Re: Are level checks still useful?

[Barbara Morris <bmorris@xxxxxxxxxx>]

On 2012/11/7 1:53 PM, Vernon Hamberg wrote:
> The key length is specified in the F-spec = I believe it has to start in
> position 1 of the record, but I can't say that for sure.

The key layout is independent of the record layout. You could have 
fields A B C D E and keys B D. If each of those fields was 10 bytes 
long, the *INPUT data structure would be 50 bytes long and the *KEY data 
structure would be 20 bytes long. It would work fine to use that 20 byte 
structure as the key for CHAIN or SETLL even though the relative 
positions of B and D are different in the key and the record.

The RPG F spec still has an entry for the start position of the key, but 
it's ignored beyond a sanity check by the RPG compiler.

Database knows where the keys are in the record. I guess there must have 
been a system some time in the past where it was possible to have a flat 
file with true program-described keys, where the key type and location 
really was determined by the RPG F spec. But that's not possible now on 
the IBM i, and I'm pretty sure it was never possible on the AS/400 or 
even the System 38. You can't do keyed access against a file that 
doesn't have "official" keys.