Merge "Add the new dictionary maker to the build."

main
Jean Chalard 2011-08-10 22:49:31 -07:00 committed by Android (Google) Code Review
commit 6f9f759ec5
14 changed files with 2472 additions and 0 deletions

View File

@ -0,0 +1,27 @@
#
# Copyright (C) 2011 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.
#
LOCAL_PATH := $(call my-dir)
include $(CLEAR_VARS)
LOCAL_SRC_FILES := $(call all-java-files-under,src)
LOCAL_SRC_FILES += $(call all-java-files-under,tests)
LOCAL_JAR_MANIFEST := etc/manifest.txt
LOCAL_MODULE_TAGS := eng
LOCAL_MODULE := makedict2
LOCAL_JAVA_LIBRARIES := junit
include $(BUILD_HOST_JAVA_LIBRARY)
include $(LOCAL_PATH)/etc/Android.mk

View File

@ -0,0 +1,21 @@
# Copyright (C) 2011 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.
LOCAL_PATH := $(call my-dir)
include $(CLEAR_VARS)
LOCAL_MODULE_TAGS := eng
LOCAL_PREBUILT_EXECUTABLES := makedict2
include $(BUILD_HOST_PREBUILT)

63
tools/makedict2/etc/makedict2 Executable file
View File

@ -0,0 +1,63 @@
#!/bin/sh
# Copyright 2011, 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.
# Set up prog to be the path of this script, including following symlinks,
# and set up progdir to be the fully-qualified pathname of its directory.
prog="$0"
while [ -h "${prog}" ]; do
newProg=`/bin/ls -ld "${prog}"`
newProg=`expr "${newProg}" : ".* -> \(.*\)$"`
if expr "x${newProg}" : 'x/' >/dev/null; then
prog="${newProg}"
else
progdir=`dirname "${prog}"`
prog="${progdir}/${newProg}"
fi
done
oldwd=`pwd`
progdir=`dirname "${prog}"`
cd "${progdir}"
progdir=`pwd`
prog="${progdir}"/`basename "${prog}"`
cd "${oldwd}"
jarfile=makedict2.jar
frameworkdir="$progdir"
if [ ! -r "$frameworkdir/$jarfile" ]
then
frameworkdir=`dirname "$progdir"`/tools/lib
libdir=`dirname "$progdir"`/tools/lib
fi
if [ ! -r "$frameworkdir/$jarfile" ]
then
frameworkdir=`dirname "$progdir"`/framework
libdir=`dirname "$progdir"`/lib
fi
if [ ! -r "$frameworkdir/$jarfile" ]
then
echo `basename "$prog"`": can't find $jarfile"
exit 1
fi
if [ "$OSTYPE" = "cygwin" ] ; then
jarpath=`cygpath -w "$frameworkdir/$jarfile"`
progdir=`cygpath -w "$progdir"`
else
jarpath="$frameworkdir/$jarfile"
fi
# need to use "java.ext.dirs" because "-jar" causes classpath to be ignored
# might need more memory, e.g. -Xmx128M
exec java -ea -Djava.ext.dirs="$frameworkdir" -jar "$jarpath" "$@"

View File

@ -0,0 +1 @@
Main-Class: com.android.inputmethod.latin.DictionaryMaker

File diff suppressed because it is too large Load Diff

View File

@ -0,0 +1,45 @@
/*
* Copyright (C) 2011 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;
import java.util.ArrayList;
/**
* Raw char group info straight out of a file. This will contain numbers for addresses.
*/
public class CharGroupInfo {
public final int mOriginalAddress;
public final int mEndAddress;
public final int mFlags;
public final int[] mCharacters;
public final int mFrequency;
public final int mChildrenAddress;
public final ArrayList<PendingAttribute> mBigrams;
public CharGroupInfo(final int originalAddress, final int endAddress, final int flags,
final int[] characters, final int frequency, final int childrenAddress,
final ArrayList<PendingAttribute> bigrams) {
mOriginalAddress = originalAddress;
mEndAddress = endAddress;
mFlags = flags;
mCharacters = characters;
mFrequency = frequency;
mChildrenAddress = childrenAddress;
mBigrams = bigrams;
}
}

View File

@ -0,0 +1,256 @@
/*
* Copyright (C) 2011 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;
import java.io.File;
import java.io.FileInputStream;
import java.io.FileNotFoundException;
import java.io.FileOutputStream;
import java.io.FileWriter;
import java.io.IOException;
import java.io.RandomAccessFile;
import java.util.Arrays;
import java.util.LinkedList;
import javax.xml.parsers.ParserConfigurationException;
import org.xml.sax.SAXException;
/**
* Main class/method for DictionaryMaker.
*/
public class DictionaryMaker {
static class Arguments {
private final static String OPTION_VERSION_2 = "-2";
private final static String OPTION_INPUT_SOURCE = "-s";
private final static String OPTION_INPUT_BIGRAM_XML = "-b";
private final static String OPTION_OUTPUT_BINARY = "-d";
private final static String OPTION_OUTPUT_XML = "-x";
private final static String OPTION_HELP = "-h";
public final String mInputBinary;
public final String mInputUnigramXml;
public final String mInputBigramXml;
public final String mOutputBinary;
public final String mOutputXml;
private void checkIntegrity() {
checkHasExactlyOneInput();
checkHasAtLeastOneOutput();
}
private void checkHasExactlyOneInput() {
if (null == mInputUnigramXml && null == mInputBinary) {
throw new RuntimeException("No input file specified");
} else if (null != mInputUnigramXml && null != mInputBinary) {
throw new RuntimeException("Both input XML and binary specified");
} else if (null != mInputBinary && null != mInputBigramXml) {
throw new RuntimeException("Cannot specify a binary input and a separate bigram "
+ "file");
}
}
private void checkHasAtLeastOneOutput() {
if (null == mOutputBinary && null == mOutputXml) {
throw new RuntimeException("No output specified");
}
}
private void displayHelp() {
MakedictLog.i("Usage: makedict2 "
+ "[-s <unigrams.xml> [-b <bigrams.xml>] | -s <binary input>] "
+ " [-d <binary output>] [-x <xml output>] [-2]\n"
+ "\n"
+ " Converts a source dictionary file to one or several outputs.\n"
+ " Source can be an XML file, with an optional XML bigrams file, or a\n"
+ " binary dictionary file.\n"
+ " Both binary and XML outputs are supported. Both can be output at\n"
+ " the same time but outputting several files of the same type is not\n"
+ " supported.");
}
public Arguments(String[] argsArray) {
final LinkedList<String> args = new LinkedList<String>(Arrays.asList(argsArray));
if (args.isEmpty()) {
displayHelp();
}
String inputBinary = null;
String inputUnigramXml = null;
String inputBigramXml = null;
String outputBinary = null;
String outputXml = null;
while (!args.isEmpty()) {
final String arg = args.get(0);
args.remove(0);
if (arg.charAt(0) == '-') {
if (OPTION_VERSION_2.equals(arg)) {
// Do nothing, this is the default
} else if (OPTION_HELP.equals(arg)) {
displayHelp();
} else {
// All these options need an argument
if (args.isEmpty()) {
throw new RuntimeException("Option " + arg + " requires an argument");
}
String filename = args.get(0);
args.remove(0);
if (OPTION_INPUT_SOURCE.equals(arg)) {
if (BinaryDictInputOutput.isBinaryDictionary(filename)) {
inputBinary = filename;
} else {
inputUnigramXml = filename;
}
} else if (OPTION_INPUT_BIGRAM_XML.equals(arg)) {
inputBigramXml = filename;
} else if (OPTION_OUTPUT_BINARY.equals(arg)) {
outputBinary = filename;
} else if (OPTION_OUTPUT_XML.equals(arg)) {
outputXml = filename;
}
}
} else {
if (null == inputBinary && null == inputUnigramXml) {
if (BinaryDictInputOutput.isBinaryDictionary(arg)) {
inputBinary = arg;
} else {
inputUnigramXml = arg;
}
} else if (null == outputBinary) {
outputBinary = arg;
} else {
throw new RuntimeException("Several output binary files specified");
}
}
}
mInputBinary = inputBinary;
mInputUnigramXml = inputUnigramXml;
mInputBigramXml = inputBigramXml;
mOutputBinary = outputBinary;
mOutputXml = outputXml;
checkIntegrity();
}
}
public static void main(String[] args)
throws FileNotFoundException, ParserConfigurationException, SAXException, IOException,
UnsupportedFormatException {
final Arguments parsedArgs = new Arguments(args);
FusionDictionary dictionary = readInputFromParsedArgs(parsedArgs);
writeOutputToParsedArgs(parsedArgs, dictionary);
}
/**
* Invoke the right input method according to args.
*
* @param args the parsed command line arguments.
* @return the read dictionary.
*/
private static FusionDictionary readInputFromParsedArgs(final Arguments args)
throws IOException, UnsupportedFormatException, ParserConfigurationException,
SAXException, FileNotFoundException {
if (null != args.mInputBinary) {
return readBinaryFile(args.mInputBinary);
} else if (null != args.mInputUnigramXml) {
return readXmlFile(args.mInputUnigramXml, args.mInputBigramXml);
} else {
throw new RuntimeException("No input file specified");
}
}
/**
* Read a dictionary from the name of a binary file.
*
* @param binaryFilename the name of the file in the binary dictionary format.
* @return the read dictionary.
* @throws FileNotFoundException if the file can't be found
* @throws IOException if the input file can't be read
* @throws UnsupportedFormatException if the binary file is not in the expected format
*/
private static FusionDictionary readBinaryFile(final String binaryFilename)
throws FileNotFoundException, IOException, UnsupportedFormatException {
final RandomAccessFile inputFile = new RandomAccessFile(binaryFilename, "r");
return BinaryDictInputOutput.readDictionaryBinary(inputFile, null);
}
/**
* Read a dictionary from a unigram XML file, and optionally a bigram XML file.
*
* @param unigramXmlFilename the name of the unigram XML file. May not be null.
* @param bigramXmlFilename the name of the bigram XML file. Pass null if there are no bigrams.
* @return the read dictionary.
* @throws FileNotFoundException if one of the files can't be found
* @throws SAXException if one or more of the XML files is not well-formed
* @throws IOException if one the input files can't be read
* @throws ParserConfigurationException if the system can't create a SAX parser
*/
private static FusionDictionary readXmlFile(final String unigramXmlFilename,
final String bigramXmlFilename) throws FileNotFoundException, SAXException,
IOException, ParserConfigurationException {
final FileInputStream unigrams = new FileInputStream(new File(unigramXmlFilename));
final FileInputStream bigrams = null == bigramXmlFilename ? null :
new FileInputStream(new File(bigramXmlFilename));
return XmlDictInputOutput.readDictionaryXml(unigrams, bigrams);
}
/**
* Invoke the right output method according to args.
*
* This will write the passed dictionary to the file(s) passed in the command line arguments.
* @param args the parsed arguments.
* @param dict the file to output.
* @throws FileNotFoundException if one of the output files can't be created.
* @throws IOException if one of the output files can't be written to.
*/
private static void writeOutputToParsedArgs(final Arguments args, final FusionDictionary dict)
throws FileNotFoundException, IOException {
if (null != args.mOutputBinary) {
writeBinaryDictionary(args.mOutputBinary, dict);
}
if (null != args.mOutputXml) {
writeXmlDictionary(args.mOutputXml, dict);
}
}
/**
* Write the dictionary in binary format to the specified filename.
*
* @param outputFilename the name of the file to write to.
* @param dict the dictionary to write.
* @throws FileNotFoundException if the output file can't be created.
* @throws IOException if the output file can't be written to.
*/
private static void writeBinaryDictionary(final String outputFilename,
final FusionDictionary dict) throws FileNotFoundException, IOException {
final File outputFile = new File(outputFilename);
BinaryDictInputOutput.writeDictionaryBinary(new FileOutputStream(outputFilename), dict);
}
/**
* Write the dictionary in XML format to the specified filename.
*
* @param outputFilename the name of the file to write to.
* @param dict the dictionary to write.
* @throws FileNotFoundException if the output file can't be created.
* @throws IOException if the output file can't be written to.
*/
private static void writeXmlDictionary(final String outputFilename,
final FusionDictionary dict) throws FileNotFoundException, IOException {
XmlDictInputOutput.writeDictionaryXml(new FileWriter(outputFilename), dict);
}
}

View File

@ -0,0 +1,602 @@
/*
* Copyright (C) 2011 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;
import java.util.ArrayList;
import java.util.Arrays;
import java.util.Collections;
import java.util.Iterator;
import java.util.LinkedList;
import java.util.List;
/**
* A dictionary that can fusion heads and tails of words for more compression.
*/
public class FusionDictionary implements Iterable<Word> {
/**
* A node of the dictionary, containing several CharGroups.
*
* A node is but an ordered array of CharGroups, which essentially contain all the
* real information.
* This class also contains fields to cache size and address, to help with binary
* generation.
*/
public static class Node {
ArrayList<CharGroup> mData;
// To help with binary generation
int mCachedSize;
int mCachedAddress;
public Node() {
mData = new ArrayList<CharGroup>();
mCachedSize = Integer.MIN_VALUE;
mCachedAddress = Integer.MIN_VALUE;
}
public Node(ArrayList<CharGroup> data) {
mData = data;
mCachedSize = Integer.MIN_VALUE;
mCachedAddress = Integer.MIN_VALUE;
}
}
/**
* A string with a frequency.
*
* This represents an "attribute", that is either a bigram or a shortcut.
*/
public static class WeightedString {
final String mWord;
final int mFrequency;
public WeightedString(String word, int frequency) {
mWord = word;
mFrequency = frequency;
}
}
/**
* A group of characters, with a frequency, shortcuts, bigrams, and children.
*
* This is the central class of the in-memory representation. A CharGroup is what can
* be seen as a traditional "trie node", except it can hold several characters at the
* same time. A CharGroup essentially represents one or several characters in the middle
* of the trie trie; as such, it can be a terminal, and it can have children.
* In this in-memory representation, whether the CharGroup is a terminal or not is represented
* in the frequency, where NOT_A_TERMINAL (= -1) means this is not a terminal and any other
* value is the frequency of this terminal. A terminal may have non-null shortcuts and/or
* bigrams, but a non-terminal may not. Moreover, children, if present, are null.
*/
public static class CharGroup {
public static final int NOT_A_TERMINAL = -1;
final int mChars[];
final ArrayList<WeightedString> mBigrams;
final int mFrequency; // NOT_A_TERMINAL == mFrequency indicates this is not a terminal.
Node mChildren;
// The two following members to help with binary generation
int mCachedSize;
int mCachedAddress;
public CharGroup(final int[] chars,
final ArrayList<WeightedString> bigrams, final int frequency) {
mChars = chars;
mFrequency = frequency;
mBigrams = bigrams;
mChildren = null;
}
public CharGroup(final int[] chars,
final ArrayList<WeightedString> bigrams, final int frequency, final Node children) {
mChars = chars;
mFrequency = frequency;
mBigrams = bigrams;
mChildren = children;
}
public void addChild(CharGroup n) {
if (null == mChildren) {
mChildren = new Node();
}
mChildren.mData.add(n);
}
public boolean isTerminal() {
return NOT_A_TERMINAL != mFrequency;
}
public boolean hasSeveralChars() {
assert(mChars.length > 0);
return 1 < mChars.length;
}
}
/**
* Options global to the dictionary.
*
* There are no options at the moment, so this class is empty.
*/
public static class DictionaryOptions {
}
public final DictionaryOptions mOptions;
public final Node mRoot;
public FusionDictionary() {
mOptions = new DictionaryOptions();
mRoot = new Node();
}
public FusionDictionary(final Node root, final DictionaryOptions options) {
mRoot = root;
mOptions = options;
}
/**
* Helper method to convert a String to an int array.
*/
static private int[] getCodePoints(String word) {
final int wordLength = word.length();
int[] array = new int[word.codePointCount(0, wordLength)];
for (int i = 0; i < wordLength; ++i) {
array[i] = word.codePointAt(i);
}
return array;
}
/**
* Helper method to add a word as a string.
*
* This method adds a word to the dictionary with the given frequency. Optional
* lists of bigrams and shortcuts can be passed here. For each word inside,
* they will be added to the dictionary as necessary.
*
* @param word the word to add.
* @param frequency the frequency of the word, in the range [0..255].
* @param bigrams a list of bigrams, or null.
*/
public void add(String word, int frequency, ArrayList<WeightedString> bigrams) {
if (null != bigrams) {
for (WeightedString bigram : bigrams) {
final CharGroup t = findWordInTree(mRoot, bigram.mWord);
if (null == t) {
add(getCodePoints(bigram.mWord), 0, null);
}
}
}
add(getCodePoints(word), frequency, bigrams);
}
/**
* Sanity check for a node.
*
* This method checks that all CharGroups in a node are ordered as expected.
* If they are, nothing happens. If they aren't, an exception is thrown.
*/
private void checkStack(Node node) {
ArrayList<CharGroup> stack = node.mData;
int lastValue = -1;
for (int i = 0; i < stack.size(); ++i) {
int currentValue = stack.get(i).mChars[0];
if (currentValue <= lastValue)
throw new RuntimeException("Invalid stack");
else
lastValue = currentValue;
}
}
/**
* Add a word to this dictionary.
*
* The bigrams, if any, have to be in the dictionary already. If they aren't,
* an exception is thrown.
*
* @param word the word, as an int array.
* @param frequency the frequency of the word, in the range [0..255].
* @param bigrams an optional list of bigrams for this word (null if none).
*/
private void add(int[] word, int frequency, ArrayList<WeightedString> bigrams) {
assert(frequency >= 0 && frequency <= 255);
Node currentNode = mRoot;
int charIndex = 0;
CharGroup currentGroup = null;
int differentCharIndex = 0; // Set by the loop to the index of the char that differs
int nodeIndex = findIndexOfChar(mRoot, word[charIndex]);
while (CHARACTER_NOT_FOUND != nodeIndex) {
currentGroup = currentNode.mData.get(nodeIndex);
differentCharIndex = compareArrays(currentGroup.mChars, word, charIndex) ;
if (ARRAYS_ARE_EQUAL != differentCharIndex
&& differentCharIndex < currentGroup.mChars.length) break;
if (null == currentGroup.mChildren) break;
charIndex += currentGroup.mChars.length;
if (charIndex >= word.length) break;
currentNode = currentGroup.mChildren;
nodeIndex = findIndexOfChar(currentNode, word[charIndex]);
}
if (-1 == nodeIndex) {
// No node at this point to accept the word. Create one.
final int insertionIndex = findInsertionIndex(currentNode, word[charIndex]);
final CharGroup newGroup = new CharGroup(
Arrays.copyOfRange(word, charIndex, word.length), bigrams, frequency);
currentNode.mData.add(insertionIndex, newGroup);
checkStack(currentNode);
} else {
// There is a word with a common prefix.
if (differentCharIndex == currentGroup.mChars.length) {
if (charIndex + differentCharIndex >= word.length) {
// The new word is a prefix of an existing word, but the node on which it
// should end already exists as is.
if (currentGroup.mFrequency > 0) {
throw new RuntimeException("Such a word already exists in the dictionary : "
+ new String(word, 0, word.length));
} else {
final CharGroup newNode = new CharGroup(currentGroup.mChars,
bigrams, frequency, currentGroup.mChildren);
currentNode.mData.set(nodeIndex, newNode);
checkStack(currentNode);
}
} else {
// The new word matches the full old word and extends past it.
// We only have to create a new node and add it to the end of this.
final CharGroup newNode = new CharGroup(
Arrays.copyOfRange(word, charIndex + differentCharIndex, word.length),
bigrams, frequency);
currentGroup.mChildren = new Node();
currentGroup.mChildren.mData.add(newNode);
}
} else {
if (0 == differentCharIndex) {
// Exact same word. Check the frequency is 0 or -1, and update.
if (0 != frequency) {
if (0 < currentGroup.mFrequency) {
throw new RuntimeException("This word already exists with frequency "
+ currentGroup.mFrequency + " : "
+ new String(word, 0, word.length));
}
final CharGroup newGroup = new CharGroup(word,
currentGroup.mBigrams, frequency);
currentNode.mData.set(nodeIndex, newGroup);
}
} else {
// Partial prefix match only. We have to replace the current node with a node
// containing the current prefix and create two new ones for the tails.
Node newChildren = new Node();
final CharGroup newOldWord = new CharGroup(
Arrays.copyOfRange(currentGroup.mChars, differentCharIndex,
currentGroup.mChars.length),
currentGroup.mBigrams, currentGroup.mFrequency, currentGroup.mChildren);
newChildren.mData.add(newOldWord);
final CharGroup newParent;
if (charIndex + differentCharIndex >= word.length) {
newParent = new CharGroup(
Arrays.copyOfRange(currentGroup.mChars, 0, differentCharIndex),
bigrams, frequency, newChildren);
} else {
newParent = new CharGroup(
Arrays.copyOfRange(currentGroup.mChars, 0, differentCharIndex),
null, -1, newChildren);
final CharGroup newWord = new CharGroup(
Arrays.copyOfRange(word, charIndex + differentCharIndex,
word.length), bigrams, frequency);
final int addIndex = word[charIndex + differentCharIndex]
> currentGroup.mChars[differentCharIndex] ? 1 : 0;
newChildren.mData.add(addIndex, newWord);
}
currentNode.mData.set(nodeIndex, newParent);
}
checkStack(currentNode);
}
}
}
/**
* Custom comparison of two int arrays taken to contain character codes.
*
* This method compares the two arrays passed as an argument in a lexicographic way,
* with an offset in the dst string.
* This method does NOT test for the first character. It is taken to be equal.
* I repeat: this method starts the comparison at 1 <> dstOffset + 1.
* The index where the strings differ is returned. ARRAYS_ARE_EQUAL = 0 is returned if the
* strings are equal. This works BECAUSE we don't look at the first character.
*
* @param src the left-hand side string of the comparison.
* @param dst the right-hand side string of the comparison.
* @param dstOffset the offset in the right-hand side string.
* @return the index at which the strings differ, or ARRAYS_ARE_EQUAL = 0 if they don't.
*/
private static int ARRAYS_ARE_EQUAL = 0;
private static int compareArrays(final int[] src, final int[] dst, int dstOffset) {
// We do NOT test the first char, because we come from a method that already
// tested it.
for (int i = 1; i < src.length; ++i) {
if (dstOffset + i >= dst.length) return i;
if (src[i] != dst[dstOffset + i]) return i;
}
if (dst.length > src.length) return src.length;
return ARRAYS_ARE_EQUAL;
}
/**
* Helper class that compares and sorts two chargroups according to their
* first element only. I repeat: ONLY the first element is considered, the rest
* is ignored.
* This comparator imposes orderings that are inconsistent with equals.
*/
static private class CharGroupComparator implements java.util.Comparator {
public int compare(Object o1, Object o2) {
final CharGroup c1 = (CharGroup)o1;
final CharGroup c2 = (CharGroup)o2;
if (c1.mChars[0] == c2.mChars[0]) return 0;
return c1.mChars[0] < c2.mChars[0] ? -1 : 1;
}
public boolean equals(Object o) {
return o instanceof CharGroupComparator;
}
}
final static private CharGroupComparator CHARGROUP_COMPARATOR = new CharGroupComparator();
/**
* Finds the insertion index of a character within a node.
*/
private static int findInsertionIndex(final Node node, int character) {
final List data = node.mData;
final CharGroup reference = new CharGroup(new int[] { character }, null, 0);
int result = Collections.binarySearch(data, reference, CHARGROUP_COMPARATOR);
return result >= 0 ? result : -result - 1;
}
/**
* Find the index of a char in a node, if it exists.
*
* @param node the node to search in.
* @param character the character to search for.
* @return the position of the character if it's there, or CHARACTER_NOT_FOUND = -1 else.
*/
private static int CHARACTER_NOT_FOUND = -1;
private static int findIndexOfChar(final Node node, int character) {
final int insertionIndex = findInsertionIndex(node, character);
if (node.mData.size() <= insertionIndex) return CHARACTER_NOT_FOUND;
return character == node.mData.get(insertionIndex).mChars[0] ? insertionIndex
: CHARACTER_NOT_FOUND;
}
/**
* Helper method to find a word in a given branch.
*/
public static CharGroup findWordInTree(Node node, final String s) {
int index = 0;
final StringBuilder checker = new StringBuilder();
CharGroup currentGroup;
do {
int indexOfGroup = findIndexOfChar(node, s.codePointAt(index));
if (CHARACTER_NOT_FOUND == indexOfGroup) return null;
currentGroup = node.mData.get(indexOfGroup);
checker.append(new String(currentGroup.mChars, 0, currentGroup.mChars.length));
index += currentGroup.mChars.length;
if (index < s.length()) {
node = currentGroup.mChildren;
}
} while (null != node && index < s.length());
if (!s.equals(checker.toString())) return null;
return currentGroup;
}
/**
* Recursively count the number of character groups in a given branch of the trie.
*
* @param node the parent node.
* @return the number of char groups in all the branch under this node.
*/
public static int countCharGroups(final Node node) {
final int nodeSize = node.mData.size();
int size = nodeSize;
for (int i = nodeSize - 1; i >= 0; --i) {
CharGroup group = node.mData.get(i);
if (null != group.mChildren)
size += countCharGroups(group.mChildren);
}
return size;
}
/**
* Recursively count the number of nodes in a given branch of the trie.
*
* @param node the node to count.
* @result the number of nodes in this branch.
*/
public static int countNodes(final Node node) {
int size = 1;
for (int i = node.mData.size() - 1; i >= 0; --i) {
CharGroup group = node.mData.get(i);
if (null != group.mChildren)
size += countNodes(group.mChildren);
}
return size;
}
// 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 nodes. Number of nodes : " + countNodes(root));
// MakedictLog.i("Number of groups : " + countCharGroups(root));
//
// final HashMap<String, ArrayList<Node>> repository =
// new HashMap<String, ArrayList<Node>>();
// mergeTailsInner(repository, root);
//
// MakedictLog.i("Number of different pseudohashes : " + repository.size());
// int size = 0;
// for (ArrayList<Node> 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(Node a, Node 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) {
// CharGroup aGroup = a.data.get(i);
// CharGroup bGroup = b.data.get(i);
// if (aGroup.frequency != bGroup.frequency) return false;
// if (aGroup.alternates == null && bGroup.alternates != null) return false;
// if (aGroup.alternates != null && !aGroup.equals(bGroup.alternates)) return false;
// if (!Arrays.equals(aGroup.chars, bGroup.chars)) return false;
// if (!isEqual(aGroup.children, bGroup.children)) return false;
// }
// return true;
// }
// static private HashMap<String, ArrayList<Node>> mergeTailsInner(
// final HashMap<String, ArrayList<Node>> map, final Node node) {
// final ArrayList<CharGroup> branches = node.data;
// final int nodeSize = branches.size();
// for (int i = 0; i < nodeSize; ++i) {
// CharGroup group = branches.get(i);
// if (null != group.children) {
// String pseudoHash = getPseudoHash(group.children);
// ArrayList<Node> similarList = map.get(pseudoHash);
// if (null == similarList) {
// similarList = new ArrayList<Node>();
// map.put(pseudoHash, similarList);
// }
// boolean merged = false;
// for (Node similar : similarList) {
// if (isEqual(group.children, similar)) {
// group.children = similar;
// merged = true;
// break;
// }
// }
// if (!merged) {
// similarList.add(group.children);
// }
// mergeTailsInner(map, group.children);
// }
// }
// return map;
// }
// private static String getPseudoHash(final Node node) {
// StringBuilder s = new StringBuilder();
// for (CharGroup g : node.data) {
// s.append(g.frequency);
// for (int ch : g.chars){
// s.append(Character.toChars(ch));
// }
// }
// return s.toString();
// }
/**
* Iterator to walk through a dictionary.
*
* This is purely for convenience.
*/
public static class DictionaryIterator implements Iterator<Word> {
private static class Position {
public Iterator<CharGroup> pos;
public int length;
public Position(ArrayList<CharGroup> groups) {
pos = groups.iterator();
length = 0;
}
}
final StringBuilder mCurrentString;
final LinkedList<Position> mPositions;
public DictionaryIterator(ArrayList<CharGroup> root) {
mCurrentString = new StringBuilder();
mPositions = new LinkedList<Position>();
final Position rootPos = new Position(root);
mPositions.add(rootPos);
}
@Override
public boolean hasNext() {
for (Position p : mPositions) {
if (p.pos.hasNext()) {
return true;
}
}
return false;
}
@Override
public Word next() {
Position currentPos = mPositions.getLast();
mCurrentString.setLength(mCurrentString.length() - currentPos.length);
do {
if (currentPos.pos.hasNext()) {
final CharGroup currentGroup = currentPos.pos.next();
currentPos.length = currentGroup.mChars.length;
for (int i : currentGroup.mChars)
mCurrentString.append(Character.toChars(i));
if (null != currentGroup.mChildren) {
currentPos = new Position(currentGroup.mChildren.mData);
mPositions.addLast(currentPos);
}
if (currentGroup.mFrequency >= 0)
return new Word(mCurrentString.toString(), currentGroup.mFrequency,
currentGroup.mBigrams);
} else {
mPositions.removeLast();
currentPos = mPositions.getLast();
mCurrentString.setLength(mCurrentString.length() - mPositions.getLast().length);
}
} while(true);
}
@Override
public void remove() {
throw new UnsupportedOperationException("Unsupported yet");
}
}
/**
* Method to return an iterator.
*
* This method enables Java's enhanced for loop. With this you can have a FusionDictionary x
* and say : for (Word w : x) {}
*/
@Override
public Iterator<Word> iterator() {
return new DictionaryIterator(mRoot.mData);
}
}

View File

@ -0,0 +1,40 @@
/*
* Copyright (C) 2011 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;
/**
* Wrapper to redirect log events to the right output medium.
*/
public class MakedictLog {
private static void print(String message) {
System.out.println(message);
}
public static void d(String message) {
print(message);
}
public static void e(String message) {
print(message);
}
public static void i(String message) {
print(message);
}
public static void w(String message) {
print(message);
}
}

View File

@ -0,0 +1,32 @@
/*
* Copyright (C) 2011 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;
/**
* A not-yet-resolved attribute.
*
* An attribute is either a bigram or an shortcut.
* All instances of this class are always immutable.
*/
public class PendingAttribute {
public final int mFrequency;
public final int mAddress;
public PendingAttribute(final int frequency, final int address) {
mFrequency = frequency;
mAddress = address;
}
}

View File

@ -0,0 +1,26 @@
/*
* Copyright (C) 2011 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;
/**
* Simple exception thrown when a file format is not recognized.
*/
public class UnsupportedFormatException extends Exception {
public UnsupportedFormatException(String description) {
super(description);
}
}

View File

@ -0,0 +1,65 @@
/*
* Copyright (C) 2011 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;
import com.android.inputmethod.latin.FusionDictionary.WeightedString;
import java.util.ArrayList;
/**
* Utility class for a word with a frequency.
*
* This is chiefly used to iterate a dictionary.
*/
public class Word implements Comparable<Word> {
final String mWord;
final int mFrequency;
final ArrayList<WeightedString> mBigrams;
public Word(String word, int frequency, ArrayList<WeightedString> bigrams) {
mWord = word;
mFrequency = frequency;
mBigrams = bigrams;
}
/**
* Three-way comparison.
*
* A Word x is greater than a word y if x has a higher frequency. If they have the same
* frequency, they are sorted in lexicographic order.
*/
@Override
public int compareTo(Word w) {
if (mFrequency < w.mFrequency) return 1;
if (mFrequency > w.mFrequency) return -1;
return mWord.compareTo(w.mWord);
}
/**
* Equality test.
*
* Words are equal if they have the same frequency, the same spellings, and the same
* attributes.
*/
@Override
public boolean equals(Object o) {
if (!(o instanceof Word)) return false;
Word w = (Word)o;
return mFrequency == w.mFrequency && mWord.equals(w.mWord)
&& mBigrams.equals(w.mBigrams);
}
}

View File

@ -0,0 +1,215 @@
/*
* Copyright (C) 2011 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;
import com.android.inputmethod.latin.FusionDictionary.WeightedString;
import java.io.IOException;
import java.io.InputStream;
import java.io.Writer;
import java.util.ArrayList;
import java.util.HashMap;
import java.util.TreeSet;
import javax.xml.parsers.ParserConfigurationException;
import javax.xml.parsers.SAXParser;
import javax.xml.parsers.SAXParserFactory;
import org.xml.sax.Attributes;
import org.xml.sax.SAXException;
import org.xml.sax.helpers.DefaultHandler;
/**
* Reads and writes XML files for a FusionDictionary.
*
* All functions in this class are static.
*/
public class XmlDictInputOutput {
private static final String WORD_TAG = "w";
private static final String BIGRAM_TAG = "bigram";
private static final String FREQUENCY_ATTR = "f";
private static final String WORD_ATTR = "word";
/**
* SAX handler for a unigram XML file.
*/
static private class UnigramHandler extends DefaultHandler {
// Parser states
private static final int NONE = 0;
private static final int START = 1;
private static final int WORD = 2;
private static final int BIGRAM = 4;
private static final int END = 5;
private static final int UNKNOWN = 6;
final FusionDictionary mDictionary;
int mState; // the state of the parser
int mFreq; // the currently read freq
final HashMap<String, ArrayList<WeightedString>> mBigramsMap;
/**
* Create the handler.
*
* @param dict the dictionary to construct.
* @param bigrams the bigrams as a map. This may be empty, but may not be null.
*/
public UnigramHandler(FusionDictionary dict,
HashMap<String, ArrayList<WeightedString>> bigrams) {
mDictionary = dict;
mBigramsMap = bigrams;
mState = START;
mFreq = 0;
}
@Override
public void startElement(String uri, String localName, String qName, Attributes attrs) {
if (WORD_TAG.equals(localName)) {
mState = WORD;
for (int attrIndex = 0; attrIndex < attrs.getLength(); ++attrIndex) {
final String attrName = attrs.getLocalName(attrIndex);
if (FREQUENCY_ATTR.equals(attrName)) {
mFreq = Integer.parseInt(attrs.getValue(attrIndex));
}
}
} else {
mState = UNKNOWN;
}
}
@Override
public void characters(char[] ch, int start, int length) {
if (WORD == mState) {
final String word = String.copyValueOf(ch, start, length);
mDictionary.add(word, mFreq, mBigramsMap.get(word));
}
}
@Override
public void endElement(String uri, String localName, String qName) {
if (WORD == mState) mState = START;
}
}
/**
* SAX handler for a bigram XML file.
*/
static private class BigramHandler extends DefaultHandler {
private final static String BIGRAM_W1_TAG = "bi";
private final static String BIGRAM_W2_TAG = "w";
private final static String BIGRAM_W1_ATTRIBUTE = "w1";
private final static String BIGRAM_W2_ATTRIBUTE = "w2";
private final static String BIGRAM_FREQ_ATTRIBUTE = "p";
String mW1;
final HashMap<String, ArrayList<WeightedString>> mBigramsMap;
public BigramHandler() {
mW1 = null;
mBigramsMap = new HashMap<String, ArrayList<WeightedString>>();
}
@Override
public void startElement(String uri, String localName, String qName, Attributes attrs) {
if (BIGRAM_W1_TAG.equals(localName)) {
mW1 = attrs.getValue(uri, BIGRAM_W1_ATTRIBUTE);
} else if (BIGRAM_W2_TAG.equals(localName)) {
String w2 = attrs.getValue(uri, BIGRAM_W2_ATTRIBUTE);
int freq = Integer.parseInt(attrs.getValue(uri, BIGRAM_FREQ_ATTRIBUTE));
WeightedString bigram = new WeightedString(w2, freq / 8);
ArrayList<WeightedString> bigramList = mBigramsMap.get(mW1);
if (null == bigramList) bigramList = new ArrayList<WeightedString>();
bigramList.add(bigram);
mBigramsMap.put(mW1, bigramList);
}
}
public HashMap<String, ArrayList<WeightedString>> getBigramMap() {
return mBigramsMap;
}
}
/**
* Reads a dictionary from an XML file.
*
* This is the public method that will parse an XML file and return the corresponding memory
* representation.
*
* @param unigrams the file to read the data from.
* @return the in-memory representation of the dictionary.
*/
public static FusionDictionary readDictionaryXml(InputStream unigrams, InputStream bigrams)
throws SAXException, IOException, ParserConfigurationException {
final SAXParserFactory factory = SAXParserFactory.newInstance();
factory.setNamespaceAware(true);
final SAXParser parser = factory.newSAXParser();
final BigramHandler bigramHandler = new BigramHandler();
if (null != bigrams) parser.parse(bigrams, bigramHandler);
final FusionDictionary dict = new FusionDictionary();
final UnigramHandler unigramHandler =
new UnigramHandler(dict, bigramHandler.getBigramMap());
parser.parse(unigrams, unigramHandler);
return dict;
}
/**
* Reads a dictionary in the first, legacy XML format
*
* This method reads data from the parser and creates a new FusionDictionary with it.
* The format parsed by this method is the format used before Ice Cream Sandwich,
* which has no support for bigrams or shortcuts.
* It is important to note that this method expects the parser to have already eaten
* the first, all-encompassing tag.
*
* @param xpp the parser to read the data from.
* @return the parsed dictionary.
*/
/**
* Writes a dictionary to an XML file.
*
* The output format is the "second" format, which supports bigrams and shortcuts.
*
* @param destination a destination stream to write to.
* @param dict the dictionary to write.
*/
public static void writeDictionaryXml(Writer destination, FusionDictionary dict)
throws IOException {
final TreeSet<Word> set = new TreeSet<Word>();
for (Word word : dict) {
set.add(word);
}
// TODO: use an XMLSerializer if this gets big
destination.write("<wordlist format=\"2\">\n");
for (Word word : set) {
destination.write(" <" + WORD_TAG + " " + WORD_ATTR + "=\"" + word.mWord + "\" "
+ FREQUENCY_ATTR + "=\"" + word.mFrequency + "\">");
if (null != word.mBigrams) {
destination.write("\n");
for (WeightedString bigram : word.mBigrams) {
destination.write(" <" + BIGRAM_TAG + " " + FREQUENCY_ATTR + "=\""
+ bigram.mFrequency + "\">" + bigram.mWord + "</" + BIGRAM_TAG + ">\n");
}
destination.write(" ");
}
destination.write("</" + WORD_TAG + ">\n");
}
destination.write("</wordlist>\n");
destination.close();
}
}

View File

@ -0,0 +1,55 @@
/*
* Copyright (C) 2011 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;
import com.android.inputmethod.latin.FusionDictionary.Node;
import java.util.ArrayList;
import junit.framework.TestCase;
/**
* Unit tests for BinaryDictInputOutput.
*/
public class BinaryDictInputOutputTest extends TestCase {
public void setUp() throws Exception {
super.setUp();
}
public void tearDown() throws Exception {
super.tearDown();
}
// Test the flattened array contains the expected number of nodes, and
// that it does not contain any duplicates.
public void testFlattenNodes() {
final FusionDictionary dict = new FusionDictionary();
dict.add("foo", 1, null);
dict.add("fta", 1, null);
dict.add("ftb", 1, null);
dict.add("bar", 1, null);
dict.add("fool", 1, null);
final ArrayList<Node> result = BinaryDictInputOutput.flattenTree(dict.mRoot);
assertEquals(4, result.size());
while (!result.isEmpty()) {
final Node n = result.remove(0);
assertFalse("Flattened array contained the same node twice", result.contains(n));
}
}
}