1

I need to write code in python which will convert packed decimal data to zoned decimal or decimal data.If anybody already have written function for it please help me with it.

Thanks in Advance.

user3118158
  • 93
  • 2
  • 7
  • Can you explain where the packed-decimal data is coming from, and why they can't just give you character data? Have you looked at the questions in the packed-decimal tag? – Bill Woodger Oct 08 '15 at 10:14
  • Do you really want to convert Zoned-Decimal; Zoned-Decimal is different between the Mainframe (-121 = 12J) and Ascii-Cobols (121 = 12Q) – Bruce Martin Oct 09 '15 at 22:41
  • 1
    Have a look at the Stingray project (https://sourceforge.net/projects/stingrayreader/?source=directory) it may have conversion routines – Bruce Martin Oct 09 '15 at 22:46
  • In case you do not know Packed-decimal is represented like (123 = x'123c'). It can be processed in two way - Convert it from hex to string (i.e. convert x123c ==> '123c') which can be converted to a number very easily or manipulate the nybbles. – Bruce Martin Oct 09 '15 at 22:51
  • Thank you Bill and sorry for late response.Actually I am getting this Packed decimal data from mainframe data files.I got utility written in java which unpacks PD data into long and packs it again.But this utility will fail while running with COMP-3 packed data.So trying to resolve it. – user3118158 Oct 12 '15 at 06:57

1 Answers1

1

It depends on how are you storing the packaged data, but definitely the best way to do it is coding a COBOL decoder from COMP-3 to numeric or alphanumeric, It will simply be:

  01 WS-YOUR-COMP-DATA PIC 9(xx) COMP-3.
  01 WS-NUMERIC-VAR        9(xx).
...
  PROCEDURE DIVISION.
*  ---> here you read your data
*  ---> then convert it:
       MOVE WS-YOUR-COMP-DATA TO WS-NUMERIC-VAR.
*  ---> and finally, you send it or write it into your dataset/queue/DB.

Although, if you can't nor want to do it, here are some options determined by how are you saving the data:

  • If your data keep its original encoding (IBM EBCDIC) it will be:
import ebcdic

file = open("your-file.txt",'r')

for line in file:
    #Usually the cp1141 is for Austria, Germany, Switzerland characters.
    bline = line.encode('cp1141')
    print(bline.hex())
file.close()

INPUT:

       FD FILEO
           RECORDING MODE IS F.
       01 FILEO-REC        PIC S9(10) COMP-3.
       01 FILEO-REC1       PIC S9(09) COMP-3.
       01 FILEO-REC2       PIC  9(09) COMP-3.
       01 FILEO-REC3       PIC  9(10) COMP-3.
       WORKING-STORAGE SECTION.
       01 FS-FILEO         PIC 9(02).
           88 FS-FILEO-OK              VALUE 00.
       01  WS-COUNT        PIC 9(02).
       01  WS-I            PIC 9(02).
       01 WS-MY-VAR OCCURS 0 TO 10 TIMES DEPENDING ON WS-COUNT
                           PIC S9(10) COMP-3.
       
       PROCEDURE DIVISION.
       MAIN-PROCEDURE.
           MOVE 4          TO WS-COUNT
           MOVE 0123456789 TO WS-MY-VAR(1)
           MOVE 123456789  TO WS-MY-VAR(2)
           MOVE ZEROES     TO WS-MY-VAR(3)
           MOVE -123456789 TO WS-MY-VAR(4)

      *Now we test with each of the four option in every record structure.

OUTPUT with S9(10):

00123456789c
00123456789c
00000000000c
00123456789d

OUTPUT with S9(09):

123456789c
123456789c
000000000c
123456789d

OUTPUT with 9(09):

123456789f
123456789f
000000000f
123456789f

OUTPUT with 9(10):

00123456789f
00123456789f
00000000000f
00123456789f

As you can see, the character printed are exactly the same, but a letter is added at the end. This character could be 'd' if the number is negative signed, 'c' if it is positive signed or 'f' if the number is not signed. These final character only appear as filler when the length is odd. Ex. 9(07) is odd, 7 divided by 2 is 3.5, rounded to 4 and filling the 0.5 space with the letter. Anyway, I saw in my test that OpenCOBOL don't follow this rule and fill ever.

To find your encoding cpxxxx value look in this table or in your emulate terminal or COBOL IDE (openCOBOL indicate it in the bottom right corner).

  • If your data was translated to ASCII or UTF-8 you will have problems to return it to EBCDIC because some bytes could be omitted, anyhow the Python program just will change bline = line.encode('cp1141') for bline = bytes(line, 'ascii')

I let you both programs python and COBOL in this GitHub repo.

Axel Persan
  • 61
  • 7