JSON diff and patch

jd is a commandline utility and Go library for diffing and patching JSON and YAML values. It supports a native jd format (similar to unified format) as well as JSON Merge Patch (RFC 7386) and a subset of JSON Patch (RFC 6902). Try it out at http://play.jd-tool.io/.

Example

Diff jd a.json b.json:

{"foo":["bar","baz"]}

{"foo":["bar","bam","boom"]}

Output:

@ ["foo",1]
  "bar"
- "baz"
+ "bam"
+ "boom"
]

Features

1. Human-friendly format, similar to Unified Diff.
2. Produces a minimal diff between array elements using LCS algorithm.
3. Adds context before and after when modifying an array to prevent bad patches.
4. Create and apply structural patches in jd, patch (RFC 6902) and merge (RFC 7386) patch formats.
5. Translates between patch formats.
6. Includes Web Assembly-based UI (no network calls).

Installation

GitHub Action:

    - name: Diff A and B
      id: diff
      uses: josephburnett/[email protected]
      with:
        args: a.json b.json
    - name: Print the diff
      run: echo '${{ steps.diff.outputs.output }}'
    - name: Check the exit code
      run: if [ "${{ steps.diff.outputs.exit_code }}" != "1" ]; then exit 1; fi

To get the jd commandline utility:

• run brew install jd, or
• run go install github.com/josephburnett/jd/v2/jd@latest, or
• visit https://github.com/josephburnett/jd/releases/latest and download the pre-built binary for your architecture/os, or
• run in a Docker image jd(){ docker run --rm -i -v $PWD:$PWD -w $PWD josephburnett/jd "$@"; }.

To use the jd web UI:

• visit http://play.jd-tool.io/, or
• run jd -port 8080 and visit http://localhost:8080.

Note: to include the UI when building from source, use the Makefile.

Command line usage

Usage: jd [OPTION]... FILE1 [FILE2]
Diff and patch JSON files.

Prints the diff of FILE1 and FILE2 to STDOUT.
When FILE2 is omitted the second input is read from STDIN.
When patching (-p) FILE1 is a diff.

Options:
  -color       Print color diff.
  -p           Apply patch FILE1 to FILE2 or STDIN.
  -o=FILE3     Write to FILE3 instead of STDOUT.
  -opts='[]'   JSON array of options. Supports global options and PathOptions.
               Global: ["SET"], ["MULTISET"], [{"precision":0.1}], [{"setkeys":["id"]}], ["DIFF_ON"], ["DIFF_OFF"]
               PathOptions target specific paths: [{"@":["path"],"^":["SET"]}]
               Example: [{"@":["users"],"^":["SET"]},{"@":["scores",0],"^":[{"precision":0.1}]}]
  -set         Treat arrays as sets. Same as -opts='["SET"]'.
  -mset        Treat arrays as multisets (bags). Same as -opts='["MULTISET"]'.
  -setkeys     Keys to identify set objects. Same as -opts='[{"setkeys":["key1","key2"]}]'.
  -yaml        Read and write YAML instead of JSON.
  -port=N      Serve web UI on port N
  -precision=N Maximum absolute difference for numbers to be equal.
               Same as -opts='[{"precision":N}]'. Example: -precision=0.00001
  -f=FORMAT    Read and write diff in FORMAT "jd" (default), "patch" (RFC 6902) or
               "merge" (RFC 7386)
  -t=FORMATS   Translate FILE1 between FORMATS. Supported formats are "jd",
               "patch" (RFC 6902), "merge" (RFC 7386), "json" and "yaml".
               FORMATS are provided as a pair separated by "2". E.g.
               "yaml2json" or "jd2patch".

Examples:
  jd a.json b.json
  cat b.json | jd a.json
  jd -o patch a.json b.json; jd patch a.json
  jd -set a.json b.json
  jd -f patch a.json b.json
  jd -f merge a.json b.json
  jd -opts='[{"@":["items"],"^":["SET"]}]' a.json b.json
  jd -opts='[{"@":["temperature"],"^":[{"precision":0.1}]}]' a.json b.json

Command Line Option Details

setkeys This option determines what keys are used to decide if two objects 'match'. Then the matched objects are compared, which will return a diff if there are differences in the objects themselves, their keys and/or values. You shouldn't expect this option to mask or ignore non-specified keys, it is not intended as a way to 'ignore' some differences between objects.

PathOptions: Targeted Comparison Options

PathOptions allow you to apply different comparison semantics to specific paths in your JSON/YAML data. This enables precise control over how different parts of your data are compared.

PathOption Syntax:

{"@": ["path", "to", "target"], "^": [options]}

• @ (At): JSON path array specifying where to apply the option
• ^ (Then): Array of options to apply at that path

Supported Options:

• "SET": Treat array as a set (ignore order and duplicates)
• "MULTISET": Treat array as a multiset (ignore order, count duplicates)
• {"precision": N}: Numbers within N are considered equal
• {"setkeys": ["key1", "key2"]}: Match objects by specified keys
• "DIFF_ON": Enable diffing at this path (default behavior)
• "DIFF_OFF": Disable diffing at this path, ignore all changes

Examples:

Treat specific array as a set while others remain as lists:

jd -opts='[{"@":["tags"],"^":["SET"]}]' a.json b.json

Apply precision to specific temperature field:

jd -opts='[{"@":["sensor","temperature"],"^":[{"precision":0.1}]}]' a.json b.json

Multiple PathOptions - SET on one path, precision on another:

jd -opts='[{"@":["items"],"^":["SET"]}, {"@":["price"],"^":[{"precision":0.01}]}]' a.json b.json

Target specific array index:

jd -opts='[{"@":["measurements", 0],"^":[{"precision":0.05}]}]' a.json b.json

Apply to root level:

jd -opts='[{"@":[],"^":["SET"]}]' a.json b.json

Ignore specific fields (deny-list approach):

jd -opts='[{"@":["timestamp"],"^":["DIFF_OFF"]}, {"@":["metadata","generated"],"^":["DIFF_OFF"]}]' a.json b.json

Allow-list approach - ignore everything except specific fields:

jd -opts='[{"@":[],"^":["DIFF_OFF"]}, {"@":["userdata"],"^":["DIFF_ON"]}]' a.json b.json

Nested override - ignore parent but include specific child:

jd -opts='[{"@":["config"],"^":["DIFF_OFF"]}, {"@":["config","user_settings"],"^":["DIFF_ON"]}]' a.json b.json

Library usage

Note: import only release commits (v2.Y.Z) because master can be unstable.

Note: the v2 library replaces the v1 (lib) library. V2 adds diff context, minimal array diffs and hunk-level metadata. However the format is not backward compatable. You should use v2.

import (
	"fmt"
	jd "github.com/josephburnett/jd/v2"
)

func ExampleJsonNode_Diff() {
	a, _ := jd.ReadJsonString(`{"foo":["bar"]}`)
	b, _ := jd.ReadJsonString(`{"foo":["baz"]}`)
	fmt.Print(a.Diff(b).Render())
	// Output:
	// @ ["foo",0]
	// [
	// - "bar"
	// + "baz"
	// ]
}

func ExampleJsonNode_Patch() {
	a, _ := jd.ReadJsonString(`["foo"]`)
	diff, _ := jd.ReadDiffString(`
@ [1]
  "foo"
+ "bar"
]
`)
	b, _ := a.Patch(diff)
	fmt.Print(b.Json())
	// Output:
	// ["foo","bar"]
}

func ExamplePathOptions() {
	// Apply SET semantics to specific array path
	a, _ := jd.ReadJsonString(`{"tags":["red","blue","green"], "items":[1,2,3]}`)
	b, _ := jd.ReadJsonString(`{"tags":["green","red","blue"], "items":[3,2,1]}`)
	
	// Only treat "tags" as a set, "items" remain as list
	opts, _ := jd.ReadOptionsString(`[{"@":["tags"],"^":["SET"]}]`)
	diff := a.Diff(b, opts...)
	fmt.Print(diff.Render())
	// Output:
	// @ ["items",0]
	// [
	// + 3
	// + 2
	//   1
	// @ ["items",3]
	//   1
	// - 2
	// - 3
	// ]
}

func ExampleMultiplePathOptions() {
	a, _ := jd.ReadJsonString(`{"temp":20.12, "pressure":1013.25, "tags":["A","B","C"]}`)
	b, _ := jd.ReadJsonString(`{"temp":20.15, "pressure":1013.30, "tags":["C","A","B"]}`)
	
	// Apply precision to temp, exact match to pressure, SET semantics to tags
	opts, _ := jd.ReadOptionsString(`[
		{"@":["temp"],"^":[{"precision":0.1}]},
		{"@":["tags"],"^":["SET"]}
	]`)
	diff := a.Diff(b, opts...)
	fmt.Print(diff.Render())
	// Output:
	// @ ["pressure"]
	// - 1013.25
	// + 1013.3
}

func ExampleSelectiveDiffing() {
	a, _ := jd.ReadJsonString(`{"userdata":"important","system":"ignore1","timestamp":"2023-01-01"}`)
	b, _ := jd.ReadJsonString(`{"userdata":"changed","system":"ignore2","timestamp":"2023-01-02"}`)
	
	// Allow-list approach: ignore everything except userdata
	opts, _ := jd.ReadOptionsString(`[
		{"@":[],"^":["DIFF_OFF"]},
		{"@":["userdata"],"^":["DIFF_ON"]}
	]`)
	diff := a.Diff(b, opts...)
	fmt.Print(diff.Render())
	// Output:
	// @ ["userdata"]
	// - "important"
	// + "changed"
}

func ExampleNestedOverride() {
	a, _ := jd.ReadJsonString(`{"config":{"system":"val1","user_settings":"setting1"}}`)
	b, _ := jd.ReadJsonString(`{"config":{"system":"val2","user_settings":"setting2"}}`)
	
	// Ignore config changes except for user_settings
	opts, _ := jd.ReadOptionsString(`[
		{"@":["config"],"^":["DIFF_OFF"]},
		{"@":["config","user_settings"],"^":["DIFF_ON"]}
	]`)
	diff := a.Diff(b, opts...)
	fmt.Print(diff.Render())
	// Output:
	// @ ["config","user_settings"]
	// - "setting1"
	// + "setting2"
}

Diff Language (v2)

The jd v2 diff format is a human-readable structural diff format with context and metadata support.

Format Overview

A diff consists of:

• Options header (optional): Shows the options used to create the diff
• Metadata lines (optional): Start with ^ and specify hunk-level metadata
• Diff hunks: Start with @ and specify the path, followed by changes and context

Options Header

When options are provided to jd, they are displayed at the beginning of the diff to show how it was produced. Each option appears on its own line starting with ^ :

^ "SET"
^ {"precision":0.001}
@ ["items",{}]
- "old-item"
+ "new-item"

This feature helps understand:

• Whether arrays were treated as sets ("SET") or multisets ("MULTISET")
• What precision was used for number comparisons ({"precision":N})
• Which keys identify set objects ({"setkeys":["key1","key2"]})
• Path-specific options ({"@":["path"],"^":["OPTION"]})
• Whether merge semantics were applied ("MERGE")
• If color output was requested ("COLOR")

The options header is informational and helps with debugging diff behavior. Note that diffs with options headers can still be parsed and applied as patches.

EBNF Grammar

Diff ::= OptionsHeader* (MetadataLine | DiffHunk)*

OptionsHeader ::= '^' SP JsonValue NEWLINE

MetadataLine ::= '^' SP JsonObject NEWLINE

DiffHunk ::= '@' SP JsonArray NEWLINE
             ContextLine*
             (RemoveLine | AddLine)*
             ContextLine*

ContextLine ::= SP SP JsonValue NEWLINE

RemoveLine ::= '-' SP JsonValue NEWLINE

AddLine ::= '+' SP JsonValue NEWLINE

JsonArray ::= '[' (PathElement (',' PathElement)*)? ']'

PathElement ::= JsonString        // Object key: "foo"
              | JsonNumber        // Array index: 0 
              | EmptyObject       // Set marker: {}
              | EmptyArray        // List marker: [] 
              | ObjectWithKeys    // Set keys: {"id":"value"}
              | ArrayWithObject   // Multiset: [{}] or [{"id":"value"}]

Note: Railroad diagram at /ebnf.png needs updating for v2 format.

Path Elements Reference

Element	Description	Example Path
"key"	Object field access	["user","name"]
0, 1, etc.	Array index access	["items",0]
{}	Treat array as set (ignore order/duplicates)	["tags",{}]
[]	Explicit list marker	["values",[]]
{"id":"val"}	Match objects by specific key	["users",{"id":"123"}]
[{}]	Treat as multiset (ignore order, count duplicates)	["counts",[{}]]
[{"key":"val"}]	Match multiset objects by key	["items",[{"id":"456"}]]

Line Types

• @ [path]: Diff hunk header specifying the location
• ^ {metadata}: Metadata for the following hunks (inherits downward)
• value: Context lines (spaces) - elements that provide context
• - value: Remove lines - values being removed
• + value: Add lines - values being added

Core Examples

Simple Object Change

@ ["name"]
- "Alice"
+ "Bob"

Array Element with Context

@ ["items",1]
  "apple"
+ "banana" 
  "cherry"

Set Operations (Ignore Order)

@ ["tags",{}]
- "urgent"
+ "completed"
+ "reviewed"

Object Identification by Key

@ ["users",{"id":"123"},"status"]
- "pending"
+ "active"

Multiset Operations

@ ["scores",[{}]]
- 85
- 92
+ 88
+ 95
+ 95

Advanced Examples

Merge Patch Metadata

^ {"Merge":true}
@ ["config"]
- {"timeout":30,"retries":3}
+ {"timeout":60,"retries":5,"debug":true}

Complex List Context

@ ["matrix",1,2]
  [[1,2,3],[4,5,6]]
- 6
+ 9
  [7,8,9]
]

Nested Set with PathOptions

@ ["department","employees",{"employeeId":"E123"},"projects",{}]
- "ProjectA"
+ "ProjectB" 
+ "ProjectC"

Multiple Hunks with Inheritance

^ {"Merge":true}
@ ["user","preferences"] 
+ {"theme":"dark","notifications":true}
@ ["user","lastLogin"]
+ "2023-12-01T10:30:00Z"

Integration with PathOptions

The path syntax directly corresponds to PathOption targeting:

• Diff path ["users",{}] ↔ PathOption {"@":["users"],"^":["SET"]}
• Diff path ["items",{"id":"123"}] ↔ PathOption with SetKeys targeting
• Diff path ["scores",[{}]] ↔ PathOption {"@":["scores"],"^":["MULTISET"]}

This allows fine-grained control over how different parts of your data structures are compared and diffed.

Cookbook

Use git diff to produce a structural diff:

Option 1: Direct external tool (simple, one-off usage)

git difftool -yx jd @ -- foo.json

Use this when you want a quick structural diff without any setup. Works immediately if jd is in your PATH.

Option 2: Configure as git diff driver (integrated, regular usage)

# One-time setup
git config diff.jd.command 'jd --git-diff-driver'
echo "*.json diff=jd" >> .gitattributes

# Then use with any git diff command:
git diff foo.json
git difftool -t jd foo.json

Use this approach if you regularly work with JSON files. It integrates jd into git's diff system, automatically using structural diffs for JSON files across all git commands. The .gitattributes file can be committed to share this behavior with your team.

Example output from either approach:

@ ["foo"]
- "bar"
+ "baz"

See what changes in a Kubernetes Deployment:

kubectl get deployment example -oyaml > a.yaml
kubectl edit deployment example
# change cpu resource from 100m to 200m
kubectl get deployment example -oyaml | jd -yaml a.yaml

output (showing all changes):

@ ["metadata","annotations","deployment.kubernetes.io/revision"]
- "2"
+ "3"
@ ["metadata","generation"]
- 2
+ 3
@ ["metadata","resourceVersion"]
- "4661"
+ "5179"
@ ["spec","template","spec","containers",0,"resources","requests","cpu"]
- "100m"
+ "200m"
@ ["status","conditions",1,"lastUpdateTime"]
- "2021-12-23T09:40:39Z"
+ "2021-12-23T09:41:49Z"
@ ["status","conditions",1,"message"]
- "ReplicaSet \"nginx-deployment-787d795676\" has successfully progressed."
+ "ReplicaSet \"nginx-deployment-795c7f5bb\" has successfully progressed."
@ ["status","observedGeneration"]
- 2
+ 3

Focus only on spec changes by ignoring metadata and status:

kubectl get deployment example -oyaml | jd -yaml -opts='[
  {"@":["metadata"],"^":["DIFF_OFF"]},
  {"@":["status"],"^":["DIFF_OFF"]}
]' a.yaml

output (spec changes only):

@ ["spec","template","spec","containers",0,"resources","requests","cpu"]
- "100m"
+ "200m"

Allow-list approach - ignore everything except spec:

kubectl get deployment example -oyaml | jd -yaml -opts='[
  {"@":[],"^":["DIFF_OFF"]},
  {"@":["spec"],"^":["DIFF_ON"]}
]' a.yaml

Apply these changes to another deployment:

# Create a focused patch file containing only the CPU change
kubectl get deployment example -oyaml | jd -yaml -opts='[
  {"@":["metadata"],"^":["DIFF_OFF"]},
  {"@":["status"],"^":["DIFF_OFF"]}
]' a.yaml > cpu-patch
kubectl patch deployment example2 --type json --patch "$(jd -t jd2patch cpu-patch)"