Smart DFSORT Tricks

Create DFSORT Symbols from COBOL COPYs

The cobdfsym REXX program shown below allows you to create DFSORT Symbols from a COBOL
copybook. Here's how:
Wrap the COBOL copybook in a dummy program and compile the program to get a compiled
listing.
Use the cobdfsym REXX program to process the compiled listing and create a DFSORT symbols
data set.
Use that DFSORT symbols data set in a SYMNAMES DD statement for your DFSORT or
ICETOOL jobs.
When you upload the cobdfsym program, watch out for the following:
Make sure the cobdfsym statements were not converted to uppercase. The cobdfsym statements
must be mixed case as shown below.
Make sure the REXX operators (for example, || for concatenation) were not translated to other
characters. REXX must be able to recognize the operators.
cobdfsym translates COBOL data types to DFSORT formats according to the table shown here. The
actual translations performed can be summarized as follows:
'Group' to CH
'Grp-VarLen' to CH
'Display' to CH
'Disp-Num' without 'LEADING' and without 'SEPARATE' to ZD
'Disp-Num' with 'LEADING' and without 'SEPARATE' to CLO
'Disp-Num' with 'SEPARATE' and without 'LEADING' to CST
'Disp-Num' with 'LEADING' and with 'SEPARATE' to FS
'Packed-Dec' to PD
'Binary' with S9 to FI
'Binary' without S9 to BI
'Comp-1' to FL
'Comp-2' to FL
Anything else to CH
cobdfsym converts COBOL 88 values to DFSORT symbols as follows:
'literal' to 'literal' (for example, 'APPROVED'). When a COBOL statement sets more than one
literal, cobdfsym generates a DFSORT Symbols statement that sets the symbol to the first literal.
decimal number to decimal number (for example, 14). When a COBOL statement sets more than
one decimal number, cobdfsym generates a DFSORT Symbols statement that sets the symbol to
the first decimal number.
SPACES to ' ' ( blank).
ZERO to 0 (decimal zero).
LOW-VALUE to X'00' (binary zero).
If any of the translated symbols do not meet your needs, you can fine-tune them by editing the DFSORT
Symbols data set produced by cobdfsym.
Here's an example of a job to do the COBOL compile and REXX processing. //SYSLIB must point to
the library that contains your COBOL copybook. //SYSPROC must point to the library in which you've
Page 1 of 22
5/2/2006 file://C:\DOCUME~1\gelewyc\LOCALS~1\Temp\DZKO70W6.htm
stored cobdfsym. //SYMNAMES must point to the output data set or member for the DFSORT Symbols
data set to be produced by cobdfsym; this data set must have, or will be given, RECFM=FB and
LRECL=80.
Be sure to specify MAP in the PARM field.
//*****************************************************************
//* GENERATE SYMBOLIC NAMES FOR DFSORT USING COBOL COPYBOOK
//*****************************************************************
//COBCOMP EXEC PGM=IGYCRCTL,
// PARM=('APOST,RENT,NOSEQ,MAP',
// 'BUF(20K),OPT(STD),TERM,LIB')
//SYSLIB DD DSN=CLIB,DISP=SHR -- library with COBOL copybook
//SYSPRINT DD DSN=&&COBLIST,DISP=(,PASS),UNIT=SYSDA,
// SPACE=(TRK,(10,10))
//SYSTERM DD SYSOUT=*
//SYSIN DD *
IDENTIFICATION DIVISION.
PROGRAM-ID. DUMMYPGM.
ENVIRONMENT DIVISION.
DATA DIVISION.
WORKING-STORAGE SECTION.
COPY DCLSYM01. -- COBOL COPY statement
PROCEDURE DIVISION.
GOBACK.
//SYSLIN DD DUMMY
//SYSUT1 DD SPACE=(CYL,(10),,,ROUND),UNIT=SYSDA
//SYSUT2 DD SPACE=(CYL,(10),,,ROUND),UNIT=SYSDA
//SYSUT3 DD SPACE=(CYL,(10),,,ROUND),UNIT=SYSDA
//SYSUT4 DD SPACE=(CYL,(10),,,ROUND),UNIT=SYSDA
//SYSUT5 DD SPACE=(CYL,(10),,,ROUND),UNIT=SYSDA
//SYSUT6 DD SPACE=(CYL,(10),,,ROUND),UNIT=SYSDA
//SYSUT7 DD SPACE=(CYL,(10),,,ROUND),UNIT=SYSDA
//****************************************************************
//* INTERPRET COBOL LISTING AND CREATE DFSORT SYMNAMES DATA SET
//****************************************************************
//SYMNAMES EXEC PGM=IKJEFT1A,
// COND=(04,LT),
// PARM='%COBDFSYM'
//SYSPROC DD DSN=RLIB,DISP=SHR -- library with cobdfsym REXX program
//SYSTSPRT DD SYSOUT=*
//SYSTSIN DD DUMMY
//COBLIST DD DSN=&&COBLIST,DISP=(OLD,PASS)
//SYMNAMES DD DSN=DFSORT.SYMBOLS(DCLSYM01),DISP=SHR -- DFSORT symbols
As an example, if the DCLSYM01 copybook looked like this:
01 PACKAGE-RECORD.
05 PACKAGE-HEADER.
10 PACKAGE-HEADER-1 PIC X(13).
10 FILLER PIC X.
10 PACKAGE-HEADER-2 PIC X(01).
10 FILLER PIC X.
10 PACKAGE-SEQUENCE PIC 9(08) COMP-3.
05 CUSTOMER-GROUP.
10 CG-NAME PIC X(30).
Page 2 of 22
5/2/2006 file://C:\DOCUME~1\gelewyc\LOCALS~1\Temp\DZKO70W6.htm
10 CG-COUNT PIC 9(10) COMP-3.
10 CG-DATE PIC 9(06) COMP.
10 CG-TIME PIC 9(08) COMP.
10 CG-TYPE PIC S9(02) COMP.
10 CG-LIMIT PIC S9(07).
10 CG-STATUS PIC X(08).
88 APPROVED VALUE 'APPROVED'.
88 DENIED VALUE 'DENIED '.
88 PENDING VALUE SPACES.
10 CG-COUNTY-NO PIC 99.
88 DUTCHESS VALUE 14.
88 KINGS VALUE 24.
88 NOCOUNTY VALUE ZERO.
the DFSORT Symbols created in DFSORT.SYMBOLS(DCLSYM01) would look like this:
PACKAGE-RECORD,1,84,CH
PACKAGE-HEADER,1,21,CH
PACKAGE-HEADER-1,1,13,CH
PACKAGE-HEADER-2,15,1,CH
PACKAGE-SEQUENCE,17,5,PD
CUSTOMER-GROUP,22,63,CH
CG-NAME,22,30,CH
CG-COUNT,52,6,PD
CG-DATE,58,4,BI
CG-TIME,62,4,BI
CG-TYPE,66,2,FI
CG-LIMIT,68,7,ZD
CG-STATUS,75,8,CH
APPROVED,'APPROVED'
DENIED,'DENIED '
PENDING,' '
CG-COUNTY-NO,83,2,ZD
DUTCHESS,14
KINGS,24
NOCOUNTY,0
You could use these symbols in a DFSORT job by specifying:
//S1 EXEC PGM=ICEMAN
//SYMNAMES DD DSN=DFSORT.SYMBOLS(DCLSYM01),DISP=SHR
...
You could use these symbols in an ICETOOL job by specifying:
//S2 EXEC PGM=ICETOOL
//SYMNAMES DD DSN=DFSORT.SYMBOLS(DCLSYM01),DISP=SHR
...
Here's the cobdfsym REXX program:
Page 3 of 22
5/2/2006 file://C:\DOCUME~1\gelewyc\LOCALS~1\Temp\DZKO70W6.htm
/*REXX - COBDFSYM : Create DFSORT symbols from COBOL listing
*** Freeware courtesy of SEB IT Partner and IBM ***
trace r
*/
call read_coblist
call fix_duplicates
call put_symnames
exit
Put_symnames:
/* Write generated symbol definitions */
do i = 1 to nf
queue dnam.i','dval.i
say dnam.i','dval.i
end
/* Write appended symbol definitions */
do i = 1 to na
queue dapp.i
say dapp.i
end
queue ''
'EXECIO * DISKW SYMNAMES (FINIS'
return
Put_line:
/* Analyze Data Division Map line */
parse var line linenr level dataname .
parse var dataname dataname '.' .
if dataname = 'FILLER' then Return
if level = 'PROGRAM-ID' then Return
if level = 88 then Do
nf = nf + 1
dnam.nf = dataname
dval.nf = d88.linenr
dlvl.nf = lev
Return
end
blk = substr(line,64,4)
if level = 1 then nf = 0
hexoff = substr(line,79,3) || substr(line,83,3)
if hexoff = ' ' then hexoff = '000000'
parse var line 92 asmdef datatyp .
if datatyp = 'Group' | datatyp = 'Grp-VarLen'
then parse var asmdef . 'CL' len
else do
len = left(asmdef,length(asmdef)-1)
if right(asmdef,2) = '1H' then len = 2
if right(asmdef,2) = '1F' then len = 4
if right(asmdef,2) = '2F' then len = 8
end
select
when datatyp = 'Group' then typ = 'CH'
when datatyp = 'Grp-VarLen' then typ = 'CH'
when datatyp = 'Display' then typ = 'CH'
when datatyp = 'Disp-Num' then typ = 'ZD'
when datatyp = 'Packed-Dec' then typ = 'PD'
when datatyp = 'Binary' then typ = 'FI'
when datatyp = 'Comp-1' then typ = 'FL'
Page 4 of 22
5/2/2006 file://C:\DOCUME~1\gelewyc\LOCALS~1\Temp\DZKO70W6.htm
when datatyp = 'Comp-2' then typ = 'FL'
otherwise typ = 'CH'
end
if typ = 'FI' then do
if s9.linenr /= 'Y' then typ = 'BI'
end
else do
if typ = 'ZD' then
if sp.linenr = 'Y' then
if ld.linenr = 'Y' then typ = 'FS'
else typ = 'CST'
else
if ld.linenr = 'Y' then typ = 'CLO'
end
off = 1 + x2d(hexoff)
nf = nf + 1
dnam.nf = dataname
dval.nf = off','len','typ
dlvl.nf = lev
Return
Read_COBLIST:
l88 = 0
lx = 0
na = 0
'EXECIO * DISKR COBLIST (FINIS'
parse pull line
do until substr(line,2,16) = ' LineID PL SL '
parse pull line
end
/* Process program text lines */
do until substr(line,2,16) /= ' LineID PL SL '
parse pull line
do until left(line,1) = '1'
call Check_Code_line
parse pull line
end
parse pull line
end
/* Skip lines */
do until substr(line,2,18) = 'LineID Data Name'
parse pull line
end
/* Process Data Division Map lines */
do until substr(line,2,18) /= 'LineID Data Name'
parse pull line
do until left(line,1) = '1'
call Put_line
parse pull line
end
parse pull line
parse pull line
end
/* Skip rest */
do until queued() = 0
parse pull line
end
Return
Page 5 of 22
5/2/2006 file://C:\DOCUME~1\gelewyc\LOCALS~1\Temp\DZKO70W6.htm
Fix_Duplicates:
/* Append _n to any duplicate data names */
nd = 0
tdup. = ''
Do i = 1 to nf
nam = dnam.i
parse var tdup.nam flag i1
if flag = '' then do
tdup.nam = '0' i
iterate
end
if flag = '0' then do
nd = nd + 1
td1.nd = i1 i
tdup.nam = '1' nd
iterate
end
td1.i1 = td1.i1 i
End
Do id = 1 to nd
parse var td1.id i tail
n = 0
Do while i /= ''
n = n + 1
dnam.i = dnam.i || '_' || n
parse var tail i tail
End
End
Return
Check_code_line:
/* Analyze program text line , capture 88 VALUE clauses */
/* Capture S9, LEADING, SEPARATE parameters */
/* Make append lines from *+ comments */
parse var line 4 linenr 10 flag . 19 . 25 stmt 91
if linenr = '' then return
linenr = linenr + 0
if left(stmt,2) = '*+' then do
na = na + 1
dapp.na = substr(stmt,3)
return
end
if left(stmt,1) = '*' then return
if left(stmt,1) = '/' then return
if lastpos('.',stmt) = 0 then do
parse pull line
if left(line,1) = '1' then parse pull line
if substr(line,2,16) = ' LineID PL SL ' then parse pull line
parse var line 4 x1 10 x2 . 19 . 25 stmt2 91
stmt = stmt||stmt2
end
parse var stmt w1 .
if w1 = '88' then do
l88 = linenr
if l88 /= 0 then do
parse var stmt . 'VALUE' tail
if tail /= 0 then do
parse var tail value '.' .
Page 6 of 22
5/2/2006 file://C:\DOCUME~1\gelewyc\LOCALS~1\Temp\DZKO70W6.htm
d88.l88 = strip(value)
if left(d88.l88,6) = 'SPACES'
then d88.l88 = ''' '''
if left(d88.l88,4) = 'ZERO'
then d88.l88 = '0'
if left(d88.l88,9) = 'LOW-VALUE'
then d88.l88 = 'X''00'''
l88 = 0
end
end
return
end
else do
lx = linenr
if lx /= 0 then do
parse var stmt x1 x2 x3
if pos(' S9',x3) /=0 then s9.lx = 'Y'
if pos(' LEADING',x3) /=0 then ld.lx ='Y'
if pos(' SEPARATE',x3) /=0 then sp.lx = 'Y'
lx = 0
end
end
Return
Back to top
Keep the last n records
Keeping the first n records of an input data set with DFSORT is pretty easy. You just use the
STOPAFT=n operand of OPTION or the ENDREC operand of OUTFIL. For example, to keep the first
10 records, you can use either:
OPTION COPY,STOPAFT=10
or:
OPTION COPY
OUTFIL ENDREC=10
But keeping the last n records of an input data set is a bit more challenging since DFSORT doesn't have
any built-in functions to do that. The trick is to attach a sequence number to the records and sort them in
descending order by the sequence number. That way, the last records end up at the front of the file and
you can use STOPAFT or ENDREC to select the last n records. Unfortunately, those last n records will
be in reverse order, so you have to sort them again in ascending order by the sequence number to get
them back in their original order.
It's less complicated then it sounds as illustrated by this DFSORT/ICETOOL job to get the last 10
records for an input data set with RECFM=FB and LRECL=80:
//S1 EXEC PGM=ICETOOL
//TOOLMSG DD SYSOUT=*
//DFSMSG DD SYSOUT=*
Page 7 of 22
5/2/2006 file://C:\DOCUME~1\gelewyc\LOCALS~1\Temp\DZKO70W6.htm
//IN DD DSN=... FB input file
//TEMP DD DSN=&T1,UNIT=SYSDA,SPACE=(CYL,(5,5)),DISP=(,PASS)
//OUT DD DSN=... FB output data set
//TOOLIN DD *
SORT FROM(IN) USING(CTL1)
SORT FROM(TEMP) TO(OUT) USING(CTL2)
/*
//CTL1CNTL DD *
* Add sequence numbers. Use them to reverse the order of
* the records.
INREC FIELDS=(1,80,SEQNUM,8,BI)
SORT FIELDS=(81,8,BI,D)
* Get the last 10 records
OUTFIL FNAMES=TEMP,ENDREC=10
/*
//CTL2CNTL DD *
* Use the sequence numbers to put the last 10 records back in their
* original order.
SORT FIELDS=(81,8,BI,A)
* Remove the sequence numbers.
OUTREC FIELDS=(1,80)
/*
Here's another example. This one gets the last 525 records from a VB input data set:
//S2 EXEC PGM=ICETOOL
//TOOLMSG DD SYSOUT=*
//DFSMSG DD SYSOUT=*
//IN DD DSN=... VB input file
//TEMP DD DSN=&T1,UNIT=SYSDA,SPACE=(CYL,(5,5)),DISP=(,PASS)
//OUT DD DSN=... VB output data set
//TOOLIN DD *
SORT FROM(IN) USING(CTL1)
SORT FROM(TEMP) TO(OUT) USING(CTL2)
/*
//CTL1CNTL DD *
* Add sequence numbers. Use them to reverse the order of
* the records.
INREC FIELDS=(1,4,5:SEQNUM,8,BI,13:5)
SORT FIELDS=(5,8,BI,D)
* Get the last 525 records
OUTFIL FNAMES=TEMP,ENDREC=525
/*
//CTL2CNTL DD *
* Use the sequence numbers to put the last 525 records back in their
* original order.
SORT FIELDS=(5,8,BI,A)
* Remove the sequence numbers.
OUTREC FIELDS=(1,4,13)
/*
Back to top
Set RC=12 or RC=4 if file is empty, has more than n records, etc
Page 8 of 22
5/2/2006 file://C:\DOCUME~1\gelewyc\LOCALS~1\Temp\DZKO70W6.htm
The following related questions have been asked by various customers:
I need to check if a data set is empty or not. If it's empty I want to skip all other steps. Is there any
way I can check for empty data sets?
I would like to skip certain steps in my job if a file is empty. This would be easiest if a utility
would generate a non-zero RC if the input file is empty.
I have several datasets that I need to sort together. How can I terminate if the total count of
records in these data sets is greater than 5000.
I have a file that always has a header and trailer record and may or may not have data records. Is
there any way to check if there are no data records in the file?
ICETOOL can easily satisfy all of these requests. You can use ICETOOL's COUNT operator to set a
return code of 12 or 4 if a specified data set is EMPTY, NOTEMPTY, HIGHER(n), LOWER(n),
EQUAL(n) or NOTEQUAL(n), where n is a specified number of records (for example, 5000). This
makes it easy to control the execution of downstream operators or steps using JCL facilities like IF or
COND. If you use the RC4 operand, ICETOOL sets RC=4; otherwise it sets RC=12.
For example, in the following ICETOOL job, the EMPTY operand of COUNT is used to stop STEP2
from being executed if the IN data set is empty. ICETOOL sets RC=12 if the IN data set is empty, or
RC=0 if the IN data set is not empty. ICETOOL only reads one record to determine if the data set is
empty or not empty, regardless of how many records there are in the data set.
//STEP1 EXEC PGM=ICETOOL
//TOOLMSG DD SYSOUT=*
//DFSMSG DD SYSOUT=*
//IN DD DSN=...
//TOOLIN DD *
* SET RC=12 IF THE 'IN' DATA SET IS EMPTY, OR
* SET RC=0 IF THE 'IN' DATA SET IS NOT EMPTY
COUNT FROM(IN) EMPTY
/*
// IF STEP1.RC = 0 THEN
//*** STEP2 WILL RUN IF 'IN' IS NOT EMPTY
//*** STEP2 WILL NOT RUN IF 'IN' IS EMPTY
//STEP2 EXEC ...
...
// ENDIF
In this next example, the HIGHER(5000) operand of COUNT is used to skip a SORT operation if the
count of records in three data sets is greater than 5000. ICETOOL sets RC=12 if the CONCAT data sets
have a total record count greater than 5000, or a RC=0 if the total record count is less than or equal to
5000. MODE STOP (the default) tells ICETOOL not to execute the SORT operation if the COUNT
operation sets RC=12.
//SRT1 EXEC PGM=ICETOOL
//TOOLMSG DD SYSOUT=*
//DFSMSG DD SYSOUT=*
//CONCAT DD DSN=...
// DD DSN=...
// DD DSN=...
//OUT DD DSN=...
//TOOLIN DD *
* SORT THE 'CONCAT' DATA SETS ONLY IF THEY
* HAVE LE 5000 RECORDS
MODE STOP
COUNT FROM(CONCAT) HIGHER(5000)
Page 9 of 22
5/2/2006 file://C:\DOCUME~1\gelewyc\LOCALS~1\Temp\DZKO70W6.htm
SORT FROM(CONCAT) TO(OUT) USING(CTL1)
/*
//CTL1CNTL DD *
SORT FIELDS=(25,8,BI,A)
/*
In this last example, the EMPTY operand of COUNT is used to stop S2 from being executed if the
INPUT data set doesn't have any data records between the header and trailer. An OMIT statement is
used with COUNT to delete the header and trailer record, leaving only the data records as a subset of the
INPUT data set. ICETOOL sets RC=4 (because the RC4 operand is specified) if the subset of records is
empty, or RC=0 if the subset of records is not empty.
//S2 EXEC PGM=ICETOOL
//TOOLMSG DD SYSOUT=*
//DFSMSG DD SYSOUT=*
//INPUT DD DSN=...
//TOOLIN DD *
* SET RC=4 IF 'INPUT' ONLY HAS A HEADER AND TRAILER, OR
* SET RC=0 IF 'INPUT' HAS DATA RECORDS.
COUNT FROM(INPUT) EMPTY USING(HDTL) RC4
/*
//HDTLCNTL DD *
OMIT COND=(1,6,CH,EQ,C'HEADER',OR,1,7,CH,EQ,C'TRAILER')
/*
// IF S1.RC = 0 THEN
//*** S2 WILL RUN IF 'IN' IS NOT EMPTY
//*** S2 WILL NOT RUN IF 'IN' IS EMPTY
//S2 EXEC ...
...
// ENDIF
Back to top
FB to VB conversion
DFSORT makes it easy to do FB to VB conversion and VB to FB conversion..
The FTOV operand of OUTFIL can be used to change fixed-length (e.g. FB) input records to variable-
length (e.g. VB) output records. If FTOV is specified without OUTREC, the entire FB input record is
used to build the VB output record. If FTOV is specified with OUTREC, the specified fields from the
FB input record are used to build the VB output record. The VB output records will consist of a 4-byte
RDW followed by the FB data. All output data sets for which FTOV is used must have or will be given
variable-length record formats.
Here's an example of FB to VB conversion:
OUTFIL FNAMES=FBVB1,FTOV
OUTFIL FNAMES=FBVB2,FTOV,OUTREC=(1,10,C'=',21,10)
The VB output records for the FBVB1 data set will contain a 4-byte RDW followed by the FB input
record.
The VB output records for the FBVB2 data set will contain a 4-byte RDW followed by the characters
Page 10 of 22
5/2/2006 file://C:\DOCUME~1\gelewyc\LOCALS~1\Temp\DZKO70W6.htm
from input positions 1-10, an '=' character, and the characters from input positions 21-30.
All of the VB output records created with FTOV will consist of:
RDW + input fields
Since all of the input fields from the FB input records are the same, all of the VB output records will be
the same length. But if you have trailing characters such as blanks, asterisks, binary zeros, etc, you can
create true VB output records with different lengths by using the VLTRIM=byte option of OUTFIL.
VLTRIM=byte can be used with FTOV to remove trailing bytes of the specified type from the end of the
VB output records. Here are some examples:
OUTFIL FNAMES=FBVB3,FTOV,VLTRIM=C'*'
OUTFIL FNAMES=FBVB4,FTOV,VLTRIM=X'40'
OUTFIL FNAMES=FBVB5,FTOV,VLTRIM=X'00'
FBVB3 will contain VB output records without trailing asterisks. FBVB4 will contain VB output
records without trailing blanks. FBVB5 will contain VB output records without trailing binary zeros.
To further illustrate how FTOV and VLTRIM=byte work, say you have the following 17-byte FB data
records that you want to convert to VB data records:
123456***********
0003*************
ABCDEFGHIJ*****22
*****************
If you use:
OUTFIL FTOV
the following VB output records will be written (4-byte RDW followed by data):
Length | Data
21 123456***********
21 0003*************
21 ABCDEFGHIJ*****22
21 *****************
but if you use:
OUTFIL FTOV,VLTRIM=C'*'
the following VB output records will be written (4-byte RDW followed by data):
Length | Data
10 123456
8 0003
21 ABCDEFGHIJ*****22
Page 11 of 22
5/2/2006 file://C:\DOCUME~1\gelewyc\LOCALS~1\Temp\DZKO70W6.htm
5 *
Back to top
Change all zeros in your records to spaces
The following related questions have been asked by various customers:
How can I replace all the X'00' bytes in a file with X'40' bytes?
Is there a way to take a large file and replace the low values (X'00') interspersed throughout the
file with spaces (X'40') instead? The low values can show up anywhere; they aren't in fixed
positions.
Can you tell me how to replace all the occurrences of one character (say 'a') with another
character (say 'b') in a sequential file.
Translation features of INREC, OUTREC and OUTFIL OUTREC make it easy to satisfy all of these
requests. The TRAN=ALTSEQ operand can be used to change any character in an input file to another
character in the output file within a specified field or throughout the entire record. The ALTSEQ
statement is used to specify the from and to characters, and TRAN=ALTSEQ is used to specify which
field or fields the ALTSEQ changes are to be applied to.
Here's how you could change all low values (X'00') to spaces (X'40'), in an FB data set with an LRECL
of 60:
ALTSEQ CODE=(0040)
OUTREC FIELDS=(1,60,TRAN=ALTSEQ)
Here's how you could change all 'a' (X'81') and 'x' (X'A7') characters to 'b' (X'82') and 'y' (X'A8')
characters, respectively, in a VB input data set with any LRECL:
ALTSEQ CODE=(81A7,82A8)
OUTREC FIELDS=(1,4,5,TRAN=ALTSEQ)
Of course, you can make your changes to specified fields instead of to the entire record. This comes in
handy when you have mixed character and numeric fields in your records and want to avoid making
changes to the numeric fields. For example, if you had an FB input file that had characters in bytes 1-20,
a PD field in bytes 21-25, and characters in bytes 26-80, you could changes zeros to spaces in the
character fields using:
ALTSEQ CODE=(0040)
OUTREC FIELDS=(1,20,TRAN=ALTSEQ, CH - change zeros to spaces
21,5, PD field - no change
26,55,TRAN=ALTSEQ) CH - change zeros to spaces
By not using TRAN=ALTSEQ for the PD field, we avoid changing PD values incorrectly, such as from
X'000000001C' (P'1') to X'404040401C' (P'404040401').
Back to top
Only include records with today's date
Page 12 of 22
5/2/2006 file://C:\DOCUME~1\gelewyc\LOCALS~1\Temp\DZKO70W6.htm
A customer asked the following question:
Can I make use of DFSORT to sort all of the records from a VSAM file to a sequential file based on the
current date? I have a C'yyyymmdd' date field starting at position 27 of each record. This job will be run
every day. Each day I want only the records for that day to be copied into the sequential file from the
VSAM file.
You can useINCLUDE and OMIT to select records using a variety of formats for today's date like
C'yyyymmdd', C'yyyy/mm/dd', +yyyymmdd, C'yyyyddd', C'yyyy/ddd', +yyyyddd, C'yymmdd' and so
on, as explained here.
The new DATE1 operand corresponds to a C'yyyymmdd' constant for today's date. So the following
control statement will include only those records with a C'yyyymmdd' date in positions 27-34 equal to
today's date:
INCLUDE COND=(27,8,CH,EQ,DATE1)
As some other examples: for date values in the form C'yyyy/mm/dd', you could use the DATE1(/)
constant; for date values in the form C'yyyy-mm', you could use the DATE2(-) constant; for date values
in the form P'yyyyddd', you could use the DATE3P constant; and for date values in the form
Z'yymmdd' (2-digit year date), you could use the Y'DATE1' constant.
Of course, you can use the other comparison operators (NE, GT, GE, LT, LE) as well as EQ. For
example, you could use GT to select records with dates after today, or LT to select records with dates
before today..
Back to top
Include records using relative dates
New! z/OS DFSORT V1R5 PTF UK90007 and DFSORT R14 PTF UK90006 (April, 2006) provide
built-in functions for relative dates like the current date - 1 day or the current date + 30 days. See User
Guide for PTFs UK90007 and UK90006 for complete details on using past date constants and future
date constants.
*****
Although DFSORT has built-in INCLUDE/OMIT constants for the current date in various forms, it does
not have built-in functions for relative dates like the current date - 1 day or the current date + 30 days.
However, if you want to use INCLUDE/OMIT with relative date constants, you can create those relative
date constants as DFSORT Symbols with REXX, and then use the symbols in your INCLUDE/OMIT
statements.
Here's an example of a job to only include records where the 'yyyymmdd' date in positions 1-8 of the
input file is equal to or greater than yesterday's date (that is, ge current date - 1 day).
//TSOBATCH EXEC PGM=IKJEFT1A,DYNAMNBR=200
//SYSEXEC DD DSN=userid.REXXD.REXX,DISP=SHR
//SYSPRINT DD SYSOUT=*
//SYSTSPRT DD SYSOUT=*
//OUTFILE DD DSN=&&S1,UNIT=SYSDA,SPACE=(TRK,(1,1)),
// DISP=(,PASS),LRECL=80,RECFM=FB,DSORG=PS
//SYSTSIN DD *
%REXXRD
//S1 EXEC PGM=ICEMAN
Page 13 of 22
5/2/2006 file://C:\DOCUME~1\gelewyc\LOCALS~1\Temp\DZKO70W6.htm
//SYSOUT DD SYSOUT=*
//SYMNAMES DD DSN=&&S1,DISP=(OLD,PASS)
//SORTIN DD DSN=.... input file
//SORTOUT DD DSN=... output file
//SYSIN DD *
OPTION COPY
INCLUDE COND=(1,8,CH,GE,RELDATE)
/*
The REXX code in userid.REXXD.REXX creates a DFSORT Symbol with date-1 as 'yyyymmdd' in the
&&S1 data set (must have RECFM=FB and LRECL=80) as follows:
RELDATE,C'yyyymmdd'
and RELDATE can then be used as a constant in INCLUDE/OMIT.
Here's the REXX code in userid.REXXD.REXX(REXXRD):
/* REXX */
"EXECIO 0 DISKW OUTFILE(OPEN" /* OPEN FILE */
QUEUE "RELDATE,C"||"'"||DATE('S',DATE('B')-1,'B')||"'"
"EXECIO 1 DISKW OUTFILE (FINIS"
"FREE FI(OUTFILE)"
You can change -1 to any -n or +n value you need.
Back to top
Insert date and time of run into records
The following related questions have been asked by various customers:
Is there any way to dynamically insert the current date, like 'yyyy-mm-dd', in my records?
Something like:
SORT FIELDS=(1,6,CH,A),FORMAT=CH
OUTREC FIELDS=(1:C'2002-03-11',11:1,6)
But with the current date in positions 1-10 instead of a hardcoded constant?
How can I save the current system date (any format) into a sequential file, in such a way that all
the file will ever contain is the said date (this will be used in another step)?
You can use INREC, OUTREC and OUTFIL OUTREC to insert the current date and time in a variety of
formats into your records, as explained here.
Both the DATE1(c) and DATE=(4MDc) operands correspond to a C'yyyycmmcdd' constant for today's
date where c is any separator character you like except blank. So either of the following pairs of control
statements will sort your records on input positions 1-6 and reformat them with today's date in the form
C'yyyy-mm-dd' in output positions 1-10, and input positions 1-6 in output positions 11-16.
SORT FIELDS=(1,6,CH,A),FORMAT=CH
OUTREC FIELDS=(1:DATE1(-),11:1,6)
Page 14 of 22
5/2/2006 file://C:\DOCUME~1\gelewyc\LOCALS~1\Temp\DZKO70W6.htm
or
SORT FIELDS=(1,6,CH,A),FORMAT=CH
OUTREC FIELDS=(1:DATE=(4MD-),11:1,6)
You could insert the current time as well as the current date in your records to produce a timestamp. For
example:
OUTREC FIELDS=(DATE3,TIME1,1,6)
would produce a character timestamp in output positions 1-12 of the form:
yyyydddhhmmss
Date constants can be produced in a variety of other character, zoned decimal and packed decimal
formats as well such as C'yyyy-mm', Z'yyyymmdd' and P'yyddd'. Time constants can also be produced
in a variety of other character, zoned decimal and packed decimal formats as well such as C'hh:mm',
Z'hhmmssxx' and P'hhmmss'.
If, as in the second question above, you wanted to produce just one record containing the date, you could
select from a variety of date formats. For example, if you wanted to create a record with just C'dddyy',
you could do it with OUTREC as follows:
//DATERCD EXEC PGM=ICEMAN
//SYSOUT DD SYSOUT=*
//SORTIN DD *
DUMMY RECORD
//SORTOUT DD DSN=...
//SYSIN DD *
OPTION COPY
OUTREC FIELDS=(YDDDNS=(DY))
/*
Back to top
Change uppercase to lowercase or lowercase to uppercase
A customer asked the following question:
Can DFSORT be used change a field from lowercase to uppercase?
Translation features of INREC, OUTREC and OUTFIL OUTREC make it easy to change the case of
characters in your fields. The TRAN=LTOU operand can be used to change lowercase EBCDIC letters
(a-z) to uppercase EBCDIC letters (A-Z) anywhere in a field. The TRAN=UTOL operand can be used to
change uppercase EBCDIC letters to lowercase EBCDIC letters anywhere in a field.
Here's how you could change lowercase letters to uppercase letters in a 100-byte character field starting
at position 51 and in a 40-byte character field starting in position 301, in an FB data set with an LRECL
of 500:
Page 15 of 22
5/2/2006 file://C:\DOCUME~1\gelewyc\LOCALS~1\Temp\DZKO70W6.htm
OUTREC FIELDS=(1,50, 1-50: no change
51,100,TRAN=LTOU, 51-150: lowercase to uppercase
151,150, 151-300: no change
301,40,TRAN=LTOU, 301-340: lowercase to uppercase
341,160) 341-500: no change
Of course, you could change the case in the entire record as well. For example, here's how you could
change uppercase to lowercase in the records of an FB data set with an LRECL of 200:
OUTREC FIELDS=(1,200,TRAN=UTOL)
And here's how you could change uppercase to lowercase in the records of a VB data set with any
LRECL:
OUTREC FIELDS=(1,4,5,TRAN=UTOL)
Back to top
Generate JCL to submit to the internal reader
A customer asked the following question:
I have a program that generates a data set with a list of file names that can vary from run to run. I'd like
to use ICEGENER to copy the records from the files in the list into one output file. For example, the
generated data set might have these file names:
RUN001.OUTPUT
RUN001.RESULTS
RUN005.OUTPUT
RUN005.RESULTS
so I'd want my output file to have the records from RUN001.OUTPUT, RUN001.RESULTS,
RUN005.OUTPUTand RUN005.RESULTS in that order. Can DFSORT help me do this?
We can copy the input data sets with an ICEGENER JOB like the following:
//CPYFILES JOB (XXX,005),'PRGMR',CLASS=A,MSGCLASS=H,
// MSGLEVEL=(1,1),TIME=(,15)
//S1 EXEC PGM=ICEGENER
//SYSPRINT DD SYSOUT=*
//SYSUT2 DD DSN=&OUT,DISP=(,PASS),SPACE=(CYL,(5,5)),UNIT=SYSDA
//SYSIN DD DUMMY
//SYSUT1 DD DISP=SHR,DSN=file1
// DD DISP=SHR,DSN=file2
...
// DD DISP=SHR,DSN=filen
But since we have to get the file1, file2, ..., filen names from the data set containing the list of file
Page 16 of 22
5/2/2006 file://C:\DOCUME~1\gelewyc\LOCALS~1\Temp\DZKO70W6.htm
names, we'll actually have to generate the JCL shown above and have the internal reader submit it for
execution. That's the tricky part, but here's what we'll do:
We'll generate all of the JCL statements, except for the SYSUT1 concatenation statements, with
OUTFIL's HEADER1 operand. The HEADER1 output is written before the data records and can
contain as many output records as you like (/ is used to start a new record).
Since HEADER1 is a report parameter, the report data set is normally given a RECFM of FBA
and a 1-byte ANSI carriage control character (for example, '1' for page eject) is placed in position
1 of the report. For our purposes, we don't want that ANSI character in position 1 because it
would give us invalid JCL (e.g. '1//CPYFILES ...' instead of '//CPYFILES ...". So we'll use
OUTFIL's REMOVECC operand remove the ANSI character from each JCL statement we
generate.
We'll generate the SYSUT1 concatenation statements with OUTFIL's OUTREC operand. We'll
change the file name in positions 1-44 of each input record into a DD statement referencing that
file name. But as shown in the JCL we're trying to generate above, we need 'SYSUT1' as the
ddname for the first DD and blanks as the ddname for the second and subsequent DD. To make
this happen, we'll use an OUTREC statement with a SEQNUM operand to add a sequence
number to the input records, and then use CHANGE in OUTFIL OUTREC to determine if the
sequence number is 1 or not so we can set up the correct ddname.
Here's our DFSORT job to generate the ICEGENER JCL we need and have the internal reader submit it:
//GENJOB EXEC PGM=ICEMAN
//SYSOUT DD SYSOUT=*
//SORTIN DD DSN=... list of file names
//IRDR DD SYSOUT=(A,INTRDR) internal reader
//SYSIN DD *
OPTION COPY
* Add sequence numbers to file list so we can identify the
* first file in the list and use '//SYSUT1 DD' for it.
* We'll use '// DD' for the second and subsequent files
* in the list.
OUTREC FIELDS=(1,44, file name from list
51:SEQNUM,3,ZD) sequence number
* Generate the JCL for output to the internal reader
OUTFIL FNAMES=IRDR,
* Remove the carriage control character from the "report".
REMOVECC,
* Generate JOB, EXEC, SYSPRINT, SYSUT2 and SYSIN statements.
HEADER1=('//CPYFILES JOB (XXX,005),''PRGMR'',',
'CLASS=A,MSGCLASS=H,',/,
'// MSGLEVEL=(1,1),TIME=(,15)',/,
'//S1 EXEC PGM=ICEGENER',/,
'//SYSPRINT DD SYSOUT=*',/,
'//SYSUT2 DD DSN=&OUT,DISP=(,PASS),SPACE=(CYL,(5,5)),',
'UNIT=SYSDA',/,
'//SYSIN DD DUMMY'),
* Generate //SYSUT1 DD DISP=SHR,DSN=name1
* // DD DISP=SHR,DSN=name2
* ...
OUTREC=(C'//',
51,3,CHANGE=(6,C'001',C'SYSUT1'), 'SYSUT1' for first name
NOMATCH=(C' '), blanks for subsequent names
C' DD DISP=SHR,DSN=',
1,44, file name from list
80:X) ensure record length is 80
/*
Page 17 of 22
5/2/2006 file://C:\DOCUME~1\gelewyc\LOCALS~1\Temp\DZKO70W6.htm
If you'd rather not encode your JOB, etc statements as constants, you can do the same thing in several
DFSORT steps that we did above in one DFSORT step, like this:
//GENJCL1 EXEC PGM=ICEMAN
//SYSOUT DD SYSOUT=*
//SORTIN DD DATA,DLM=$$
//CPYFILES JOB (XXX,005),'PRGMR',CLASS=A,MSGCLASS=H,
// MSGLEVEL=(1,1),TIME=(,15)
//S1 EXEC PGM=ICEGENER
//SYSPRINT DD SYSOUT=*
//SYSUT2 DD DSN=&OUT,DISP=(,PASS),SPACE=(CYL,(5,5)),UNIT=SYSDA
//SYSIN DD DUMMY
$$
//SORTOUT DD DSN=&T1,UNIT=SYSDA,SPACE=(CYL,(1,1)),DISP=(,PASS)
//SYSIN DD *
* Copy JOB, EXEC, SYSPRINT, SYSUT2 and SYSIN statements.
OPTION COPY
/*
//GENJCL2 EXEC PGM=ICEMAN
//SYSOUT DD SYSOUT=*
//SYSUDUMP DD SYSOUT=*
//SORTIN DD DSN=... list of file names
//TEMP DD DSN=&T1,DISP=(MOD,PASS)
//SYSIN DD *
OPTION COPY
* Add sequence numbers to file list so we can identify the
* first file in the list and use '//SYSUT1 DD' for it.
* We'll use '// DD' for the second and subsequent files
* in the list.
OUTREC FIELDS=(1,44, file name from list
51:SEQNUM,3,ZD) sequence number
* Generate //SYSUT1 DD DISP=SHR,DSN=name1
* // DD DISP=SHR,DSN=name2
* ...
OUTFIL FNAMES=TEMP,
OUTREC=(C'//',
51,3,CHANGE=(6,C'001',C'SYSUT1'), 'SYSUT1' for first name
NOMATCH=(C' '), blanks for subsequent names
C' DD DISP=SHR,DSN=',
1,44, file name from list
80:X) ensure record length is 80
/*
//SUBJCL EXEC PGM=ICEMAN
//SYSOUT DD SYSOUT=*
//SORTIN DD DSN=&T1,DISP=(OLD,PASS)
//SORTOUT DD SYSOUT=(A,INTRDR) internal reader
//SYSIN DD *
* Submit the JCL to the internal reader
OPTION COPY
/*
Back to top
Page 18 of 22
5/2/2006 file://C:\DOCUME~1\gelewyc\LOCALS~1\Temp\DZKO70W6.htm
Change a C sign to an F sign in PD values
New! z/OS DFSORT V1R5 PTF UK90007 and DFSORT R14 PTF UK90006 (April, 2006) provide
built-in functions that allow you to select C or F for the sign of positive PD values or change from one to
the other. See User Guide for PTFs UK90007 and UK90006 for complete details on the new TO=PDC
and TO=PDF parameters.
*****
A customer asked the following question:
I have three 5-byte packed decimal fields starting in positions 1, 6 and 11. The positive values have a C
sign (e.g X'123456789C'), but I need them to have an F sign. The negative values have a D sign which I
don't want to change. Can DFSORT change the sign from C to F for the positive values?
Translation features of INREC, OUTREC and OUTFIL OUTREC make it easy to change the C sign to
an F sign in your positive PD values. The new TRAN=ALTSEQ operand can be used to change a
character anywhere in a field to another character. The ALTSEQ statement is used to specify the from
and to characters, and TRAN=ALTSEQ is used to specify which field or fields the ALTSEQ changes
are to be applied to.
In this case, we only want to apply TRAN=ALTSEQ to the sign (last) byte of each of the three PD
fields, that is, positions 5, 10 and 15. Here are the control statements we'd use:
OUTREC FIELDS=(1,4,5,1,TRAN=ALTSEQ,
6,4,10,1,TRAN=ALTSEQ,
11,4,15,1,TRAN=ALTSEQ)
ALTSEQ CODE=(0C0F,1C1F,2C2F,3C3F,4C4F,
5C5F,6C6F,7C7F,8C8F,9C9F)
Back to top
Totals by key and grand totals
A customer asked the following question:
I have a file with ZD data for which I want totals by key and grand totals. The 10-byte CH key starts in
position 1 and the 12-byte ZD field starts in position 19. Here's an example of what the input might look
like:
(Key) (Data)
052500ABCDGRC 00000002246K
053500ABCDGRC 00000828784K
054500ABCDGRC 00000000155H
052501ABCDGRC 02019572110B
053501ABCDGRC 00121859976B
054501ABCDGRC 00515208642F
052502ABCDGRC 02638364444R
053502ABCDGRC 00131222671P
054502ABCDGRC 00824793989R
And here's what I'd want the sorted output to look like:
(Key) (052 total) (053 total) (054 total) (grand total)
Page 19 of 22
5/2/2006 file://C:\DOCUME~1\gelewyc\LOCALS~1\Temp\DZKO70W6.htm
500ABCDGRC 00000002246K 00000828784K 00000000155H 00000830874O
501ABCDGRC 02019572110B 00121859976B 00515208642F 02656640729{
502ABCDGRC 02638364444R 00131222671P 00824793989R 03594381106N
Note that the (headings) shown above are just for reference; they shouldn't actually appear in the output
data set.
I know an indirect method of doing this which will take 5 passes over the data, but is there a way to do it
in a single pass with DFSORT?
The trick here is to use IFTHEN clauses to reformat the records for each type of key (052, 053 and 054)
with slots for all of the keys and the grand total. The actual key will go into its slot and the grand total
slot, and zeros will go into the slots for the other keys. That will allow us to use SUM to produce the key
totals and grand totals. Here's a DFSORT job that produces the requested output:
//S1 EXEC PGM=ICEMAN
//SYSOUT DD SYSOUT=*
//SORTIN DD DSN=... input file
//SORTOUT DD DSN=... output file
//SYSIN DD *
* Reformat 052 records with a slot for 052, 053, 054 and grand total.
* 052 and total will have the 052 values. Others will be Z'0'.
INREC IFTHEN=(WHEN=(1,3,CH,EQ,C'052'),
BUILD=(1:4,10,16:19,12,30:12C'0',44:12C'0',58:19,12)),
* Reformat 053 records with a slot for 052, 053, 054 and grand total.
* 053 and total will have the 053 values. Others will be Z'0'.
IFTHEN=(WHEN=(1,3,CH,EQ,C'053'),
BUILD=(1:4,10,16:12C'0',30:19,12,44:12C'0',58:19,12)),
* Reformat 054 records with a slot for 052, 053, 054 and grand total.
* 054 and total will have the 054 values. Others will be Z'0'.
IFTHEN=(WHEN=(1,3,CH,EQ,C'054'),
BUILD=(1:4,10,16:12C'0',30:12C'0',44:19,12,58:19,12))
* Sort on the key
SORT FIELDS=(1,10,CH,A)
* Sum the 052, 053, 054 and grand total fields.
SUM FORMAT=ZD,FIELDS=(16,12,30,12,44,12,58,12)
/*
Back to top
Join fields from two files record-by-record
A customer asked the following question:
I want to merge two files laterally record by record. For example:
If FILE1 contains:
AAAAA
BBBBB
CCCCC
Page 20 of 22
5/2/2006 file://C:\DOCUME~1\gelewyc\LOCALS~1\Temp\DZKO70W6.htm
and FILE2 contains:
11111
22222
33333
then my output file should contain:
AAAAA11111
BBBBB22222
CCCCC33333
Can DFSORT do this?
Normally, we're asked how to join files on a key, but in this case there is no key; we just want to join the
files record-by-record. Well, no problem, we'll just create a key we can use by adding a sequence
number (1, 2, ...) to the records in file1 and a sequence number (1, 2, ...) to the records in file2. Then we
can join the fields from the 1 records, from the 2 records, and so on using the same technique we used
for joining files on a key.
Below is an ICETOOL job that can do this join by using the SPLICE operator. SPLICE helps you
perform various file join and match operations.
//S1 EXEC PGM=ICETOOL
//TOOLMSG DD SYSOUT=*
//DFSMSG DD SYSOUT=*
//IN1 DD DSN=... file1
//IN2 DD DSN=... file2
//TMP1 DD DSN=&&TEMP1,DISP=(MOD,PASS),SPACE=(TRK,(5,5)),UNIT=SYSDA
//OUT DD DSN=... output file
//TOOLIN DD *
* Reformat the IN1 data set so it can be spliced
COPY FROM(IN1) TO(TMP1) USING(CTL1)
* Reformat the IN2 data set so it can be spliced
COPY FROM(IN2) TO(TMP1) USING(CTL2)
* Splice records with matching sequence numbers.
SPLICE FROM(TMP1) TO(OUT) ON(11,8,PD) WITH(6,5) USING(CTL3)
/*
//CTL1CNTL DD *
* Use OUTREC to create: |f1fld|blank|seqnum|
OUTREC FIELDS=(1:1,5,11:SEQNUM,8,PD)
/*
//CTL2CNTL DD *
* Use OUTREC to create: |blank|f2fld|seqnum|
OUTREC FIELDS=(6:1,5,11:SEQNUM,8,PD)
/*
//CTL3CNTL DD *
* Use OUTFIL OUTREC to remove the sequence number
OUTFIL FNAMES=OUT,OUTREC=(1,10)
/*
Back to top
Page 21 of 22
5/2/2006 file://C:\DOCUME~1\gelewyc\LOCALS~1\Temp\DZKO70W6.htm
Page 22 of 22
5/2/2006 file://C:\DOCUME~1\gelewyc\LOCALS~1\Temp\DZKO70W6.htm