MVSFORUMS.com

Disco_Stu · Beginner Joined: 04 Feb 2003 Posts: 19 Topics: 7

Here is my situation:

I am concatenating 5 datasets as input. Each is of the same format.

I need to put these through a syncsort that will sum all duplicate detail records based on positions 1-63. The final result will be one header record, sort/summed detail reocrds, and a trailer record with updated totals. This will look like

kolusu · Posted: Wed Oct 29, 2003 11:55 am Post subject:

Disco_stu,

I really don't see the use of an exit in here unless you are doing some complex programming logic. Every thing you asked for can be done using sort with just one pass.The simplest and easiest solution(option 2) would be to pull off the header and trailer records and then sum sort on the detail records. Ofcourse this would involve an extra pass of the data.

There is a way to do all in just one pass. I need the following details.

1. How do you distinguish a header record, a detail record and the trailer record?

2. what are the contents in the positions(154, 163 & 172 ) on the header and trailer record?

3. where does the money field start on the trailer record as you said it is at a different position

4. what is the recfm,lrecl of the files?

5. Please post a sample of input records and the desired output

Btw I don't think you can sum on 9 bytes on packed decimal field.Sort can handle a PD length up to 8 (15 digits and a sign), so if the length is really 9 (17 digits and a sign), that's a problem, unless the first byte of the PD field is always X'00'.

Let me know these details and I will post the solution.

Kolusu

Frank Yaeger · Posted: Wed Oct 29, 2003 12:26 pm Post subject:

Kolusu said

Disco_Stu · Beginner Joined: 04 Feb 2003 Posts: 19 Topics: 7

I really don't see the use of an exit in here unless you are doing some complex programming logic. 

I need to update the trailer record with the new record count. if records will end up being combined, the count will decrease. I also want to merge all the trailer records into one. I dont want the final output to have more than one header or trailer record. Maybe there is an easier way to do this...

1. How do you distinguish a header record, a detail record and the trailer record? 

header - 'CC' in pos 1-2
detail - 'DL' in pos 1-2
trailer - 'XX' in pos 1-2

2. what are the contents in the positions(154, 163 & 172 ) on the header and trailer record? 

header - spaces

trailer - various money buckets, but the positions arent the same as the detail reocrd, so they dont match up with the detail layout, hence soc7

3. where does the money field start on the trailer record as you said it is at a different position

start at positions 162,171,180; each defined as PIC S9(15)V99 comp-3

4. what is the recfm,lrecl of the files?

293 bytes, fixed block

5. Please post a sample of input records and the desired output 

I'll use an example here. i figured real data would be too big. I have 2 input datasets here. each has a header (CC) record, a few detail (DL) records, and a trailer (XX) record.
The header record is all spaces.
The deail reocord is made up of:
Positions 3-5 are my key.
At the end, 13-15 is a money field.
Trailer record has a record count in positions 3-5 and a total money field in positions 10-12.

Dataset 1

CC
DLAAABBBBBBB001
DLAAABBBBBBB000
DLAAABBBBBBB003
XX003AAAA004AAA

Dataset 2
CC
DLAAABBBBBBB002
DLAAABBBBBBB002
DLCCCBBBBBBB002
XX003AAAA006AAA

Output
CC
DLAAABBBBBBB008
DLCCCBBBBBBB002
XX002AAAA010AAA

All recs with key 'AAA' are combined. New trailer record with updated record count and money amt is created.

Btw I don't think you can sum on 9 bytes on packed decimal field.Sort can handle a PD length up to 8 (15 digits and a sign), so if the length is really 9 (17 digits and a sign), that's a problem, unless the first byte of the PD field is always X'00'. 

See above for field definition. It's been working so far. but since the field is so big, the first byte is probably always X'00', like you said. I dont really have any freedom to change the field size.

Depends on which "Sort" you mean - DFSORT can handle a PD length up to 16 bytes for SUM.

I am using SyncSort

kolusu · Posted: Wed Oct 29, 2003 2:16 pm Post subject:

Disco_stu,

The following JCL will give you the desired results. The only thing I was not able to do is the updation of the trailer record count.A brief explanation of the JOb.

I am using the change parm to change the contents of fields for header and trailer records.For every header & trailer record we are going to change the bytes 154, 163, 172 to have zero in packed decimal format.

I am also adding another 3 fields for which I am using the change operator to pad zeroes for every header & detail record. But for the trailer record I copy the contents from the bytes 162,171,180.

Now I summing the contents in 154, 163, 172, as well as the contents which we added i.e bytes 294,303,312

Now all the detailed records will be summed in the bytes 154 thru 180 and the trailer records will be summed in the bytes 294 thru 320.

Now using outrec fields I am adding the summed value in 155 bytes with the value in 295 bytes. similarly for the values in 164 bytes with the value in 304 .similary for the values in 173 with the value in 313. I am adding on 8 bytes as Sort can handle a PD length up to 8 and I padded a X'00' in front to the fields.

Now using change parm on outfil outrec I am padding spaces for the header record.

I haven't tested this code as I don't have the time to set up the input data.But you get the idea as to how it is being done.

kolusu · Posted: Wed Oct 29, 2003 2:22 pm Post subject:

Frank,

Thanks for pointing that for summation DFSORT can handle upto 16 bytes of PD field.It is the same for syncsort also.But the max length on inrec/outrec fields is 8 for PD for syncsort.Does DFSORT allow more than 8 bytes for PD field on outrec?? I get a numeric field error if try to convert a zoned decimal to PD field with length of 9 bytes.

Kolusu

Disco_Stu · Beginner Joined: 04 Feb 2003 Posts: 19 Topics: 7

Thanks. There's alot of stuff in there I've never seen before. I'm gonna have to go through it and figure out the details. But from your description i get the basic idea.

Unfortunately, the count definitely needs to get updated. Otherwise i'll get out of balance errors down the line.

But you definitly exposed me to some new synsort code and some new ideas, and i think i'm going to be able to apply some of the things i see here to other processes, so thank you for your time.

Frank Yaeger · Posted: Wed Oct 29, 2003 3:18 pm Post subject:

Kolusu said

Disco_Stu · Beginner Joined: 04 Feb 2003 Posts: 19 Topics: 7

One last question if you have the time...

whats the difference here between the OUTREC control statement

kolusu · Posted: Wed Oct 29, 2003 4:35 pm Post subject:

Disco_Stu,

Yes I am reformating the output twice. I need to process the output record twice as I want sum up the quantities and then also changes the value to spaces on the header record.

Basically this is what happens after summing of the records

kolusu · Posted: Thu Oct 30, 2003 9:12 am Post subject:

Disco_stu,

I need small clarification from you about trailer record. Where does the count of records fit in? i.e the position of the record count.

For that position what are the contents on the header and detail records.

You can have the updated record count also with some adjustments. Are you willing increase the LRECL of the output dataset??

Kolusu

Disco_Stu · Beginner Joined: 04 Feb 2003 Posts: 19 Topics: 7

This is what the trailer record looks like.

kolusu · Posted: Thu Oct 30, 2003 10:21 am Post subject:

Disco_stu,

If the alphanumeric fileds in position 23 for 4 bytes on the header and detail records are not that important then you can add a seqnum to outrec fields and in the end subtract 2 from it so that you will get the record count.