Render jinja2 templates on the command line with shell environment variables
pip install envtpl
Say you have a configuration file called whatever.conf that looks like this
foo = 123
bar = "abc"
You can use envtpl to set foo and bar from the command line by creating a file called whatever.conf.tpl
foo = {{ FOO }}
bar = "{{ BAR }}"
If you run
FOO=123 BAR=abc envtpl < whatever.conf.tpl > whatever.conf
you'll get back the original whatever.conf.
You can also specify default values
foo = {{ FOO | default(123) }}
bar = "{{ BAR | default("abc") }}"
Running
FOO=456 envtpl < whatever.conf.tpl > whatever.conf
will generate
foo = 456
bar = "abc"
This is all standard Jinja2 syntax, so you can do things like
{% if BAZ is defined %}
foo = 123
{% else %}
foo = 456
{% endif %}
bar = "abc"
If an environment variable is missing, envtpl will throw an error
$ echo '{{ FOO }} {{ BAR }}' | FOO=123 envtpl
Error: 'BAR' is undefined
You can change this behaviour to insert empty strings instead by passing the --allow-missing flag.
Instead of reading from stdin and writing to stdout, you can pass the input filename as an optional positional argument,
and set the output filename with the --output-file (-o) argument.
envtpl -o whatever.conf whatever.conf.tpl
As a convenience, if you don't specify an output filename and the input filename ends with .tpl, the output filename will be the input filename without the .tpl extension, i.e.
envtpl whatever.conf.tpl
# is equivalent to
envtpl -o whatever.conf whatever.conf.tpl
By default, envtpl will delete the input template file. You can keep it by passing the --keep-template flag.
There's a special environment(prefix='') function that you can use as a kind of wildcard variable. If you have hello.tpl
hello = {{ FOO }}
{% for key, value in environment('MY_') %}{{ key }} = {{ value }}
{% endfor %}
and compile it using
FOO=world MY_baz=qux MY_foo=bar envtpl hello.tpl
You end up with
hello = world
baz = qux
foo = bar
If you need more complex data structures you can pass in JSON as a string and use the from_json filter to turn it into an object you can use in your template:
FOO='[{"v": "hello"}, {"v": "world"}]' envtpl <<< '{% for x in FOO | from_json %}{{ x.v }}{% endfor %}'
gives
helloworld
and
FOO='{"bar": "baz"}' envtpl <<< '{{ (FOO | from_json).bar }}'
renders
baz
I use this script quite a lot in Docker images. Here is an example.
In the CMD script I set up the runtime configuration using environment variables, e.g.
#!/bin/bash
# start_container
envtpl /etc/redis.conf.tpl
redis-server
This is the use case I've optimised for, so that's why envtpl by default will delete the original template file.