Class BasePage

java.lang.Object
org.apache.derby.impl.store.raw.data.BasePage
All Implemented Interfaces:
TypedFormat, DerbyObserver, Page
Direct Known Subclasses:
CachedPage

abstract class BasePage extends Object implements Page, DerbyObserver, TypedFormat
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.

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.

BasePage implements Observer to watch the ContainerHandle which notifies its Observers when it is closing.
MT - mutable

  • Field Details

    • auxObj

      private AuxObject auxObj
      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.
    • identity

      protected PageKey identity
      this page's identity
      MT - immutable - content dynamic : single thread required
    • headers

      private StoredRecordHeader[] headers
      In-memory slot table, array of StoredRecordHeaders.
      MT - Immutable - Content Dynamic : Single thread required.
    • recordCount

      private int recordCount
    • owner

      protected BaseContainerHandle owner
      Page owner during exclusive access. MT - mutable : single thread required, provided by Lockable single thread required.
    • nestedLatch

      private int nestedLatch
      Count of times a latch is held nested during an abort
    • inClean

      protected boolean inClean
    • preLatch

      protected boolean preLatch
      Used to determine latch state of a page. MT - mutable There are 3 latch states for a page: UNLATCHED - (owner == null) PRELATCH - (owner != null) && preLatch LATCHED - (owner != null) && !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.
    • lastLog

      private LogInstant lastLog
      Instant of last log record that updated this page.
      MT - mutable : latched
    • repositionNeededAfterVersion

      private long repositionNeededAfterVersion
      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.
    • pageVersion

      private long pageVersion
      Version of the page.
      MT - mutable : single thread required - The page must be latched to access this variable or the page muts be in the noidentiy state.
    • pageStatus

      private byte pageStatus
      Status of the page
    • VALID_PAGE

      public static final byte VALID_PAGE
      Values for pageStatus flag page goes thru the following transition: VALID_PAGE <-> deallocated page -> free page <-> 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.
      See Also:
    • INVALID_PAGE

      public static final byte INVALID_PAGE
      See Also:
    • INIT_PAGE_REUSE

      public static final int INIT_PAGE_REUSE
      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
      See Also:
    • INIT_PAGE_OVERFLOW

      public static final int INIT_PAGE_OVERFLOW
      See Also:
    • INIT_PAGE_REUSE_RECORDID

      public static final int INIT_PAGE_REUSE_RECORDID
      See Also:
    • LOG_RECORD_DEFAULT

      public static final int LOG_RECORD_DEFAULT
      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.
      See Also:
    • LOG_RECORD_FOR_UPDATE

      public static final int LOG_RECORD_FOR_UPDATE
      See Also:
    • LOG_RECORD_FOR_PURGE

      public static final int LOG_RECORD_FOR_PURGE
      See Also:
    • InvalidRecordHandle

      private static final RecordHandle InvalidRecordHandle
  • Constructor Details

    • BasePage

      protected BasePage()
      Create a new, empty page.
  • Method Details

    • initialize

      protected void initialize()
      Initialized the BasePage.

      Initialize the object, ie. perform work normally perfomed in constructor. Called by setIdentity() and createIdentity().

    • initializeHeaders

      protected void initializeHeaders(int numRecords)
      Must be called by a sub-class before calling setHeaderAtSlot.
    • fillInIdentity

      protected void fillInIdentity(PageKey key)
    • clearIdentity

      public void clearIdentity()
    • cleanPageForReuse

      protected void cleanPageForReuse()
      Initialized this page for reuse or first use
    • getIdentity

      public Object getIdentity()
      OK to hand object outside to cache..
    • getInvalidRecordHandle

      public final RecordHandle getInvalidRecordHandle()
      Description copied from interface: Page
      Return an invalid record handle.

      Specified by:
      getInvalidRecordHandle in interface Page
      Returns:
      an invalid record handle.
    • MakeRecordHandle

      public static final RecordHandle MakeRecordHandle(PageKey pkey, int recordHandleConstant) throws StandardException
      Throws:
      StandardException
    • makeRecordHandle

      public final RecordHandle makeRecordHandle(int recordHandleConstant) throws StandardException
      Description copied from interface: Page
      Return a record handle for the given constant record id.

      Return a record handle that doesn't represent a record but rather has a special meaning. Used for special cases like creating a key specific to the page, but not specific to a row on the page.

      See RecordHandle interface for a list of "special record handles."

      Specified by:
      makeRecordHandle in interface Page
      Parameters:
      recordHandleConstant - the special recordId
      Returns:
      The created record handle.
      Throws:
      StandardException - if input is not a special record identifier.
      See Also:
    • getPageNumber

      public final long getPageNumber()
      Description copied from interface: Page
      Return the page number of this page.

      Page numbers are unique within a container and start at ContainerHandle.FIRST_PAGE_NUMBER and increment by 1 regardless of the page size.


      MT - Latched

      Specified by:
      getPageNumber in interface Page
      Returns:
      The page number of this page.
      See Also:
    • getPageKey

      public final PageKey getPageKey()
      Description copied from interface: Page
      Return the page key of this page.


      MT - Latched

      Specified by:
      getPageKey in interface Page
      Returns:
      The page key of this page.
      See Also:
    • getRecordHandle

      public final RecordHandle getRecordHandle(int recordId)
      Description copied from interface: Page
      Get a record handle from a previously stored record id.

      Get a record handle from a previously stored record identifier that was obtained from a RecordHandle.


      MT - Latched

      Specified by:
      getRecordHandle in interface Page
      Parameters:
      recordId - previously stored recordId.
      Returns:
      A valid record handle or null if the record no longer exists.
      See Also:
    • getRecordHandleAtSlot

      public final RecordHandle getRecordHandleAtSlot(int slot)
      Description copied from interface: Page
      Get the record handle of row at slot.

      Get the record handle of a record on a latched page using its slot number.
      MT - latched

      Specified by:
      getRecordHandleAtSlot in interface Page
      Parameters:
      slot - the slot number
      Returns:
      the record handle.
    • recordExists

      public final boolean recordExists(RecordHandle handle, boolean ignoreDelete) throws StandardException
      Description copied from interface: Page
      does the record still exist on the page?

      If "ignoreDelete" is true and the record handle represents a record on the page (either marked deleted or not) return true. If "ignoreDelete" is false return true if the record handle represents a record on the page and the record is not marked as deleted. Return false otherwise.
      MT - Latched

      Specified by:
      recordExists in interface Page
      Parameters:
      handle - handle of the record to look for.
      ignoreDelete - if true, then routine will return true even if the row is marked deleted.
      Returns:
      boolean indicating if the record still exists on the page.
      Throws:
      StandardException - recordHandle is not a valid record handle
      See Also:
    • fetchFromSlot

      public RecordHandle fetchFromSlot(RecordHandle rh, int slot, Object[] row, FetchDescriptor fetchDesc, boolean ignoreDelete) throws StandardException
      Description copied from interface: Page
      Fetch a record located in the passed in slot.

      Fetch a record located in the passed in slot and fill-in the passed in StorebleRow and the Object columns contained within. If row is null then the record is locked but is not fetched.
      This interface allows the caller to either return a deleted row or not. If "ignoreDelete" is set to true, fetch the record regardless of whether it is deleted or not (same as above fetchFromSlot). However, if "ignoreDelete" is set to false and the and the slot correspond to a deleted row, null is returned.
      If a non-null Qualifier list is provided then the qualifier array will be applied to the row and the row will only be returned if the row qualifies, otherwise null will be returned. Values in the columns of row may or may not be altered while trying to apply the qualifiers, if null is returned the state of the columns is undefined. If a null Qualifier list is provided then no qualification is applied.
      If a non-null record handle is passed in, it is assumed that the record handle corresponds to the record in the slot. If record handle is null, a record handle will be manufactured and returned if the record is not deleted or if "ignoreDelete" is true. This parameter is here for the case where the caller have already manufactured the record handle for locking or other purposes so it would make sense for the page to avoid creating a new record handle object if possible.

      Specified by:
      fetchFromSlot in interface Page
      Parameters:
      rh - the record handle of the row. If non-null it must refer to the same record as the slot.
      slot - the slot number
      row - Row to be filled in with information from record.
      fetchDesc - A structure to efficiently carry a set of parameters needed to describe the fetch, these include: validColumns - A bit map of which columns in the row to be fetched. ValidColumns will not be changed by RawStore. qualifier_list - A list of Qualifiers to apply to the row to see if the row should be returned. An array of qualifiers which restrict whether or not the row should be returned by the fetch. Rows for which any one of the qualifiers returns false are not returned by the fetch. If null, no qualification is done and the requested columns of the rows are returned. Qualifiers can only reference columns which are included in the scanColumnList. The column id that a qualifier returns is the column id the table, not the column id in the partial row being returned. qualifier_scratch_space - An array of int's that matches the size of the row[] array. Used to process qualifiers, if no qualifiers are input then array need not be input. Passed in rather than allocated so that space can be allocated a single time in a scan. If not passed in then raw store will allocate and deallocate per call.
      ignoreDelete - if true, return row regardless of whether it is deleted or not. If false, only return non-deleted row.
      Returns:
      A handle to the record.
      Throws:
      StandardException - Standard Derby error policy
      See Also:
    • fetchFieldFromSlot

      public final RecordHandle fetchFieldFromSlot(int slot, int fieldId, Object column) throws StandardException
      Description copied from interface: Page
      Fetch a single field from a deleted or non-deleted record. Fills in the passed in Object column with the field identified by fieldid if column is not null, otherwise the record is locked but not fetched.
      The fieldId of the first field is 0. If the fieldId is >= the number of fields on the record, column is restored to null

      Locking Policy
      No locks are obtained. It is up to the caller to obtain the correct locks.
      It is guaranteed that the page latch is not released by this method

      Specified by:
      fetchFieldFromSlot in interface Page
      Parameters:
      slot - is the slot number
      fieldId - is the column id
      column - is to be filled in with information from the record.
      Returns:
      the Handle to the record that is locked
      Throws:
      StandardException - Standard Derby error policy
      See Also:
    • getSlotNumber

      public final int getSlotNumber(RecordHandle handle) throws StandardException
      Description copied from interface: Page
      Get the slot number.

      Get the slot number of a record on a latched page using its record handle.

      Note The slot number is only good for as long as the page is latched.
      MT - latched

      Specified by:
      getSlotNumber in interface Page
      Parameters:
      handle - the record handle
      Returns:
      the slot number
      Throws:
      StandardException - Record does not exist on this page.
      See Also:
    • getNextSlotNumber

      public final int getNextSlotNumber(RecordHandle handle) throws StandardException
      Description copied from interface: Page
      Find slot for record with an id greater than the passed in identifier.

      Find the slot for the first record on the page with an id greater than the passed in identifier.
      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 client's 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 (assuming a PAGE_REUSABLE_RECORD_ID container), 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.
      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.
      This method returns the record regardless of its deleted status.
      MT - latched

      Specified by:
      getNextSlotNumber in interface Page
      Parameters:
      handle - record handle to find the next higher id.
      Returns:
      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.
      Throws:
      StandardException - Record does not exist on this page.
      See Also:
    • insertAtSlot

      public RecordHandle insertAtSlot(int slot, Object[] row, FormatableBitSet validColumns, LogicalUndo undo, byte insertFlag, int overflowThreshold) throws StandardException
      Description copied from interface: Page
      Insert a record at the specified slot.

      All records that occupy FIRST_SLOT_NUMBER to (slot - 1) are not moved.
      All records that occupy slot to (recordCount() - 1) are moved up one slot.
      The new record is inserted at the specified slot.
      If slot == FIRST_SLOT_NUMBER, then the new record will be inserted at the first slot.
      If slot == recordCount(), then the record is inserted in a new slot, no records are moved.
      If slot is > recordCount() or if slot < FIRST_SLOT_NUMBER, an exception will be thrown.

      Space Policy
      If the row will not fit on a page then:

      • an exception is thrown if the page has no other rows, this is an indication that the row could never fit on a page in this container.
      • null is returned if there are other rows on the page, this is an indication that the row can potentially be inserted successfully onto an empty page.

      Locking Policy
      Calls the lockRecordForWrite() method of the LockingPolicy object passed to the openContainer() call before the record is inserted.
      MT - latched

      Specified by:
      insertAtSlot in interface Page
      Parameters:
      slot - The specified slot
      row - The row version of the data
      validColumns - a bit map of which columns in the row is valid. ValidColumns will not be changed by RawStore.
      undo - if logical undo may be necessary, a function pointer to the access code where the logical undo logic resides. Null if logical undo is not necessary.
      insertFlag - if INSERT_UNDO_WITH_PURGE set, then the undo of this insert will purge the row rather than mark it as deleted, which is the default bahavior for insertAtSlot and insert.
      Returns:
      A RecordHandle representing the new record, or null if the row will not fit on a non-empty page.
      Throws:
      StandardException - Standard Derby error policy
      See Also:
    • insertNoOverflow

      protected RecordHandle insertNoOverflow(int slot, Object[] row, FormatableBitSet validColumns, LogicalUndo undo, byte insertFlag, int overflowThreshold) throws StandardException
      Throws:
      StandardException
    • insert

      public final RecordHandle insert(Object[] row, FormatableBitSet validColumns, byte insertFlag, int overflowThreshold) throws StandardException
      Description copied from interface: Page
      Insert a record anywhere on the page.

      Locking Policy
      Calls the lockRecordForWrite() method of the LockingPolicy object passed to the openContainer() call before the record is inserted.
      MT - latched

      Specified by:
      insert in interface Page
      Parameters:
      row - The row version of the data
      validColumns - a bit map of which columns in the row is valid. ValidColumns will not be changed by RawStore.
      insertFlag - see values for insertFlag below.
      Returns:
      A RecordHandle representing the new record.
      Throws:
      StandardException - Standard Derby error policy
      See Also:
    • insertAllowOverflow

      public RecordHandle insertAllowOverflow(int slot, Object[] row, FormatableBitSet validColumns, int startColumn, byte insertFlag, int overflowThreshold, RecordHandle nextPortionHandle) throws StandardException
      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.
      Throws:
      StandardException - Standard Derby error policy
    • insertLongColumn

      protected RecordHandle insertLongColumn(BasePage mainChainPage, LongColumnException lce, byte insertFlag) throws StandardException
      Routine to insert a long column.

      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.

      Parameters:
      mainChainPage - The parent page with row piece containing column that will eventually point to this long column chain.
      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.
      insertFlag - flags for insert operation.
      Returns:
      The recordHandle of the first piece of the long column chain.
      Throws:
      StandardException - Standard exception policy.
    • preDirty

      public abstract void preDirty()
      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.
    • updateOverflowDetails

      public abstract void updateOverflowDetails(RecordHandle handle, RecordHandle overflowHandle) throws StandardException
      Update the overflow pointer for a long row
      MT - latched - page latch must be held
      Parameters:
      handle - handle of the record for long row
      overflowHandle - the overflow (continuation) pointer for the long row
      Throws:
      StandardException - Standard Derby error policy
    • updateFieldOverflowDetails

      public abstract void updateFieldOverflowDetails(RecordHandle handle, RecordHandle overflowHandle) throws StandardException
      Update the overflow pointer for a long column
      MT - latched - page latch must be held
      Parameters:
      handle - handle of the record for long row
      overflowHandle - the overflow (continuation) pointer for the long row
      Throws:
      StandardException - Standard Derby error policy
    • appendOverflowFieldHeader

      public abstract int appendOverflowFieldHeader(DynamicByteArrayOutputStream logBuffer, RecordHandle overflowHandle) throws StandardException, IOException
      Append an overflow pointer to a partly logged row, to point to a long column that just been logged.
      MT - latched - page latch must be held
      Parameters:
      logBuffer - The buffer that contains the partially logged row.
      overflowHandle - the overflow (continuation) pointer to the beginning of the long column
      Throws:
      StandardException - Standard Derby error policy
      IOException
    • getOverflowPageForInsert

      public abstract BasePage getOverflowPageForInsert(int slot, Object[] row, FormatableBitSet validColumns, int startColumn) throws StandardException
      Throws:
      StandardException
    • getNewOverflowPage

      protected abstract BasePage getNewOverflowPage() throws StandardException
      Throws:
      StandardException
    • updateAtSlot

      public final RecordHandle updateAtSlot(int slot, Object[] row, FormatableBitSet validColumns) throws StandardException
      Description copied from interface: Page
      Update the complete record identified by the slot.

      Locking Policy

      Calls the lockRecordForWrite() method of the LockingPolicy object passed to the openContainer() call before the record is undeleted. If record already deleted, an exception is thrown.
      It is guaranteed that the page latch is not released by this method

      Specified by:
      updateAtSlot in interface Page
      Parameters:
      slot - is the slot number
      validColumns - a bit map of which columns in the row is valid. ValidColumns will not be changed by RawStore.
      Returns:
      a Handle to the updated record.
      Throws:
      StandardException - Standard Derby error policy
      StandardException - StandardException.newException(SQLState.UPDATE_DELETED_RECORD if the record is already deleted
      StandardException - StandardException.newException(SQLState.CONTAINER_READ_ONLY if the container is read only
      See Also:
    • doUpdateAtSlot

      public abstract void doUpdateAtSlot(RawTransaction t, int slot, int id, Object[] row, FormatableBitSet validColumns) throws StandardException
      Throws:
      StandardException
    • updateFieldAtSlot

      public RecordHandle updateFieldAtSlot(int slot, int fieldId, Object newValue, LogicalUndo undo) throws StandardException
      Description copied from interface: Page
      Update a field within the record, replacing its current value with the stored representation of newValue. Record is identified by slot. If the field does not exist then it is added to the record, but only if (fieldId - 1) exists.
      RESOLVE right now it throws an exception if fieldId is not already on the record, not add the next one as advertised.

      Locking Policy

      Calls the lockRecordForWrite() method of the LockingPolicy object passed to the openContainer() call before the record is updated.
      It is guaranteed that the page latch is not released by this method

      Specified by:
      updateFieldAtSlot in interface Page
      Parameters:
      slot - is the slot number
      fieldId - is the column id
      newValue - has the new colum value to be stored in the record
      undo - if logical undo may be necessary, a function pointer to the access code where the logical undo logic resides. Null if logical undo is not necessary.
      Returns:
      a Handle to the updated record.
      Throws:
      StandardException - Standard Derby error policy
      StandardException - StandardException.newException(SQLState.UPDATE_DELETED_RECORD if the record is already deleted
      StandardException - StandardException.newException(SQLState.CONTAINER_READ_ONLY if the container is read only
      See Also:
    • fetchNumFields

      public final int fetchNumFields(RecordHandle handle) throws StandardException
      Description copied from interface: Page
      Fetch the number of fields in a record.

      Locking Policy

      No locks are obtained.
      MT - latched

      Specified by:
      fetchNumFields in interface Page
      Parameters:
      handle - record handle to deleted or non-deleted record
      Returns:
      the number of fields in the record
      Throws:
      StandardException - Standard exception policy.
      See Also:
    • fetchNumFieldsAtSlot

      public int fetchNumFieldsAtSlot(int slot) throws StandardException
      Description copied from interface: Page
      Fetch the number of fields in a record.

      Locking Policy

      No locks are obtained.
      It is guaranteed that the page latch is not released by this method

      Specified by:
      fetchNumFieldsAtSlot in interface Page
      Parameters:
      slot - is the slot number
      Returns:
      the number of fields in the record
      Throws:
      StandardException - Standard exception policy.
      See Also:
    • deleteAtSlot

      public RecordHandle deleteAtSlot(int slot, boolean delete, LogicalUndo undo) throws StandardException
      Description copied from interface: Page
      Mark the record at slot as deleted or undeleted according to delete flag.

      Locking Policy

      Calls the lockRecordForWrite() method of the LockingPolicy object passed to the openContainer() call before the record is deleted. If record already deleted, and an attempt is made to delete it, an exception is thrown. If record not deleted, and an attempt is made to undelete it, an exception is thrown.
      MT - latched

      Specified by:
      deleteAtSlot in interface Page
      Parameters:
      slot - the slot number
      delete - true if this record is to be deleted, false if this deleted record is to be marked undeleted
      undo - logical undo logic if necessary
      Returns:
      a Handle to the deleted/undeleted record.
      Throws:
      StandardException - Standard exception policy.
      StandardException - StandardException.newException(SQLState.UPDATE_DELETED_RECORD if an attempt to delete a record that is already deleted
      StandardException - StandardException.newException(SQLState.UNDELETE_RECORD if an attempt to undelete a record that is not deleted
      See Also:
    • purgeAtSlot

      public void purgeAtSlot(int slot, int numpurges, boolean needDataLogged) throws StandardException
      Purge one or more rows on a non-overflow page.
      Specified by:
      purgeAtSlot in interface Page
      Parameters:
      slot - the starting slot number
      numpurges - number of slots to purge. If <= 0, just returns as a no-op.
      needDataLogged - if set to true data is logged for purges else only headers.
      Throws:
      StandardException - Standard exception policy.
      See Also:
    • purgeRowPieces

      protected abstract void purgeRowPieces(RawTransaction t, int slot, RecordHandle headRowHandle, boolean needDataLogged) throws StandardException
      Purge all the overflow columns and overflow rows of the record at slot.
      Throws:
      StandardException - Standard exception policy.
    • copyAndPurge

      public void copyAndPurge(Page destPage, int src_slot, int num_rows, int dest_slot) throws StandardException
      Description copied from interface: Page
      move rows from one page to another, purging in the process.

      Move from this page slot[src_slot] to slot[src_slot+num_rows-1] to destPage slot[dest_slot] to slot[dest_slot + num_rows - 1], in that order. Both this page and destPage must be latched and from the same container with the same page and record format.
      Slot[src_slot] to slot[src_slot+numrows-1] will be purged from this page. RecordId on the dest page will be brand new and not in any particular order or range. RecordId of the purged rows in this page is never reused. Deleted and undeleted rows are copied over just the same. Exception will be thrown if this page does not have all the rows in the moved over range.
      RESOLVE: reserve space now not copied over because in btree, a row never shrinks. When this routine is called by heap or by some page which will have shrunken row, then we need to add that
      DestPage must have at least dest_slot row occupying slot[0] to slot[dest_slot-1]. DestPage must have enough space to take the copied over data. Rows that occupied slot number > dest_slot will be moved up the slot (I.e., slot[dest_slot] -> slot[dest_slot + num_rows]).
      If this operation rolls back, this page (the src page) will get the rows back and the dest page will purge the rows that were copied - this is as if the rows were inserted into the dest page with INSERT_UNDO_WITH_PURGE.

      Locking Policy

      Calls the lockRecordForWrite() method of the LockingPolicy object passed to the openContainer() call before the rows are copied over and bore the records are purged. I.e, for num_rows moved, there will be 2*num_rows calls to lockRecordForWrite.

      Use with caution
      As with a normal purge, no space is reserved on this page for rollback of the purge, so you must commit before inserting any rows onto this page - unless those inserts are INSERT_UNDO_WITH_PURGE.

      Specified by:
      copyAndPurge in interface Page
      Parameters:
      destPage - the page to copy to
      src_slot - start copying from this slot
      num_rows - copy and purge this many rows from this page
      dest_slot - copying into this slot of destPage
      Throws:
      StandardException - Standard exception policy.
      See Also:
    • unlatch

      public void unlatch()
      Unlatch the page.
      Specified by:
      unlatch in interface Page
      See Also:
    • isLatched

      public final boolean isLatched()
      Check whether the page is latched.
      Specified by:
      isLatched in interface Page
      Returns:
      true if the page is latched, false otherwise
      See Also:
    • recordCount

      public final int recordCount()
      Description copied from interface: Page
      Return the number of records on the page. The returned count includes rows that are deleted, i.e. it is the same as the number of slots on the page.
      MT - latched
      Specified by:
      recordCount in interface Page
      See Also:
    • internalDeletedRecordCount

      protected abstract int internalDeletedRecordCount()
      get record count without checking for latch
    • internalNonDeletedRecordCount

      protected int internalNonDeletedRecordCount()
      get record count without checking for latch
    • nonDeletedRecordCount

      public int nonDeletedRecordCount()
      Description copied from interface: Page
      Return the number of records on this page that are not marked as deleted.
      MT - latched
      Specified by:
      nonDeletedRecordCount in interface Page
      See Also:
    • shouldReclaimSpace

      public boolean shouldReclaimSpace(int num_non_deleted_rows, int slot_just_deleted) throws StandardException
      Is this page/deleted row a candidate for immediate reclaim space.

      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.

      Also used by access methods after undo of an insert.

      Will return true if the number of non-deleted rows on the page is <= "num_non_deleted_rows". For instance 0 means schedule reclaim only if all rows are deleted, 1 if all rows but one are deleted.

      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.

      Specified by:
      shouldReclaimSpace in interface Page
      Parameters:
      num_non_deleted_rows - threshold number of non-deleted rows to schedule reclaim space.
      slot_just_deleted - row on page to check for long row/long column
      Returns:
      true if a reclaim space should be scheduled post commit on this page, false otherwise.
      Throws:
      StandardException - Standard exception policy.
    • isDeletedOnPage

      protected final boolean isDeletedOnPage(int slot)
    • isDeletedAtSlot

      public boolean isDeletedAtSlot(int slot) throws StandardException
      Description copied from interface: Page
      Test if a record is deleted.

      Locking Policy
      No locks are obtained.
      It is guaranteed that the page latch is not released by this method

      Specified by:
      isDeletedAtSlot in interface Page
      Parameters:
      slot - slot of record to be tested.
      Throws:
      StandardException - Standard exception policy.
      See Also:
    • setAuxObject

      public void setAuxObject(AuxObject obj)
      Set the aux object.
      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.
      Specified by:
      setAuxObject in interface Page
      See Also:
    • getAuxObject

      public AuxObject getAuxObject()
      Get the aux object.
      MT - latched - It is required the caller throws away the returned reference when the page is unlatched.
      Specified by:
      getAuxObject in interface Page
      See Also:
    • setRepositionNeeded

      public void setRepositionNeeded()
      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).
      Specified by:
      setRepositionNeeded in interface Page
    • isRepositionNeeded

      public boolean isRepositionNeeded(long version)
      Check if a B-tree scan positioned on this page needs to reposition.
      Specified by:
      isRepositionNeeded in interface Page
      Parameters:
      version - the last version on which the B-tree scan had a valid position on this page
      Returns:
      true if a repositioning is needed because the row on the current position may have been removed from this page after the specified version; false otherwise
    • update

      public void update(DerbyObservable obj, Object arg)
      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.
      MT - latched
      Specified by:
      update in interface DerbyObserver
      Parameters:
      obj - The object which is being observed
      arg - Extra information being passed to the callback
      See Also:
    • getPageId

      public final PageKey getPageId()
      Get the Page identifer
      MT - RESOLVE
    • setExclusive

      void setExclusive(BaseContainerHandle requester) throws StandardException
      Get an exclusive latch on the page.
      MT - thread safe
      Throws:
      StandardException - Standard Derby policy.
    • setExclusiveNoWait

      boolean setExclusiveNoWait(BaseContainerHandle requester) throws StandardException
      Get an exclusive latch on the page, but only if I don't have to wait.
      MT - thread safe
      Throws:
      StandardException
    • releaseExclusive

      protected void releaseExclusive()
      Release the exclusive latch on the page.
      MT - latched
    • preLatch

      private void preLatch(BaseContainerHandle requester)
      Move page state from UNLATCHED to PRELATCH. setExclusive*() routines do the work of completing the latch - using the preLatch status.
      Parameters:
      requester - handle to the container requesting ownership
    • setHeaderAtSlot

      protected final void setHeaderAtSlot(int slot, StoredRecordHeader rh)
      Must be called by any non-abstract sub-class to initialise the slot table.
    • bumpRecordCount

      protected final void bumpRecordCount(int number)
    • getHeaderAtSlot

      public final StoredRecordHeader getHeaderAtSlot(int slot)
    • entireRecordOnPage

      public abstract boolean entireRecordOnPage(int slot) throws StandardException
      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.
      MT - latched
      Throws:
      StandardException - Standard Derby error policy
    • recordHeaderOnDemand

      public abstract StoredRecordHeader recordHeaderOnDemand(int slot)
    • checkSlotOnPage

      private final void checkSlotOnPage(int slot) throws StandardException
      Is the given slot number on the page?
      MT - latched
      Throws:
      StandardException
    • setDeleteStatus

      public int setDeleteStatus(int slot, boolean delete) throws StandardException, IOException
      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.
      Any sub-class must call this method when deleting a record.
      MT - latched
      Throws:
      StandardException - Standard Derby error policy
      IOException - IO error accessing page
    • deallocatePage

      public void deallocatePage() throws StandardException
      Mark this page as being deallocated
      Throws:
      StandardException - Derby Standard error policy
    • initPage

      public void initPage(int initFlag, long pageOffset) throws StandardException
      Mark this page as being allocated and initialize it to a pristine page
      Throws:
      StandardException - Derby Standard error policy
    • findRecordById

      public int findRecordById(int recordId, int slotHint)
      Find the slot for the record with the passed in identifier.
      This method returns the record regardless of its deleted status.
      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.
      MT - latched
      Parameters:
      recordId - record id of the record to search for.
      slotHint - "hint" about which slot the record might be in.
    • findNextRecordById

      private int findNextRecordById(int recordId)
      Find the slot for the first record on the page with an id greater than the passed in identifier.
      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.
      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.
      This method returns the record regardless of its deleted status.
      MT - latched
      Parameters:
      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.
    • copyInto

      private void copyInto(BasePage srcPage, int src_slot, int num_rows, int dest_slot) throws StandardException
      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.
      Throws:
      StandardException
      See Also:
    • removeAndShiftDown

      protected void removeAndShiftDown(int slot)
      Remove record at slot.

      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.
      MT - latched

    • shiftUp

      protected StoredRecordHeader shiftUp(int low)
      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.
      MT - latched
    • compactRecord

      public void compactRecord(RecordHandle handle) throws StandardException
      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.

      Locking Policy

      No locks are obtained.
      MT - latched

      NOTE : CAVEAT
      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.

      Parameters:
      handle - Handle to deleted or non-deleted record
      Throws:
      StandardException - Standard Derby error policy
      See Also:
    • getLastLogInstant

      public final LogInstant getLastLogInstant()
    • clearLastLogInstant

      protected final void clearLastLogInstant()
    • updateLastLogInstant

      protected final void updateLastLogInstant(LogInstant instant)
    • getPageVersion

      public final long getPageVersion()
      Return the current page version.
      Specified by:
      getPageVersion in interface Page
      Returns:
      page version number
    • bumpPageVersion

      protected final long bumpPageVersion()
      increment the version by one and return the new version.
    • setPageVersion

      public final void setPageVersion(long v)
      set it when the page is read from disk.
      MT - single thread required - Only called while the page has no identity which requires that only a single caller can be accessing it.
    • setPageStatus

      protected void setPageStatus(byte status)
      Set page status based on passed in status flag.
    • getPageStatus

      public byte getPageStatus()
      Get the page status, one of the values in the above page status flag
    • restoreRecordFromSlot

      protected abstract boolean restoreRecordFromSlot(int slot, Object[] row, FetchDescriptor fetchDesc, RecordHandle rh, StoredRecordHeader recordHeader, boolean isHeadRow) throws StandardException
      Read the record at the given slot into the given row.

      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[].

      If there are more columns than entries in row[] then it just stops after every entry in row[] is full.

      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() ).

      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.
      MT - latched, page is latched when this methods is called.

      Parameters:
      slot - the slot number
      row - (out) filled in sparse row
      fetchDesc - A set of information about the fetch: what columns to fetch, any qualifiers, ...
      rh - the record handle for the row at top level, and is used in OverflowInputStream to lock the row for Blobs/Clobs.
      isHeadRow - Is the head row portion of the row, false if a long row and the 2-N'th portion of the long row.
      Returns:
      false if a qualifier_list is provided and the row does not qualifier (no row read in that case), else true.
      Throws:
      StandardException - Standard Derby error policy
    • restorePortionLongColumn

      protected abstract void restorePortionLongColumn(OverflowInputStream fetchStream) throws StandardException, IOException
      Read portion of a log record at the given slot into the given byteHolder.
      MT - latched, page is latched when this methods is called.
      Throws:
      StandardException - Standard Derby error policy
      IOException
    • newRecordId

      public abstract int newRecordId() throws StandardException
      Create a new record identifier.
      MT - latched, page is latched when this methods is called.
      Throws:
      StandardException - Standard Derby error policy
    • newRecordIdAndBump

      public abstract int newRecordIdAndBump() throws StandardException
      Create a new record identifier, and bump to next recordid.
      MT - latched, page is latched when this methods is called.
      Throws:
      StandardException - Standard Derby error policy
    • newRecordId

      protected abstract int newRecordId(int recordId) 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.
      MT - latched, page is latched when this methods is called.
      Throws:
      StandardException - Standard Derby error policy
    • spaceForCopy

      public abstract boolean spaceForCopy(int num_rows, int[] spaceNeeded) throws StandardException
      Is there space for copying this many rows which takes this many bytes on the page
      MT - latched, page is latched when this methods is called.
      Throws:
      StandardException - Standard Derby policy.
    • getTotalSpace

      public abstract int getTotalSpace(int slot) throws StandardException
      Return the total number of bytes used, reserved, or wasted by the record at this slot.
      MT - latched, page is latched when this methods is called.
      Throws:
      StandardException - Standard Derby policy.
    • getReservedCount

      public abstract int getReservedCount(int slot) throws IOException
      Return the total number of bytes reserved by the record at this slot.
      MT - latched, page is latched when this methods is called.
      Throws:
      IOException - Thrown by InputStream methods potential I/O errors
    • getRecordLength

      public abstract int getRecordLength(int slot) throws IOException
      Get the stored length of a record. This must match the amount of data written by logColumn and logField.
      MT - latched - page latch must be held
      Throws:
      IOException
    • restoreRecordFromStream

      public abstract void restoreRecordFromStream(LimitObjectInput in, Object[] row) throws StandardException, IOException
      Restore a storable row from a InputStream that was used to store the row after a logRecord call.
      MT - latched - page latch must be held
      Throws:
      StandardException - Standard Derby error policy
      IOException - object exceeds the available data in the stream.
    • logRecord

      public abstract void logRecord(int slot, int flag, int recordId, FormatableBitSet validColumns, OutputStream out, RecordHandle headRowHandle) throws StandardException, IOException
      Log a currently stored record to the output stream. The logged version of the record must be readable by storeRecord.
      MT - latched - page latch must be held
      Parameters:
      slot - Slot number the record is stored in.
      flag - LOG_RECORD_*, the reason for logging the record.
      recordId - Record identifier of the record.
      validColumns - which columns needs to be logged
      out - Where to write the logged form.
      headRowHandle - the recordHandle of the head row piece, used for post commit cleanup for update.
      Throws:
      StandardException - Standard Derby error policy
      IOException
    • logRow

      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 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.
      MT - latched - page latch must be held
      Parameters:
      slot - Slot number the record will be stored in.
      forInsert - True if the row is being logged for an insert, false for an update.
      recordId - Record identifier of the record.
      row - The row version of the record.
      validColumns - FormatableBitSet of which columns in row are valid, null indicates all are valid
      out - Where to write the logged form.
      startColumn - The first column that is being logged in this row. This is used when logging portion of long rows.
      insertFlag - To indicate whether the insert would allow overflow.
      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.
      realSpaceOnPage - Being used in conjunction with realStartColumn, to indicate the real free space left on the page.
      Throws:
      StandardException - Standard Derby error policy
      IOException
    • logField

      public abstract void logField(int slot, int fieldNumber, OutputStream out) throws StandardException, IOException
      Log a currently stored field. The logged version of the field must be readable by storeField.
      MT - latched - page latch must be held
      Parameters:
      slot - Slot number the record is stored in.
      fieldNumber - Number of the field (starts at 0).
      out - Where to write the logged form.
      Throws:
      StandardException - Standard Derby error policy
      IOException
    • logColumn

      public abstract void logColumn(int slot, int fieldId, Object column, DynamicByteArrayOutputStream out, int overflowThreshold) throws StandardException, IOException
      Log a to be stored column.
      MT - latched - page latch must be held
      Parameters:
      slot - slot of the current record
      fieldId - field number of the column being updated
      column - column version of the field.
      out - Where to write the logged form.
      Throws:
      StandardException - Standard Derby error policy
      IOException
    • logLongColumn

      public abstract int logLongColumn(int slot, int recordId, Object column, DynamicByteArrayOutputStream out) throws StandardException, IOException
      Log a to be stored long column. return -1 when done.
      MT - latched - page latch must be held
      Parameters:
      slot - slot of the current record
      recordId - the id of the long column record
      column - column version of the field.
      out - Where to write the logged form.
      Throws:
      StandardException - Standard Derby error policy
      IOException
    • storeRecord

      public abstract void storeRecord(LogInstant instant, int slot, boolean forInsert, ObjectInput in) 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.
      MT - latched - page latch must be held
      Throws:
      StandardException - Standard Derby error policy
      IOException - Thrown by InputStream methods potential I/O errors while writing the page
    • storeField

      public abstract void storeField(LogInstant instant, int slot, int fieldId, 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.
      MT - latched - page latch must be held
      Throws:
      StandardException - Standard Derby error policy
      IOException - Thrown by InputStream methods and potential I/O errors while writing the page.
    • reserveSpaceForSlot

      public abstract void reserveSpaceForSlot(LogInstant instant, int slot, int spaceToReserve) throws StandardException, IOException
      Reserve the required number of bytes for the record in the specified slot.
      MT - latched - page latch must be held
      Throws:
      StandardException - Standard Derby error policy
      IOException - Thrown by InputStream methods and potential I/O errors while writing the page.
    • skipField

      public abstract void skipField(ObjectInput in) throws StandardException, IOException
      Skip a previously stored field written by logField or logColumn.
      MT - latched - page latch must be held
      Throws:
      StandardException - Standard Derby error policy
      IOException - Thrown by InputStream methods
    • skipRecord

      public abstract void skipRecord(ObjectInput in) throws StandardException, IOException
      Throws:
      StandardException
      IOException
    • setDeleteStatus

      public abstract void setDeleteStatus(LogInstant instant, int slot, boolean delete) throws StandardException, IOException
      Set the delete status of a record from the page.
      MT - latched - page latch must be held
      Parameters:
      slot - the slot to delete or undelete
      delete - set delete status to this value
      Throws:
      StandardException - Standard Derby error policy
      IOException - IO error accessing page
    • purgeRecord

      public abstract void purgeRecord(LogInstant instant, int slot, int recordId) throws StandardException, IOException
      Purge a record from the page.
      MT - latched - page latch must be held
      Parameters:
      slot - the slot to purge
      recordId - the id of the record that is to be purged
      Throws:
      StandardException - Standard Derby error policy
      IOException - Thrown by potential I/O errors while writing the page.
    • compactRecord

      protected abstract void compactRecord(RawTransaction t, int slot, int recordId) throws StandardException
      Subclass implementation of compactRecord.
      Throws:
      StandardException - Standard Derby error policy
      See Also:
    • setPageStatus

      public abstract void setPageStatus(LogInstant instant, byte status) throws StandardException
      Set the page status underneath a log record
      MT - latched - page latch must be held
      Parameters:
      instant - the log instant of the log record
      status - the page status
      Throws:
      StandardException - Standard Derby error policy
    • initPage

      public abstract void initPage(LogInstant instant, byte status, int recordId, boolean overflow, boolean reuse) 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
      Throws:
      StandardException - Standard Derby error policy
    • setReservedSpace

      public abstract void setReservedSpace(LogInstant instant, int slot, int value) throws StandardException, IOException
      Set the reserved space for this row to value.
      Throws:
      StandardException - Standard Derby error policy
      IOException
    • isOverflowPage

      public abstract boolean isOverflowPage()
      Return true if the page is an overflow page, false if not. For implementation that don't have overflow pages, return false.
    • allowInsert

      public abstract boolean allowInsert()
      Returns false if an insert is not to be allowed in the page.
    • unfilled

      public abstract boolean unfilled()
      Returns true if an insert is allowed in the page and the page is relatively unfilled - let specific implementation decide what relatively unfilled means
    • setContainerRowCount

      public abstract void setContainerRowCount(long count)
      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.
    • getPageArray

      protected abstract byte[] getPageArray() throws StandardException
      Throws:
      StandardException
    • slotTableToString

      protected String slotTableToString()
      Debugging, print slot table information