Can I export a variable to the environment from a Bash script without sourcing it?

Question

Suppose that I have this script:

export.bash:

#! /usr/bin/env bash
export VAR="HELLO, VARIABLE"

When I execute the script and try to access to the $VAR, I don't get any value!

echo $VAR

Is there a way to access the $VAR by just executing export.bash without sourcing it?

Answer 1

Let's say I want to have a command that sets an environment varible FOO in the current shell. The example just shows the process, it's not the best way to achieve the same result.

I put the following in a file called foo.sh:

foo() {
  local value="${1:-bar}"
  source <(echo "export FOO=$value")
}

export -f foo

Then I can run the following commands:

$ source foo.sh
$ foo && echo $FOO
bar
$ foo baz && echo $FOO
baz

If I want the command to be generally available, I could put it in my shell's rc file, e.g. .bashrc.

A slightly more involved example could be something like the following that sets my AWS_PROFILE environment variable to something I chose from the list of valid values. The example requires the fzf command to be installed.:

list-profiles() {
    local filter="${1:-}"
    aws configure list-profiles | sort -f | grep -v ^default$ | grep "$filter"
}

set-profile() {
    local query="${1:+--query $1}"
    if RES=$(list-profiles | fzf --preview="" -1 $query); then
        echo "export AWS_PROFILE=$RES"
    fi
}

sp() {
    source <(set-profile "${1:-}") && aws sts get-caller-identity
}

export -f sp

If I run sp without parameters, it will show a list of profiles to choose from. If I run sp foo it will only show profiles that contain foo in their name. If there is only one valid result, it will automatically pick that one.

Answer 2

Others have already answered your question, but here's a little advice. Since you have to call the script with a leading dot so that set variables remain valid for the calling shell, I solved this by having the script always inform me if I didn't start it with a leading dot. This way I am reminded to use a leading dot.

#!/bin/sh
if [ $0 = "NAME_OF_SCRIPT_WITH_ITS_PATH" ]
then
  # If the script was called without a leading dot, print a hint.
  echo "Important:"
  echo "  This script must be called with a leading dot."
  echo "  Otherwise, the environment variables will not be added to the calling shell."
else
  # If the script was called with a leading dot, set the environment variables.
  # In this case $0 corresponds to the name of your shell.
  # For example $0 is "/bin/bash" for bash.
  export FOO=TRUE
  export BAR=FALSE
  echo "Environment variables have been set."
fi

NAME_OF_SCRIPT_WITH_ITS_PATH must be replaced with the name of your script file and its path. Basically it is the value that a command like echo $0 would output inside the script if the script is not run with a leading dot.

Answer 3

If you want to set a variable for the calling shell there is a robust approach using c unix sockets. The result will be tested like this:

$ ./a.out &
[1] 5363
$ ./b.out 123;a=`./b.out`;echo ${a}
123
[1]+  Done                    ./a.out

See a.c, b.c, and a.h from my github.

Answer 4

simple naive approach that works:

script 1:

echo "something" > /tmp/myvar

script 2:

myvar=$(cat /tmp/myvar)

Answer 5

Export environment variables using script file

Problem:
When you run a script, it executes in a child shell and returns back to the parent shell after execution. Exporting variables only works down the child shells, you can't export a child shell's variable back to the parent shell.

Solution:
From your script file invoke a child shell along with variables that you want to export, this will create a new child shell with your variables exported.

script.sh ->

bash -c 'export VAR=variable; exec bash'

: : CIPH3R

Answer 6

I had similar problem calling ssh-agent -s in a script called by option -e in rsync.

In the script eval $(ssh-agent -s) don't preserve the environment variables for the next call.

rsync -e 'source ssh-check-agent.sh -p 8022' does not work, so I made a workaround. In the script I saved the variables in a temporal file after call ssh-agent like:

echo "export SSH_AUTH_SOCK=$SSH_AUTH_SOCK;" > /tmp/ssh-check-agent.vars
echo "export SSH_AGENT_PID=$SSH_AGENT_PID;" >> /tmp/ssh-check-agent.vars

and after in the script that calls rsync (backup.sh) I call:

source /tmp/ssh-check-agent.vars

The problem is that script that calls rsync must be called by source (source backup.sh).

I know that is not the question (I use two times source), but I put here if someone has similar problem with rsync.

Answer 7

This workaround is somehow hinted to elsewhere, but maybe not that clearly:

In your script, after setting the variable, start a new shell, rather than return.

My use cases is that I have a number of terminals open and in some of them I want some values for some variables, while in others I want other values.

As using source may be harder to remember, a small advantage of this approach is when it takes a while to realize that you forgot to use source, and you have to start from scratch.

(For me it makes more sense to use source script, as the missing variables are noticed immediately.)

Answer 8

I don't think this can be done, but I found a workaround using alias. It will only work when you place your script in your scripts directory. Otherwise your alias will have an invalid name.

The only point to the workaround is to be able to have a function inside a file with the same name and not have to bother sourcing it before using it. Add the following code to file ~/.bashrc:

alias myFunction='unalias myFunction && . myFunction && myFunction "$@"'

You can now call myFunction without sourcing it first.

Answer 9

Maybe you can add a function in ~/.zshrc or ~/.bashrc.

# set my env
[ -s ~/.env ] && export MYENV=`cat ~/.env`
function myenv() { [[ -s ~/.env ]] && echo $argv > ~/.env && export MYENV=$argv }

Because of the use of a variable outside, you can avoid the use of a script file.

Answer 10

I found an interesting and neat way to export environment variables from a file:

In file env.vars:

foo=test

Test script:

eval `cat env.vars`
echo $foo         # => test
sh -c 'echo $foo' # =>

export eval `cat env.vars`
echo $foo         # => test
sh -c 'echo $foo' # => test

# a better one. "--" stops processing options,
# key=value list given as parameters
export -- `cat env.vars`
echo $foo         # => test
sh -c 'echo $foo' # => test

Answer 11

Another workaround that, depends on the case, it could be useful: creating another bash script that inherits the exported variable. It is a particular case of Keith Thompson's answer, will all of those drawbacks.

File export.bash:

# !/bin/bash
export VAR="HELLO, VARIABLE"
bash

Now:

./export.bash
echo $VAR

Answer 12

The answer is no, but for me I did the following

The script:

myExport

#! \bin\bash
export $1

An alias in my .bashrc file:

alias myExport='source myExport'

Still you source it, but maybe in this way it is more useable and it is interesting for someone else.

Answer 13

In order to export out the VAR variable first, the most logical and seemly working way is to source the variable:

. ./export.bash

or

source ./export.bash

Now when echoing from the main shell, it works:

echo $VAR
HELLO, VARIABLE

We will now reset VAR:

export VAR=""
echo $VAR

Now we will execute a script to source the variable then unset it:

./test-export.sh
HELLO, VARIABLE
--
.

The code: file test-export.sh

#!/bin/bash
# Source env variable
source ./export.bash

# echo out the variable in test script
echo $VAR

# unset the variable
unset VAR
# echo a few dotted lines
echo "---"
# now return VAR which is blank
echo $VAR

Here is one way:

Please note: The exports are limited to the script that execute the exports in your main console - so as far as a cron job I would add it like the console like below... for the command part still questionable: here is how you would run in from your shell:

On your command prompt (so long as the export.bash file has multiple echo values)

IFS=$'\n'; for entries in $(./export.bash); do  export $entries;  done; ./v1.sh
HELLO THERE
HI THERE

File cat v1.sh

#!/bin/bash
echo $VAR
echo $VAR1

Now so long as this is for your usage - you could make the variables available for your scripts at any time by doing a Bash alias like this:

myvars ./v1.sh
HELLO THERE
HI THERE

echo $VAR

.

Add this to your .bashrc file:

function myvars() {
    IFS=$'\n';
    for entries in $(./export.bash); do  export $entries;  done;

    "$@";

    for entries in $(./export.bash); do variable=$(echo $entries|awk -F"=" '{print $1}'); unset $variable;
    done
}

Source your .bashrc file and you can do like the above any time...

Anyhow back to the rest of it...

This has made it available globally then executed the script...

Simply echo it out and run export on the echo!

File export.bash

#!/bin/bash
echo "VAR=HELLO THERE"

Now within script or your console run:

export "$(./export.bash)"

Try:

echo $VAR
HELLO THERE

Multiple values so long as you know what you are expecting in another script using the above method:

File export.bash

#!/bin/bash
echo "VAR=HELLO THERE"
echo "VAR1=HI THERE"

File test-export.sh

#!/bin/bash

IFS=$'\n'
for entries in $(./export.bash); do
    export $entries
done

echo "round 1"
echo $VAR
echo $VAR1

for entries in $(./export.bash); do
    variable=$(echo $entries|awk -F"=" '{print $1}');
    unset $variable
done

echo "round 2"
echo $VAR
echo $VAR1

Now the results

./test-export.sh
round 1
HELLO THERE
HI THERE
round 2


.

And the final final update to auto assign, read the VARIABLES:

./test-export.sh
Round 0 - Export out then find variable name -
Set current variable to the variable exported then echo its value
$VAR has value of HELLO THERE
$VAR1 has value of HI THERE
round 1 - we know what was exported and we will echo out known variables
HELLO THERE
HI THERE
Round 2 - We will just return the variable names and unset them
round 3 - Now we get nothing back

The script:

File test-export.sh

#!/bin/bash

IFS=$'\n'
echo "Round 0 - Export out then find variable name - "
echo "Set current variable to the variable exported then echo its value"
for entries in $(./export.bash); do
    variable=$(echo $entries|awk -F"=" '{print $1}');
    export $entries
    eval current_variable=\$$variable
    echo "\$$variable has value of $current_variable"
done


echo "round 1 - we know what was exported and we will echo out known variables"
echo $VAR
echo $VAR1

echo "Round 2 - We will just return the variable names and unset them "
for entries in $(./export.bash); do
    variable=$(echo $entries|awk -F"=" '{print $1}');
    unset $variable
done

echo "round 3 - Now we get nothing back"
echo $VAR
echo $VAR1

Answer 14

Is there any way to access to the $VAR by just executing export.bash without sourcing it ?

Quick answer: No.

But there are several possible workarounds.

The most obvious one, which you've already mentioned, is to use source or . to execute the script in the context of the calling shell:

$ cat set-vars1.sh 
export FOO=BAR
$ . set-vars1.sh 
$ echo $FOO
BAR

Another way is to have the script, rather than setting an environment variable, print commands that will set the environment variable:

$ cat set-vars2.sh
#!/bin/bash
echo export FOO=BAR
$ eval "$(./set-vars2.sh)"
$ echo "$FOO"
BAR

A third approach is to have a script that sets your environment variable(s) internally and then invokes a specified command with that environment:

$ cat set-vars3.sh
#!/bin/bash
export FOO=BAR
exec "$@"
$ ./set-vars3.sh printenv | grep FOO
FOO=BAR

This last approach can be quite useful, though it's inconvenient for interactive use since it doesn't give you the settings in your current shell (with all the other settings and history you've built up).

Answer 15

Execute

set -o allexport

Any variables you source from a file after this will be exported in your shell.

source conf-file

When you're done execute. This will disable allexport mode.

set +o allexport