Random Number Generator

[SiouxCityElvis]

Anyone ever created a Random Number Generator?
I'm using RMCOBOL-85 version 8 on Linux. I thought maybe the newest release might have a Random Number Generator sub-program(C$name), but I don't think so...

Thanks.
-David

 

[k5tm]

Try this link:

http://groups.google.com/groups?dq=...e=UTF-8&selm=35a2ab06.0@news1.ibm.net&rnum=20

Just a word to the wise: random number generation is not easy.

Tom Morrison

http://www.liant.com

 

[WMK]

You may want to ask your "vendor of choice" (listening Tom) exactly WHEN (if ever) they plan on implementing the Intrinsic Functions module which has been a part of the REQUIRED "High-Level" FIPS COBOL definition since 1990 (or so) and has been an optional part of the High-level ISO/ANSI Standards since then.

It is a REQUIRED part of the ANSI/ISO 2002 COBOL Standard

...

And it includes a RANDOM function (which is only PSEUDO-random - if you don't "vary" the seed value)

 

[webrabbit]

In any multi-tasking system such as Windows NT^tm or Unix, the low-order bits of the time clock make and excellent random seed.

 

[k5tm]

Bill,

Actually, David is the first customer in my recollection that has asked for RANDOM. Our customer base has been notably quite about any intrinsic functions; they have other concerns that more directly impact the marketability of their software. I have had a discussion with the compiler gurus about REVERSED, my personal favorite.



Tom Morrison

http://www.liant.com

 

[SiouxCityElvis]

I saw some stuff out there for different types of COBOL on the web about Random Number generators i.e. "seed", "pseudo-random" as mentioned above by WMK.

Seems simple to do in Java. Just curious why the Intrinsic Functions module or other sub-program (C$...) does not include random number generation in RM-COBOL. Random Number generation seems to be a pretty common necessity. Learned about it in classes back in my school days and have seen it used in Java in other work places.

On a more positive note, I did get to use C$DELAY yesterday and was happy about its results!
I'll check that link out. Thanks.

Webrabbit:
What's the efficiency of using the low order bits of the system's clock for the seed in a random number? What if you have 2 transactions, for example, hitting your application one right after the other within milliseconds of each other? For each you need a distinct random number generated. Since they are within milliseconds of each other, wouldn't you end up with the same random number on each?

 

[SiouxCityElvis]

Tom,
Do you know how to access the Linux/Unix clock's lower bit values as suggested by Webrabbit above from my COBOL program?

I used that link you suggested. It has various Math functions in it including the RN Generator, but I'm required to send it a value for setting the format value of the seed variable to be used in the Random function later.

So, in order to make that seed a random number, I thought maybe Webrabbit's suggestion would work.

Thanks.
-David

 

[k5tm]

A common characteristic of pseudo-random number generators is the ability (necessity) to specify the initial seed value. Often, one wants random number sequences that may be reliably repeated!

I do not know, off the top of my head, what system call returns the system clock.

May I ask...what are you trying to accomplish with a random number?

Tom Morrison

http://www.liant.com