Duplicate numbers using $RANDOM in Bash

Question

I wrote a script with bash in while loop. The code:

number=0
while [ 1500 -gt $number ]
do
  var="abcdefghijklmnopqrstuvxyz"
  letter1="${var:$(( RANDOM % ${#var} )):1}"
  letter2="${var:$(( RANDOM % ${#var} )):1}"
  a=$RANDOM
  b=$RANDOM
  c=$(( $a * $b))
  echo "$letter1$letter2 $c" >> a.txt
  number=$(( 1 + $number ))
done

but now I see duplicate numbers in result:

Result:
ab 15474
at 15474
yh 15474
gd 15474
re 18696
jg 18696

The numbers are duplicate.

I guess the $RANDOM changes after a invariant time and my script starts again the the while loop faster than the $RANDOM changes.

Can you help me an other randomization way?

Not the most efficient, but you could read from `/dev/urandom` and filter for only-letters — knittl, Aug 29 '19 at 08:25
Was RANDOM changed in any way before this code is executed? `man bash` warns that while RANDOM should produce a new random number each time it is referenced, it will lose this special property if it is unset at some point. Also, I [can't reproduce](https://ideone.com/l0AKzs) — Aaron, Aug 29 '19 at 08:44
@knittl how can i use the /dev/urandom . im not pro at linux. i use debian — Ali emaditaj, Aug 29 '19 at 08:55

j23 · Accepted Answer · 2019-08-29T10:19:03.817

3

I have tried your script and it worked fine for me.

There are many ways to generate random number in bash. One is using the /dev/random special device file. The /dev/random uses the noise collected from the device drivers and other sources to generate random data. The od (octal dump) command can extract a number of bytes and displays their decimal equivalent.

od -A n -t d -N 1 /dev/urandom

Here, -t d specifies that the output format should be signed decimal; -N 1 says to read one byte from /dev/urandom.

One other way is using the jot command:

jot -r 10 1 1000

Here we generate 10 numbers between 1 and 1000. -r specifies to generate random number.

edited Aug 29 '19 at 10:19

answered Aug 29 '19 at 10:13

j23

3,139
1
6
13

how to install jot . i didnt see that in comand list and apt install list – Ali emaditaj Aug 29 '19 at 10:37
Try this `sudo apt-get install athena-jot ` – j23 Aug 29 '19 at 10:41
how to save this random result to a Variable(i used google translate for the Variable )(the $ in bash) – Ali emaditaj Aug 29 '19 at 10:59
@Aliemaditaj you can use command substitution (using `resultVariable=$(command)` ) to write the result to the variable. For example, `a=$(jot -r 2 1 1000)` or `a=$(od -A n -t d -N 1 /dev/urandom)` will store the results in `a` and then `echo $a` will print the result. – j23 Aug 29 '19 at 11:09
it dont works at high numbers . please see this screenshot: https://ibb.co/5LDzBbq – Ali emaditaj Aug 29 '19 at 14:13
Pictures of text are pretty universally loathed. Maybe at this point create a new question with your *actual* requirements instead of heaping on follow-up questions in comments. – tripleee Aug 29 '19 at 14:25

oliv · Answer 2 · 2019-08-29T11:12:24.953

1

You could use tr together with /dev/urandom:

tr -dc 'a-z' </dev/urandom | head -c 2; echo

Set the number of characters you want with the -c option of the head function.

To generate random digit, use another set in the tr command:

tr -dc '0-9' </dev/urandom | head -c 4; echo

edited Aug 29 '19 at 11:12

answered Aug 29 '19 at 10:55

oliv

12,690
25
45

@tripleee I don't get your point. Do you mean that the output isn't random? or that there isn't digit? – oliv Aug 29 '19 at 11:00
That there are no digits. You have solved the first half of the question but not the second. – tripleee Aug 29 '19 at 11:07
@tripleee ok, kind of misunderstood the question, answer updated – oliv Aug 29 '19 at 11:13

score 0 · Answer 3 · answered Aug 29 '19 at 10:53

@j23 has given a very reasonable operational answer. By way of possible explanation of the behaviour you observed, the number you are printing is not a single $RANDOM value, but the product of two such values. Pairs of consecutive outputs of a pseudorandom number generator (PRNG) are not necessarily as independent as you might like. E.g., Matlab randn had a correlation problem in 2006 (arXiv:math/0603058).

tripleee · Answer 4 · 2019-08-30T03:50:34.093

Something like this?

#!/bin/bash
export LC_ALL=C
for((i=0; i<1500; ++i)); do
    IFS='' read -n 4 -d '' bytes
    # https://stackoverflow.com/questions/28476611/ord-and-chr-a-file-in-bash
    printf -v a %u "'${bytes:2:1}"
    a=$((a%255))
    printf -v b %u "'${bytes:3:1}"
    b=$((b%255))
    printf "%s %s\n" "$(
        tr '\000-\011\013-\140\173-\377' 'a-za-za-za-za-za-za-za-za-z' <<<"${bytes:0:2}"
    )" $((${a#-}*${b#-}))
done</dev/urandom

The %u conversion oddly creates really big numbers when the character code is above 0x80, and the modulo 255 of that creates a negative number, so I had to do some nonobvious workarounds to fix that. Perhaps you could come up with a less contorted way to unpack two bytes into an unsigned 15-bit number.

Here's an update which gets results in the range 200000-1000000 in the second column. It needs two additional random bytes and then performs a modulo and addition on the result to bring it into the correct range. This is outside the reach of the Bash built-in arithmetic so I used bc instead.

#!/bin/bash
export LC_ALL=C
for((i=0; i<1500; ++i)); do
    IFS='' read -n 6 -d '' bytes
    # https://stackoverflow.com/questions/28476611/ord-and-chr-a-file-in-bash
    printf -v a %u "'${bytes:2:1}"
    a=$((a%255))
    printf -v b %u "'${bytes:3:1}"
    b=$((b%255))
    printf -v c %u "'${bytes:4:1}"
    c=$((c%255))
    printf -v d %u "'${bytes:5:1}"
    d=$((d%255))
    printf "%s %s\n" "$(
        tr '\000-\011\013-\140\173-\377' 'a-za-za-za-za-za-za-za-za-z' <<<"${bytes:0:2}"
    )" $(bc <<<"((${a#-}*${b#-}*${c#-}*${d#-})%800000)+200000")
done</dev/urandom

This is getting pretty complex, though. If Bash is not essential for you, try this Python 3 script.

from random import choice, randrange
from string import ascii_lowercase

for r in range(1500):
  print('{0}{1} {2}'.format(
    choice(ascii_lowercase),
    choice(ascii_lowercase),
    200000+randrange(999999800000)))

I guess you want bigger numbers than 15-bit actually, but your examples all look like they are less than 32768. What is the expected range actually? — tripleee, Aug 29 '19 at 10:59
Just use `printf -v variable` in the last statement in the loop as well. — tripleee, Aug 29 '19 at 14:22

Duplicate numbers using $RANDOM in Bash

4 Answers4