Google Git
Sign in
apache / derby / dbed020d3949dd7707a6eca23741aa5ba155ece1 / . / java / engine / org / apache / derby / impl / store / raw / data / BasePage.java
blob: 3ffb3bb193767eb1341de2d55365d993cb5db10b [file] [log] [blame]
/*
Derby - Class org.apache.derby.impl.store.raw.data.BasePage
Licensed to the Apache Software Foundation (ASF) under one or more
contributor license agreements. See the NOTICE file distributed with
this work for additional information regarding copyright ownership.
The ASF licenses this file to you 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 org.apache.derby.impl.store.raw.data;
import org.apache.derby.iapi.reference.SQLState;
import org.apache.derby.iapi.services.io.FormatableBitSet;
import org.apache.derby.iapi.services.io.DynamicByteArrayOutputStream;
import org.apache.derby.shared.common.sanity.SanityManager;
import org.apache.derby.iapi.services.io.LimitObjectInput;
import org.apache.derby.iapi.services.io.TypedFormat;
import org.apache.derby.iapi.services.monitor.DerbyObservable;
import org.apache.derby.iapi.services.monitor.DerbyObserver;
import org.apache.derby.iapi.error.StandardException;
import org.apache.derby.iapi.store.raw.AuxObject;
import org.apache.derby.iapi.store.raw.ContainerHandle;
import org.apache.derby.iapi.store.raw.ContainerKey;
import org.apache.derby.iapi.store.raw.FetchDescriptor;
import org.apache.derby.iapi.store.raw.Page;
import org.apache.derby.iapi.store.raw.PageKey;
import org.apache.derby.iapi.store.raw.RecordHandle;
import org.apache.derby.iapi.store.raw.xact.RawTransaction;
import org.apache.derby.iapi.store.raw.log.LogInstant;
import org.apache.derby.iapi.store.access.conglomerate.LogicalUndo;
import org.apache.derby.iapi.util.InterruptStatus;
import java.io.IOException;
import java.io.OutputStream;
import java.io.ObjectInput;
/**
This class implements all the the generic locking behaviour for a Page.
It leaves method used to log and store the records up to sub-classes.
It is intended that the object can represent multiple pages from different
containers during its lifetime.
<P>
A page contains a set of records, which can be accessed by "slot",
which defines the order of the records on the page, or by "id" which
defines the identity of the records on the page. Clients access
records by both slot and id, depending on their needs.
<P>
BasePage implements Observer to watch the ContainerHandle which notifies
its Observers when it is closing.
<BR>
MT - mutable
**/
abstract class BasePage implements Page, DerbyObserver, TypedFormat
{
/**
auxiliary object
MT - mutable - content dynamic : single thread required. This reference is
set while the page is latched and returned to callers of page while the page is latched.
For correct MT behaviour it is assumed that the caller discards any reference to an
auxiliary object once the page is unlatched. The reference mya be cleared while
the page is latched, or while the page is being cleaned from the cache. In the latter
case the cache manager ensures that only a single thread can access this object.
*/
private AuxObject auxObj;
/**
this page's identity
<BR>
MT - immutable - content dynamic : single thread required
*/
protected PageKey identity;
/**
In-memory slot table, array of StoredRecordHeaders.
<BR>
MT - Immutable - Content Dynamic : Single thread required.
*/
private StoredRecordHeader[] headers; // in memory slot table
private int recordCount;
/**
Page owner during exclusive access.
MT - mutable : single thread required, provided by Lockable single thread required.
*/
protected BaseContainerHandle owner;
/**
Count of times a latch is held nested during an abort
*/
private int nestedLatch;
protected boolean inClean; // is the page being cleaned
/**
* Used to determine latch state of a page.
*
* MT - mutable
*
* There are 3 latch states for a page:
*
* UNLATCHED - (owner == null)
* PRELATCH - (owner != null) &amp;&amp; preLatch
* LATCHED - (owner != null) &amp;&amp; !preLatch
*
* A page may be "cleaned" while it is either UNLATCHED, or PRELATCH, but
* it must wait for it to be not LATCHED.
*
* A page may move from UNLATCHED to PRELATCH, while being cleaned.
* A page must wait for !inClean before it can move from PRELATCH to
* LATCHED.
**/
protected boolean preLatch;
/**
Instant of last log record that updated this page.
<BR> MT - mutable : latched
*/
private LogInstant lastLog;
/**
* The oldest version where we know that any record id that was on the
* page at that version, must still be on the page. This is used by the
* B-tree code to decide whether or not it needs to reposition when
* resuming a scan.
*/
private long repositionNeededAfterVersion;
/**
Version of the page.
<BR> MT - mutable : single thread required - The page must be latched to access
this variable or the page muts be in the noidentiy state.
*/
private long pageVersion = 0; // version of the page
/**
Status of the page
*/
private byte pageStatus;
/**
Values for pageStatus flag
page goes thru the following transition:
VALID_PAGE &lt;-&gt; deallocated page -&gt; free page &lt;-&gt; VALID_PAGE
deallocated and free page are both INVALID_PAGE as far as BasePage is concerned.
When a page is deallocated, it transitioned from VALID to INVALID.
When a page is allocated, it trnasitioned from INVALID to VALID.
*/
public static final byte VALID_PAGE = 1;
public static final byte INVALID_PAGE = 2;
/**
Init page flag.
INIT_PAGE_REUSE - set if page is being initialized for reuse
INIT_PAGE_OVERFLOW - set if page will be an overflow page
INIT_PAGE_REUSE_RECORDID - set if page is being reused and its record
id can be reset to RecordHandle.FIRST_RECORD_ID, rather
to 1+ next recordId on the page
*/
public static final int INIT_PAGE_REUSE = 0x1;
public static final int INIT_PAGE_OVERFLOW = 0x2;
public static final int INIT_PAGE_REUSE_RECORDID = 0x4;
/**
Log Record flag. Why the before image of this record is being logged
LOG_RECORD_FOR_UPDATE - set if the record is being logged for update.
LOG_RECORD_DEFAULT - for non update.
LOG_RECORD_FOR_PURGE - set if the record is being logged for purges
and no data required to ve logged.
The other cases (copy, purge, delete), we don't need to distinguish,
leave no bit set.
*/
public static final int LOG_RECORD_DEFAULT = 0x0;
public static final int LOG_RECORD_FOR_UPDATE = 0x1;
public static final int LOG_RECORD_FOR_PURGE = 0x2;
/**
** Create a new, empty page.
**/
protected BasePage()
{
}
/**
Initialized the BasePage.
<p>
Initialize the object, ie. perform work normally perfomed in
constructor. Called by setIdentity() and createIdentity().
*/
protected void initialize()
{
setAuxObject(null);
identity = null;
recordCount = 0;
clearLastLogInstant();
repositionNeededAfterVersion = 0;
if (SanityManager.DEBUG)
{
if (nestedLatch != 0)
SanityManager.THROWASSERT("nestedLatch is non-zero in initialize - value = " + nestedLatch);
if (inClean)
SanityManager.THROWASSERT("inClean is true in initialize");
if (preLatch)
SanityManager.THROWASSERT("preLatch is true in initialize");
}
}
/**
Must be called by a sub-class before calling setHeaderAtSlot.
*/
protected void initializeHeaders(int numRecords)
{
if (SanityManager.DEBUG)
{
if (recordCount != 0)
SanityManager.THROWASSERT(
"record count = " + recordCount +
" before initSlotTable is called");
}
headers = new StoredRecordHeader[numRecords];
}
/*
** Cacheable methods
*/
protected void fillInIdentity(PageKey key) {
if (SanityManager.DEBUG) {
SanityManager.ASSERT(identity == null);
SanityManager.ASSERT(repositionNeededAfterVersion == 0);
}
identity = key;
repositionNeededAfterVersion = pageVersion;
}
public void clearIdentity() {
if (SanityManager.DEBUG) {
SanityManager.ASSERT(!isLatched());
}
identity = null;
cleanPageForReuse();
}
/**
Initialized this page for reuse or first use
*/
protected void cleanPageForReuse()
{
setAuxObject(null);
recordCount = 0;
repositionNeededAfterVersion = 0;
}
/**
OK to hand object outside to cache..
*/
public Object getIdentity() {
return identity;
}
/*
** Methods of Page
*/
private static final RecordHandle InvalidRecordHandle =
new RecordId(
new PageKey(
new ContainerKey(0,0), ContainerHandle.INVALID_PAGE_NUMBER),
RecordHandle.INVALID_RECORD_HANDLE);
public final RecordHandle getInvalidRecordHandle()
{
// a static invalid record handle
return InvalidRecordHandle;
}
public static final RecordHandle MakeRecordHandle(PageKey pkey, int recordHandleConstant)
throws StandardException
{
if (recordHandleConstant >= RecordHandle.FIRST_RECORD_ID)
{
throw StandardException.newException(
SQLState.DATA_CANNOT_MAKE_RECORD_HANDLE,
recordHandleConstant);
}
return new RecordId(pkey, recordHandleConstant);
}
public final RecordHandle makeRecordHandle(int recordHandleConstant)
throws StandardException
{
return MakeRecordHandle(getPageId(), recordHandleConstant);
}
/** @see Page#getPageNumber */
public final long getPageNumber() {
if (SanityManager.DEBUG) {
SanityManager.ASSERT(isLatched(), "page is not latched.");
SanityManager.ASSERT(identity != null, "identity is null.");
}
return identity.getPageNumber();
}
public final PageKey getPageKey() {
if (SanityManager.DEBUG) {
SanityManager.ASSERT(isLatched(), "page is not latched.");
SanityManager.ASSERT(identity != null, "identity is null.");
}
return identity;
}
public final RecordHandle getRecordHandle(int recordId) {
if (SanityManager.DEBUG) {
SanityManager.ASSERT(isLatched());
}
int slot = findRecordById(recordId, FIRST_SLOT_NUMBER);
if (slot < 0)
return null;
return getRecordHandleAtSlot(slot);
}
public final RecordHandle getRecordHandleAtSlot(int slot) {
return getHeaderAtSlot(slot).getHandle(getPageId(), slot);
}
/**
@see Page#recordExists
@exception StandardException recordHandle is not a valid record handle
*/
public final boolean recordExists(RecordHandle handle, boolean ignoreDelete)
throws StandardException
{
if (SanityManager.DEBUG) {
SanityManager.ASSERT(isLatched());
}
if (handle.getId() < RecordHandle.FIRST_RECORD_ID)
{
throw StandardException.newException(
SQLState.DATA_INVALID_RECORD_HANDLE, handle);
}
if (handle.getPageNumber() != getPageNumber())
return false;
int slot = findRecordById(handle.getId(), handle.getSlotNumberHint());
return (slot >= FIRST_SLOT_NUMBER &&
(ignoreDelete || !isDeletedAtSlot(slot)));
}
public RecordHandle fetchFromSlot(
RecordHandle rh,
int slot,
Object[] row,
FetchDescriptor fetchDesc,
boolean ignoreDelete)
throws StandardException
{
if (SanityManager.DEBUG) {
SanityManager.ASSERT(isLatched());
if (rh != null)
SanityManager.ASSERT(getSlotNumber(rh) == slot);
}
checkSlotOnPage(slot);
StoredRecordHeader recordHeader = getHeaderAtSlot(slot);
if (rh == null)
rh = recordHeader.getHandle(getPageId(), slot);
if (!ignoreDelete && recordHeader.isDeleted())
return null;
/*
SanityManager.DEBUG_PRINT("fetchFromSlot", "before.");
SanityManager.showTrace(new Throwable());
SanityManager.DEBUG_PRINT("fetchFromSlot", "fetchDesc = " + fetchDesc);
if (fetchDesc != null)
{
SanityManager.DEBUG_PRINT("fetchFromSlot",
";fetchDesc.getMaxFetchColumnId() = " +
fetchDesc.getMaxFetchColumnId() +
";fetchDesc.getValidColumns() = " +
fetchDesc.getValidColumns() +
";fetchDesc.getQualifierList() = " +
fetchDesc.getQualifierList()
);
}
*/
return(
restoreRecordFromSlot(
slot, row, fetchDesc, rh, recordHeader, true) ? rh : null);
}
/**
@exception StandardException Standard Derby error policy
@see Page#fetchFieldFromSlot
*/
public final RecordHandle fetchFieldFromSlot(
int slot,
int fieldId,
Object column)
throws StandardException
{
// need to allocate row with fieldId cols because of sparse row change
// needs to be RESOLVED
Object[] row = new Object[fieldId + 1];
row[fieldId] = column;
FetchDescriptor fetchDesc =
new FetchDescriptor(fieldId + 1, fieldId);
return(fetchFromSlot(null, slot, row, fetchDesc, true));
}
/**
@exception StandardException Record does not exist on this page.
@see Page#getSlotNumber
*/
public final int getSlotNumber(RecordHandle handle)
throws StandardException
{
if (SanityManager.DEBUG) {
SanityManager.ASSERT(isLatched());
}
int slot = findRecordById(handle.getId(), handle.getSlotNumberHint());
if (slot < 0)
{
throw StandardException.newException(
SQLState.RAWSTORE_RECORD_VANISHED, handle);
}
return slot;
}
/**
@exception StandardException Record does not exist on this page.
@see Page#getNextSlotNumber
*/
public final int getNextSlotNumber(RecordHandle handle)
throws StandardException
{
if (SanityManager.DEBUG) {
SanityManager.ASSERT(isLatched());
}
int slot = findNextRecordById(handle.getId());
return slot;
}
/** @see Page#insertAtSlot
@exception StandardException Standard Derby error policy
*/
public RecordHandle insertAtSlot(
int slot,
Object[] row,
FormatableBitSet validColumns,
LogicalUndo undo,
byte insertFlag,
int overflowThreshold)
throws StandardException
{
if (SanityManager.DEBUG) {
if (overflowThreshold == 0)
SanityManager.THROWASSERT("overflowThreshold cannot be 0");
}
if ((insertFlag & Page.INSERT_DEFAULT) == Page.INSERT_DEFAULT) {
return (insertNoOverflow(slot, row, validColumns, undo, insertFlag, overflowThreshold));
} else {
if (SanityManager.DEBUG) {
if (undo != null)
SanityManager.THROWASSERT("logical undo with overflow allowed on insert " + undo.toString());
}
return (insertAllowOverflow(slot,
row, validColumns, 0, insertFlag, overflowThreshold, (RecordHandle) null));
}
}
protected RecordHandle insertNoOverflow(
int slot,
Object[] row,
FormatableBitSet validColumns,
LogicalUndo undo,
byte insertFlag,
int overflowThreshold)
throws StandardException
{
if (SanityManager.DEBUG) {
SanityManager.ASSERT(isLatched());
}
if (!owner.updateOK())
{
throw StandardException.newException(
SQLState.DATA_CONTAINER_READ_ONLY);
}
if (slot < FIRST_SLOT_NUMBER || slot > recordCount)
{
throw StandardException.newException(
SQLState.DATA_SLOT_NOT_ON_PAGE);
}
if (!allowInsert())
return null;
RawTransaction t = owner.getTransaction();
// logical operations not allowed in internal transactions.
if (undo != null) {
t.checkLogicalOperationOk();
}
int recordId;
RecordHandle handle;
do {
// loop until we get a new record id we can get a lock on.
// If we can't get the lock without waiting then assume the record
// id is owned by another xact. The current heap overflow
// algorithm makes this likely, as it first try's to insert a row
// telling raw store to fail if it doesn't fit on the page getting
// a lock on an id that never makes it to disk. The inserting
// transaction will hold a lock on this "unused" record id until
// it commits. The page can leave the cache at this point, and
// the inserting transaction has not dirtied the page (it failed
// after getting the lock but before logging anything), another
// inserting transaction will then get the same id as the
// previous inserter - thus the loop on lock waits.
//
// The lock we request indicates that this is a lock for insert,
// which the locking policy may use to perform locking concurrency
// optimizations.
recordId = newRecordIdAndBump();
handle = new RecordId(getPageId(), recordId, slot);
} while(!owner.getLockingPolicy().lockRecordForWrite(
t, handle,
true /* lock is for insert */,
false /* don't wait for grant */));
owner.getActionSet().actionInsert(t, this, slot, recordId, row, validColumns,
undo, insertFlag, 0, false, -1, (DynamicByteArrayOutputStream) null, -1, overflowThreshold);
// at this point the insert has been logged and made on the physical
// page the in-memory manipulation of the slot table is also performed
// by the PageActions object that implements actionInsert.
return handle;
}
/** @see Page#insert
@exception StandardException Standard Derby error policy
*/
public final RecordHandle insert(
Object[] row,
FormatableBitSet validColumns,
byte insertFlag,
int overflowThreshold)
throws StandardException
{
if (SanityManager.DEBUG) {
if (overflowThreshold == 0)
SanityManager.THROWASSERT("overflowThreshold much be greater than 0");
}
if (((insertFlag & Page.INSERT_DEFAULT) == Page.INSERT_DEFAULT)) {
return (insertAtSlot(recordCount, row, validColumns,
(LogicalUndo) null, insertFlag, overflowThreshold));
} else {
return (insertAllowOverflow(recordCount, row, validColumns, 0,
insertFlag, overflowThreshold, (RecordHandle) null));
}
}
/**
Insert a row allowing overflow.
If handle is supplied then the record at that hanlde will be updated
to indicate it is a partial row and it has an overflow portion.
@exception StandardException Standard Derby error policy
*/
public RecordHandle insertAllowOverflow(
int slot,
Object[] row,
FormatableBitSet validColumns,
int startColumn,
byte insertFlag,
int overflowThreshold,
RecordHandle nextPortionHandle)
throws StandardException
{
BasePage curPage = this;
if (!curPage.owner.updateOK())
{
throw StandardException.newException(
SQLState.DATA_CONTAINER_READ_ONLY);
}
// Handle of the first portion of the chain
RecordHandle headHandle = null;
RecordHandle handleToUpdate = null;
RawTransaction t = curPage.owner.getTransaction();
for (;;) {
if (SanityManager.DEBUG) {
SanityManager.ASSERT(curPage.isLatched());
}
if (!curPage.allowInsert())
return null;
// 'this' is the head page
if (curPage != this)
slot = curPage.recordCount;
boolean isLongColumns = false;
int realStartColumn = -1;
int realSpaceOnPage = -1;
DynamicByteArrayOutputStream logBuffer = null;
// allocate new record id and handle
int recordId = curPage.newRecordIdAndBump();
RecordHandle handle =
new RecordId(curPage.getPageId(), recordId, slot);
if (curPage == this) {
// Lock the row, if it is the very first portion of the record.
if (handleToUpdate == null) {
while (!owner.getLockingPolicy().lockRecordForWrite(
t, handle,
true /* lock is for insert */,
false /* don't wait for grant */)) {
// loop until we get a new record id we can get a lock
// on. If we can't get the lock without waiting then
// assume the record id is owned by another xact. The
// current heap overflow algorithm makes this likely,
// as it first try's to insert a row telling raw store
// to fail if it doesn't fit on the page getting a lock
// on an id that never makes it to disk. The
// inserting transaction will hold a lock on this
// "unused" record id until it commits. The page can
// leave the cache at this point, and the inserting
// transaction has not dirtied the page (it failed
// after getting the lock but before logging anything),
// another inserting transaction will then get the
// same id as the previous inserter - thus the loop on
// lock waits.
//
// The lock we request indicates that this is a lock
// for insert, which the locking policy may use to
// perform locking concurrency optimizations.
// allocate new record id and handle
recordId = curPage.newRecordIdAndBump();
handle =
new RecordId(curPage.getPageId(), recordId, slot);
}
}
headHandle = handle;
}
do {
// do this loop at least once. If we caught a long Column,
// then, we redo the insert with saved logBuffer.
try {
startColumn =
owner.getActionSet().actionInsert(
t, curPage, slot, recordId,
row, validColumns, (LogicalUndo) null,
insertFlag, startColumn, false,
realStartColumn, logBuffer, realSpaceOnPage,
overflowThreshold);
isLongColumns = false;
} catch (LongColumnException lce) {
// we caught a long column exception
// three things should happen here:
// 1. insert the long column into overflow pages.
// 2. append the overflow field header in the main chain.
// 3. continue the insert in the main data chain.
logBuffer = new DynamicByteArrayOutputStream(lce.getLogBuffer());
// step 1: insert the long column ... use the same
// insertFlag as the rest of the row.
RecordHandle longColumnHandle =
insertLongColumn(curPage, lce, insertFlag);
// step 2: append the overflow field header to the log buffer
int overflowFieldLen = 0;
try {
overflowFieldLen +=
appendOverflowFieldHeader((DynamicByteArrayOutputStream)logBuffer, longColumnHandle);
} catch (IOException ioe) {
// YYZ: revisit... ioexception, insert failed...
return null;
}
// step 3: continue the insert in the main data chain
// need to pass the log buffer, and start column to the next insert.
realStartColumn = lce.getNextColumn() + 1;
realSpaceOnPage = lce.getRealSpaceOnPage() - overflowFieldLen;
isLongColumns = true;
}
} while (isLongColumns);
if (handleToUpdate != null) {
// update the recordheader on the previous page
updateOverflowDetails(handleToUpdate, handle);
}
// all done
if (startColumn == -1) {
if (curPage != this)
curPage.unlatch();
if (nextPortionHandle != null) {
// need to update the overflow details of the last portion
// to point to the existing portion
updateOverflowDetails(handle, nextPortionHandle);
}
return headHandle;
}
handleToUpdate = handle;
BasePage nextPage =
curPage.getOverflowPageForInsert(
slot, row, validColumns,startColumn);
if (curPage != this)
curPage.unlatch();
curPage = nextPage;
}
}
/**
* Routine to insert a long column.
* <p>
* This code inserts a long column as a linked list of rows on overflow
* pages. This list is pointed to by a small pointer in the main page
* row column. The operation does the following:
* allocate new overflow page
* insert single row filling overflow page
* while (more of column exists)
* allocate new overflow page
* insert single row with next piece of row
* update previous piece to point to this new piece of row
*
* Same code is called both from an initial insert of a long column and
* from a subsequent update that results in a long column.
*
* @return The recordHandle of the first piece of the long column chain.
*
* @param mainChainPage The parent page with row piece containing column
* that will eventually point to this long column
* chain.
* @param lce The LongColumnException thrown when we recognized
* that the column being inserted was "long", this
* structure is used to cache the info that we have
* read so far about column. In the case of an insert
* of the stream it will have a copy of just the first
* page of the stream that has already been read once.
* @param insertFlag flags for insert operation.
*
*
* @exception StandardException Standard exception policy.
**/
protected RecordHandle insertLongColumn(
BasePage mainChainPage,
LongColumnException lce,
byte insertFlag)
throws StandardException
{
Object[] row = new Object[1];
row[0] = lce.getColumn();
RecordHandle firstHandle = null;
RecordHandle handle = null;
RecordHandle prevHandle = null;
BasePage curPage = mainChainPage;
BasePage prevPage = null;
boolean isFirstPage = true;
// undo inserts as purges of all pieces of the overflow column
// except for the 1st overflow page pointed at by the main row.
//
// Consider a row with one column which is a long column
// that takes 2 pages for itself plus an entry in main parent page.
// the log records in order for this look something like:
// insert overflow page 1
// insert overflow page 2
// update overflow page 1 record to have pointer to overflow page 2
// insert main row (which has pointer to overflow page 1)
//
// If this insert gets aborted then something like the following
// happens:
// main row is marked deleted (but ptr to overflow 1 still exists)
// update is aborted so link on page 2 to page 1 is lost
// overflow row on page 2 is marked deleted
// overflow row on page 1 is marked deleted
//
// There is no way to reclaim page 2 later as the abort of the update
// has now lost the link from overflow page 1 to overflow 2, so
// the system has to do it as part of the abort of the insert. But,
// it can't for page 1 as the main page will attempt to follow
// it's link in the deleted row during it's space reclamation and it
// can't tell the difference
// between a row that has been marked deleted as part of an aborted
// insert or as part of a committed delete. When it follows the link
// it could find no page and that could be coded against, but it could
// be that the page is now used by some other overflow row which would
// lead to lots of different kinds of problems.
//
// So the code leaves the 1st overflow page to be cleaned up with the
// main page row is purged, but goes ahead and immediately purges all
// the segments that will be lost as part of the links being lost due
// to aborted updates.
byte after_first_page_insertFlag =
(byte) (insertFlag | Page.INSERT_UNDO_WITH_PURGE);
// when inserting a long column startColumn is just used
// as a flag. -1 means the insert is complete, != -1 indicates
// more inserts are required.
int startColumn = 0;
RawTransaction t = curPage.owner.getTransaction();
do {
// in this loop, we do 3 things:
// 1. get a new overflow page
// 2. insert portion of a long column
// 3. update previous handle, release latch on previous page
if (!isFirstPage) {
prevPage = curPage;
prevHandle = handle;
}
// step 1. get a new overflow page
curPage = (BasePage) getNewOverflowPage();
if (SanityManager.DEBUG) {
SanityManager.ASSERT(curPage.isLatched());
SanityManager.ASSERT(curPage.allowInsert());
}
int slot = curPage.recordCount;
int recordId = curPage.newRecordId();
handle = new RecordId(curPage.getPageId(), recordId, slot);
if (isFirstPage)
firstHandle = handle;
// step 2: insert column portion
startColumn =
owner.getActionSet().actionInsert(
t, curPage, slot, recordId, row, (FormatableBitSet)null,
(LogicalUndo) null,
(isFirstPage ? insertFlag : after_first_page_insertFlag),
startColumn, true, -1, (DynamicByteArrayOutputStream) null,
-1, 100);
// step 3: if it is not the first page, update previous page,
// then release latch on prevPage
if (!isFirstPage) {
// for the previous page, add an overflow field header,
// and update the record header to show 2 fields
prevPage.updateFieldOverflowDetails(prevHandle, handle);
prevPage.unlatch();
prevPage = null;
} else {
isFirstPage = false;
}
} while (startColumn != (-1)) ;
if (curPage != null) {
curPage.unlatch();
curPage = null;
}
return (firstHandle);
}
/**
The page or its header is about to be modified.
Loggable actions use this to make sure the page gets cleaned if a
checkpoint is taken after any log record is sent to the log stream but
before the page is actually dirtied.
*/
public abstract void preDirty();
/**
Update the overflow pointer for a long row
<BR> MT - latched - page latch must be held
@param handle handle of the record for long row
@param overflowHandle the overflow (continuation) pointer for the long row
@exception StandardException Standard Derby error policy
*/
public abstract void updateOverflowDetails(RecordHandle handle, RecordHandle overflowHandle)
throws StandardException;
/**
Update the overflow pointer for a long column
<BR> MT - latched - page latch must be held
@param handle handle of the record for long row
@param overflowHandle the overflow (continuation) pointer for the long row
@exception StandardException Standard Derby error policy
*/
public abstract void updateFieldOverflowDetails(RecordHandle handle, RecordHandle overflowHandle)
throws StandardException;
/**
Append an overflow pointer to a partly logged row,
to point to a long column that just been logged.
<BR> MT - latched - page latch must be held
@param logBuffer The buffer that contains the partially logged row.
@param overflowHandle the overflow (continuation) pointer
to the beginning of the long column
@exception StandardException Standard Derby error policy
*/
public abstract int appendOverflowFieldHeader(DynamicByteArrayOutputStream logBuffer, RecordHandle overflowHandle)
throws StandardException, IOException;
public abstract BasePage getOverflowPageForInsert(
int slot,
Object[] row,
FormatableBitSet validColumns,
int startColumn)
throws StandardException;
protected abstract BasePage getNewOverflowPage()
throws StandardException;
/** @see Page#updateAtSlot
@exception StandardException Standard Derby error policy
@exception StandardException StandardException.newException(SQLState.UPDATE_DELETED_RECORD
if the record is already deleted
@exception StandardException StandardException.newException(SQLState.CONTAINER_READ_ONLY
if the container is read only
*/
public final RecordHandle updateAtSlot(
int slot,
Object[] row,
FormatableBitSet validColumns)
throws StandardException
{
if (SanityManager.DEBUG) {
SanityManager.ASSERT(isLatched());
}
if (!owner.updateOK())
{
throw StandardException.newException(
SQLState.DATA_CONTAINER_READ_ONLY);
}
if (isDeletedAtSlot(slot))
{
throw StandardException.newException(
SQLState.DATA_UPDATE_DELETED_RECORD);
}
RecordHandle handle = getRecordHandleAtSlot(slot);
RawTransaction t = owner.getTransaction();
doUpdateAtSlot(t, slot, handle.getId(), row, validColumns);
return handle;
}
public abstract void doUpdateAtSlot(
RawTransaction t,
int slot,
int id,
Object[] row,
FormatableBitSet validColumns)
throws StandardException;
/** @see Page#updateFieldAtSlot
@exception StandardException Standard Derby error policy
@exception StandardException StandardException.newException(SQLState.UPDATE_DELETED_RECORD
if the record is already deleted
@exception StandardException StandardException.newException(SQLState.CONTAINER_READ_ONLY
if the container is read only
*/
public RecordHandle updateFieldAtSlot(
int slot,
int fieldId,
Object newValue,
LogicalUndo undo)
throws StandardException
{
if (SanityManager.DEBUG) {
SanityManager.ASSERT(isLatched());
SanityManager.ASSERT(newValue != null);
}
if (!owner.updateOK())
{
throw StandardException.newException(
SQLState.DATA_CONTAINER_READ_ONLY);
}
if (isDeletedAtSlot(slot))
{
throw StandardException.newException(
SQLState.DATA_UPDATE_DELETED_RECORD);
}
RawTransaction t = owner.getTransaction();
RecordHandle handle = getRecordHandleAtSlot(slot);
owner.getActionSet().actionUpdateField(t, this, slot,
handle.getId(), fieldId, newValue, undo);
return handle;
}
/** @see Page#fetchNumFields
@exception StandardException Standard exception policy.
*/
public final int fetchNumFields(RecordHandle handle)
throws StandardException
{
if (SanityManager.DEBUG) {
SanityManager.ASSERT(isLatched());
}
return fetchNumFieldsAtSlot(getSlotNumber(handle));
}
/** @see Page#fetchNumFieldsAtSlot
@exception StandardException Standard exception policy.
*/
public int fetchNumFieldsAtSlot(int slot)
throws StandardException
{
if (SanityManager.DEBUG) {
SanityManager.ASSERT(isLatched());
}
return getHeaderAtSlot(slot).getNumberFields();
}
/**
@see Page#deleteAtSlot
@param slot the slot number
@param delete true if this record is to be deleted, false if this
deleted record is to be marked undeleted
@param undo logical undo logic if necessary
@exception StandardException Standard exception policy.
@exception StandardException StandardException.newException(SQLState.UPDATE_DELETED_RECORD
if an attempt to delete a record that is already deleted
@exception StandardException StandardException.newException(SQLState.UNDELETE_RECORD
if an attempt to undelete a record that is not deleted
*/
public RecordHandle deleteAtSlot(int slot, boolean delete, LogicalUndo undo)
throws StandardException
{
if (SanityManager.DEBUG) {
SanityManager.ASSERT(isLatched());
}
if (!owner.updateOK())
{
throw StandardException.newException(
SQLState.DATA_CONTAINER_READ_ONLY);
}
if (delete)
{
if (isDeletedAtSlot(slot))
{
throw StandardException.newException(
SQLState.DATA_UPDATE_DELETED_RECORD);
}
}
else // undelete a deleted record
{
if (!isDeletedAtSlot(slot))
{
throw StandardException.newException(
SQLState.DATA_UNDELETE_RECORD);
}
}
RawTransaction t = owner.getTransaction();
// logical operations not allowed in internal transactions.
if (undo != null) {
t.checkLogicalOperationOk();
}
RecordHandle handle = getRecordHandleAtSlot(slot);
owner.getActionSet().actionDelete(t, this, slot, handle.getId(), delete, undo);
// delete/undelete the record in the stored version
// and in the in memory version performed by the PageActions item
return handle;
}
/**
Purge one or more rows on a non-overflow page.
@see Page#purgeAtSlot
@exception StandardException Standard exception policy.
*/
public void purgeAtSlot(int slot, int numpurges, boolean needDataLogged)
throws StandardException
{
if (SanityManager.DEBUG)
{
SanityManager.ASSERT(isLatched());
if (isOverflowPage())
SanityManager.THROWASSERT(
"purge committed deletes on an overflow page. Page = " +
this);
}
if (numpurges <= 0)
return;
if (!owner.updateOK())
{
throw StandardException.newException(
SQLState.DATA_CONTAINER_READ_ONLY);
}
if ((slot < 0) || ((slot+numpurges) > recordCount))
{
throw StandardException.newException(
SQLState.DATA_SLOT_NOT_ON_PAGE);
}
RawTransaction t = owner.getTransaction();
// lock the records to be purged
int[] recordIds = new int[numpurges];
PageKey pageId = getPageId(); // RESOLVE: MT problem ?
for (int i = 0; i < numpurges; i++)
{
recordIds[i] = getHeaderAtSlot(slot + i).getId();
// get row lock on head row piece
RecordHandle handle = getRecordHandleAtSlot(slot);
owner.getLockingPolicy().lockRecordForWrite(t, handle, false, true);
// Before we purge these rows, we need to make sure they don't have
// overflow rows and columns. Only clean up long rows and long
// columns if this is not a temporary container, otherwise, just
// lose the space.
if (owner.isTemporaryContainer() || entireRecordOnPage(slot+i))
continue;
// row[slot+i] has overflow rows and/or long columns, reclaim
// them in a loop.
RecordHandle headRowHandle =
getHeaderAtSlot(slot+i).getHandle(pageId, slot+i);
purgeRowPieces(t, slot+i, headRowHandle, needDataLogged);
}
owner.getActionSet().actionPurge(t, this, slot, numpurges, recordIds, needDataLogged);
}
/**
Purge all the overflow columns and overflow rows of the record at slot.
@exception StandardException Standard exception policy.
*/
protected abstract void purgeRowPieces(RawTransaction t, int slot,
RecordHandle headRowHandle,
boolean needDataLogged)
throws StandardException;
/** @see Page#copyAndPurge
@exception StandardException Standard exception policy.
*/
public void copyAndPurge(Page destPage, int src_slot, int num_rows, int dest_slot)
throws StandardException
{
if (SanityManager.DEBUG) {
SanityManager.ASSERT(isLatched());
}
if (num_rows <= 0)
{
throw StandardException.newException(SQLState.DATA_NO_ROW_COPIED);
}
if (!owner.updateOK())
{
throw StandardException.newException(
SQLState.DATA_CONTAINER_READ_ONLY);
}
if ((src_slot < 0) || ((src_slot + num_rows) > recordCount))
{
throw StandardException.newException(
SQLState.DATA_SLOT_NOT_ON_PAGE);
}
if (SanityManager.DEBUG) {
// first copy into the destination page, let it do the work
// if no problem, then purge from this page
SanityManager.ASSERT((destPage instanceof BasePage), "must copy from BasePage to BasePage");
}
BasePage dpage = (BasePage)destPage;
// make sure they are from the same container - this means they are of
// the same size and have the same page and record format.
PageKey pageId = getPageId(); // RESOLVE: MT problem ?
if (!pageId.getContainerId().equals(dpage.getPageId().getContainerId()))
{
throw StandardException.newException(
SQLState.DATA_DIFFERENT_CONTAINER,
pageId.getContainerId(),
dpage.getPageId().getContainerId());
}
int[] recordIds = new int[num_rows];
RawTransaction t = owner.getTransaction();
// lock the records to be purged and calculate total space needed
for (int i = 0; i < num_rows; i++)
{
RecordHandle handle = getRecordHandleAtSlot(src_slot+i);
owner.getLockingPolicy().lockRecordForWrite(t, handle, false, true);
recordIds[i] = getHeaderAtSlot(src_slot + i).getId();
}
// first copy num_rows into destination page
dpage.copyInto(this, src_slot, num_rows, dest_slot);
// Now purge num_rows from this page
// Do NOT purge overflow rows, if it has such a thing. This operation
// is called by split and if the key has overflow, spliting the head
// page does not copy over the remaining pieces, i.e.,the new head page
// still points to those pieces.
owner.getActionSet().actionPurge(
t, this, src_slot, num_rows, recordIds, true);
}
/**
Unlatch the page.
@see Page#unlatch
*/
public void unlatch() {
if (SanityManager.DEBUG) {
SanityManager.ASSERT(isLatched(),
"unlatch() attempted on page that is not latched.");
}
releaseExclusive();
}
/**
* Check whether the page is latched.
*
* @return <code>true</code> if the page is latched, <code>false</code>
* otherwise
* @see Page#isLatched
*/
public final synchronized boolean isLatched() {
if (SanityManager.DEBUG) {
SanityManager.ASSERT(identity != null);
}
return owner != null;
}
/** @see Page#recordCount */
public final int recordCount() {
if (SanityManager.DEBUG) {
SanityManager.ASSERT(
isLatched(), "page not latched on call to recordCount()");
}
return recordCount;
}
/**
get record count without checking for latch
*/
protected abstract int internalDeletedRecordCount();
/**
get record count without checking for latch
*/
protected int internalNonDeletedRecordCount()
{
// deallocated or freed page, don't count
if (pageStatus != VALID_PAGE)
return 0;
int deletedCount = internalDeletedRecordCount();
if (deletedCount == -1) {
int count = 0;
int maxSlot = recordCount;
for (int slot = FIRST_SLOT_NUMBER ; slot < maxSlot; slot++) {
if (!isDeletedOnPage(slot))
count++;
}
return count;
} else {
if (SanityManager.DEBUG) {
int delCount = 0;
int maxSlot = recordCount;
for (int slot = FIRST_SLOT_NUMBER ; slot < maxSlot; slot++) {
if (recordHeaderOnDemand(slot).isDeleted())
delCount++;
}
if (delCount != deletedCount)
{
SanityManager.THROWASSERT(
"incorrect deleted row count. Should be: " + delCount +
", instead got: " + deletedCount +
", maxSlot = " + maxSlot +
", recordCount = " + recordCount +
"\npage = " + this);
}
}
return (recordCount - deletedCount);
}
}
/** @see Page#nonDeletedRecordCount */
public int nonDeletedRecordCount() {
if (SanityManager.DEBUG) {
SanityManager.ASSERT(isLatched());
}
return internalNonDeletedRecordCount();
}
/**
* Is this page/deleted row a candidate for immediate reclaim space.
* <p>
* Used by access methods after executing a delete on "slot_just_deleted"
* to ask whether a post commit should be queued to try to reclaim space
* after the delete commits.
* <p>
* Also used by access methods after undo of an insert.
* <p>
* Will return true if the number of non-deleted rows on the page is
* &lt;= "num_non_deleted_rows". For instance 0 means schedule reclaim
* only if all rows are deleted, 1 if all rows but one are deleted.
* <p>
* Will return true if the row just deleted is either a long row or long
* column. In this case doing a reclaim space on the single row may
* reclaim multiple pages of free space, so better to do it now rather
* than wait for all rows on page to be deleted. This case is to address
* the worst case scenario of all rows with long columns, but very short
* rows otherwise. In this case there could be 1000's of rows on the
* main page with many gigabytes of data on overflow pages in deleted space
* that would not be reclaimed until all rows on the page were deleted.
*
* @return true if a reclaim space should be scheduled post commit on this
* page, false otherwise.
*
* @param num_non_deleted_rows threshold number of non-deleted rows to
* schedule reclaim space.
* @param slot_just_deleted row on page to check for long row/long column
*
* @exception StandardException Standard exception policy.
**/
public boolean shouldReclaimSpace(
int num_non_deleted_rows,
int slot_just_deleted)
throws StandardException
{
if (SanityManager.DEBUG)
{
SanityManager.ASSERT(isLatched());
}
boolean ret_val = false;
if (!isOverflowPage())
{
// only return true for non-overflow pages. Overflow pages
// will be reclaimed as part of reclaiming the rows on the
// head pages.
if (internalNonDeletedRecordCount() <= num_non_deleted_rows)
{
// all user rows on this page are marked deleted, so
// reclaim this page.
ret_val = true;
}
else
{
// always reclaim if there rows span more than one page,
// this picks up long rows and blob/clob columns. This
// tells us likely to reclaim 1 or more pages.
if (!entireRecordOnPage(slot_just_deleted))
{
ret_val = true;
}
}
}
return(ret_val);
}
// no need to check for slot on page, call already checked
protected final boolean isDeletedOnPage(int slot)
{
return getHeaderAtSlot(slot).isDeleted();
}
/** @see Page#isDeletedAtSlot
@exception StandardException Standard exception policy.
*/
public boolean isDeletedAtSlot(int slot)
throws StandardException
{
if (SanityManager.DEBUG) {
SanityManager.ASSERT(isLatched());
}
checkSlotOnPage(slot);
return isDeletedOnPage(slot);
}
/**
Set the aux object.
<BR> MT - single thread required. Calls via the Page interface will have the
page latched, thus providing single threadedness. Otherwise calls via this class
are only made when the class has no-identity, thus only a single thread can see the object.
@see Page#setAuxObject
*/
public void setAuxObject(AuxObject obj)
{
if (SanityManager.DEBUG) {
SanityManager.ASSERT((identity == null) || isLatched());
}
if (auxObj != null) {
auxObj.auxObjectInvalidated();
}
auxObj = obj;
}
/**
Get the aux object.
<BR> MT - latched - It is required the caller throws away the returned reference
when the page is unlatched.
@see Page#getAuxObject
*/
public AuxObject getAuxObject()
{
if (SanityManager.DEBUG) {
SanityManager.ASSERT(isLatched());
}
return auxObj;
}
/**
* Set a hint in this page to make B-tree scans positioned on it
* reposition before they continue. This method is typically called
* when rows are removed from a B-tree leaf page (for instance in a
* page split).
*/
public void setRepositionNeeded() {
if (SanityManager.DEBUG) {
SanityManager.ASSERT(isLatched());
}
repositionNeededAfterVersion = getPageVersion();
}
/**
* Check if a B-tree scan positioned on this page needs to reposition.
*
* @param version the last version on which the B-tree scan had a valid
* position on this page
* @return {@code true} if a repositioning is needed because the row
* on the current position may have been removed from this page after
* the specified version; {@code false} otherwise
*/
public boolean isRepositionNeeded(long version) {
if (SanityManager.DEBUG) {
SanityManager.ASSERT(isLatched());
}
return repositionNeededAfterVersion > version;
}
/*
** Methods of Observer.
*/
/**
This object is set to observe the BaseContainerHandle it was obtained by,
that handle will notify its observers when it is being closed. In that case
we will release the latch on the page held by that container.
<BR>
MT - latched
@see DerbyObserver#update
*/
public void update(DerbyObservable obj, Object arg) {
if (SanityManager.DEBUG) {
SanityManager.ASSERT(isLatched());
SanityManager.ASSERT(obj == owner);
}
releaseExclusive();
}
/*
** Implementation specific methods
*/
/**
Get the Page identifer
<BR> MT - RESOLVE
*/
public final PageKey getPageId() {
if (SanityManager.DEBUG) {
SanityManager.ASSERT(identity != null);
}
return identity;
}
/**
Get an exclusive latch on the page.
<BR>
MT - thread safe
@exception StandardException Standard Derby policy.
*/
void setExclusive(BaseContainerHandle requester)
throws StandardException {
RawTransaction t = requester.getTransaction();
// In some cases latches are held until after a commit or an abort
// (currently internal and nested top transactions.
// If this is the case then during an abort a latch
// request will be made for a latch that is already held.
// We do not allow the latch to be obtained multiple times
// because i) latches are exclusive for simplicity and
// performance, ii) holding a page latched means that the page is
// on the container handle's obervers list, if we latched it twice
// then the paeg would have to be on the list twice, which is not supported
// since the page has value equality. To be on the list twice reference
// equality would be required, which would mean pushing a ReferenceObservable
// object for every latch; iii) other unknown reasons :-)
synchronized (this)
{
// need synchronized block because owner may be set to null in the
// middle if another thread is in the process of unlatching the
// page
if ((owner != null) && (t == owner.getTransaction())) {
if (t.inAbort()) {
//
nestedLatch++;
return;
}
// A transaction tries to double latch the page while not in
// abort: if we let it wait, it would hang forever, so it's
// better to throw an exception. This should never occur unless
// the code is wrong. If it still does, we'd better tear down
// the connection just in case, so the error has session level
// severity;
throw StandardException.newException(
SQLState.DATA_DOUBLE_LATCH_INTERNAL_ERROR, identity);
}
while (owner != null) {
try {
// Expect notify from releaseExclusive().
wait();
} catch (InterruptedException ie) {
InterruptStatus.setInterrupted();
}
}
preLatch(requester);
// latch granted, but cleaner may "own" the page.
if (SanityManager.DEBUG) {
SanityManager.ASSERT(isLatched(), "page not latched");
}
// Latch will be granted, even if cleaner "owns" the page.
// Wait here unil cleaner is done. This is safe as now we own the
// latch, and have yet to do anything to the in-memory data
// structures.
//
// Previously we would wait in lockEvent, but that caused the code
// to block on I/O while holding the locking system monitor.
while (inClean)
{
try
{
// Expect notify from clean() routine.
wait();
}
catch (InterruptedException ie)
{
InterruptStatus.setInterrupted();
}
}
// no clean taking place, so safe to move to full LATCHED state.
preLatch = false;
}
}
/**
Get an exclusive latch on the page, but only if I don't have to wait.
<BR>
MT - thread safe
*/
boolean setExclusiveNoWait(BaseContainerHandle requester) throws StandardException {
RawTransaction t = requester.getTransaction();
// comment in setExclusive()
synchronized (this)
{
if ((owner != null) && (t == owner.getTransaction())) {
if (t.inAbort()) {
//
nestedLatch++;
return true;
}
}
// Pre-latch the page if no one already has latched it or requested
// a latch. Otherwise, give up and return false.
if (owner == null) {
preLatch(requester);
} else {
return false;
}
// Latch will be granted, even if cleaner "owns" the page.
// Wait here unil cleaner is done. This is safe as now we own the
// latch, and have yet to do anything to the in-memory data
// structures.
//
// Previously we would wait in lockEvent, but that caused the code
// to block on I/O while holding the locking system monitor.
while (inClean)
{
//if (SanityManager.DEBUG)
// SanityManager.DEBUG_PRINT("setExclusiveNoWait", "in while loop.");
try
{
// Expect notify from clean() routine.
wait();
}
catch (InterruptedException ie)
{
InterruptStatus.setInterrupted();
}
}
// no clean taking place, so safe to move to full LATCHED state.
preLatch = false;
}
if (SanityManager.DEBUG) {
SanityManager.ASSERT(isLatched(), "page not latched");
}
return true;
}
/**
Release the exclusive latch on the page.
<BR>
MT - latched
*/
protected synchronized void releaseExclusive() {
if (SanityManager.DEBUG) {
if (!isLatched())
{
SanityManager.THROWASSERT(
"releaseExclusive failed, nestedLatch = " + nestedLatch);
}
}
if (nestedLatch > 0) {
nestedLatch--;
return;
}
owner.deleteObserver(this);
owner = null;
notifyAll();
}
/**
* Move page state from UNLATCHED to PRELATCH. setExclusive*() routines do
* the work of completing the latch - using the preLatch status.
*
* @param requester handle to the container requesting ownership
*/
private void preLatch(BaseContainerHandle requester) {
if (SanityManager.DEBUG) {
SanityManager.ASSERT(!isLatched(),
"Attempted to pre-latch a latched page");
}
owner = requester;
// make sure the latch is released if the container is closed
requester.addObserver(this);
preLatch = true;
}
/*
** Manipulation of the in-memory version of the slot table.
*/
/**
Must be called by any non-abstract sub-class to initialise the slot
table.
*/
protected final void setHeaderAtSlot(int slot, StoredRecordHeader rh) {
if (slot < headers.length)
{
// check that array "cache" of headers is big enough.
if (rh != null)
{
headers[slot] = rh;
}
}
else
{
// need to grow the array, just allocate new array and copy.
StoredRecordHeader[] new_headers = new StoredRecordHeader[slot + 1];
System.arraycopy(headers, 0, new_headers, 0, headers.length);
headers = new_headers;
headers[slot] = rh;
}
}
protected final void bumpRecordCount(int number) {
recordCount += number;
}
public final StoredRecordHeader getHeaderAtSlot(int slot) {
if (slot < headers.length)
{
StoredRecordHeader rh = headers[slot];
return((rh != null) ? rh : recordHeaderOnDemand(slot));
}
else
{
return recordHeaderOnDemand(slot);
}
}
/**
Returns true if the entire record of that slot fits inside of this
page. Returns false if part of the record on this slot overflows to
other pages, either due to long row or long column.
<BR>
MT - latched
@exception StandardException Standard Derby error policy
*/
public abstract boolean entireRecordOnPage(int slot)
throws StandardException;
public abstract StoredRecordHeader recordHeaderOnDemand(int slot);
/**
Is the given slot number on the page?
<BR>
MT - latched
*/
private final void checkSlotOnPage(int slot)
throws StandardException {
if (SanityManager.DEBUG) {
SanityManager.ASSERT(isLatched());
}
if (slot >= FIRST_SLOT_NUMBER && slot < recordCount)
{
return;
}
throw StandardException.newException(SQLState.DATA_SLOT_NOT_ON_PAGE);
}
/**
Mark the record at the passed in slot as deleted.
return code comes from StoredRecordHeader class:
return 1, if delete status from not deleted to deleted
return -1, if delete status from deleted to not deleted
return 0, if status unchanged.
<BR>
<B>Any sub-class must call this method when deleting a record.</B>
<BR>
MT - latched
@exception StandardException Standard Derby error policy
@exception IOException IO error accessing page
*/
public int setDeleteStatus(int slot, boolean delete) throws StandardException, IOException {
if (SanityManager.DEBUG) {
// latch check performed in checkSlotOnPage
checkSlotOnPage(slot);;
}
return (getHeaderAtSlot(slot).setDeleted(delete));
}
/**
Mark this page as being deallocated
@exception StandardException Derby Standard error policy
*/
public void deallocatePage() throws StandardException
{
if (SanityManager.DEBUG) {
SanityManager.ASSERT(isLatched());
}
if (!owner.updateOK())
{
throw StandardException.newException(
SQLState.DATA_CONTAINER_READ_ONLY);
}
RawTransaction t = owner.getTransaction();
owner.getActionSet().actionInvalidatePage(t, this);
}
/**
Mark this page as being allocated and initialize it to a pristine page
@exception StandardException Derby Standard error policy
*/
public void initPage(int initFlag, long pageOffset)
throws StandardException
{
if (SanityManager.DEBUG) {
SanityManager.ASSERT(isLatched());
}
if (!owner.updateOK())
{
throw StandardException.newException(
SQLState.DATA_CONTAINER_READ_ONLY);
}
RawTransaction t = owner.getTransaction();
owner.getActionSet().actionInitPage(
t, this, initFlag, getTypeFormatId(), pageOffset);
}
/**
Find the slot for the record with the passed in identifier.
<BR>
This method returns the record regardless of its deleted status.
<BR>
The "slotHint" argument is a hint about what slot the record id might
be in. Callers may save the last slot where the record was across
latch/unlatches to the page, and then pass that slot back as a hint -
if the page has not shuffled slots since the last reference then the
hint will succeed and a linear search is saved. If the caller has
no idea where it may be, then FIRST_SLOT_NUMBER is passed in and a
linear search is performed.
<BR>
MT - latched
@param recordId record id of the record to search for.
@param slotHint "hint" about which slot the record might be in.
*/
public int findRecordById(int recordId, int slotHint) {
if (SanityManager.DEBUG) {
SanityManager.ASSERT(isLatched());
}
if (slotHint == FIRST_SLOT_NUMBER)
slotHint = recordId - RecordHandle.FIRST_RECORD_ID;
int maxSlot = recordCount();
if ((slotHint > FIRST_SLOT_NUMBER) &&
(slotHint < maxSlot) &&
(recordId == getHeaderAtSlot(slotHint).getId())) {
return(slotHint);
} else {
for (int slot = FIRST_SLOT_NUMBER; slot < maxSlot; slot++) {
if (recordId == getHeaderAtSlot(slot).getId()) {
return slot;
}
}
}
return -1;
}
/**
Find the slot for the first record on the page with an id greater than
the passed in identifier.
<BR>
Returns the slot of the first record on the page with an id greater
than the one passed in. Usefulness of this functionality depends on the
clients use of the raw store interfaces. If all "new" records are
always inserted at the end of the page, and the raw store continues
to guarantee that all record id's will be allocated in increasing order
on a given page, then a page is always sorted
in record id order. For instance current heap tables function this
way. If the client ever inserts at a particular slot number, rather
than at the "end" then the record id's will not be sorted.
<BR>
In the case where all record id's are always sorted on a page, then
this routine can be used by scan's which "lose" their position because
the row they have as a position was purged. They can reposition their
scan at the "next" row after the row that is now missing from the table.
<BR>
This method returns the record regardless of its deleted status.
<BR>
MT - latched
@param recordId record id of the first record on the page with a
record id higher than the one passed in. If no
such record exists, -1 is returned.
*/
private int findNextRecordById(int recordId)
{
if (SanityManager.DEBUG)
{
SanityManager.ASSERT(isLatched());
}
int maxSlot = recordCount();
for (int slot = FIRST_SLOT_NUMBER; slot < maxSlot; slot++)
{
if (getHeaderAtSlot(slot).getId() > recordId)
{
return(slot);
}
}
return -1;
}
/**
Copy num_rows from srcPage, src_slot into this page starting at dest_slot.
This is destination page of the the copy half of copy and Purge.
@see Page#copyAndPurge
*/
private void copyInto(BasePage srcPage, int src_slot, int num_rows,
int dest_slot)
throws StandardException
{
if ((dest_slot < 0) || dest_slot > recordCount)
{
throw StandardException.newException(
SQLState.DATA_SLOT_NOT_ON_PAGE);
}
RawTransaction t = owner.getTransaction();
// get num_rows row locks, need to predict what those recordIds will be
int[] recordIds = new int[num_rows];
PageKey pageId = getPageId(); // RESOLVE - MT problem ?
// get new recordIds for the rows from this page
// RESOLVE: we should also record the amount of reserved space
for (int i = 0; i < num_rows; i++)
{
if (i == 0)
recordIds[i] = newRecordId();
else
recordIds[i] = newRecordId(recordIds[i-1]);
RecordHandle handle = new RecordId(pageId, recordIds[i], i);
owner.getLockingPolicy().lockRecordForWrite(t, handle, false, true);
}
// RESOLVE: need try block here to invalidate self and crash the system
owner.getActionSet().actionCopyRows(t, this, srcPage,
dest_slot, num_rows, src_slot,
recordIds);
}
/**
Remove record at slot.
<p>
Remove the slot at the in-memory slot table, i.e.,
slots from 0 to deleteSlot-1 is untouched, deleteSlot is removed from
in memory slot table, deleteSlot+1 .. recordCount()-1 move to
down one slot.
<BR>
MT - latched
*/
protected void removeAndShiftDown(int slot)
{
if (SanityManager.DEBUG) {
SanityManager.ASSERT(isLatched());
SanityManager.ASSERT(slot >= 0 && slot < recordCount);
}
// just copy the records down in the array (copying over the slot
// entry that is being eliminated) and null out the last entry,
// it is ok for the array to be larger than necessary.
//
// source of copy: slot + 1
// dest of copy: slot
// length of copy: (length of array - source of copy)
System.arraycopy(
headers, slot + 1, headers, slot, headers.length - (slot + 1));
headers[headers.length - 1] = null;
recordCount--;
}
/**
Shift all records in the in-memory slot table up one slot,
starting at and including the record in slot 'low'
A new slot is added to accomdate the move.
<BR>
MT - latched
*/
protected StoredRecordHeader shiftUp(int low)
{
if (SanityManager.DEBUG)
{
SanityManager.ASSERT(isLatched());
if ((low < 0) || (low > recordCount))
{
SanityManager.THROWASSERT(
"shiftUp failed, low must be between 0 and recordCount." +
" low = " + low + ", recordCount = " + recordCount +
"\n page = " + this);
}
}
if (low < headers.length)
{
// just copy the records up in the array (copying over the slot
// entry that is being eliminated) and null out the entry at "low",
// it is ok for the array to be shorter than necessary.
//
// This code throws away the "last" entry in
// the array, which will cause a record header cache miss if it
// is needed. This delays the object allocation of a new array
// object until we really need that entry, vs. doing it on the
// insert.
//
// source of copy: low
// dest of copy: low + 1
// length of copy: (length of array - dest of copy)
// adding in the middle
System.arraycopy(
headers, low, headers, low + 1, headers.length - (low + 1));
headers[low] = null;
}
return(null);
}
/**
Try to compact this record. Deleted record are treated the same way as
nondeleted record. This page must not be an overflow page. The record
may already have been purged from the page.
<P>
<B>Locking Policy</B>
<P>
No locks are obtained.
<BR>
MT - latched
<P>
<B>NOTE : CAVEAT </B><BR>
This operation will physically get rid of any reserved space this
record may have, or it may compact the record by merging strung out row
pieces together. Since the freed reserved space is immediately usable
by other transactions which latched the page, it is only safe to use
this operation if the caller knows that it has exclusive access to the
page for the duration of the transaction, i.e., effectively holding a
page lock on the page, AND that the record has no uncommitted
updates.
@param handle Handle to deleted or non-deleted record
@see ContainerHandle#compactRecord
@exception StandardException Standard Derby error policy
*/
public void compactRecord(RecordHandle handle) throws StandardException
{
if (SanityManager.DEBUG) {
SanityManager.ASSERT(isLatched());
}
if (!owner.updateOK())
{
throw StandardException.newException(
SQLState.DATA_CONTAINER_READ_ONLY);
}
if (handle.getId() < RecordHandle.FIRST_RECORD_ID)
{
throw StandardException.newException(
SQLState.DATA_INVALID_RECORD_HANDLE, handle);
}
if (handle.getPageNumber() != getPageNumber())
{
throw StandardException.newException(
SQLState.DATA_WRONG_PAGE_FOR_HANDLE, handle);
}
if (isOverflowPage())
{
throw StandardException.newException(
SQLState.DATA_UNEXPECTED_OVERFLOW_PAGE, handle);
}
int slot = findRecordById(handle.getId(), handle.getSlotNumberHint());
if (slot >= 0)
{
compactRecord(owner.getTransaction(), slot, handle.getId());
}
// else record gone, no compaction necessary
}
/*
** Methods that read/store records/fields based upon calling methods
** a sub-calls provides to do the actual storage work.
*/
/*
** Page LastLog Instant control
*/
public final LogInstant getLastLogInstant()
{
return lastLog;
}
protected final void clearLastLogInstant() {
lastLog = null;
}
protected final void updateLastLogInstant(LogInstant instant)
{
if (SanityManager.DEBUG) {
SanityManager.ASSERT(isLatched());
}
// we should not null out the log instant even if this page is being
// updated by a non-logged action, there may have been logged action
// before this and we don't want to loose that pointer
if (instant != null)
lastLog = instant;
}
/*
** Page Version control
*/
/**
Return the current page version.
*/
public final long getPageVersion()
{
return pageVersion;
}
/**
increment the version by one and return the new version.
*/
protected final long bumpPageVersion()
{
if (SanityManager.DEBUG) {
SanityManager.ASSERT(isLatched());
}
return ++pageVersion;
}
/**
set it when the page is read from disk.
<BR> MT - single thread required - Only called while the page has no identity which
requires that only a single caller can be accessing it.
*/
public final void setPageVersion(long v)
{
pageVersion = v;
}
/**
Set page status based on passed in status flag.
*/
protected void setPageStatus(byte status)
{
pageStatus = status;
}
/**
Get the page status, one of the values in the above page status flag
*/
public byte getPageStatus()
{
return pageStatus;
}
/*
** abstract methods that an implementation must provide.
**
** <BR> MT - latched, page is latched when these methods are called.
*/
/**
* Read the record at the given slot into the given row.
* <P>
* This reads and initializes the columns in the row array from the raw
* bytes stored in the page associated with the given slot. If validColumns
* is non-null then it will only read those columns indicated by the bit
* set, otherwise it will try to read into every column in row[].
* <P>
* If there are more columns than entries in row[] then it just stops after
* every entry in row[] is full.
* <P>
* If there are more entries in row[] than exist on disk, the requested
* excess columns will be set to null by calling the column's object's
* restoreToNull() routine
* (ie. ((Object) column).restoreToNull() ).
* <P>
* If a qualifier list is provided then the row will only be read from
* disk if all of the qualifiers evaluate true. Some of the columns may
* have been read into row[] in the process of evaluating the qualifier.
*
* <BR> MT - latched, page is latched when this methods is called.
*
*
* @param slot the slot number
* @param row (out) filled in sparse row
* @param fetchDesc A set of information about the fetch: what
* columns to fetch, any qualifiers, ...
* @param rh the record handle for the row at top level,
* and is used in OverflowInputStream to lock the
* row for Blobs/Clobs.
* @param isHeadRow Is the head row portion of the row, false if
* a long row and the 2-N'th portion of the long
* row.
*
* @return false if a qualifier_list is provided and the row does not
* qualifier (no row read in that case), else true.
*
* @exception StandardException Standard Derby error policy
**/
protected abstract boolean restoreRecordFromSlot(
int slot,
Object[] row,
FetchDescriptor fetchDesc,
RecordHandle rh,
StoredRecordHeader recordHeader,
boolean isHeadRow)
throws StandardException;
/**
Read portion of a log record at the given slot into the given byteHolder.
<BR> MT - latched, page is latched when this methods is called.
@exception StandardException Standard Derby error policy
*/
protected abstract void restorePortionLongColumn(OverflowInputStream fetchStream)
throws StandardException, IOException;
/**
Create a new record identifier.
<BR> MT - latched, page is latched when this methods is called.
@exception StandardException Standard Derby error policy
*/
public abstract int newRecordId() throws StandardException;
/**
Create a new record identifier, and bump to next recordid.
<BR> MT - latched, page is latched when this methods is called.
@exception StandardException Standard Derby error policy
*/
public abstract int newRecordIdAndBump() throws StandardException;
/**
Create a new record identifier, the passed in one is the last one created.
Use this method to collect and reserve multiple recordIds in one
stroke. Given the same input recordId, the subclass MUST return the
same recordId every time.
<BR> MT - latched, page is latched when this methods is called.
@exception StandardException Standard Derby error policy
*/
protected abstract int newRecordId(int recordId) throws StandardException;
/**
Is there space for copying this many rows which takes this many bytes
on the page
<BR> MT - latched, page is latched when this methods is called.
@exception StandardException Standard Derby policy.
*/
public abstract boolean spaceForCopy(int num_rows, int[] spaceNeeded)
throws StandardException;
/**
Return the total number of bytes used, reserved, or wasted by the
record at this slot.
<BR> MT - latched, page is latched when this methods is called.
@exception StandardException Standard Derby policy.
*/
public abstract int getTotalSpace(int slot) throws StandardException;
/**
Return the total number of bytes reserved by the
record at this slot.
<BR> MT - latched, page is latched when this methods is called.
@exception IOException Thrown by InputStream methods potential I/O errors
*/
public abstract int getReservedCount(int slot) throws IOException;
/*
** Methods that our super-class (BasePage) requires we implement.
** Here we only implement the methods that correspond to the logical
** operations that require logging, any other methods that are storage
** specific we leave to our sub-class.
**
** All operations that are logged must bump this page's version number
** and update this page's last log instant.
** These should be sanity checked on each logAndDo (similarly, it should
** be checked in CompensationOperation.doMe)
*/
/*
** Methods that any sub-class must implement. These allow generic log operations.
*/
/**
Get the stored length of a record. This must match the amount of data
written by logColumn and logField.
<BR> MT - latched - page latch must be held
*/
public abstract int getRecordLength(int slot) throws IOException;
/**
Restore a storable row from a InputStream that was used to
store the row after a logRecord call.
<BR> MT - latched - page latch must be held
@exception StandardException Standard Derby error policy
@exception IOException object exceeds the available data in the stream.
*/
public abstract void restoreRecordFromStream(
LimitObjectInput in,
Object[] row)
throws StandardException, IOException;
/**
Log a currently stored record to the output stream.
The logged version of the record must be readable by storeRecord.
<BR> MT - latched - page latch must be held
@param slot Slot number the record is stored in.
@param flag LOG_RECORD_*, the reason for logging the record.
@param recordId Record identifier of the record.
@param validColumns which columns needs to be logged
@param out Where to write the logged form.
@param headRowHandle the recordHandle of the head row piece, used
for post commit cleanup for update.
@exception StandardException Standard Derby error policy
*/
public abstract void logRecord(int slot, int flag, int recordId,
FormatableBitSet validColumns, OutputStream out,
RecordHandle headRowHandle)
throws StandardException, IOException;
/**
Log the row that will be stored at the given slot to the given OutputStream.
The logged form of the Row must be readable by storeRecord.
<BR> MT - latched - page latch must be held
@param slot Slot number the record will be stored in.
@param forInsert True if the row is being logged for an insert,
false for an update.
@param recordId Record identifier of the record.
@param row The row version of the record.
@param validColumns FormatableBitSet of which columns in row are valid,
null indicates all are valid
@param out Where to write the logged form.
@param startColumn The first column that is being logged in this row.
This is used when logging portion of long rows.
@param insertFlag To indicate whether the insert would allow overflow.
@param realStartColumn This is used when a long column is detected.
Portion of the row may already be logged and stored
in the 'out' buffer. After we log the long column,
the saved buffer was passed here, and we need to
continue to log the row. realStartColumn is the starting
column for the continuation of the logRow operation.
Pass in (-1) if realStartColumn is not significant.
@param realSpaceOnPage Being used in conjunction with realStartColumn,
to indicate the real free space left on the page.
@exception StandardException Standard Derby error policy
*/
public abstract int logRow(
int slot,
boolean forInsert,
int recordId,
Object[] row,
FormatableBitSet validColumns,
DynamicByteArrayOutputStream out,
int startColumn,
byte insertFlag,
int realStartColumn,
int realSpaceOnPage,
int overflowThreshold)
throws StandardException, IOException;
/**
Log a currently stored field.
The logged version of the field must be readable by storeField.
<BR> MT - latched - page latch must be held
@param slot Slot number the record is stored in.
@param fieldNumber Number of the field (starts at 0).
@param out Where to write the logged form.
@exception StandardException Standard Derby error policy
*/
public abstract void logField(int slot, int fieldNumber, OutputStream out)
throws StandardException, IOException;
/**
Log a to be stored column.
<BR> MT - latched - page latch must be held
@param slot slot of the current record
@param fieldId field number of the column being updated
@param column column version of the field.
@param out Where to write the logged form.
@exception StandardException Standard Derby error policy
*/
public abstract void logColumn(
int slot,
int fieldId,
Object column,
DynamicByteArrayOutputStream out,
int overflowThreshold)
throws StandardException, IOException;
/**
Log a to be stored long column. return -1 when done.
<BR> MT - latched - page latch must be held
@param slot slot of the current record
@param recordId the id of the long column record
@param column column version of the field.
@param out Where to write the logged form.
@exception StandardException Standard Derby error policy
*/
public abstract int logLongColumn(
int slot,
int recordId,
Object column,
DynamicByteArrayOutputStream out)
throws StandardException, IOException;
/**
Read a previously stored record written by logRecord or logRow and store
it on the data page at the given slot with the given record identifier.
Any previously stored record must be replaced.
<BR> MT - latched - page latch must be held
@exception StandardException Standard Derby error policy
@exception IOException Thrown by InputStream methods potential I/O errors
while writing the page
*/
public abstract void storeRecord(LogInstant instant, int slot, boolean forInsert, ObjectInput in)
throws StandardException, IOException;
/**
Read a previously stored field written by logField or logColumn and store
it on the data page at thge given slot with the given record identifier
and field number. Any previously stored field is replaced.
<BR> MT - latched - page latch must be held
@exception StandardException Standard Derby error policy
@exception IOException Thrown by InputStream methods and potential I/O errors
while writing the page.
*/
public abstract void storeField(LogInstant instant, int slot,
int fieldId,
ObjectInput in)
throws StandardException, IOException;
/**
Reserve the required number of bytes for the record in the specified slot.
<BR> MT - latched - page latch must be held
@exception StandardException Standard Derby error policy
@exception IOException Thrown by InputStream methods and potential I/O errors
while writing the page.
*/
public abstract void reserveSpaceForSlot(LogInstant instant, int slot, int spaceToReserve)
throws StandardException, IOException;
/**
Skip a previously stored field written by logField or logColumn.
<BR> MT - latched - page latch must be held
@exception StandardException Standard Derby error policy
@exception IOException Thrown by InputStream methods
*/
public abstract void skipField(ObjectInput in)
throws StandardException, IOException;
public abstract void skipRecord(ObjectInput in) throws StandardException, IOException;
/**
Set the delete status of a record from the page.
<BR> MT - latched - page latch must be held
@param slot the slot to delete or undelete
@param delete set delete status to this value
@exception StandardException Standard Derby error policy
@exception IOException IO error accessing page
*/
public abstract void setDeleteStatus(LogInstant instant, int slot, boolean delete)
throws StandardException, IOException;
/**
Purge a record from the page.
<BR> MT - latched - page latch must be held
@param slot the slot to purge
@param recordId the id of the record that is to be purged
@exception StandardException Standard Derby error policy
@exception IOException Thrown by potential I/O errors
while writing the page.
*/
public abstract void purgeRecord(LogInstant instant, int slot,
int recordId)
throws StandardException, IOException;
/**
Subclass implementation of compactRecord.
@see BasePage#compactRecord
@exception StandardException Standard Derby error policy
*/
protected abstract void compactRecord(RawTransaction t, int slot, int recordId)
throws StandardException;
/**
Set the page status underneath a log record
<BR> MT - latched - page latch must be held
@param instant the log instant of the log record
@param status the page status
@exception StandardException Standard Derby error policy
*/
public abstract void setPageStatus(LogInstant instant, byte status)
throws StandardException;
/**
initialize a page for the first time or for reuse
All subtypes are expected to overwrite this method if it has something to clean up
@exception StandardException Standard Derby error policy
*/
public abstract void initPage(LogInstant instant, byte status,
int recordId, boolean overflow, boolean reuse)
throws StandardException;
/**
Set the reserved space for this row to value.
@exception StandardException Standard Derby error policy
*/
public abstract void setReservedSpace(LogInstant instant, int slot, int value)
throws StandardException, IOException;
/**
Return true if the page is an overflow page, false if not.
For implementation that don't have overflow pages, return false.
*/
public abstract boolean isOverflowPage();
/**
Returns false if an insert is not to be allowed in the page.
*/
public abstract boolean allowInsert();
/**
Returns true if an insert is allowed in the page and the page is
relatively unfilled - let specific implementation decide what
relatively unfilled means
*/
public abstract boolean unfilled();
/**
Set the number of rows in the container - the page uses this to decide
whether it needs to aggressive set the container's row count when it
changes.
*/
public abstract void setContainerRowCount(long count);
/*
* returns the page data array, that is actually written to the disk.
*/
protected abstract byte[] getPageArray() throws StandardException;
/*
** Debugging methods
*/
/** Debugging, print slot table information */
protected String slotTableToString()
{
String str = null;
if (SanityManager.DEBUG)
{
StoredRecordHeader rh;
str = "";
for (int slot = FIRST_SLOT_NUMBER; slot < recordCount; slot++) {
rh = getHeaderAtSlot(slot);
if (rh != null)
str += "Slot " + slot + " recordId " + rh.getId();
else
str += "Slot " + slot + " null record";
str += "\n";
}
}
return str;
}
}