Compare counts and abend by 10% difference

[jself]

Hi,
I need some advice on syncsort. I have two files (FILEA - main, FILEB - new). These are bank files with router numbers on them. We are trying to avoid a problem where FILEB comes in bad and only has 500 records and should have had 20,000 records. The information should be pretty much the same but FILEB might have say 100 more records or 100 less records. We would like to compare FILEA - 20,000 RECORDS, FILEB 20,002 RECORDS. Abend if there are more than say 1% difference in FILEA and FILEB.
Anyone have an example of this code?
Thanks for the help.

[jself]

I have the code to do the compare and get the differences in the two files.
I really just need to be able to count FILEA & FILEB and do the calculation.

[BillyBoyo]

Can you post the code you have, please?

[jself]

Here is what I have so far.
SORTJNF1
00000
11111
22222
33333
44444
55555
66666
77777
88888
99999

SORTFNF2
00000
11111
22222
33333
44444
55555
66666
77777

000021 //SORT03 EXEC PGM=SORT
000022 //SYSUDUMP DD SYSOUT=Y
000023 //SYSOUT DD SYSOUT=*
000024 //SORTJNF1 DD DSN=TSO####.KEEP.KEEP.SORTJNF1,
000025 // DISP=SHR
000026 //SORTJNF2 DD DSN=TSO####.KEEP.KEEP.SORTJNF2,
000027 // DISP=SHR
000028 //SORTOUT DD DSN=LIFET.####.#030.JUNK.JUNK,
000029 // DISP=(NEW,CATLG,DELETE)
000031 //SYSIN DD *
000032 JOINKEYS FILES=F1,FIELDS=(1,5,A)
000033 JOINKEYS FILES=F2,FIELDS=(1,5,A)
000034 JOIN UNPAIRED,ONLY
000035 REFORMAT FIELDS=(F1:1,5)
000036 SORT FIELDS=COPY
000037 /*

Thank you for reviewing.

I started another step that reads in the ouput from the above and its as follows.
SORT FIELDS=COPY
OUTFIL REMOVECC,NODETAIL,
TRAILER1=('NUMBER OF RECORDS ARE:',COUNT=(M11,LENGTH=8))

[BillyBoyo]

You don't need a second step. There is a "Main Task" for JOINKEYS, which gives you a full run at anything you can normally have in a SORT/MERGE/COPY step, just with the result of the JOINKEYS as input.

You are counting the mismatches, but you won't be able to tell the percentage if you don't count the originals as well, or do you have a count already?

[jself]

No, I don't have the count for the originals or the new file. Some how I need to count the 10 records from the original file
10 records * 10 percent = 1 record that is allowed to be different.
Join shows 2 differences > 1 allowed then abend.

[jself]

To make things easier I might be able to convince others to use a set amount instead of calculating the difference.
so, in above example still use the join to get the differences.
then count the differences.
2 records > hard coded amount of differences allowed 1 ; then abend.