RE: IFS in RPG

>It seems to me that with either method (API's and
>a lot of parsing) or "integrated" file i/o and
>parsing, you'd have to be changing your programs
>and/or file formats a lot.  Is that the case?

Not much.  The work comes in finding out the new rules.  I never get to
throw out the old rules, alas.  I haven't yet perfected a foolproof
table-driven pattern matching algorithm that performs quickly enough to deal
with a few million records a pop, so there's code involved for each 'record
type.'  The decoder stores the decoded messages in a dynamic array and
passes back the message count:

mainline
read block
  eval count = decode(block: messages: recordType)
  for i = 1 to count
    process messages(i)
  end
loop

The decode procedure is nothing more than a traffic cop, directing control
to the appropriate decoding routine.  There's one routine for each record
type.  In my case, the I/O really is trivial, and the vast bulk of the
processing goes to decoding the messages.  The beauty of doing it this way
is that my decoding is device independent.  It doesn't matter if the raw
data comes from the IFS, a CD-ROM or a tape drive (yes I have to deal with
all of these.)  The I/O bit is all that changes; the decoding service
program doesn't care one whit.

Cheers!
  --buck