bash variable as a command: echo the command before execution and save the result to a variable

Question

I am executing a chain of curl commands:

1. I need to echo the command before the execution.
2. Execute the command and save the result to a bash variable.
3. Get values from the result of the execution and execute the next curl with that values.

This is how it looks like:

#  -----> step 1 <-----
URL="https://web.example.com:8444/hello.html"
CMD="curl \
        --insecure \
        --dump-header - \
        \"$URL\""

echo $CMD && eval $CMD
OUT="<result of the curl command???>"

# Set-Cookie: JSESSIONID=5D5B29689EFE6987B6B17630E1F228AD; Path=/; Secure; HttpOnly
JSESSIONID=$(echo $OUT | grep JSESSIONID | awk '{ s = ""; for (i = 2; i <= NF; i++) s = s $i " "; print s }' | xargs)

# Location: https://web.example.com:8444/oauth2/authorization/openam
URL=$(echo $OUT | grep Location | awk '{print $2}')

#  -----> step 2 <-----
CMD="curl \
        --insecure \
        --dump-header - \
        --cookie \"$JSESSIONID\" \
        \"$URL\""
echo $CMD && eval $CMD
OUT="<result of the curl command???>"
...

#  -----> step 3 <-----
...

I only have a problem with the step 2: save the full result of the curl command to a variable in order to I can parse it.

I have tried it many different way, non of them works:

• OUT="eval \$CMD"
• OUT=\$$CMD
• OUT=$($CMD)
• ...

What I missed?

Answer 1

For very basic commands, OUT=$($CMD) should work. The problem with this is, that strings stored in variables are processed differently than strings entered directly. For instance, echo "a" prints a, but var='"a"'; echo $a prints "a" (note the quotes). Because of that and other reasons, you shouldn't store commands in variables.

In bash, you can use arrays instead. By the way: The naming convention for regular variables is NOT ALLCAPS, as such names might accidentally collide with special variables. Also, you can probably drastically simplifiy your grep | awk | xargs.

url="https://web.example.com:8444/hello.html"
cmd=(curl --insecure --dump-header - "$url")
printf '%q ' "${cmd[@]}"; echo
out=$("${cmd[@]}")
# Set-Cookie: JSESSIONID=5D5B29689EFE6987B6B17630E1F228AD; Path=/; Secure; HttpOnly
jsessionid=$(awk '{$1=""; printf "%s%s", d, substr($0,2); d=FS}' <<< "$out")
# Location: https://web.example.com:8444/oauth2/authorization/openam
url=$(awk '/Location/ {print $2}' <<< "$out")

#  -----> step 2 <-----
cmd=(curl --insecure --dump-header - --cookie "$jsessionid" "$url")
printf '%q ' "${cmd[@]}"; echo
out=$("${cmd[@]}")

#  -----> step 3 <-----
...

If you have more steps than that, wrap the repeating part into a function, as suggested by Charles Duffy.

Answer 2

Easy Mode: Use set -x

Bash has a built-in feature, xtrace, which tells it to log every command to the file descriptor named in the variable BASH_XTRACEFD (by default, file descriptor 2, stderr).

#!/bin/bash
set -x
url="https://web.example.com:8444/hello.html"
output=$(curl \
  --insecure \
  --dump-header - \
  "$url")

echo "Output of curl follows:"
echo "$output"

...will provide logs having the form of:

+ url=https://web.example.com:8444/hello.html
++ curl --insecure --dump-header - https://web.example.com:8444/hello.html
+ output=Whatever
+ echo 'Output of curl follows:'
+ echo Whatever

...where the + is based on the contents of the variable PS4, which can be modified to have more information. (I often use and suggest PS4=':${BASH_SOURCE}:$LINENO+' to put the source filename and line number in each logged line).

Doing It By Hand

If that's not acceptable, you can write a function.

log_and_run() {
  { printf '%q ' "$@"; echo; } >&2
  "$@"
}

output=$(log_and_run curl --insecure --dump-header - "$url")

...will write your curl command line to stderr before storing its output in $output. Note when writing that output that you need to use quotes: echo "$output", not echo $output.

Answer 3

I guess OUT=$(eval $CMD) will do what you want.