Tip: Using BASH variables in jello queries

[kellyjonbrazil]

Here's how to use BASH variables in your jello queries. Do you have any other methods?

Updating an existing Object with an Object contained in a BASH variable

Referencing the variable via BASH

Let's say you have the following JSON file:

$ cat file.json | jello
{
  "nodes": {
    "node1": {
      "ip": "10.10.10.11",
      "memory": 1024,
      "cpus": 1
    },
    "node2": {
      "ip": "10.10.10.12",
      "memory": 1024,
      "cpus": 1
    }
  }
}

And you want to insert the following Object under nodes:

{
  "node3":{
    "ip":"10.10.10.13",
    "memory":1024,
    "cpus":1
  }
}

You can do it in BASH with the following:

$ node='{
  "node3":{
    "ip":"10.10.10.13",
    "memory":1024,
    "cpus":1
  }
}'
$ cat file.json | jello "_.nodes.update($node); _"
{
  "nodes": {
    "node1": {
      "ip": "10.10.10.11",
      "memory": 1024,
      "cpus": 1
    },
    "node2": {
      "ip": "10.10.10.12",
      "memory": 1024,
      "cpus": 1
    },
    "node3": {
      "ip": "10.10.10.13",
      "memory": 1024,
      "cpus": 1
    }
  }
}

But notice I used double quotes there? I did that so I could reference $node without any funky escaping. If we need to put single quotes around our query, then we can do this another way:

Referencing the variable within jello via os.getenv()

Let's do the same thing, but this time we'll surround our query in single quotes. When we do that, we no longer have access to the $node variable, since it will become a string literal. One way around this is to export the NODE variable to the environment and then reference it within jello using os.getenv(). os is conveniently already imported so we can use it right away:

$ export NODE='{
  "node3": {
    "ip":"10.10.10.13",
    "memory":1024,
    "cpus":1
  }
}'
$ cat file.json | jello '\
new = json.loads(os.getenv("NODE"))
_.nodes.update(new)
_'
{
  "nodes": {
    "node1": {
      "ip": "10.10.10.11",
      "memory": 1024,
      "cpus": 1
    },
    "node2": {
      "ip": "10.10.10.12",
      "memory": 1024,
      "cpus": 1
    },
    "node3": {
      "ip": "10.10.10.13",
      "memory": 1024,
      "cpus": 1
    }
  }
}

Notice the use of json.loads() there? Since environment variables are always assigned as strings, we need to convert the string into a dictionary somehow. json.loads() is an easy way to do it, though there are other methods like using ast or even eval.

Do you know of any other cool tricks or ways to accomplish this task? I'm also thinking about possibly adding a --var argument to jello in the future so you can assign a variable directly in the jello command instead of using the environment. I'll have to noodle on that to see if it's something users really want.