Re: Updating a triggered record, part deux

Hi James,

> Be that as it may, why are you complicating things by using pointers to
> get things into structures? If your program is always going to be
> triggering on the same file, you can probably get away with defining the
> data structures directly on the trigger buffer, and even if you can't you
> can certainly just define one big alphanumeric field on the trigger
> buffer, then use substrings on it.

A few reasons why I recommend the pointer approach:

1) Hard-coding the fields in the data structure has already broken during
previous release upgrades.  But, you have no promise from IBM that it will
only break during a release upgrade!  They pass you numbers in the data
structure that tell you where the field data is -- nowhere does it say
that the number will always be the same.  They make no promises that this
position won't break with a PTF, or even during day-to-day use of the
program.

2) The substring approach would work, assuming that you used variables for
the substring positions -- but that would be an order of magnitude more
complicated than the pointer approach.

3) When you use the pointer approach, you can use the external definitions
for your files.  That means that you don't have to re-code the data
structure positions every time you change your file, you can just
recompile the trigger and let it pick up the new definitions -- and that
in turn makes maintenance MUCH simpler.

> Besides which, if you're using pointers, and doing it right, then you
> should be mapping the structures on the data where it is, in the trigger
> buffer, rather than moving it around (which defeats anything you could
> possibly gain by using pointers),

Who said anything about moving it around?  The pointer should be set once,
when the program starts, to put the external definition into the right
position of the trigger buffer...

> in which case, you never moved it out of the trigger buffer, and should
> therefore have no need to move it back in.

There's no need to move it back in...  the way he had his code written,
changing a field in the after image would've changed the actual trigger
buffer.  He just didn't understand.

Just my recommendation, of course...