Re: Finding invalid data in zoned numeric fields with SQL

The quoted message [and others posted in this and another recent thread] may lead one to incorrectly infer that the only valid sign nibble for BCD [binary coded decimal] values are the hex digits D & F [i.e. 0xD & 0xF]. All of the values 0xA through 0xF are valid sign nibbles. On IBM i 6.1 [and prior iterations of that] OS, the "preferred" sign nibbles are defined with the 0xD for negative and 0xF for positive, but the other noted digits are valid & supported.
http://en.wikipedia.org/wiki/Binary-coded_decimal

The SQL can not be used easily to effect Is_Numeric(numeric_column) because an error with the data is diagnosed before the function is performed. The HEX() function is an exception as one [the only AFaIK] function which enables access to the unmapped data.

Regards, Chuck

Jim Essinger wrote:

Thanks all. Several suggestions worked well. I chose the 2nd
statement that Elvis came up with as it seemed the easiest to
understand for maintenance. I was hoping that a Zoned(myField),
DEC(myField) or a Char(myField) would return a null that could
be tested (Where Zoned(myField) is null).
What I get is an data decimal error when SQL tries to retrieve
the records.

On Mon, Jul 13, 2009 at 1:55 PM, Elvis Budimlic wrote:

Not overly ugly for a one time thing... perhaps:

select rrn(a),hex(myField), f1 from myLib/myFile a
where
substring(hex(myField), 1,1) <> 'F' and
substring(hex(myField), 3,1) <> 'F' and
substring(hex(myField), 5,1) <> 'F' and
substring(hex(myField), 7,1) <> 'F' and
substring(hex(myField), 9,1) <> 'F' and
substring(hex(myField),11,1) <> 'F' and
substring(hex(myField),13,1) <> 'F' and
substring(hex(myField),15,1) not in ('F','D')

or

select rrn(a),hex(myField), f1 from myLib/myFile a
where 'F' not in
(substring(hex(myField), 1,1),
substring(hex(myField), 3,1),
substring(hex(myField), 5,1),
substring(hex(myField), 7,1),
substring(hex(myField), 9,1),
substring(hex(myField),11,1),
substring(hex(myField),13,1))
AND
substring(hex(myField),15,1) not in ('F','D')

Jim Essinger wrote:

Nice thought! The concern I have is if the first position of
the field has an "F" value in the left nibble for that byte,
that does not mean the rest of the field has valid values.
I think I am going to have to test all bytes for valid
values, if SQL does not have nice function to do it
for me. That could get a little ugly with an 8 digit field.

Yes - I believe the software vendor did create this with DDS,
and even thought it is defined as Zoned, it somehow allows
invalid data to be added to the file in 4 fields.