Joined: 28 Dec 2006 Posts: 6 Location: Italy, Trieste
Hi all,
I use the IBM VS COBOL II Release 4.0 compiler.
I have the following need for a Cobol program:
read a sequential dataset with record-format and record-length undefined (for "undefined" I mean: I know its at runtime moment, not in compile time).
Other information:
- I can assume that the record-format is only or FB or VB.
- I can call Assembler modules.
- I cannot change the version of compiler.
- I know the DDNAME of the dataset, and it is defined in the JCL of execution of the Cobol
In my knowledge it's impossible satisfy this goal only in Cobol (I mean: in VS COBOL II).
In the web I founded a lot of Assembler routine (called from a Cobol) that use DYNALLOC macro, but these routines are not useful for me, since I must define in the Cobol the FD with the exact lenght of the dataset.
My main questions:
a) Is it true my knowledge that it's impossible satisfy this goal only in Cobol (I mean: in VS COBOL II)?
b) If answer to a) is "yes", somebody has already developed an Assembler routines (callable from a Cobol) that reachs the goal?
c) If answer to b) is "no", somebody has already developed an Cobol examples that reachs the goal?
and, last question (but only if there is an aswer to first three)
d) Somebody has already developed an Assembler routines (callable from a Cobol) or a Cobol that writes an undefined dataset?
Joined: 22 Nov 2005 Posts: 700 Location: Troy, Michigan USA
Gabriele Soranzo,
I am absolutely 100% that with the correct assembler routines this is possible, but if you have access to these types of assembler subroutines, why not just let the subroutine deal with ALL of the I/O and just pass back the record.
Another question: What are you going to do with a record of varying lengths, how are you going to process this?
Joined: 18 Nov 2006 Posts: 3156 Location: Tucson AZ
Gabriele Soranzo wrote:
I have the following need for a Cobol program:
read a sequential dataset with record-format and record-length undefined (for "undefined" I mean: I know its at runtime moment, not in compile time).
Other information:
- I can assume that the record-format is only or FB or VB.
- I can call Assembler modules.
- I cannot change the version of compiler.
- I know the DDNAME of the dataset, and it is defined in the JCL of execution of the Cobol
Is This the third or fourth one of these questions that have surfaced here in the last couple of cycles?
Why won't will these people just look and see if the question has been answered yet?
Assembler (subroutines) can identify the format and structure of the dataset prior to actually opening it, like David said:
Quote:
with the correct assembler routines this is possible
.
Look before asking, better to ask for search instructions than to just blindly ask for answers, that is what defines a programmer from a coder....
Joined: 23 Nov 2006 Posts: 19244 Location: Inside the Matrix
Hello Gabriele,
Depending on just what you are going to do with the data once you read it, and that you know the dcb info before you start the process, you might copy the "undefined" file into a fixed length file that has an lrecl larger than the largest possible input record can be. Depending on the range of data lengths you intend to support, you might have a few of these modules so that "everything" doesn't need an intermediate of maximum possible length.
When your program runs, it will have "standard" records to process and then it can proceed. One of the parms you would pass it is the data length is has to work with.
You could make the copy with SORT or IEBGENER.
While this is not as elegant and would cost an extra pass of the data, it would keep all of your code in standard COBOL with no calling of assembler modules that aren't yet on your system or trying to chase/modify control blocks in your program.
Also, if you provide some details as to what your program is supposed to do (other than just read undefined records), we may be able to provide better suggestions.
1.Read the dataset into a variable larger than largest of the record.
2.Then perform a loop from the end of the record looking for value not equal to spaces and retreive the actual length.
eg:
PERFORM VARYING TRX036-OPER-STMT-DATA-LEN
FROM 590 BY -1 UNTIL
FC-NAR-OPER-STMT-VALUES(TRX036-OPER-STMT-DATA-LEN:1)
NOT = SPACES
Joined: 28 Dec 2006 Posts: 6 Location: Italy, Trieste
Hi all,
thanks for your answers.
I adopted the suggestion of Dave, and I have developed an Assembler routines that just open-read-close a generic sequential dataset (the routine receive in input only the DDNAME, all other information (recfm and lrecl are not needs).
I list below the solution (there is the Assembler routines and the Cobol programs used for Unit Test), maybe this code will be used by others members...I spent an entire day for investigate on the web the same question but I wanted a confirm that it was impossible solve the problem in "pure" Cobol (I mean: without Assembler).
For William: I was (very) forced in the time, I had read several threads on this topic, but I was not sure that it's definitly impossible reach my goal without Assembler, I didn't wanted violate the rules of this forum (but I think that I have not done it). In every case I appreciate the cooperative approach and I consider that your comment should appears a personal dispute, and I am definitly not interested to do it. I somebody had developed already a solution for me it was welcome. In every case the next time I will be more precise in my requests.
For the below code:
- the Assembler code use two custom macro (MINIZIA and MFINE) that I "forced" to used for programming convention, just substitute it with the standard start and end of an Assembler routine.
- there are not remarks...it should be in in Italian
- the routine receives in input always P$COMMAND (CL8'OPEN'/CL8'READ'/CL8'CLOSE) and in output always P$RETCOD(C'00' success, other errors) and P$RETMSG(if success it's space, otherwise it is a short error message),
for OPEN in input P$DDNAME, in output P$RECLNG, P$BLKSIZ, p$RECFMT, P$DSNAME,
fore READ in input nothing, in output P$RECLNG, p$EOF ('N' not eof, 'Y' eof)
for CLOSE in input nothing, in output nothing.
The Cobol code used for unit test receives in input from PARM the DDNAME (if it's not defined the DDNAME 'INFILE' is used.
Regards,
Gabriele
Code:
****BASED ON R9****
DCBD DSORG=PS
*
****BASED ON RA****
DSJFCB DSECT
IEFJFCBN
*
****BASED ON R8****
P$PARAM DSECT
P$COMMAND DS CL8 * OPEN/READ/CLOSE
P$DDNAME DS CL8 * DCBDDNAM
P$BUFLNG DS H * DCBBUFL
P$RECLNG DS H * DCBLRECL
P$BLKSIZ DS H * DCBBLKSI
P$RECFMT DS BL1 * DCBRECFM
P$DSNAME DS CL44
P$EOF DS CL1
P$RETCOD DS CL2
P$RETMSG DS CL60
****BASED ON R7****
P$PARAMR DSECT
P$RECVAL DS CL32760
*
GWDSEQ1A MINIZIA RBASE=(R3,R4),DATEUPD=?DATEUPD,TIMEUPD=?TIMEUPD 00002500
*
SAVE (14,12)
BALR 12,0
USING *,12
*
B START
*
JFCB RDJFCB INFILE,MF=L
*
START EQU *
*
ST R1,SAVER1
USING P$PARAM,R8
L R8,0(,R1)
*
MVC P$RECLNG,=H'0'
MVI P$EOF,C'N'
MVI P$DSNAME,C' '
MVC P$DSNAME+1(L'P$DSNAME-1),P$DSNAME
MVC P$RETCOD,=C'00'
MVI P$RETMSG,C' '
MVC P$RETMSG+1(L'P$RETMSG-1),P$RETMSG
*
CLC P$COMMAND,READ
BE C_READ
*
CLC P$COMMAND,OPEN
BE C_OPEN
*
CLC P$COMMAND,CLOSE
BE C_CLOSE
*
MVC P$RETCOD,=C'01'
MVC P$RETMSG(20),=CL20'COMANDO INVALIDO'
MVI P$RETMSG+20,C'('
MVC P$RETMSG+21(8),P$COMMAND
MVI P$RETMSG+29,C')'
B END_KO Return to Calling program
*
END_EOF EQU *
*
MVI P$EOF,C'Y'
B END_OK
*
END_KO EQU *
*
LA RF,1
B END_X
*
END_OK EQU *
*
SR RF,RF
B END_X
END_X EQU *
*
MFINE RC=(RF)
*
***********************************************************************
C_OPEN EQU *
*
CLC P$DDNAME,=CL8' '
BE C_OPEN2
*
USING IHADCB,R9
LA R9,INFILE
MVC DCBDDNAM,P$DDNAME
*
SR R1,R1
LH R1,=F'32760'
STH R1,DCBBUFL
*
RDJFCB MF=(E,JFCB)
*
USING INFMJFCB,RA
LA RA,IN$JFCB
MVC P$DSNAME,JFCBDSNM
*
LA R9,INFILE
USING IHADCB,R9
*
C_OPEN2 EQU *
*
OPEN (INFILE,(INPUT))
TM DCBOFLGS,X'10'
BZ ERR_IO
LA R9,INFILE
*
MVC P$RECLNG,DCBLRECL
MVC P$BLKSIZ,DCBBLKSI
MVC P$RECFMT,DCBRECFM
*
B END_OK
*
***********************************************************************
C_READ EQU *
*
L R1,SAVER1
USING P$PARAMR,R7
L R7,4(,R1)
GET INFILE,P$RECVAL
LA R9,INFILE
MVC P$RECLNG,DCBLRECL
B END_OK
*
***********************************************************************
C_CLOSE EQU *
*
CLOSE INFILE
FREEPOOL INFILE
B END_OK
*
ERR_IO EQU *
*
MVC P$RETCOD,=C'02'
MVC P$RETMSG(20),=CL20'ERRORE I/O'
MVI P$RETMSG+20,C'('
MVC P$RETMSG+21(8),P$COMMAND
MVI P$RETMSG+29,C')'
B END_KO
*
INFILE DCB DDNAME=INFILE, *
DSORG=PS, *
EODAD=END_EOF, *
EXLST=IN$EXLST, *
MACRF=GM, *
SYNAD=ERR_IO
*
D$RECVAL DC CL80' '
*
SAVER1 DC F'0'
*
READ DC CL8'READ '
OPEN DC CL8'OPEN '
CLOSE DC CL8'CLOSE '
*
IN$EXLST DS 0F
DC X'87'
DC AL3(IN$JFCB)
IN$JFCB DS 0F
DC CL176' '
*
END
Joined: 18 Nov 2006 Posts: 3156 Location: Tucson AZ
Gabriele,
Gabriele Soranzo wrote:
For William: I was (very) forced in the time, I had read several threads on this topic, but I was not sure that it's definitly impossible reach my goal without Assembler, I didn't wanted violate the rules of this forum (but I think that I have not done it). In every case I appreciate the cooperative approach and I consider that your comment should appears a personal dispute, and I am definitly not interested to do it. I somebody had developed already a solution for me it was welcome. In every case the next time I will be more precise in my requests.
I apologize for the flame, The synchronicity of your inquiry and several others just rang my bell. You were right, as was David, assembler is the only way this can be done. You have NOT violated any rules of this forum, sometimes the "planets" just don't align as we wish they would.
Again I am sorry; keep asking and contributing to this forum; everything added is potentially helpful to somebody (providing he/she looks )
Bill
BTW, thank you (again) for following up and "closing" the thread.
Joined: 28 Dec 2006 Posts: 6 Location: Italy, Trieste
Hi William,
don't worry...in true also I have criteria for distinguish "coder" and "developer"...just I don't use this forum for describes it ;)...and I understand the relevance of the "planets" !
I hope and I think that we have good benefits using this forum,
I apologize for the flame, The synchronicity of your inquiry and several others just rang my bell. You were right, as was David, assembler is the only way this can be done. You have NOT violated any rules of this forum, sometimes the "planets" just don't align as we wish they would.
