8.3.Database Configuration for Unicode

Below are instructions on the configuration of the Unicode enabled drivers and databases for testing. Typically the first task to be performed is the creation of a Unicode enabled Database, which for most databases means configuring them to store data using the UTF8 encoding.

8.3.1.Oracle 8 & 9

The Oracle 9i, 8i and 8.0 databases store Unicode data in the UTF8 encoding scheme, which is an ASCII compatible multibyte encoding for Unicode.

Database Confguration

Using the Oracle ‘Database Configuration Assistant’ wizard follow the options for creating a new database, selecting the ‘custom’ option when presented and you will during the configuration of this Custom database be allowed to ‘Change the Character Set’, at which point this can be changed to UTF8.

To check the character set in use by your database, execute the following query in SQL*Plus:

SQL> SELECT parameter, value FROM nls_database_parameters
   WHERE parameter = 'NLS_CHARACTERSET';
------------------    ---------------------

Unicode support is dependent on the Unicode features available through the Oracle Call Interface (OCI). OCI 8.1.5 supports inputting Unicode data into a database and retrieving Unicode data from a database.

The Following Oracle Data types can be using for storing Unicode data:


Driver Configuration

The Oracle configuration parameter for control character sets is the NLS_LANG environment variable, which should be set to the correct character set for your client. Oracle 8.1.7 claims to be capable of dynamically determining the character set in use on the client and does not require the NLS_LANG to be set, but it is not a bad idea to set it anyway.

Additional information on Oracle Unicode support can be found from otn.oracle.com

8.3.2.Informix 9.x

Database Configuration

When a database is created, the DB_LOCALE in effect at the time is stored in the system catalog and used throughout the lifetime of the database. Using DBACCESS, create a new database with DB_LOCALE set beforehand to ‘EN_US.UTF8’. Specifying UTF8 as the codeset allows the creation of schema objects with names which can contain multibyte characters.

The database locale being used by a database can be viewed in DBACCESS using the menu commands: Database > Info > NLS.

The documentation for Informix GLS states that DB_LOCALE is also used to correctly interpret the locale sensitive datatypes NCHAR and NVARCHAR. The code set specified in DB_LOCALE specifies which characters are valid in any character column as well as the names of database objects.

Setting the Client Locale

The codeset to be used by an Informix client application is specified as part of the client locale. The client locale takes the form:


An Informix 9 Lite driver or agent should use UTF-8 as the codeset. The language and territory should not matter; so it should be possible, for example, to use French (fr_fr) or American English (en_us). For Informix clients on Windows, the client locale is typically set through SetNet32. Rather than rely on the SetNet32 settings, our agent or Lite driver instead sets the client locale at runtime. For an Informix Lite driver on Windows, you must manually add an entry to the registry to set the client locale under the entry for the appropriate DSN in the ODBC.INI hive, add the value:

ClientLocale:REG_SZ:<client locale>

This example uses a codeset number (57372) rather than a codeset name (UTF8) to specify UTF-8 as the codeset. Either form can be used. The registry file included in an Informix client installation lists the supported code sets and the correspondence between codeset names and numbers.

Driver Configuration

For an Informix agent (on Windows or Unix), specify the CLIENT_LOCALE environment variable setting in the [Environment INFORMIX…] section of the rulebook. For an Informix Unix Lite driver, set the CLIENT_LOCALE environment variable appropriately.

8.3.3.Sybase 12.5 +

Database Configuration

The pre-requisites for Unicode with Sybase are:

  • Sybase Active Server (ASE) Version 12.5 or later. (Unicode support is NOT enabled for Version 12.0);

  • Default character set for the Sybase Server needs to be "UTF-8".

To set this:

  1. Make sure the Sybase SQL Server is not running. (Cancel it from the "Services" screen).

  2. Run "Server Configuration" from the "Sybase" entry in the Task Menu bar, or run SYCONFIG.EXE directly;

  3. Either Create a new Active Server or Configure an existing Active Server. (Selection is via Pushbuttons on the dialog box);

  4. For either method, select the "Language" pushbutton;

  5. Select the "Character Set" pushbutton;

  6. Select the "Set Default" pushbutton;

  7. Select "Unicode 3.0.1 UTF-8" from the list box.

If this entry is not available, you will have to add it. From the Character Set selection dialog box, select the "Add / Delete" pushbutton. Select the character set from the list box of those available. Select the "Add" pushbutton (or the "Add All" pushbutton to make all character sets available). Select OK. One the default character set has been selected, select "OK" and "Exit". Start (or restart) the Sybase SQL-Server.

Driver Configuration

There is no need to set anything at the Client end. The character set in use is actually set using Sybase locale functions at connection time. However, it may be useful to ensure that "utf8" is one of the enabled character sets for the relevant platform, in the file [SYBASE]/locales/locales.dat.

8.3.4.Progress 9.1 (SQL-92)

Database Configuration

The Progress database can be run in the UTF-8 Unicode codepage. The sql-92 client can be also be run in unicode. The SQL-92 server uses the codepage of the connected database as its internal codepage. Conversion between the database codepage and the SQL-92 client codepage is done by the server. There are no specific functions provided for converting between codepages within an ESQL-92 program.

The easiest method to create to Progress Unicode enabled database is to use the ‘proutil’ program to convert an existing database to utf8 format using the following command:

proutil <db-name> -C convchar convert utf-8

Multibyte characters can be used in character and varchar fields. Character string literals and the arguments to string functions can also be multibyte characters. There are some provisos for specific functions noted in the documentation. In addition, when the SQL-92 language element syntax requires single quotes, double quotes, parentheses, or braces, the requirement is for the single-byte ASCII encoding of these characters, and other encodings are not equivalent. The string operators in Progress SQL-92 consider the unit of length to be the character count, not a byte count or a column count.

When a column of type CHAR or VARCHAR is created the maximum length specified is a number of characters so the actual number of bytes storage required depends on the database codepage. The length of character data returned in the sqlda is in bytes not characters.

Driver Configuration

For ESQL-92 clients the internal codepage is determined by the value of the client's SQL_CLIENT_CHARSET environment variable, if set. Otherwise, the internal codepage is that of the client's locale. There is a similar environment variable that controls the codepage of messages sent by the database server.

8.3.5.DB/2 v7.x

Database Configuration

Using the DB/2 ‘Control Center’ create a new database instance using the wizard provide. During the create of this database you will be prompted to specify the locale for the new database, which should be set to a code set type of UTF-8. Unicode data can be stored in the following DB/2 datatypes:





Driver Configuration

There are no specific environment variables that need to be set for the DB/2 Driver to handle Unicode data. One special consideration when inserting Unicode data into the daatbase though is that you cannot insert literal Unicode values into the database. Instead these values have to be inserted as bound parameters as follows:

        F4 DBCLOB(100));
         Successfully connected to DSN 'UO_db2'.
  In: StatementHandle = 0x00751860, ParameterNumber = 1,
      InputOutputtype = SQL_PARAM_INPUT=1, ValueType = SQL_C_WCHAR=-8,
      ParameterType = SQL_WCHAR=-8, ColumnSize = 0, DecimalDigits = 0,
      ParameterValuePtr = "?????", BufferLength = 0,
      StrLen_or_IndPtr = SQL_NTS=-3, SQL_LEN_DATA_AT_EXEC = FALSE,
      Buffer Size = 600
   Return:       SQL_SUCCESS=0
   In: Statementhandle = 0x00751860, StatementText = "insert into utest(f1) 
values(?)", Statementlength = 31
   Return:       SQL_SUCCESS=0
   In: Statementhandle = 0x00751860, StatementText = "select * from utest", 
Statementlength = 19
   Return:       SQL_SUCCESS=0
Get Data All:
"F1", "F2", "F3", "F4"
"АБВГД               ", <Null>, <Null>, NO DATA
1 row fetched from 4 columns.

This is because the Graphic string data types are compatible only with other graphic string data types, and never with numeric, character string, or datetime data types.

Note that additional Unicode support has been added to the DB/2 agent for VARCHAR, LONGVARCHAR, CLOB & BLOB types, although a specific Patch (FIXPAK7) is required from IBM to obtain this support in DB/2 v 7.2 databases and FIXPAK 3 & 7 are required for DB/2 v7.1 databases.

The application code page must be set to UTF-8, which can be done by issuing the command:

      db2set DB2CODEPAGE=1208

on the client (DB2 Lite) or server (DB2 agent) as appropriate.

8.3.6.MS SQLServer 2000

There are no Unicode-specific settings for SQLServer. When creating a Database, the collation type for the database can be specified but there is no UTF8 or Unicode specific setting, and a wide (Unicode) language type like Chinese or similar has to be selected, after which wide (Unicode) data can be inserted into the SQLServer wide character types NCHAR & NVARCHAR.

8.3.7.Operational Notes

If you are debugging a unicode connection, you can expect to see this in the request broker log - note the serveropts field:

14:08:11 using mapping: db2:*:*:*:*:*:*
14:08:11 using [generic_db2] ServerProgram=db2_mv
14:08:11 connect params: domain=DB2 db=sample serveropts=W readonly=0

The Unicode parameters that are supplied to the server options cannot be displayed properly in the broker log so the above will be seen; this is normal behaviour.