how to pick random file name from a list of filenames in a text file.

pdklinux79 · 06-19-2008, 04:13 PM

My problem is :

ls -l /data1/cur/*.cdt >>fi.txt

this file has list of column names

i want some way to pick a random name from this fi.txt.

how to do it in bash shell scripting?

thanks

unSpawn · 06-19-2008, 05:34 PM

Could try

Code:

a=($(< fi.txt)); rand () { i=`od -cN10 /dev/random | sha1sum`; i=${i// */}; i="${i//[^0-9]/}"; n=${#a[@]}; n=${#n}; i="${i:0:$n}"; [ $i -le ${#a[@]} ] && echo "${i}" || rand; }; for (( n = 0 ; n < 10; n++)); do echo ${a[${RANDOM:2:3}]}; done

. Looks a bit wack but that's because $RANDOM isn't as random as it should be. Hmm. Spose an easier method will come your way RSN ;-p

Mr. C. · 06-19-2008, 05:35 PM

Code:

awk "NR==$(($RANDOM % $(wc -l fi.txt| awk '{print $1}')))" fi.txt

or to break it down into multiple stages:

Code:

nlines=$(wc -l fi.txt| awk '{print $1}')
rand=$((RANDOM % $nlines))
awk "NR==$rand" fi.txt

unSpawn · 06-19-2008, 05:38 PM

Yeah, *way* easier.

rjlee · 06-19-2008, 06:15 PM

Quote:

Originally Posted by pdklinux79

My problem is :

ls -l /data1/cur/*.cdt >>fi.txt

this file has list of column names

i want some way to pick a random name from this fi.txt.

how to do it in bash shell scripting?

thanks

Just take the first line when sorted randomly:

Code:

sort -R fi.txt | head -n 1

You can pipe that through cut if you want just the filename.

Of course, this isn't truly random since it hashes the keys, so you will get the same chance of each file regardless of the number of times it occurs. Also, it's only pseudorandom, as are all answers not seeding from /dev/urandom.

Mr. C. · 06-19-2008, 06:18 PM

Quote:

Originally Posted by rjlee

Just take the first line when sorted randomly:

Code:

sort -R fi.txt | head -n 1

Ouch! Sorting is one of the most expensive operations possible.

kenoshi · 06-19-2008, 06:54 PM

Quote:

Originally Posted by Mr. C.

Code:

awk "NR==$(($RANDOM % $(wc -l fi.txt| awk '{print $1}')))" fi.txt

or to break it down into multiple stages:

Code:

nlines=$(wc -l fi.txt| awk '{print $1}')
rand=$((RANDOM % $nlines))
awk "NR==$rand" fi.txt

This will work only if $rand doesn't return a 0, which it does eventually. Also, you don't need the awk for nlines.

E.g.

Code:

rand=$((RANDOM % $(wc -l < fi.txt)))
if [ "$rand" = "0" ]; then ((rand++)) ; fi
awk "NR==$rand" fi.txt

Mr. C. · 06-19-2008, 07:13 PM

The redirection is a clever way to avoid the filename as part of the output. Good call.

I ignored the 0 case, but shouldn't have. Instead, because the line numbers are 1 based vs. 0 based, we can bias by one:

rand=$((1 + (RANDOM % $nlines)))

With 10 lines, mod will give 0 to 9. Biasing by 1, yields 1 to 10, exactly what we want.

pdklinux79 · 06-20-2008, 11:38 AM

Thanks a ton guys... Its working and my friday at work is a cool breeze. Have agood weekend.

rjlee · 06-20-2008, 02:46 PM

Quote:

Originally Posted by Mr. C.

Ouch! Sorting is one of the most expensive operations possible.

Ah, but it's easier to read.

And depending on the size of the list of files, speed may not be an issue. Slow but readable and reliable is sometimes better than highly optimised, and if speed were essential then I would be using C or assembler…