File: //usr/include/pqxx/dbtransaction.hxx
/*-------------------------------------------------------------------------
*
* FILE
* pqxx/dbtransaction.hxx
*
* DESCRIPTION
* definition of the pqxx::dbtransaction abstract base class.
* pqxx::dbransaction defines a real transaction on the database
* DO NOT INCLUDE THIS FILE DIRECTLY; include pqxx/dbtransaction instead.
*
* Copyright (c) 2004-2015, Jeroen T. Vermeulen <jtv@xs4all.nl>
*
* See COPYING for copyright license. If you did not receive a file called
* COPYING with this source code, please notify the distributor of this mistake,
* or contact the author.
*
*-------------------------------------------------------------------------
*/
#ifndef PQXX_H_DBTRANSACTION
#define PQXX_H_DBTRANSACTION
#include "pqxx/compiler-public.hxx"
#include "pqxx/compiler-internal-pre.hxx"
#include "pqxx/transaction_base"
namespace pqxx
{
enum readwrite_policy
{
read_only,
read_write
};
/// Abstract base class responsible for bracketing a backend transaction
/**
* @addtogroup transaction Transaction classes
*
* Use a dbtransaction-derived object such as "work" (transaction<>) to enclose
* operations on a database in a single "unit of work." This ensures that the
* whole series of operations either succeeds as a whole or fails completely.
* In no case will it leave half-finished work behind in the database.
*
* Once processing on a transaction has succeeded and any changes should be
* allowed to become permanent in the database, call commit(). If something
* has gone wrong and the changes should be forgotten, call abort() instead.
* If you do neither, an implicit abort() is executed at destruction time.
*
* It is an error to abort a transaction that has already been committed, or to
* commit a transaction that has already been aborted. Aborting an already
* aborted transaction or committing an already committed one has been allowed
* to make errors easier to deal with. Repeated aborts or commits have no
* effect after the first one.
*
* Database transactions are not suitable for guarding long-running processes.
* If your transaction code becomes too long or too complex, please consider
* ways to break it up into smaller ones. There's no easy, general way to do
* this since application-specific considerations become important at this
* point.
*
* The actual operations for beginning and committing/aborting the backend
* transaction are implemented by a derived class. The implementing concrete
* class must also call Begin() and End() from its constructors and destructors,
* respectively, and implement do_exec().
*/
class PQXX_LIBEXPORT PQXX_NOVTABLE dbtransaction : public transaction_base
{
public:
virtual ~dbtransaction();
protected:
dbtransaction(
connection_base &,
const std::string &IsolationString,
readwrite_policy rw=read_write);
explicit dbtransaction(
connection_base &,
bool direct=true,
readwrite_policy rw=read_write);
/// Start a transaction on the backend and set desired isolation level
void start_backend_transaction();
/// Sensible default implemented here: begin backend transaction
virtual void do_begin(); //[t1]
/// Sensible default implemented here: perform query
virtual result do_exec(const char Query[]);
/// To be implemented by derived class: commit backend transaction
virtual void do_commit() =0;
/// Sensible default implemented here: abort backend transaction
/** Default implementation does two things:
* <ol>
* <li>Clears the "connection reactivation avoidance counter"</li>
* <li>Executes a ROLLBACK statement</li>
* </ol>
*/
virtual void do_abort() PQXX_OVERRIDE; //[t13]
static std::string fullname(const std::string &ttype,
const std::string &isolation);
private:
/// Precomputed SQL command to run at start of this transaction
std::string m_StartCmd;
};
} // namespace pqxx
#include "pqxx/compiler-internal-post.hxx"
#endif