You can use the piecewise capabilities provided by OCIBindByName() or OCIBindByName2(), OCIBindByPos() or OCIBindByPos2(), OCIDefineByPos() or OCIDefineByPos2(), OCIStmtGetPieceInfo(), and OCIStmtSetPieceInfo() to perform inserts, updates or fetches involving column data of the LONG, RAW, LONG RAW, and VARCHAR2 data types.

You can use following Oracle internal data types 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 it is 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 data type 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 Database 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.

Logical rowids are primary key-based logical identifiers for the rows of index-organized tables (IOTs).

To use columns of the UROWID data type, the value of the COMPATIBLE initialization parameter must be set to 8.1 or later.

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)

• SQLT_AFC (CHAR)

• SQLT_AVC (CHARZ)

• SQLT_VST (OCI String)

• SQLT_RDD (ROWID descriptor)

Prior to the addition of these data types with release 10.1, all numeric values in an Oracle Database were stored in the Oracle NUMBER format. These new binary floating point types do 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 data types:

• 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, use external types SQLT_BFLOAT and SQLT_BDOUBLE in conjunction with the BINARY_FLOAT and BINARY_DOUBLE data types.

If you do use it as an external data type, Oracle Database returns numeric values in its internal 21-byte binary format and expects this format on input. The following discussion is included for completeness only.

Oracle Database stores values of the NUMBER data type 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 zeros 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 data type code 2 in the dty parameter of an OCIDefineByPos() or OCIDefineByPos2() 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 must know the number of bytes returned, use the VARNUM external data type instead of NUMBER.

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 Database 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 Database returns an "overflow on conversion" error.

The number is represented in the host system's floating-point format. Normally the length is either 4 or 8 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 Database can represent numbers with greater precision than floating-point representations.

The STRING data type is most useful for C language programs.

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 5-3 shows several examples of the VARNUM values returned for numbers in a table.

You can store up to 2 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.

The first 2 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.

A date in binary format contains 7 bytes, as shown in Table 5-4.

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 data type, 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.

If you are using objects in an Oracle database, you can work with a special OCIDate data type using a set of predefined OCI functions.

The maximum length of a RAW column is 2000 bytes. If the init.ora parameter max_string_size = standard (default value), the maximum length of a RAW can be 2000 bytes. If the init.ora parameter max_string_size = extended, the maximum length of a RAW can be 32767 bytes.

When RAW data in an Oracle Database 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'. To input a character string in your program to a RAW column in an Oracle Database 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.

If you are using objects in an Oracle database, you can work with a special OCIRaw data type using a set of predefined OCI functions.

However, the first 2 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 data type.

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

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 Database 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 Database returns an "overflow on conversion" error.

The first 4 bytes of a LONG VARCHAR contain the length of the item. So, the maximum length of a stored item is 2^31-5 bytes.

The length is contained in the first four bytes. The maximum length is 2^31-5 bytes.

CHAR strings are compared using blank-padded comparison semantics.

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

char my_num[] = "123.45";

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

The C language representation of a REF is a variable declared to be of type OCIRef *. The SQLT_REF data type 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 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 by 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 using a ROWID descriptor, which you can use as a bind or define variable.

They are represented natively, that is, in the host system's floating-point format.

These external types were added in release 10.1 to externally represent the BINARY_FLOAT and BINARY_DOUBLE internal data types. Thus, performance for the internal types is 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 types.

Table 5-6 shows the data conversions for LOBs. For example, the external character data types (VARCHAR, CHAR, LONG, and LONG VARCHAR) convert to the internal CLOB data type, whereas the external raw data types (RAW, VARRAW, LONG RAW, and LONG VARRAW) convert to the internal BLOB data type.

You can also use one of the character data types for the host variable used in a fetch or insert operation from or to a datetime or interval column. Oracle Database does the conversion between the character data type and datetime or interval data type for you (see Table 5-7.

Table 5-8 shows the supported conversions between internal numeric data types 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). An (I/O) implies that the conversion is valid for input and output (binds and defines).

Table 5-9 shows the supported conversions between all relevant internal types and numeric 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). An (I/O) implies that the conversion is valid for input and output (binds and defines).

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 data type in a bind or define operation, enabling you to control data conversions between Oracle Database 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_SIGNED8, 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 5-11 illustrates the mappings between SQLT and OCI_TYPECODE types.