Get non-monotonically increasing fields in Bash - bash

Question

Let's say I have a file with multiple columns and I want to get several fields but they may be not in increasing order. Field indexes are in an array, indexes can be in any order or not order at all and the number of indexes is unknown, for example:
arr=(1 3 2) #indexes, unknown length
echo 'c1 c2 c3' | cut -d " " -f "${arr[*]}"
The output of that is
c1 c2 c3
but I want
c1 c3 c2
So it seems cut is sorting the fields before reading them, I don't want that. I am not restricted to cut, any other command can be used.
However, I am restricted to this, rather old, version of bash:
GNU bash, version 2.05b.0(1)-release (i586-suse-linux)
Copyright (C) 2002 Free Software Foundation, Inc.
EDIT Solved thanks to Benjamin W and Glenn Jackman
echo "1 2 3" | awk -v fields="${arr[*]}" 'BEGIN{ n = split(fields,f) } { for (i=1; i<=n; ++i) printf "%s%s", $f[i], (i<n?OFS:ORS) }'
It is important to reference the array with '*' instead of '#'.

Answer 1

This may or may not work with bash 2.05:
arr=(1 3 2)
set -f # disable filename generation
while read line; do
set -- $line # unquoted: taking advantage of word splitting,
# store the words as positional parameters
for i in "${arr[#]}"; do
printf "%s " "${!i}" # indirect variable expansion
done
echo
done < file
Or, perl
$ cat file
c1 c2 c3
$ perl -slane '
BEGIN {#a = map {$_ - 1} split " ", $arr}
print join " ", #F[#a]
' -- -arr="${arr[*]}" file
c1 c3 c2

Answer 2

Using awk
$ arr=(1 3 2)
$ echo 'c1 c2 c3' | awk -v arr="${arr[*]}" '
BEGIN {
split(arr, idx," ");
}
{
for(i=1; i<=length(idx); ++i)
printf("%s ",$idx[i])} ;
END {
printf("\n")
}
'
First, split arr by ' ' and assign to idx
Then, print based on each index i

Answer 3

Using awk:
$ cat file
a b c d
a b c d
a b c d
a b c d
$ awk -v ord="1 4 3 2" 'BEGIN { split(ord, order, " ") }
{
split($0, line, FS)
for (i = 1; i <= length(order); ++i)
$i = line[order[i]]
print
}' file
a d c b
a d c b
a d c b
a d c b
The order is given by the ord variable passed on the command line. This variable is assumed to hold as many space-separated values as are available in the input file.
In the BEGIN block, an array, order, is created from ord by splitting it on spaces.
In the default block, the current input line is split into the array line on FS (whitespace by default). The fields are then rearranged according to the order array and then the re-constructed line is printed out.
No test is made that the passed-in value in ord is sane. If the input has N columns, it must contain all the integers from 1 to N is some order.

Answer 4

Abstract your read from your print so you can name the parts and order them accordingly.
$: cat x
c1 c2 c3
c1 c2 c3
c1 c2 c3
c1 c2 c3
c1 c2 c3
$: cut -f 1,3,2 x |
> while read a b c
> do printf "$a $c $b\n"
> done
c1 c3 c2
c1 c3 c2
c1 c3 c2
c1 c3 c2
c1 c3 c2
This puts the read loop into the bash interpreter, which isn't as fast as a binary, but doesn't require another tool that you were already using.
I don't see much point in using awk if you have perl, so if the file is big enough you need a faster solution, try this:
perl -a -n -e 'print join " ", #F[0,2,1],"\n"' x
Assumes a lot, and adds a space before the newline, but should give you a working place to start.