the
// dictionary itself does not have a size limit, each node array must still be within 16MB
// of all its children and parents. As long as this is ensured, the dictionary file may
// grow to any size.
// Leave the choice of the optimal node order to the flattenTree function.
MakedictLog.i("Flattening the tree...");
ArrayList<PtNodeArray> flatNodes = BinaryDictEncoderUtils.flattenTree(dict.mRootNodeArray);
MakedictLog.i("Computing addresses...");
BinaryDictEncoderUtils.computeAddresses(dict, flatNodes, formatOptions);
MakedictLog.i("Checking PtNode array...");
if (MakedictLog.DBG) BinaryDictEncoderUtils.checkFlatPtNodeArrayList(flatNodes);
// Create a buffer that matches the final dictionary size.
final PtNodeArray lastNodeArray = flatNodes.get(flatNodes.size() - 1);
final int bufferSize = lastNodeArray.mCachedAddressAfterUpdate + lastNodeArray.mCachedSize;
mBuffer = new byte[bufferSize];
MakedictLog.i("Writing file...");
for (PtNodeArray nodeArray : flatNodes) {
BinaryDictEncoderUtils.writePlacedNode(dict, this, nodeArray, formatOptions);
}
if (MakedictLog.DBG) BinaryDictEncoderUtils.showStatistics(flatNodes);
mOutStream.write(mBuffer, 0, mPosition);
MakedictLog.i("Done");
close(); close();
} }
@Override
public void setPosition(final int position) {
if (mBuffer == null || position < 0 || position >= mBuffer.length) return;
mPosition = position;
}
@Override
public int getPosition() {
return mPosition;
}
@Override
public void writePtNodeCount(final int ptNodeCount) {
final int countSize = BinaryDictIOUtils.getPtNodeCountSize(ptNodeCount);
if (1 == countSize) {
mBuffer[mPosition++] = (byte) ptNodeCount;
} else if (2 == countSize) {
mBuffer[mPosition++] = (byte) ((ptNodeCount >> 8) & 0xFF);
mBuffer[mPosition++] = (byte) (ptNodeCount & 0xFF);
} else {
throw new RuntimeException("Strange size from getGroupCountSize : " + countSize);
}
}
@Override
public void writePtNodeFlags(final PtNode ptNode, final int parentAddress,
final FormatOptions formatOptions) {
final int childrenPos = BinaryDictEncoderUtils.getChildrenPosition(ptNode, formatOptions);
mBuffer[mPosition++] = BinaryDictEncoderUtils.makePtNodeFlags(ptNode, mPosition,
childrenPos, formatOptions);
}
@Override
public void writeParentPosition(final int parentPosition, final PtNode ptNode,
final FormatOptions formatOptions) {
if (parentPosition == FormatSpec.NO_PARENT_ADDRESS) {
mPosition = BinaryDictEncoderUtils.writeParentAddress(mBuffer, mPosition,
parentPosition, formatOptions);
} else {
mPosition = BinaryDictEncoderUtils.writeParentAddress(mBuffer, mPosition,
parentPosition - ptNode.mCachedAddressAfterUpdate, formatOptions);
}
}
@Override
public void writeCharacters(final int[] codePoints, final boolean hasSeveralChars) {
mPosition = CharEncoding.writeCharArray(codePoints, mBuffer, mPosition);
if (hasSeveralChars) {
mBuffer[mPosition++] = FormatSpec.PTNODE_CHARACTERS_TERMINATOR;
}
}
@Override
public void writeFrequency(final int frequency) {
if (frequency >= 0) {
mBuffer[mPosition++] = (byte) frequency;
}
}
@Override
public void writeChildrenPosition(final PtNode ptNode, final FormatOptions formatOptions) {
final int childrenPos = BinaryDictEncoderUtils.getChildrenPosition(ptNode, formatOptions);
if (formatOptions.mSupportsDynamicUpdate) {
mPosition += BinaryDictEncoderUtils.writeSignedChildrenPosition(mBuffer, mPosition,
childrenPos);
} else {
mPosition += BinaryDictEncoderUtils.writeChildrenPosition(mBuffer, mPosition,
childrenPos);
}
}
@Override
public void writeShortcuts(final ArrayList<WeightedString> shortcuts) {
if (null == shortcuts || shortcuts.isEmpty()) return;
final int indexOfShortcutByteSize = mPosition;
mPosition += FormatSpec.PTNODE_SHORTCUT_LIST_SIZE_SIZE;
final Iterator<WeightedString> shortcutIterator = shortcuts.iterator();
while (shortcutIterator.hasNext()) {
final WeightedString target = shortcutIterator.next();
final int shortcutFlags = BinaryDictEncoderUtils.makeShortcutFlags(
shortcutIterator.hasNext(),
target.mFrequency);
mBuffer[mPosition++] = (byte)shortcutFlags;
final int shortcutShift = CharEncoding.writeString(mBuffer, mPosition, target.mWord);
mPosition += shortcutShift;
}
final int shortcutByteSize = mPosition - indexOfShortcutByteSize;
if (shortcutByteSize > 0xFFFF) {
throw new RuntimeException("Shortcut list too large");
}
mBuffer[indexOfShortcutByteSize] = (byte)((shortcutByteSize >> 8) & 0xFF);
mBuffer[indexOfShortcutByteSize + 1] = (byte)(shortcutByteSize & 0xFF);
}
@Override
public void writeBigrams(final ArrayList<WeightedString> bigrams, final FusionDictionary dict) {
if (bigrams == null) return;
final Iterator<WeightedString> 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
- (mPosition + FormatSpec.PTNODE_ATTRIBUTE_FLAGS_SIZE);
int bigramFlags = BinaryDictEncoderUtils.makeBigramFlags(bigramIterator.hasNext(),
offset, bigram.mFrequency, unigramFrequencyForThisWord, bigram.mWord);
mBuffer[mPosition++] = (byte) bigramFlags;
mPosition += BinaryDictEncoderUtils.writeChildrenPosition(mBuffer, mPosition,
Math.abs(offset));
}
}
@Override
public void writeForwardLinkAddress(final int forwardLinkAddress) {
mBuffer[mPosition++] = (byte) ((forwardLinkAddress >> 16) & 0xFF);
mBuffer[mPosition++] = (byte) ((forwardLinkAddress >> 8) & 0xFF);
mBuffer[mPosition++] = (byte) (forwardLinkAddress & 0xFF);
}
} }