使用 Python 子进程运行 `eval($ something)` 命令

Question

I have the following subprocess/command I run inside terminal:

$ eval $(docker-machine env machine-name)

I assumed the python equivalent would be:

sp = subprocess.run(["eval", "$(docker-machine env {})".format(self.project_machine_name)])

or perhaps:

sp = subprocess.run(["eval", "$(docker-machine", "env", "{})".format(self.project_machine_name)])

However, neither of the above seem to work.

I was wondering, how would I run the above eval within a python script...

NB . For brevity, assume self.project_machine_name = 'machine-name'

This is not the same as calling a sub-process as it needs to be eval() .

Answer 1

try

sp = subprocess.run(['docker-machine', 'env', self.project_machine_name])

More info and additional arguments - in the docs

EDIT: if you want the output, use subprocess.check_output() instead.

Answer 2

It depends what you're trying to do but I may be doing something similar. I'm building my dockers in minikube and have a python script that I use to build those dockers.

Before I run the docker build I need to make sure the right environment variables are set, meaning I need to run

eval $(minikube -p minikube docker-env)

Which generates a bunch of statements that you're meant to run in the shell (typically with an eval)

export DOCKER_TLS_VERIFY="1"
export DOCKER_HOST="tcp://192.168.49.2:2376"
export DOCKER_CERT_PATH="/home/aluchko/.minikube/certs"
export MINIKUBE_ACTIVE_DOCKERD="minikube"

# To point your shell to minikube's docker-daemon, run:
# eval $(minikube -p minikube docker-env)

To be able to trigger my builds in python I need these environment variables set when I my docker build commands.

Now there's basically two ways to do this (aside from doing the eval in a wrapper around your script). Both ways involve getting the that string with the export commands.

output = subprocess.check_output(['minikube', '-p', 'minikube', 'docker-env'])

1. Parse the results of the minikube (or docker-machine command) and set those variables in the shell for future subprocesses. You can be a bit sloppy with the RE since you know the output of the docker-machine command. The shell=True is just for the echo.

    output = subprocess.check_output(['minikube', '-p', 'minikube', 'docker-env']) import re export_re = re.compile('export ([AZ\\_]+)="(.*)"\\\\n') export_pairs = export_re.findall(output.decode("UTF-8")) build_env = os.environ.copy() for k,v in export_pairs: build_env[k] = v subprocess.run('echo $DOCKER_HOST', shell=True, env=build_env)

2. A bit sloppier, and more prone to injection attacks, just sandwich the strings together so all your commands are basically a script that starts with a bunch of exports, this isn't nearly as clean and locks you into using the shell=True subprocess functionality, but it does work.

    output = subprocess.check_output(['minikube', '-p', 'minikube', 'docker-env']) extended_output = output + b"\\necho $DOCKER_HOST" print(extended_output) subprocess.run(extended_output, shell=True)