A REPL to make exploring data with JQ easier.

I'm a huge fan of JQ and use it in a lot of small utilities or to explore JSON APIs from the command line, and I often found myself doing things like this:

aws ec2 describe-images | jq '.Images'

# Hmm, that's still large
aws ec2 describe-images | jq '.Images | keys'

aws ec2 describe-images | jq '.Images | .Tags'

i.e. I was using jq as a tool to explore a complex JSON data structure -- but each invokation of aws ec2 describe-images took 5 to 15 seconds which which made the process of building a jq filter quite jaring.

Now, I could have just piped the result of the aws command to a file and then invoked JQ on that many times, and to start with that's what I did. But it turned out that each of the Images above has differente keys, so finding the error with jq alone was painful, so in another terminal I fired up ipython, loaded that JSON file into a python dictionary and started exploring the data that way. Somehow it got suggested that a REPL for JQ would be the right tool for this job - and thus the seed for this tool was planted. (P.S. Samir and James: this is all your fault for egging me on)

Sort of. I'm working on it slowly.

I am using this project as excuse and reason to learn Go so it will take me a while to get it functional and bug free. And then even longer so that other pepole can use it.

$ aws ec2 describe-images --owner self | jq-repl
  0 » type
$out[0] = "object"

  1 » keys
$out[1] = [
    "Images"
]

  2 » .Images[0]
$out[2] = {
  "VirtualizationType": "hvm",
  "Name": "leader 2015-11-05T16-50-35Z",
  "Tags": [
    {
      "Value": "2015-11-05T16:50:35Z",
      "Key": "build_date"
    }
  ],
  "Hypervisor": "xen",
  "SriovNetSupport": "simple",
  "ImageId": "ami-abc01234",
  "State": "available",
  "BlockDeviceMappings": [
    {
      "DeviceName": "/dev/sda1",
      "Ebs": {
        "DeleteOnTermination": true,
        "SnapshotId": "snap-01234fed",
        "VolumeSize": 16,
        "VolumeType": "gp2",
        "Encrypted": false
      }
    },
  ],
  "Architecture": "x86_64",
  "RootDeviceType": "ebs",
  "RootDeviceName": "/dev/sda1",
  "CreationDate": "2015-11-05T16:55:15.000Z",
  "Public": false,
  "ImageType": "machine",
  "Description": "My AMI"
}

So far all fairly mundane. This is where I think things start to get interesting - you will be able to refer back to previous results.

  3 » $_ | .Name
$out[3] = "leader 2015-11-05T16-50-35Z"

We define some special variables.

• $_

  The previous result. Equivalent to $out[-1]

• $__

  The result before last. Equivalent to $out[-2]

• $out

  An array of all previous results. You can use negative indcies to count backwards from the end.

• Accept JSON input on stdin
• Accept JSON input from a file
• Sepcify a command to run to get input
• Be able to re-execute command to refresh input
• Autocompletion
• Refer to previous results
• Build up complex filters by refering to previous results.
• Automated tests for the JqRepl methods

• JQ souce code and anything it needs to compile

It doesn't do much of anything yet. But to build it you will need to do something like this:

curl -fL https://github.com/stedolan/jq/releases/download/jq-1.5/jq-1.5.tar.gz | tar -zx
cd jq-1.5
./configure --disable-maintainer-mode --prefix=$PWD/BUILD
# We could run `make install` but we only actually need these components.
make install-libLTLIBRARIES install-includeHEADERS
go test ./...

I have no idea if this will work on platforms other than OSX right now. I will work on that later once I have some basic functionality