Get environment variables by prefix

[koenpunt]

I like the idea of parsing the variables into objects, although parsing environment variables like TWITTER_ACCESS_TOKEN_SECRET result is something like:

twitter:
  { access:
     { token:
       { key: 'xxxx',
         secret: 'xxxx' } }
    consumer:
     { key: 'xxxxxxx',
       secret: 'xxxxxxx' } }

Where I previously could use a JSON object like this:

var client = new Tuiter(config.twitter);

I now have to supply every property separately:

var client = new Tuiter({
  consumer_key: config.twitter.consumer.key,
  consumer_secret: config.twitter.consumer.secret,
  access_token_key: config.twitter.access.token.key,
  access_token_secret: config.twitter.access.token.secret
});

And I know with environment variables would it be the same:

var client = new Tuiter({
  consumer_key: config.twitter..TWITTER_CONSUMER_KEY,
  consumer_secret: process.env.TWITTER_CONSUMER_SECRET,
  access_token_key: process.env.TWITTER_ACCESS_TOKEN_KEY,
  access_token_secret: process.env.TWITTER_ACCESS_TOKEN_SECRET
});

But I think it would be nice to fetch some variables by prefix, but not completely teared apart.

Something in the likes of:

var env = require('environmental');
var twitter = env.filter('twitter');
// Where twitter is
{
  consumer_key: 'xxxxx',
  consumer_secret: 'xxxxx',
  access_token_key: 'xxxxx',
  access_token_secret: 'xxxxx'
}

[kvz]

I see a few options:

1. access the var via process.env
2. name the var TWITTER_KEYS_CONSUMER vs TWITTER_CONSUMER_KEY
3. name the var TWITTER_CONSUMERKEY vs TWITTER_CONSUMER_KEY
4. make the delimiter configurable. eg. __ and then name the var TWITTER__CONSUMER_KEY vs TWITTER_CONSUMER_KEY
5. provide a method that let's you flat/unflatten on a level that the developer chooses (your solution)

For me personally, 3 has worked well enough. But I think if Environmental needs to address this, I feel most for 4.

What do you think?

[koenpunt]

The problem is that in this case the library (Tuiter) expects the keys of the configuration to be like

{
  consumer_key: 'xxxxx',
  consumer_secret: 'xxxxx',
  access_token_key: 'xxxxx',
  access_token_secret: 'xxxxx'
}

so this rules out option 3.

Customizable delimiter will convenient anyway.

Another option could be to limit nesting with a numeric value. When set to 1, TWITTER_ACCESS_TOKEN_SECRET results in:

twitter: {
   access_token_secret: 'xxx'
}

When set to 2 it results in:

twitter: {
   access: {
      token_secret: 'xxx'
   }
}

Etc.

[kvz]

I like the idea of making the delimiter and nesting level configurable! Would you also be interested in hacking on that?