JJ is a command line utility that provides a fast and simple way to retrieve or update values from JSON documents. It's powered by GJSON and SJSON under the hood.
It's fast because it avoids parsing irrelevant sections of json, skipping over values that do not apply, and aborts as soon as the target value has been found or updated.
brew install tidwall/jj/jj
make
Or download a pre-built binary for Linux, OSX, Windows, or FreeBSD.
$ jj -h
usage: jj [-v value] [-purOD] [-i infile] [-o outfile] keypath
examples: jj keypath read value from stdin
or: jj -i infile keypath read value from infile
or: jj -v value keypath edit value
or: jj -v value -o outfile keypath edit value and write to outfile
options:
-v value Edit JSON key path value
-p Make json pretty, keypath is optional
-u Make json ugly, keypath is optional
-r Use raw values, otherwise types are auto-detected
-n Do not output color or extra formatting
-O Performance boost for value updates
-D Delete the value at the specified key path
-l Output array values on multiple lines
-i infile Use input file instead of stdin
-o outfile Use output file instead of stdout
keypath JSON key path (like "name.last")
JJ uses a path syntax for finding values.
Get a string:
$ echo '{"name":{"first":"Tom","last":"Smith"}}' | jj name.last
Smith
Get a block of JSON:
$ echo '{"name":{"first":"Tom","last":"Smith"}}' | jj name
{"first":"Tom","last":"Smith"}
Try to get a non-existent key:
$ echo '{"name":{"first":"Tom","last":"Smith"}}' | jj name.middle
null
Get the raw string value:
$ echo '{"name":{"first":"Tom","last":"Smith"}}' | jj -r name.last
"Smith"
Get an array value by index:
$ echo '{"friends":["Tom","Jane","Carol"]}' | jj friends.1
Jane
There's support for JSON Lines using the ..
path prefix.
Which when specified, treats the multi-lined document as an array.
For example:
{"name": "Gilbert", "age": 61}
{"name": "Alexa", "age": 34}
{"name": "May", "age": 57}
..# >> 4
..1 >> {"name": "Alexa", "age": 34}
..#.name >> ["Gilbert","Alexa","May"]
..#[name="May"].age >> 57
The path syntax for setting values has a couple of tiny differences than for getting values.
The -v value
option is auto-detected as a Number, Boolean, Null, or String.
You can override the auto-detection and input raw JSON by including the -r
option.
This is useful for raw JSON blocks such as object, arrays, or premarshalled strings.
Update a value:
$ echo '{"name":{"first":"Tom","last":"Smith"}}' | jj -v Andy name.first
{"name":{"first":"Andy","last":"Smith"}}
Set a new value:
$ echo '{"name":{"first":"Tom","last":"Smith"}}' | jj -v 46 age
{"age":46,"name":{"first":"Tom","last":"Smith"}}
Set a new nested value:
$ echo '{"name":{"first":"Tom","last":"Smith"}}' | jj -v relax task.today
{"task":{"today":"relax"},"name":{"first":"Tom","last":"Smith"}}
Replace an array value by index:
$ echo '{"friends":["Tom","Jane","Carol"]}' | jj -v Andy friends.1
{"friends":["Tom","Andy","Carol"]}
Append an array:
$ echo '{"friends":["Tom","Jane","Carol"]}' | jj -v Andy friends.-1
{"friends":["Tom","Andy","Carol","Andy"]}
Set an array value that's past the bounds:
$ echo '{"friends":["Tom","Jane","Carol"]}' | jj -v Andy friends.5
{"friends":["Tom","Andy","Carol",null,null,"Andy"]}
Set a raw block of JSON:
$ echo '{"name":"Carol"}' | jj -r -v '["Tom","Andy"]' friends
{"friends":["Tom","Andy"],"name":"Carol"}
Start new JSON document:
$ echo '' | jj -v 'Sam' name.first
{"name":{"first":"Sam"}}
Delete a value:
$ echo '{"age":46,"name":{"first":"Tom","last":"Smith"}}' | jj -D age
{"name":{"first":"Tom","last":"Smith"}}
Delete an array value by index:
$ echo '{"friends":["Andy","Carol"]}' | jj -D friends.0
{"friends":["Carol"]}
Delete last item in array:
$ echo '{"friends":["Andy","Carol"]}' | jj -D friends.-1
{"friends":["Andy"]}
The -O
option can be used when the caller expects that a value at the
specified keypath already exists.
Using this option can speed up an operation by as much as 6x, but slow down as much as 20% when the value does not exist.
For example:
echo '{"name":{"first":"Tom","last":"Smith"}}' | jj -v Tim -O name.first
The -O
tells jj that the name.first
likely exists so try a fasttrack operation first.
The -p
flag will make the output json pretty.
$ echo '{"name":{"first":"Tom","last":"Smith"}}' | jj -p name
{
"first": "Tom",
"last": "Smith"
}
Also the keypath is optional when the -p
flag is specified, allowing for the entire json document to be made pretty.
$ echo '{"name":{"first":"Tom","last":"Smith"}}' | jj -p
{
"name": {
"first": "Tom",
"last": "Smith"
}
}
The -u
flag will compress the json into the fewest characters possible by squashing newlines and spaces.
A quick comparison of jj to jq. The test json file is 180MB file of 206,560 city parcels in San Francisco.
Tested on a 2015 Macbook Pro running jq 1.5 and jj 1.0.0
jq:
$ time cat citylots.json | jq -cM .features[10000].properties.LOT_NUM
"091"
real 0m5.486s
user 0m4.870s
sys 0m0.686s
jj:
$ time cat citylots.json | jj -r features.10000.properties.LOT_NUM
"091"
real 0m0.354s
user 0m0.161s
sys 0m0.321s
jq:
$ time cat citylots.json | jq -cM '.features[10000].properties.LOT_NUM="12A"' > /dev/null
real 0m13.579s
user 0m16.484s
sys 0m1.310s
jj:
$ time cat citylots.json | jj -O -v 12A features.10000.properties.LOT_NUM > /dev/null
real 0m0.431s
user 0m0.201s
sys 0m0.295s
Josh Baker @tidwall
JJ source code is available under the MIT License.