Class OpenConglomerate
- Direct Known Subclasses:
OpenHeap
The access method is page based and contained in a single container maintained by raw store.
-
Field Summary
FieldsModifier and TypeFieldDescriptionprivate ContainerHandle
private boolean
private boolean
private Conglomerate
The following group of fields are all basic input parameters which are provided by the calling code when doing any sort of operation requiring an open conglomerate (openScan(), open(), openCostController(), ...).private DynamicCompiledOpenConglomInfo
private boolean
private int
private LockingPolicy
private int
private Transaction
private TransactionManager
private OpenConglomerateScratchSpace
scratch space used for stuff like templates, export rows, ...private boolean
convenience boolean's for various mode's -
Constructor Summary
Constructors -
Method Summary
Modifier and TypeMethodDescriptionvoid
Check consistency of a conglomerate.void
close()
Close the container.void
final Conglomerate
final ContainerHandle
abstract int[]
final boolean
getHold()
Request set of properties associated with a table.final int
final Transaction
Get information about space used by the conglomerate.void
getTableProperties
(Properties prop) Request the system properties associated with a table.final TransactionManager
Public Accessors of This class:init
(ContainerHandle open_container, Conglomerate conglomerate, int[] format_ids, int[] collation_ids, TransactionManager xact_manager, Transaction rawtran, boolean hold, int openmode, int lock_level, LockingPolicy locking_policy, DynamicCompiledOpenConglomInfo dynamic_info) Open the container.final boolean
isClosed()
final boolean
protected boolean
isKeyed()
protected boolean
is the open btree table locked?final boolean
boolean
latchPage
(RowPosition pos) Latch the page containing the current RowPosition.boolean
Latch the page containing the current RowPosition, and reposition scan.boolean
lockPositionForRead
(RowPosition pos, RowPosition aux_pos, boolean moveForwardIfRowDisappears, boolean waitForLock) Lock row at given row position for read.boolean
lockPositionForWrite
(RowPosition pos, boolean waitForLock) Lock the row at the given position for write.protected abstract RowLocation
Return an "empty" row location object of the correct type.reopen()
Open the container.void
Unlock the record after a previous request to lock it.
-
Field Details
-
init_conglomerate
The following group of fields are all basic input parameters which are provided by the calling code when doing any sort of operation requiring an open conglomerate (openScan(), open(), openCostController(), ...). These are just saved values from what was initially input. -
init_xact_manager
-
init_rawtran
-
init_openmode
private int init_openmode -
init_lock_level
private int init_lock_level -
init_dynamic_info
-
init_hold
private boolean init_hold -
init_locking_policy
-
useUpdateLocks
private boolean useUpdateLocksconvenience boolean's for various mode's -
forUpdate
private boolean forUpdate -
getBaseTableLocks
private boolean getBaseTableLocks -
runtime_mem
scratch space used for stuff like templates, export rows, ... -
container
-
-
Constructor Details
-
OpenConglomerate
public OpenConglomerate()
-
-
Method Details
-
newRowLocationTemplate
Return an "empty" row location object of the correct type.- Returns:
- The empty Rowlocation.
- Throws:
StandardException
- Standard exception policy.
-
getFormatIds
public abstract int[] getFormatIds() -
latchPageAndRepositionScan
Latch the page containing the current RowPosition, and reposition scan.Upon return the scan will hold a latch on the page to continue the scan on. The scan will positioned on the record, just before the next record to return. Note that for both hold cursor and read uncommitted support this routine handles all cases of either the current position "dissappearing" (either the row and/or page). The row and/or page can disappear by deleted space being reclaimed post commit of that delete, and for some reason the code requesting the reposition does not have locks which prevented the space reclamation. Both hold cursor and read uncommitted scans are examples of ways the caller will not prevent space reclamation from claiming the position. This implementation also automatically updates the RowPosition to point at the slot containing the current RowPosition. This slot value is only valid while the latch is held.
- Returns:
- true if scan had to reposition because a row disappeared.
- Throws:
StandardException
- Standard exception policy.
-
latchPage
Latch the page containing the current RowPosition.
This implementation also automatically updates the RowPosition to point at the slot containing the current RowPosition. This slot value is only valid while the latch is held.
If the row pointed to by
pos
does not exist (including the case where the page itself does not exist), the page will not be latched, andpos.current_page
will be set tonull
.- Parameters:
pos
- the position to a row on the page that should be latched- Returns:
true
if the page was successfully latched, orfalse
otherwise- Throws:
StandardException
- Standard exception policy.
-
lockPositionForRead
public boolean lockPositionForRead(RowPosition pos, RowPosition aux_pos, boolean moveForwardIfRowDisappears, boolean waitForLock) throws StandardException Lock row at given row position for read.
This routine requests a row lock NOWAIT on the row located at the given RowPosition. If the lock is granted NOWAIT the routine will return true. If the lock cannot be granted NOWAIT, then the routine will release the latch on "page" and then it will request a WAIT lock on the row.
This implementation: Assumes latch held on current_page. If the current_rh field of RowPosition is non-null it is assumed that we want to lock that record handle and that we don't have a slot number. If the current_rh field of RowPosition is null, it is assumed the we want to lock the indicated current_slot. Upon return current_rh will point to the record handle associated with current_slot.
After waiting and getting the lock on the row, this routine will fix up RowPosition to point at the row locked. This means it will get the page latch again, and it will fix the current_slot to point at the waited for record handle - it may have moved while waiting on the lock.
When this method returns, the page holding the row pointed to by the
RowLocation
is latched. This is however not the case ifmoveForwardIfRowDisappears
isfalse
and the row has disappeared. Then the latch will be released before the method returns, andpos.current_page
will be set tonull
.- Parameters:
pos
- Position to lock.aux_pos
- If you have to give up latch to get lock, then also unlock this position if it is non-null.moveForwardIfRowDisappears
- If true, then this routine must handle the case where the row id we are waiting on disappears when the latch is released. If false, and the row disappears, the latch will be released again and false is returned.waitForLock
- if true wait for lock, if lock can't be granted NOWAIT, else if false, throw a lock timeout exception if the lock can't be granted without waiting.- Returns:
- true if lock granted without releasing the latch, else return false.
- Throws:
StandardException
- Standard exception policy.
-
lockPositionForWrite
Lock the row at the given position for write.
The page pointed to by the
RowPosition
is assumed to be latched when this method is called. If the lock cannot be obtained without waiting, the latch will be released and re-obtained when the lock has been acquired.If the latch was released while waiting for the lock, and the row does not exist after the lock is obtained, the latch will be released again before the method returns, and
pos.current_page
will be set tonull
.- Throws:
StandardException
-
unlockPositionAfterRead
Unlock the record after a previous request to lock it.Unlock the record after a previous call to lockRecordForRead(). It is expected that RowPosition contains information used to lock the record, Thus it is important if using a single RowPosition to track a scan to call unlock before you move the position forward to the next record.
Note that this routine assumes that the row was locked forUpdate if the OpenConglomerate is forUpdate, else it assumes the record was locked for read.
- Throws:
StandardException
- Standard exception policy.
-
getInternalTablePropertySet
Request set of properties associated with a table.Returns a property object containing all properties that the store knows about, which are stored persistently by the store. This set of properties may vary from implementation to implementation of the store.
This call is meant to be used only for internal query of the properties by jbms, for instance by language during bulk insert so that it can create a new conglomerate which exactly matches the properties that the original container was created with. This call should not be used by the user interface to present properties to users as it may contain properties that are meant to be internal to jbms. Some properties are meant only to be specified by jbms code and not by users on the command line.
Note that not all properties passed into createConglomerate() are stored persistently, and that set may vary by store implementation.
- Parameters:
prop
- Property list to add properties to. If null, routine will create a new Properties object, fill it in and return it.- Throws:
StandardException
- Standard exception policy.
-
getTableProperties
Request the system properties associated with a table.Request the value of properties that are associated with a table. The following properties can be requested: derby.storage.pageSize derby.storage.pageReservedSpace derby.storage.minimumRecordSize derby.storage.initialPages
To get the value of a particular property add it to the property list, and on return the value of the property will be set to it's current value. For example: get_prop(ConglomerateController cc) { Properties prop = new Properties(); prop.put("derby.storage.pageSize", ""); cc.getTableProperties(prop); System.out.println( "table's page size = " + prop.getProperty("derby.storage.pageSize"); }
- Parameters:
prop
- Property list to fill in.- Throws:
StandardException
- Standard exception policy.
-
getXactMgr
Public Accessors of This class: -
getRawTran
-
getContainer
-
getOpenMode
public final int getOpenMode() -
getConglomerate
-
getHold
public final boolean getHold() -
isForUpdate
public final boolean isForUpdate() -
isClosed
public final boolean isClosed() -
isUseUpdateLocks
public final boolean isUseUpdateLocks() -
getRuntimeMem
-
checkConsistency
Check consistency of a conglomerate.Checks the consistency of the data within a given conglomerate, does not check consistency external to the conglomerate (ie. does not check that base table row pointed at by a secondary index actually exists).
There is no checking in the default implementation, you must override to get conglomerate specific consistency checking.
- Throws:
StandardException
- Standard exception policy.
-
debugConglomerate
- Throws:
StandardException
-
getSpaceInfo
Get information about space used by the conglomerate.- Throws:
StandardException
-
isKeyed
protected boolean isKeyed() -
isTableLocked
protected boolean isTableLocked()is the open btree table locked? -
init
public ContainerHandle init(ContainerHandle open_container, Conglomerate conglomerate, int[] format_ids, int[] collation_ids, TransactionManager xact_manager, Transaction rawtran, boolean hold, int openmode, int lock_level, LockingPolicy locking_policy, DynamicCompiledOpenConglomInfo dynamic_info) throws StandardException Open the container.Open the container, obtaining necessary locks. Most work is actually done by RawStore.openContainer().
- Throws:
StandardException
- Standard exception policy.
-
reopen
Open the container.Open the container, obtaining necessary locks. Most work is actually done by RawStore.openContainer(). Will only reopen() if the container is not already open.
- Throws:
StandardException
- Standard exception policy.
-
close
Close the container.Handles being closed more than once.
- Throws:
StandardException
- Standard exception policy.
-