Setting global environment vars from Python or Bash

Question

#!/usr/bin/env python
from os import environ, path, name as os_name, getuid
from sys import exit
from fileinput import input
update_environ = lambda: environ.update(dict(env.split('=')
 for env in
 open('/etc/environment', 'r').readlines()))
if os_name != 'nt' and getuid() != 0:
 print "Rerun {0} as `sudo`".format(__file__)
 exit(1)
def custom_vars(e_vars):
 if not path.isfile('custom_vars.txt'):
 return e_vars
 with open('custom_vars.txt', 'r') as f:
 for line in f.readline():
 var = line.split()
 e_vars[var[0]] = var[1]
def append_environment(e_vars):
 for line in input('/etc/environment', inplace=True):
 var, _, val = line.partition('=')
 nline = '{var}={val}'.format(var=var, val=e_vars.pop(var, val))
 if nline and nline != '=' and nline != '\n' and nline != '=\n':
 print nline # inplace editing
 if e_vars:
 lines = filter(lambda l: l != '=\n' and nline != '\n',
 ['{var}={val}\n'.format(var=k, val=e_vars[k])
 for k in e_vars])
 with open('/etc/environment', 'a') as f:
 f.writelines(lines)
if __name__ == '__main__':
 environment_vars = {'NOM_REPOS': path.join(path.expanduser("~"), 'NOM'),
 'NOM_API': path.join('/var', 'www')}
 environment_vars = custom_vars(environment_vars)
 append_environment(environment_vars)

I have written a similar script in Bash; with identical functionality:

#!/usr/bin/env bash
sudo sed -i "/^NOM*/d" /etc/environment
for i in `env | grep NOM | cut -f1 -d=`; do
 unset "$i";
done
function set_var()
{
 VAR="1ドル"
 VAL="2ドル"
 sudo sed -i "/^$VAR/d" /etc/environment
 printf "$VAR=$VAL\n" | sudo tee -a /etc/environment > /dev/null
 # make the variable available for use in this script and custom_env_vars.sh
 export "$VAR=$VAL"
}
set_var NOM_REPOS "$HOME/NOM"
set_var NOM_API /var/www
if [ -f custom_env_vars.sh ]; then
 . ./custom_env_vars.sh
else
 echo "You can set custom environment variables in 'custom_env_vars.sh'"
fi

• Gareth Rees: Can you give some context? What is the purpose of this code? What are you trying to achieve?
• A T: I have a variety of scripts; the first set env vars; the others utilise these vars to clone/update repositories, install/update dependencies + OS and configure the system (e.g.: deploy the .confs for my web and application servers). The env vars are also used by my remote log daemons.
• Gareth Rees: That's a bit vague. Can you describe a use case in detail?

Answer 1

env | grep '^NOM' would more exactly align with the sed command.

ensure you're exactly matching the variable name:

sudo sed -i "/^$VAR=/d" /etc/environment
# .................^

Use %-formatters for printf:

printf '%s="%s"\n" "$VAR" "$VAL"

In the function, use local to limit the scope of VAR and VAL

---

had another thought about that function. Change

sudo sed -i "/^$VAR/d" /etc/environment
printf "$VAR=$VAL\n" | sudo tee -a /etc/environment > /dev/null

to

tmpfile=$(mktemp)
grep -v "^$VAR=" /etc/environment > $tmpfile
echo "$VAR=$VAL" >> $tmpfile
sudo mv $tmpfile /etc/environment

That minimized any possibility that the /etc/environment file contains "invalid" data -- it overwrites the previous good file with the new good file all at once.