UNIQUE WHERE NOT NULL - Explained

I always had confusion over the meaning of indexes created with UNIQUE WHERE NOT NULL clause on DB2 for z/OS. The manuals don't seem to do a good job explaining it. I understand it now and here is how it works.

It means its ok to have multiple rows with null values in columns included in the index, but its not ok to have rows where the combination of the columns included in the index is not unique and all the columns are not null. I tested this on a three column index. I was able to insert rows rows shown below just fine, the only duplicate exception I got was where column a, b, c were the same value for two different rows and none of the 3 columns contained a null value.

  COL1  COL2  COL3
  ----  ----  ----
   A     B     C
   A     B     -
   A     B     -
   A     -     -
   A     -     -
   -     -     -
   -     -     -

CREATE UNIQUE WHERE NOT NULL INDEX PRE.IDX1
       ON PRE.TEST
       (COL1   ASC
       ,COL2   ASC
       ,COL3   ASC)
       USING STOGROUP ASTERISK
             PRIQTY 12
             SECQTY 10800
             ERASE NO
       FREEPAGE 0
       PCTFREE 0
       BUFFERPOOL BP1
       CLOSE YES
       PIECESIZE 2097152  K;

CREATE TABLE PRE.TEST
       (COL1     CHARACTER(1)   FOR SBCS DATA WITH DEFAULT NULL
       ,COL2     CHARACTER(1)   FOR SBCS DATA WITH DEFAULT NULL
       ,COL3     CHARACTER(1)   FOR SBCS DATA WITH DEFAULT NULL
       )
       IN DB.TS;