3 Datatypes

Using External Datatype Codes

An external datatype code indicates to Oracle how a host variable represents data in your program. This determines how the data is converted when returned to output variables in your program, or how it is converted from input (bind) variables to Oracle column values. For example, if you want to convert a NUMBER in an Oracle column to a variable-length character array, you specify the VARCHAR2 external datatype code in the OCIDefineByPos() call that defines the output variable.

To convert a bind variable to a value in an Oracle column, specify the external datatype code that corresponds to the type of the bind variable. For example, if you want to input a character string such as 02-FEB-65 to a DATE column, specify the datatype as a character string and set the length parameter to nine.

It is always the programmer's responsibility to make sure that values are convertible. If you try to insert the string "MY BIRTHDAY" into a DATE column, you will get an error when you execute the statement.

LONG, RAW, LONG RAW, VARCHAR2

You can use the piecewise capabilities provided by OCIBindByName(), OCIBindByPos(), OCIDefineByPos(), OCIStmtGetPieceInfo() and OCIStmtSetPieceInfo() to perform inserts, updates or fetches involving column data of these types.

Character Strings and Byte Arrays

You can use five Oracle internal datatypes to specify columns that contain characters or arrays of bytes: CHAR, VARCHAR2, RAW, LONG, and LONG RAW.

CHAR, VARCHAR2, and LONG columns normally hold character data. RAW and LONG RAW hold bytes that are not interpreted as characters, for example, pixel values in a bit-mapped graphic image. Character data can be transformed when passed through a gateway between networks. Character data passed between machines using different languages, where single characters may be represented by differing numbers of bytes, can be significantly changed in length. Raw data is never converted in this way.

It is the responsibility of the database designer to choose the appropriate Oracle internal datatype for each column in the table. The OCI programmer must be aware of the many possible ways that character and byte-array data can be represented and converted between variables in the OCI program and Oracle tables.

When an array holds characters, the length parameter for the array in an OCI call is always passed in and returned in bytes, not characters.

UROWID

The Universal ROWID (UROWID) is a datatype that can store both logical and physical rowids of Oracle tables. Logical rowids are primary key-based logical identifiers for the rows of Index-Organized Tables (IOTs).

To use columns of the UROWID datatype, the value of the COMPATIBLE initialization parameter must be set to 8.1 or higher.

The following host variables can be bound to Universal ROWIDs:

• SQLT_CHR (VARCHAR2)

• SQLT_VCS (VARCHAR)

• SQLT_STR (NULL-terminated string)

• SQLT_LVC (LONG VARCHAR)

• SLQT_AFC (CHAR)

• SQLT_AVC (CHARZ)

• SQLT_VST (OCI String)

• SQLT_RDD (ROWID descriptor)

BINARY_FLOAT and BINARY_DOUBLE

The BINARY_FLOAT and BINARY_DOUBLE datatypes represent single-precision and double-precision floating point values that mostly conform to the IEEE754 standard for Floating Point Arithmetic.

Prior to the addition of these datatypes, all numeric values in an Oracle database were stored in the Oracle NUMBER format. These new binary floating point types will not replace Oracle NUMBER. Rather, they are alternatives to Oracle NUMBER that provide the advantage of using less disk storage.

These internal types are represented by the following codes:

• SQLT_IBFLOAT for BINARY_FLOAT.

• SQLT_IBDOUBLE for BINARY_DOUBLE.

All the following host variables can be bound to BINARY_FLOAT and BINARY_DOUBLE datatypes:

• SQLT_BFLOAT (native float)

• SQLT_BDOUBLE (native double)

• SQLT_INT (integer)

• SQLT_FLT (float)

• SQLT_NUM (Oracle NUMBER)

• SQLT_UIN (unsigned)

• SQLT_VNU (VARNUM)

• SQLT_CHR (VARCHAR2)

• SQLT_VCS (VARCHAR)

• SQLT_STR (NULL-terminated String)

• SQLT_LVC (LONG VARCHAR)

• SQLT_AFC (CHAR)

• SQLT_AVC (CHARZ)

• SQLT_VST (OCIString)

For best performance, you are advised to use external types SQLT_BFLOAT and SQLT_BDOUBLE in conjunction with the BINARY_FLOAT and BINARY_DOUBLE datatypes.

VARCHAR2

The VARCHAR2 datatype is a variable-length string of characters with a maximum length of 4000 bytes.

NUMBER

You should not need to use NUMBER as an external datatype. If you do use it, Oracle returns numeric values in its internal 21-byte binary format and will expect this format on input. The following discussion is included for completeness only.

Oracle stores values of the NUMBER datatype in a variable-length format. The first byte is the exponent and is followed by 1 to 20 mantissa bytes. The high-order bit of the exponent byte is the sign bit; it is set for positive numbers and it is cleared for negative numbers. The lower 7 bits represent the exponent, which is a base-100 digit with an offset of 65.

To calculate the decimal exponent, add 65 to the base-100 exponent and add another 128 if the number is positive. If the number is negative, you do the same, but subsequently the bits are inverted. For example, -5 has a base-100 exponent = 62 (0x3e). The decimal exponent is thus (~0x3e) -128 - 65 = 0xc1 -128 -65 = 193 -128 -65 = 0.

Each mantissa byte is a base-100 digit, in the range 1..100. For positive numbers, the digit has 1 added to it. So, the mantissa digit for the value 5 is 6. For negative numbers, instead of adding 1, the digit is subtracted from 101. So, the mantissa digit for the number -5 is 96 (101 - 5). Negative numbers have a byte containing 102 appended to the data bytes. However, negative numbers that have 20 mantissa bytes do not have the trailing 102 byte. Because the mantissa digits are stored in base 100, each byte can represent 2 decimal digits. The mantissa is normalized; leading zeroes are not stored.

Up to 20 data bytes can represent the mantissa. However, only 19 are guaranteed to be accurate. The 19 data bytes, each representing a base-100 digit, yield a maximum precision of 38 digits for an Oracle NUMBER.

If you specify the datatype code 2 in the dty parameter of an OCIDefineByPos() call, your program receives numeric data in this Oracle internal format. The output variable should be a 21-byte array to accommodate the largest possible number. Note that only the bytes that represent the number are returned. There is no blank padding or NULL termination. If you need to know the number of bytes returned, use the VARNUM external datatype instead of NUMBER.

INTEGER

The INTEGER datatype converts numbers. An external integer is a signed binary number; the size in bytes is system dependent. The host system architecture determines the order of the bytes in the variable. A length specification is required for input and output. If the number being returned from Oracle is not an integer, the fractional part is discarded, and no error or other indication is returned. If the number to be returned exceeds the capacity of a signed integer for the system, Oracle returns an "overflow on conversion" error.

FLOAT

The FLOAT datatype processes numbers that have fractional parts or that exceed the capacity of an integer. The number is represented in the host system's floating-point format. Normally the length is either four or eight bytes. The length specification is required for both input and output.

The internal format of an Oracle number is decimal, and most floating-point implementations are binary; therefore Oracle can represent numbers with greater precision than floating-point representations.

STRING

The NULL-terminated STRING format behaves like the VARCHAR2 format, except that the string must contain a NULL terminator character. This datatype is most useful for C language programs.

VARNUM

The VARNUM datatype is like the external NUMBER datatype, except that the first byte contains the length of the number representation. This length does not include the length byte itself. Reserve 22 bytes to receive the longest possible VARNUM. Set the length byte when you send a VARNUM value to Oracle Database.

Table 3-3 shows several examples of the VARNUM values returned for numbers in a table.

LONG

The LONG datatype stores character strings longer than 4000 bytes. You can store up to two gigabytes (2^31-1 bytes) in a LONG column. Columns of this type are used only for storage and retrieval of long strings. They cannot be used in functions, expressions, or WHERE clauses. LONG column values are generally converted to and from character strings.

Do not create tables with LONG columns. Use LOB columns (CLOB, NCLOB, or BLOB) instead. LONG columns are supported only for backward compatibility.

Oracle also recommends that you convert existing LONG columns to LOB columns. LOB columns are subject to far fewer restrictions than LONG columns. Furthermore, LOB functionality is enhanced in every release, but LONG functionality has been static for several releases.

VARCHAR

The VARCHAR datatype stores character strings of varying length. The first two bytes contain the length of the character string, and the remaining bytes contain the string. The specified length of the string in a bind or a define call must include the two length bytes, so the largest VARCHAR string that can be received or sent is 65533 bytes long, not 65535.

DATE

The DATE datatype can update, insert, or retrieve a date value using the Oracle internal date binary format. A date in binary format contains seven bytes, as shown in Table 3-4.

The century and year bytes (bytes 1 and 2) are in excess-100 notation. The first byte stores the value of the year, which is 1992, as an integer, divided by 100, giving 119 in excess-100 notation. The second byte stores year modulo 100, giving 192. Dates Before Common Era (BCE) are less than 100. The era begins on 01-JAN-4712 BCE, which is Julian day 1. For this date, the century byte is 53, and the year byte is 88. The hour, minute, and second bytes are in excess-1 notation. The hour byte ranges from 1 to 24, the minute and second bytes from 1 to 60. If no time was specified when the date was created, the time defaults to midnight (1, 1, 1).

When you enter a date in binary format using the DATE external datatype, the database does not do consistency or range checking. All data in this format must be carefully validated before input.

When a DATE column is converted to a character string in your program, it is returned using the default format mask for your session, or as specified in the INIT.ORA file.

RAW

The RAW datatype is used for binary data or byte strings that are not to be interpreted by Oracle, for example, to store graphics character sequences. The maximum length of a RAW column is 2000 bytes.

When RAW data in an Oracle table is converted to a character string in a program, the data is represented in hexadecimal character code. Each byte of the RAW data is returned as two characters that indicate the value of the byte, from '00' to 'FF'. If you want to input a character string in your program to a RAW column in an Oracle table, you must code the data in the character string using this hexadecimal code.

You can use the piecewise capabilities provided by OCIDefineByPos(), OCIBindByName(), OCIBindByPos(), OCIStmtGetPieceInfo(), and OCIStmtSetPieceInfo() to perform inserts, updates, or fetches involving RAW (or LONG RAW) columns.

VARRAW

The VARRAW datatype is similar to the RAW datatype. However, the first two bytes contain the length of the data. The specified length of the string in a bind or a define call must include the two length bytes, so the largest VARRAW string that can be received or sent is 65533 bytes, not 65535. For converting longer strings, use the LONG VARRAW external datatype.

LONG RAW

The LONG RAW datatype is similar to the RAW datatype, except that it stores raw data with a length up to two gigabytes (2^31-1 bytes).

UNSIGNED

The UNSIGNED datatype is used for unsigned binary integers. The size in bytes is system dependent. The host system architecture determines the order of the bytes in a word. A length specification is required for input and output. If the number being output from Oracle is not an integer, the fractional part is discarded, and no error or other indication is returned. If the number to be returned exceeds the capacity of an unsigned integer for the system, Oracle returns an "overflow on conversion" error.

LONG VARCHAR

The LONG VARCHAR datatype stores data from and into an Oracle LONG column. The first four bytes of a LONG VARCHAR contain the length of the item. So, the maximum length of a stored item is 2^31-5 bytes.

LONG VARRAW

The LONG VARRAW datatype is used to store data from and into an Oracle LONG RAW column. The length is contained in the first four bytes. The maximum length is 2^31-5 bytes.

CHAR

The CHAR datatype is a string of characters, with a maximum length of 2000. CHAR strings are compared using blank-padded comparison semantics.

CHARZ

The CHARZ external datatype is similar to the CHAR datatype, except that the string must be NULL-terminated on input, and Oracle places a NULL-terminator character at the end of the string on output. The NULL terminator serves only to delimit the string on input or output; it is not part of the data in the table.

On input, the length parameter must indicate the exact length, including the NULL terminator. For example, if an array in C is declared as

char my_num[] = "123.45";

then the length parameter when you bind my_num must be seven. Any other value would return an error for this example.

The following new external datatypes were introduced with or after release 8.0. These datatypes are not supported when you connect to an Oracle release 7 server.

Named Datatypes: Object, VARRAY, Nested Table

Named datatypes are user-defined types which are specified with the CREATE TYPE command in SQL. Examples include object types, varrays, and nested tables. In the OCI, named datatype refers to a host language representation of the type. The SQLT_NTY datatype code is used when binding or defining named datatypes.

In a C application, named datatypes are represented as C structs. These structs can be generated from types stored in the database by using the Object Type Translator. These types correspond to OCI_TYPECODE_OBJECT.

REF

This is a reference to a named datatype. The C language representation of a REF is a variable declared to be of type OCIRef *. The SQLT_REF datatype code is used when binding or defining REFs.

Access to REFs is only possible when an OCI application has been initialized in object mode. When REFs are retrieved from the server, they are stored in the client-side object cache.

To allocate a REF for use in your application, you should declare a variable to be a pointer to a REF, and then call OCIObjectNew(), passing OCI_TYPECODE_REF as the typecode parameter.

ROWID Descriptor

The ROWID datatype identifies a particular row in a database table. ROWID can be a select-list item in a query, such as:

SELECT ROWID, ename, empno FROM emp

In this case, you can use the returned ROWID in further DELETE statements.

If you are performing a SELECT for UPDATE, the ROWID is implicitly returned. This ROWID can be read into a user-allocated ROWID descriptor using OCIAttrGet() on the statement handle and used in a subsequent UPDATE statement. The prefetch operation fetches all ROWIDs on a SELECT for UPDATE; use prefetching and then a single row fetch.

You access rowids through the use of a ROWID descriptor, which you can use as a bind or define variable.

LOB Descriptor

A LOB (Large Object) stores binary or character data up to 128 terabytes in length. Binary data is stored in a BLOB (Binary LOB), and character data is stored in a CLOB (Character LOB) or NCLOB (National Character LOB).

LOB values may or may not be stored inline with other row data in the database. In either case, LOBs have the full transactional support of the database server. A database table stores a LOB locator that points to the LOB value, which may be in a different storage space.

When an OCI application issues a SQL query which includes a LOB column or attribute in its select-list, fetching the result(s) of the query returns the locator, rather than the actual LOB value. In OCI, the LOB locator maps to a variable of type OCILobLocator.

The OCI functions for LOBs take a LOB locator as one of their arguments. The OCI functions assume that the locator has already been created, whether or not the LOB to which it points contains data.

Bind and define operations are performed on the LOB locator, which is allocated with the OCIDescriptorAlloc() function.

The locator is always fetched first using SQL or OCIObjectPin(), and then operations are performed using the locator. The OCI functions never take the actual LOB value as a parameter.

The datatype codes available for binding or defining LOBs are:

• SQLT_BLOB - a binary LOB datatype.

• SQLT_CLOB - a character LOB datatype.

The NCLOB is a special type of CLOB with the following requirements:

• To write into or read from an NCLOB, the user must set the character set form (csfrm) parameter to be SQLCS_NCHAR.

• The amount (amtp) parameter in calls involving CLOBs and NCLOBs is always interpreted in terms of characters, rather than bytes, for fixed-width character sets.

  See Also:

  "LOB and BFILE Functions in OCI"

Datetime and Interval Datatype Descriptors

The datetime and interval datatype descriptors are briefly summarized here.

TIMESTAMP(fractional_seconds_precision) 

TIMESTAMP(fractional_seconds_precision) WITH TIME ZONE

TIMESTAMP(fractional_seconds_precision) WITH LOCAL TIME ZONE

INTERVAL YEAR(year_precision) TO MONTH

INTERVAL DAY (day_precision) TO SECOND(fractional_seconds_precision)

Native Float and Native Double

The native float (SQLT_BFLOAT) and native double (SQLT_BDOUBLE) datatypes represent the single-precision and double-precision floating point values. They are represented natively, that is, in the host system's floating point format.

Note that these new external types were added to externally represent the BINARY_FLOAT and BINARY_DOUBLE internal datatypes. Thus, performance for the new internal types will be best when used in conjunction with external types native float and native double respectively. This draws a clear distinction between the existing representation of floating point values (SQLT_FLT) and these new types.

C Object-Relational Datatype Mappings

OCI supports Oracle-defined C datatypes for mapping user-defined datatypes to C representations (for example, OCINumber, OCIArray). OCI provides a set of calls to operate on these datatypes, and to use these datatypes in bind and define operations, in conjunction with OCI external datatypes.

Data Conversions for LOB Datatype Descriptors

Table 3-6, "Data Conversions for LOBs" shows the data conversions for LOBs:

Data Conversions for Datetime and Interval Datatypes

You can also use one of the character datatypes for the host variable used in a fetch or insert operation from or to a datetime or interval column. Oracle will do the conversion between the character datatype and datetime/interval datatype for you.

Table 3-7, "Data Conversion for Datetime and Interval Types" is next:

Datetime and Date Upgrading Rules

OCI has full forward and backward compatibility between a client application and the database server as far as the datetime and date columns are concerned.

Data Conversion for BINARY_FLOAT and BINARY_DOUBLE in OCI

Table 3-8 shows the supported conversions between internal numerical datatypes and all relevant external types. An (I) implies that the conversion is valid for input only (binds), and (O) implies that the conversion is valid for output only (defines), while an (I/O) implies that the conversion is valid for input as well as output (binds and defines).

Table 3-9 shows the supported conversions between all relevant internal types and numerical external types. An (I) implies that the conversion is valid for input only (only for binds), and (O) implies that the conversion is valid for output only (only for defines), while an (I/O) implies that the conversion is valid for input as well as output (binds and defines).

Relationship Between SQLT and OCI_TYPECODE Values

Oracle recognizes two different sets of datatype code values. One set is distinguished by the SQLT_ prefix, the other by the OCI_TYPECODE_ prefix.

The SQLT typecodes are used by OCI to specify a datatype in a bind or define operation, enabling you to control data conversions between Oracle and OCI client applications. The OCI_TYPECODE types are used by Oracle's type system to reference or describe predefined types when manipulating or creating user-defined types.

In many cases there are direct mappings between SQLT and OCI_TYPECODE values. In other cases, however, there is not a direct one-to-one mapping. For example OCI_TYPECODE_SIGNED16, OCI_TYPECODE_SIGNED32, OCI_TYPECODE_INTEGER, OCI_TYPECODE_OCTET, and OCI_TYPECODE_SMALLINT are all mapped to the SQLT_INT type.

Table 3-11 illustrates the mappings between SQLT and OCI_TYPECODE types.

Notes:

1. n is the size of the string in bytes

2. These are floating point numbers, the precision is given in terms of binary digits. b is the precision of the number in binary digits.

3. This is equivalent to a NUMBER with no decimal places.

4. i is the size of the number in bytes, set as part of an OCI call.

5. p is the precision of the number in decimal digits; s is the scale of the number in decimal digits.

6. Can only be part of a named collection type.