|
|
DbSequence::get |
#include <db_cxx.h>int DbSequence::get(DbTxn *txnid, int32_t delta, db_seq_t *retp, u_int32_t flags);
The DbSequence::get method returns the next available element in the sequence and changes the sequence value by delta. The value of delta must be greater than zero. If there are enough cached values in the sequence handle then they will be returned. Otherwise the next value will be fetched from the database and incremented (decremented) by enough to cover the delta and the next batch of cached values.
For maximum concurrency a non-zero cache size should be specified prior to opening the sequence handle and DB_TXN_NOSYNC should be specified for each DbSequence::get method call.
By default, sequence ranges do not wrap; to cause the sequence to wrap around the beginning or end of its range, specify the DB_SEQ_WRAP flag to the DbSequence::set_flags method.
The DbSequence::get method will return EINVAL if the record in the database is not a valid sequence record, or the sequence has reached the beginning or end of its range and is not configured to wrap.
If the underlying database handle was opened in a transaction, calling DbSequence::get may result in changes to the sequence object; these changes will be automatically committed in a transaction internal to the Berkeley DB library. If the thread of control calling DbSequence::get has an active transaction, which holds locks on the same database as the one in which the sequence object is stored, it is possible for a thread of control calling DbSequence::get to self-deadlock because the active transaction's locks conflict with the internal transaction's locks. For this reason, it is often preferable for sequence objects to be stored in their own database.
Copyright (c) 1996-2005 Sleepycat Software, Inc. - All rights reserved.