I'm using the next code to execute a bash script that defines a lot of environment variables:
#!/usr/bin/python
# -*- coding: utf-8 -*-
import subprocess
# Code used to get the value of variable before the call to my script
command = "echo $MYDIR"
ps = subprocess.Popen(command, shell=True, stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
my_dir = ps.communicate()[0]
print "my_dir", my_dir
subprocess.Popen(["/bin/sh", "/home/user/env_file"])
# Code used to get the value of established variable
command = "echo $MYDIR"
ps = subprocess.Popen(command, shell=True, stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
my_dir = ps.communicate()[0]
print "my_dir", my_dir
The content of my /home/user/env_file is the next:
#!/bin/bash
export MYDIR=/home/user
export MYDIR2=/home/user2
export MYDIR3=/home/user3
export MYDIR4=/home/user4
export MYDIR5=/home/user5
export MYDIR6=/home/user6
export MYDIR7=/home/user7
export MYDIR8=/home/user8
export MYDIR9=/home/user9
export MYDIR10=/home/user10
export MYDIR11=/home/user11
export MYDIR12=/home/user12
export MYDIR13=/home/user13
export MYDIR14=/home/user14
export MYDIR15=/home/user15
export MYDIR16=/home/user16
When I execute the Python code I do not get any value for my_dir variable. How can I perform this?
Best regards.
The only way for a process to influence its parent's environment (absent ugly hacks abusing development/debugging facilities) is with that parent process's direct involvement (which is why one runs eval "$(ssh-agent)" -- evaling its output is, in this case, the cooperation needed from the parent to let ssh-agent set environment variables in the process that started it). Consider the following shell wrapper:
#!/usr/bin/env bash
# Save this script as "envvar-extraction-wrapper"
exec {orig_stdout}>&1 # create a backup of stdout
exec >&2 # and then use stderr while sourcing the child script
source "$@" # run arbitrary script with arbitrary arguments *in this shell*
# note that this means that if it runs "exit", we abort prematurely
while read -r varname; do
printf '%s\0%s\0' "$varname" "${!varname}" >&$orig_stdout
done < <(compgen -e)
...called from the following Python:
import subprocess, itertools, os
ps = subprocess.Popen(['envvar-extraction-wrapper', '/home/user/env_file'],
stdout=subprocess.PIPE, stderr=subprocess.STDOUT)
(stdout, stderr) = ps.communicate()
items_iter = iter(stdout.split('\0'))
for (k,v) in itertools.izip(items_iter, items_iter):
os.environ[k]=v
The bash script, env_file is executed in its own process, thus changes to the enviroment variables here does not get reflected back to the python script's process. If you can redesign your scripts to avoid this situation, it would be best. Otherwise, we have to resort to some hack and here is one of them.
#!/usr/bin/python
# -*- coding: utf-8 -*-
import os
import subprocess
import tempfile
env_file = "/tmp/env_file"
with open(env_file) as f:
original_script_contents = f.read()
# Duplicate the env_file and add an echo statement at the end
with tempfile.NamedTemporaryFile() as temp_script:
temp_script.write(original_script_contents)
temp_script.write('\necho $MYDIR')
temp_script.flush()
my_dir = subprocess.check_output(["/bin/sh", temp_script.name])
print 'my_dir =', my_dir
This hack involves duplicate the contents of env_file into another temporary file, then add an echo statement at the end of the duplicate file, execute it and collect the output.
Also, my assumption is the original script does not output anything. If it does, you have to parse the output and find what you are looking for.
