Jku

Jku is a command-line tool to filter and/or modify a JSON stream. It is heavily inspired by jq.

$ jku
Filter and modify a JSON stream
Usage: jku [-p] [-s <SEP>] [-f <FILTER>] [-t <TRANSFORM>]

Options:
  -f, --filter     An expression to filter by               [string]
  -t, --transform  Transform expression                     [string]
  -p, --pretty     Pretty-print resulting JSON              [boolean]  [default: false]
  -s, --separator  The separator with which to join arrays  [string]  [default: "\t"]

Please supply a filter and/or a transform.

Installation

$ npm install -g jku

Example Usage

$ cat foo.json
{ "foo": 1, "bar": "a" }
{ "foo": 2, "bar": "b" }
{ "foo": 3 }

$ cat foo.json | jku -f '_.bar' -t '_.foo'
1
2

The above filters for rows which have a truthy bar value and outputs the value of the foo key.

Filtering and Transforming Expressions

Expressions are evaluated as Javascript code, with _ representing the current JSON object.

Filtering

By specifying a filter expression, jku will only emit JSON objects where the expression is truthy.

Transforming

Based on the type of the value generated from transform expression, jku will output to different formats:

Scalar values, e.g. String, Number, will return the string representation of the value.

$ echo '{ "foo" : 1 }' | jku -t '_.foo'
1

$ echo '{ "foo" : "bar" }' | jku -t '_.foo'
bar

Array value will join the elements of the array with a tab. The separator can be configured by supplying the -s option.

$ echo '{ "foo" : [1,2,3] }' | jku -t '_.foo'
1	2	3

Object value will output the JSON representation of the object. Pretty-printing can be enabled by supplying the -p option.

$ echo '{ "foo" : "bar" }' | jku -t '{ baz: _.foo }'
{"baz":"bar"}

More examples

Filter by number of keys

$ echo '{ "foo": 1, "bar": 2, "baz": 3 }' | jku -p -f 'Object.keys(_).length > 2'
{
  "foo": 1,
  "bar": 2,
  "baz": 3
}

Generate CSV

$ cat foo.json
{ "foo": 1, "bar": "a" }
{ "foo": 2, "bar": "b" }
{ "foo": 3 }

$ cat foo.json | jku -s ',' -t '[_.foo, _.bar]'
1,a
2,b
3,

Convert timestamps to human readable

The following outputs new JSON documents with the timestamp replaced by a human readable string and the count represented in hexadecimal form. This is an example of how arbitrary Javascript functions can be used in expressions.

$ cat foo.json
{ "timestamp": 1372068657151, "count": 17 }
{ "timestamp": 1372068659841, "count": 136 }
{ "timestamp": 1372068668142, "count": 255 }

$ cat foo.json | jku -t '_.timestamp = new Date(_.timestamp), _.count = _.count.toString(16), _'
{"timestamp":"2013-06-24T10:10:57.151Z","count":"11"}
{"timestamp":"2013-06-24T10:10:59.841Z","count":"88"}
{"timestamp":"2013-06-24T10:11:08.142Z","count":"ff"}

Contribute

Some things I would really like to see:

• Multi-line JSON support. I guess a streaming JSON parser would be useful.
• CoffeeScript syntax support

In any case, make sure to run the test examples with make test. All examples in this README are runnable, so simply add an example to include it in the test suite.