Re: Flat file, same field, different positions

[Scott Klement <klemscot@xxxxxxxxxxxx>]

We *used* to do something similar to this.   Our old EDI software (which
we scrapped 5 years ago) was Sterling Software's GENTRAN: Basic for DOS.
It would output the EDI data in a file where all of the EDI elements had
been expanded to fixed-length fields in a flat file.
We would upload that file to our AS/400 and process it in an RPG program.

The way I implemented that was I wrote a program that ran on the PC that
would generate RPG source for all of the different segment types.  and
make a big SELEC group that moved the record data to the appopriate data
structure based on the segment ID.

Then I'd upload that source member to the iSeries...  add the code that
manipulated the data, etc.

As I recall, the fields in the data structure were all set to alphanumeric
so that I didn't have to deal with decimal data errors.   I had
subroutines (though, now I'd use subprocedures) that would do the data
conversion when I needed it.

To avoid problems in the namespace, I used the segment names followed
by a number to name the fields... so I'd have something like this:

     D***************************************************
     D*  Segment N9 -- Reference Number
     D***************************************************
     D dsSegN9         DS
     D* N901  = Reference Number Qualifier
     D  wkN901                18     19
     D* N902  = Reference Number
     D  wkN902                20     49
     D* N903  = Free-form Description
     D  wkN903                50     94
     D* N904  = Date
     D  wkN904                95    100
     D* N905  = Time
     D  wkN905               101    108
     D* N906  = Time Code
     D  wkN906               109    110

I'm not sure if this longwinded tale helps you... but who knows?


>
> There are dozens of fields on each record type.  I'm doing maintenance and
> want to make this RPG IV so I can use reasonable names, etc.  Unfortunately,
> I need to use FIXNBR(*ZONED *INPUTPACKED) in order for it to run as-is.
> This generates tens of thousands of messages in the job log as each
> 'invalid' number is 'fixed'.  Understand that for a 500 record type, the
> program only references the 500 fields.
>
> I want something like:
>      d struc0001       ds
>      d  sensType                      3p 0
>      d  sensID                        7p 0
>      d  roType                        3p 0
>      d  roID                          7p 0
>
>      d struc0360       ds
>      d  sensType                      3p 0
>      d  sensID                        7p 0
>      d  roType                        3p 0
>      d  roID                          7p 0
>
> But that won't compile because the same field is in a different DS.  I could
> use qualified, but as far as the calculations go, eval sensType=sensType0001
> isn't much different from eval sensType=struc0001.sensType.  I still have to
> copy all the fields by hand as it were.
>
> I have two goals:
> 1) I don't want to use FIXNBR; so I'm separating the fields
>    into separate structures
> 2) I'd rather use one set of names to describe the same fields;
>    not SEA500, SEA625, SEA364, but SEA.
>
> Since the input data is packed, I can't readily use the old standby I specs
> because I can't specify the hex codes needed to recognise the packed fields.
> I'm stuck, aren't I?  It gets worse when having to support this back to
> V4R5.  Just thought I'd post a tale of woe on a rainy Northeast US Friday.
>   --buck