search for a field in a sequential file

gchitra · Posted: Mon Oct 22, 2007 10:26 pm

I need to develop a pgm to look up for a field in a sequential file. The pgm will do a find for the field in the file and if it finds, the entire record will be written to the output file. It should then skip to the next record and continue the process. Please let me know what READ statement should I use to search for the string or field.

Thanks in advance
Chitra

Craq Giegerich · Posted: Mon Oct 22, 2007 10:35 pm

socker_dad · Posted: Mon Oct 22, 2007 11:39 pm

File Aid is a lot more efficient than a program for this application.

Craq Giegerich · Posted: Mon Oct 22, 2007 11:42 pm

Sort would be good, it's a simple task.

dick scherrer · Posted: Mon Oct 22, 2007 11:46 pm

Hello,

For my $.02, it depends on knowing all of the "rules".

If the only requirement is to select/copy some records with a particular value, the utility route has appeal.

If there are other "things to do" or the process may grow, i'd probably use code.

Craq Giegerich · Posted: Mon Oct 22, 2007 11:49 pm

Sounds to me like a homework assignment.

Craq Giegerich · Posted: Tue Oct 23, 2007 12:15 am

dick scherrer · Posted: Tue Oct 23, 2007 12:31 am

gchitra · Posted: Tue Oct 23, 2007 8:09 am

Thanks for all your replies. I know we can use the READ statement to read a sequential file. Can you tell me how I do the search to look up for the particular field?

Thanks
Chitra

dick scherrer · Posted: Tue Oct 23, 2007 8:18 am

Hello,

If you post your input record format and describe what value(s) you want to search for, we can offer suggestions on how to code it.

Until you explain the rules, we cannot offer much.

balakrishna reddy · Posted: Tue Oct 23, 2007 2:29 pm

Hi Chitra,

As dick said

gchitra · Posted: Tue Oct 23, 2007 10:04 pm

The input file is nearly 30,000 in length and the string that I need to search (Eg: MPD800) will be anywhere after 180 and there could be more than one occurrence of the same string in the same record. So as soon as the first occurrence is found, the pgm should move to the next record. The record that has the matching string should be written to the output file.

Let me know if you need any other information. Thanks for all your help.

dick scherrer · Posted: Tue Oct 23, 2007 10:16 pm

Hello,

To do what you want, i'd suggest you define your input record with a "placeholder" field of 180. Define the remainder of the record as one large field of whatever the length of the record is minus the placeholder (rec-lth - 180).

Code a loop that will start at the beginning of the large field and using reference modification, compare the search value against the data in the large field. Increment the displacement in the large field until you find the value or reach the end of the data.

When there is a match, write the record to the output file. When there is no match, continue with the next read.

As you write the code there may be questions. If there are, post a reply here showing your code and what needs clarification.

gchitra · Posted: Tue Oct 23, 2007 11:49 pm

Thank you very much, Dick.

The search field can be anywhere after the position 180. It can be in 181 or 485. So can you please tell me what we are trying to accomplish by defining a field of 180 in the record layout? Moreover, the file is VB.

Can you also please tell me the keyword to be used to search the string?
I hope it should be something like this:

count=0
READ INPUT-FILE
perform until count = n (this count will be no: of records that should have the search string. if there are more than n records, process only the first n records)
Search for the string (I don't know the keyword for this)
if found,
move to the ws-output-rec
write the output rec
count=count+1.
read the input file
end-perform

Please let me know if it is wrong.

Thanks in advance
Chitra

dick scherrer · Posted: Wed Oct 24, 2007 12:27 am

Hello Chitra,

gchitra · Posted: Wed Oct 24, 2007 6:08 pm

Thanks a lot, Dick. I think this gave me an idea of how to go about with the program. But I may come back to you if I need any futher clarification.

Till then have a nice day and good luck!!!!

Thanks
Chitra

dick scherrer · Posted: Wed Oct 24, 2007 8:43 pm

Hello Chitra,

gchitra · Posted: Wed Oct 24, 2007 11:14 pm

Thank you very much, Dick.

gchitra · Posted: Fri Nov 16, 2007 1:25 am

Hi Dick,

I used the same logic that you suggested: Following is the code:

300-READ-FILE.
READ SKELCMPI INTO WS-SKELCOMPI-REC
MOVE 1 TO WS-DISPL
PERFORM 400-PROCESS-FILE THRU 400-PROCESS-FILE-EXIT
UNTIL MATCH-FOUND OR WS-DISPL = 30145.

300-READ-FILE-EXIT.
EXIT.

400-PROCESS-FILE.
IF WS-SKELCOMPI-LAST(WS-DISPL:6) = WS-UTI-NUM
MOVE 'Y' TO WS-MATCH-SW
PERFORM 500-WRITE-FILE THRU 500-WRITE-FILE-EXIT
COMPUTE WS-COUNT = WS-COUNT + 1
ELSE
COMPUTE WS-DISPL = WS-DISPL + 1
END-IF.

But as the record length (VB) is 30279, it takes almost 30min to find the record that matches the search string. The search string is in the 3rd record (i found it manually), still it takes more time.

I could not think of any better logic. Can you suggest one that increases the performance?

Thanks
Chitra

gcicchet · Posted: Fri Nov 16, 2007 4:30 am

Hi,
as mentioned earlier, FILEAID if you have it, will do what you want.

See code below

Gerry

dick scherrer · Posted: Fri Nov 16, 2007 4:44 am

Hello Chitra,

gchitra · Posted: Sat Nov 17, 2007 12:29 am

Thanks Gerry & Dick. I don't think we have a FILEAID utility, but we do have FILEAID option in the ISPF menu panel.

Dick, you are right. Most of the time is spent in looking thru the rest of the file. I manipulated the input file so that first and second record has the search string, only to check if the output is as expected. But as the search starts from byte 125 and moves one byte at a time and the string appears somewhere after byte 450, the job after running for almost 3hrs, finally abended with S322.

I can not think of anything that could better the performance. I would appreciate, if any of you can suggest a better logic.

Thanks
Chitra

dick scherrer · Posted: Sat Nov 17, 2007 1:55 am

Hello Chitra,

I may have missed the answer to this, but how many records will be in the file?

When you say 3 hours, i suspect that is wall-time. How much cpu was used?

Which sort product is used on your system?