#include <connection_base.hxx>
Inheritance diagram for pqxx::connection_base:
Capabilities | |
Some functionality is only available in certain versions of the backend, or only when speaking certain versions of the communications protocol that connects us to the backend. This includes clauses for SQL statements that were not accepted in older database versions, but are required in newer versions to get the same behaviour. | |
enum | capability { cap_prepared_statements, cap_create_table_with_oids, cap_nested_transactions, cap_cursor_scroll, cap_cursor_with_hold, cap_cursor_update, cap_table_column, cap_end } |
Session capabilities. More... | |
bool | supports (capability c) const throw () |
Does this connection seem to support the given capability? | |
int | protocol_version () const throw () |
What version of the PostgreSQL protocol is this connection using? | |
int | server_version () const throw () |
What version of the PostgreSQL server are we connected to? | |
Public Member Functions | |
void | disconnect () throw () |
Explicitly close connection. | |
bool | is_open () const throw () |
Is this connection open at the moment? | |
void | trace (FILE *) throw () |
Enable tracing to a given output stream, or NULL to disable. | |
void | set_client_encoding (const PGSTD::string &Encoding) |
Set client-side character encoding. | |
void | set_variable (const PGSTD::string &Var, const PGSTD::string &Value) |
Set session variable. | |
PGSTD::string | get_variable (const PGSTD::string &) |
Read session variable. | |
PGSTD::string | adorn_name (const PGSTD::string &) |
Suffix unique number to name to make it unique within session context. | |
Activation | |
Connections can be temporarily deactivated, or they can break because of overly impatient firewalls dropping TCP connections. Where possible, libpqxx will try to re-activate these when resume using them, or you can wake them up explicitly. You probably won't need this feature, but you should be aware of it. | |
void | activate () |
Explicitly activate deferred or deactivated connection. | |
void | deactivate () |
Explicitly deactivate connection. | |
void | inhibit_reactivation (bool inhibit) |
Disallow (or permit) connection recovery. | |
void | simulate_failure () |
Make the connection fail. | |
Error/warning output | |
Whenever the database has a warning or error to report, it will call a noticer to process the associated message. The default noticer sends the text of the message to standard error output, but you may choose to select a different noticer for the connection. | |
PGSTD::auto_ptr< noticer > | set_noticer (PGSTD::auto_ptr< noticer > N) throw () |
Set handler for postgresql errors or warning messages. | |
noticer * | get_noticer () const throw () |
void | process_notice (const char[]) throw () |
Invoke notice processor function. The message should end in newline. | |
void | process_notice (const PGSTD::string &) throw () |
Invoke notice processor function. Newline at end is recommended. | |
Connection properties | |
These are probably not of great interest, since most are derived from information supplied by the client program itself, but they are included for completeness. | |
const char * | dbname () |
Name of database we're connected to, if any. | |
const char * | username () |
Database user ID we're connected under, if any. | |
const char * | hostname () |
Address of server, or NULL if none specified (i.e. default or local). | |
const char * | port () |
Server port number we're connected to. | |
int | backendpid () const throw () |
Process ID for backend process. | |
int | sock () const throw () |
Socket currently used for connection, or -1 for none. Use with care! | |
Notifications and Triggers | |
int | get_notifs () |
Check for pending trigger notifications and take appropriate action. | |
int | await_notification () |
Wait for a trigger notification notification to come in. | |
int | await_notification (long seconds, long microseconds) |
Wait for a trigger notification to come in, or for given timeout to pass. | |
Prepared statements | |
PostgreSQL supports prepared SQL statements, i.e. statements that can be registered under a client-provided name, optimized once by the backend, and executed any number of times under the given name. Prepared statement definitions are not sensitive to transaction boundaries; a statement defined inside a transaction will remain defined outside that transaction, even if the transaction itself is subsequently aborted. Once a statement has been prepared, only closing the connection or explicitly "unpreparing" it can make it go away. Use the transaction classes' exec_prepared() functions to execute a prepared statement.
| |
prepare::declaration | prepare (const PGSTD::string &name, const PGSTD::string &definition) |
Define a prepared statement. | |
void | unprepare (const PGSTD::string &name) |
Drop prepared statement. | |
void | prepare_now (const PGSTD::string &name) |
Request that prepared statement be registered with the server. | |
Transactor framework | |
See the transactor class template for more about transactors. To use the transactor framework, encapsulate your transaction code in a class derived from an instantiation of the pqxx::transactor template. Then, to execute it, create an object of your transactor class and pass it to one of the perform() functions here. The perform() functions may create and execute several copies of the transactor before succeeding or ultimately giving up. If there is any doubt over whether execution succeeded (this can happen if the connection to the server is lost just before the backend can confirm success), it is no longer retried and an in_doubt_error is thrown.
Take care: no member functions will ever be invoked on the transactor object you pass into perform(). The object you pass in only serves as a "prototype" for the job to be done. The perform() function will copy-construct transactors from the original you passed in, executing the copies only. The original object remains "clean" in its original state. | |
template<typename TRANSACTOR> | |
void | perform (const TRANSACTOR &T, int Attempts) |
Perform the transaction defined by a transactor-based object. | |
template<typename TRANSACTOR> | |
void | perform (const TRANSACTOR &T) |
Perform the transaction defined by a transactor-based object. | |
Protected Member Functions | |
connection_base (connectionpolicy &) | |
void | init () |
void | close () throw () |
void | wait_read () const |
void | wait_read (long seconds, long microseconds) const |
void | wait_write () const |
Friends | |
class | subtransaction |
class | prepare::declaration |
class | transaction_base |
class | largeobject |
class | trigger |
class | pipeline |
class | cursor_base |
class | dbtransaction |
class | internal::reactivation_avoidance_exemption |
This is the first class to look at when you wish to work with a database through libpqxx. Depending on the implementing concrete child class, a connection can be automatically opened when it is constructed, or when it is first used. The connection is automatically closed upon destruction, if it hasn't already been closed manually.
To query or manipulate the database once connected, use one of the transaction classes (see pqxx/transaction_base.hxx) or preferably the transactor framework (see pqxx/transactor.hxx).
If a network connection to the database server fails, the connection will be restored automatically (although any transaction going on at the time will have to be aborted). This also means that any information set in previous transactions that is not stored in the database, such as connection-local variables defined with PostgreSQL's SET command, will be lost. Whenever you create such state, either do it within each transaction that may need it, or if at all possible, use specialized functions made available by libpqxx. Always avoid raw queries if libpqxx offers a dedicated function for the same purpose.
Session capabilities.
pqxx::connection_base::connection_base | ( | connectionpolicy & | ) | [explicit, protected] |
void pqxx::connection_base::disconnect | ( | ) | throw () |
Explicitly close connection.
bool pqxx::connection_base::is_open | ( | ) | const throw () |
Is this connection open at the moment?
void pqxx::connection_base::activate | ( | ) |
Explicitly activate deferred or deactivated connection.
Use of this method is entirely optional. Whenever a connection is used while in a deferred or deactivated state, it will transparently try to bring itself into an activated state. This function is best viewed as an explicit hint to the connection that "if you're not in an active state, now would be a good time to get into one." Whether a connection is currently in an active state or not makes no real difference to its functionality. There is also no particular need to match calls to activate() with calls to deactivate(). A good time to call activate() might be just before you first open a transaction on a lazy connection.
void pqxx::connection_base::deactivate | ( | ) |
Explicitly deactivate connection.
Like its counterpart activate(), this method is entirely optional. Calling this function really only makes sense if you won't be using this connection for a while and want to reduce the number of open connections on the database server. There is no particular need to match or pair calls to deactivate() with calls to activate(), but calling deactivate() during a transaction is an error.
void pqxx::connection_base::inhibit_reactivation | ( | bool | inhibit | ) |
Disallow (or permit) connection recovery.
A connection whose underlying socket is not currently connected to the server will normally (re-)establish communication with the server whenever needed, or when the client program requests it (although for reasons of integrity, never inside a transaction; but retrying the whole transaction may implicitly cause the connection to be restored). In normal use this is quite a convenient thing to have and presents a simple, safe, predictable interface.
There is at least one situation where this feature is not desirable, however. Although most session state (prepared statements, session variables) is automatically restored to its working state upon connection reactivation, temporary tables and so-called WITH HOLD cursors (which can live outside transactions) are not.
Cursors that live outside transactions are automatically handled, and the library will quietly ignore requests to deactivate or reactivate connections while they exist; it does not want to give you the illusion of being back in your transaction when in reality you just dropped a cursor. With temporary tables this is not so easy: there is no easy way for the library to detect their creation or track their lifetimes.
So if your program uses temporary tables, and any part of this use happens outside of any database transaction (or spans multiple transactions), some of the work you have done on these tables may unexpectedly be undone if the connection is broken or deactivated while any of these tables exists, and then reactivated or implicitly restored before you are finished with it.
If this describes any part of your program, guard it against unexpected reconnections by inhibiting reconnection at the beginning. And if you want to continue doing work on the connection afterwards that no longer requires the temp tables, you can permit it again to get the benefits of connection reactivation for the remainder of the program.
inhibit | should reactivation be inhibited from here on? |
void pqxx::connection_base::simulate_failure | ( | ) |
Make the connection fail.
Breaks the connection in some unspecified, horrible, dirty way to enable failure testing.
Do not use this in normal programs. This is only meant for testing.
auto_ptr< pqxx::noticer > pqxx::connection_base::set_noticer | ( | PGSTD::auto_ptr< noticer > | N | ) | throw () |
Set handler for postgresql errors or warning messages.
The use of auto_ptr implies ownership, so unless the returned value is copied to another auto_ptr, it will be deleted directly after the call. This may be important when running under Windows, where a DLL cannot free memory allocated by the main program. The auto_ptr will delete the object from your code context, rather than from inside the library.
If a noticer exists when the connection_base is destructed, it will also be deleted.
N | New message handler; must not be null or equal to the old one |
noticer* pqxx::connection_base::get_noticer | ( | ) | const throw () |
void pqxx::connection_base::process_notice | ( | const | char[] | ) | throw () |
Invoke notice processor function. The message should end in newline.
void pqxx::connection_base::process_notice | ( | const PGSTD::string & | ) | throw () |
Invoke notice processor function. Newline at end is recommended.
void pqxx::connection_base::trace | ( | FILE * | ) | throw () |
Enable tracing to a given output stream, or NULL to disable.
const char * pqxx::connection_base::dbname | ( | ) |
Name of database we're connected to, if any.
const char * pqxx::connection_base::username | ( | ) |
Database user ID we're connected under, if any.
const char * pqxx::connection_base::hostname | ( | ) |
Address of server, or NULL if none specified (i.e. default or local).
const char * pqxx::connection_base::port | ( | ) |
Server port number we're connected to.
int pqxx::connection_base::backendpid | ( | ) | const throw () |
Process ID for backend process.
Use with care: connections may be lost and automatically re-established without your knowledge, in which case this process ID may no longer be correct. You may, however, assume that this number remains constant and reliable within the span of a successful backend transaction. If the transaction fails, which may be due to a lost connection, then this number will have become invalid at some point within the transaction.
int pqxx::connection_base::sock | ( | ) | const throw () |
Socket currently used for connection, or -1 for none. Use with care!
Query the current socket number. This is intended for event loops based on functions such as select() or poll(), where multiple file descriptors are watched.
Please try to stay away from this function. It is really only meant for event loops that need to wait on more than one file descriptor. If all you need is to block until a trigger notification arrives, for instance, use await_notification(). If you want to issue queries and retrieve results in nonblocking fashion, check out the pipeline class.
bool pqxx::connection_base::supports | ( | capability | c | ) | const throw () |
Does this connection seem to support the given capability?
Don't try to be smart by caching this information anywhere. Obtaining it is quite fast (especially after the first time) and what's more, a capability may "suddenly" appear or disappear if the connection is broken or deactivated, and then restored. This may happen silently any time no backend transaction is active; if it turns out that the server was upgraded or restored from an older backup, or the new connection goes to a different backend, then the restored session may have different capabilities than were available previously.
Some guesswork is involved in establishing the presence of any capability; try not to rely on this function being exactly right. Older versions of libpq may not detect any capabilities.
int pqxx::connection_base::protocol_version | ( | ) | const throw () |
What version of the PostgreSQL protocol is this connection using?
The answer can be 0 (when there is no connection, or the libpq version being used is too old to obtain the information); 2 for protocol 2.0; 3 for protocol 3.0; and possibly higher values as newer protocol versions are taken into use.
If the connection is broken and restored, the restored connection could possibly a different server and protocol version. This would normally happen if the server is upgraded without shutting down the client program, for example.
Requires libpq version from PostgreSQL 7.4 or better.
int pqxx::connection_base::server_version | ( | ) | const throw () |
What version of the PostgreSQL server are we connected to?
The result is a bit complicated: each of the major, medium, and minor release numbers is written as a two-digit decimal number, and the three are then concatenated. Thus server version 7.4.2 will be returned as the decimal number 70402. If there is no connection to the server, of if the libpq version is too old to obtain the information, zero is returned.
void pqxx::connection_base::set_client_encoding | ( | const PGSTD::string & | Encoding | ) |
Set client-side character encoding.
Search the PostgreSQL documentation for "multibyte" or "character set encodings" to find out more about the available encodings, how to extend them, and how to use them. Not all server-side encodings are compatible with all client-side encodings or vice versa.
Encoding | Name of the character set encoding to use |
void pqxx::connection_base::set_variable | ( | const PGSTD::string & | Var, | |
const PGSTD::string & | Value | |||
) |
Set session variable.
Set a session variable for this connection, using the SET command. If the connection to the database is lost and recovered, the last-set value will be restored automatically. See the PostgreSQL documentation for a list of variables that can be set and their permissible values. If a transaction is currently in progress, aborting that transaction will normally discard the newly set value. Known exceptions are nontransaction (which doesn't start a real backend transaction) and PostgreSQL versions prior to 7.3.
Var | Variable to set | |
Value | Value vor Var to assume: an identifier, a quoted string, or a number. |
string pqxx::connection_base::get_variable | ( | const PGSTD::string & | ) |
Read session variable.
Will try to read the value locally, from the list of variables set with the set_variable function. If that fails, the database is queried.
int pqxx::connection_base::get_notifs | ( | ) |
Check for pending trigger notifications and take appropriate action.
All notifications found pending at call time are processed by finding any matching triggers and invoking those. If no triggers matched the notification string, none are invoked but the notification is considered processed.
Exceptions thrown by client-registered trigger handlers are reported, but not passed on outside this function.
int pqxx::connection_base::await_notification | ( | ) |
Wait for a trigger notification notification to come in.
The wait may also be terminated by other events, such as the connection to the backend failing. Any pending or received notifications are processed as part of the call.
int pqxx::connection_base::await_notification | ( | long | seconds, | |
long | microseconds | |||
) |
Wait for a trigger notification to come in, or for given timeout to pass.
The wait may also be terminated by other events, such as the connection to the backend failing. Any pending or received notifications are processed as part of the call.
pqxx::prepare::declaration pqxx::connection_base::prepare | ( | const PGSTD::string & | name, | |
const PGSTD::string & | definition | |||
) |
Define a prepared statement.
To declare parameters to this statement, add them by calling the function invocation operator (operator()
) on the returned object. See prepare_param_declaration and prepare::param_treatment for more about how to do this.
The statement's definition can refer to a parameter using the parameter's positional number n in the definition. For example, the first parameter can be used as a variable "$1", the second as "$2" and so on.
One might use a prepared statement as in the following example. Note the unusual syntax associated with parameter definitions and parameter passing: every new parameter is just a parenthesized expression that is simply tacked onto the end of the statement!
using namespace pqxx; void foo(connection_base &C) { C.prepare("findtable", "select * from pg_tables where name=$1") ("varchar", treat_string); work W(C); result R = W.prepared("findtable")("mytable").exec(); if (R.empty()) throw runtime_error("mytable not found!"); }
For better performance, prepared statements aren't really registered with the backend until they are first used. If this is not what you want, e.g. because you have very specific realtime requirements, you can use the prepare_now()
function to force immediate preparation.
name | unique identifier to associate with new prepared statement | |
definition | SQL statement to prepare |
void pqxx::connection_base::unprepare | ( | const PGSTD::string & | name | ) |
Drop prepared statement.
void pqxx::connection_base::prepare_now | ( | const PGSTD::string & | name | ) |
Request that prepared statement be registered with the server.
If the statement had already been fully prepared, this will do nothing.
If the connection should break and be transparently restored, then the new connection will again defer registering the statement with the server. Since connections are never restored inside backend transactions, doing this once at the beginning of your transaction ensures that the statement will not be re-registered during that transaction. In most cases, however, it's probably better not to use this and let the connection decide when and whether to register prepared statements that you've defined.
void pqxx::connection_base::perform | ( | const TRANSACTOR & | T, | |
int | Attempts | |||
) |
Perform the transaction defined by a transactor-based object.
Invokes the given transactor, making at most Attempts attempts to perform the encapsulated code. If the code throws any exception other than broken_connection, it will be aborted right away.
T | The transactor to be executed. | |
Attempts | Maximum number of attempts to be made to execute T. |
void pqxx::connection_base::perform | ( | const TRANSACTOR & | T | ) |
Perform the transaction defined by a transactor-based object.
T | The transactor to be executed. |
string pqxx::connection_base::adorn_name | ( | const PGSTD::string & | ) |
Suffix unique number to name to make it unique within session context.
Used internally to generate identifiers for SQL objects (such as cursors and nested transactions) based on a given human-readable base name.
void pqxx::connection_base::init | ( | ) | [protected] |
void pqxx::connection_base::close | ( | ) | throw () [protected] |
void pqxx::connection_base::wait_read | ( | ) | const [protected] |
void pqxx::connection_base::wait_read | ( | long | seconds, | |
long | microseconds | |||
) | const [protected] |
void pqxx::connection_base::wait_write | ( | ) | const [protected] |
friend class subtransaction [friend] |
friend class prepare::declaration [friend] |
friend class transaction_base [friend] |
friend class largeobject [friend] |
friend class trigger [friend] |
friend class pipeline [friend] |
friend class cursor_base [friend] |
friend class dbtransaction [friend] |
friend class internal::reactivation_avoidance_exemption [friend] |