Comments on: Expression to insert value to a table column in SQL

By: philpl1jb

philpl1jb — Sun, 20 Jan 2013 01:37:58 +0000

Great.

By: TomLiotta

TomLiotta — Sun, 20 Jan 2013 00:31:21 +0000

…though it could indeed be replaced by overflow/underflow if the data isn’t good. — Tom

By: TomLiotta

TomLiotta — Sun, 20 Jan 2013 00:29:40 +0000

SQL0406 is “Conversion error on assignment to column &2.” It’s a conversion error (as noted in the question). It goes away when the CAST() is done. — Tom

By: philpl1jb

philpl1jb — Sun, 20 Jan 2013 00:24:44 +0000

Apparently not. This was a question about error SQL0406 which is, I believer, an overflow situation.

By: TomLiotta

TomLiotta — Sat, 19 Jan 2013 22:12:59 +0000

Error handling for possible overflow/underflow is always a good idea. Can’t argue against it. However, the values in A, B, C and D should sanity checked at the point of entry (or update). If values that will be trouble are not allowed into the file in the first place, later tests can be minimized for processing efficiency. This current process is only updating FLD1, so it might be rational to _assume_ that the existing values will work. Maybe. — Tom

By: philpl1jb

philpl1jb — Sat, 19 Jan 2013 20:16:48 +0000

Don’t know if a case can be used in the update..might be two updates with where clauses .. 99999etc where A*B*C/10**D > 9999999999999.99 and the other updates the field with the actual value.

By: philpl1jb

philpl1jb — Sat, 19 Jan 2013 20:12:11 +0000

so I would probably do a case filling the field with 9s when the result is > 9999999999999.99

By: philpl1jb

philpl1jb — Sat, 19 Jan 2013 20:07:56 +0000

…SET FLD1 = A*B*C/cast(10**D as int)Will fail with divide by zero error when D is a negative number…Doesn’t matter what you do on the right, process will fail with error SQL0406 when updating FLD1 when the answer has more than 13 significant digits as in 1234567890123.45 * 1234567890123. * 12345678.90123 / 10 * 1

By: TomLiotta

TomLiotta — Sat, 19 Jan 2013 07:59:30 +0000

I should have thought of this immediately. The exponentiation is causing an intermediate floating-point result to be created, and that causes floating-point to be carried through each subsequent intermediate result. Try it this way:

…SET FLD1 = A*B*C/cast(10**D as int)

By casting as an INTEGER, the division works much better and final conversion to the FLD1 data type is fairly straightforward.

Tom

By: philpl1jb

philpl1jb — Fri, 18 Jan 2013 23:50:16 +0000

Sorry .. had half my mind on work .. and that’s the most of it.