DB2 10.5 for Linux, UNIX, and Windows

PC version of IXF file format

The PC version of IXF (PC/IXF) file format is a database manager adaptation of the Integration Exchange Format (IXF) data interchange architecture. The IXF architecture was specifically designed to enable the exchange of relational database structures and data. The PC/IXF architecture allows the database manager to export a database without having to anticipate the requirements and idiosyncrasies of a receiving product. Similarly, a product importing a PC/IXF file need only understand the PC/IXF architecture; the characteristics of the product which exported the file are not relevant. The PC/IXF file architecture maintains the independence of both the exporting and the importing database systems.

The IXF architecture is a generic relational database exchange format that supports a rich set of relational data types, including some types that might not be supported by specific relational database products. The PC/IXF file format preserves this flexibility; for example, the PC/IXF architecture supports both single-byte character string (SBCS) and double-byte character string (DBCS) data types. Not all implementations support all PC/IXF data types; however, even restricted implementations provide for the detection and disposition of unsupported data types during import.

In general, a PC/IXF file consists of an unbroken sequence of variable-length records. The file contains the following record types in the order shown:

One header record of record type H
One table record of record type T
Multiple column descriptor records of record type C (one record for each column in the table)
Multiple data records of record type D (each row in the table is represented by one or more D records).

A PC/IXF file might also contain application records of record type A, anywhere after the H record. These records are permitted in PC/IXF files to enable an application to include additional data, not defined by the PC/IXF format, in a PC/IXF file. A records are ignored by any program reading a PC/IXF file that does not have particular knowledge about the data format and content implied by the application identifier in the A record.

Every record in a PC/IXF file begins with a record length indicator. This is a 6-byte right-aligned character representation of an integer value specifying the length, in bytes, of the portion of the PC/IXF record that follows the record length indicator; that is, the total record size minus 6 bytes. Programs reading PC/IXF files should use these record lengths to locate the end of the current record and the beginning of the next record. H, T, and C records must be sufficiently large to include all of their defined fields, and, of course, their record length fields must agree with their actual lengths. However, if extra data (for example, a new field), is added to the end of one of these records, pre-existing programs reading PC/IXF files should ignore the extra data, and generate no more than a warning message. Programs writing PC/IXF files, however, should write H, T and C records that are the precise length needed to contain all of the defined fields.

If a PC/IXF file contains LOB Location Specifier (LLS) columns, each LLS column must have its own D record. D records are automatically created by the export utility, but you will need to create them manually if you are using a third party tool to generate the PC/IXF files. Further, an LLS is required for each LOB column in a table, including those with a null value. If a LOB column is null, you will need to create an LLS representing a null LOB.

The D record entry for each XML column will contain two bytes little endian indicating the XML data specifier (XDS) length, followed by the XDS itself.

For example, the following XDS:

   XDS FIL="a.xml" OFF="1000" LEN="100" SCH="RENATA.SCHEMA" />

will be represented by the following bytes in a D record:

   0x3D 0x00 XDS FIL="a.xml" OFF="1000" LEN="100" SCH="RENATA.SCHEMA" />

PC/IXF file records are composed of fields which contain character data. The import and export utilities interpret this character data using the CPGID of the target database, with two exceptions:

The IXFADATA field of A records.
The code page environment of character data contained in an IXFADATA field is established by the application which creates and processes a particular A record; that is, the environment varies by implementation.
The IXFDCOLS field of D records.
The code page environment of character data contained in an IXFDCOLS field is a function of information contained in the C record which defines a particular column and its data.

Numeric fields in H, T, and C records, and in the prefix portion of D and A records should be right-aligned single-byte character representations of integer values, filled with leading zeros or blanks. A value of zero should be indicated with at least one (right-aligned) zero character, not blanks. Whenever one of these numeric fields is not used, for example IXFCLENG, where the length is implied by the data type, it should be filled with blanks. These numeric fields are:

   IXFHRECL, IXFTRECL, IXFCRECL, IXFDRECL, IXFARECL,
   IXFHHCNT, IXFHSBCP, IXFHDBCP, IXFTCCNT, IXFTNAML,
   IXFCLENG, IXFCDRID, IXFCPOSN, IXFCNAML, IXFCTYPE,
   IXFCSBCP, IXFCDBCP, IXFCNDIM, IXFCDSIZ, IXFDRID

Note: The database manager PC/IXF file format is not identical to the System/370.