Data files that use a Unicode encoding (UTF-16 or UTF-8) may contain a byte-order mark (BOM) in the first few bytes of the file. For a data file that uses the character set UTF16, the values {0xFE,0xFF} in the first two bytes of the file are the BOM indicating that the file contains big-endian data. The values {0xFF,0xFE} are the BOM indicating that the file contains little-endian data.
If the first primary data file uses the UTF16 character set and it also begins with a BOM, then that mark is read and interpreted to determine the byte order for all primary data files. SQL*Loader reads and interprets the BOM, skips it, and begins processing data with the byte immediately after the BOM. The BOM setting overrides any BYTEORDER specification for the first primary data file. BOMs in data files other than the first primary data file are read and used for checking for byte-order conflicts only. They do not change the byte-order setting that SQL*Loader uses in processing the data file.
In summary, the precedence of the byte-order indicators for the first primary data file is as follows:
BOM in the first primary data file, if the data file uses a Unicode character set that is byte-order dependent (UTF16) and a BOM is present
BYTEORDER parameter value, if specified before the INFILE parameters
The byte order of the system where SQL*Loader is running
For a data file that uses a UTF8 character set, a BOM of {0xEF,0xBB,0xBF} in the first 3 bytes indicates that the file contains UTF8 data. It does not indicate the byte order of the data, because data in UTF8 is not byte-order dependent. If SQL*Loader detects a UTF8 BOM, then it skips it but does not change any byte-order settings for processing the data files.
SQL*Loader first establishes a byte-order setting for the first primary data file using the precedence order just defined. This byte-order setting is used for all primary data files. If another primary data file uses the character set UTF16 and also contains a BOM, then the BOM value is compared to the byte-order setting established for the first primary data file. If the BOM value matches the byte-order setting of the first primary data file, then SQL*Loader skips the BOM, and uses that byte-order setting to begin processing data with the byte immediately after the BOM. If the BOM value does not match the byte-order setting established for the first primary data file, then SQL*Loader issues an error message and stops processing.
If any LOBFILEs or secondary data files are specified in the control file, then SQL*Loader establishes a byte-order setting for each LOBFILE and secondary data file (SDF) when it is ready to process the file. The default byte-order setting for LOBFILEs and SDFs is the byte-order setting established for the first primary data file. This is overridden if the BYTEORDER parameter is specified with a LOBFILE or SDF. In either case, if the LOBFILE or SDF uses the UTF16 character set and contains a BOM, the BOM value is compared to the byte-order setting for the file. If the BOM value matches the byte-order setting for the file, then SQL*Loader skips the BOM, and uses that byte-order setting to begin processing data with the byte immediately after the BOM. If the BOM value does not match, then SQL*Loader issues an error message and stops processing.
In summary, the precedence of the byte-order indicators for LOBFILEs and SDFs is as follows:
BYTEORDER parameter value specified with the LOBFILE or SDF
The byte-order setting established for the first primary data file