Jaybird 2.1 JDBC driver


Jaybird 2.1 JDBC driver Java Programmer's Manual The contents of this Documentation are subject to the Public Documentation License Version 1.0 (the “License”); you may only use this Documentation if you comply with the terms of this License. A copy of the License is available at http://www.firebirdsql.org/manual/licenses-pdl-text.html. The Original Documentation is _________________. The Initial Writer of the Original Documentation is Roman Rokytskyy, Copyright (C) 2004-2008. All Rights Reserved. (Initial Writer contact: roman@rokytskyy.de). Contributor(s): ______________________________________. Portions created by ______ are Copyright (C) _________ [Insert year(s)]. All Rights Reserved. (Contributor contact(s): ________________ [Insert hyperlink/alias]). Table Of Contents Table of Contents 1. Introduction.....................................................................................................7 Jaybird architecture.........................................................................................7 Supported Servers...........................................................................................8 Supported Specifications................................................................................9 Distribution package.......................................................................................9 Quality Assurance.........................................................................................10 Useful resources............................................................................................10 2. Obtaining a connection..................................................................................11 Obtaining connection java.sql.DriverManager.............................................11 Driver types..................................................................................................15 Connection Pooling......................................................................................17 The javax.sql.ConnectionPoolDataSource implementation.........................17 Using FBConnectionPoolDataSource with JNDI.........................................19 The javax.sql.DataSource implementation...................................................21 The javax.sql.XADataSource implementation.............................................22 3. Handling exceptions......................................................................................25 Working with exceptions..............................................................................25 Warnings.......................................................................................................28 java.sql.SQLException in Jaybird................................................................28 SQL states.....................................................................................................29 Useful Firebird error codes...........................................................................30 4. Executing statements.....................................................................................35 The java.sql.Statement interface...................................................................35 Statement behind the scenes.........................................................................39 The java.sql.PreparedStatement interface.....................................................40 Prepared statement pooling...........................................................................44 The java.sql.CallableStatement interface......................................................46 Batch Updates...............................................................................................52 Escaped Syntax.............................................................................................53 5. Working with result sets...............................................................................55 ResultSet properties......................................................................................55 ResultSet manipulation.................................................................................57 6. Using transactions.........................................................................................61 JDBC transactions........................................................................................61 Auto-commit mode.......................................................................................62 Read-only Transactions................................................................................64 Transaction Isolation Levels.........................................................................65 Savepoints.....................................................................................................66 Transaction Parameter Buffer.......................................................................67 Table Reservation.........................................................................................70 7. Working with Services..................................................................................73 ServiceManager............................................................................................74 Backup and restore.......................................................................................74 User management.........................................................................................79 Database maintenance..................................................................................80 Database statistics.........................................................................................85 8. Working with Events.....................................................................................87 Database events............................................................................................87 Posting the events.........................................................................................88 Subscribing for events..................................................................................89 Appendix A. Extended connection properties...................................................91 Appendix B. Data Type Conversion Table.......................................................93 Appendix C. Connection Pool Properties..........................................................95 Standard JDBC Properties............................................................................95 Pool Properties..............................................................................................96 Runtime Pool Properties...............................................................................96 Firebird-specific Properties..........................................................................96 Non-standard parameters..............................................................................97 Appendix D. Character Encodings....................................................................99 Encodings Types...........................................................................................99 Encodings in Java.......................................................................................100 Available Encodings...................................................................................102 Appendix E. Supported JDBC Scalar Functions.............................................105 Numeric Functions......................................................................................106 String Functions..........................................................................................107 String Functions (continued)......................................................................108 Time and Date Functions............................................................................109 Time and Date Functions (continued)........................................................109 System Functions........................................................................................110 Conversion Functions.................................................................................110 Java Programmer's Manual 4 Chapter 1. 1. Introduction Jaybird is a JCA/JDBC driver suite to connect to Firebird database server. When Borland released an open-source version of the InterBase RDBMS, it included sources for a type 3 JDBC driver called InterClient. However due to some inherent limitations of the InterBase (and later Firebird) client library, it was decided that the type 3 driver was a dead end, and the Firebird team developed a pure Java implementation of the wire protocol. This implementation became the basis for Jaybird, a pure Java driver for Firebird relational database. This driver is based on both the new JCA standard for application server connections to enterprise information systems and the well known JDBC standard. The JCA standard specifies an architecture in which an application server can cooperate with a driver so that the application server manages transactions, security, and resource pooling, and the driver supplies only the connection functionality. Jaybird architecture The Jaybird driver consists of three layers, each of which is responsible for its part of the functionality. The component diagram depicting the Jaybird internal structure contains two additional components: “pool” and “JMX manager”. • The GDS layer represents a Java translation of the Firebird API. It is represented by two classes from org.firebirdsql.gds package: GDS interface and GDSFactory. GDS factory class is responsible for instantiating an implementation of the GDS interface depending of the type of driver used. Implementation of the GDS interface determines the type of the driver that will be used. • The JCA layer represents the heart of the driver. Here all connection and transaction management happens. Additionally this layer adapts the GDS API and proxies the calls to the GDS implementation. • The JDBC layer is an implementation of the JDBC specification. • The Pool component represents implementation of ConnectionPoolDataSource, DataSource and XADataSource interfaces from the JDBC 2.0 Optional Package. The pool implementation uses ManagedConnectionFactory to create physical connections to the database. • The Manager component represents a JMX 1.0 compatible implementation that uses The Services API to manage the database and the server itself. Currently only calls to create and drop database are available, but in the future a full range of services will be made public: database backup/restore, user management, statistics gathering, etc. Supported Servers Jaybird 2.1 supports all current Firebird servers, however no support for the optimized wire protocol from the Firebird 2.1 is being implemented. Jaybird versions 2.x in current distribution won't work with InterBase and Yaffil servers, however the compatibility might be restored in the future. Supported Specifications Jaybird supports the following specifications: Chapter 1. Introduction 6 id Architecture GDS JDBC JCA ManagedConnection ManagedConnectionFactory Type 4 Type 2 Pool JMX Manager«delegate»«delegate» Illustration 1.1 Jaybird internal structure. Specification Details JDBC 3.0 Driver passes the complete JDBC compatibility test suite, though some features are not implemented. It is not officially JDBC compliant, because of the high certification costs. JDBC 2.0 Optional Package (formerly Standard Extension API) Jaybird provides an implementation of following interfaces from javax.sql.* package: ● The ConnectionPoolDataSource implementation provides connection and prepared statement pooling. ● The DataSource implementation provides seamless integration with major web and application servers. ● The XADataSource implementation provides means to use driver in distributed transactions. JCA 1.0 Jaybird provides an implementation of javax.resource.spi.ManagedConnectionFactory and related interfaces. CCI interfaces are not supported. JTA 1.0.1 The driver provides implementation of the javax.transaction.xa.XAResource interface via the JCA framework and a javax.sql.XADataSource implementation. JAAS 1.0 JMX 1.2 Jaybird provides an MBean that allows creating and dropping databases via JMX agent. Distribution package Jaybird driver has compile-time and run-time dependencies to JCA 1.0, JTA 1.0.1, JAAS 1.0 and JDBC 2.0 Optional Package. Additionally, if Log4J classes are found in the class path, it is possible to enable extensive logging inside the driver. Following file groups can be found in distribution package: File name Description jaybird-2.1.6.jar An archive containing the JDBC driver, the JCA connection manager, the Services API and event management classes. jaybird-full-2.1.6.jar Same as above but also the connection pooling classes. jaybird21.dll Precompiled version of the JNI library for Type 2 and Embedded Server drivers for 32-bit Windows platform. libjaybird21.so Precompiled version of the JNI library for Type 2 and Embedded Server drivers for 32-bit Linux platforms. Quality Assurance The Jaybird team uses JUnit test cases to assure the quality of the released driver. Also during development unit tests are extensively used. It is not allowed to commit a code to the CVS until it passes all existing unit tests. Also each Chapter 1. Introduction 7 reproducible bug usually gets its own test case. This guarantees that a clean check out from the CVS can be compiled and will not contain any previously discovered and fixed bug. Currently there are more than 450 test cases covering most of the driver code. Additionally, before the driver is released, is is required to pass JDBC compatibility suite (CTS), which currently contains 1216 test cases. Unfortunately Firebird does not support all features used by the CTS, so some test cases from the original CTS suite were excluded from run. Useful resources JDBC For extensive JDBC documentation, see the “Documentation” section of Sun's website http://java.sun.com/products/jdbc/. Firebird General information about the Firebird database is available from the Firebird web site (http://www.firebirdsql.org). Information about using SQL in Firebird, see the “Language Reference” and “Developer's Guide” documents, that are available for download from the “Main Downloads” section of the IBPhoenix web site. Jaybird Support A new resource JaybirdWiki has become available. It can be found at http://jaybirdwiki.firebirdsql.org. This is a place where the community shares information about different aspects of Jaybird usage, configuration examples for different applications/servers, tips and tricks, FAQ, etc. Chapter 1. Introduction 8 Chapter 2. 2. Obtaining a connection Jaybird is regular JDBC driver and supports two primary ways to obtain connection: via java.sql.DriverManager and via javax.sql.DataSource interface. Obtaining connection java.sql.DriverManager java.sql.DriverManager historically was the first connection factory in Java. It is based on the concept of the JDBC URL, a string that uniquely identifies JDBC driver to use and the database to which user wants to connect. Additionally there is possibility to specify additional connection parameters, like user name and password. JDBC URL consists of three parts that are presented on Illustration 2.1. Illustration 2.1: Structure of the JDBC URL First part, "jdbc:firebirdsql:" is always fixed and specifies the so called protocol and subprotocol for the JDBC connection. In other words, the type of the localhost/3050:c:/database/example.fdbfirebirdsql:jdbc: JDBC protocol JDBC subprotocol, identifies driver to use RDBMS specific part, identifies the database to which driver must connect, in our case that is /: connection that the application wants to obtain, in our case it is a connection to a Firebird database. Example of obtaining the connection is shown on Illustration 2.2. package hello; import java.sql.*; public class HelloServer { public static void main(String[] args) throws Exception { Class.forName("org.firebirdsql.jdbc.FBDriver"); Connection connection = DriverManager.getConnection( "jdbc:firebirdsql:localhost/3050:c:/db/employee.fdb", "SYSDBA", "masterkey"); // do something here } } Illustration 2.2: Simple example shows how to obtain JDBC connection. The first line of this code is important – it tells JVM to load the Jaybird 2.1 JDBC driver. According to JDBC specification, at this point driver registers itself in java.sql.DriverManager and tells it for which protocol it is responsible for. There are two ways to register JDBC driver: • Possibility 1. The application loads the driver's class. The JDBC specification requires that during class initialization the driver performs the registration itself. Class.forName("org.firebirdsql.jdbc.FBDriver"); • Possibility 2. The JDBC driver is listed in a jdbc.drivers system property. For example in your ~/.hotjava/properties file you can specify following line: jdbc.drivers=foo.Driver:org.firebirdsql.jdbc.FBDriver Alternatively you can specify the value of this property during JVM startup: java -Djdbc.drivers=org.firebirdsql.jdbc.FBDriver -classpath jaybird-full-2.1.6.jar;C:/myproject/classes my.company.SomeJavaExample The second statement of the example tells the java.sql.DriverManager to open database connection to the Firebird server running on the host where Java code is executed, and the path to the database is c:/database/employee.fdb. Database specification consists of the name of the server where the database server resides, optionally you can specify a port to which the driver will connect (by default port 3050 is used). The server name can be specified either using its IP address (for example 192.168.0.5) or using its DNS name (for example fb- server.mycompany.com or just fb-server). After the server name and port, path to the database is specified. The format in which the path is specified depends on the platform where the Firebird server Chapter 2. Obtaining a connection 10 runs. On Windows it must include the drive letter and path, for example "c:/database/employee.gdb", which points to the employee database that can be found in a root directory of drive C:. Java allows to use either "/" or "\\" as path separator on the Windows platform. On Unix and Linux platform, you can use only "/" as the path separator. If you are using Firebird 1.5 or higher, you can specify a database alias instead of the absolute database path. For more information about using aliases see the documentation of the Firebird server. Specifying extended properties What if we want to specify additional connection parameters, a client encoding, for example? JDBC specification provides another method that allows to specify additional connection properties (Illustration 2.3). package hello; import java.sql.*; import java.util.*; public class HelloServerWithEncoding { public static void main(String[] args) throws Exception { Class.forName("org.firebirdsql.jdbc.FBDriver"); Properties props = new Properties(); props.setProperty("user", "SYSDBA"); props.setProperty("password", "masterkey"); props.setProperty("encoding", "UNICODE_FSS"); Connection connection = DriverManager.getConnection( "jdbc:firebirdsql:localhost/3050:C:/employee.gdb", props); // do something here } } Illustration 2.3: Obtaining JDBC connection with additional connection properties. Additional properties, for example SQL role for the connection can be added to the props map. The list of all available extended properties can be found in Ошибка: источник перекрестной ссылки не найден. However, not in every place you can use the above described method. Jaybird provides a possibility to specify extended properties in the JDBC URL. Illustration 2.4 shows the specification for specifying extended JDBC properties in the URL. jdbc:firebirdsql:host[/port]:? ::= [&] ::= [=] Illustration 2.4: Extended JDBC URL format. In this case extended properties are passed together with the URL using the HTTP-like parameter passing scheme: first comes main part of the URL, then “?”, Chapter 2. Obtaining a connection 11 then name-value pairs separated with “&”. Code in Illustration 2.6 is equivalent to the previous example. import java.sql.*; ... Class.forName("org.firebirdsql.jdbc.FBDriver"); Connection connection = DriverManager.getConnection( "jdbc:firebirdsql:localhost/3050:C:/employee.gdb" + "?encoding=UNICODE_FSS", "SYSDBA", "masterkey"); Illustration 2.5: Example of specifying extended properties in JDBC URL. Obtaining a connection via javax.sql.DataSource JDBC 2.0 specification introduced a new mechanism to obtain database connection without requiring the application to know any specifics of the underlying JDBC driver. The application is required to know a logical name under which application can find an instance of the javax.sql.DataSource interface using Java Naming and Directory Interface (JNDI). This is a common way to obtain connections in web and application servers. In order to obtain a connection via DataSource object, you can use code showed on Illustration 2.6. This code assumes that you have correctly configured JNDI properties. For more information about configuring JNDI please refer to the documentation provided with your web or application server. package hello; import java.sql.*; import javax.sql.*; import javax.naming.*; public class HelloServerJNDI { public static void main(String[] args) throws Exception { InitialContext ctx = new InitialContext(); DataSource ds = (DataSource)ctx.lookup("jdbc/SomeDB"); Connection connection = ds.getConnection(); try { // do something here... } finally { connection.close(); } } } Illustration 2.6: Typical way to obtain JDBC connection via JNDI. Usually binding between the DataSource object and its JNDI name happens in the configuration of your web or application server. However under some circumstances (e.g. you are developing your own JNDI-enabled application server/framework) you have to do this yourself. You can use the code snippet for Chapter 2. Obtaining a connection 12 this purpose showed on Illustration 2.7. DataSource implementation supports all connection properties available to the DriverManager interface, but also it supports additional properties that control connection pooling. For more information on this topic please read the “Connection Pooling” chapter. Driver types As it was mentioned in the Chapter 3. Jaybird Architecture, Jaybird supports multiple implementations of the GDS interface. The original Jaybird distribution contains two main categories of the GDS implementation: pure Java implementation of the Firebird wire protocol and a JNI proxy that can use a dynamically linked library with a compatible API. Below you find the list of existing types and their short configuration description with the corresponding JDBC URLs that should be used to obtain the connection of desired type. The type of the JDBC driver for the javax.sql.DataSource is configured via corresponding property. PURE_JAVA type The PURE_JAVA driver type uses pure Java implementation of the Firebird wire protocol. This type is recommended for connecting to a remote database server using TCP/IP sockets. No installation is required except adding the JDBC driver to the class path. This type of driver provides best performance when connecting to the remote server. In order to obtain connection using the PURE_JAVA driver type you have to use JDBC URL that was shown on Illustration 2.2: jdbc:firebirdsql:host[/port]: When using javax.sql.DataSource implementation, you can specify either "PURE_JAVA" or "TYPE4" driver type, however this type is used by default. NATIVE and LOCAL types The NATIVE and LOCAL driver types use a JNI proxy to access the Firebird client library and requires installation of the Firebird client. The NATIVE driver type is used to access the remote database server, the LOCAL type accesses the database server running on the same host by means of IPC. Performance of NATIVE driver is approximately 10% lower compared to the PURE_JAVA driver, but LOCAL type has up to 30% higher performance compared to the Chapter 2. Obtaining a connection 13 PURE_JAVA driver when connecting the server on the same host. This is mostly due to the fact that TCP/IP stack is not involved in this mode. In order to instantiate a connection using the NATIVE JDBC driver to connect to a remote server you have to use the following JDBC URL with new subprotocol: jdbc:firebirdsql:native:host[/port]: When connecting to a local database server using the LOCAL driver, you should use following: jdbc:firebirdsql:local: Additionally to the Firebird client library installation, driver requires a JNI proxy to be available to the JVM. The JNI proxy is a platform-dependent dynamically linked library that translates GDS calls into Firebird API calls. Windows On Windows, the JNI proxy is represented by a dynamically linked library (DLL) jaybird21.dll. You have to make this library available through PATH environment variable. Alternatively you can specify the directory containing this DLL in java.library.path system property. For example, if you put library in the current directory you have to use the following command to start Java: java -Djava.library.path=. com.mycompany.MyClass Linux On Linux JNI proxy is represented by a shared library libjaybird21.so. It must be available through the LD_PATH environment variable. Usually shared libraries are stored in the /usr/lib/ directory; however you will need root permissions to copy libjaybird21.so there. Alternatively you can specify directory containing the proxy in java.library.path Java system property. See Windows example above for more details. Limitations Firebird client library is not thread-safe when connecting to a local database server using IPC. Jaybird provides the necessary synchronization in Java code using a static object instance. However, this static object instance is local to the classloader that has loaded the Jaybird classes. In order to guarantee correct synchronization , the Jaybird driver must be loaded by the top-most classloader. For example, when using the Type 2 JDBC driver with a web or application server, you have to add the Jaybird classes to the main classpath (for example, to the lib/ directory of your web or application server), but not to the web or J2EE application, e.g. the WEB-INF/lib directory. EMBEDDED type The Embedded server JDBC driver is the Type 2 JDBC driver that rather than using the Firebird client library, loads Firebird embedded server library instead. This is the highest performance type of JDBC driver for accessing local databases, as the Java code accesses the database file directly. In order to obtain a connection via DriverManager you have to use following URL: Chapter 2. Obtaining a connection 14 jdbc:firebirdsql:embedded: jdbc:firebirdsql:embedded:host[/port]: When host and, optionally, port is specified, embedded server acts as client library (i.e. you get the same Type 2 behavior as you would get with using “native”). Installation of the JNI proxy is same as described in the “NATIVE and LOCAL types” chapter. Limitations The Firebird embedded server for Linux is not thread safe. Jaybird provides the needed synchronization in Java code, similar to the one described for the Type 2 JDBC driver. This implies the same restrictions on the classloader that will load the Jaybird classes. The Firebird embedded server for Windows opens databases in exclusive mode. This means that this particular database is accessible only to one Java virtual machine. There is no exclusive mode on the POSIX platform. When the same database file is accessed by multiple JVM instances, database will be corrupted! Connection Pooling Each time a connection is opened via DriverManager, a new physical connection to server is opened. It is closed when the connection is closed. In order to avoid the overhead of creating connections, the DataSource implementation can maintain a cache of open physical connections that can be reused between user sessions. The javax.sql.ConnectionPoolDataSource implementation FBConnectionPoolDataSource is an implementation of the javax.sql.ConnectionPoolDataSource interface, which is used by an application to obtain PooledConnection objects. A PooledConnection instance represents a physical connection to a database and is a source of logical connection. Closing a logical connection returns the physical connection back into the pool. Additionally, the logical connection provides caching of prepared statements, which improves application performance even more than connection pooling. Usually the connection pool is specified in web or application server configuration. However, you can instantiate it also inside the application. Chapter 2. Obtaining a connection 15 package hello; import java.sql.*; import javax.sql.*; import org.firebirdsql.pool.*; public class HelloConnectionPool { public static void main(String[] args) throws Exception { org.firebirdsql.pool.FBConnectionPoolDataSource pool = new org.firebirdsql.pool.FBConnectionPoolDataSource(); pool.setMaxPoolSize(5); pool.setMinPoolSize(2); pool.setMaxStatements(10); pool.setMaxIdleTime(30 * 60 * 60); pool.setDatabase("localhost/3050:C:/db/employee.fdb"); pool.setUser("SYSDBA"); pool.setPassword("masterkey"); // obtain a physical connection to the database PooledConnection pooledCon = pool.getPooledConnection(); // obtain a wrapped connection Connection connection = pooledCon.getConnection(); try { // do something here... } finally { // release the connection back to pool connection.close(); } } } Illustration 2.8.: Example of instantiating and using the javax.sql.ConnectionPoolDataSource implementation The code on Illustration 2.8 we perform the following steps: 1. Create a connection pool object. In this example we create instance implementing the javax.sql.ConnectionPoolDataSource interface. 2. Now we specify the pooling properties: maximum of 5 physical connections, with minimum of 2, and each connection will maintain a cache of 10 prepared statements of the same type (i.e. with the same SQL). Connections in the pool that are idle for more than half an hour (30 * 60 * 60 seconds) are closed automatically. 3. After specifying the pooling properties we set the database connection properties. In our case that is only database path, user name and password, but also any other supported property can be set here. 4. Having configured the data source, we obtain the physical connection to the database. Our data source implementation will check the internal connection pool and will open a new physical connection to the database if the pool is empty. An instance of javax.sql.PooledConnection represents a physical connection to the database. Calling the PooledConnection.close() method Chapter 2. Obtaining a connection 16 6 5 4 3 2 1 will close the physical connection to the database and will remove this connection from the pool. 5. Now we obtain regular JDBC connection to the database and perform the needed work. 6. At the end of processing we close the JDBC connection, but note that we do not close the physical connection, but simply forget the reference to it. Please pay especial attention to the steps 4, 5 and 6. They show the typical approach of using the JDBC connections in case of connection pooling. The step 4 is to some extent optional – if we use javax.sql.DataSource connection factory, it provides already wrapped JDBC connections doing the step 4 implicitly. However it must be a rule for an application to use the try/finally block to release the connection. In the XXX chapter we will discuss the transaction boundaries and how they can influence that connection handling, but for the code running in the J2EE environment the try/finally guarantees that connections are never leaked in the application code, the container will take care for the rest. List of all pool-related properties can be found in “Pool Properties” and “Runtime Pool Properties”. Using FBConnectionPoolDataSource with JNDI Connection pooling is tightly coupled with the Java Naming and Directory Interface, which provides a network-transparent hierarchical mapping of the symbolic references to objects. As it was showed in Illustration 2.6, pooled connections are obtained from JNDI using a symbolic reference, a JNDI name. When an application binds an object into JNDI, typically following happens: • If object implements java.io.Serializable interface, object is directly bound to the specified name. If application accesses the JNDI from the local JVM, a reference to the object bound in JNDI is returned. If application accesses the JNDI from remote JVM, a serialized copy of an object is sent over the wire to the remote node, where it is deserialized and returned to the application • If object implements javax.naming.Referencable interface, JNDI implementation binds the so-called reference instead of an object. Reference contains all necessary information to reconstruct the object regardless of the JVM in which this operation happens. This is performed with the help of so- called object factories. Object factory knows how to convert instance of javax.naming.Reference into an appropriate object. • If object implements none of the above mentioned interfaces, the behavior is undefined. Usually JNDI provider allows to access bind objects in local JVM, but when access happens in remote JVM, an exception is thrown. FBConnectionPoolDataSource implements both java.io.Serializable and javax.naming.Referencable interfaces. Illustration 2.9 shows how to create Chapter 2. Obtaining a connection 17 and bind the JNDI reference for a FBConnectionPoolDataSource class: package hello; import javax.naming.*; import org.firebirdsql.pool.*; public class HelloBindJndi { public static void main(String[] args) throws Exception { Reference ref = new Reference( "org.firebirdsql.pool.FBConnectionPoolDataSource"); ref.add(new StringRefAddr("maxPoolSize", "5")); ref.add(new StringRefAddr("minPoolSize", "2")); ref.add(new StringRefAddr("maxStatements", "10")); ref.add(new StringRefAddr("maxIdleTime", "108000")); ref.add(new StringRefAddr("database", "localhost/3050:C:/db/employee.fdb)); ref.add(new StringRefAddr("user", "SYSDBA")); ref.add(new StringRefAddr("password", "masterkey")); Context ctx = new InitialContext(); ctx.bind("jdbc/test", ref); } } Illustration 2.9.: Example of initializing and binding FBConnectionPoolDataSource reference. 1. Create reference instance for the FBConnectionPoolDataSource. According to the JNDI specification we could specify another type here, however the identifier specified here is used later by the object factory to check whether it is responsible for materializing the specified reference. Our object factory accepts only references with the ID equal to "org.firebirdsql.pool.FBConnectionPoolDataSource". 2. Fill the pooling properties as reference addresses. 3. Fill the database connection properties. Steps 2 and 3 look quite strange from the programming point of view, especially compared to the Illustration 2.8. However, this approach is very elegant if we consider reading the configuration from the file. In this case we no longer have to use Java reflection to set needed properties – object factory does it for us. 4. Create JNDI initial context and bind the reference to the specified name. Chapter 2. Obtaining a connection 18 3 4 1 2 Illustration 2.10 Shows how to access the FBConnectionDataSource bound in previous example: package hello; import java.util.*; import javax.naming.*; import org.firebirdsql.pool.*; public class HelloLookupJndiFactory { public static void main(String[] args) throws Exception { Hashtable props = new Hashtable(); props.put( "java.naming.factory.initial", "com.sun.jndi.fscontext.RefFSContextFactory"); props.put( "java.naming.factory.object", "org.firebirdsql.pool.FBConnectionPoolDataSource"); Context ctx = new InitialContext(props); FBConnectionPoolDataSource pool = (FBConnectionPoolDataSource)ctx.lookup("jdbc/test"); } } Illustration 2.10.: Example of accessing the FBConnectionPoolDataSource via reference. 1. Create an environment for the JNDI initial context. 2. Specify the “ java.naming.factory.initial” property. Our example uses Sun file system JNDI provider. In J2EE environment this property should match the one used by J2EE container. Additionally one has to specify the “java.naming.provider.url” when accessing remote JVM. 3. Specify the “ java.naming.factory.object” property. In our case it contains only one object factory – our pool class itself. In J2EE environment one has to configure the environment correctly. 4. Create JNDI context and perform the JNDI lookup. The javax.sql.DataSource implementation The example before showed how to work with the Jaybird 2.1 connection pool. However, the javax.sql.ConnectionPoolDataSource is usually not accessible to the application code, as it provides the ability to manipulate physical connections. In a J2EE environment application accesses the instance of javax.sql.DataSource interfaces instead. This is usually done by wrapping the connection pool by a simple implementation of the latter interface. Jaybird 2.1 provides such implementation in org.firebirdsql.pool.SimpleDataSource class that takes javax.sql.ConnectionPoolDataSource as parameter in constructor. Additionally Jaybird 2.1 provides a class that can be used in the same was as the FBConnectionPoolDataSource. Illustration 2.11 shows how to instantiate FBWrappingDataSource implementation in a client application. Chapter 2. Obtaining a connection 19 4 3 2 1 package hello; import java.sql.*; import javax.sql.*; import org.firebirdsql.pool.*; public class HelloConnectionPool { public static void main(String[] args) throws Exception { org.firebirdsql.pool.FBWrappingDataSource pool = new org.firebirdsql.pool.FBWrappingDataSource(); pool.setMaxPoolSize(5); pool.setMinPoolSize(2); pool.setMaxStatements(10); pool.setMaxIdleTime(30 * 60 * 60); pool.setDatabase("localhost/3050:C:/db/employee.gdb"); pool.setUser("SYSDBA"); pool.setPassword("masterkey"); // no step 4 as in previous example // obtain a wrapped connection Connection connection = pool.getConnection(); try { // do something here... } finally { // release the connection back to pool connection.close(); } } } Illustration 2.11.: Example of instantiating and using the javax.sql.DataSource implementation This class is called “wrapping” because it wraps the connection pool and delegates all calls to the underlying implementation. As you can see, the code is very similar to the Illustration 2.8, only few places are different. Please note, that there is no longer Step 4, the wrapper does this automatically in getConnection() method used in Step 5. The javax.sql.XADataSource implementation JDBC 2.0 specification introduced the javax.sql.XADataSource interface that should be used to access connections that can participate in distributed transactions with JTA-compatible transaction coordinator. This gives applications possibility to use two-phase commit to synchronize multiple resource managers. Jaybird 2.1 does not have separate class, but FBConnectionPoolDataSource also implements the javax.sql.XADataSource interface. For information how to instantiate this class please see Illustration 2.8. Applications usually do not need to access the javax.sql.XADataSource directly, this is task for a J2EE container. Chapter XXX contains more detailed description of distributed transactions and contains code to access and manipulate connections that participate in distributed transactions. Chapter 2. Obtaining a connection 20 6 5 4 3 2 1 Chapter 2. Obtaining a connection 21 Chapter 3. 3. Handling exceptions An exception handling is probably the most important aspect that directly affects the stability of the application. Correct handling of the error cases guarantees correct functioning of the client code as well as the database server. Additionally, all methods of the interfaces defined in the JDBC specification throw instances of java.sql.SQLException to notify about all error conditions that happen during request processing. The SQLException is checked exception, which forces Java programmer to either handle it with the try/catch clause or redeclare it in the method signature. Working with exceptions The exception handling becomes even more important if we consider that this topic is either ignored or presented in incorrect form in the most JDBC tutorials. The official JDBC tutorial from Sun Microsystems briefly mentions that exceptions should be handled by using try/catch blocks only at the end of the course, but neither reasons of doing this nor the best practices are presented. There are few reasons to think about exception handling in your applications before you start coding. First of all, it is very hard to change the exception handling pattern in the existing code. The changes will affect all layers above the place where the changes in exception handling are made and the new application must be thoroughly tested after the change. Another reason was already mentioned on the beginning of this chapter – instances of java.sql.SQLException is the only way for the RDBMS server to notify about the error condition that happened during request processing. By checking the error code which is sent with the exception application can try to recover from the error. And the last but not the least issue is the resource management. When exception happens in the method, the execution flow of Java code differs from the normal one and only correctly coded application will ensure that all allocated resources will be released. The resources in our case are JDBC connections, statement, prepared statement and callable statement objects, result sets, etc. All these objects not only take memory in the Java Virtual Machine in which application runs, but also consume memory on the server, which in worst cases can lead to Denial-of-Service attack. A good exception handling strategy requires you do distinguish three kinds of error conditions: • errors that database access layer can detect and correctly handle; for example, the application might decide to re-execute the business transaction if database server returned a deadlock error; • errors that database access layer can detect, but is unable to handle; usually those are all database errors that do not have special handling routines; • errors that database access layer cannot detect without additional code unrelated to the functionality of this layer; basically, all runtime exceptions fall into this category. The handling strategy then consists of • processing the selected error codes for cases described above; • converting the generic SQLException into generic business error in the application (this can be throwing some generic exception defined in the application, but can also be an entry in the application event log and short message that asks to retry the operation later); • some emergency tactics, since the error that happened (e.g. NullPointerException or OutOfMemoryError) was not considered while the application was created, thus possibly leaving it in an unknown state; further operating should be considered dangerous and the corresponding execution branch has to be halted. The problem can be solved if resource allocation and deallocation happens in the same code block and is protected with try/finally block and the code to recover from error conditions should use try/catch blocks. Example of such error and resource handling code is presented on Illustration 3.1. Chapter 3. Handling exceptions 24 PreparedStatement updateSales = null; String updateString = "update COFFEES " + "set SALES = ? where COF_NAME like ?"; updateSales = con.prepareStatement(updateString); try { int [] salesForWeek = {175, 150, 60, 155, 90}; String [] coffees = {"Colombian", "French_Roast", "Espresso", "Colombian_Decaf", "French_Roast_Decaf"}; int len = coffees.length; for(int i = 0; i < len; i++) { updateSales.setInt(1, salesForWeek[i]); updateSales.setString(2, coffees[i]); try { updateSales.executeUpdate(); } catch(SQLException ex) { if (ex.getErrorCode() == ...) // do something else throw new BusinessDBException(ex); } } } finally { updateSales.close(); } Illustration 3.1.: Typical resource allocation and error handling patterns The nested try/catch block shows you an example of handling a deadlock error if it happens (first scenario according to our classification), otherwise the exception is converted and passed to the upper layers (second scenario). As you see, there is no special treatment to the third scenario. A possible bug in the JDBC driver could have generated runtime exception in the PreparedStatement.executeUpdate() method, which would lead to the statement handle leakage if no try/finally block is used to do the resource cleanup. As a rule of thumb, the “try” keyword should go right after the resource was allocated and the “finally” keyword should be placed right before the resource is freed. Such coding practice might look weird, because on the first sight the whole purpose of using the PreparedStatement is neglected – statement is prepared, used only once and then deallocated. However, when this practice is combined with the connection and statement pooling, it brings enormous advantage to the application code. The code becomes much more manageable – resource allocations and deallocations happen in the same method and software developer must not remember the places where the same prepared statement might be used – statement pool will either reuse the statement or it will prepare a new one, if it detects that all pooled prepared statements are currently in use. As a side effect, Chapter 3. Handling exceptions 25 application will always use the minimum number of statements handles, which in turn reduces the used resources on the server side. Warnings Some errors returned by the Firebird server are treated as warnings. They are converted into instances of java.sql.SQLWarning class in the JDBC layer. These exceptions are not thrown from the driver methods, but added to a connection instance. Currently no warning is added to Statement or ResultSet objects. Each next warning is appended to the tail of the warning chain. In order to read the warning chain, use the code presented on Illustration 3.2. import java.sql.*; .... SQLWarning warning = connection.getWarnings(); while (warning != null) { .... // do something with the warning warning = warning.getNextWarning(); } Illustration 3.2.: Example how to work with the exceptions In order to clear existing warning, call Connection.clearWarnings() method. java.sql.SQLException in Jaybird An SQLException is a special exception that is thrown by the JDBC connectivity component in case of an error. Each instance of this exception is required to carry the vendor error code (if applicable) and a SQL state according to the X/Open SQLstate or SQL 2003 specifications. When multiple SQL errors happened, they are joined into a chain. Usually the most recent exception is thrown to the application, the exceptions that happened before can be obtained via SQLException.getNextException() method. Unfortunately the JDBC specification does not provide a usable exception hierarchy that would allow application to react on the error situations using regular exception handling rather than checking the error code. Only two subclasses are defined in JDBC 3.0 specification: • java.sql.DataTruncation exception is thrown when data truncation error happens. • java.sql.BatchUpdateException exception is thrown when batch of the statement did not execute successfully and contains the result of batch execution. Upcoming JDBC 4.0 will address this issue by introducing better hierarchy of the exceptions and it will be supported by the next versions of Jaybird. Each of three layers in Jaybird use exceptions most appropriate to the specific layer. • org.firebirdsql.gds.GDSException is an exception that directly corresponding to the error returned by the database engine. Instances of this Chapter 3. Handling exceptions 26 class are thrown by the GDS implementations. Upper layers either convert these exceptions into the ones appropriate to that layer or catch them if driver can handle the error condition. • Subclasses of javax.resource.ResourceException are thrown by the JCA layer when an error happens in the JCA-related code. Upper layer converts this exception into a subclass of java.sql.SQLException. If the ResourceException was caused by the GDSException, latter is extracted during conversion preserving the error code. If ResourceException was caused by an error condition not related to an error returned by the database engine, error code of the SQLException remains 0. • Subclasses of javax.transaction.XAException are thrown when an XA protocol error happens in JCA layer. Similar to the previous case, XAException can wrap the GDSException, which are extracted during exception conversion to preserve the error code. • Subclasses of java.sql.SQLException are thrown by the JDBC layer. Driver has also few subclasses that might be interesting to the application: • org.firebirdsql.jdbc.FBDriverConsistencyCheckException – this exception is thrown when driver detects an internal inconsistent state. SQL state is SQL_STATE_GENERAL_ERROR. • org.firebirdsql.jdbc.FBDriverNotCapableException – this exception is thrown when an unsupported method is called. SQL state is SQL_STATE_DRIVER_NOT_CAPABLE. • org.firebirdsql.jdbc.FBSQLParseException – this exception is thrown when incorrect escaped syntax is detected. SQL state is SQL_STATE_INVALID_ESCAPE_SEQ. • org.firebirdsql.jdbc.field.TypeConversionException – this exception is thrown when the driver is asked to perform a type conversion that is not defined in the JDBC specification. For a table of allowed type conversions see Ошибка: источник перекрестной ссылки не найден. SQL states Jaybird supports the SQL states from the X/Open standard, however only few states nicely map into the Firebird error codes. Below is the table containing the reported SQL states. Constant name Constant value SQL_STATE_INVALID_CON_ATTR "01S00" SQL_STATE_NO_ROW_AVAIL "01S06" SQL_STATE_GENERAL_ERROR "HY00" SQL_STATE_DRIVER_NOT_CAPABLE "HYC00" SQL_STATE_INVALID_COLUMN "HY02" SQL_STATE_INVALID_PARAM_TYPE "HY105" SQL_STATE_INVALID_ARG_VALUE "HY009" Chapter 3. Handling exceptions 27 Constant name Constant value SQL_STATE_WRONG_PARAM_NUM "07001" SQL_STATE_NO_RESULT_SET "07005" SQL_STATE_INVALID_CONVERSION "07006" SQL_STATE_CONNECTION_CLOSED "08003" SQL_STATE_CONNECTION_FAILURE_IN_TX "08007" SQL_STATE_COMM_LINK_FAILURE "08S01" SQL_STATE_INVALID_ESCAPE_SEQ "22025" Application can use the SQL state codes in the error handling routines which should handle errors that are returned from different databases. But since there is little agreement between RDBMS vendors, this method can be used only for very coarse error distinction. Useful Firebird error codes Contrary to the SQL states, the Firebird native error codes are extremely useful to determine the type of an error that happened. Here you can find a short list of error codes, symbolic names of a corresponding constant in a org.firebirdsql.gds.ISCConstants class, the error message and short explanation of an error. DDL Errors Happen during execution of the DDL requests and two primary error codes are used in Firebird while executing the DDL operations. There are few other rare cases not mentioned here, but the corresponding error messages contain enough information to understand the reason of an error. 335544351L isc_no_meta_update "unsuccessful metadata update" Error happens when the requested DDL operation cannot be completed, for example application tries to define a primary key that will exceed the maximum allowed key size. 335544510L isc_lock_timeout In combination with isc_obj_in_use (335544453L) means that the DDL command tries to modify an object that is used in some other place, usually in another transaction. The complete error message will contain the name of the locked object. 335544569L isc_dsql_error If the third error code is either isc_dsql_datatype_err or isc_dsql_command_err, then additional error codes and arguments specify the reason why the operation had failed. Lock Errors Lock errors are reported by Firebird primarily when application tries to modify a record which is already modified by a concurrent transaction. Depending on the transaction parameters such error can be reported either right after detecting it or after waiting some defined timeout hoping that concurrent transaction will either Chapter 3. Handling exceptions 28 commit or rollback and eventually release the resource. More information on transaction locking modes can be found in Chapter 6, Using transactions. 335544345L isc_lock_conflict "lock conflict on no wait transaction" This error happens when a “no wait” transaction needs to acquire a lock but finds another concurrent transaction holding a lock. Instead of waiting the predefined timeout hoping that concurrent transaction will either commit or rollback, an error is returned to notify an application about the situation. 335544510L isc_lock_timeout "lock time-out on wait transaction" Similar to the isc_deadlock, but is returned when the lock timeout that was specified for the current transaction expired while waiting for a lock. Another source of this error are DDL operations that try to obtain a lock on a database object that is currently used in some other place. 335544336L isc_deadlock "deadlock" Two transactions experience a deadlock when each of them has a lock on a resource on which another one is trying to obtain a lock. Referential Integrity Errors Referential integrity constrains ensure that database remains in a consistent state after the DML operation and/or whole transaction is completed. Three primary error codes are returned when the defined constraints are violated. The error messages are self-explanatory. 335544665L isc_unique_key_viola tion violation of PRIMARY or UNIQUE KEY constraint "{0}" on table "{1}" 335544558L isc_check_constraint Operation violates CHECK constraint {0} on view or table {1} 335544466L isc_foreign_key violation of FOREIGN KEY constraint "{0}" on table "{1}" DSQL Errors This group contains secondary codes for the primary error code isc_dsql_error (335544569L), that has a message "Dynamic SQL Error". 335544573L isc_dsql_datatype_err "Data type unknown" Usually this error is reported during DDL operation when the specified data type is either unknown or cannot be used in the specified statement. However it also can happen in DML operation, e.g. when an ORDER BY clause contains unknown Chapter 3. Handling exceptions 29 collation. isc_dsql_command_err "Invalid command" Error happens either during parsing the specified SQL request or by handling the DDL command. Other Errors This table contains other errors that might be interesting to the application developer, however they do not fall into any of the previous categories. 335544321L isc_arith_except "arithmetic exception, numeric overflow, or string truncation" Happens at runtime when an arithmetic exception happens, like division by zero or the numeric overflow (e.g. number does not fit the 64 bits limit). Another source of these errors are all string operations, like string concatenation producing too long string, impossibility to transliterate characters between character sets, etc. Future versions of Firebird will provide a secondary code to distinguish the exact reason of an error. 335544348L isc_no_cur_rec "no current record for fetch operation" Happens when application asks Firebird to fetch a record, but no record is available for fetching. Java applications should never get this error, since checks in the JDBC driver prevent application from executing fetch operation on the server side. 335544374L isc_stream_eof "attempt to fetch past the last record in a record stream" Application tries to execute fetch operation after all records had been already fetched. Similar to the previous error, Java application should not get this error due to the checks that happen before issuing the fetch request to the server. 335544517L isc_except "exception {0}" An custom exception has been raised on the server. Java application can examine the underlying GDSException to extract the exception message. 335544721L isc_network_error Unable to complete network request to host "{0}" This error is thrown when Java application cannot establish connection to the database server due to a network issues, e.g. host name is specified incorrectly, Firebird had not been started on the remote host, firewall configuration prevents client Chapter 3. Handling exceptions 30 from establishing the connection, etc. Chapter 3. Handling exceptions 31 Chapter 4. 4. Executing statements After obtaining a connection, the next thing to do is to execute the SQL statement. JDBC specification distinguishes three kinds of statements – regular statements that are execute fixed SQL requests, prepared statements used to executed SQL code with parameters in it and callable statements that are used to execute stored procedures. The java.sql.Statement interface The java.sql.Statement interface is the simplest interface to execute SQL statements. It distinguishes only three types of them: • statements that return results, or, in other words, queries; • statements that change the state of the database but return no results; • INSERT statements that return the values of the columns which were generated by the database engine while inserting the record. Let's check one of the typical usages showed on Illustration 4.1. In general the usage pattern of the statement consists of three steps Statement stmt = connection.createStatement(); try { ResultSet rs = stmt.executeQuery( "SELECT firstName, lastName FROM users" + " WHERE userId = 5"); rs.next(); String firstName = rs.getString(1); String lastName = rs.getString(2); } finally { stmt.close(); } Illustration 4.1: Typical way to execute query to get information about the user. : 1. Create a Statement object by calling the createStatement() method of the Connection object. 2. Use the Statement object by calling its methods, in our case we execute simple query SELECT firstName, lastName FROM users WHERE userId = 5. Processing of the query result will be discussed in details in the next chapter. 3. Close the statement to release all allocated resources. In our example this is done in the finally section of the try/finally block. The fact that connection object is a factory for the statement objects puts a constraint on the object lifetime – statements are bound to the connection; when the connection is closed, all statements that were created by that connection become invalid and the resources allocated by them are released. However, despite that fact that the resources are finally released, it is strongly recommended to use the try/finally block, to guarantee that resources are released as soon as possible because of the reasons that will be discussed later. Statement can be executed using the following methods: • Statement.executeQuery(String) – executes a SELECT statement and returns a result set. If the specified statement is not a SELECT statement, an SQLException is thrown after the statement execution. • Statement.executeUpdate(String) – executes INSERT, UPDATE, DELETE or DDL1 statements and returns the number of updated rows. If the specified statement is a query, an SQLException is thrown. • Statement.execute(String) – executes a statement and returns true when the statement returned a result set, otherwise an update was executed and false is returned. You can use Statement.getResultSet() method to get the result of the executed query or you can use Statement.getUpdateCount() when you have executed update statement. Statement is closed by calling the Statement.close() method. After this the statement object is invalid and cannot be used anymore. 1 DDL – Data Definition Language. This term is used to group all statements that are used to manipulate database schema, i.e. creation of tables, indices, views, etc. Chapter 4. Executing statements 34 3 2 1 It is also allowed to use the same object to execute different types of queries one after other. The Illustration 4.2 contains a short example in which application first performs a select to find the ID of the user 'Joe Doe', and if the record is found, it enables his account. Statement stmt = connection.createStatement(); try { ResultSet rs = stmt.executeQuery( "SELECT userId FROM users " + "WHERE lastName = 'Doe' AND firstName = 'Joe'"); if (rs.next()) { int userId = rs.getInt(1); int rowsUpdated = stmt.executeUpdate( "UPDATE accounts SET accountEnabled = 1 " + "WHERE userId = " + userId); if (rowsUpdated == 0) rowsUpdated = stmt.executeUpdate( "INSERT INTO accounts (userId, enabled) " + "VALUES (" + userId + ", 1)"); if (rowsUpdated != 1) throw new SomeException( "User was not updated correctly."); } } finally { stmt.close(); } Illustration 4.2: Using the same statement object multiple times to enable user account. The way the code is constructed is quite dangerous because of the result set lifetime constraints that are put by the JDBC specification, please read the chapter where result sets are discussed for more details. However, here it is done intentionally to emphasize that the single object is used to execute SELECT and UPDATE/INSERT statements. It also shows how to check whether the executed statement modified expected number of rows – application first tries to update the account and only if no rows were updated, it inserts new record into the accounts table. When application needs to execute DDL statements, it is recommended to use the Statement.execute(String) method, as in this case amount of modified records makes little sense. Illustration 4.3 shows an example of creating database Chapter 4. Executing statements 35 tables using the above mentioned method. Statement stmt = connection.createStatement(); try { DatabaseMetaData metaData = connection.getMetaData(); ResultSet tables = metaData.getTables( null, null, "customer", new String[]{"TABLE"}); if (!tables.next()) stmt.execute("CREATE TABLE customer(" + "customerId INTEGER NOT NULL PRIMARY KEY, " + "firstName VARCHAR(20) NOT NULL, " + "lastName VARCHAR(40) NOT NULL)"); } finally { stmt.close(); } Illustration 4.3: Example of creating database tables. First, application checks the existence of the table in the database by calling the DatabaseMetaData.getTables(String, String, String, String[]) method that returns a result set describing the database tables matching the specified search pattern. First two parameters of this method, the database catalog and schema names, are set to null as Firebird supports neither catalogs nor schemas. Third parameter is the table name search pattern, in our case we search for the table "customer". Last parameter is the list of table types to check, in our case we check for the "TABLE" type. Other tables types are "SYSTEM TABLE" and "VIEW". After that application checks if the result set is empty by calling the ResultSet.next() method. If no "customer" table was found, application creates new table with three columns. As it was already mentioned, the Statement.execute(String) method can also be used to execute statements of the unknown type. Statement stmt = connection.createStatement(); try { boolean hasResultSet = stmt.execute(sql); if (hasResultSet) { ResultSet rs = stmt.getResultSet(); ... } else { int updateCount = stmt.getUpdateCount(); ... } } finally { stmt.close(); } It is worth mentioning, that according to the JDBC specification getResultSet() and getUpdateCount() methods can be only called once per result, and in case of using Firebird, that means once per executed statement, since Firebird does not support multiple results from a single statement. Calling the methods the second time will cause an exception. Chapter 4. Executing statements 36 Statement behind the scenes The example on Illustration 4.2 requires us to discuss the statement object dynamics, its life cycle and how it affects other subsystems in details. Statement dynamics When the Java application executes statement, a lot more operations happen behind the scenes: 1. A new statement object is allocated on the server. Firebird returns to the client a 32-bit identifier of the allocated object, a statement handle, that must be used in next operations. 2. An SQL statement is compiled into an executable form and is associated with the specified statement handle. 3. Jaybird asks server to describe the statement and Firebird returns the information about the statement type and possible statement input parameters (we will discuss this with prepared statements) and output parameters, namely a result set columns. 4. If no parameters are required for the statement, Jaybird tells Firebird to execute statement passing the statement handle into corresponding method. After this Jaybird has to make a decision depending on the operation that was called. • If Statement.execute() method was used, Jaybird only checks the statement type to decide whether it should return true, telling the application that there is a result set for this operation, or false, if statement did not return any result set. • If Statement.executeUpdate() method was called Jaybird asks Firebird to give the information about the number of affected rows. This method can be called only if the statement type tells that no query can be returned by the statement. When it is called for queries, an exception is thrown despite the fact that the statement was successfully executed on the server. • If Statement.executeQuery() method was called and statement type tells that result set can be returned, Jaybird constructs a ResultSet object and returns it to the application. No additional checks, like whether result set contains rows, are performed, as it is the responsibility of the ResultSet object. If this method is used for the statements that do not return result set, an exception is thrown despite the fact that the statement was successfully executed on the server. When the application does not need to know how many rows were modified, it should use the execute() method instead of executeUpdate() one. This saves additional call to the server to get the number of modified rows and significantly increases the performance in the situations where network latency is comparable with the statement execution times. The execute() method is also the only method that can be used when the application does not know what kind of statement is being executed (for example, an application that allows the user to enter SQL statements to execute). After using the statement object application should close it. Two different possibilities exist – to close the result set object associated with the statement handle and to close the object completely. Chapter 4. Executing statements 37 If, for example, we want to reuse the statement object for another query, it is not necessary to completely release the allocated structures. Jaybird is required only to compile a new statement before using it, in other words we can skip the step 1. This saves us one round-trip to the server over the network, which might improve the application performance. If we close the statement completely, the allocated statement handle is no longer usable. Jaybird could allocate a new statement handle, however the implementation does not allow to use the Statement object after close() method has been called to comply with the JDBC specification. Statement lifetime and DDL The step 2 in the previous chapter is probably the most important, and usually, most expensive part of the statement execution life cycle. As it was already told, when Firebird server receives the “prepare statement” call, it parses the SQL statement and converts it into the executable form – BLR representation. BLR, a Binary Language Representation, contains low-level commands to traverse the database tables, conditions that are used to filter records, defines the order in which records are accessed, indices that are used to improve the performance, etc. When a BLR is prepared, it holds the references to all database object definitions that are used during that statement execution. This mechanism preserves the database schema consistency, it saves the statement objects from the “surprises” like accessing the database table that is been accessed by some application. However, holding a reference on the database objects has one very unpleasant effect – it is not possible to upgrade the database schema, if there are active connections to the database with open statements referencing the objects being upgraded. In other words, if two application are running and one is trying to modify the table, view, procedure or trigger definition while another one is accessing those objects, the first application will receive an error 335544453 “object is in use”. Therefore it is strongly recommended to close the statement as soon as it is no longer needed. This invalidates the BLR and release all references to the database objects, making them available for the modification. Special care should be taken when the statement pooling is used (check the details on page 41). In this case statements are not released even if the close() method is called. The only possibility to close the pooled statements is to close the pooled connections. Please check the corresponding chapter for more information. The java.sql.PreparedStatement interface As we have seen, Jaybird already performs internal optimization when it comes to multiple statement execution – it can reuse the allocated statement handle in subsequent calls. However this improvement is very small and sometimes can even be neglected when compared to the time needed to compile the SQL statement into the BLR form. The PreparedStatement interface addresses such inefficiency. An object that implements this interface represents a precompiled statement that can be executed multiple times. If we use the execution flow described in the “Statement Chapter 4. Executing statements 38 dynamics” chapter, it allows to go directly to the step 4 for the subsequent executions. However, executing the same statement with the same values makes little sense, unless we want to fill the table with the same data, which usually is not the case. Therefore, JDBC provides support for the parametrized statements – SQL statements where literals are replaced with the question marks. Application is required to fill the parameters before executing the statement. Our first example in this chapter can be rewritten the way it is showed on Illustration 4.4. At the first glance the code became more complicated without any visible advantage. PreparedStatement stmt1 = connection.prepareStatement( "SELECT userId FROM users WHERE " + "lastName = ? AND firstName = ?"); try { stmt1.setString(1, "Doe"); stmt1.setString(2, "Joe"); ResultSet rs = stmt1.executeQuery(); if (rs.next()) { int userId = rs.getInt(1); PreparedStatement stmt2 = connection.prepareStatement( "UPDATE accounts SET accountEnabled = 1 " + "WHERE userId = ?" ); int rowsUpdated; try { stmt2.setInt(1, userId); rowsUpdated = stmt2.executeUpdate(); if (rowsUpdated == 0) { PreparedStatement stmt3 = connection.prepareStatement( "INSERT INTO accounts " + "(userId, enabled) VALUES (?, 1)"); try { stmt3.setInt(1, userId); rowsUpdated = stmt3.executeUpdate(); } finally { stmt3.close(); } if (rowsUpdated != 1) throw new SomeException( "User was not updated correctly."); } finally { stmt2.close(); } } } finally { stmt1.close(); } Illustration 4.4: Example for user account update rewritten using prepared statements. Chapter 4. Executing statements 39 • First, instead of using the one statement object we have to use three – one per statement. • Second, before executing the statement we have to set parameters first. As it is showed on the example, parameters are referenced by their position. The PreparedStatement interface provides setter methods for all primitive types in Java as well as for some widely used SQL data types (BLOBs, CLOBs, etc.). The NULL value is set by calling the PreparedStatement.setNull(int) method. • Third, we are forced now to use three nested try/finally blocks, which makes code less readable. So, where's the advantage? First of all, we can redesign our application to prepare those statements before calling that code (for example in a constructor) and close them when application is shut down. In this case the code is even more compact (see Illustration 4.5). Unfortunately the application is now responsible for prepared statement management. When connection is closed, the prepared statement object will be invalidated, but the application will not be notified about this fact. Also, if the application uses similar statements in different parts of the application, the refactoring might affect many classes, possibly destabilizing the Chapter 4. Executing statements 40 code. So, the refactoring on the Illustration 4.5 is not something we want to do. // prepared statement management PreparedStatement queryStmt = connection.prepareStatement(queryStr); PreparedStatement updateStmt = connection.prepareStatement(updateStr); PreparedStatement insertStmt = connection.prepareStatement(insertStr); ...................... // query management queryStmt.clearParameters(); queryStmt.setString(1, "Doe"); queryStmt.setString(2, "Joe"); ResultSet rs = queryStmt.executeQuery(); if (rs.next()) { int userId = rs.getInt(1); updateStmt.clearParameters(); updateStmt.setInt(1, userId); int rowsUpdated = updateStmt.executeUpdate(); if (rowsUpdated == 0) { insertStmt.clearParameters(); insertStmt.setInt(1, userId); rowsUpdated = insertStmt.executeUpdate(); if (rowsUpdated != 1) throw new SomeException( "User was not updated correctly."); } ...................... // prepared statement cleanup insertStmt.close(); updateStmt.close(); queryStmt.close(); Illustration 4.5: Rewritten example to let application manage prepared statements The answer to the advantage question is hidden in the prepareStatement(String) call. Since the same statement can be used for different parameter values, connection object has a possibility to perform prepared statement caching. JDBC driver can ignore the request to close the prepared statement, save it internally and reuse it each time application asks to prepare an SQL statement that is known to the connection. Prepared statement pooling As it was mentioned before, connection pooling was introduced to reduce the time needed to obtain a connection. Despite its rich features, the operation of obtaining new connection in Firebird is cheap – usually connection pooling in an application that heavily opens and closes connection brings ~5% of performance. However there is one more way to speed the application. Execution of statements in Firebird always happens in three steps: Chapter 4. Executing statements 41 • Compile the SQL statement into an internal BLR representation, save it in an internal structure and assign a statement handle to a compiled statement. • Execute the statement. First application checks whether the compiled statement has any parameters and sets them if needed. After that statement is executed using special call. • Obtain results of the statement execution, for example number of updated rows or a result set. The obvious approach is to save the compiled statement and use it later. JDBC specification already contains a java.sql.PreparedStatement interface exactly for such purposes. Application prepares statement and uses it multiple times. This approach works fine within the context of the same connection. However, when connection pooling is used, application can no longer cache prepared statements, since they are bound to the connection it obtained from the pool. Application must close all prepared statements before giving connection back to pool, but even if it does not do this, connection pool will perform this automatically according to the JDBC specification. All advantages of the prepared statements are undone – when application obtains next connection from the pool, it must re-prepare statements. Solution to the problem is to allow connection pool do statement pooling internally. In this case all code remains compliant with the JDBC specification saving all advantages of prepared statements. Jaybird 2.1 connection pool has maxStatements property that controls the behavior of the prepared statement pooling: • If property is set to 0, no statement pooling is performed. • If property is set to value n>0, connection pool will save maximum n java.sql.PreparedStatement objects for the same SQL statement per connection. Number of pooled PreparedStatement objects corresponding to different SQL statements is not limited. If application needs more prepared statements simultaneously (i.e. it prepares new statement before releasing the one being currently in use), connection pool transparently passes call to the connection object without pooling those statements. In other words, the request to prepare statement is always satisfied immediately, but only n prepared statements will remain pooled, rest will be deallocated when PreparedStatement.close() method is called. Following limitations apply: • Firebird can have approx. 20.000 active statement handles per connection. Additional care should be used when specifying the value of maxStatements property. • Neither java.sql.Statement nor java.sql.CallableStatement objects are pooled. • A compiled statement in Firebird contains references on database objects (tables, views, procedures, etc.) needed to execute that SQL command. When connection pool is used, those references are not released preventing any structure modification of the database objects used in the SQL statement. Chapter 4. Executing statements 42 Therefore, if structure modification is needed, connection pool with enabled statement pooling must be shut down. Prepared statement pooling – advantages and drawbacks Originally statement pooling was introduced to preserve the PreparedStatement advantages when connection pooling is used. Since the lifetime of the statement object is bound to the connection object, prepared statement must be closed before the connection is released to the connection pool. When FBConnectionPoolDataSource or FBWrappingDataSource classes are used to obtain database connections, prepared statement pooling is there for granted. No application modification is needed to enable it, but the performance improvement, depending on the application, might reach up to 50%. Connections that were obtained via DriverManager do not provide statement pooling and application must handle prepared statements itself. The biggest drawback of the statement pooling is the fact that statement handle is not released even when application does not need the statement, which in turn prevents database administrator to upgrade the database schema. To overcome this issue two additional methods were added to the connection pooling classes as well as the possibility to switch the statement pooling off. The restart() method defined in the FBConnectionPoolDataSource and in the FBWrappingDataSource classes. This method closes all open connection residing in the pool. Connections that are currently used in the application are marked as “pending for close” and are deallocated as soon as application returns them to the pool. This algorithm guarantees that eventually all connections will be closed and statements will be deallocated without closing the working applications. The only requirement for successful database schema upgrade is that the application does not “lock” the database objects before the upgrade happens. Unfortunately there is no easy application design guidelines that would guarantee the hot schema upgrade. The restart() method can also be used during database schema upgrade when the Firebird ClassicServer is used. There is an old issue related to the architecture of the ClassicServer – each instance of the database engine that is serving the application connection caches the metadata information. So, even if the application did not have any open statement and the database schema upgrade was successful, open connections will not notice the change. The restart() method softly closes all open connections that are not in use and ensures that new connections obtained from the pool will use new ClassicServer instances with fresh metadata information. The shutdown() method defined in the pool classes that marks the pool as invalid and closes all open connection regardless whether they are currently in use or not. This method can be used if a short-time application down time is acceptable. To continue functioning application must construct a new pool and replace the old one since shutdown() method invalidates the pool object. The java.sql.CallableStatement interface The CallableStatement interface extends PrepatedStatement with methods for executing an retrieving results from stored procedures. It was introduced in JDBC specification in order to unify access to the stored procedures across the Chapter 4. Executing statements 43 database system. The main difference to PreparedStatement is that the procedure call is specified using the portable escaped syntax procedure call ::= {[?=]call } params ::= [, ...] Illustration 4.6: Unified escaped syntax for stored procedure execution. Each stored procedure is allowed to take zero or more input parameters, similar to the PreparedStatement interface. After being executed, procedure can either return data in the output parameters or it can return a result set that can be traversed. Though the interface is generic enough to support also database engines that can return both and have multiple result sets. These features are of no interest to Jaybird users, since Firebird does not support them. The IN and OUT parameters are specified in one statement. The syntax above does not allow to specify the type of the parameter, therefore additional facilities are needed to tell the driver which parameter is will contain output values, the rest are considered to be IN parameters. Firebird stored procedures Firebird stored procedures represent a piece of code written in the PSQL language that allows SQL statement execution at the native speed of the engine and provides capabilities for a limited execution flow control. The PSQL language is not general purpose language therefore its capabilities are limited when it comes to interaction with other systems. Firebird stored procedures can be classified as follow: • Procedures that do not return any results. These are stored procedures that do not contain the RETURNS keyword in their header. • Procedures that return only a single row of results. These are stored procedures that contain the RETURNS keyword in their header, but do not contain the SUSPEND keyword in their procedure body. These procedures can be viewed as a functions that return multiple values. These procedures are executed by using the EXECUTE PROCEDURE statement. • Procedures that return result sets, also called “selectable stored procedures”. These are stored procedures that contain the RETURNS keyword in their header and the SUSPEND keyword in their procedure body, usually within a loop. Selectable procedures are executed using the "SELECT * FROM myProcedure(...)" SQL statement. It is also allowed to use EXECUTE PROCEDURE statement, however this call might produce strange results, since for selectable procedures is is equivalent to executing a SELECT statement, but doing only one fetch after the select. If procedure implementation relies on the fact that all rows that it returns must be fetched, the logic will be broken. Consider the following stored procedure that returns factorial of the specified number. Chapter 4. Executing statements 44 CREATE PROCEDURE factorial( max_value INTEGER ) RETURNS ( factorial INTEGER ) AS DECLARE VARIABLE temp INTEGER; DECLARE VARIABLE counter INTEGER; BEGIN counter = 0; temp = 1; WHILE (counter <= max_value) DO BEGIN IF (counter = 0) THEN temp = 1; ELSE temp = temp * row_num; counter = counter + 1; END END Illustration 4.7.: Source code for the procedure that multiplies two integers. This procedure can be executed using the EXECUTE PROCEDURE call. When it is done in isql, the output looks as follow SQL> EXECUTE PROCEDURE factorial(5); FACTORIAL ============ 120 Illustration 4.8.: Output of the EXECUTE PROCEDURE call in isql. Now let's modify this procedure to return each intermediate result to the client. Chapter 4. Executing statements 45 CREATE PROCEDURE factorial( max_value INTEGER ) RETURNS ( row_num INTEGER, factorial INTEGER ) AS DECLARE VARIABLE temp INTEGER; DECLARE VARIABLE counter INTEGER; BEGIN counter = 0; temp = 1; WHILE (counter <= max_value) DO BEGIN IF (row_num = 0) THEN temp = 1; ELSE temp = temp * row_num; factorial = temp; row_num = counter; counter = counter + 1; SUSPEND; END END Illustration 4.9.: Modified procedure that returns each intermediate result. If you create this procedure using the isql command line tool and then issue the "SELECT * FROM test_selectable(5)" statement, the output will be like this: SQL> SELECT * FROM factorial(5); ROW_NUM FACTORIAL ============ ============ 0 1 1 1 2 2 3 6 4 24 5 120 Illustration 4.10.: Output of the modified procedure. Using the CallableStatement Let's see how the procedures defined above can be accessed from Java. First, we can execute this procedure showed on Illustration 4.7 using the EXECUTE PROCEDURE statement and PreparedStatement, however this approach requires some more code for result set handling. Chapter 4. Executing statements 46 PreparedStatement stmt = connection.prepareStatement( "EXECUTE PROCEDURE factorial(?)"); stmt.setInt(1, 2); ResultSet rs = stmt.executeQuery(); rs.next(); // move cursor to the first row int result = rs.getInt(1); Illustration 4.11.: Example of using the PreparedStatement to call executable procedure. However, standard was of calling stored procedures in JDBC is to use the CallableStatement. Note that the call should be specified using the escaped syntax, but native Firebird EXECUTE PROCEDURE syntax is also supported. CallableStatement stmt = connection.prepareCall( "{call factorial(?,?)}"); stmt.setInt(1, 2); stmt.registerOutParameter(2, Types.INTEGER); stmt.execute(); int result = stmt.getInt(2); Illustration 4.12.: Accessing the executable procedure via CallableStatement. Please note the difference in the number of parameters used in the examples. The first example contained only IN parameter on position 1 and the OUT parameter was returned in the ResultSet on the first position, so it was accessed via index 1. The latter example additionally contains the OUT parameter in the call. We have used the CallableStatement.registerOutParameter method to tell the driver that the second parameter in our call is an OUT parameter of type INTEGER. Parameters that were not marked as OUT are considered by Jaybird as IN parameters. Finally the "EXECUTE PROCEDURE factorial(?)" SQL statement is prepared and executed. After executing the procedure call we get the result from the appropriate getter method. It is worth mentioning that the stored procedure call preparation happens in the CallableStatement.execute method, and not in the prepareCall method of the Connection object. Reason for this potential deviation from the specification is that Firebird does not allow to prepare a procedure without specifying parameters and set them only after the statement is prepared. It seems that this part of the JDBC specification is modeled after the Oracle RDBMS and a workaround for this issue had to be delivered. Another side effect of this issue is, that it is allowed to intermix input and output parameters, for example in the “IN, OUT, IN, OUT, OUT, IN” order. Not that it makes much sense to do this, but it might help in some cases when porting applications from another database server. It is also allowed to use a procedure call parameter both as an input and output parameter. It is recommended to use this only when porting applications from the database servers that allow INOUT parameter types, such as Oracle. The actual stored procedure call using the CallableStatement is equivalent to the call using the prepared statement as it was showed in the first example. There Chapter 4. Executing statements 47 is no measurable performance differences when using the callable statement interface. Also the JDBC specification allows another syntax for the stored procedure calls: CallableStatement stmt = connection.prepareCall( "{?= call factorial(?}"); stmt.registerOutParameter(1, Types.INTEGER); stmt.setInt(2, 2); stmt.execute(); int result = stmt.getInt(1); Illustration 4.13.: Calling stored procedure using different syntax. Note, that input parameters have now indices 2 and 3, and not 1 and 2 as in the previous example. This syntax seems to be more intuitive, as it looks like a function call. It is also possible to use this syntax for stored procedures that return more than one parameter by combining code from the second and the last examples. Firebird stored procedures can also return result sets. This is achieved by using the SUSPEND keyword inside the procedure body. This keyword returns the current values of the output parameters as a single row to the client. The following example is more complex and shows a stored procedure that computes a set of factorial of the numbers up to the specified number of rows. The SELECT SQL statement is the natural way of accessing the selectable procedures in Firebird. You “select” from such procedures using the Statement or PreparedStatement objects. With minor issues it is also possible to access selectable stored procedures through the CallableStatement interface. The escaped call must include all IN and OUT parameters. After the call is prepared, parameters are set the same way. However application must explicitly tell the driver that selectable procedure is used and access to the result set is desired. This is done by calling a Jaybird- specific method as showed in the example below. When this is not done, application has access only to the first row of the result set. The getter methods from the CallableStatement interface will provide you access only to the first row of the result set. In order to get access to the complete result set you have to either call the executeQuery method or the execute method followed by getResultSet method. Chapter 4. Executing statements 48 import java.sql.*; import org.firebirdsql.jdbc.*; ... CallableStatement stmt = connection.prepareCall( "{call factorial(?, ?, ?)}"); FirebirdCallableStatement fbStmt = (FirebirdCallableStatement)stmt; fbStmt.setSelectableProcedure(true); stmt.setInt(1, 5); stmt.registerOutParameter(2, Types.INTEGER); // first OUT stmt.registerOutParameter(3, Types.INTEGER); // second OUT ResultSet rs = stmt.executeQuery(); while(rs.next()) { int firstCol = rs.getInt(1); // first OUT int secondCol = rs.getInt(2); // second OUT int anotherSecondCol = stmt.getInt(3); // second OUT } Illustration 4.14.: Example of using selectable stored procedure via escaped syntax Note that OUT parameter positions differ when they are accessed through the ResultSet interface (the firstCol and secondCol variables in our example). They are numbered in the order of their appearance in the procedure call starting with 1. When OUT parameter is accessed through the CallableStatement interface (the anotherSecondCol parameter in our example), the registered position should be used. In this case the result set can be used for navigation only. Describing Output and Input Parameters The PreparedStatement.getMetaData method is used to obtain description of the columns that will be returned by the prepared SELECT statement. The method returns an instance of java.sql.ResultSetMetaData interface that among other descriptions provides the following: • column type, name of the type, its scale and precision if relevant; • column name, its label and the display size; • name of the table, to which this column belongs; • information whether the column is read-only or writable, whether it contains signed numbers, whether it can contains NULL values, etc. Additionally, the JDBC 3.0 specification defines a new interface java.sql.ParameterMetaData that provides similar information for the input parameters of both PreparedStatement and CallableStatement objects. Note, due to the implementation specifics of the escaped syntax support for callable statements, it is not allowed to call getParameterMetaData before all OUT parameters are registered. Otherwise driver will try to prepare a procedure with an incorrect number of parameters and the database server will generate an error. Chapter 4. Executing statements 49 Batch Updates Batch updates are intended to group multiple update operations to be submitted to a database server to be processed at once. Firebird does not provide support for such functionality, but Jaybird emulates it by issuing separate update commands. Batch Updates with java.sql.Statement interface The Statement interface defines three methods for batch updates: addBatch, executeBatch and clearBatch. It is allowed to add arbitrary INSERT/UPDATE/DELETE or DDL statement to the batch group. Adding a statement that returns result set is an error. Statement stmt = connection.createStatement(); stmt.addBatch("UPDATE products " + "SET amount = amount – 1 WHERE id = 1"); stmt.addBatch("INSERT INTO orders(id, amount) VALUES(1, 1)"); int[]updateCounts = stmt.executeBatch(); Illustration 4.15.: Example of batch updates using Statement object The JDBC specification recommends to turn the auto-commit mode off to guarantee standard behavior for all databases. The specification explicitly states that behavior in auto-commit case is implementation defined. Jaybird executes a batch in a single transaction, i.e. the “all-or-nothing” principle. A new transaction is started before the batch execution and is committed if there were no exception during batch execution, or is rolled back if at least one batch command generated an error. The Statement.executeBatch method submits the job to the database server. In case of successful execution of the complete batch, it returns an array of integers containing update counts for each of the commands. Possible values are: • 0 or positive value – an update count for the corresponding update/DDL statement. • Statement.SUCCESS_NO_INFO – driver does not have any information about the update count, but it knows that statement was executed successfully. The Statement.executeBatch method closes the current result set if one is open. After successful execution the batch is cleared. Calling execute, executeUpdate and executeQuery before the batch is executed does not have any effect on the currently added batch statements. If at least one statement from the batch fails, a java.sql.BatchUpdateException is thrown. Jaybird will stop executing statements from batch after the first error. In auto-commit mode it will also rollback the transaction. An application can obtain update counts for the already executed statements using getUpdateCounts method of the BatchUpdateException class. The returned array will always contain fewer entries than there were statements in the batch. Batch Updates with java.sql.PreparedStatement and java.sql.CallableStatement Chapter 4. Executing statements 50 Using batch updates with a prepared statement is conceptually similar to the java.sql.Statement approach. Main difference is that only one statement can be used. PreparedStatement stmt = connection.prepareStatement( "INSERT INTO products(id, name) VALUES(?, ?)"); stmt.setInt(1, 1); stmt.setString(2, "apple"); stmt.addBatch(); stmt.setInt(1, 2); stmt.setString(2, "orange"); stmt.addBatch(); int[] updateCounts = stmt.executeBatch(); Illustration 4.16.: Example of batch updates with PreparedStatement PreparedStatement stmt = connection.prepareCall( "{call add_product(?, ?)"); stmt.setInt(1, 1); stmt.setString(2, "apple"); stmt.addBatch(); stmt.setInt(1, 2); stmt.setString(2, "orange"); stmt.addBatch(); int[] updateCounts = stmt.executeBatch(); Illustration 4.17.: Example of batch updates with CallableStatement Escaped Syntax Escaped syntax was introduced as a portable JDBC-specific syntax to represent parts of the SQL language that are usually implemented differently by database vendors. Also, the escaped syntax is used to define features that might not be implemented by the database server, but have an appropriate implementation in the driver. The JDBC specification defines escaped syntax for the following • scalar functions • date and time literals • outer joins • calling stored procedures • escape characters for LIKE clauses Jaybird implements the escaped syntax support for all cases except the last one which will be addressed in next releases. Scalar Functions Escaped syntax for the scalar function call is defined as Chapter 4. Executing statements 51 {fn (argument list)} For example {fn concat('Firebird', 'Java')} concatenates these two words into 'FirebirdJava' literal. “Ошибка: источник перекрестной ссылки не найден” provides a list of supported scalar functions. Date and Time Literals It is allowed to include date and time literals in SQL statements. In order to guarantee that each database will interpret the literal identically, the JDBC specification provides following syntax to specify them: Date literal escaped syntax: {d 'yyyy-mm-dd'} Time literal escaped syntax: {t 'hh:mm:ss'} Timestamp literal syntax (fractional seconds part '.f...' can be omitted): {ts 'yyyy-mm-dd hh:mm:ss.f...'} Outer Joins Due to the various approaches to specify outer joins (for instance, the Oracle “(+)” syntax), the JDBC specification provides the following syntax: {oj } where the outer join is specified as ::= {LEFT|RIGHT|FULL} OUTER JOIN {
| } ON An example SQL statement would look like this: SELECT * FROM {oj tableA a LEFT OUTER JOIN tableB b ON a.id = b.id} Stored Procedures The escaped syntax for stored procedures is described in details in the XXX where CallableStatetent is covered. LIKE Escaped Characters The percent sign (%) and underscore (_) characters are wild cards in LIKE clause of the SQL statement. In order to interpret them literally they must be preceded by the backslash character (\) that is called the escape character. The escaped syntax for this case identifies which character is used as an escape character: {escape ''} Chapter 4. Executing statements 52 Chapter 5. 5. Working with result sets When a SELECT statement is executed, the results of the query processing are returned through the implementation of the java.sql.ResultSet interface. ResultSet properties ResultSet Types JDBC 3.0 specification defines three types of result sets • TYPE_FORWARD_ONLY – the result set is not scrollable, cursor can move only forward. When the TRANSACTION_READ_COMMITTED isolation level is used, the result set will return all rows that are satisfying the search condition at the moment of the ResultSet.next() call. In other cases result set will return only rows that were visible at the moment of the transaction start. • TYPE_SCROLL_INSENSITIVE – the result set is scrollable, the cursor can move back and forth, can be positioned on the specified row. Only rows satisfying the condition at the time of query execution are visible. • TYPE_SCROLL_SENSITIVE, is not supported by Firebird and Jaybird. Driver allows application to ask for this type of result set, however according to the JDBC specification, the type is “downgraded” to the previous type and corresponding warning is added to the connection object. Due to a missing support of scrollable cursors in Firebird, their support (TYPE_SCROLL_INSENSITIVE result set type) is implemented by fetching the complete result set to the client. Scrolling happens in the memory on the client. This can have adverse effect on the system memory usage and performance when the result set is large. ResultSet Concurrency Result set concurrency specifies whether the result set object can be updated directly or a separate SQL request should be used to update the row. Result sets that allow direct modification using the ResultSet.updateXXX methods are usually used in GUI applications which allow in-place editing of the underlying result set. Type of result set concurrency is specified during statement creation and cannot be changed later. Jaybird supports two types of result set concurrency: • CONCUR_READ_ONLY is available for all types of result sets. It tells the driver that direct update of the result set is not possible and all ResultSet.updateXXX methods should throw an exception. • CONCUR_UPDATABLE is supported only under certain conditions that are needed for the driver to correctly construct a DML request that will modify exactly one row. These conditions are: • the SELECT statement that generated the result set references only one table; • all columns that are not referenced by the SELECT statement allow NULL values, otherwise it won't be possible to insert new rows; • the SELECT statement does not contain DISTINCT predicate, aggregate functions, joined tables or stored procedures; • the SELECT statement references all columns from the table primary key definition or the RDB$DB_KEY column. ResultSet Holdability Result set holdability tells driver whether result sets should be kept open across commits. ResultSet.HOLD_CURSORS_OVER_COMMIT tells the driver to keep the result set object open, while ResultSet.CLOSE_CURSORS_AT_COMMIT tells driver to close them on commit. This property is available only in JDBC 3.0 specification. When application calls Connection.commit(), the Firebird server closes all open result sets. It is not possible to tell the server to keep result set open over commit unless “commit retaining” mode is used. This mode is global for the complete connection and is not suitable for holdability control on the statement level. Also this mode is believed to have an undesired side-effect for read-write transactions as it inhibits garbage collection. Because of these two reasons “commit retaining” is not used in Jaybird during normal execution. Applications are able to commit the transaction keeping the result sets open by executing a "COMMIT RETAIN" SQL statement. ResultSet manipulation ResultSet objects are created when either Statement.executeQuery(String) or Statement.getResultSet() methods are called. The latter is used in combination with the Statement.execute(String) method and can be called only once per result set (see Illustration 5.1 and Illustration 5.2 respectively). Chapter 5. Working with result sets 54 Note, current implementation does not allow to call getResultSet() method after using the executeQuery(String) method of the Statement class. The JDBC 3.0 specification is unclear on this topic and JDBC drivers of different vendors treat it differently. Statement stmt = connection.createStatement(); ResultSet rs = stmt.executeQuery("SELECT * FROM myTable"); Illustration 5.1.: Using Statement.executeQuery(String) method. Statement stmt = connection.createStatement(); boolean hasResultSet = stmt.execute("SELECT * FROM myTable"); if (hasResultSet) { ResultSet rs = stmt.getResultSet(); ... } Illustration 5.2.: Using Statement.getResultSet() method. Accessing the values in the result set Depending on the type of the result set it is possible to move the cursor either forward only (Illustration 5.3) or using absolute and relative positioning (Illustration 5.4). Values of the result set are obtained by calling the corresponding getter method depending on the type of column. For example the ResultSet.getInt(1) method returns the value of the first column as an int value. If value of the column is not integer, driver tries to convert it according to the “Data Type Conversion Table” specified in Ошибка: источник перекрестной ссылки не найден. If conversion is not possible, an exception is thrown. There are two possibilities to obtain data from the result set columns: by column name or by column position. Position of the first column is 1. Names supplied to getter methods are case-insensitive. The search first happens in the column aliases, and if no match found, driver checks the original column names. If there is more then one column matching the specified name (even if the original names were quoted), the first match is taken. When getters for primitive types are used and original value in the result set is NULL, driver returns a default value for that type. For example getInt() method will return 0. In order to know whether the value is really 0 or NULL, you have to call ResultSet.wasNull(...) method. Getters that return object values (getString, getDate, getObject, etc.) will correctly report a null value for the columns containing NULL. Chapter 5. Working with result sets 55 Statement forwardStatement = connection.createStatement(); ResultSet rs = forwardStatement.executeQuery( "SELECT id, name, price FROM myTable"); while(rs.next()) { int id = rs.getInt(1); String name = rs.getString("name"); double price = rs.getDouble(3); } Illustration 5.3.: Example of using forward-only result sets. Updating records in the result set Scrollable cursors are especially useful when result of some query is displayed by the application which also allows the user to directly edit the data and post the changes to the database. Statement scrollStatement = connection.createStatement( ResultSet.TYPE_SCROLL_INSENSITIVE, ResultSet.CONCUR_UPDATABLE); ResultSet rs = scrollStatement.executeQuery( "SELECT id, name, price FROM myTable"); rs.absolute(1); // move to the first row rs.updateString(2, anotherName); // update the name rs.updateRow(); // post changes to the db rs.moveToInsertRow(); rs.updateInt(1, newId); rs.updateString(2, newName); rs.updateDouble(3, newPrice); rs.insertRow(); rs.moveToCurrentRow(); rs.relative(-2); Illustration 5.4.: Example of using scrollable and updatable result sets. The code example on Illustration 5.4 shows how to update first row, insert new one and after that move two records backwards. Also an application can update the current row using so called “positioned updates” on named cursors. This technique can be used only with forward-only cursors, since application can update only the row to which the server-side cursor points to. In case of scrollable cursors the complete result set is fetched to the client and then the server-side cursor is closed. The example on Illustration 5.5 shows how to use positioned updates. First of all application has to specify the name of the cursor and the list of the columns that will be updated before the query is executed. This name is later used in the UPDATE statement as showed on the example. Chapter 5. Working with result sets 56 Statement selectStmt = connection.createStatement(); selectStmt.setCursorName("someCursor"); Statement updateStmt = connection.createStatement(); ResultSet rs = selectStmt.executeQuery( "SELECT id, name, price FROM myTable " + "FOR UPDATE OF myColumn"); while(rs.next()) { ... if (someCondition) { updateStmt.executeUpdate("UPDATE myTable " + "SET myColumn = myColumn + 1 " + "WHERE CURRENT OF " + rs.getCursorName()); } } Illustration 5.5.: Example of using the positioned updates. Closing the result set A result set is closed by calling the ResultSet.close() method. This releases the associated server resources and makes the ResultSet object available for garbage collection. It is strongly recommended to explicitly close result sets in auto-commit more or ResultSet.TYPE_SCROLL_INSENSITIVE result sets, because this releases memory used for the cached data. The result set object is also closed automatically, when the statement that created it is closed or re-executed. In auto-commit mode, the result set is closed automatically if any statement is executed over the same connection. Chapter 5. Working with result sets 57 Chapter 6. 6. Using transactions Transactions are used to group SQL statements into a single block that satisfies so called ACID properties: atomicity, consistency, isolation and durability. In other words, all statements executed within transaction will either succeed and their results will be permanently stored in the database or the effect of the statement execution will be undone. JDBC transactions Firebird supports multiple concurrent transactions over the same database connection. This allows applications that work via native Firebird API to save the number of network connections, which in turn saves the resources on the server2. This model however cannot be applied to each database engine in the world and designers of the JDBC API have chosen a model where each database connection has one and only one active transaction associated with it. Also, unlike the Firebird model, where transactions require explicit start, JDBC specification requires the driver to start transaction automatically as soon as transactional context is needed. The Illustration 6.1 shows a very simple example of using transactions in JDBC where a hypothetical intruder that increases salary of each employee twice and uses explicit transaction control in JDBC. It also tries to hide his own identity and if the operations succeed, he commits the transaction, otherwise he rolls the changes back. 2 Additionally, before the InterBase was open-sourced, this allowed application developers to create multi-threaded application without need to purchase additional user licenses. Connection connection = ... connection.setAutoCommit(false); Statement stmt = connection.createStament(); try { stmt.executeUpdate("UPDATE employee " + " SET salary = salary * 2"); // ... do some more changes to database // to hide the identity of the person // that messed up the salary information // by deleting the audit trails logs, etc. stmt.executeUpdate("DELETE FROM audit_trails"); connection.commit(); } catch(SQLException ex) { connection.rollback(); } Illustration 6.1.: Example of explicit transaction control. In order to do this, application first switches the auto-commit mode off (see below for more information), then creates a java.sql.Statement object, executes an UPDATE statement. Please note, that there is no explicit transaction start, new transaction will be started right before executing the statement (step 2). If we have database where not only referential integrity is preserved, but also reasonable security rules are encoded in the triggers, it will raise an error preventing cleaning the audit trails information. In this case intruder chooses to undo all the changes he made, so that nobody notices anything. But if no security rules are implemented, he commits the transaction. When connection uses explicit transaction control, each transaction must be finished by calling the commit or rollback methods of the Connection object before the connection is closed. If a transaction was not finished, but the close method is called, the active transaction is rolled back automatically. This also happens when transaction was not finished and connection was not closed explicitly and that Connection object became eligible for garbage collection. In this case, the close method is implicitly invoked by the class finalizer, which in turn rolls the transaction back. Auto-commit mode Each newly created connection by default has the auto-commit property set on. In other words, the duration of the transaction is limited by the duration of statement execution, or speaking formal language – transaction is ended when the statement is completed. The point when statement executions is considered complete is defined in the specification according to the table on Illustration 6.2. Chapter 6. Using transactions 60 4 3 2 1 • For insert, update, delete and DDL statements, the statement is complete as soon as it has finished executing. • For select statements, statement is complete when the associated result set is closed. The result set is closed as soon as one of the following occurs: • all of the rows have been retrieved • the associated Statement object is re-executed • another Statement object is executed on the same connection • For CallableStatement objects, the statement is complete, when all of the associated result sets have been closed. Illustration 6.2.: Rules when the statement is completed in auto-commit mode If there is an ongoing transaction and the value of the auto-commit property is changed, the current transaction is committed. Note, when a connection is obtained via javax.sql.DataSource object and container managed transactions are used (for example, the application is executing inside an EJB container), it is an error to call setAutoCommit method. Special care should be taken when using multiple statements in auto-commit mode. The JDBC 2.0 specification did not define the rules for the statement completion as well as it did not define the behavior of multiple ResultSet objects created using the same Connection object in auto-commit mode. Since Firebird does not allow the result set to remain open after the transaction ends, Jaybird 1.5.x and below cached the complete result set in memory when the SELECT statements were executed and corresponding transaction was committed. It had an adverse effect on allocated memory when the result set is big, especially when it contains BLOB fields. The JDBC 3.0 specification addressed this unclear situation (see Illustration 6.2) and Jaybird 2.1 was improved to correctly handle them. It also allowed to improve the memory footprint – the driver no longer caches non-scrollable and non-holdable result sets in memory. However, some Java applications that do not conform the current JDBC specification might no longer work with Jaybird 2.1 and above unless additional steps are taken. The piece of code on Illustration 6.3 works perfectly with explicit transaction control. However, it won't work correctly with a driver that complies with JDBC 3.0 specification, while the selectStmt and updateStmt object were created by the same connection object (step 1). When the UPDATE is executed in step 3, the result set produced by the SELECT statement must be closed before the execution. When Java application tries to fetch the next record by calling the rs.next() method, it will receive an SQLException with a message “Result set object is closed”. The only correct solution to this situation is to fix the application by either using explicit transaction control, or by using two connection objects, one for SELECT statement and one for UPDATE statement. Chapter 6. Using transactions 61 Statement selectStmt = connection.createStatement(); Statement updateStmt = connection.createStatement(); ResultSet rs = selectStmt.executeQuery( "SELECT * FROM myTable"); while(rs.next()) { int id = rs.getInt(1); String name = rs.getString(2); updateStmt.executeUpdate("UPDATE anotherTable SET " + " name = '" + name + "', " + " WHERE id = " + id); } Illustration 6.3.: Non-compliant usage of nested statements in auto-commit mode. Unfortunately not all applications can be changed either because there is no source code available or, simply, because any change in the code requires complete release testing of the software. The Jaybird 2.1 introduces new connection parameter defaultHoldable which makes result sets holdable by default. The holdable result sets will be fully cached in memory, but won't be closed automatically when transaction ends3. Please note, this property affects also the cases with explicit transaction control. Read-only Transactions A transaction can be declared read-only to reduce the possibility of lock conflicts. In general, this makes little sense for Firebird, because of its multi-generational architecture, where readers do not block writers and vice versa. However, in some cases it can be useful. It is not allowed to connect with a read-write transaction to a database located on a read-only media, for example, a CD-ROM. The reason is that, in order to guarantee consistency of the read-write transactions, Firebird has to increase the transaction identifier when transaction ends, and to store the new value on the so- called Transaction Inventory Page even if no changes were made in that transaction. This requirement can be relaxed if transaction is declared read-only and the engine ensures that no data can be modified. Another reason is that long running read-write transactions inhibit the process of collecting garbage, i.e. a process of identifying previous versions of the database records that are no longer needed and releasing the occupied space for the new versions. Without garbage collection the database size will grow very fast and the speed of the database operations will decrease, because the database engine will have to check all available record versions to determine the appropriate one. Therefore, if you are sure that application won't modify the database in the transaction, use the setReadOnly method of the java.sql.Connection object to tell the server that the transaction is read-only. Transaction Isolation Levels 3 Other cases, e.g. closing the statement object or the connection object will still ensure that the result set object is closed. If you need result sets that can be “detached” from the statement object that created them, please check the javax.sql.RecordSet implementations. Chapter 6. Using transactions 62 3 2 1 To address the performance issue, the isolation property is relaxed. The ANSI/ISO SQL standard defines four such levels, each next one weaker than the previous. These isolation levels were reflected in the JDBC specification (Illustration 6.4) JDBC isolation level Description TRANSACTION_SERIALIZABLE Transactions with this isolation level prohibit the phantom reads, the situation when one transaction reads all rows satisfying the WHERE condition, another transaction inserts a row satisfying that condition, and first transaction re- executes the statement. TRANSACTION_REPEATABLE_ READ This isolation level prevents the non-repeatable reads, a situation when a row is read in one transaction, then modified in another transaction, and later re-read in the first transaction. In this case different values had been read within the same transaction. TRANSACTION_READ_COMMIT TED Transactions with this isolation level can see only committed records. However, it does not prevent so-called non-repeatable reads and phantom reads. TRANSACTION_READ_UNCOM MITTED The weakest isolation level, or better to say level with no isolation. Such transactions can see the not yet committed changes to the data in the database from the concurrently running transactions. Illustration 6.4.: JDBC transaction isolation levels and their characteristics. Firebird, however, defines other isolation levels: read_committed, concurrency and consistency. Only the read_committed isolation level can be mapped to the same level defined by the ANSI/ISO SQL standard. The dirty reads are prevented, non-repeatable reads as well as phantom reads can occur. The concurrency isolation level is stronger than repeatable read isolation defined in ANSI/SQL standard and satisfies the requirements of a serializable isolation level, however, unlike RDBMSes with locking concurrency control, it guarantees better performance. And finally Firebird provides a consistency isolation level which in combination with table reservation feature guarantees the deadlock-free execution of transactions. A transaction will be prevented from starting if there is already another one with the overlapping sets of the reserved tables. This isolation level guarantees truly serial history of transaction execution. In order to satisfy the JDBC specification Jaybird provides a following default mapping of the JDBC transaction isolation levels into Firebird isolation levels: • TRANSACTION_READ_COMMITTED is mapped to read_committed isolation level in Firebird – any changes made inside a transaction are not visible outside a transaction until the transaction is committed. A transaction in Chapter 6. Using transactions 63 read-committed mode sees all committed changes made by other transactions even if that happened after start of the current transaction. • TRANSACTION_REPEATABLE_READ is mapped to concurrency isolation level in Firebird – any changes made inside this transaction are not visible outside a transaction until the transaction is committed. A transaction in repeatable-read sees only those changes that were committed before the transaction started. Any committed change in another transaction that happened after the start of this transaction is not visible in this transaction. • TRANSACTION_SERIALIZABLE is mapped into consistency isolation level in Firebird – any modification to a table happens in serial way: all transactions wait until the current modification is done. This mode can be considered as a traditional pessimistic locking scheme, but the lock is placed on the whole table. See Chapter “Table Reservation” on page 68 for more information. The mapping is specified in the isc_tpb_mapping.properties file that can be found in the Jaybird archive and can be overridden via the connection properties • via the tpbMapping property that specifies the path to the PropertiesResourceBundle with the new mapping of the isolation level; • via the direct specification of the JDBC transaction isolation level. Illustration 6.5 contains an example of such operation, the values in the mapping are described in Chapter “Transaction Parameter Buffer” on page 65. • via the connection pool configuration. Properties props = new Properties(); props.setProperty("user", "SYSDBA"); props.setProperty("password", "masterkey"); props.setProperty("TRANSACTION_READ_COMMITTED", "isc_tpb_read_committed,isc_no_rec_version," + "isc_tpb_write,isc_tpb_nowait"); Connection connection = DriverManager.getConnection( "jdbc:firebirdsql:localhost/3050:c:/example.fdb", props); Illustration 6.5.: Overriding the default isolation level mapping. The overridden mapping is used for all transactions started within the database connection. If the default mapping is overridden via the connection pool configuration, it will be used for all connections created by the pool. Savepoints Savepoints provide finer-grained control over transactions by providing intermediate steps within a larger transaction. Once a savepoint has been set, transaction can be rollback to that point without affecting preceding work. In order to set a savepoint, use following code: Chapter 6. Using transactions 64 Statement stmt = connection.createStatement(); stmt.executeUpdate( "INSERT INTO myTable(id, name) VALUES (1, 'John')"); Savepoint savePoint1 = connection.setSavepoint("savepoint_1"); stmt.executeUpdate( "UPDATE myTable SET name = 'Ann' WHERE id = 1"); ... connection.rollback(savePoint1); // at this point changes done by second update are undone Illustration 6.6.:Example of using savepoints Note, rolling back to the savepoint automatically releases and invalidates any savepoints that were created after the released savepoint. If the savepoint is no longer needed, you can use the Connection.releaseSavepoint method to release system resources. After releasing a savepoint it is no longer possible to rollback the current transaction to that savepoint. Attempt to call the rollback method will result in an SQLException. Savepoints that have been created within a transaction are automatically released when transaction is committed or rolled back. Transaction Parameter Buffer The behavior of the Firebird transactions is internally controlled by the Transaction Parameter Buffer (TPB), which specifies different transaction properties: • the transaction isolation level; • the transaction's read-only or read-write mode; • the lock conflict resolution mode – wait or no wait; • and, finally, the table reservations – their names and reservation modes. The TPB is automatically generated depending on the transaction isolation level specified for the java.sql.Connection object and usually there is no need to manipulate the TPB directly. Additionally, if the connection is set to read-only mode, this is reflected in the TPB by appropriate constant. However, the lock resolution mode as well as table reservations cannot be specified by using the standard JDBC interfaces. For the cases where this is needed, Jaybird provides an extension of the JDBC standard. Chapter 6. Using transactions 65 FirebirdConnection fbConnection = (FirebirdConnection)connection; TransactionParameterBuffer tpb = fbConnection.createTransactionParameterBuffer(); tpb.addArgument(TransactionParameterBuffer.READ_COMMITTED); tpb.addArgument(TransactionParameterBuffer.REC_VERSION); tpb.addArgument(TransactionParameterBuffer.WRITE); tpb.addArgument(TransactionParameterBuffer.WAIT); fbConnection.setTransactionParameters(tpb); Illustration 6.7.: Example of specifying custom TPB. The Illustration 6.7 presents an example of populating the TPB with custom parameters. Isolation level As it was already mentioned before, Firebird supports three isolation levels: read_committed, concurrency and consistency which are represented by the appropriate constants in the TransactionParameterBuffer class. The isolation level specifies the way the database engine processes the record versions on read operations. The concurrency isolation level is also often called SNAPSHOT and the consistency - SNAPSHOT TABLE STABILITY isolation levels. In consistency and concurrency modes Firebird database engine loads the different versions of the same record from disk and checks the “timestamps” of each version and compares it with the “timestamp” of the current transaction. The record version with the highest timestamp that is however lower or equal to the timestamp of the current transaction is returned to the application. This effectively returns the version of the record that was when the current transaction started and guarantees that neither non-repeatable reads not phantom reads can ever occur. In read_committed mode Firebird database engine, however, accesses the record version with the highest timestamp, for which the corresponding transaction is marked as committed. This prevents engine from reading the record versions which were modified in the concurrent transactions that are not yet committed or were rolled back for whatever reasons. However, such mode allows non- repeatable reads as well as phantom reads if the concurrent transaction that modified records or inserted new ones had been committed. The read_committed isolation mode requires another constant that specifies the behavior of the transaction when it meets a record version with the timestamp which belongs to the currently running transaction which is not yet committed. The most application require the TransactionParameterBuffer.REC_VERSION mode, which is shown on Illustration 6.7. In this mode database engine fetches the latest committed version as described before. The TransactionParameterBuffer.NO_REC_VERSION constant tells database engine to report the lock conflict when uncommitted record version is met while fetching data from the database. The outcome of the operation is then controlled by the lock resolution mode (see page 67). Read-only transactions Chapter 6. Using transactions 66 The read-only or read-write transaction mode is controlled by two constants: • TranscationParameterBuffer.READ and • TransactionParameterBuffer.WRITE When the read-write mode is specified, database engine stores the “timestamp” of new transaction in the database even no modification will be made in the transaction. The “timestamp” affects the garbage collection process, since the database engine cannot release records that were modified in the transactions with higher “timestamps” even when these record versions are no longer needed (in other words, when there are already newer versions of the records). Thus, the long-running read-write transaction inhibits the garbage collection even when no modifications were done in it. Therefore, it is recommended to set the read-only mode for the transaction when it is used for read operations. Lock resolution mode The RDBMS systems that use pessimistic locking for the concurrency control lock the records regardless of the operation type, read or write. When application tries to read a record from the database, database engine tries to obtain a “read lock” to that record. If the operation succeeds and application later tries to update the record, the lock is upgraded to the “write lock”. And finally, if the resource is already locked for write, concurrent transactions cannot lock it for reading, since the system cannot allow the transaction to make a decision based on data that might be rolled back later. This in approach significantly decreases concurrency. However, the databases systems that employ the record versioning mechanism do not have such restriction because each transaction “sees” its own version of the record – the only possible conflict happens when two concurrent transactions try to obtain “write lock” for the same database record. Firebird belongs to the latter, and on read_committed and concurrency isolation levels it behaves appropriately – there are no lock conflicts between readers and writers, and only writers competing for the same resource raise a lock conflict. However, on the consistency isolation level Firebird emulates the behavior of the systems with pessimistic locking – read operation will conflict with write. Even more, the locks are obtained for the whole tables (see “Table Reservation” chapter for details). The Illustration 6.8 summarizes the above said in a table for Firebird 2.0. It shows that read-committed or repeatable read transactions conflict only when they simultaneously update the same rows. In contrast, a consistency transaction conflicts with any transaction running in read-write mode, e.g. as soon as a consistency transaction gets write access to a table, other read-write transactions are not allowed to make changes in that tables. Chapter 6. Using transactions 67 Read-committed Concurrency Consistency read-write read-only read-write read-only Read-committed Concurrency read-write some updates may conflict - conflict conflict read-only - - - - Consistency read-write conflict - conflict conflict read-only conflict - conflict - Illustration 6.8.: Lock conflicts within one table depending on the isolation level. Table Reservation Table reservation allows to specify the database tables and the corresponding access modes at the beginning of the transaction. When transaction is started, engine tries to obtain the requested locks for the specified tables and proceeds only when all of them were successfully obtained. Such behavior allows to create a deadlock-free execution histories4. The table reservation is specified via TPB and includes the table to lock, the lock mode (read or write) and lock type (shared, protected and exclusive). Connection connection = ... TransactionParameterBuffer tpb = connection.createTransactionParameterBuffer(); tpb.addArgument(TransactionParameterBuffer.CONSISTENCY); tpb.addArgument(TransactionParameterBuffer.WRITE); tpb.addArgument(TransactionParameterBuffer.NOWAIT); tpb.addArgument(TransactionParameterBuffer.LOCK_WRITE, "TEST_LOCK"); tpb.addArgument(TransactionParameterBuffer.PROTECTED); connection.setTransactionParameters(tpb); // next transaction will lock TEST_LOCK table for writing // in protected mode Illustration 6.9.: Example of using table reservation facility in Firebird. The Illustration 6.8 shows an example of reserving the TEST_LOCK table for writing in a protected mode. The code does the following: 1. Create a new instance of TransactionParameterBuffer class. 2. Populate the TPB. The first three statements were described in the chapter “Transaction Parameter Buffer”. The fourth call specifies that the application wants to obtain a lock on the table TEST_LOCK for writing. The 4 This approach follows the two-phase locking protocol, where all locks are acquired on the beginning of the transaction and are released only when transaction is finished. Chapter 6. Using transactions 68 3 2 1 fifth call specifies the type of the lock to obtain, in our case the protected lock. 3. Set the new TPB to be used for the next transaction. The lock mode to the table specified in the TPB can be either ● TransactionParameterBuffer.LOCK_READ for read-only access to the table; ● or TransactionParameterBuffer.LOCK_WRITE for read-write access to the table. The lock type can be either ● TransactionParameterBuffer.SHARED for shared access to the table; ● or, TransactionParameterBuffer.PROTECTED for protected access to the table; The TransactionParameterBuffer.EXCLUSIVE mode was introduced in latter versions of Firebird, however it behaves like PROTECTED mode for all read-write transactions. The lock conflict table depends on the isolation level of the transactions and has the following properties: ● LOCK_WRITE mode always conflicts with another LOCK_WRITE mode regardless of the lock type and transaction isolation mode; ● LOCK_WRITE always conflicts with another LOCK_READ mode if both transactions have consistency isolation, but has no conflict with shared- read locks it if another transaction has either concurrency or read_committed isolation level; ● LOCK_READ mode never conflicts with LOCK_READ mode. Chapter 6. Using transactions 69 Chapter 7. 7. Working with Services Additionally to the normal database connections Firebird features the server-wide connections. These are used to perform various administrative tasks in Firebird, e.g. database backup, maintenance, statistics. The set of API calls to perform such tasks are known under the name “Services API”. Additionally client application can use the Services API to get some limited information about the server environment and configuration. The actual execution of the Services API calls can be viewed as a tasks triggered from the client application to be executed on server. The parameters passed in the calls are internally used to construct the string similar to the one that is passed to command-line tools. Later this string is passed into entry routine of the gbak, gfix, gsec or gstat utility. The output of the utility, which in normal case is printed to standard out, is in this case transmitted over the network to the client application. Jaybird JCA/JDBC driver hides the complexity of the original API providing a set of interfaces and their implementations to perform the administrative tasks regardless of the usage mode (i.e. remote server and embedded engine, wire protocol and access via native client library). This chapter describes the available Java API for the administrative tasks. All classes and interfaces described below are defined in the org.firebirdsql.management package. Each management class works as a standalone object and does not require an open connection to the server. ServiceManager The ServiceManager interface and the FBServiceManager class are defined as the common superclasses providing setters and getters for common properties as well as some common routines. The following properties can be specified: Name Type Description host java.lang.String Name or the IP address of the host to which we make the Service API request. Required. port int Port to which we make the request, 3050 by default. database java.lang.String Path to the database. The meaning of the property depends on the service being invoked and will be described in each of chapters below. user java.lang.String Name of the user on behalf of which which all Service API calls will be executed. Required. password java.lang.String Password corresponding to the specified user. Required. logger java.io.OutputStream Stream into which the output of the remote service will be written to. Optional. The last parameter requires some more information. The calls to all Services API routines are asynchronous. The client application can start the call, but there are no other means to find out whether execution of the service call is finished or not except reading the output of the service call – EOF in this case means that execution is finished. The FBServiceManager converts the asynchronous calls into synchronous by constantly polling the service output stream. If logger property is specified the received data are copied into the specified OutputStream, otherwise it is simply ignored and the EOF-marker is being watched. This behavior can be changed by overriding the appropriate method in the FBServiceManager class and/or subclasses. The only requirement is to detach from the service manager when it is no longer needed. Backup and restore The backup and restore routines are defined in the BackupManager interface and are implemented in the FBBackupManager class. Additionally to the setters and getters described in the previous section the following methods are used to specify the backup and restore paths and properties: Name Type Description database String For backup operation it specifies the path to the database to backup. For restore operation it specifies the path to the database into which the backup file will be restored. In case when multi-file database should be created, use addRestorePath(String, int) method instead. backupPath String Path to the backup file. For backup operation specifies the path and the file name of the newly created backup file. If multi-file backup files are to Chapter 7. Working with Services 72 be created, use addBackupPath(String, int) method instead. For restore operations specifies path to the single backup file. If database should be restored from multi-file backup, please use the addBackupPath(String) method instead. restorePageBufferCount int Number of pages that will be cached of this particular database. Should be used only for restore operation. restorePageSize int Size of the database page. Should be used only for restore operation. Valid values depend on the Firebird version, but should be one of the 1024, 2048, 4096 or 8192. restoreReadOnly boolean Set to true if the database should be restored in read-only mode. restoreReplace boolean Set to true if restore should replace the existing database with the one from backup. Warning! It is easy to drop the only existing database if the backup can't be restored, as the existing database is first deleted and only after that the restore process starts. To avoid such situation it is recommended to restore database into some dummy file first and then use file system commands to replace the existing database with the newly created one. verbose boolean Be verbose when writing to the log. The service called on the server will produce lots of output that will be written to the output stream specified in logger property. Additionally to the properties following methods are used to configure the paths to backup and database files when multi-file backup or restore operations are used. Method Description addBackupPath(String) Add a path to a backup file from a multi-file backup. Should be used for restore operation only. addBackupPath(String, int) Add a path to the multi-file backup. The second parameter specifies the ,maximum size of the particular file in bytes. Should be used for backup operation only. addRestorePath(String, int) Add a path for the multi-file database. The second parameter specifies the maximum size of the database file in pages (in other words, the maximum size in bytes can be obtained by multiplying this value by restorePageSize parameter) clearBackupPaths() Clear all the specified backup paths. This method also clears the path specified in backupPath property. clearRestorePaths() Clear all the specified restore paths. This method also clears the path specified in the database Chapter 7. Working with Services 73 Method Description property. Please note, all paths that are specified are paths specifications on the remote server. This has following implications: a) it is not possible to backup to the local or network drive unless it is mounted on the remote server; b) it is not possible to restore from the local or network drive unless it is mounted on the remote server. After specifying all the needed properties application developer can use backupDatabase(), backupMetadata() and restoreDatabase() methods to perform the backup and restore tasks. These methods will block until the operation is finished. If the logger property was set, the output of the service will be written into the specified output stream, otherwise it will be ignored5. // backup the database BackupManager backupManager = new FBBackupManager(); backupManager.setHost("localhost"); backupManager.setPort(3050); backupManager.setUser("SYSDBA"); backupManager.setPassword("masterkey"); backupManager.setLogger(System.out); backupManager.setVerbose(true); backupManager.setDatabase("C:/database/employee.fdb"); backupManager.setBackupPath("C:/database/employee.fbk"); backupManager.backupDatabase(); ... // and restore it back BackupManager restoreManager = new FBBackupManager(); restoreManager.setHost("localhost"); restoreManager.setPort(3050); restoreManager.setUser("SYSDBA"); restoreManager.setPassword("masterkey"); restoreManager.setLogger(System.out); restoreManager.setVerbose(true); restoreManager.setRestoreReplace(true); // attention!!! restoreManager.setDatabase("C:/database/employee.fdb"); restoreManager.setBackupPath("C:/database/employee.fbk"); backupManager.restoreDatabase(); Illustration 7.1.: Example of backup and restore process. The methods backupDatabase(int) and restoreDatabase(int) provide a possibility to specify additional backup and restore options that cannot be 5 The output of the service is always transferred over the network regardless whether the logger property is set or not. Additionally to providing a possibility to the user to track the service progress it acts also as a signal of operation completion – in this case the Java code will receive an EOF marker. Chapter 7. Working with Services 74 specified via the properties of this class. The parameter value is bitwise combination of the following constants: Constant Description BACKUP_CONVERT Backup external files as tables. By default external tables are not backed up, only references to the external files with data are stored in the backup file. When this option is used, the backup will store the external table as if they were regular tables. After restore the tables will remain regular tables. BACKUP_EXPAND No data compression. The gbak utility uses RLE compression for the strings in backup file. Using this option tells it to write strings in their full length, possibly fully consisting of empty characters, etc. BACKUP_IGNORE_CHECKSUMS Ignore checksums. The backup utility can't backup a database with page checksum errors. Such database is considered corrupted and the completeness and correctness of the backup cannot be guaranteed. However in some cases such errors can be ignored, e.g. when the index page is corrupted. In such case the data in the database are ok and the error disappears when the database is restored and index is recreated. Use this option only when checksum errors are detected and can't be corrected without full backup/restore cycle. Ensure that the restored database contains correct data afterwards. BACKUP_IGNORE_LIMBO Ignore in-limbo transactions. The backup utility can't backup database with in- limbo transactions. When such situation appears, backup has to wait until the decision about the outcome of the in-limbo transaction. After some timeout exception is thrown and backup is aborted. This option allows to workaround this situation – the gbak looks for the most recent committed version of the record and writes it into backup. BACKUP_METADATA_ONLY Backup metadata only. When this option is specified, the backup utility backups only the metadata information (e.g. table an/or view structure, stored procedures, etc.), but no data are backed up. This allows restoring a clean database from the backup. BACKUP_NO_GARBAGE_COLLECT Do not collect garbage during backup. The backup process reads all records in the tables one by one. When cooperative garbage collection is enabled6 the transaction that accesses the latest version of the record is also responsible for marking 6 Cooperative garbage collection can be switched off in Firebird 2.0 SuperServer architecture by corresponding configuration option. It can't be switched off in ClassicServer architecture and in previous Firebird versions. Chapter 7. Working with Services 75 Constant Description the previous versions as garbage. This process is time consuming and might be switched off when creating backup, where the most recent version will be read. Later operator can restore the database from the backup. In database with many backversions of the records the backup-restore cycle can be faster than traditional garbage collection. BACKUP_NON_TRANSPORTABLE Use non-transportable backup format. By default gbak creates so-called transportable backup where it does not make difference whether it is later restored on the platform with big or little endianess. By using this option a non-transportable format will be used which allows restoring the database only on the same architecture. BACKUP_OLD_DESCRIPTIONS Save old style metadata descriptions. Actually no real information exist for this option, by default it is switched off. RESTORE_DEACTIVATE_INDEX Deactivate indexes during restore. By default indexes are created at the beginning of the restore process and they are updated with each record being restored from the backup file. On a big tables it is more efficient first to store data in the database and to update the index afterwards. When this option is specified, the indexes will be restored in the inactive state. Application is however required to activate indexes afterwards, it won't happen automatically. RESTORE_NO_SHADOW Do not restore shadow database. If the shadow database is configured, an absolute path to the shadow is stored in the backup file. If such backup file is restored on a different system where such path does not exist (e.g. moving database from Windows to Linux or otherwise), restore process will fail. Using this option allows to overcome such situation. RESTORE_NO_VALIDITY Do not restore validity constraints. This option is usually needed when the validity constrains (e.g. NOT NULL constraints) were added after the data were already in the database but the database contains records that do not satisfy such constraints7. When this option is specified, the validity constraints won't be restored. This allows to recover the data and perform cleanup tasks. Application and/or administrators are responsible for restoring the validity constrains afterwards. 7 All versions of Firebird including 2.0 allow to define validity constraints despite the table(s) contain data that do not satisfy them. Only the new records will be validated, and it is responsibility of the database administrator to ensure the validity of existing ones. Chapter 7. Working with Services 76 Constant Description RESTORE_ONE_AT_A_TIME Commit after completing restore of each table. By default all data are restored in one transaction. If for some reason the complete restore is not possible, using this option will allow to restore at least some of the data. RESTORE_USE_ALL_SPACE Do not reserve 20% on each page for the future versions, useful for read-only databases. Example of using these options can be found on Illustration 7.2. BackupManager restoreManager = new FBBackupManager(); restoreManager.setHost("localhost"); restoreManager.setPort(3050); restoreManager.setUser("SYSDBA"); restoreManager.setPassword("masterkey"); restoreManager.setLogger(System.out); restoreManager.setVerbose(true); restoreManager.setRestoreReplace(true); // attention!!! restoreManager.setDatabase("C:/database/employee.fdb"); restoreManager.setBackupPath("C:/database/employee.fbk"); // restore database with no indexes, // validity constraints and shadow database backupManager.restoreDatabase( BackupManager.RESTORE_DEACTIVATE_INDEX | BackupManager.RESTORE_NO_VALIDITY | BackupManager.RESTORE_NO_SHADOW | BackupManager.RESTORE_ONE_AT_A_TIME); Illustration 7.2.: Example of using extended options for restore. User management The next service available is the user management. The routines are defined in the UserManager interface and are implemented in the FBUserManager class. Additionally, there is an User interface providing getters and setters for properties of a user account on the server and corresponding implementation in the FBUser8 class. The available properties of the FBUser class are: Name Type Description userName String Unique name of the user on the Firebird server. Required. Maximum length is 31 byte. password String Corresponding password. Getter return value only if the password had been set firstName String First name of the user. Optional. middleName String Middle name of the user. Optional. lastName String Last name of the user. Optional. userId int ID of the user on Unix. Optional. 8 The class implementation is simple bean publishing the properties via getters and setters. You can replace it with any other implementation of User interface. Chapter 7. Working with Services 77 Name Type Description groupId int ID of the group on Unix. Optional. The management class, FBUserManager has following methods to manipulate the user accounts on the server: Method Description getUsers():Map Method delivers a map containing user names as keys and instances of FBUser class as values containing all users that are registered on the server. The instances of FBUser class do not contain passwords, the corresponding property is null. addUser(User) Register the user account on the server. updateUser(User) Update the user account on the server. deleteUser(User) Delete the user account on the server. The Illustration 7.3 shows an example of using the FBUserManager class. UserManager userManager = new FBUserManager(); userManager.setHost("localhost"); userManager.setPort(3050); userManager.setUser("SYSDBA"); userManager.setPassword("masterkey"); User user = new FBUser(); user.setUserName("TESTUSER123"); user.setPassword("test123"); user.setFirstName("John"); user.setMiddleName("W."); user.setLastName("Doe"); userManager.add(user); Illustration 7.3.: Example of FBUserManager class usage. Database maintenance The database maintenance is something that everybody would prefer to avoid, and, contrary to the backup/restore and user management procedures, there little automation that can be done here. Usually the maintenance tasks are performed on the server by the database administrator, but some routines are needed to perform the automated database upgrade or perform periodic checks of the database validity. This chapter describes the methods declared in the MaintenanceManager interface and its implementation, the FBMaintenanceManager class. Database shutdown and restart One of the most often used maintenance operations is database shutdown and/or bringing it back online. When the database was shutdown only the user that initiated the shutdown, either SYSDBA or database owner, can connect to the database and perform other tasks, e.g. metadata modification or database validation and repair. Chapter 7. Working with Services 78 The database shutdown is performed by shutdownDatabase(int, int) method. The first parameter is the shutdown mode, the second – maximum allowed time for operation. There are three shutdown modes: Shutdown mode Description SHUTDOWN_ATTACH The shutdown process is initiated and it is not possible to obtain a new connection to the database but the currently open connection are fully functional. When after the maximum allowed time for operation there are still open connections to the database, the shutdown process is aborted. SHUTDOWN_TRANSACTIONAL The shutdown process is started and it is not possible to start new transactions and open new connections to the database. The transactions that were running at the time of shutdown initiation are fully functional. When after the maximum allowed time for operation there are still running transactions, the shutdown process is aborted. If no running transactions are found, the currently open connections are allowed to disconnect. SHUTDOWN_FORCE The shutdown process is started and will be completed before or when the maximum allowed time for operation is reached. New connections and transactions are not prohibited during the wait. After that any running transaction won't be able to commit. After database was shutdown, the owner of the database or SYSDBA can connect to it and perform maintenance tasks, e.g. migration to the new data model9, validation of the database, changing the database file configuration. To bring system back online use the bringDatabaseOnline() method. Shadow configuration Database shadow is an in-sync copy of the database that is usually stored on a different hard disk, possibly on remote computer10, which can be used as a primary database if the main database server crashes. Shadows can be defined using CREATE SHADOW SQL command and are characterized by a mode parameter: ● in the AUTO mode database continues operating even if shadow becomes unavailable (disk or file system failure, remote node is not accessible, etc.) ● in the MANUAL mode all database operations are halted until the problem is fixed. Usually it means that DBA has to kill the unavailable shadow and define a new one. 9 Until Firebird 2.0 adding a foreign key constraint required an exclusive access to the database. 10 Currently possible only on Unix platforms by using the NFS shares. Chapter 7. Working with Services 79 The MaintenanceManager provides a killUnavailableShadows() method to kill the unavailable shadows. This is equivalent to the gfix -kill command. Additionally, if the main database becomes unavailable, DBA can decide to switch to the shadow database. In this case the shadow must be activated before use. To activate the shadow use the activateShadowFile() method. Please note, that in this case the database property of the MaintenanceManager must point to the shadow file which must be located on the local file system of the server to which the management class is connected. Database validation and repair The Firebird server does its best to keep the database file in a consistent form. In particular this is achieved by a special algorithm called careful writes which guarantees that the server writes data on disk in such a manner than despite the server crash the database file always remains in consistent form. Unfortunately it is still possible that under certain conditions, e.g. crash of the file system or hardware failure, the database file might become corrupted. Firebird server can detect such cases including ● Orphan pages. These are the database pages that were allocated for subsequent write, but due to a crash were not used. Such pages have to be marked as unused to return storage space back to the application; ● Corrupted pages. These are the database pages that were caused by the operating system or hardware failures. The MaintenanceManager class provides a validateDatabase() method to perform simple health check of the database, and releasing the orphan pages if needed. It also reports presence of the checksum errors. The output of the routine is written to the output stream configured in the logger property. The validateDatabase(int) method can be used to customize the validation process: Validation mode Description VALIDATE_READ_ONLY Perform read-only validation. In this case the database file won't be repaired, only the presence of the database file errors will be reported. Can be used for periodical health-check of the database. VALIDATE_FULL Do a full check on record and pages structures, releasing unassigned record fragments. VALIDATE_IGNORE_CHECKSUM Ignore checksums during repair operations. The checksum error means that the database page was overwritten in a random order and the data stored on it are corrupted. When this option is specified, the validation process will succeed even if checksum errors are present. In order to repair the corrupted database use the markCorruptRecords() method which marks the corrupted records as unavailable. This method is equivalent to Chapter 7. Working with Services 80 gfix -mend command. After this operation database can be backed up and restored to a different place. Note, the presence of the checksum errors and subsequent use of markCorruptedRecords() method will mark all corrupted data as unused space. You have to perform careful check after backup/restore cycle to assess the caused damage. Limbo transactions The limbo transactions are transactions that were prepared for commit but were never committed. This can happen when, for example, database was accessed by JTA-enabled applications from Java11. The in-limbo transactions affect the normal database operation, since the records that were modified in that transactions are not available – Firebird does not know whether the new version will be committed or rolled back and blocks access to them. Also the in-limbo transactions prevent the garbage collection, since garbage collector does not know whether it can discard the record versions of the in-limbo transaction. Jaybird contains needed functionality to allow the JTA-enabled transaction coordinator to recover the in-limbo transactions and either commit them or perform a rollback. For the cases when this is not possible MaintenanceManager provides following methods to perform this in interactive mode: Method Description listLimboTransactions() Method lists all IDs of the in-limbo transactions to the output stream specified in logger property. Application has to either parse the output to commit or rollback the transactions in automated mode or it should present the output to the user and let him/her make a decision. commitTransaction(int) Commit the transaction with the specified ID. rollbackTransaction(int) Rollback the transaction with the specified ID. Sweeping the database The in-limbo transactions are not the only kind of transactions that prevent garbage collection. Another type are transactions are those that were finished by “rollback” and the changes made in such transactions were not automatically undone by using internal savepoint mechanism, e.g. when there were a lot of changes made in the transaction (e.g. 10,000 records and more). Such transactions are marked as “rollback” transactions on Transaction Inventory Page and prevent advancing the so-called Oldest Interesting Transaction (OIT) – ID of the oldest transaction which created record versions that are relevant to any of the currently running transactions. On each access to the records, Firebird has to check all the record versions between the current transaction and the OIT, which leads to performance degradation on large databases. In order to solve the issue Firebird periodically starts the database sweeping process, that traverses all database records, removes the changes made by the rolled back transactions and moves forward the OIT12. 11 Another reason for limbo transactions are multidatabase transactions which can be initiated via native Firebird API. However, since Jaybird does not provide methods to initiate them, we do not consider them in this manual. 12 For more information please read article by Ann Harrison “Firebird for the Database Expert: Chapter 7. Working with Services 81 The sweep process is controlled by a threshold parameter – a difference between the Next Transaction and OIT, by default it equal to 20,000. While this value is ok for the average database, DBA might decide to increase or decrease the number to fit the database usage scenario. Alternatively DBA can trigger the sweep process manually regardless of the current difference between Next Transaction and OIT. The MaintenanceManager provides following methods to help with database sweeping: Method Description setSweepThreshold(int) Set the threshold between Next Transaction and OIT that will trigger the automatic sweep process. Default value is 20,000. sweepDatabase() Perform the sweep regardless of the current difference between Next Transaction and OIT. Other database properties There are few other properties of the database that can be set via MaintenanceManager: Method Description setDatabaseAccessMode(int) Change the access mode of the database. Possible values are: ● ACCESS_MODE_READ_ONLY to make database read-only; ● ACCESS_MODE_READ_WRITE to allow writes into the database. Please note, only read-only databases can be placed on read-only media, the read-write databases even if accessed within read-only transactions. setDatabaseDialect(int) Change the database SQL dialect. The allowed values can be either 1 or 3. setDefaultCacheBuffer(int) Change the number of database pages to cache. Default value is 2048. setForcedWrites(boolean) Change the forced writes setting for the database. When the forced writes are switched off, database engine does not enforce flushing pending changes to disk and they are kept in OS cache. When later the RDBMS changes the same page later, the write happens in memory, which in many cases increases the performance. However, in case of OS or hardware crash the database will be corrupted. setPageFill(int) Set the page fill factor. Firebird leaves 20% of free space on each database page for the future record versions. It is possible to tell Firebird not to reserve the space, makes sense for read-only databases, since more data fit the page, which increases the performance. Possible values are: ● PAGE_FILL_FULL – do not reserve additional space for future versions; Episode 4 - OAT, OIT, & Sweep”, available, for example, at http://www.ibphoenix.com/main.nfs?page=ibp_expert4 Chapter 7. Working with Services 82 Method Description ● PAGE_FILL_RESERVE – reserve the free space for future record versions. Database statistics The last but not least are the StatisticsManager interface and corresponding implementation in the FBStatisticsManager class, which allow to obtain the statistical information for the database, like page size, values of OIT and Next transactions, database dialect, database page allocation and its distribution. The following methods provide the functionality equivalent to the gstat command line tool, the output of the commands is written to the output stream specified in the logger property. It is the responsibility of the application to correctly parse the text output if needed. Method Description getDatabaseStatistics() Get complete statistics about the database. getDatabaseStatistics(int) Get the statistical information for the specified options. Possible values are (bit mask, can be combined together): ● DATA_TABLE_STATISTICS ● SYSTEM_TABLE_STATISTICS ● INDEX_STATISTICS ● RECORD_VERSION_STATISTICS getHeaderPage() Get information from the header page (e.g. page size, OIT, OAT and Next transaction values, etc.) getTableStatistics(String[]) Get statistic information for the specified tables. This method allows to limit the reported statistical information to a single or couple of the tables, not for the whole database. Chapter 7. Working with Services 83 Chapter 8. 8. Working with Events Until recently Firebird (InterBase) was, probably, the only RDBMS that provided asynchronous notification of the connected application about some system events. Where applications using other RDBMSs were required to reread the database tables to check for the changes, those using Firebird were able to avoid it – the triggers in the database could post an event in case of change. And even more, the event can be so specific that application would need to reread very limited set of records, possibly only one. This chapter describes the event mechanism in Firebird and the common usage scenarios. Database events The event is a message generated in a trigger or stored procedure that is delivered to subscribed application. The event is characterized only by name which is assigned when event is posted, therefore two different events must have two different names. The applications that subscribe for events are required to specify the event names in which they are interested, no wildcards are allowed; and applications either provide a callback function that will be invoked in case of event or are required to poll for the posted events periodically. The events are delivered to the application only if the transaction that generated the event was committed. Also Firebird does not provide any guarantees about the time of event delivery – it depends on the load of the RDBMS, application load, network delays between application and RDBMS. The database engine will continue operating even if no application subscribe to events or when the subscribed application crashed in the meantime. It can also happen that multiple transactions will be committed before the events are delivered to the client system. But even in such case the callback function will be invoked only once, and only the event name and the count of the events will be passed as parameters. Same applies to periodical polling – application will receive event name and count of the events since last polling. Internally Firebird can be thought to store the subscription information in a table where columns contain event names, rows correspond to the subscribed applications and the cells contain the count of the particular event for a particular application. When event is posted in trigger or stored procedure, Firebird check the subscription information and increases the event count for the subscribed applications. Another thread checks the table periodically and notifies the application about all new events relevant to the particular application. Such mechanism allows Firebird to keep the event notification table very small13 and to reduce the number of messages sent to the application. Note, it is not possible to pass any parameters with the event, e.g. an ID of the modified records. It is also not possible to encode such information in the event names – the wildcards are not supported. For such cases applications should maintain a change tracking table where the IDs of the modified records are stored and the event mechanism is used to tell the application that new records were added to the table. Posting the events The events are posted from the trigger or stored procedure language using the POST_EVENT command. It is possible to create a stored procedure with a sole purpose of posting events (Illustration 8.1). CREATE PROCEDURE sp_post_event(event_name VARCHAR(72)) AS BEGIN POST_EVENT :event_name; END Illustration 8.1.: Example of posting events from PSQL code Firebird 2.0 features new command EXECUTE BLOCK which allows to execute PSQL statements within the DSQL code (Illustration 8.2). Statement stmt = connection.createStatement(); try { stmt.execute( "EXECUTE BLOCK AS BEGIN POST_EVENT 'some_evt'; END"); } finally { stmt.close(); } Illustration 8.2.: Using EXECUTE BLOCK to post events. Subscribing for events 13 For example, the effective size for 100 applications subscribed for 100 different events is about 40k in memory. Chapter 8. Working with Events 86 The design of the classes and interfaces in the org.firebirdsql.event package is similar to the Services API support – there is central manager-class that establishes a database connection and provides service methods to work with the events, a callback interface that applications must implement to use the asynchronous event notification and an interface representing a database event with two properties – event name and occurrence count. Applications have to configure following properties before starting using the implementation EventManager interface: Name Type Description host java.lang.String Name or the IP address of the host to which we subscribe for events. Required. port int Port to which we connect to, 3050 by default. database java.lang.String Path to the database. The path is specified for the remote host but must be absolute. Required. user java.lang.String Name of the user on behalf of which we connect to the database. Required. password java.lang.String Password corresponding to the specified user. Required. After configuring these properties application has to invoke the connect() method to establish a physical connection to the database. At this point the EventManager is ready to receive the event notifications. Now the application developer has two choices: use asynchronous event notification or use the methods that will block until event is delivered or timeout occurs. Asynchronous event notification The asynchronous event notification uses a separate daemon thread to wait for the event notifications and to deliver the events to the registered listeners. The listeners are added using the addEventListener(String, EventListener) method, where the first parameter contains the name of the event to which application registers to and the second parameter – an instance of EventListener interface that will be notified about the events. It is allowed to use the same instance of EventListener interface to listen on different events. The Illustration 8.3 show an example of using the asynchronous event notification. Chapter 8. Working with Events 87 EventManager eventManager = new FBEventManager(); eventManager.setHost("localhost"); eventManager.setUser("SYSDBA"); eventManager.setPassword("masterkey"); eventManager.setDatabase("c:/database/employee.fdb"); eventManager.connect(); eventManager.addEventListener("test_event", new EventListener() { public void eventOccurred(DatabaseEvent event){ System.out.println("Event [" + event.getEventName() + "] occured " + event.getEventCount() + " time(s)"); } } ); Illustration 8.3.: Example of registering an event listener for asynchronous event notification. Using blocking methods Alternatively application can use the synchronous methods, one that blocks until the named event is received – the waitForEvent(String) method, or the one will block until the named event is received or timeout specified in the second parameter occurs – the waitForEvent(String, int) method. The Illustration 8.4 shows an example of using the blocking methods. EventManager eventManager = new FBEventManager(); eventManager.setHost("localhost"); eventManager.setUser("SYSDBA"); eventManager.setPassword("masterkey"); eventManager.setDatabase("c:/database/employee.fdb"); eventManager.connect(); int eventCount = eventManager.waitForEvent("test_event", 10 * 1000); System.out.println( "Received " + eventCount + " event(s) during 10 sec."); Illustration 8.4.: Example of blocking waiting for event with a specified timeout. Chapter 8. Working with Events 88 Appendix A. Extended connection properties Jaybird allows using arbitrary Database Parameters Block entries. The current Firebird API has almost 70 DPB parameters, however only few of them are interesting for regular users. This chapter provides a list of most useful DPB parameters and short explanation to each of them. DPB Parameter and its Alias(es) Explanation isc_dpb_user_name user_name, user Name of the user for the connection. isc_dpb_password password Password corresponding to the specified user. isc_dpb_lc_ctype lc_ctype, encoding Character encoding for the connection. This property tells the database server the encoding in which it expects character content. For list of the available encodings see “Ошибка: источник перекрестной ссылки не найден”. CharSet, localEncoding Character set for the connection. Similar to the previous property, however instead of Firebird- specific name allows using standard Java encoding name. DPB Parameter and its Alias(es) Explanation isc_dpb_num_buffers num_buffers Number of database pages that will be cached. Default values are 75 for ClassicServer and 2048 for SuperServer. isc_dpb_sql_role_name sql_role_name Name of the SQL role for the specified connection. isc_dpb_sql_dialect sql_dialect SQL dialect, by default 3, can be 1 and 2. isc_dpb_set_db_readonly set_db_readonly Set the database into read-only state. isc_dpb_set_db_sql_dialect set_db_sql_dialect Set the SQL dialect of the database. isc_dpb_set_db_charset set_db_charset Set the default character set of the database. isc_dpb_socket_buffer_size socket_buffer_size Jaybird specific parameter. Tells Jaybird Type 4 driver the size of the socket buffer. Should be used on the systems where default socket buffer provided by JVM is not correct isc_dpb_blob_buffer_size blob_buffer_size Jaybird specific parameter. Tells the driver the size of the buffer that is used to transfer BLOB content. It is recommended to keep the value equal to n* (and preferably also socket buffer size). isc_dpb_use_stream_blobs use_stream_blobs Jaybird specific parameter. Tells the driver to create stream BLOBs. By default segmented BLOBs are created. isc_dpb_paranoia_mode paranoia_mode Jaybird specific parameter. Tells the driver to throw exception in the situation not covered by the specification. For example, the JDBC specification does not say whether it is allowed to call the close() method twice. In “paranoia mode” Jaybird will throw an exception while in normal mode it will simply ignore the second call. isc_dpb_use_standard_udf use_standard_udf Jaybird specific parameter. Tells the JDBC driver to assume that standard UDF library is registered in the database when converting escaped function calls. See “Ошибка: источник перекрестной ссылки не найден” for more details. Extended connection properties 90 Appendix B. Data Type Conversion Table Mapping between JDBC, Firebird and Java Types The below table describes a mapping of the JDBC data types defined in java.sql.Types class to the Firebird data types. Also, for each JDBC data type a class instance of which is returned by ResultSet.getObject method is provided. JDBC Type Firebird Type Java Object Type CHARCHAR String VARCHARVARCHAR String LONGVARCHAR BLOB SUB_TYPE 1 String NUMERICNUMERIC java.math.BigDecimal DECIMALDECIMAL java.math.BigDecimal SMALLINTSMALLINT java.lang.Short INTEGER INTEGER java.lang.Integer BIGINTBIGINT java.lang.Long REALREAL java.lang.Float FLOAT FLOAT java.lang.Double DOUBLE DOUBLE PRECISION java.langDouble JDBC Type Firebird Type Java Object Type LONGVARBINARY BLOB SUB_TYPE 0 byte[] DATE DATE java.sql.Date TIME TIME java.sql.Time TIMSTAMP TIMESTAMP java.sql.Timestamp BLOB BLOB SUB_TYPE < 0 java.sql.Blob Data Type Conversions This table specifies the compatible conversions between the Firebird and Java types. S M A L L I N T I N T E G E R B I G I N T R E A L F L O A T D O U B L E D E C I M A L N U M E R I C C H A R V A R C H A R B L O B S U B _ T Y P E 1 B L O B S U B _ T Y P E 0 B L O B S U B T Y P E > 0 D A T E T I M E T I M E S T A M P String x x x x x x x x x x x x x x x x java.math.- BigDecimal x x x x x x x x x x Boolean x x x x x x x x x x Integer x x x x x x x x x x Long x x x x x x x x x x Float x x x x x x x x x x Double x x x x x x x x x x byte[] x x x Blob x x x Date x x Time x Timestamp x x x Data Type Conversion Table 92 Appendix C. Connection Pool Properties This chapter contains the list of properties of the ConnectionPoolDataSource, DataSource and XADataSource interface implementations. Standard JDBC Properties This group contains properties defined in the JDBC specification and should be standard to all connection pools. Property Description maxIdleTime Maximum time in milliseconds after which an idle connection in the pool is closed. maxPoolSize Maximum number of open physical connections. minPoolSize Minimum number of open physical connections. If value is greater than 0, corresponding number of connections will be opened when first connection is obtained. maxStatements Maximum size of the prepared statement pool. If zero, statement pooling is switched off. When the application requests more statements than can be kept in the pool, Jaybird will allow creating those statements, however closing them would not return them back to the pool, but rather immediately release the resources. Pool Properties This group of properties are specific to the Jaybird implementation of the connection pooling classes. Property Description blockingTimeout Maximum time in milliseconds during which application can be blocked waiting for a connection from the pool. If no free connection can be obtained, an exception is thrown. retryInterval Period in which the pool will try to obtain a new connection while blocking the application. pooling Allows to switch connection pooling off. statementPooling Allows to switch statement pooling off. pingStatement Statement that will be used to “ping” the JDBC connection, in other words, to check if it is still alive. This statement must always succeed. The default SQL statement for the Firebird database is "SELECT CAST(1 AS INTEGER) FROM rdb$database". pingInterval Time during which connection is believed to be valid in any case. The pool “pings” the connection before giving it to the application only if more than specified amount of time passed since last “ping”. isolation Default transaction isolation level. All connections returned from the pool will have this isolation level. One of: • TRANSACTION_READ_COMMITTED • TRANSACTION_REPEATABLE_READ • TRANSACTION_SERIALIZABLE transactionIsolat ionLevel Integer value from java.sql.Connection interface corresponding to the transaction isolation level specified in isolation property. Runtime Pool Properties This group contains read-only properties that provide information about the state of the pool. Property Description freeSize Tells how many free connections are in the pool. Value is between 0 and totalSize. workingSize Tells how many connections were taken from the pool and are currently used in the application. totalSize Total size of open connection. At the pool creation – 0, after obtaining first connection – between minPoolSize and maxPoolSize. Firebird-specific Properties This group contains properties that specify parameters of the connections that are obtained from this data source. Commonly used parameters have the corresponding getter and setter methods, rest of the Database Parameters Block parameters can be set using setNonStandardProperty setter method. Connection Pool Properties 94 Property Description database Path to the database in the format [host/port:]/path/to/database.fdb type Type of the driver to use. Possible values are: • PURE_JAVA or TYPE4 for type 4 JDBC driver • NATIVE or TYPE2 for type 2 JDBC driver • EMBEDDED for using embedded version of the Firebird. • ORACLE for accessing Oracle-mode Firebird. blobBufferSize Size of the buffer used to transfer BLOB content. Maximum value is 64k-1. socketBufferSize Size of the socket buffer. Needed on some Linux machines to fix performance degradation. charSet Character set for the connection. Similar to encoding property, but accepts Java names instead of Firebird ones. encoding Character encoding for the connection. See Firebird documentation for more information. userName Name of the user that will be used by default. password Corresponding password. roleName SQL role to use. tpbMapping TPB mapping for different transaction isolation modes. Non-standard parameters Many of the above mentioned Firebird parameters have have a corresponding DPB entry. However, not every DPB entry has a corresponding getter/setter. This was done intentionally, Firebird provides almost 70 DPB parameters, but for most of the applications only few are needed. The remaining parameters are used by specialized applications (usually server or database management software) for setting some default values of the database, controlling the cache buffers on the server, etc. Creating a corresponding getter/setter for each of them simply does not make sense. For those Java applications that still need non-standard connectivity parameters, DataSource and ConnectionPoolDataSource implementations provides a getter and two setters: • getNonStandardProperty(String name) method returns a non-standard property specified by name parameter. If this property was not previously set, null is returned. • setNonStandardProperty(String name, String value) method sets the property specified by the first parameter to a value contained in the second parameter. • setNonStandardProperty(String nameValuePair) method provides a possibility to set a DPB parameter using following syntax: dataSource.setNonStandardProperty("isc_dpb_sql_dialect=3"); The parameter syntax of the last method is not very common in Java code, it would be much more natural to use two-parameter setter. However, it has a Connection Pool Properties 95 specialized use, because there's no possibility to use two-parameter setter method in configuration files. Usually, when setting a configuration parameter of a data source, web-containers use the Java reflection API and consider only those setters that take one parameter. For instance, in the Tomcat server the configuration parameter would look like this: nonStandardProperty sql_dialect=3 Syntax of the parameter is [][{=|:|}[]] where is the name of the DPB parameter, and is its value. The two are separated by any combination of whitespace and either whitespace or “=” (equal sign) or “:” (colon) characters. Considering the aliases described in Ошибка: источник перекрестной ссылки не найден For example following values are equivalent: isc_dpb_sql_dialect 3 isc_dpb_sql_dialect : 3 sql_dialect : 3 sql_dialect=3 Connection Pool Properties 96 Appendix D. Character Encodings Character encodings and their correct use with Firebird RDBMS from Java is an important topic, that initially seems to be complex, but in fact can be formulated by just a few rules. This appendix provides information on this topic. Encodings Types Firebird uses character encodings in two different areas: ● The database encoding defines a character set in which all CHAR, VARCHAR and BLOB SUB_TYPE 1 fields are physically stored on the disk. There is default database encoding that is specified during database creation. It is also possible to specify character sets on per column basis. ● The client connection encoding defines a character set in which client will send and expects to receive character data. This encoding might or might not match the database default encoding. The Firebird RDBMS performs translation between character sets of the client connection and the character set of the content. The list of allowed character sets as well as the allowed translations between them are specified in the fbintl shared library14 located in the intl/ directory of the Firebird installation. There is also a special character set NONE that tells RDBMS not to interpret the contents of the character field. 14 On Windows this library is represented by fbintl.dll, on Linux – libfbintl.so. Firebird uses following algorithm when performing translations: ● If source and target character sets match, send the content unchanged. ● If the target character set is “NONE”, send source content unchanged. ● If the source character set is “NONE”, send source content unchanged. ● If there is a direct translation rule between source and target character sets, use that rule. ● If there is no direct translation rule, check if there is rule to translate the source character set into the UTF8 character set and a rule to translate from UTF8 into the target character set. If yes, use these two rules for translation. ● If no suitable translation rule can be found, throw an exception. Encodings in Java Java programming language is based on the Unicode character set and uses the UTF-16 encoding, in which each character is represented by one or two 16-bit units. The Firebird, on the other side, is not based on Unicode and allows different character sets to be assigned to different database objects. Additionally, Firebird requires a connection character set to be specified in connection options, which forces Firebird to convert data from the character set of the database object into the character set of the client application. There are two boundary cases that we will consider here, one when Firebird database was created with default15 character set UTF8, another when Firebird database was created without specifying the character set (i.e. character set NONE). The UTF8 in Firebird 2.0 is a Unicode character set that uses UTF-8 encoding and occupies from one to four 8-bit units. Firebird supported Unicode character set for a long time, however its implementation was deficient – it did not support proper uppercasing and correct sorting. These issues were addressed in the Firebird 2.0 release and at the moment nothing prevents developers from using Unicode in the database and on the client side, which greatly simplifies the internationalization and localization of the applications. Jaybird JDBC driver properly supports Firebird 2.0 since version 2.1.016. The UTF8 character set Software developer must ensure two things to enable use of Unicode characters in the database and the application: ● the database objects must be defined with the UTF8 character set; this can be done by either creating database with default UTF8 character set or by adding CHARACTER SET UTF8 clause to the column or domain definitions. ● the encoding connection property in the JDBC driver has to be set to UTF8; this can be done several ways: the he easiest one is to add the appropriate parameter to the JDBC URL (Illustration D.1). Another 15 The default character set simplifies the explanation, since we do not have to consider the cases when different columns with different character sets are used within the same connection. The statements made here, obviously, can be applied to those cases as well. 16 Jaybird provided character set conversion from the very beginning, however Jaybird versions prior to 2.1.0 know nothing about improved Unicode support in Firebird 2.0 because Jaybird 2.0.0 was released almost a year before Firebird 2.0 was released. Character Encodings 98 possibility is to use appropriate method of the DriverManager class (Illustration D.2). The applications that use DataSource interface to obtain the database connections also have access to the encoding property17. Class.forName("org.firebirdsql.jdbc.FBDriver"); Connection connection = DriverManager.getConnection( "jdbc:firebirdsql:localhost/3050:employee?encoding=UTF8", "SYSDBA", "masterkey"); Illustration D.1.: Specifying the connection encoding in JDBC URL. Class.forName("org.firebirdsql.jdbc.FBDriver"); Properties props = new Properties(); props.setProperty("user", "SYSDBA"); props.setProperty("password", "masterkey"); props.setProperty("encoding", "UTF8"); Connection connection = DriverManager.getConnection( "jdbc:firebirdsql:localhost/3050:employee", props); Illustration D.2.: Specifying connection encoding in the connection properties. There are few limitations related to using the UTF8 character set: ● it is not possible to create Unicode columns longer than 8191 Unicode characters; this limitation is caused by the fact that the longest possible VARCHAR column can occupy 32765 bytes (32767 for CHAR columns) and single UTF8 character can occupy maximum four bytes. ● it is not possible to index Unicode columns longer than 1023 characters; this limitation is caused by the fact that the longest index key cannot be longer than a quarter of the database page, which in Firebird 2.0 can be maximum 16k and the before mentioned fact that each UTF8 character can occupy up to four bytes. Also it should be mentioned, that using Unicode character set might cause noticeable performance degradation when database is used over the wide-area networks. This mainly applies to the cases when Cyrillic characters are stored in the database, since each character would occupy two bytes of the network packet, which in turn might cause additional roudtrips to the server to fetch data. The NONE character set Java introduces additional complexity when the NONE character set is used. The reason for this is that Java internally stores all strings in Unicode format, and the application must define the character encoding for the byte contents to the JVM. When the NONE character set is used, the driver does not know how to interpret the received data. The only choice that is left to the driver is to construct a string using the default character set of the JVM, which usually matches the regional settings of the operating system and can be accessed from within the JVM through the file.encoding system property. 17 See http://jaybirdwiki.firebirdsql.org/... for configuration examples of the most popular application servers. Character Encodings 99 It is clear that the conversion using default character set that happens inside the JVM can lead to errors when the same content is accessed from two or more different Java Virtual Machines that have different configuration. One application running on the computer with, for example, Russian regional settings saves the Russian text (the default character set of the JVM is Cp1251) and another application running on computer with German regional settings (default character set is Cp1252) will read in such case some special or accented characters. However, when all client applications run same OS with the same regional settings in most cases will not have any severe consequences (except probably wrong sorting order or uppercasing on the server side). On Linux and other Unix platforms it might have more severe consequences as it is very common that regional settings are not configured and that the default “C” locale is used and the non-ASCII characters will be replaced with question marks (“?”). Therefore, application should use NONE character encoding as an encoding for a database and a connection only when at least one of the following is met: ● Database will contain only ASCII characters ● It is guaranteed that all Java Virtual Machines accessing the database will have the same default encoding that can correctly handle all characters stored in the database. As a partial workaround one can specify the encoding that should be used to interpret bytes coming from the server in the charSet connection property. The following rules are used when interpreting the encoding and charSet properties: ● When only encoding property specified, driver uses the default mapping between server and Java encodings. When encoding property is not set or set to NONE and charSet property is not set, the default JVM encoding is used to interpret bytes coming from the server. ● When only charSet property is specified, driver uses the reverse mapping to specify the connection encoding for the server and interprets byte stream according to the value of the property. ● When both encoding and charSet property are specified, driver sets the connection encoding according to the value of the encoding property, but interprets the byte stream according to the charSet property. The last case is most powerful, but also is the most dangerous in use. When used properly, it can solve the problems with the legacy databases; when used incorrectly, one can easily trash the content of the database. Available Encodings The below table lists the available character encodings in the default Firebird distribution and their mapping to the Java ones: Firebird encoding Java encoding Size in bytes Comments NONE- 1 Raw bytes, no interpretation of the content is possible. ASCIIASCII 1 - Character Encodings 100 Firebird encoding Java encoding Size in bytes Comments BIG_5 Big5 2 Traditional Chinese DOS437 Cp437 1 MS-DOS: United States, Australia, New Zealand, South Africa DOS737 Cp737 1 MS-DOS: Greek DOS775 Cp775 1 MS-DOS: Baltic DOS850 Cp850 1 MS-DOS: Latin-1 DOS852 Cp852 1 MS-DOS: Latin-2 DOS857 Cp857 1 IBM: Turkish DOS858 Cp858 1 IBM: Latin-1 + Euro DOS860 Cp860 1 MS-DOS: Portuguese DOS861 Cp861 1 MS-DOS: Icelandic DOS862 Cp862 1 IBM: Hebrew DOS863 Cp863 1 MS-DOS: Canadian French DOS864 Cp864 1 IBM: Arabic DOS865 Cp865 1 MS-DOS: Nordic DOS866 Cp866 1 IBM: Cyrillic DOS869 Cp869 1 IBM: Modern Greek EUCJ_0208 EUC_JP 2 JIS X 0201, 0208, 0212, EUC encoding, Japanese GB_2312 EUC_CN 2 GB2312, EUC encoding, Simplified Chinese ISO8859_1 ISO-8859-1 1 ISO 8859-1, Latin alphabet No. 1 ISO8859_2 ISO-8859-2 1 ISO 8859-2 ISO8859_3 ISO-8859-3 1 ISO 8859-3 ISO8859_4 ISO-8859-4 1 ISO 8859-4 ISO8859_5 ISO-8859-5 1 ISO 8859-5 ISO8859_6 ISO-8859-6 1 ISO 8859-6 ISO8859_7 ISO-8859-7 1 ISO 8859-7 ISO8859_8 ISO-8859-8 1 ISO 8859-8 ISO8859_9 ISO-8859-9 1 ISO 8859-9 ISO8859_13 ISO-8859-13 1 ISO 8859-13 KSC_5601 MS949 2 Windows Korean UNICODE_FSS UTF-8 3 8-bit Unicode Transformation Format. Deprecated in FB 2.0. WIN1250 Cp1250 1 Windows Eastern European WIN1251 Cp1251 1 Windows Cyrillic WIN1252 Cp1252 1 Windows Latin-1 WIN1253 Cp1253 1 Windows Greek WIN1254 Cp1254 1 Windows Turkish Character Encodings 101 Firebird encoding Java encoding Size in bytes Comments WIN1255 Cp1255 1 - WIN1256 Cp1256 1 - WIN1257 Cp1257 1 - UTF8 UTF-8 4 8-bit Unicode Transformation Format. Recommended since Firebird 2.0. Character Encodings 102 Appendix E. Supported JDBC Scalar Functions The JDBC API supports the escaped syntax for numeric, string, time, date, system and conversion functions. Jaybird will try to provide an equivalent of the JDBC function using the built-in capabilities of the Firebird database. When no equivalent is available, Jaybird will pass the function call “as is” to the database assuming that it contains all necessary UDF declarations . Not all functions described in the JDBC specification have corresponding built-in functions in Firebird, but are available in the standard UDF library ib_udf18 shipped with Firebird. Jaybird provides a connection parameter use_standard_udf telling the driver to assume that functions from that UDF are available in the database. In this case Jaybird will convert all JDBC function calls into the corresponding calls of the UDF functions. Below you will find the list of JDBC functions and whether they have a corresponding equivalent in the “built-in” and in the “UDF” modes. 18 On Windows platform it is represented by the ib_udf.dll, on Linux it is represented by the libib_udf.so. Numeric Functions JDBC Jaybird built-in UDF mode Description ABS(number) m l Absolute value of number. ACOS(float) m l Arccosine, in radians, of float. ASIN(float) m l Arcsine, in radians, of float. ATAN(float) m l Arctangent, in radians, of float. ATAN2(float1, float2) m l Arctangent, in radians, of float2 / float1. CEILING(number) m l Smallest integer >= number. COS(float) m l Cosine of float radians. COT(float) m l Cotangent of float radians. DEGREES(number) m m Degrees in number radians. EXP(float) m m Exponential function of float. FLOOR(number) m l Largest integer <= number. LOG(float) m l Base e logarithm of float. LOG10(float) m l Base 10 logarithm of float. MOD(integer1, integer2) m l Remainder for integer1 / integer2. PI() m l The constant pi. POWER(number, power) m m number raised to (integer) power. RADIANS(number) m m Radians in number degrees. RAND(integer) m RAND() 19 Random floating point for seed integer. ROUND(number, places) m m number rounded to places places. SIGN(number) m m -1 to indicate number is < 0; 0 to indicate number is = 0; 1 to indicate number is > 0. SIN(float) m l Sine of float radians. SQRT(float) m l Square root of float. TAN(float) m l Square root of float. TRUNCATE(number, places) m m number truncated to places places. 19 Standard UDF library provides RAND() function taking no parameters. The random number generator is seeded by the current time. There is no function where the seed can be specified. Supported JDBC Scalar Functions 104 Legend: m – not available in this mode; l – available in this mode. String Functions JDBC Jaybird built-in UDF mode Description ASCII(string) m l Integer representing the ASCII code value of the leftmost character in string. CHAR(code) m l Character with ASCII code value code, where code is between 0 and 255. CONCAT(strin1 g, string2) l l Character string formed by appending string2 to string1; if a string is null, the result is DBMS- dependent. DIFFERENCE(st ring1, string2) m m Integer indicating the difference between the values returned by the function SOUNDEX for string1 and string2. INSERT(string 1, start, length, string2) m m A character string formed by deleting length characters from string1 beginning at start, and inserting string2 into string1 at start. LCASE(string) m l Converts all uppercase characters in string to lowercase. LEFT(string, count) l l The count leftmost characters from string. LENGTH(string ) m l 20 Number of characters in string, excluding trailing blanks. LOCATE(string 1, string2[, start]) m m Position in string2 of the first occurrence of string1, searching from the beginning of string2; if start is specified, the search begins from position start. 0 is returned if string2 does not contain string1. Position 1 is the first character in string2. LTRIM(string) m l Characters of string with leading blank spaces removed. Legend: m – not available in this mode; l – available in this mode. String Functions (continued) JDBC Jaybird built-in UDF mode Description REPEAT(string , count) m m A character string formed by repeating string count times. 20 The trailing blanks are also counted. Supported JDBC Scalar Functions 105 JDBC Jaybird built-in UDF mode Description REPLACE(strin g1, string2, string3) m m Replaces all occurrences of string2 in string1 with string3. RIGHT(string, count) m m The count rightmost characters in string. RTRIM(string) m l The characters of string with no trailing blanks. SOUNDEX(strin g) m m A character string, which is data source-dependent, representing the sound of the words in string; this could be a four-digit SOUNDEX code, a phonetic representation of each word, etc. SPACE(count) m m A character string consisting of count spaces. SUBSTRING(str ing, start, length) l l A character string formed by extracting length characters from string beginning at start. UCASE(string) l l Converts all lowercase characters in string to uppercase. Legend: m – not available in this mode; l – available in this mode. Supported JDBC Scalar Functions 106 Time and Date Functions JDBC Jaybird built-in UDF mode Description CURDATE() l l The current date as a date value. CURTIME() l l The current local time as a time value. DAYNAME(date) m m A character string representing the day component of date; the name for the day is specific to the data source. DAYOFMONTH(da te) l l An integer from 1 to 31 representing the day of the month in date. DAYOFWEEK(dat e) m m An integer from 1 to 7 representing the day of the week in date; 1 represents Sunday. DAYOFYEAR(dat e) m m An integer from 1 to 366 representing the day of the year in date. HOUR(time) l l An integer from 0 to 23 representing the hour component of time. MINUTE(time) l l An integer from 0 to 59 representing the minute component of time. MONTH(date) l l An integer from 1 to 12 representing the month component of date. MONTHNAME(dat e) m m A character string representing the month component of date; the name for the month is specific to the data source. NOW() l l A timestamp value representing the current date and time. QUARTER(date) m m An integer from 1 to 4 representing the quarter in date; 1 represents January 1 through March 31. Legend: m – not available in this mode; l – available in this mode. Time and Date Functions (continued) JDBC Jaybird built-in UDF mode Description SECOND(time) l l An integer from 0 to 59 representing the second component of time. TIMESTAMPADD( interval, count, timestamp) m m A timestamp calculated by adding count number of interval(s) to timestamp. TIMESTAMPDIFF m m An integer representing the number Supported JDBC Scalar Functions 107 JDBC Jaybird built-in UDF mode Description (interval, timestamp1, timestamp2) of interval by which timestamp2 is greater than timestamp1. WEEK(date) l l An integer from 1 to 53 representing the week of the year in date. YEAR(date) l l An integer representing the year component of date. Legend: m – not available in this mode; l – available in this mode. System Functions JDBC Jaybird built-in UDF mode Description DATABASE() m m Name of the database. IFNULL(expres sion, value) l l value if expression is null; expression if expression is not null. USER() m m User name in the DBMS. Legend: m – not available in this mode; l – available in this mode. Conversion Functions JDBC Jaybird built-in UDF mode Description CONVERT(value , SQL type) l l value converted to SQL type where SQL type may be one of the following SQL types: • BIGINT • CHAR • DATE • DECIMAL • DOUBLE PRECISION • FLOAT • INTEGER • REAL • SMALLINT • TIME • TIMESTAMP • VARCHAR Legend: m – not available in this mode; l – available in this mode. Supported JDBC Scalar Functions 108
还剩102页未读

继续阅读

下载pdf到电脑,查找使用更方便

pdf的实际排版效果,会与网站的显示效果略有不同!!

需要 5 金币 [ 分享pdf获得金币 ] 0 人已下载

下载pdf

pdf贡献者

bnpysse

贡献于2014-02-15

下载需要 5 金币 [金币充值 ]
亲,您也可以通过 分享原创pdf 来获得金币奖励!