Bash

dtmc · 03-14-2012, 10:01 AM

I've written a small script to generate random words, which seems to work reasonably well except for the fact that it's generating far too many 'a's (that's the single letter 'a', not words starting with 'a'. It's about 1 'a' in every 6 words generated). Could someone have a look and suggest where I may be going wrong?

For information, if it's relevant, I'm using Bash 4.2.20 on Arch.

Many thanks,

David Shaw

#!/bin/bash

if [ ! -e 'wordlist' ]; then
aspell dump master | sort | uniq > wordlist
fi

LINES=`cat 'wordlist' | wc -l`

for WORD in 1 2 3 4; do
RAND=`od -N4 -t u4 /dev/urandom | head -1 | cut -d' ' -f2`
LINE=$(( RAND % LINES + 1 ))
head -$LINE 'wordlist' | tail -1
done

jhwilliams · 03-14-2012, 10:35 AM

The problem is coming when RAND is set empty, and so LINE gets the value 1. ( (0 % anything) + 1 is always 1. ) The value of the wordlist at line 1 is 'a'.

Code:

+ RAND=
+ echo 'RAND is '
RAND is

You should scrutinize this line -- probably break it up a bit -- and determine under what conditions that behaviour is ocurring.

Code:

#Needs fixing
RAND=`od -N4 -t u4 /dev/urandom | head -1 | cut -d' ' -f2`

David the H. · 03-14-2012, 12:10 PM

Please use [code][/code] tags around your code and data, to preserve formatting and to improve readability. Please do not use quote tags, colors, or other fancy formatting.

bash has a built-in RANDOM environment variable, BTW, so unless you need industrial-strength randomness you can remove the complex external command.

You can also re-seed RANDOM at any time by simply setting it to any string e.g.

Code:

RANDOM=$( date '+%s' )

Edit: I just remembered this as well:

Code:

shuf -n 4 wordlist

This will output four random lines from the file.

Edit2: And if you want to stick with the script, use this to print a line number instead of head+tail:

Code:

sed -n "${LINE}p" wordlist

dtmc · 03-14-2012, 02:49 PM

Many thanks for the replies :-)

@jhwilliams
D'oh! Never even considered that. I knew it would be *something* simple

@David the H.
The problem with using the built in is that it only returns 32768 possible values and there's well over 100,000 entries in wordlist but other than that, yeah, it's far easier. shuf looks like the way to go, though and I guess I'll end up using it - but I still want to fix my script as a learning exercise.

Again, to both, many thanks

David Shaw