DataType Conversions in Db2

Question

I am currently stuck on concatenating three fields from a table into a single string. These three fields are of different datatypes.

Select 
    CASE COALESCE(CHAR_COLUMN,'XXX') WHEN 'XXX' 
        THEN 'CHAR_COLUMN is null' 
    else 'CHAR_COLUMN='''||CHAR_COLUMN||'''' END
    ||' and '||
    CASE COALESCE(DT_COLUMN,TIMESTAMP('1980-01-01-00.00.00')) 
        WHEN TIMESTAMP('1980-01-01-00.00.00') THEN 'DT_COLUMN is null' 
    else 'DT_COLUMN='''||DT_COLUMN||'''' END
    ||' and '||
    CASE COALESCE(NUM_COLUMN,111) WHEN 111 
        THEN 'NUM_COLUMN is null' 
    else 'NUM_COLUMN='''||NUM_COLUMN||'''' END
from 
S_DATATABLE

This works perfectly fine in DB2/AIX64 9.1.7 but not in DB2 z/OS 10.1.5.
Error
when ran separately for numeric column

An unexpected token ",111" was found following ",111".  Expected tokens may include:  "CONCAT || / MICROSECONDS MICROSECOND SECONDS SECOND MINUTES".  SQLSTATE=42601

when ran separately for date column

SQL0171N  The data type, length or value of the argument for the parameter in position "2" of routine "||" is incorrect. Parameter name: "||".  SQLSTATE=42815       

Please suggest what changes need to be done for this DB2 version. Thanks in advance.

Answer 1

First of all, regardless of the DB2 version concatenation requires character operands; you cannot concatenate a string and an integer -- DB2 will attempt to implicitly convert non-character data types to characters. It is best if you do explicit conversion to avoid errors.

Secondly, your SQL seems unnecessarily complex. Instead of

CASE COALESCE(NUM_COLUMN,111) 
     WHEN 111 
        THEN 'NUM_COLUMN is null' 
     else 'NUM_COLUMN='''||NUM_COLUMN||'''' 
END

you can simply do this:

CASE WHEN NUM_COLUMN IS NULL 
        THEN 'NUM_COLUMN is null' 
     ELSE 'NUM_COLUMN='||VARCHAR(NUM_COLUMN)
END

Note that in your original code you compare NUM_COLUMN with a character literal, which will also cause implicit conversion. Not all DB2 platforms support implicit conversion between all data types, so once again, do not rely on it but use explicit conversion instead.