blob: 87111f29d65fc2e5b39330ea87f17765368130ff [file] [log] [blame]
/*
* Copyright (C) 2010 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.gallery3d.data;
import android.content.Context;
import android.text.format.DateFormat;
import android.text.format.DateUtils;
import com.android.gallery3d.common.Utils;
import com.android.gallery3d.util.GalleryUtils;
import java.util.ArrayList;
import java.util.Collections;
import java.util.Comparator;
public class TimeClustering extends Clustering {
@SuppressWarnings("unused")
private static final String TAG = "TimeClustering";
// If 2 items are greater than 25 miles apart, they will be in different
// clusters.
private static final int GEOGRAPHIC_DISTANCE_CUTOFF_IN_MILES = 20;
// Do not want to split based on anything under 1 min.
private static final long MIN_CLUSTER_SPLIT_TIME_IN_MS = 60000L;
// Disregard a cluster split time of anything over 2 hours.
private static final long MAX_CLUSTER_SPLIT_TIME_IN_MS = 7200000L;
// Try and get around 9 clusters (best-effort for the common case).
private static final int NUM_CLUSTERS_TARGETED = 9;
// Try and merge 2 clusters if they are both smaller than min cluster size.
// The min cluster size can range from 8 to 15.
private static final int MIN_MIN_CLUSTER_SIZE = 8;
private static final int MAX_MIN_CLUSTER_SIZE = 15;
// Try and split a cluster if it is bigger than max cluster size.
// The max cluster size can range from 20 to 50.
private static final int MIN_MAX_CLUSTER_SIZE = 20;
private static final int MAX_MAX_CLUSTER_SIZE = 50;
// Initially put 2 items in the same cluster as long as they are within
// 3 cluster frequencies of each other.
private static int CLUSTER_SPLIT_MULTIPLIER = 3;
// The minimum change factor in the time between items to consider a
// partition.
// Example: (Item 3 - Item 2) / (Item 2 - Item 1).
private static final int MIN_PARTITION_CHANGE_FACTOR = 2;
// Make the cluster split time of a large cluster half that of a regular
// cluster.
private static final int PARTITION_CLUSTER_SPLIT_TIME_FACTOR = 2;
private Context mContext;
private ArrayList<Cluster> mClusters;
private String[] mNames;
private Cluster mCurrCluster;
private long mClusterSplitTime =
(MIN_CLUSTER_SPLIT_TIME_IN_MS + MAX_CLUSTER_SPLIT_TIME_IN_MS) / 2;
private long mLargeClusterSplitTime =
mClusterSplitTime / PARTITION_CLUSTER_SPLIT_TIME_FACTOR;
private int mMinClusterSize = (MIN_MIN_CLUSTER_SIZE + MAX_MIN_CLUSTER_SIZE) / 2;
private int mMaxClusterSize = (MIN_MAX_CLUSTER_SIZE + MAX_MAX_CLUSTER_SIZE) / 2;
private static final Comparator<SmallItem> sDateComparator =
new DateComparator();
private static class DateComparator implements Comparator<SmallItem> {
@Override
public int compare(SmallItem item1, SmallItem item2) {
return -Utils.compare(item1.dateInMs, item2.dateInMs);
}
}
public TimeClustering(Context context) {
mContext = context;
mClusters = new ArrayList<Cluster>();
mCurrCluster = new Cluster();
}
@Override
public void run(MediaSet baseSet) {
final int total = baseSet.getTotalMediaItemCount();
final SmallItem[] buf = new SmallItem[total];
final double[] latLng = new double[2];
baseSet.enumerateTotalMediaItems(new MediaSet.ItemConsumer() {
@Override
public void consume(int index, MediaItem item) {
if (index < 0 || index >= total) return;
SmallItem s = new SmallItem();
s.path = item.getPath();
s.mediaType = item.getMediaType();
s.dateInMs = item.getDateInMs();
item.getLatLong(latLng);
s.lat = latLng[0];
s.lng = latLng[1];
buf[index] = s;
}
});
ArrayList<SmallItem> items = new ArrayList<SmallItem>(total);
for (int i = 0; i < total; i++) {
if (buf[i] != null) {
items.add(buf[i]);
}
}
Collections.sort(items, sDateComparator);
int n = items.size();
long minTime = 0;
long maxTime = 0;
for (int i = 0; i < n; i++) {
long t = items.get(i).dateInMs;
if (t == 0) continue;
if (minTime == 0) {
minTime = maxTime = t;
} else {
minTime = Math.min(minTime, t);
maxTime = Math.max(maxTime, t);
}
}
setTimeRange(maxTime - minTime, n);
for (int i = 0; i < n; i++) {
compute(items.get(i));
}
compute(null);
int m = mClusters.size();
mNames = new String[m];
for (int i = 0; i < m; i++) {
mNames[i] = mClusters.get(i).generateCaption(mContext);
}
}
@Override
public int getNumberOfClusters() {
return mClusters.size();
}
@Override
public ArrayList<Path> getCluster(int index) {
ArrayList<SmallItem> items = mClusters.get(index).getItems();
ArrayList<Path> result = new ArrayList<Path>(items.size());
for (int i = 0, n = items.size(); i < n; i++) {
result.add(items.get(i).path);
}
return result;
}
@Override
public String getClusterName(int index) {
return mNames[index];
}
private void setTimeRange(long timeRange, int numItems) {
if (numItems != 0) {
int meanItemsPerCluster = numItems / NUM_CLUSTERS_TARGETED;
// Heuristic to get min and max cluster size - half and double the
// desired items per cluster.
mMinClusterSize = meanItemsPerCluster / 2;
mMaxClusterSize = meanItemsPerCluster * 2;
mClusterSplitTime = timeRange / numItems * CLUSTER_SPLIT_MULTIPLIER;
}
mClusterSplitTime = Utils.clamp(mClusterSplitTime, MIN_CLUSTER_SPLIT_TIME_IN_MS, MAX_CLUSTER_SPLIT_TIME_IN_MS);
mLargeClusterSplitTime = mClusterSplitTime / PARTITION_CLUSTER_SPLIT_TIME_FACTOR;
mMinClusterSize = Utils.clamp(mMinClusterSize, MIN_MIN_CLUSTER_SIZE, MAX_MIN_CLUSTER_SIZE);
mMaxClusterSize = Utils.clamp(mMaxClusterSize, MIN_MAX_CLUSTER_SIZE, MAX_MAX_CLUSTER_SIZE);
}
@Override
public int getClusterImageCount(int index) {
// TODO Auto-generated method stub
return mClusters.get(index).mPhotoCount;
}
@Override
public int getClusterVideoCount(int index) {
// TODO Auto-generated method stub
return mClusters.get(index).mVideoCount;
}
private void compute(SmallItem currentItem) {
if (currentItem != null) {
int numClusters = mClusters.size();
int numCurrClusterItems = mCurrCluster.size();
boolean geographicallySeparateItem = false;
boolean itemAddedToCurrentCluster = false;
// Determine if this item should go in the current cluster or be the
// start of a new cluster.
if (numCurrClusterItems == 0) {
mCurrCluster.addItem(currentItem);
} else {
SmallItem prevItem = mCurrCluster.getLastItem();
if (isGeographicallySeparated(prevItem, currentItem)) {
mClusters.add(mCurrCluster);
geographicallySeparateItem = true;
} else if (numCurrClusterItems > mMaxClusterSize) {
splitAndAddCurrentCluster();
} else if (timeDistance(prevItem, currentItem) < mClusterSplitTime) {
mCurrCluster.addItem(currentItem);
itemAddedToCurrentCluster = true;
} else if (numClusters > 0 && numCurrClusterItems < mMinClusterSize
&& !mCurrCluster.mGeographicallySeparatedFromPrevCluster) {
mergeAndAddCurrentCluster();
} else {
mClusters.add(mCurrCluster);
}
// Creating a new cluster and adding the current item to it.
if (!itemAddedToCurrentCluster) {
mCurrCluster = new Cluster();
if (geographicallySeparateItem) {
mCurrCluster.mGeographicallySeparatedFromPrevCluster = true;
}
mCurrCluster.addItem(currentItem);
}
}
} else {
if (mCurrCluster.size() > 0) {
int numClusters = mClusters.size();
int numCurrClusterItems = mCurrCluster.size();
// The last cluster may potentially be too big or too small.
if (numCurrClusterItems > mMaxClusterSize) {
splitAndAddCurrentCluster();
} else if (numClusters > 0 && numCurrClusterItems < mMinClusterSize
&& !mCurrCluster.mGeographicallySeparatedFromPrevCluster) {
mergeAndAddCurrentCluster();
} else {
mClusters.add(mCurrCluster);
}
mCurrCluster = new Cluster();
}
}
}
private void splitAndAddCurrentCluster() {
ArrayList<SmallItem> currClusterItems = mCurrCluster.getItems();
int numCurrClusterItems = mCurrCluster.size();
int secondPartitionStartIndex = getPartitionIndexForCurrentCluster();
if (secondPartitionStartIndex != -1) {
Cluster partitionedCluster = new Cluster();
for (int j = 0; j < secondPartitionStartIndex; j++) {
partitionedCluster.addItem(currClusterItems.get(j));
}
mClusters.add(partitionedCluster);
partitionedCluster = new Cluster();
for (int j = secondPartitionStartIndex; j < numCurrClusterItems; j++) {
partitionedCluster.addItem(currClusterItems.get(j));
}
mClusters.add(partitionedCluster);
} else {
mClusters.add(mCurrCluster);
}
}
private int getPartitionIndexForCurrentCluster() {
int partitionIndex = -1;
float largestChange = MIN_PARTITION_CHANGE_FACTOR;
ArrayList<SmallItem> currClusterItems = mCurrCluster.getItems();
int numCurrClusterItems = mCurrCluster.size();
int minClusterSize = mMinClusterSize;
// Could be slightly more efficient here but this code seems cleaner.
if (numCurrClusterItems > minClusterSize + 1) {
for (int i = minClusterSize; i < numCurrClusterItems - minClusterSize; i++) {
SmallItem prevItem = currClusterItems.get(i - 1);
SmallItem currItem = currClusterItems.get(i);
SmallItem nextItem = currClusterItems.get(i + 1);
long timeNext = nextItem.dateInMs;
long timeCurr = currItem.dateInMs;
long timePrev = prevItem.dateInMs;
if (timeNext == 0 || timeCurr == 0 || timePrev == 0) continue;
long diff1 = Math.abs(timeNext - timeCurr);
long diff2 = Math.abs(timeCurr - timePrev);
float change = Math.max(diff1 / (diff2 + 0.01f), diff2 / (diff1 + 0.01f));
if (change > largestChange) {
if (timeDistance(currItem, prevItem) > mLargeClusterSplitTime) {
partitionIndex = i;
largestChange = change;
} else if (timeDistance(nextItem, currItem) > mLargeClusterSplitTime) {
partitionIndex = i + 1;
largestChange = change;
}
}
}
}
return partitionIndex;
}
private void mergeAndAddCurrentCluster() {
int numClusters = mClusters.size();
Cluster prevCluster = mClusters.get(numClusters - 1);
ArrayList<SmallItem> currClusterItems = mCurrCluster.getItems();
int numCurrClusterItems = mCurrCluster.size();
if (prevCluster.size() < mMinClusterSize) {
for (int i = 0; i < numCurrClusterItems; i++) {
prevCluster.addItem(currClusterItems.get(i));
}
mClusters.set(numClusters - 1, prevCluster);
} else {
mClusters.add(mCurrCluster);
}
}
// Returns true if a, b are sufficiently geographically separated.
private static boolean isGeographicallySeparated(SmallItem itemA, SmallItem itemB) {
if (!GalleryUtils.isValidLocation(itemA.lat, itemA.lng)
|| !GalleryUtils.isValidLocation(itemB.lat, itemB.lng)) {
return false;
}
double distance = GalleryUtils.fastDistanceMeters(
Math.toRadians(itemA.lat),
Math.toRadians(itemA.lng),
Math.toRadians(itemB.lat),
Math.toRadians(itemB.lng));
return (GalleryUtils.toMile(distance) > GEOGRAPHIC_DISTANCE_CUTOFF_IN_MILES);
}
// Returns the time interval between the two items in milliseconds.
private static long timeDistance(SmallItem a, SmallItem b) {
return Math.abs(a.dateInMs - b.dateInMs);
}
}
class SmallItem {
Path path;
long dateInMs;
double lat, lng;
int mediaType;
}
class Cluster {
@SuppressWarnings("unused")
private static final String TAG = "Cluster";
private static final String MMDDYY_FORMAT = "MMddyy";
public int mPhotoCount = 0;
public int mVideoCount = 0;
// This is for TimeClustering only.
public boolean mGeographicallySeparatedFromPrevCluster = false;
private ArrayList<SmallItem> mItems = new ArrayList<SmallItem>();
public Cluster() {
}
public void addItem(SmallItem item) {
if(item.mediaType == MediaObject.MEDIA_TYPE_IMAGE) {
mPhotoCount++;
} else if(item.mediaType == MediaObject.MEDIA_TYPE_VIDEO) {
mVideoCount++;
}
mItems.add(item);
}
public int size() {
return mItems.size();
}
public SmallItem getLastItem() {
int n = mItems.size();
return (n == 0) ? null : mItems.get(n - 1);
}
public ArrayList<SmallItem> getItems() {
return mItems;
}
public String generateCaption(Context context) {
int n = mItems.size();
long minTimestamp = 0;
long maxTimestamp = 0;
for (int i = 0; i < n; i++) {
long t = mItems.get(i).dateInMs;
if (t == 0) continue;
if (minTimestamp == 0) {
minTimestamp = maxTimestamp = t;
} else {
minTimestamp = Math.min(minTimestamp, t);
maxTimestamp = Math.max(maxTimestamp, t);
}
}
if (minTimestamp == 0) return "";
String caption;
String minDay = DateFormat.format(MMDDYY_FORMAT, minTimestamp)
.toString();
String maxDay = DateFormat.format(MMDDYY_FORMAT, maxTimestamp)
.toString();
if (minDay.substring(4).equals(maxDay.substring(4))) {
// The items are from the same year - show at least as
// much granularity as abbrev_all allows.
caption = DateUtils.formatDateRange(context, minTimestamp,
maxTimestamp, DateUtils.FORMAT_ABBREV_ALL);
// Get a more granular date range string if the min and
// max timestamp are on the same day and from the
// current year.
if (minDay.equals(maxDay)) {
int flags = DateUtils.FORMAT_ABBREV_MONTH | DateUtils.FORMAT_SHOW_DATE;
// Contains the year only if the date does not
// correspond to the current year.
String dateRangeWithOptionalYear = DateUtils.formatDateTime(
context, minTimestamp, flags);
String dateRangeWithYear = DateUtils.formatDateTime(
context, minTimestamp, flags | DateUtils.FORMAT_SHOW_YEAR);
if (!dateRangeWithOptionalYear.equals(dateRangeWithYear)) {
// This means both dates are from the same year
// - show the time.
// Not enough room to display the time range.
// Pick the mid-point.
long midTimestamp = (minTimestamp + maxTimestamp) / 2;
caption = DateUtils.formatDateRange(context, midTimestamp,
midTimestamp, DateUtils.FORMAT_SHOW_TIME | flags);
}
}
} else {
// The items are not from the same year - only show
// month and year.
int flags = DateUtils.FORMAT_NO_MONTH_DAY
| DateUtils.FORMAT_ABBREV_MONTH | DateUtils.FORMAT_SHOW_DATE;
caption = DateUtils.formatDateRange(context, minTimestamp,
maxTimestamp, flags);
}
return caption;
}
}