How to make a while-loop with awk.

Lets say I have a variable number of columns:

1 4 3
2 4 4
3 5 3

Now I want to add all elements of column 1, 2 and 3 and then divide them by the number of elements in each.

columns= the number of columns which is given as an argument to my script

i = the column we are adding for the moment

this is some pseudo code for what I want to be done:

while [columns > 0]
do
awk '{sum+=$i} END {print sum/NR "\t"}'
i++
column--
done

So when executed the finished result will be:

2 6.5 3.333333

But I don't get my while loop to function. Please help me. This is my code:

#############################
while [ "$columns" -gt 0 ]
do
awk_cmd=`awk -v i=$1 '{sum+=$i} END {print sum/NR}'`
echo "$data | $awk_cmd"
i=`expr $i + 1`
columns=`expr $columns - 1`
done
#############################

awk -f temp.awk tempfile

{ for(i=1;i<=NF;i++) { sum[i]+=$i } }
END { for (i in sum) {printf("%f\t", sum[i]/NR) } }

Btw, the second column should not be 6.5, but 4.33

ranj,
when iterating through an array 'sum' with the index of 'i', the order of the iteration is indeterminant, i.e. don't assume that 'i' will go from '1' to 'NF'.

I didnt know that. So, you mean the values could be printed in any order? Will keep that in mind!!!

baghera,
I believe we've already gone through this exercise once in one of the previous threads. What is it exactly that you're not trying to do differently?
If you're try to calculate the average value per column for all the rows - this solution has also been provided as part of the previous thread.
Try to understand the previous solution and/or adjust it to your 'new' [???] requirement.

The difference between now and then is that I need it to be a bash/bourne script. The previous thread was for a awk-program. But I had to change it to be a bash/bourne shell script. To be honest, I don't know the difference really, but I got some uggly looks from my friend when I showed him the awk-program that some of you helped me with. So I turned back to bourne/bash.

So my script so far is:

cat $1 | grep "results" | grep "get numbers" | xargs -n $2

It is somewhat more advanced then the command above, but this is only so you get the whole picture. It produces columns with values from files named $1 and the numbers of columns are determined by $2.

Now, since $2 is a variable I don't know how many columns there is going to be. It is not going to be many, lets say around 2-10 should suffice. I need, as I previous stated go through each column adding them and then dividing them with the number of elements in each column. So if the output from the following command above is:

1 4
2 5
3 6
4 7
5 8

The result should be:

3 6

Which is the average of each of the columns.

I've done something like this, but it doesn't seem to work,

while [ $j > 0 ]
do
cat list | awk '{sum+=$i} END {print sum/NR}'
i=$[$i+1]
j=$[$j-1]
done

awk seems to have problem with $i. But I need this to be a variable but awk says:

awk: illegal field $(), name "i"
input record number 1, file
source line number 1

I solved the problem:

i="1"
j=$2

while [ "$j" -gt  0 ]
  do
  echo "$data" | awk -v k=$i '{sum+=$k} END {print sum/NR}'
  i=$[$i+1]
  j=$[$j-1]
done

data='cat myList | xargs -n $2'