How LOAD builds indexes while loading data

At the same time the indexes are being built, LOAD checks for duplicate values of any unique index key. If LOAD finds any duplicate values, none of the corresponding rows are loaded. Error messages identify the input records that produce duplicates; optionally, the records are copied to a discard data set. At the end of the job, a summary report lists all errors that are found.

For unique indexes, any two null values are assumed to be equal, unless the index was created with the UNIQUE WHERE NOT NULL clause. In that case, if the key is a single column, it can contain any number of null values, although its other values must be unique.

Neither the loaded table nor its indexes contain any of the records that might have produced an error. Using the error messages, you can identify faulty input records, correct them, and load them again. If you use a discard data set, you can correct the records there and add them to the table with LOAD RESUME.