/*
* Copyright 2004-2014 H2 Group. Multiple-Licensed under the MPL 2.0,
* and the EPL 1.0 (http://h2database.com/html/license.html).
* Initial Developer: H2 Group
*/
package org.h2.index;
import org.h2.result.SearchRow;
import org.h2.store.Data;
import org.h2.store.Page;
/**
* A page that contains index data.
*/
public abstract class PageBtree extends Page {
/**
* This is a root page.
*/
static final int ROOT = 0;
/**
* Indicator that the row count is not known.
*/
static final int UNKNOWN_ROWCOUNT = -1;
/**
* The index.
*/
protected final PageBtreeIndex index;
/**
* The page number of the parent.
*/
protected int parentPageId;
/**
* The data page.
*/
protected final Data data;
/**
* The row offsets.
*/
protected int[] offsets;
/**
* The number of entries.
*/
protected int entryCount; //对于PageBtreeNode就是分隔点(rows)的个数(也等于子节点个数-1)，对于PageBtreeLeaf就代表行数
/**
* The index data
*/
protected SearchRow[] rows;
/**
* The start of the data area.
*/
protected int start;
/**
* If only the position of the row is stored in the page
*/
protected boolean onlyPosition;
/**
* Whether the data page is up-to-date.
*/
protected boolean written;
/**
* The estimated memory used by this object.
*/
private final int memoryEstimated;
PageBtree(PageBtreeIndex index, int pageId, Data data) {
this.index = index;
this.data = data;
setPos(pageId);
memoryEstimated = index.getMemoryPerPage(); //最开始时为0
}
/**
* Get the real row count. If required, this will read all child pages.
*
* @return the row count
*/
abstract int getRowCount();
/**
* Set the stored row count. This will write the page.
*
* @param rowCount the stored row count
*/
abstract void setRowCountStored(int rowCount);
/**
* Find an entry.
*
* @param compare the row
* @param bigger if looking for a larger row
* @param add if the row should be added (check for duplicate keys)
* @param compareKeys compare the row keys as well
* @return the index of the found row
*/
//折半查找
//返回的下标会用于org.h2.store.Page的insert方法中，使得insert到rows时就按照建立索引时指定的排序方式排好了，
//比如CREATE index IF NOT EXISTS idx_name ON IndexTestTable(name desc)，因为为name指定了desc，所以rows数组是降序的
//此方法的返回值是: 0<=x<=entryCount
int find(SearchRow compare, boolean bigger, boolean add, boolean compareKeys) { //只有addRow和remove时compareKeys为true，find时compareKeys为false
if (compare == null) {
return 0;
}
int l = 0, r = entryCount;
int comp = 1;
while (l < r) {
int i = (l + r) >>> 1; //除以2，从中间元素开始
SearchRow row = getRow(i);
//如果索引字段是降序的，比如CREATE index IF NOT EXISTS idx_name ON IndexTestTable(name desc)
//那么此时会按降序比较，如果row<compare，那么得到的结果是row>compare，此时rows数组中的元素是降序排列的。
comp = index.compareRows(row, compare);
if (comp == 0) {
//增加新记录时，如果是唯一索引，当两条记录相等时，
//再根据不同数据库的兼容模式来判断记录中的null字段情况
//1. uniqueIndexSingleNull 返回false, 抛出DuplicateKeyException
//2. uniqueIndexSingleNullExceptAllColumnsAreNull
// 如果要比较的记录包含非null字段，返回false, 抛出DuplicateKeyException，否则允许通过
//3. 其他情况: 如果要比较的记录含null字段，返回true，否则返回false, 抛出DuplicateKeyException
if (add && index.indexType.isUnique()) {
if (!index.containsNullAndAllowMultipleNull(compare)) {
throw index.getDuplicateKeyException(compare.toString());
}
}
if (compareKeys) {
comp = index.compareKeys(row, compare);
if (comp == 0) {
return i;
}
}
}
//第i个元素大于要比较的元素时，往左移动，
//或者相等时，如果bigger为false，说明要找更小的，往左移动
if (comp > 0 || (!bigger && comp == 0)) {
r = i;
} else {
l = i + 1;
}
}
return l;
}
/**
* Add a row if possible. If it is possible this method returns -1,
* otherwise the split point. It is always possible to add one row.
*
* @param row the row to add
* @return the split point of this page, or -1 if no split is required
*/
abstract int addRowTry(SearchRow row);
/**
* Find the first row.
*
* @param cursor the cursor
* @param first the row to find
* @param bigger if the row should be bigger
*/
abstract void find(PageBtreeCursor cursor, SearchRow first, boolean bigger);
/**
* Find the last row.
*
* @param cursor the cursor
*/
abstract void last(PageBtreeCursor cursor);
/**
* Get the row at this position.
*
* @param at the index
* @return the row
*/
SearchRow getRow(int at) {
SearchRow row = rows[at];
if (row == null) {
row = index.readRow(data, offsets[at], onlyPosition, true);
memoryChange();
rows[at] = row;
} else if (!index.hasData(row)) {
row = index.readRow(row.getKey());
memoryChange();
rows[at] = row;
}
return row;
}
/**
* The memory usage of this page was changed. Propagate the change if
* needed.
*/
protected void memoryChange() {
// nothing to do
}
/**
* Split the index page at the given point.
*
* @param splitPoint the index where to split
* @return the new page that contains about half the entries
*/
abstract PageBtree split(int splitPoint);
/**
* Change the page id.
*
* @param id the new page id
*/
void setPageId(int id) {
changeCount = index.getPageStore().getChangeCount();
written = false;
index.getPageStore().removeFromCache(getPos());
setPos(id);
index.getPageStore().logUndo(this, null);
remapChildren();
}
/**
* Get the first child leaf page of a page.
*
* @return the page
*/
abstract PageBtreeLeaf getFirstLeaf();
/**
* Get the first child leaf page of a page.
*
* @return the page
*/
abstract PageBtreeLeaf getLastLeaf();
/**
* Change the parent page id.
*
* @param id the new parent page id
*/
void setParentPageId(int id) {
index.getPageStore().logUndo(this, data);
changeCount = index.getPageStore().getChangeCount();
written = false;
parentPageId = id;
}
/**
* Update the parent id of all children.
*/
abstract void remapChildren();
/**
* Remove a row.
*
* @param row the row to remove
* @return null if the last row didn't change,
* the deleted row if the page is now empty,
* otherwise the new last row of this page
*/
abstract SearchRow remove(SearchRow row);
/**
* Free this page and all child pages.
*/
abstract void freeRecursive();
/**
* Ensure all rows are read in memory.
*/
protected void readAllRows() {
for (int i = 0; i < entryCount; i++) {
SearchRow row = rows[i];
if (row == null) {
row = index.readRow(data, offsets[i], onlyPosition, false);
rows[i] = row;
}
}
}
/**
* Get the estimated memory size.
*
* @return number of double words (4 bytes)
*/
@Override
public int getMemory() {
// need to always return the same value for the same object (otherwise
// the cache size would change after adding and then removing the same
// page from the cache) but index.getMemoryPerPage() can adopt according
// to how much memory a row needs on average
return memoryEstimated;
}
@Override
public boolean canRemove() {
if (changeCount >= index.getPageStore().getChangeCount()) {
return false;
}
return true;
}
// 我加上的
public String tree() {
return tree("");
}
// 我加上的
public abstract String tree(String p);
}