RE: Single record access really required (was RE: Views and Indexes)

["Joe Pluta" <joepluta@xxxxxxxxxxxxxxxxx>]

> From: rob@xxxxxxxxx
> 
> The problem with that logic is then it is required in every Query, etc
> that
> uses it.  Therefore if you had a view with
> SELECT iim.IPROD, iim.IDESC, iim.ICLAS, MyUDF(IGLOVR, iim.ICLAS,
:cust) as
> GL#
> FROM iim
> All this complication would be hidden from every use of the GL#.  Not
only
> from query, but also not buried in your application programs.  Sort of
> like
> externalizing I/O.

Actually, I would externalize the I/O, Rob.  This particular code would
definitely be located in a server somewhere; I wouldn't repeat it. And
you're still hung up on queries - a large amount of business programming
involves intermediate results that are never really seen by end users.

In any case, I've shown you my code, which I wrote in about 30 seconds.
However, you still haven't shown me how the SQL version is written.  You
have a magic "MyUDF" which I'd like to see the code to.

I'm also confused about something else: weren't you complaining about
how having DDS means you need another source member?  Doesn't creating
UDFs like this mean you have different source members for each field?
(I could be wrong, I'm not that clear on the whole UDF thing.)


> The second part is not all that clunky if you consider that it could
be a
> subrocedure called after every write.  I would think that one
subprocedure
> should work for all your files.

"I would think".  I'd be interested to see the code for this.  Then I'd
want to see just how much overhead is added to every write if I had a
common subroutine.  Then I'd weigh that against the need to do the
check. I'd think very carefully about whether or not my programming
practices are such as to make checking every WRITE a necessity.  Then
I'd weigh the extra coding and overhead against the added database
integrity.

As always, it's a business decision.

Joe