5 Globalization Support

Character sets and languages

Choosing a database character set determines what languages can be represented in the database.

A group of characters, such as alphabetic characters, ideographs, symbols, punctuation marks, and control characters, can be encoded as a character set. An encoded character set assigns unique numeric codes to each character in the character repertoire. The numeric codes are called code points or encoded values.

Character sets can be single-byte or multibyte. Single-byte 7-bit encoding schemes can define up to 128 characters and normally support just one language. Single-byte 8-bit encoding schemes can define up to 256 characters and often support a group of related languages. Multibyte encoding schemes are needed to support ideographic scripts used in Asian languages like Chinese or Japanese because these languages use thousands of characters. These encoding schemes use either a fixed number or a variable number of bytes to represent each character. Unicode is a universal encoded character set that enables information from any language to be stored using a single character set. Unicode provides a unique code value for every character, regardless of the platform, program, or language.

Client operating system and application compatibility

The database character set is independent of the operating system. On an English operating system, you can create and run a database with a Japanese character set. However, when an application in the client operating system accesses the database, the client operating system must be able to support the database character set with appropriate fonts and input methods. For example, you cannot insert or retrieve Japanese data on the English Windows operating system without first installing a Japanese font and input method. Another way to insert and retrieve Japanese data is to use a Japanese operating system remotely to access the database server.

If all client applications use the same character set, then that character set is usually the best choice for the database character set. When client applications use different character sets, the database character set should be a superset of all the application character sets. This ensures that every character is represented when converting from an application character set to the database character set.

Performance and storage implications

For best performance, choose a character set that avoids character set conversion and uses the most efficient encoding for the languages desired. Single-byte character sets result in better performance than multibyte character sets, and they also are the most efficient in terms of space requirements. However, single-byte character sets limit how many languages you can support.

Character sets and replication

All databases in a replication scheme must have the same database character set. No character set conversion occurs during replication.

Monolingual linguistic sorts

TimesTen compares character strings in two steps for monolingual sorts. The first step compares the major value of the entire string from a table of major values. Usually, letters with the same appearance have the same major value. The second step compares the minor value from a table of minor values. The major and minor values are defined by TimesTen. TimesTen defines letters with accent and case differences as having the same major value but different minor values.

Monolingual linguistic sorting is available only for single-byte and Unicode database character sets. If a monolingual linguistic sort is specified when the database character set is non-Unicode multibyte, then the default sort order is the binary sort order of the database character set.

For a list of supported sorts, see "NLS_SORT "in the Oracle TimesTen In-Memory Database Reference.

Multilingual linguistic sorts

TimesTen provides multilingual linguistic sorts so that you can sort data for multiple languages in one sort. Multilingual linguistic sort is based on the ISO/OEC 14651 - International String Ordering and the Unicode Collation algorithm standards. This framework enables the database to handle languages that have complex sorting rules, such as those in Asian languages, as well as providing linguistic support for databases with multilingual data.

In addition, multilingual sorts can handle canonical equivalence and supplementary characters. Canonical equivalence is a basic equivalence between characters or sequences of characters. For example, ç is equivalent to the combination of c and ,.

For example, TimesTen supports a monolingual French sort (FRENCH), but you can specify a multilingual French sort (FRENCH_M). _M represents the ISO 14651 standard for multilingual sorting. The sorting order is based on the GENERIC_M sorting order and can sort accents from right to left. TimesTen recommends using a multilingual linguistic sort if the tables contain multilingual data. If the tables contain only French, then a monolingual French sort may have better performance because it uses less memory. It uses less memory because fewer characters are defined in a monolingual French sort than in a multilingual French sort. There is a trade-off between the scope and the performance of a sort.

For a list of supported multilingual sorts, see "NLS_SORT" in the Oracle TimesTen In-Memory Database Reference.

Case-insensitive and accent-insensitive linguistic sorts

Operations inside a database are sensitive to the case and the accents of the characters. Sometimes you might need to perform case-insensitive or accent-insensitive comparisons.

To specify a case-insensitive or accent-insensitive sort:

• Append _CI to a TimesTen sort name for a case-insensitive sort. For example:

  BINARY_CI: accent-sensitive and case-insensitive binary sort

  GENERIC_M_CI: accent-sensitive and case-insensitive GENERIC_M sort

• Append _AI to a TimesTen sort name for an accent-insensitive and case-insensitive sort. For example:

  BINARY_AI: accent-insensitive and case-insensitive binary sort

  FRENCH_M_AI: accent-insensitive and case-insensitive FRENCH_M sort

Performing a linguistic sort

The NLS_SORT data store connection attribute indicates which collating sequence to use for linguistic comparisons. The NLS_SORT value affects the SQL string comparison operators and the ORDER BY clause.

You can use the ALTER SESSION statement to change the value of NLS_SORT:

ALTER SESSION SET NLS_SORT=SWEDISH;
SELECT product_name
  FROM product
  ORDER BY product_name;

PRODUCT NAME
------------
aerial
Antenne
Lcd
ächzen
Ähre

You can also override the NLS_SORT setting by using the NLSSORT SQL function to perform a linguistic sort:

SELECT * FROM test ORDER BY NLSSORT(name,'NLS_SORT=SPANISH');

For more extensive examples of using NLSSORT, see "NLSSORT" in the Oracle TimesTen In-Memory Database SQL Reference.

Using linguistic indexes

You can create a linguistic index to achieve better performance during linguistic comparisons. A linguistic index requires storage for the sort key values.

To create a linguistic index, use a statement similar to the following:

CREATE INDEX german_index ON employees
(NLSSORT(employee_id, 'NLS_SORT=GERMAN'));

The optimizer chooses the appropriate index based on the values for NLSSORT and NLS_SORT.

You must create multiple linguistic indexes if you want more than one linguistic sort on a column. For example, if you want both GERMAN and GERMAN_CI sorts against the same column, create two linguistic indexes.

For more information, see "CREATE INDEX" in the Oracle TimesTen In-Memory Database SQL Reference.

Backward compatibility using TIMESTEN8

TIMESTEN8 is a restricted database character set that specifies behavior from TimesTen releases before 7.0. It is supported for backward compatibility only.

TIMESTEN8 has the following restrictions:

• There is no support for character set conversion of any kind. This includes:

  • Conversions between the application and the database. If DatabaseCharacterSet is TIMESTEN8, then ConnectionCharacterSet must also be TIMESTEN8.

  • Conversions between CHAR/VARCHAR2 data and NCHAR/NVARCHAR2 data.

• Sorting for CHAR and VARCHAR data types is limited to binary ordering. NLS_SORT=BINARY is the only sort allowed.

• TIMESTEN8 is not supported in IMDB Cache.

During database creation, customers should select the database character set matching the actual encoding of the data being stored in CHAR and VARCHAR2 columns whenever possible. Select TIMESTEN8 only when backwards compatibility to existing TimesTen data is required.

Object migration and character sets

The ttMigrate utility tags each object it saves with the object's character set. By default, ttMigrate stores object data in the database character set, but you can choose a different character set by using the -saveAsCharset option. You can specify this option in create mode (-c) or append mode (-a).

When you use ttMigrate to restore an object, its data is implicitly converted to the database character set of the target database if needed. Character set conversion can result in data loss if the database character set of the target database cannot represent all of the data that it receives.

If you know that the data is in encoded in the database character set of the target database, you can use the -noCharsetConversion option. This option can be specified only in restore mode (-r). If you use the -noCharsetConversion option, ttMigrate treats the data as if it is in the database character set of the target database.

When you restore untagged character data from a database that was created before release 7.0 into a database from release 7.0 and later, ttMigrate treats the data as if it is in the database character set of the target database.

The ttMigrate utility issues a warning whenever there is an implicit or explicit character set conversion while saving or restoring data.

Migration and length semantics

The ttMigrate utility saves length semantic information about CHAR and VARCHAR2 columns. It restores the length semantic information when restoring objects into databases created in TimesTen release 7.0 or later.

When objects are migrated back into a TimesTen release before 7.0, columns with character semantics are converted to byte semantics and the column length is adjusted to match the byte length of the original columns.

When objects are migrated from a release before 7.0 to release 7.0 and later, byte semantics is used.

Migrating linguistic indexes

The ttMigrate utility supports migration of linguistic indexes into TimesTen releases that support them. When migrating back to a TimesTen release before 7.0, ttMigrate issues a warning indicating that the linguistic indexes cannot be restored. Migration of the table proceeds without the linguistic indexes.

Migrating cache group tables

You cannot restore cache group tables containing NCHAR/NVARCHAR2 columns to a release before 7.0. Releases before 7.0 do not allow these data types in cache group tables.