696
votes

TL;DR: How do I export a set of key/value pairs from a text file into the shell environment?


For the record, below is the original version of the question, with examples.

I'm writing a script in bash which parses files with 3 variables in a certain folder, this is one of them:

MINIENTREGA_FECHALIMITE="2011-03-31"
MINIENTREGA_FICHEROS="informe.txt programa.c"
MINIENTREGA_DESTINO="./destino/entrega-prac1"

This file is stored in ./conf/prac1

My script minientrega.sh then parses the file using this code:

cat ./conf/$1 | while read line; do
    export $line
done

But when I execute minientrega.sh prac1 in the command line it doesn't set the environment variables

I also tried using source ./conf/$1 but the same problem still applies

Maybe there is some other way to do this, I just need to use the environment variables of the file I pass as the argument of my script.

30
This is a great question but is phrased way too specifically, with particular variable names ("MINIENTREGA_FECHALIMITE"? what does that mean?) and numbers (3). The general question is simply, "How do I export a set of key/value pairs from a text file into the shell environment". - Dan Dascalescu
Also, this has already been answered on unix.SE and is arguably more on-topic there. - Dan Dascalescu

30 Answers

282
votes

Problem with your approach is the export in the while loop is happening in a sub shell, and those variable will not be available in current shell (parent shell of while loop).

Add export command in the file itself:

export MINIENTREGA_FECHALIMITE="2011-03-31"
export MINIENTREGA_FICHEROS="informe.txt programa.c"
export MINIENTREGA_DESTINO="./destino/entrega-prac1"

Then you need to source in the file in current shell using:

. ./conf/prac1

OR

source ./conf/prac1
1125
votes

This might be helpful:

export $(cat .env | xargs) && rails c

Reason why I use this is if I want to test .env stuff in my rails console.

gabrielf came up with a good way to keep the variables local. This solves the potential problem when going from project to project.

env $(cat .env | xargs) rails

I've tested this with bash 3.2.51(1)-release


Update:

To ignore lines that start with #, use this (thanks to Pete's comment):

export $(grep -v '^#' .env | xargs)

And if you want to unset all of the variables defined in the file, use this:

unset $(grep -v '^#' .env | sed -E 's/(.*)=.*/\1/' | xargs)

Update:

To also handle values with spaces, use:

export $(grep -v '^#' .env | xargs -d '\n')

on GNU systems -- or:

export $(grep -v '^#' .env | xargs -0)

on BSD systems.


From this answer you can auto-detect the OS with this:

export-env.sh

#!/bin/sh

## Usage:
##   . ./export-env.sh ; $COMMAND
##   . ./export-env.sh ; echo ${MINIENTREGA_FECHALIMITE}

unamestr=$(uname)
if [ "$unamestr" = 'Linux' ]; then

  export $(grep -v '^#' .env | xargs -d '\n')

elif [ "$unamestr" = 'FreeBSD' ]; then

  export $(grep -v '^#' .env | xargs -0)

fi

561
votes

-o allexport enables all following variable definitions to be exported. +o allexport disables this feature.

set -o allexport
source conf-file
set +o allexport
224
votes
set -a
. ./env.txt
set +a

If env.txt is like:

VAR1=1
VAR2=2
VAR3=3
...

Explanations -a is equivalent to allexport. In other words, every variable assignment in the shell is exported into the environment (to be used by multiple child processes). More information can be found in the Set builtin documentation:

-a     Each variable or function that is created or modified is given the export attribute and marked for export to the environment of subsequent commands.

Using ‘+’ rather than ‘-’ causes these options to be turned off. The options can also be used upon invocation of the shell. The current set of options may be found in $-.

55
votes

I found the most efficient way is:

export $(xargs < .env)

Explanation

When we have a .env file like this:

key=val
foo=bar

run xargs < .env will get key=val foo=bar

so we will get an export key=val foo=bar and it's exactly what we need!

Limitation

  1. It doesn't handle cases where the values have spaces in them. Commands such as env produce this format. – @Shardj
39
votes

The allexport option is mentioned in a couple of other answers here, for which set -a is the shortcut. Sourcing the .env really is better than looping over lines and exporting because it allows for comments, blank lines, and even environment variables generated by commands. My .bashrc includes the following:

# .env loading in the shell
dotenv () {
  set -a
  [ -f .env ] && . .env
  set +a
}

# Run dotenv on login
dotenv

# Run dotenv on every new directory
cd () {
  builtin cd $@
  dotenv
}
27
votes
eval $(cat .env | sed 's/^/export /')
22
votes

Here is another sed solution, which does not run eval or require ruby:

source <(sed -E -n 's/[^#]+/export &/ p' ~/.env)

This adds export, keeping comments on lines starting with a comment.

.env contents

A=1
#B=2

sample run

$ sed -E -n 's/[^#]+/export &/ p' ~/.env
export A=1
#export B=2

I found this especially useful when constructing such a file for loading in a systemd unit file, with EnvironmentFile.

18
votes

I have upvoted user4040650's answer because it's both simple, and it allows comments in the file (i.e. lines starting with #), which is highly desirable for me, as comments explaining the variables can be added. Just rewriting in the context of the original question.

If the script is callled as indicated: minientrega.sh prac1, then minientrega.sh could have:

set -a # export all variables created next
source $1
set +a # stop exporting

# test that it works
echo "Ficheros: $MINIENTREGA_FICHEROS"

The following was extracted from the set documentation:

This builtin is so complicated that it deserves its own section. set allows you to change the values of shell options and set the positional parameters, or to display the names and values of shell variables.

set [--abefhkmnptuvxBCEHPT] [-o option-name] [argument …] set [+abefhkmnptuvxBCEHPT] [+o option-name] [argument …]

If no options or arguments are supplied, set displays the names and values of all shell variables and functions, sorted according to the current locale, in a format that may be reused as input for setting or resetting the currently-set variables. Read-only variables cannot be reset. In POSIX mode, only shell variables are listed.

When options are supplied, they set or unset shell attributes. Options, if specified, have the following meanings:

-a Each variable or function that is created or modified is given the export attribute and marked for export to the environment of subsequent commands.

And this as well:

Using ‘+’ rather than ‘-’ causes these options to be turned off. The options can also be used upon invocation of the shell. The current set of options may be found in $-.

15
votes

Improving on Silas Paul's answer

exporting the variables on a subshell makes them local to the command.

(export $(cat .env | xargs) && rails c)

14
votes

SAVE=$(set +o | grep allexport) && set -o allexport && . .env; eval "$SAVE"

This will save/restore your original options, whatever they may be.

Using set -o allexport has the advantage of properly skipping comments without a regex.

set +o by itself outputs all your current options in a format that bash can later execute. Also handy: set -o by itself, outputs all your current options in human-friendly format.

14
votes

Not exactly sure why, or what I missed, but after running trough most of the answers and failing. I realized that with this .env file:

MY_VAR="hello there!"
MY_OTHER_VAR=123

I could simply do this:

source .env
echo $MY_VAR

Outputs: Hello there!

Seems to work just fine in Ubuntu linux.

13
votes

Here's my variant:

  with_env() {
    (set -a && . ./.env && "$@")
  }

compared with the previous solutions:

  • it does not leak variables outside scope (values from .env are not exposed to caller)
  • does not clobber set options
  • returns exit code of the executed command
  • uses posix compatible set -a
  • uses . instead of source to avoid bashism
  • command is not invoked if .env loading fails
with_env rails console
12
votes

The shortest way I found:

Your .env file:

VARIABLE_NAME="A_VALUE"

Then just

. ./.env && echo ${VARIABLE_NAME}

Bonus: Because it's a short one-liner, it's very useful in package.json file

  "scripts": {
    "echo:variable": ". ./.env && echo ${VARIABLE_NAME}"
  }
10
votes

Simpler:

  1. grab the content of the file
  2. remove any blank lines (just incase you separated some stuff)
  3. remove any comments (just incase you added some...)
  4. add export to all the lines
  5. eval the whole thing

eval $(cat .env | sed -e /^$/d -e /^#/d -e 's/^/export /')

Another option (you don't have to run eval (thanks to @Jaydeep)):

export $(cat .env | sed -e /^$/d -e /^#/d | xargs)

Lastly, if you want to make your life REALLY easy, add this to your ~/.bash_profile:

function source_envfile() { export $(cat $1 | sed -e /^$/d -e /^#/d | xargs); }

(MAKE SURE YOU RELOAD YOUR BASH SETTINGS!!! source ~/.bash_profile or.. just make a new tab/window and problem solved) you call it like this: source_envfile .env

8
votes

The problem with source is that it requires the file to have a proper bash syntax, and some special characters will ruin it: =, ", ', <, >, and others. So in some cases you can just

source development.env

and it will work.

This version, however, withstands every special character in values:

set -a
source <(cat development.env | \
    sed -e '/^#/d;/^\s*$/d' -e "s/'/'\\\''/g" -e "s/=\(.*\)/='\1'/g")
set +a

Explanation:

  • -a means that every bash variable would become an environment variable
  • /^#/d removes comments (strings that start with #)
  • /^\s*$/d removes empty strings, including whitespace
  • "s/'/'\\\''/g" replaces every single quote with '\'', which is a trick sequence in bash to produce a quote :)
  • "s/=\(.*\)/='\1'/g" converts every a=b into a='b'

As a result, you are able to use special characters :)

To debug this code, replace source with cat and you'll see what this command produces.

7
votes

You can use your original script to set the variables, but you need to call it the following way (with stand-alone dot):

. ./minientrega.sh

Also there might be an issue with cat | while read approach. I would recommend to use the approach while read line; do .... done < $FILE.

Here is a working example:

> cat test.conf
VARIABLE_TMP1=some_value

> cat run_test.sh
#/bin/bash
while read line; do export "$line";
done < test.conf
echo "done"

> . ./run_test.sh
done

> echo $VARIABLE_TMP1
some_value
7
votes
t=$(mktemp) && export -p > "$t" && set -a && . ./.env && set +a && . "$t" && rm "$t" && unset t

How it works

  1. Create temp file.
  2. Write all current environment variables values to the temp file.
  3. Enable exporting of all declared variables in the sources script to the environment.
  4. Read .env file. All variables will be exported into current environment.
  5. Disable exporting of all declared variables in the sources script to the environment.
  6. Read the contents of the temp file. Every line would have declare -x VAR="val" that would export each of the variables into environment.
  7. Remove temp file.
  8. Unset the variable holding temp file name.

Features

  • Preserves values of the variables already set in the environment
  • .env can have comments
  • .env can have empty lines
  • .env does not require special header or footer like in the other answers (set -a and set +a)
  • .env does not require to have export for every value
  • one-liner
6
votes

Building on other answers, here is a way to export only a subset of lines in a file, including values with spaces like PREFIX_ONE="a word":

set -a
. <(grep '^[ ]*PREFIX_' conf-file)
set +a
6
votes

I work with docker-compose and .env files on Mac, and wanted to import the .env into my bash shell (for testing), and the "best" answer here was tripping up on the following variable:

.env

NODE_ARGS=--expose-gc --max_old_space_size=2048

Solution

So I ended up using eval, and wrapping my env var defs in single quotes.

eval $(grep -v -e '^#' .env | xargs -I {} echo export \'{}\')

Bash Version

$ /bin/bash --version
GNU bash, version 3.2.57(1)-release (x86_64-apple-darwin18)
Copyright (C) 2007 Free Software Foundation, Inc.
6
votes

If env supports the -S option one may use newlines or escape characters like \n or \t (see env):

env -S "$(cat .env)" command

.env file example:

KEY="value with space\nnewline\ttab\tand
multiple
lines"

Test:

env -S "$(cat .env)" sh -c 'echo "$KEY"'
5
votes

My .env:

#!/bin/bash
set -a # export all variables

#comments as usual, this is a bash script
USER=foo
PASS=bar

set +a #stop exporting variables

Invoking:

source .env; echo $USER; echo $PASS

Reference https://unix.stackexchange.com/questions/79068/how-to-export-variables-that-are-set-all-at-once

4
votes

My requirements were:

  • simple .env file without export prefixes (for compatibility with dotenv)
  • supporting values in quotes: TEXT="alpha bravo charlie"
  • supporting comments prefixed with # and empty lines
  • universal for both mac/BSD and linux/GNU

Full working version compiled from the answers above:

  set -o allexport
  eval $(grep -v '^#' .env | sed 's/^/export /')
  set +o allexport
3
votes

I have issues with the earlier suggested solutions:

  • @anubhava's solution makes writing bash friendly configuration files very annoying very fast, and also - you may not want to always export your configuration.
  • @Silas Paul solution breaks when you have variables that have spaces or other characters that work well in quoted values, but $() makes a mess out of.

Here is my solution, which is still pretty terrible IMO - and doesn't solve the "export only to one child" problem addressed by Silas (though you can probably run it in a sub-shell to limit the scope):

source .conf-file
export $(cut -d= -f1 < .conf-file)
3
votes

Modified from @Dan Kowalczyk

I put this in ~/.bashrc.

set -a
. ./.env >/dev/null 2>&1
set +a

Cross-compatible very well with Oh-my-Zsh's dotenv plugin. (There is Oh-my-bash, but it doesn't have dotenv plugin.)

3
votes

I use this:

source <(cat .env \
  | sed -E '/^\s*#.*/d' \
  | tr '\n' '\000' \
  | sed -z -E 's/^([^=]+)=(.*)/\1\x0\2/g' \
  | xargs -0 -n2 bash -c 'printf "export %s=%q;\n" "${@}"' /dev/null)

First Removing comments:

sed -E '/^\s*#.*/d'

Then converting to null delimiters instead of newline:

tr '\n' '\000'

Then replacing equal with null:

sed -z -E 's/^([^=]+)=(.*)/\1\x0\2/g'

Then printing pairs to valid quoted bash exports (using bash printf for %q):

xargs -0 -n2 bash -c 'printf "export %s=%q;\n" "${@}"' /dev/null

Then finally sourcing all of that.

It should work for just about all cases with all special characters.

3
votes

Use shdotenv

dotenv support for shell and POSIX-compliant .env syntax specification
https://github.com/ko1nksm/shdotenv

eval "$(shdotenv)"

Usage

Usage: shdotenv [OPTION]... [--] [COMMAND [ARG]...]

  -d, --dialect DIALECT  Specify the .env dialect [default: posix]
                           (posix, ruby, node, python, php, go, rust, docker)
  -s, --shell SHELL      Output in the specified shell format [default: posix]
                           (posix, fish)
  -e, --env ENV_PATH     Location of the .env file [default: .env]
                           Multiple -e options are allowed
  -o, --overload         Overload predefined environment variables
  -n, --noexport         Do not export keys without export prefix
  -g, --grep PATTERN     Output only those that match the regexp pattern
  -k, --keyonly          Output only variable names
  -q, --quiet            Suppress all output
  -v, --version          Show the version and exit
  -h, --help             Show this message and exit

Requirements

shdotenv is a single file shell script with embedded awk script.

  • POSIX shell (dash, bash, ksh, zsh, etc)
  • awk (gawk, nawk, mawk, busybox awk)
2
votes

White spaces in the value

There are many great answers here, but I found them all lacking support for white space in the value:

DATABASE_CLIENT_HOST=host db-name db-user 0.0.0.0/0 md5

I have found 2 solutions that work whith such values with support for empty lines and comments.

One based on sed and @javier-buzzi answer:

source <(sed -e /^$/d -e /^#/d -e 's/.*/declare -x "&"/g' .env)

And one with read line in a loop based on @john1024 answer

while read -r line; do declare -x "$line"; done < <(egrep -v "(^#|^\s|^$)" .env)

The key here is in using declare -x and putting line in double quotes. I don't know why but when you reformat the loop code to multiple lines it won't work — I'm no bash programmer, I just gobbled together these, it's still magic to me :)

2
votes

First, create an environment file that will have all the key-value pair of the environments like below and named it whatever you like in my case its env_var.env

MINIENTREGA_FECHALIMITE="2011-03-31"
MINIENTREGA_FICHEROS="informe.txt programa.c"
MINIENTREGA_DESTINO="./destino/entrega-prac1"

Then create a script that will export all the environment variables for the python environment like below and name it like export_env.sh

#!/usr/bin/env bash

ENV_FILE="$1"
CMD=${@:2}

set -o allexport
source $ENV_FILE
set +o allexport

$CMD

This script will take the first argument as the environment file then export all the environment variable in that file and then run the command after that.

USAGE:

./export_env.sh env_var.env python app.py
2
votes

Here's my take on this. I had the following requirements:

  • Ignore commented lines
  • Allow spaces in the value
  • Allow empty lines
  • Ability to pass a custom env file while defaulting to .env
  • Allow exporting as well as running commands inline
  • Exit if env file doesn't exist
source_env() {
  [ "$#" -eq 1 ] && env="$1" || env=".env"
  [ -f "$env" ] || { echo "Env file $env doesn't exist"; return 1; }
  eval $(grep -Ev '^#|^$' "$env" | sed -e 's/=\(.*\)/="\1/g' -e 's/$/"/g' -e 's/^/export /')
}

Usage after saving the function to your .bash_profile or equivalent:

source_env                # load default .env file
source_env .env.dev       # load custom .env file
(source_env && COMMAND)   # run command without saving vars to environment

Inspired by Javier and some of the other comments.