new experimental page store

h2/src/docsrc/text/_docs_ja.utf8.txt

@@ -2275,7 +2275,7 @@ ANALYZEを実行する (詳細はドキュメントをご覧下さい)
 理論上のデータの制限は現在256GBです。これはBLOBとCLOBデータを含まない状態での値です: BLOB、CLOBの最大値はそれぞれ256GB以上の設定が可能です。インデックスデータのサイズ制限も同様に256GBです。
 
 @faq_1032_p
-#The maximum number of rows per table is 2'147'483'648. #FAT、FAT32ファイルシステムの最大ファイルサイズは4GBです。よって、FATやFAT32を利用する場合は、データの制限は4GBということになります。
+#The maximum number of rows per table is 2'147'483'648.
 
 @faq_1033_p
 FAT、FAT32ファイルシステムの最大ファイルサイズは4GBです。よって、FATやFAT32を利用する場合は、データの制限は4GBということになります。
@@ -2332,7 +2332,7 @@ FAT、FAT32ファイルシステムの最大ファイルサイズは4GBです。
 試験的に考慮された箇所は以下の通り:
 
 @faq_1051_li
-#The PostgreSQL server #他のデータベースとの互換モード (一部の特徴のみ提供される)
+#The PostgreSQL server
 
 @faq_1052_li
 他のデータベースとの互換モード (一部の特徴のみ提供される)

h2/src/main/org/h2/index/Page.java

@@ -42,4 +42,9 @@ class Page {
      */
     static final int TYPE_DATA_OVERFLOW_WITH_MORE = 7;
     
+    /**
+     * This is a root page.
+     */
+    static final int ROOT = 0;
+
 }

h2/src/main/org/h2/index/PageData.java

@@ -29,12 +29,12 @@ abstract class PageData {
     /**
      * the page number.
      */
-    protected final int pageId;
+    protected int pageId;
     
     /**
      * The page number of the parent.
      */
-    protected final int parentPageId;
+    protected int parentPageId;
     
     /**
      * The number of entries.
@@ -42,9 +42,9 @@ abstract class PageData {
     protected int entryCount;
     
     /**
-     * If the page has unwritten changes.
+     * The row keys.
      */
-    protected boolean changed;
+    protected int[] keys;
     
     PageData(PageScanIndex index, int pageId, int parentPageId, DataPageBinary data) {
         this.index = index;
@@ -53,6 +53,28 @@ abstract class PageData {
         this.data = data;
     }
     
+    /**
+     * Find an entry by key.
+     * 
+     * @param key the key (may not exist)
+     * @return the matching or next index
+     */
+    int find(int key) {
+        int l = 0, r = entryCount;
+        while (l < r) {
+            int i = (l + r) >>> 1;
+            int k = keys[i];
+            if (k > key) {
+                r = i;
+            } else if (k == key) {
+                return i;
+            } else {
+                l = i + 1;
+            }
+        }
+        return l;
+    }
+    
     /**
      * Read the data.
      */
@@ -72,11 +94,66 @@ abstract class PageData {
      * 
      * @return the cursor
      */
-    abstract Cursor find();
+    abstract Cursor find() throws SQLException;
 
     /**
      * Write the page.
      */
     abstract void write() throws SQLException;
 
+    /**
+     * Get the key at this position.
+     * 
+     * @param index the index
+     * @return the key
+     */
+    int getKey(int index) {
+        return keys[index];
+    }
+
+    /**
+     * Split the index page at the given point.
+     * 
+     * @param session the session
+     * @param splitPoint the index where to split
+     * @return the new page that contains about half the entries
+     */
+    abstract PageData split(int splitPoint) throws SQLException;
+
+    /**
+     * Change the page id.
+     * 
+     * @param id the new page id
+     */
+    void setPageId(int id) {
+        this.pageId = id;
+    }
+
+    int getPageId() {
+        return pageId;
+    }
+
+    /**
+     * Get the last key of a page.
+     * 
+     * @return the last key
+     */
+    abstract int getLastKey() throws SQLException;
+
+    /**
+     * Get the first child leaf page of a page.
+     * 
+     * @return the page
+     */
+    abstract PageDataLeaf getFirstLeaf() throws SQLException;
+
+    /**
+     * Change the parent page id.
+     * 
+     * @param id the new parent page id
+     */
+    void setParentPageId(int id) {
+        this.parentPageId = id;
+    }
+    
 }

h2/src/main/org/h2/index/PageDataLeaf.java

@@ -15,14 +15,14 @@ import org.h2.store.DataPageBinary;
 /**
  * A leaf page that contains data of one or multiple rows.
  * Format:
- * <ul><li>0-3: parent page id
+ * <ul><li>0-3: parent page id (0 for root)
  * </li><li>4-4: page type
- * </li><li>5-5: entry count
- * </li><li>only if there is overflow: 6-9: overflow page id
- * </li><li>list of offsets (2 bytes each)
+ * </li><li>5-6: entry count
+ * </li><li>only if there is overflow: 7-10: overflow page id
+ * </li><li>list of key / offset pairs (4 bytes key, 2 bytes offset)
  * </li></ul>
  * The format of an overflow page is:
- * <ul><li>0-3: parent page id
+ * <ul><li>0-3: parent page id (0 for root)
  * </li><li>4-4: page type
  * </li><li>only if there is overflow: 5-8: next overflow page id
  * </li><li>data
@@ -57,13 +57,15 @@ class PageDataLeaf extends PageData {
     void read() throws SQLException {
         data.setPos(4);
         int type = data.readByte();
-        entryCount = data.readByte() & 255;
+        entryCount = data.readShortInt();
         offsets = new int[entryCount];
+        keys = new int[entryCount];
         rows = new Row[entryCount];
         if (type == Page.TYPE_DATA_LEAF_WITH_OVERFLOW) {
             overflowPageId = data.readInt();
         }
         for (int i = 0; i < entryCount; i++) {
+            keys[i] = data.readInt();
             offsets[i] = data.readShortInt();
         }
         start = data.length();
@@ -83,11 +85,12 @@ class PageDataLeaf extends PageData {
             type = Page.TYPE_DATA_LEAF_WITH_OVERFLOW;
         }
         data.writeByte((byte) type);
-        data.writeByte((byte) entryCount);
+        data.writeShortInt(entryCount);
         if (overflowPageId != 0) {
             data.writeInt(overflowPageId);
         }
         for (int i = 0; i < entryCount; i++) {
+            data.writeInt(keys[i]);
             data.writeShortInt(offsets[i]);
         }
         for (int i = 0; i < entryCount; i++) {
@@ -115,32 +118,65 @@ class PageDataLeaf extends PageData {
      * @return the split point of this page, or 0 if no split is required
      */
     int addRow(Row row) throws SQLException {
-        if (entryCount >= 255) {
-            return entryCount / 2;
-        }
         int rowLength = row.getByteCount(data);
         int last = entryCount == 0 ? index.getPageStore().getPageSize() : offsets[entryCount - 1];
         int offset = last - rowLength;
-        if (offset < start + 2) {
+        int[] newOffsets = new int[entryCount + 1];
+        int[] newKeys = new int[entryCount + 1];
+        Row[] newRows = new Row[entryCount + 1];
+        int x;
+        if (entryCount == 0) {
+            x = 0;
+        } else {
+            x = find(row.getPos());
+            System.arraycopy(offsets, 0, newOffsets, 0, x);
+            System.arraycopy(keys, 0, newKeys, 0, x);
+            System.arraycopy(rows, 0, newRows, 0, x);
+            if (x < entryCount) {
+                System.arraycopy(offsets, x, newOffsets, x + 1, entryCount - x);
+                System.arraycopy(keys, x, newKeys, x + 1, entryCount - x);
+                System.arraycopy(rows, x, newRows, x + 1, entryCount - x);
+            }
+        }
+        entryCount++;
+        start += 6;
+        newOffsets[x] = offset;
+        newKeys[x] = row.getPos();
+        newRows[x] = row;
+        offsets = newOffsets;
+        keys = newKeys;
+        rows = newRows;
+        if (offset < start) {
             if (entryCount > 0) {
+                int todoSplitAtLastInsertionPoint;
                 return entryCount / 2;
             }
-            offset = start + 2;
+            offset = start + 6;
             overflowPageId = index.getPageStore().allocatePage();
+            int todoWriteOverflow;
+        }
+        write();
+        return 0;
+    }
+    
+    private void removeRow(int index) throws SQLException {
+        entryCount--;
+        if (entryCount <= 0) {
+            Message.getInternalError();
         }
-        changed = true;
-        entryCount++;
         int[] newOffsets = new int[entryCount];
+        int[] newKeys = new int[entryCount];
         Row[] newRows = new Row[entryCount];
-        System.arraycopy(offsets, 0, newOffsets, 0, entryCount - 1);
-        System.arraycopy(rows, 0, newRows, 0, entryCount - 1);
-        start += 2;
-        newOffsets[entryCount - 1] = offset;
-        newRows[entryCount - 1] = row;
+        System.arraycopy(offsets, 0, newOffsets, 0, index);
+        System.arraycopy(keys, 0, newKeys, 0, index);
+        System.arraycopy(rows, 0, newRows, 0, index);
+        System.arraycopy(offsets, index + 1, newOffsets, index, entryCount - index);
+        System.arraycopy(keys, index + 1, newKeys, index, entryCount - index);
+        System.arraycopy(rows, index + 1, newRows, index, entryCount - index);
+        start -= 6;
         offsets = newOffsets;
+        keys = newKeys;
         rows = newRows;
-        write();
-        return 0;
     }
 
     Cursor find() {
@@ -158,6 +194,7 @@ class PageDataLeaf extends PageData {
         if (r == null) {
             data.setPos(offsets[index]);
             r = this.index.readRow(data);
+            r.setPos(keys[index]);
             rows[index] = r;
         }
         return r;
@@ -167,4 +204,31 @@ class PageDataLeaf extends PageData {
         return entryCount;
     }
     
+    PageData split(int splitPoint) throws SQLException {
+        int newPageId = index.getPageStore().allocatePage();
+        PageDataLeaf p2 = new PageDataLeaf(index, newPageId, parentPageId, index.getPageStore().createDataPage());
+        for (int i = splitPoint; i < entryCount;) {
+            p2.addRow(getRow(splitPoint));
+            removeRow(splitPoint);
+        }
+        return p2;
+    }
+
+    int getLastKey() throws SQLException {
+        int todoRemove;
+        return getRow(entryCount - 1).getPos();
+    }
+
+    public PageDataLeaf getNextPage() throws SQLException {
+        if (parentPageId == Page.ROOT) {
+            return null;
+        }
+        PageDataNode next = (PageDataNode) index.getPage(parentPageId);
+        return next.getNextPage(keys[entryCount - 1]);
+    }
+
+    PageDataLeaf getFirstLeaf() {
+        return this;
+    }
+
 }

h2/src/main/org/h2/index/PageDataNode.java

@@ -9,6 +9,7 @@ package org.h2.index;
 import java.sql.SQLException;
 
 import org.h2.result.Row;
+import org.h2.result.SearchRow;
 import org.h2.store.DataPageBinary;
 
 /**
@@ -16,35 +17,129 @@ import org.h2.store.DataPageBinary;
  * Format:
  * <ul><li>0-3: parent page id
  * </li><li>4-4: page type
- * </li><li>5-5: entry count
- * </li><li>6- entries: 4 bytes leaf page id, 4 bytes key
+ * </li><li>5-6: entry count
+ * </li><li>7-10: rightmost child page id
+ * </li><li>11- entries: 4 bytes leaf page id, 4 bytes key
  * </li></ul>
  */
 class PageDataNode extends PageData {
 
-    // optimization
-    // int childrenEntryCount;
+    /**
+     * The page ids of the children.
+     */
+    int[] childPageIds;
     
     PageDataNode(PageScanIndex index, int pageId, int parentPageId, DataPageBinary data) {
         super(index, pageId, parentPageId, data);
+        int todoOptimizationChildrenEntryCount;
     }
 
     void read() {
-        int todo;
+        data.setPos(5);
+        entryCount = data.readShortInt();
+        childPageIds = new int[entryCount + 1];
+        childPageIds[entryCount] = data.readInt();
+        keys = new int[entryCount];
+        for (int i = 0; i < entryCount; i++) {
+            childPageIds[i] = data.readInt();
+            keys[i] = data.readInt();
+        }
+    }
+
+    void write() throws SQLException {
+        data.reset();
+        data.writeInt(parentPageId);
+        data.writeByte((byte) Page.TYPE_DATA_NODE);
+        data.writeShortInt(entryCount);
+        data.writeInt(childPageIds[entryCount]);
+        for (int i = 0; i < entryCount; i++) {
+            data.writeInt(childPageIds[i]);
+            data.writeInt(keys[i]);
+        }
+        index.getPageStore().writePage(pageId, data);
     }
 
     int addRow(Row row) throws SQLException {
-        int todo;
+        int x = find(row.getPos());
+        PageData page = index.getPage(childPageIds[x]);
+        int splitPoint = page.addRow(row);
+        if (splitPoint == 0) {
             return 0;
         }
+        int pivot = page.getKey(splitPoint);
+        PageData page2 = page.split(splitPoint);
+        int[] newKeys = new int[entryCount + 1];
+        int[] newChildPageIds = new int[entryCount + 2];
+        System.arraycopy(keys, 0, newKeys, 0, x);
+        System.arraycopy(childPageIds, 0, newChildPageIds, 0, x);
+        if (x < entryCount) {
+            System.arraycopy(keys, x, newKeys, x + 1, entryCount - x);
+            System.arraycopy(childPageIds, x, newChildPageIds, x + 1, entryCount - x + 1);
+        }
+        newKeys[x] = pivot;
+        newChildPageIds[x] = page2.getPageId();
+        keys = newKeys;
+        childPageIds = newChildPageIds;
+        entryCount++;
+        int maxEntries = (index.getPageStore().getPageSize() - 11) / 8;
+        if (entryCount >= maxEntries) {
+            int todoSplitAtLastInsertionPoint;
+            return entryCount / 2;
+        }
+        write();
+        return 0;
+    }
+
+    Cursor find() throws SQLException {
+        int child = childPageIds[0];
+        return index.getPage(child).find();
+    }
 
-    Cursor find() {
+    PageData split(int splitPoint) throws SQLException {
         int todo;
         return null;
     }
 
-    void write() throws SQLException {
-        int todo;
+    /**
+     * Initialize the page.
+     * 
+     * @param page1 the first child page
+     * @param pivot the pivot key
+     * @param page2 the last child page
+     */
+    void init(PageData page1, int pivot, PageData page2) {
+        entryCount = 1;
+        childPageIds = new int[] { page1.getPageId(), page2.getPageId() };
+        keys = new int[] { pivot };
+    }
+    
+    int getLastKey() throws SQLException {
+        int todoRemove;
+        return index.getPage(childPageIds[entryCount]).getLastKey();
+    }
+
+    /**
+     * Get the next leaf page.
+     * 
+     * @param key the last key of the current page
+     * @return the next leaf page
+     */
+    public PageDataLeaf getNextPage(int key) throws SQLException {
+        int i = find(key) + 1;
+        if (i > entryCount) {
+            if (parentPageId == Page.ROOT) {
+                return null;
+            }
+            PageDataNode next = (PageDataNode) index.getPage(parentPageId);
+            return next.getNextPage(keys[entryCount - 1]);
+        }
+        PageData page = index.getPage(childPageIds[i]);
+        return page.getFirstLeaf();
+    }
+    
+    PageDataLeaf getFirstLeaf() throws SQLException {
+        int child = childPageIds[0];
+        return index.getPage(child).getFirstLeaf();
     }
 
 }

h2/src/main/org/h2/index/PageScanCursor.java

@@ -38,14 +38,17 @@ class PageScanCursor implements Cursor {
     }
 
     public boolean next() throws SQLException {
-        int todo;
-        if (index < current.getEntryCount()) {
+        if (index >= current.getEntryCount()) {
+            current = current.getNextPage();
+            index = 0;
+            if (current == null) {
+                return false;
+            }
+        }
         row = current.getRow(index);
         index++;
         return true;
     }
-        return false;
-    }
 
     public boolean previous() throws SQLException {
         index--;

h2/src/main/org/h2/index/PageScanIndex.java

@@ -30,6 +30,14 @@ public class PageScanIndex extends BaseIndex implements RowIndex {
     private TableData tableData;
     private int headPos;
     
+    // TODO remember last page with deleted keys (in the root page?), 
+    // and chain such pages
+    // TODO order pages so that searching for a key 
+    // doesn't seek backwards in the file
+    private int nextKey;
+    
+    // TODO remember the row count (in the root page?)
+    
     public PageScanIndex(TableData table, int id, IndexColumn[] columns, IndexType indexType, int headPos) throws SQLException {
         initBaseIndex(table, id, table.getName() + "_TABLE_SCAN", columns, indexType);
         if (database.isMultiVersion()) {
@@ -44,11 +52,11 @@ public class PageScanIndex extends BaseIndex implements RowIndex {
         if (headPos == Index.EMPTY_HEAD || headPos >= store.getPageCount()) {
             // new table
             headPos = store.allocatePage();
-            PageDataLeaf root = new PageDataLeaf(this, headPos, 1, store.createDataPage());
+            PageDataLeaf root = new PageDataLeaf(this, headPos, Page.ROOT, store.createDataPage());
             root.write();
         } else {
-            int todo;
-            rowCount = 10;
+            int todoRowCount;
+            rowCount = getPage(headPos).getLastKey();
         }
         this.headPos = headPos;
         table.setRowCount(rowCount);
@@ -56,14 +64,36 @@ public class PageScanIndex extends BaseIndex implements RowIndex {
 
 
     public void add(Session session, Row row) throws SQLException {
-        int invalidateRowCount;
+        row.setPos((int) rowCount);
         PageData root = getPage(headPos);
-        root.addRow(row);
+        int splitPoint = root.addRow(row);
+        if (splitPoint != 0) {
+            int pivot = root.getKey(splitPoint);
+            PageData page1 = root;
+            PageData page2 = root.split(splitPoint);
+            int rootPageId = root.getPageId();
+            int id = store.allocatePage();
+            page1.setPageId(id);
+            page1.setParentPageId(headPos);
+            PageDataNode newRoot = new PageDataNode(this, rootPageId, Page.ROOT, store.createDataPage());
+            newRoot.init(page1, pivot, page2);
+            page1.write();
+            page2.write();
+            newRoot.write();
+            root = newRoot;
+        }
         rowCount++;
     }
 
-    private PageData getPage(int id) throws SQLException {
+    /**
+     * Read the given page.
+     * 
+     * @param id the page id
+     * @return the page
+     */
+    PageData getPage(int id) throws SQLException {
         DataPageBinary data = store.readPage(id);
+        data.reset();
         int parentPageId = data.readInt();
         int type = data.readByte() & 255;
         PageData result;

h2/src/tools/org/h2/build/doc/dictionary.txt

@@ -572,4 +572,4 @@ localization olivier hprof jps jstack qua processor casting brasilia leap
 daylight vision declarative shape formula webapp catalina study impact
 statisticlog activeobjects manske redeployment michael kaspersky datatext
 bleyl donald conservative offsets diabetes ansorg allocating osmond gluco 
-joachim
\ No newline at end of file
+joachim gpl
\ No newline at end of file