Interfacing bash with awk:

I am a beginner in awk and was looking at tweaking some bash script at work that interfaces with awk utility to parse the output of some system files. I made my changes based on my understanding and then better sense prevailed, and so I thought I would write a simple test program to try out the same logical pattern. But to my disappointment, the test program didn't work as expected. I am reproducing below the bash script and the awk utility script that I had coded, along with the error that I am getting. Appreciate your thoughts.

bash script
-----------

#!/bin/bash

string=$'a\nb\nc\nd\ne'
echo "$string"

awk -v input=${string} -f post.awk < file_input.txt > file_output.txt


awk script
----------

#!/bin/gawk

BEGIN {
  
  getline tmp
  print tmp > $3

}

END{

 print $1 > $3

}

In the awk script, I am trying to read the single line "Only one line" inside file_input.txt, and write it to file_output.txt inside the BEGIN block. Inside the END block, I am trying to write the string passed on command-line to the awk script, to file_output.txt.

I get the following output upon running the bash script:

a

b

c

d

awk: cmd. line:1: fatal: cannot open file 'c' for reading (No such file or directory)

Obviously, I haven't got how awk processes the command-line arguments based on the error. I was assuming that in the line

awk -v input=${string} -f post.awk < file_input.txt > file_output.txt

$1 would be input
$2 would be file_input.txt
$3 would be file_output.txt

Can someone point out where I have misplaced my assumptions?

TIA

There are many things wrong with your code:

1. You forget to quote ${string} leading to the confusing behaviour that awk wants to read the file c. Due to the missquoteing, you actually try to execute:

   $ awk -v input=a b c d -f post.awk < file_input.txt > file_output.txt
   

   Here you set the variable input=a and try to read 3 files (b, c and d). The file file_input.txt has no effect on the awk script, unless you actively process /dev/stdin in post.awk.

2. You make the mistake and believe that $n in the awk code, represents the _n_th argument passed to awk. This is not true. In awk, $n represents the n_th field of the current input record.

3. In the BEGIN block, there is no input record $0 defined. Only when an input record is defined, there are fields defined ($i, i>0). The END block, on the other hand has knowledge of the last input-record read from the input file.

You can define an input record in the BEGIN block by using getline as is, but not getline var as this does not define $0.

So, how can we make this work now.

If you just rely on simple bash, it will work the way you want. That is to say, let bash define, using redirections, what /dev/stdin and /dev/stdout is of the used command. Example:

$ binary < f1 > f2

Here, the executable binary is executed with /dev/stdin pointing to f1 and /dev/stdout pointing to f2.

So you can do the same, and write your awk program to just assume default /dev/stdin and /dev/stdout.

# post.awk
BEGIN { getline tmp; print tmp }
END   { print input }

and executing this as:

$ awk -v input="${string}" -f post.awk file_input.txt > file_output.txt

should do the trick.