awk：在awk之外调用一个函数

Question

Update:

I'm working on a modular shell script which calls usage() in different level of subprocess. The script sample.sh has the structure resembles below:

sample.sh
├─ ...
├─ usage()
├─ awk
    ├─ usage()
├─ ...

usage() displays brief information on how to use the script (eg available arguments and its descriptions). When executed for the first time, usage() is displayed in the beginning of the script. Examples of other conditions for usage() invocation include:

1. awk fails to validate input from stdio or files

2. argument --help is received

3. other user-derived failures are occurred

I'd like to call identical function, usage() , from shell and its direct child process, awk .

---

Original Question:

Function, usage() of sample.sh prints its print statement as desired.

$cat sample.sh
#!/bin/sh
awk '
    BEGIN {
        usage()
    }
    function usage() {
        print "function inside of awk"
    }
'
$./sample.sh
function inside of awk

To take out usage() from awk and put it as a local function in sample.sh~ , I tried:

$cat sample.sh~
#!/bin/sh
usage() {
    print "function outside of awk"
}    
awk '
    BEGIN {
        usage()
    }
'
$./sample.sh~
awk: calling undefined function usage
 source line number 3

As we observe, we get the error message saying "undefined function usage" in sample.sh~ . How should I improve it?

Answer 1

I find a way to call function out of awk

put below code in a file, such as call it a.sh

$ cat a.sh
usage() {
  print "function outside of awk"
 }

$ chmod +x a.sh

then you can call the function in awk

#!/bin/sh
awk '
    BEGIN {
       system(". a.sh;usage")
    }'

Answer 2

A child process cannot call a function in the parent, and vice versa. Not even in the same language.

What you can do, is start another instance of the shell script that contains your awk script. For example, save this in a script called ./script.sh :

#!/bin/sh -e

usage() {
    echo usage: ...
    exit 1
}

test "$1" = --usage && usage

awk -v usage_script="$0" '
function usage() { system(usage_script " --usage") }
BEGIN {
    usage()
}'

Make it executable, and run as ./script.sh . In this script, the usage function inside the inner awk script simply re-runs the containing shell script with the --usage script. Keep in mind that there will be 3 processes here: shell -> awk -> shell. It's not possible to do this in a single process.

Original answer

If you move the usage function definition outside of awk , then it's normal that you cannot execute from inside. If you want to execute awk commands you have to be inside of an awk interpreter, and can only use what is defined inside.

Note sure if this is what you're looking for, but you can move the function definition to another awk file and have awk process it, for example:

cat <<EOF > functions.awk
function usage() {
    print "function in functions.awk"
}
EOF
awk -f functions.awk -e '
    BEGIN {
        usage()
    }
'

Answer 3

With bash , another option is to export shell functions with export -f , which then become available to awk via the system() call:

#!/usr/bin/env bash

usage() {
  echo "function outside of awk"
}

# Export the usage() function.
export -f usage

awk '
  BEGIN { 
    system("/usr/bin/env bash -c usage") 
  }'

Note that awk invokes sh with system() , which may or may not be bash - if it IS bash (eg, on OSX), you don't strictly need the /usr/bin/env bash -c portion, but leaving it in should make it work on most platforms (that have bash).