The London Perl and Raku Workshop takes place on 26th Oct 2024. If your company depends on Perl, please consider sponsoring and/or attending.


DBD::JDBC - JDBC proxy driver for the DBI module


    use DBI;
    $dbh = DBI->connect("dbi:JDBC:hostname=$hostname;port=$port;url=$url",
                        $user, $password);

    # See the DBI module documentation.


  Perl 5.8.6 or higher
  DBI 1.48 or higher
  Convert::BER 1.31
  Java Virtual Machine compatible with JDK 1.4
  A JDBC driver
  log4j 1.2.13


DBD::JDBC is a Perl module which works in conjunction with a server written in Java to provide a DBI front end to a JDBC driver. The Perl module and Java server may be installed on different machines, as long as socket connections are allowed. The Java server portion is multi-threaded and supports multiple simultaneous connections.

This driver currently supports JDBC drivers which implement the JDBC 1.22 interface. JDBC 2.0-compatible drivers are expected to work, but no JDBC 2.0 functionality is explicitly exposed via DBD::JDBC. The $h->jdbc_func method exposes additional JDBC and driver-specific methods. Only Java methods with primitive or String parameters and return types are currently supported in this way.

The expected use for this module is as a DBI interface to databases with JDBC drivers but no DBI drivers. The implementation of this module was originally done for a non-SQL database in order to take advantage of the existing SQL parser in the database's JDBC driver.

The Java classes provided with this module also allow a Java application or servlet to create a JDBC connection and then execute a Perl script which can use that pre-existing JDBC connection via DBI. This particular functionality was implemented in order to allow Perl customizations to a Java servlet-based application. See the example in the example/ directory.


Before using DBD::JDBC, you must start the DBD::JDBC server.

Starting the server

The DBD::JDBC server is a Java application intended to be run from the command line. It may be installed, along with whatever JDBC driver you wish to use, on any host capable of accessing the database you wish to use via JDBC. Perl applications using DBD::JDBC will open a socket connection to this server. You will need to know the hostname and port where this server is running.

To start the server,

  1. Place the dbd_jdbc.jar file, log4j-1.2.13.jar, and your database's JDBC driver on the machine where you wish to run the server.

  2. Add dbd_jdbc.jar, log4j-1.2.13.jar, a file, and your database's JDBC driver to your classpath. Follow any other instructions which came with your JDBC driver. For example, a type 2 JDBC driver may require that the database's native libraries be added to your path or library path.

  3. Start the server, providing at least the required system properties on the command line:

    jdbc.drivers (required)

    This should be set to the complete class name of your JDBC driver. If you want to use more than one JDBC driver, use a colon-separated list of driver names. See the standard Javadoc documentation for java.sql.DriverManager for an example.

    dbd.port (required)

    This is the port to which this server will listen. Your Perl client applications will need to know this in order to connect.


  java -Ddbd.port=12345 com.vizdom.dbd.jdbc.Server

Here is a simple example shell script for running the server (written for bash).

  export CLASSPATH=dbd_jdbc.jar:log4j-1.2.13.jar:.:/oracle/jdbc/$CLASSPATH
  java -Djdbc.drivers=$DRIVERS -Ddbd.port=9001 com.vizdom.dbd.jdbc.Server

Setting up log4j

As of DBD:JDBC 0.70, the server uses log4j for logging. log4j is a Java logging implementation from the Apache Logging project. More information is available at

A log4j jar file and sample configuration file are included with DBD::JDBC. The log4j jar file is in the t/hsqldb directory since it's also used by hsqldb. The properties file,, is in the unzipped module directory. The sample file has all logging disabled. To enable logging, change the value of OFF in the following line = OFF

to one of the following values: FATAL, ERROR, WARN, INFO, DEBUG, TRACE, ALL. Log messages will be written to stdout. You can configure log4j to log to files; see the log4j documentation for more information on how to configure log4j.

Here's an example command line which starts the server from the unzipped module distribution (no JDBC driver is included in the classpath). Notice that "." is in the classpath in order to locate

java -Ddbd.port=9001 -classpath 'dbd_jdbc.jar;t/hsqldb/log4j-1.2.13.jar;.' com.vizdom.dbd.jdbc.Server

Connecting to the server

A dsn for DBD::JDBC has the following form:



  • $host is the hostname on which the DBD::JDBC server is running (optionally followed by :$port; if so, the port name/value pair is optional).

  • $port is the port on which the DBD::JDBC server is running.

  • $url is a complete JDBC url for your JDBC driver. You might want to test this URL in a Java application (in the same environment in which you intend to run the DBD::JDBC server) before attempting to connect from Perl. Your JDBC url may need to include your database host and port information; these values are distinct from those needed in the DBD::JDBC dsn, which are for the DBD::JDBC server, not the database.

    If the JDBC url contains the characters ';' or '=', those characters must be URL-encoded. For example,

        $url =~ s/([=;])/uc sprintf("%%%02x",ord($1))/eg;

    The driver will unescape the url before sending it to the JDBC driver. [See the driver code if you really want to know why.]

  • $charset is the character set used by your DBI application (i.e. the character set in use on whichever machine is running Perl, not the machine running the DBD::JDBC server, unless they're the same). This should be specified in the form of a valid Java character encoding name. If no character set is given, the driver defaults to ISO8859_1. See for a list of supported encodings. The character set name must be encoded in ASCII so that the server can appropriately decode it.

    Any strings sent to the DBD::JDBC server as parameter values will be converted from this encoding to Unicode (Java's native encoding). Use bind_param and a type hint to avoid character set conversion of binary data. Data being returned as strings (everything other than binary columns) will be converted to this encoding.

    The character set in use will be available as the database handle attribute 'jdbc_character_set'. Changing this attribute will have no effect on character conversion; the character set is established at connection time.


   $dsn = "dbi:JDBC:hostname=myhost;port=12345;url=jdbc:odbc:mydatasource";
   $dsn = "dbi:JDBC:hostname=myhost:12345;url=jdbc:oracle:thin:\@mydbhost:1521:test;jdbc_character_set=ASCII";

Specifying JDBC connection properties

You can specify JDBC connection properties in the JDBC URL. You can also specify connection properties as follows:

   %properties = ('user' => 'user', 
                  'password' => 'password', 
                  '' => 'dbhost', 
                  'host.port' => '7000');

   $dsn = "dbi:JDBC:hostname=host:9001;url=jdbc:opentext:db:";
   $dbh = DBI->connect($dsn, undef, undef, 
          { PrintError => 0, RaiseError => 1, jdbc_properties => \%properties })
          or die "Failed to connect: ($DBI::err) $DBI::errstr\n";

When specifying properties this way, use undef for the user and password parameters in the DBI->connect method. The DBD::JDBC server will use the JDBC method DriverManager.getConnection(String, Properties) when the username and password parameters are undefined. Otherwise, it will use the JDBC method DriverManager.getConnection(String, String, String) and ignore any other connection properties.


Reading long data

As of DBD::JDBC 0.70, the default behavior of this driver is to read long fields in their entirety. This behavior is controlled by the driver attribute jdbc_longreadall. When jdbc_longreadall is true, the DBI attributes LongReadLen and LongTruncOk will be ignored and the entire contents of long fields will be returned. To return to the default DBI behavior, set jdbc_longreadall to false.

Calling JDBC methods

JDBC methods are exposed using the $h->jdbc_func method and Java reflection. This feature is not intended to replace any existing DBI methods, merely to provide access to JDBC-specific methods with no DBI equivalent. Reflection, rather than explicit methods corresponding to methods in the JDBC API, is used in order to allow access to driver-specific methods not in the JDBC API.

The general syntax is

  $h->jdbc_func(parameter, ..., <jdbc_method_name>);

For example,

  $ret = $dbh->jdbc_func("getAutoCommit");
  $ret = $sth->jdbc_func("mycursor", "Statement.setCursorName");

The driver-specific method jdbc_func replaces the previous use of the standard DBI method func for calling JDBC methods. Since jdbc_func is a driver-specific method, the jdbc_ prefix previously required on the method name argument is no longer required. Use of the jdbc_ prefix on the method name is still supported.

The following limitations apply:

  • Only public methods can be called.

  • In general, only methods whose parameters and return values are primitive types (int, boolean, void, etc.) or Strings can be called. See below for more details.

  • DBD::JDBC doesn't know which methods are being called when reflection is used. This means that interleaving DBI methods and calls to JDBC methods using jdbc_func may leave DBD::JDBC in an inconsistent state. One example of this is $sth->rows: if you've called directly, rather than using $sth->fetch, the row count will not accurately reflect the rows in the result set.

  • Methods can only be called on the Connection, Statement, ResultSet, and ResultSetMetaData objects.

The JDBC method name is used as the jdbc_func method name. Parameters are passed as strings by default. To pass parameters of other types, pass the parameter as reference to an array in which the first element is the parameter and the second is one of the DBI SQL_XXX typecodes. For example,

    $h->jdbc_func("string parameter", [11 => SQL_INTEGER], 
             [1 => SQL_BIT], "method_name");

It is very important to use the correct typecodes for the actual parameter types of the Java method in order to enable Java reflection to locate the correct method. The method will be looked up using the java.lang.class.getMethod(String, Class[]) method, so if the parameter types don't match the actual method parameters, the method won't be found.

SQL types are mapped to Java types by mapping the DBI constants to values from java.sql.Types, then mapping the java.sql.Types values to Java types.

    DBI constant           java.sql.Types constant    Java type
    SQL_CHAR               CHAR                       String
    SQL_VARCHAR            VARCHAR                    String
    SQL_LONGVARCHAR        LONGVARCHAR                String
    SQL_BINARY             BINARY                     byte[]
    SQL_VARBINARY          VARBINARY                  byte[]
    SQL_LONGVARBINARY      LONGVARBINARY              byte[]
    SQL_BIT                BIT                        boolean
    SQL_TINYINT            TINYINT                    byte
    SQL_SMALLINT           SMALLINT                   short
    SQL_INTEGER            INTEGER                    int
    DBD::JDBC::SQL_BIGINT  BIGINT                     long
    SQL_REAL               REAL                       float
    SQL_FLOAT              FLOAT                      double
    SQL_DOUBLE             DOUBLE                     double
    SQL_NUMERIC            NUMERIC                    java.math.BigDecimal
    SQL_DECIMAL            DECIMAL                    java.math.BigDecimal
    SQL_DATE               DATE                       java.sql.Date
    SQL_TIME               TIME                       java.sql.Time
    SQL_TIMESTAMP          TIMESTAMP                  java.sql.Timestamp

The mapping from java.sql.Types to Java types is taken from Table 21.6.1 in JDBC Data Access with Java, by Hamilton, Cattell, and Fisher. See also Table 47.9.1 in JDBC API Tutorial and Reference, Second Edition, by White, Fisher, Cattell, Hamilton, and Hapner.

For SQL_DATE, SQL_TIME, and SQL_TIMESTAMP parameters, the default JDBC string representations for these types must be used.

    SQL_DATE: yyyy-mm-dd
    SQL_TIME: hh:mm:ss
    SQL_TIMESTAMP: yyyy-mm-dd hh:mm:ss.f

(The .f portion of the timestamp format is in nanoseconds and is optional.)

Possible return values from $h->jdbc_func are undef if the Java method returned null or had a void return type, 1 or 0 if the Java method had a boolean return type, or a scalar for any other return type (the Object returned by java.lang.reflect.Method.invoke will be converted to a String by calling its toString method).

You are not limited to calling methods defined by the JDBC API. Any public method defined by your JDBC driver on the available objects, with parameters and return type as described above, may be called.

jdbc_func is made available using DBI's install_method method. This means that errors are handled in the standard DBI manner, not the way they're handled for $h->func.

Connection methods

To call JDBC methods on the JDBC Connection object, use the jdbc_func method on the $dbh handle.


    $ret = $dbh->jdbc_func("jdbc_getAutoCommit");
    $dbh->jdbc_func([1 => SQL_BIT], "jdbc_setAutoCommit");
    $ret = $dbh->jdbc_func("select * from client", "jdbc_nativeSQL");
    $dbh->jdbc_func([4 => SQL_INTEGER], "jdbc_setTransactionIsolation");

Statement, ResultSet, and ResultSetMetaData methods

To call JDBC methods on the JDBC Statement, ResultSet, and ResultSetMetaData objects, use the jdbc_func method on the $sth handle and prefix the method name with one of the listed interface names. You may use either Statement or PreparedStatement to indicate the current PreparedStatement object, since DBD::JDBC uses PreparedStatements internally.

ResultSet and ResultSetMetaData methods are not available until after $sth->execute has been called.


    $ret = $sth->jdbc_func("jdbc_Statement.getMaxFieldSize");
    $sth1->jdbc_func("mycursor", "jdbc_Statement.setCursorName");
    $sth1->jdbc_func([22 => SQL_INTEGER], "jdbc_Statement.setMaxFieldSize");

    $ret = $sth1->jdbc_func("");
    $ret = $sth1->jdbc_func("cname", "jdbc_ResultSet.getString");
    $ret = $sth2->jdbc_func("eno", [5003 => SQL_INTEGER], "jdbc_ResultSet.updateInt");

    $ret = $sth1->jdbc_func([1 => SQL_INTEGER], "jdbc_ResultSetMetaData.getSchemaName");


If for some reason you reach the end of a ResultSet using $sth->jdbc_func("") rather than one of the standard DBI methods (fetch, etc.), the DBI statement handle will continue to think that it's active. You must call $sth->finish explicitly in this case.

Be aware of which JDBC methods are called by the standard DBI methods. For example, $sth->fetch calls next and reads all the columns in the current row. With some JDBC drivers, you will not be able to call $sth->fetch followed by $sth->jdbc_func("column_name", "ResultSet.getString") because all the data for the row has already been read.

If you are using a JDBC driver with scrollable result sets, please note that support for such is provided purely through jdbc_func, not through any explicit DBD::JDBC support. This means that a loop over the set, such as

    while ($row = $sth->fetch()) { 
        # do something

will cause DBD::JDBC to mark the statement handle as inactive at the end of the loop ($sth->{Active} will be false). You can still use jdbc_func to operate on the underlying ResultSet, but you can't continue to use any DBI method which requires that the statement handle be active. The following sequence seems to work, though perhaps it shouldn't:

    while ($row = $sth->fetch()) { 
        # do something
    while ($row = $sth->fetch()) { 
        # do something else

Some sort of explicit support for scrollable result sets will probably be implemented at a later date.

Closing cursors

When a statement handle goes out of scope, Perl will call its DESTROY method. This method will cause Statement.close to be called on the associated Java Statement object in the DBD::JDBC server. For many applications, this is sufficient. However, if you find that statement handles are not being destroyed quickly enough, or you are maintaining a collection of statements for repeated use, you may choose to close the ResultSet associated with the Statement explicitly using jdbc_func. Closing the ResultSet will not prevent you from executing the statement again, but it will release any database resources held by the ResultSet.

Typical usage:

    $sth = $dbh->prepare("select id from sched");
    while ($row = $sth->fetch()) { 
        # do something
    # At this point, the statement handle is no longer active, but
    # the ResultSet still exists on the server.

DBD::JDBC does not close ResultSet objects when $sth->finish is called (whether it is called implicitly when the end of the result set is reached or explicitly in your program) in order to support scrollable result sets. With a scrollable result set, reaching the end of the data does not mean that the ResultSet is unusable, so calling close would be unfortunate.

Character sets

You can find out what character set Java thinks your platform uses by examining the value of the system property file.encoding.

    System.out.println("This system uses: " + System.getProperty("file.encoding"));

Local experimentation (in the US) indicates that Windows NT uses "Cp1252" (Windows Latin-1) and Unix variants (AIX, Solaris) use "ISO8859_1".


When a JDBC exception is thrown in the server, the exception and any exceptions chained to the original are returned and placed in the jdbc_error attribute of the most-recently-used handle. This attribute will contain an array of hashrefs with keys err, errstr, and state. The first error's values will also be available via $h->err, $h->errstr, and $h->state.

    foreach $err (@{$sth->{jdbc_error}}) {
        print "Error: ($err->{err}/$err->{state}) $err->{errstr}\n";

DBI to JDBC method mappings

What follows is a guide to the JDBC methods being called when a DBI method or property is accessed. See the source code for com.vizdom.dbd.jdbc.Connection for details.

        DriverManager.getConnection(url, username, password)
        DriverManager.getConnection(url, properties)

        Connection.setAutoCommit   [if AutoCommit was specified]
        PreparedStatement.close()   [for each open statement]
        PreparedStatement.setXXX(value)   [if there are any parameters]
        [This is called automatically when a statement handle
        goes out of scope]


Statement parameters and column values

Parameter values are sent to the DBD::JDBC server as a sequence of bytes. Numeric parameters are encoded as strings rather than numbers (so 11 is sent as the two characters '1' '1').

When the DBD::JDBC server receives the parameter values, the bytes are converted to a Java String (using the character encoding specified at connection time by the jdbc_character_set value) and the PreparedStatement.setString() method is used to set the parameter value.

If a type hint (one of the SQL_XXX types you can import from the DBI module) is supplied along with a parameter in $sth->bind_param(), the type code will be mapped to the corresponding JDBC type code and passed along to the DBD::JDBC server. The JDBC type will be used to determine which PreparedStatement.setXXX method to call. The mapping from type hint to setXXX method is taken from Table 21.2, p. 394, in JDBC Data Access with Java.

        TINYINT, SMALLINT: setShort
        INTEGER: setInt
        BIGINT: setLong
        REAL: setFloat
        FLOAT, DOUBLE: setDouble
        DECIMAL, NUMERIC: setBigDecimal
        BIT: setBoolean

Type hints are required for binary data in order to avoid having binary parameter values passed through the default character conversion process. In other cases, they are generally optional and may in fact reduce performance by causing unneccessary data conversions. For example, if your database's JDBC driver passes all data to the database as strings, the JDBC driver will have to convert numbers back to strings anyway.

A call to $sth->fetch will cause the DBD::JDBC server to use the column type information from ResultSetMetaData to determine how to retrieve column data.

        Column type         Method used        
        BINARY, VARBINARY   getBytes
        LONGVARBINARY       getBinaryStream
        BLOB                getBlob().getBinaryStream
        CLOB                getClob().getCharacterStream
        LONGVARCHAR         getCharacterStream
        ARRAY               getString
        all others          getString

Once retrieved, the data is returned to Perl as a sequence of bytes. The caller may choose whether to treat the returned scalar as a character string, number, or byte string.

NUM_OF_PARAMS implementation

The statement attribute NUM_OF_PARAMS is set when $dbh->prepare is called. Since JDBC doesn't expose this information about PreparedStatements, DBD::JDBC uses a simple '?' counting method which may fail to provide an accurate count if the parameter marker is not '?' or the syntax does not conform to standard SQL (and possibly even if it does, if I've interpreted the SQL grammar poorly). Depending on this value to be accurate is not recommended, although you may find that it is sufficient for your use.


All errors generated by DBD::JDBC have IJDBC as the SQLSTATE. If a SQLException was thrown by the JDBC driver without a SQLSTATE, DBD::JDBC will set the SQLSTATE to IDRVR.

Errors generated by the Perl driver

Unsupported AutoCommit value (no error number)

If you attempt to set AutoCommit to anything other than 0 or 1, the driver will die with this error.

Error code 100

An error occurred while sending a request to the server.

Error code 101

An error occurred while receiving a response from the server.

Error code 102

There was a problem decoding a server response packet.

Error code 103

The dsn supplied to connect is missing one or more required values.

Error code 104

A connection to the server could not be established. The server may not be running, or the host or port information may be incorrect.

Error code 105

An $sth->execute call caused the server to return an invalid response. This is an internal error.

Errors generated by the Java server

Error code 1

The client requested an operation on a statement object which does not exist on the serer.

Error code 2

fetch was called on a statement which has no data. For example, this error might result if fetch is called before a statement is executed.

Error code 3

The server was asked to return the value of an unknown attribute.

Error code 4

This error code indicates that the client attempted to do something which requires a cursor (a ResultSet) on the server, but no cursor is present.

Error code 5

No metadata is currently available. This error will result if a request is made for a statement attribute at a time when no ResultSet is associated with the statement.

Error code 6

This error code indicates that the client sent a message to the server which the server does not understand.

Error code 7

The server was unable to respond to the client's request. This error would likely be sent as the result of another, undetected, error on the server.

Error code 8

This error code is used when the server wishes to send a random error string to the client. For example, arbitrary Java exceptions may be sent with this error code.

Error code 9

An error occurred during fetch. The error text will describe the actual error.

Error code 10

This error code indicates that the client's requested character encoding is not supported.

Error code 11

An error occurred while setting a statement parameter.

Error code 12

A long field was truncated during fetch.

Error code 13

A reflection request was made, but there's no object on which to call the indicated method. For example, trying to call before calling $sth->execute will cause this error to be reported, since no ResultSet exists.

Error code 14

An unknown class name was passed to $sth->jdbc_func.

Error code 15

A Java exception related to reflection was thrown. This may include, for example, NoSuchMethodException if the requested method can't be located.


See the ToDo file included with the distribution. Highlights include

  • Make the complete JDBC interface available from DBI.

  • DBI metadata methods, cancel, row cache.

  • Better handling of long fields via some sort of streaming interface.

  • JDBC 2.0 support.


perldoc DBI

For general DBI information and questions, see the DBI home page at

This site contains pointers to archives of the DBI mailing lists and list subscription information. DBI in general is primarily supported through the dbi-users mailing list.


Gennis Emerson <>


Copyright 1999-2001,2005-2006 Vizdom Software, Inc. All Rights Reserved.

This program is free software; you can redistribute it and/or modify it under the same terms as the Perl Kit, namely, under the terms of either:

      a) the GNU General Public License as published 
      by the Free Software Foundation; either version 1 
      of the License, or (at your option) any later 
      version, or
      b) the "Artistic License" that comes with the 
      Perl Kit.

This program is distributed in the hope that it will be seful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See either the GNU General Public License or the Artistic License for more details.


The tests for this module make use of hsqldb.

  Copyright (c) 2001-2005, The HSQL Development Group
  All rights reserved.

  Copyright (c) 1995-2000, The Hypersonic SQL Group.
  All rights reserved.

See the full hsqldb copyright and license statements in the t/hsqldb directory or at