Mr. Walter J. Murray,
I've read many of your posts, and I've always considered you to be an
authority on HP3000 Cobol, and still do.
Shoot'in from the hip, I was not aware that you could pass the random
function a seed value, and I stand corrected! Thank you, really!
Thats probably why I preferred to use RAND1 and RAND.
Now my ASCII is showing, I might as well run with it.
So, I'll need to look at the manual more often, because the generation
of random numbers is far too important to be left to chance.
;-)
As to the source of the random number I will still disagree with you,
and mostly because I use RAND1 and RAND, not the function..
When using RAND1, the value is derived from a quantity changed every
millisecond by the MPE system timer (Clock milliseconds). Additionally,
looking at my source example you'll see that each iteration of PLAY-LIST
makes another call to RAND1. This re-seeding is making the numbers more
random, and the way it is written you will get the same results if
PLAY-LIST is performed twice within one millisecond, yes, now the
machines are faster, and I should change my ways. Furthermore, it is a
fact that calling RAND 100 times, starting with the same initial seed,
will generate the same list of 100 numbers, thats what I mean by
patterns, or a random set of related numbers, or a generation of
numbers. To start a new generation, I re-seed using RAND1.
I agree that I have seriously misunderstood the Cobol function "random",
and now actually looking in the manual, I find the Cobol function does
use RAND1 and RAND, not the other way around as you stated, when you said:
"The techniques Michael describes using RAND and RAND1, if I am
understanding him correctly, are quite possible using the standard
RANDOM function in COBOL."
The Cobol manual states:
"For details on the RANDOM function, refer to the /Compiler Library
Reference Manual/."
In the compiler library manual it refers to RAND1 & RAND procedures, and
RAND1 does get it's value from the system timer.
OK, shooting from the hip again, I'm going to guess that when you use
"function random" with value 0, then the function (I would hope)
actually calls RAND1 to get a fresh seed to start a new generation.
Can I do this using the "function random"?
I know I can do it using RAND1 and RAND.
Cheers,
Mike.
Walter J. Murray wrote:
> Greetings,
>
> I'm afraid there are some inaccuracies in Michael's posting.
>
> First, the RANDOM function in COBOL II/iX does not use the "clock
> milliseconds" as the seed, or in any other way. You either specify the
> seed value yourself, or let it default. In COBOL II/iX, the default
> seed value is zero.
>
> The suggestion to use the return value from RANDOM to re-seed it is
> essentially what COBOL II/iX does for you automatically.
>
> The statement that you don't have control of the seed value is
> incorrect. You have the option of specifying an argument to RANDOM,
> which seeds it (or re-seeds it); or not specifying an argument, which
> causes it to return the next pseudo-random number in the sequence
> determined by the previous seed. If you never specify a seed, it's the
> same as specifying an initial seed value of zero.
>
> The techniques Michael describes using RAND and RAND1, if I am
> understanding him correctly, are quite possible using the standard
> RANDOM function in COBOL.
>
> The implication that you will get the same result if you call RANDOM
> twice within a millisecond is incorrect. As I said, the RANDOM function
> doesn't use any system timer value.
>
> If I want a reasonably unpredictable sequence of pseudo-random numbers,
> I will typically use a seed value derived from the return value from a
> call to FUNCTION CURRENT-DATE. This is what I do, for example, in a
> routine I wrote to generate a random strong password.
>
> Bottom line: Save yourself some trouble and use the RANDOM function
> provided by the COBOL language. It's standard, portable COBOL.
>
> Walter
>
> Walter J. Murray
>
>
> -----Original Message-----
> From: HP-3000 Systems Discussion [mailto:[log in to unmask]] On
> Behalf Of Michael Anderson
> Sent: Wednesday, September 15, 2010 4:53 PM
> To: [log in to unmask]
> Subject: Re: [HP3000-L] RANDOM NUMBER..
>
> Such library functions will repeat patterns using the clock milliseconds
> as the initial seed. This "ramdom" function may provide enough
> randomness
> for some, but not all.
>
> To achieve more randomness with the standard system clock algorithm
> you'll
> need to set the initial seed value first. Then use that seed value to
> generate your first random number, then push the first random number
> back
> into the seed, thus making the seed more random than just using the
> clock.
> Then generate another random number using the random seed.
> Correct me if I'm wrong. To the best of my knowledge, when using the
> single function, you do not have control of the seed value. So when a
> pattern begins to repeat you're just SOL.
>
> My point is: On MPE there are two intrinsic functions available to give
> you more control of your random number generator. They are RAND1 and
> RAND.
> You can use these two routines in various different ways to generate a
> random seed, which is then passed to RAND to generate the random number.
> Not only can you use the random number as a seed for the next random
> number, but to break any pattern you can use any three digits of the
> random number to pause for a random number of milliseconds, then get a
> new
> seed value from the system clock by calling RAND1, then continue in the
> loop. None of this is possible using the standard "random" function.
>
> Example:
> RAND1 returns the seed to be passed to the RAND intrinsic. RAND1 is
> based
> on the system timer (clock milliseconds). The random function works the
> same way, therefore you will in up with the same random number if call
> it
> twice within 1 millisecond. Worse, if you are generating many random
> numbers you will start to see a pattern.
>
> About 20 years ago I was asked to write a random employee list, using
> COBOL74 on 16 bit MPE V. I would initially call RAND1 to get the first
> seed. From that point on I used the result from RAND as the next seed,
> and
> so on.
>
> To test this routine and make sure it will work for a truly random
> subset
> of employees, I wrote a little program to gen lotto numbers, and here it
> is:
>
> ====================================
> IDENTIFICATION DIVISION.
> PROGRAM-ID. RANNUM.
> AUTHOR. MICHAEL ANDERSON.
> ****************************************************************
> * - RANNUM - is a program developed to test and display the *
> * unbiased selection of numbers, that the HP random *
> * number generator will produce. *
> * The original need for this was the random employee*
> * list. *
> ****************************************************************
>
> DATE-COMPILED.
> ENVIRONMENT DIVISION.
> CONFIGURATION SECTION.
> SOURCE-COMPUTER. HP3000.
> OBJECT-COMPUTER. HP3000.
> SPECIAL-NAMES.
> CONDITION-CODE IS CC.
> DATA DIVISION.
> WORKING-STORAGE SECTION.
> 01 RANDOM-SEED PIC S9(9) COMP VALUE 0.
> 01 RANDOM-RETURN PIC S9(9) COMP VALUE 0.
> 01 DISP-N PIC 9(9) VALUE ZERO.
> 01 FILLER REDEFINES DISP-N.
> 02 FILLER PIC 9.
> 02 FILLER PIC 99.
> 02 FILLER PIC 99.
> 02 FILLER PIC 99.
> 02 RANDOM-PICK PIC 99.
>
> 01 DISP-N18 PIC ZZZ,ZZZ,ZZZ,ZZZ,ZZZ,ZZ9.
> 01 TBL-IDX PIC S9(4) COMP VALUE 0.
> 01 IDX PIC S9(4) COMP VALUE 0.
> 01 X PIC S9(18) COMP VALUE 0.
> 01 DOLLARS PIC S9(18) COMP VALUE 0.
> 01 SIXFLAGS PIC S9(4) COMP VALUE 0.
> 01 PLAYS PIC 99 VALUE 0.
> 01 F PIC S9(4) COMP VALUE 0.
> 01 T PIC S9(4) COMP VALUE 0.
> 01 NUMF PIC 99 VALUE 0.
> 01 NUMT PIC 99 VALUE 0.
>
> 01 FROM-NUM-TABLE.
> 02 FROM-NUMBERS.
> 03 FROM-NUM OCCURS 6 TIMES PIC 99.
>
> 01 TO-NUM-TABLE.
> 02 TO-NUMBERS.
> 03 TO-NUM OCCURS 6 TIMES PIC 99.
>
> 01 NUMBER-TABLE.
> 02 RANDOM-NUMBERS.
> 03 RANDOM-NUMBER OCCURS 6 TIMES PIC 99.
>
> PROCEDURE DIVISION.
> 0000-RANNUM.
> DISPLAY "RANDOM NUMBERS.".
> DISPLAY "--------------------".
> MOVE ZERO TO PLAYS SIXFLAGS DOLLARS.
> PERFORM PLAY-LIST UNTIL DOLLARS > 10.
> GOBACK.
>
>
> PLAY-LIST.
> INITIALIZE RANDOM-NUMBERS.
> MOVE ZERO TO TBL-IDX.
> CALL INTRINSIC "RAND1" GIVING RANDOM-SEED.
> *
> * GET NUMBERS
> PERFORM VARYING X FROM 1 BY 1
> UNTIL TBL-IDX IS EQUAL TO 6
> IF DISP-N > 0
> MOVE DISP-N TO RANDOM-SEED
> END-IF
> CALL INTRINSIC "RAND" USING RANDOM-SEED
> GIVING RANDOM-RETURN
> MOVE RANDOM-RETURN TO DISP-N
>
> IF ( RANDOM-PICK > 0 AND < 55 ) AND
> ( RANDOM-PICK <> RANDOM-NUMBER(1) AND RANDOM-NUMBER(2)
> AND RANDOM-NUMBER(3) AND RANDOM-NUMBER(4)
> AND RANDOM-NUMBER(5) AND RANDOM-NUMBER(6) )
> ADD 1 TO TBL-IDX
> MOVE RANDOM-PICK TO RANDOM-NUMBER(TBL-IDX)
> END-IF
> END-PERFORM.
> *
> * SORT NUMBERS
> IF RANDOM-NUMBERS <> TO-NUMBERS
> ADD 1 TO DOLLARS
> MOVE RANDOM-NUMBERS TO FROM-NUMBERS
> INITIALIZE NUMF NUMT TO-NUMBERS X T F
> PERFORM VARYING X FROM 1 BY 1 UNTIL X > 54
> PERFORM VARYING F FROM 1 BY 1 UNTIL F > 6
> MOVE FROM-NUM(F) TO NUMF
> IF NUMF = X
> ADD 1 TO T
> MOVE FROM-NUM(F) TO TO-NUM(T)
> END-IF
> END-PERFORM
> END-PERFORM
> *
> * DISPLAY NUMBERS
> MOVE TO-NUMBERS TO RANDOM-NUMBERS
> DISPLAY RANDOM-NUMBER(1)
> ", " RANDOM-NUMBER(2)
> ", " RANDOM-NUMBER(3)
> ", " RANDOM-NUMBER(4)
> ", " RANDOM-NUMBER(5)
> ", " RANDOM-NUMBER(6).
>
>
>
>
>
>
>
>
>> The manual shows the format to be:
>>
>> 77 my-random pic v99999.
>>
>> Compute my-random = function random (30).
>>
>> My problem.. I keep getting a compile error on the word 'function'.
>>
>> 00604 011010 32 356 Q UNDEFINED DATA NAME FUNCTION.
>>
>> I'm on the latest and greatest and LAST version of the operating
>>
> system.
>
>> C.75.00.
>>
>> I've tried taking the word 'function' out, along with various other
>> things.
>>
>> From page 10-49/50 of the manual.
>>
>> Mr Kim Borgman
>> IT mgr, RNDC of Indiana.
>>
>> * To join/leave the list, search archives, change list settings, *
>> * etc., please visit http://raven.utc.edu/archives/hp3000-l.html *
>>
>>
>
> * To join/leave the list, search archives, change list settings, *
> * etc., please visit http://raven.utc.edu/archives/hp3000-l.html *
>
> * To join/leave the list, search archives, change list settings, *
> * etc., please visit http://raven.utc.edu/archives/hp3000-l.html *
>
>
* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *
|
