aboutsummaryrefslogtreecommitdiffstats
path: root/java/src
diff options
context:
space:
mode:
Diffstat (limited to 'java/src')
-rw-r--r--java/src/com/android/inputmethod/latin/makedict/BinaryDictInputOutput.java123
-rw-r--r--java/src/com/android/inputmethod/latin/makedict/FusionDictionary.java8
-rw-r--r--java/src/com/android/inputmethod/latin/userdictionary/UserDictionarySettings.java20
3 files changed, 100 insertions, 51 deletions
diff --git a/java/src/com/android/inputmethod/latin/makedict/BinaryDictInputOutput.java b/java/src/com/android/inputmethod/latin/makedict/BinaryDictInputOutput.java
index 29e7c28f3..a54344ab5 100644
--- a/java/src/com/android/inputmethod/latin/makedict/BinaryDictInputOutput.java
+++ b/java/src/com/android/inputmethod/latin/makedict/BinaryDictInputOutput.java
@@ -389,7 +389,7 @@ public final class BinaryDictInputOutput {
* @param node the node to compute the maximum size of.
* @param options file format options.
*/
- private static void setNodeMaximumSize(final Node node, final FormatOptions options) {
+ private static void calculateNodeMaximumSize(final Node node, final FormatOptions options) {
int size = getGroupCountSize(node);
for (CharGroup g : node.mData) {
final int groupSize = getCharGroupMaximumSize(g, options);
@@ -548,17 +548,17 @@ public final class BinaryDictInputOutput {
boolean changed = false;
int size = getGroupCountSize(node);
for (CharGroup group : node.mData) {
- if (group.mCachedAddress != node.mCachedAddress + size) {
+ if (group.mCachedAddress != node.mCachedAddressBeforeUpdate + size) {
changed = true;
- group.mCachedAddress = node.mCachedAddress + size;
+ group.mCachedAddress = node.mCachedAddressBeforeUpdate + size;
}
int groupSize = getGroupHeaderSize(group, formatOptions);
if (group.isTerminal()) groupSize += FormatSpec.GROUP_FREQUENCY_SIZE;
if (null == group.mChildren && formatOptions.mSupportsDynamicUpdate) {
groupSize += FormatSpec.SIGNED_CHILDREN_ADDRESS_SIZE;
} else if (null != group.mChildren) {
- final int offsetBasePoint = groupSize + node.mCachedAddress + size;
- final int offset = group.mChildren.mCachedAddress - offsetBasePoint;
+ final int offsetBasePoint = groupSize + node.mCachedAddressBeforeUpdate + size;
+ final int offset = group.mChildren.mCachedAddressBeforeUpdate - offsetBasePoint;
if (formatOptions.mSupportsDynamicUpdate) {
groupSize += FormatSpec.SIGNED_CHILDREN_ADDRESS_SIZE;
} else {
@@ -568,7 +568,7 @@ public final class BinaryDictInputOutput {
groupSize += getShortcutListSize(group.mShortcutTargets);
if (null != group.mBigrams) {
for (WeightedString bigram : group.mBigrams) {
- final int offsetBasePoint = groupSize + node.mCachedAddress + size
+ final int offsetBasePoint = groupSize + node.mCachedAddressBeforeUpdate + size
+ FormatSpec.GROUP_FLAGS_SIZE;
final int addressOfBigram = findAddressOfWord(dict, bigram.mWord);
final int offset = addressOfBigram - offsetBasePoint;
@@ -589,20 +589,48 @@ public final class BinaryDictInputOutput {
}
/**
- * Computes the byte size of a list of nodes and updates each node cached position.
+ * Initializes the cached addresses of nodes from their size.
*
* @param flatNodes the array of nodes.
* @param formatOptions file format options.
* @return the byte size of the entire stack.
*/
- private static int stackNodes(final ArrayList<Node> flatNodes,
+ private static int initializeNodesCachedAddresses(final ArrayList<Node> flatNodes,
final FormatOptions formatOptions) {
int nodeOffset = 0;
- for (Node n : flatNodes) {
- n.mCachedAddress = nodeOffset;
+ for (final Node n : flatNodes) {
+ n.mCachedAddressBeforeUpdate = nodeOffset;
+ int groupCountSize = getGroupCountSize(n);
+ int groupOffset = 0;
+ for (final CharGroup g : n.mData) {
+ g.mCachedAddress = groupCountSize + nodeOffset + groupOffset;
+ groupOffset += g.mCachedSize;
+ }
+ final int nodeSize = groupCountSize + groupOffset
+ + (formatOptions.mSupportsDynamicUpdate
+ ? FormatSpec.FORWARD_LINK_ADDRESS_SIZE : 0);
+ nodeOffset += n.mCachedSize;
+ }
+ return nodeOffset;
+ }
+
+ /**
+ * Updates the cached addresses of nodes after recomputing their new positions.
+ *
+ * @param flatNodes the array of nodes.
+ * @param formatOptions file format options.
+ * @return the byte size of the entire stack.
+ */
+ private static int updateNodeCachedAddresses(final ArrayList<Node> flatNodes,
+ final FormatOptions formatOptions) {
+ int nodeOffset = 0;
+ for (final Node n : flatNodes) {
+ n.mCachedAddressBeforeUpdate = n.mCachedAddressAfterUpdate;
int groupCountSize = getGroupCountSize(n);
int groupOffset = 0;
- for (CharGroup g : n.mData) {
+ for (final CharGroup g : n.mData) {
+ // TODO: just copy cached address after update into cached address before update
+ // when the two fields are separated.
g.mCachedAddress = groupCountSize + nodeOffset + groupOffset;
groupOffset += g.mCachedSize;
}
@@ -610,8 +638,13 @@ public final class BinaryDictInputOutput {
+ (formatOptions.mSupportsDynamicUpdate
? FormatSpec.FORWARD_LINK_ADDRESS_SIZE : 0);
if (nodeSize != n.mCachedSize) {
+ // TODO: remove this test when the addresses are separated
throw new RuntimeException("Bug : Stored and computed node size differ");
}
+ if (nodeOffset != n.mCachedAddressAfterUpdate) {
+ // TODO: remove this test when the code is well tested
+ throw new RuntimeException("Bug : Stored and computed node address differ");
+ }
nodeOffset += n.mCachedSize;
}
return nodeOffset;
@@ -627,11 +660,13 @@ public final class BinaryDictInputOutput {
*/
private static void computeParentAddresses(final ArrayList<Node> flatNodes) {
for (final Node node : flatNodes) {
- for (CharGroup group : node.mData) {
+ for (final CharGroup group : node.mData) {
if (null != group.mChildren) {
- // assign my address to children's parent address
+ // Assign my address to children's parent address
+ // Here BeforeUpdate and AfterUpdate addresses have the same value, so it
+ // does not matter which we use.
group.mChildren.mCachedParentAddress = group.mCachedAddress
- - group.mChildren.mCachedAddress;
+ - group.mChildren.mCachedAddressAfterUpdate;
}
}
}
@@ -654,9 +689,9 @@ public final class BinaryDictInputOutput {
*/
private static ArrayList<Node> computeAddresses(final FusionDictionary dict,
final ArrayList<Node> flatNodes, final FormatOptions formatOptions) {
- // First get the worst sizes and offsets
- for (Node n : flatNodes) setNodeMaximumSize(n, formatOptions);
- final int offset = stackNodes(flatNodes, formatOptions);
+ // First get the worst possible sizes and offsets
+ for (final Node n : flatNodes) calculateNodeMaximumSize(n, formatOptions);
+ final int offset = initializeNodesCachedAddresses(flatNodes, formatOptions);
MakedictLog.i("Compressing the array addresses. Original size : " + offset);
MakedictLog.i("(Recursively seen size : " + offset + ")");
@@ -665,14 +700,17 @@ public final class BinaryDictInputOutput {
boolean changesDone = false;
do {
changesDone = false;
- for (Node n : flatNodes) {
+ int nodeStartOffset = 0;
+ for (final Node n : flatNodes) {
+ n.mCachedAddressAfterUpdate = nodeStartOffset;
final int oldNodeSize = n.mCachedSize;
final boolean changed = computeActualNodeSize(n, dict, formatOptions);
final int newNodeSize = n.mCachedSize;
if (oldNodeSize < newNodeSize) throw new RuntimeException("Increased size ?!");
+ nodeStartOffset += newNodeSize;
changesDone |= changed;
}
- stackNodes(flatNodes, formatOptions);
+ updateNodeCachedAddresses(flatNodes, formatOptions);
++passes;
if (passes > MAX_PASSES) throw new RuntimeException("Too many passes - probably a bug");
} while (changesDone);
@@ -683,7 +721,7 @@ public final class BinaryDictInputOutput {
final Node lastNode = flatNodes.get(flatNodes.size() - 1);
MakedictLog.i("Compression complete in " + passes + " passes.");
MakedictLog.i("After address compression : "
- + (lastNode.mCachedAddress + lastNode.mCachedSize));
+ + (lastNode.mCachedAddressAfterUpdate + lastNode.mCachedSize));
return flatNodes;
}
@@ -701,10 +739,12 @@ public final class BinaryDictInputOutput {
private static void checkFlatNodeArray(final ArrayList<Node> array) {
int offset = 0;
int index = 0;
- for (Node n : array) {
- if (n.mCachedAddress != offset) {
+ for (final Node n : array) {
+ // BeforeUpdate and AfterUpdate addresses are the same here, so it does not matter
+ // which we use.
+ if (n.mCachedAddressAfterUpdate != offset) {
throw new RuntimeException("Wrong address for node " + index
- + " : expected " + offset + ", got " + n.mCachedAddress);
+ + " : expected " + offset + ", got " + n.mCachedAddressAfterUpdate);
}
++index;
offset += n.mCachedSize;
@@ -946,7 +986,7 @@ public final class BinaryDictInputOutput {
private static int writePlacedNode(final FusionDictionary dict, byte[] buffer,
final Node node, final FormatOptions formatOptions) {
// TODO: Make the code in common with BinaryDictIOUtils#writeCharGroup
- int index = node.mCachedAddress;
+ int index = node.mCachedAddressAfterUpdate;
final int groupCount = node.mData.size();
final int countSize = getGroupCountSize(node);
@@ -977,7 +1017,7 @@ public final class BinaryDictInputOutput {
if (group.mFrequency >= 0) groupAddress += FormatSpec.GROUP_FREQUENCY_SIZE;
final int childrenOffset = null == group.mChildren
? FormatSpec.NO_CHILDREN_ADDRESS
- : group.mChildren.mCachedAddress - groupAddress;
+ : group.mChildren.mCachedAddressAfterUpdate - groupAddress;
byte flags = makeCharGroupFlags(group, groupAddress, childrenOffset, formatOptions);
buffer[index++] = flags;
@@ -985,7 +1025,7 @@ public final class BinaryDictInputOutput {
index = writeParentAddress(buffer, index, parentAddress, formatOptions);
} else {
index = writeParentAddress(buffer, index,
- parentAddress + (node.mCachedAddress - group.mCachedAddress),
+ parentAddress + (node.mCachedAddressAfterUpdate - group.mCachedAddress),
formatOptions);
}
@@ -1055,9 +1095,9 @@ public final class BinaryDictInputOutput {
= FormatSpec.NO_FORWARD_LINK_ADDRESS;
index += FormatSpec.FORWARD_LINK_ADDRESS_SIZE;
}
- if (index != node.mCachedAddress + node.mCachedSize) throw new RuntimeException(
+ if (index != node.mCachedAddressAfterUpdate + node.mCachedSize) throw new RuntimeException(
"Not the same size : written "
- + (index - node.mCachedAddress) + " bytes out of a node that should have "
+ + (index - node.mCachedAddressAfterUpdate) + " bytes from a node that should have "
+ node.mCachedSize + " bytes");
return index;
}
@@ -1077,25 +1117,27 @@ public final class BinaryDictInputOutput {
int charGroups = 0;
int maxGroups = 0;
int maxRuns = 0;
- for (Node n : nodes) {
+ for (final Node n : nodes) {
if (maxGroups < n.mData.size()) maxGroups = n.mData.size();
- for (CharGroup cg : n.mData) {
+ for (final CharGroup cg : n.mData) {
++charGroups;
if (cg.mChars.length > maxRuns) maxRuns = cg.mChars.length;
if (cg.mFrequency >= 0) {
- if (n.mCachedAddress < firstTerminalAddress)
- firstTerminalAddress = n.mCachedAddress;
- if (n.mCachedAddress > lastTerminalAddress)
- lastTerminalAddress = n.mCachedAddress;
+ if (n.mCachedAddressAfterUpdate < firstTerminalAddress)
+ firstTerminalAddress = n.mCachedAddressAfterUpdate;
+ if (n.mCachedAddressAfterUpdate > lastTerminalAddress)
+ lastTerminalAddress = n.mCachedAddressAfterUpdate;
}
}
- if (n.mCachedAddress + n.mCachedSize > size) size = n.mCachedAddress + n.mCachedSize;
+ if (n.mCachedAddressAfterUpdate + n.mCachedSize > size) {
+ size = n.mCachedAddressAfterUpdate + n.mCachedSize;
+ }
}
final int[] groupCounts = new int[maxGroups + 1];
final int[] runCounts = new int[maxRuns + 1];
- for (Node n : nodes) {
+ for (final Node n : nodes) {
++groupCounts[n.mData.size()];
- for (CharGroup cg : n.mData) {
+ for (final CharGroup cg : n.mData) {
++runCounts[cg.mChars.length];
}
}
@@ -1205,7 +1247,7 @@ public final class BinaryDictInputOutput {
// Create a buffer that matches the final dictionary size.
final Node lastNode = flatNodes.get(flatNodes.size() - 1);
- final int bufferSize = lastNode.mCachedAddress + lastNode.mCachedSize;
+ final int bufferSize = lastNode.mCachedAddressAfterUpdate + lastNode.mCachedSize;
final byte[] buffer = new byte[bufferSize];
int index = 0;
@@ -1584,8 +1626,9 @@ public final class BinaryDictInputOutput {
buffer.position() != FormatSpec.NO_FORWARD_LINK_ADDRESS);
final Node node = new Node(nodeContents);
- node.mCachedAddress = nodeOrigin;
- reverseNodeMap.put(node.mCachedAddress, node);
+ node.mCachedAddressBeforeUpdate = nodeOrigin;
+ node.mCachedAddressAfterUpdate = nodeOrigin;
+ reverseNodeMap.put(node.mCachedAddressAfterUpdate, node);
return node;
}
diff --git a/java/src/com/android/inputmethod/latin/makedict/FusionDictionary.java b/java/src/com/android/inputmethod/latin/makedict/FusionDictionary.java
index 17d281518..5530b7a74 100644
--- a/java/src/com/android/inputmethod/latin/makedict/FusionDictionary.java
+++ b/java/src/com/android/inputmethod/latin/makedict/FusionDictionary.java
@@ -46,7 +46,13 @@ public final class FusionDictionary implements Iterable<Word> {
ArrayList<CharGroup> mData;
// To help with binary generation
int mCachedSize = Integer.MIN_VALUE;
- int mCachedAddress = Integer.MIN_VALUE;
+ // mCachedAddressBefore/AfterUpdate are helpers for binary dictionary generation. They
+ // always hold the same value except between dictionary address compression, during which
+ // the update process needs to know about both values at the same time. Updating will
+ // update the AfterUpdate value, and the code will move them to BeforeUpdate before
+ // the next update pass.
+ int mCachedAddressBeforeUpdate = Integer.MIN_VALUE;
+ int mCachedAddressAfterUpdate = Integer.MIN_VALUE;
int mCachedParentAddress = 0;
public Node() {
diff --git a/java/src/com/android/inputmethod/latin/userdictionary/UserDictionarySettings.java b/java/src/com/android/inputmethod/latin/userdictionary/UserDictionarySettings.java
index 50dda9663..64f860e7d 100644
--- a/java/src/com/android/inputmethod/latin/userdictionary/UserDictionarySettings.java
+++ b/java/src/com/android/inputmethod/latin/userdictionary/UserDictionarySettings.java
@@ -1,17 +1,17 @@
-/**
- * Copyright (C) 2013 Google Inc.
+/*
+ * 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
+ * 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
+ * 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.
+ * 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.userdictionary;