/*** TIP 00383 ***/
I need to sort an input file on 2 fields. The input file is fixed
length 25000 byte.  The problem I have is some of the records span 
multiple lines, say multiple records.  I need to keep those groups 
together when doing the sort.  Here's a sample input file,

0001AAA1111ASADFJA;DFJ          -line1
0001CCC1122DAF'SDFK'SDKF        -line2
ADSAF'DKF'A;SDKF'AD;KFD         -line3
0001BBB2222ADFASDFLJAD'KF       -line4
DADFA'DSFKA'SDKF'DKFDFD         -line5
ADFAKFD';LKDSF'AKSDFSDF         -line6
0001DDD1111DADAFSAFDFF          -line7

Sort parameters are: 3 positions from 5th byte
        and          4 positions from 8th byte

A new record is identified by '0001' at the start of the record. If it
is not '0001', that means, the record is a continuation of the
previous line and this record needs to be grouped together.

So the sample output will be:

0001AAA1111ASADFJA;DFJ          -line1
0001BBB2222ADFASDFLJAD'KF       -line4
DADFA'DSFKA'SDKF'DKFDFD         -line5
ADFAKFD';LKDSF'AKSDFSDF         -line6
0001CCC1122DAF'SDFK'SDKF        -line2
ADSAF'DKF'A;SDKF'AD;KFD         -line3
0001DDD1111DADAFSAFDFF          -line7

Any suggestion how I can accomplish this using SAS?

Solution #1
By Charles Patridge

filename rawdata "c:\download\rawdata.txt";

data test;
  length record $80 sort $ 7.;
  retain group sort;
  infile rawdata missover;
  input @1 code $4. @;
  if code = '0001' then do;
    seq = 0;
    group + 1;
    input @1 record ;
    sort = substr(record,5,3)|| substr(record,8,4);
  end;

  if code ne '0001' then input @1 record ;
  seq + 1;
run;

proc sort data=test out=testsort; by sort group seq; run;


/*** rawdata.txt looks like this
0001AAA1111ASADFJA;DFJ          -line1
0001CCC1122DAF'SDFK'SDKF        -line2
ADSAF'DKF'A;SDKF'AD;KFD         -line3
0001BBB2222ADFASDFLJAD'KF       -line4
DADFA'DSFKA'SDKF'DKFDFD         -line5
ADFAKFD';LKDSF'AKSDFSDF         -line6
0001DDD1111DADAFSAFDFF          -line7
***/
/*** end of tip 00383 ***/