[SOLVED] Can we create a new argument to a gawk program command-line?

Trd300 · 10-11-2012, 08:27 PM

I have a gawk script called prog that I made self-executable.

usage is:

Code:

prog <input>

I can use 2 kinds of input files: file containing letters or file containing numbers.
If I use letter input, prog will do the task 1
If I use number input, prog will do a different task 2.

Is it possible to add an option to the command-line specifying what kind of input file I am gonna use, which thus tell prog what task to execute?
And also a -h option showing the usage of prog?

e.g. if input file contains letters:

Code:

prog <input> -let

or if input file comtain numbers:

Code:

prog <input> -num

and to show usage:

Code:

prog -h

pan64 · 10-12-2012, 06:19 AM

do you have the source code?

grail · 10-12-2012, 10:34 AM

This should get you going:

Code:

#!/usr/bin/awk -f

BEGIN{
	if( ARGC != 3 )
		usage()

	switch(ARGV[1]){
		case "-let":
			print "we have a letters file"
			break
		case "-num":
			print "we have a numbers file"
			break
		default:
			usage()
	}
	ARGV[1]=""
	
}

1

function usage(		e)
{
	e = "Usage: script_name -[let|num] file"
	print e > "/dev/stderr"
	exit 1
}

The issue that you have is, if any switch is passed to your script that (g)awk understands, it will be used as a switch to it instead of your program, namely '-h'.
The only way around this that I know of is to always pass '--' prior to any of your options.

Here is the detail around dealing with the arguments yourself.

Trd300 · 10-13-2012, 04:02 AM

@pan64: It was just a general question. It was in my plans to create arguments to the command line, but until now I didn't know it was even possible, that's why I asked.

@grail: I didn't expect so much. And I learned a new statement. Thank you very much !

Trd300 · 10-15-2012, 01:05 AM

2 Little problems I didn't think about though.

* When the program executed properly, "Done !" displays to the terminal (to do that I used "END{print "Done !"}).
But as we have written in the BEGIN section, if ARGC != 3 the program displays an error massage described in the usage function.
The problem is that it also displays "Done !" like if the program executed normally.
To get rid off the successful message when an error occurred, I tried to change the END section by:

Code:

END{
    if(ARGC = 3){
        print "Done !"
    }
}

but it still keep displaying the END message.

* Is there a way to get rid off the internal awk error message when ARGC != 3.
If ARGC !=3 the program doesn't launch so I get the usage message, but I also get the awk error message (like input_file: No such file or directory).

Hope someone could help me !

pan64 · 10-15-2012, 01:58 AM

i would suggest you to set a variable in the BEGIN block (instead of using ARGC). And you can check that variable in the END block

David the H. · 10-15-2012, 01:58 AM

Possibly a simpler way may be something like this:

Code:

awk '( let ) { <do letter tasks> } ; ( num ) { <do number tasks> }' let=1 num=0 letterfile.txt let=0 num=1 numberfile.txt

awk allows you to pass variables as arguments before a filename.

I haven't actually tested it, but it should work in concept.

grail · 10-15-2012, 11:06 AM

Here is what the manual says:

Quote:

In such a case, if you don't want the END rule to do its job, set a variable to nonzero before the exit statement and check that variable in the END rule.

See here for further details.

Trd300 · 10-17-2012, 06:47 AM

I also tried to do the same task for a bash script. A good way seems to use the getopts function, but the tutorials I found on the web are difficult to understand.

Does someone could try to explain me with the above example?
Or any other non-getopts method.

Thanks !

pan64 · 10-17-2012, 08:32 AM

here is a step-by-step tutorial: http://wiki.bash-hackers.org/howto/getopts_tutorial
but this getopts cannot handle long options like -let or -num

grail · 10-17-2012, 11:39 AM

Maybe this is what you are looking for:

http://www.linuxquestions.org/questi...andling-34675/

Trd300 · 10-17-2012, 11:23 PM

Very helpful, thanks !

Trd300 · 10-18-2012, 06:37 AM

I read carefully both links you posted and I choose to give a try to getopts, but I got some issues.
So I started with something simple. Let's say the usage of myprog is as follow:

Code:

myprog -[l|n] <input>       #to launch the program

# OR

myprog -h          #to display the usage above

where -l and -n exclude each other, -l or -n and input are compulsory and don't take arguments.
myprog contains 3 steps. If -l is mentioned myprog does step 1, 2, 3. If -n is mentioned myprog skip step 1 and does steps 2 and 3 only.

I tried several versions of the code below without success:

Code:

#! bin/bash
input=""

while getopts "l n" opt
do
  case $opt in
    l)
      <do step 1>
      ;;
    n)
      next
      ;;
    h)
      echo "Usage: myprog -[l|n] input"
      ;;
    \?)
      echo "Invalid option: -$OPTARG" >&2
      exit1
      ;;
done

<do step 2>
<do step 3>

echo "End of the process. See ya !"

I am not really familiar with bash scripting. Is what I wrote really ugly...?

pan64 · 10-18-2012, 06:44 AM

not really ugly, but not really nice: you should write something like:

Code:

case $opt in
   h) function_h;; # usage
   l) function_l;;
   n) function_n;;
   *) function_error;;
esac

do_common_steps

and implement four functions to handle all the cases.

grail · 10-18-2012, 10:54 AM

I haven't played much with getopt, mainly as I prefer to roll my own so I have long opts as well, but I think you are supposed to include your 'h' option in the getopts call.