Re: Packed vs. Zoned

[bmorris@xxxxxxxxxx]

>From: PatrickConner@parkdalemills.com
>Date: Tue, 25 Sep 2001 10:36:40 -0400
>
>The other day I was discussing the differences/advantages of packed
>numerics vs. zoned numerics in physical files.
>This was brought up by the fact that zoned becomes packed during the
>compile unless explicitly defined.
>One of the benefits of packed is that it is supposed to perform better.
>What other reasons are there for this?
>Also, consider packed vs. binary.
>I understand that if I define a binary field in a physical file the
>compiler will convert it to packed. Is it true that the packed field is
>then converted to binary to be processed at the MI level (or somewhere
down
>there) to be reconverted back to packed on the way back up?

Patrick, RPG converts decimal types (zoned, packed, 'B'-binary) to/from
packed when doing arithmetic on them.  I don't know whether packed
fields are then re-converted to binary as part of MI-level processing.
So packed performs better than binary or zoned in an RPG program.

RPG doesn't convert 'I'-integer (known as "binary" outside the RPG world)
fields, unless it's doing calculations with other packed fields.  You
can get your zero-decimal-position binary fields from external files
and data structures to be treated as integer (rather than 'B'-binary)
types by coding EXTBININT(*YES) in your H spec.

(For compatibility reasons, RPG IV has to keep the same bizarre rules for
'B'-binary as RPG III had.  It's related to the truncation that happens
if the actual value is larger than can fit in the specified number of
digits.  The high-value for a true binary number isn't a nice tidy number
of 9's (214748364 vs 999999999).)

Rule of thumb: define all zero-decimal position variables as integer or
unsigned.  Define all non-zero-decimal position variables as packed, for
convenience (handling the decimal scaling yourself would be ugly, and
probably slower, too).

Barbara Morris