I have a file with the following contents..say

123 abc 90and /[abd] 1009

from which i only need numbers to be printed.
like 123 90 1009

using any shell command.

Thanks.

$ perl -ne 'print $1," " while /(\d+)/g; print "\n"'

Thats nice Pludi. But I'd prefer that to be done in shell itself
not possible with shell command,awk or sed?

Possible: yes. Shorter? Probably not. And the Perl command integrates into a pipe or other shell sequence just as well as awk or sed.

> echo "123 abc 90and /[abd] 1009"                                          
123 abc 90and /[abd] 1009

> echo "123 abc 90and /[abd] 1009" | tr -d "[:alpha:]|[:punct:]"
123  90  1009

> echo "123 abc 90and /[abd] 1009" | tr -d "[:alpha:]|[:punct:]" | tr -s " "
123 90 1009

No need to invoke an external program such as awk, sed or perl!

#!/usr/bin/ksh93

str="123 abc 90and /[abd] 1009"
echo "$str"

print ${str//[^0-9]/ }

$ ./demo
123 abc 90and /[abd] 1009
123 90 1009
$

All,

What if i have a huge file with nxn matrix of rows and columns and still i would like to have only numbers from every row ? I mean where no. of rows and columns are unpredictable. Any suggestions for that please.

rgds,

Srini