Python: Exporting environment variables in subprocess.Popen(..)

Question

I am having an issue with setting an environment variable on a call to subprocess.Popen. The environment variable does not seem to be getting set. Any suggestions on how to properly set environmental variables for a Python commandline call?

My goal is to run a script that uses an environmental variable determined from my Python code:

d = dict(os.environ)
d["TEST_VARIABLE"] = str(1234)
subprocess.Popen('/usr/bin/mybinary', env=d).wait()

but the script reacts as if the variable has never been set

Here is my attempt to test, using Python's interactive interpreter:

d = dict(os.environ)
d["TEST_VARIABLE"] = str(1234)
subprocess.Popen(['/bin/echo', '$TEST_VARIABLE'], env=d).wait() 

and the output is:

"$TEST_VARIABLE"
0

I thought env=d should set the environment for the subprocess, but it apparently does not. Any suggestions on how to correct this issue?

Answer 1

The substitution of environment variables on the command line is done by the shell, not by /bin/echo. So you need to run the command in a shell to get the substitution:

In [22]: subprocess.Popen('/bin/echo $TEST_VARIABLE', shell=True, env=d).wait()
1234
Out[22]: 0

That doesn't mean the environment variable is not set when shell=False, however. Even without shell=True, the executable does see the enviroment variables set by the env parameter. For example, date is affected by the TZ environment variable:

In [23]: subprocess.Popen(["date"], env={'TZ': 'America/New_York'}).wait()
Wed Oct 29 22:05:52 EDT 2014
Out[23]: 0

In [24]: subprocess.Popen(["date"], env={'TZ': 'Asia/Taipei'}).wait()
Thu Oct 30 10:06:05 CST 2014
Out[24]: 0

Answer 2

For Python 3.5 and newer, you can use unpacking generalizations, eg:

env = {
    **os.environ,
    "TEST_VARIABLE": str(1234),
}
subprocess.Popen('/usr/bin/mybinary', env=env).wait()