a.sh
#! /bin/sh
export x=/usr/local
we can do source ./a in command-line. But I need to do the export through shell script.
b.sh
#! /bin/sh
. ~/a.sh
no error... but $x in command-line will show nothing. So it didn't get export.
Any idea how to make it work?
a.sh
#! /bin/sh
export x=/usr/local
-----------
admin@client: ./a.sh
admin@client: echo $x
admin@client: <insert ....>
You can put export statements in a shell script and then use the 'source' command to execute it in the current process:
source a.sh
You can't do an export through a shell script, because a shell script runs in a child shell process, and only children of the child shell would inherit the export.
The reason for using source is to have the current shell execute the commands
It's very common to place export commands in a file such as .bashrc which a bash will source on startup (or similar files for other shells)
Another idea is that you could create a shell script which generates an export command as it's output:
shell$ cat > script.sh
#!/bin/sh
echo export foo=bar
^D
chmod u+x script.sh
And then have the current shell execute that output
shell$ `./script.sh`
shell$ echo $foo
bar
shell$ /bin/sh
$ echo $foo
bar
(note above that the invocation of the script is surrounded by backticks, to cause the shell to execute the output of the script)
Answering my own question here, using the answers above: if I have more than one related variable to export which use the same value as part of each export, I can do this:
#!/bin/bash
export TEST_EXPORT=$1
export TEST_EXPORT_2=$1_2
export TEST_EXPORT_TWICE=$1_$1
and save as e.g. ~/Desktop/TEST_EXPORTING
and finally $chmod +x ~/Desktop/TEST_EXPORTING
--
After that, running it with source ~/Desktop/TEST_EXPORTING bob
and then checking with export | grep bob should show what you expect.
source only needs read access, you only need chmod +x for files that you actually want to execute.Exporting a variable into the environment only makes that variable visible to child processes. There is no way for a child to modify the environment of its parent.
Another way you can do it (to steal/expound upon the idea above), is to put the script in ~/bin and make sure ~/bin is in your PATH. Then you can access your variable globally. This is just an example I use to compile my Go source code which needs the GOPATH variable to point to the current directory (assuming you're in the directory you need to compile your source code from):
From ~/bin/GOPATH:
#!/bin/bash
echo declare -x GOPATH=$(pwd)
Then you just do:
#> $(GOPATH)
So you can now use $(GOPATH) from within your other scripts too, such as custom build scripts which can automatically invoke this variable and declare it on the fly thanks to $(pwd).
script1.sh
shell_ppid=$PPID
shell_epoch=$(grep se.exec_start "/proc/${shell_ppid}/sched" | sed 's/[[:space:]]//g' | cut -f2 -d: | cut -f1 -d.)
now_epoch=$(($(date +%s%N)/1000000))
shell_start=$(( (now_epoch - shell_epoch)/1000 ))
env_md5=$(md5sum <<<"${shell_ppid}-${shell_start}"| sed 's/[[:space:]]//g' | cut -f1 -d-)
tmp_dir="/tmp/ToD-env-${env_md5}"
mkdir -p "${tmp_dir}"
ENV_PROPS="${tmp_dir}/.env"
echo "FOO=BAR" > "${ENV_PROPS}"
script2.sh
shell_ppid=$PPID
shell_epoch=$(grep se.exec_start "/proc/${shell_ppid}/sched" | sed 's/[[:space:]]//g' | cut -f2 -d: | cut -f1 -d.)
now_epoch=$(($(date +%s%N)/1000000))
shell_start=$(( (now_epoch - shell_epoch)/1000 ))
env_md5=$(md5sum <<<"${shell_ppid}-${shell_start}"| sed 's/[[:space:]]//g' | cut -f1 -d-)
tmp_dir="/tmp/ToD-env-${env_md5}"
mkdir -p "${tmp_dir}"
ENV_PROPS="${tmp_dir}/.env"
source "${ENV_PROPS}"
echo $FOO
./script1.sh
./script2.sh
BAR
It persists for the scripts run in the same parent shell, and it prevents collisions.
