README
¶
jsondiff
jsondiff is a Go package for computing the diff between two JSON documents as a series of RFC6902 (JSON Patch) operations, which is particularly suitable to create the patch response of a Kubernetes Mutating Webhook for example.
Usage
First, get the latest version of the library using the following command:
$ go get github.com/wI2L/jsondiff@latest
⚠ Requires Go1.14+, due to the usage of the package hash/maphash.
Example use cases
Kubernetes Dynamic Admission Controller
The typical use case within an application is to compare two values of the same type that represents the source and desired target of a JSON document. A concrete application of that would be to generate the patch returned by a Kubernetes dynamic admission controller to mutate a resource. Thereby, instead of generating the operations, just copy the source in order to apply the required changes and delegate the patch generation to the library.
For example, given the following corev1.Pod value that represents a Kubernetes demo pod containing a single container:
import corev1 "k8s.io/api/core/v1"
pod := corev1.Pod{
Spec: corev1.PodSpec{
Containers: []corev1.Container{{
Name: "webserver",
Image: "nginx:latest",
VolumeMounts: []corev1.VolumeMount{{
Name: "shared-data",
MountPath: "/usr/share/nginx/html",
}},
}},
Volumes: []corev1.Volume{{
Name: "shared-data",
VolumeSource: corev1.VolumeSource{
EmptyDir: &corev1.EmptyDirVolumeSource{
Medium: corev1.StorageMediumMemory,
},
},
}},
},
}
The first step is to copy the original pod value. The corev1.Pod type defines a DeepCopy method, which is handy, but for other types, a shallow copy is discouraged, instead use a specific library, such as ulule/deepcopier. Alternatively, if you don't require to keep the original value, you can marshal it to JSON using json.Marshal to store a pre-encoded copy of the document, and mutate the value.
newPod := pod.DeepCopy()
// or
podBytes, err := json.Marshal(pod)
if err != nil {
// handle error
}
Secondly, make some changes to the pod spec. Here we modify the image and the storage medium used by the pod's volume shared-data.
// Update the image of the webserver container.
newPod.Spec.Containers[0].Image = "nginx:1.19.5-alpine"
// Switch storage medium from memory to default.
newPod.Spec.Volumes[0].EmptyDir.Medium = corev1.StorageMediumDefault
Finally, generate the patch that represents the changes relative to the original value. Note that when the Compare or CompareOpts functions are used, the source and target parameters are first marshaled using the encoding/json package in order to obtain their final JSON representation, prior to comparing them.
import "github.com/wI2L/jsondiff"
patch, err := jsondiff.Compare(pod, newPod)
if err != nil {
// handle error
}
b, err := json.MarshalIndent(patch, "", " ")
if err != nil {
// handle error
}
os.Stdout.Write(b)
The output is similar to the following:
[{
"op": "replace",
"path": "/spec/containers/0/image",
"value": "nginx:1.19.5-alpine"
}, {
"op": "remove",
"path": "/spec/volumes/0/emptyDir/medium"
}]
The JSON patch can then be used in the response payload of you Kubernetes webhook.
Diff options
If more control over the diff behaviour is required, use the CompareOpts or CompareJSONOpts function instead. The third parameter is variadic and accept a list of functional opt-in options described below.
Note that any combination of options can be used without issues.
Operations factorization
By default, when computing the difference between two JSON documents, the package does not produce move or copy operations. To enable the factorization of value removals and additions as moves and copies, you should use the functional option Factorize(). Factorization reduces the number of operations generated, which inevitably reduce the size of the patch once it is marshaled as JSON.
For instance, given the following document:
{
"a": [1, 2, 3],
"b": { "foo": "bar" }
}
In order to obtain this updated version:
{
"a": [1, 2, 3],
"c": [1, 2, 3],
"d": { "foo": "bar" }
}
The package generates the following patch:
[
{ "op": "remove", "path": "/b" },
{ "op": "add", "path": "/c", "value": [1, 2, 3] },
{ "op": "add", "path": "/d", "value": { "foo": "bar" } }
]
If we take the previous example and generate the patch with factorization enabled, we then get a different patch, containing copy and move operations instead:
[
{ "op": "copy", "from": "/a", "path": "/c" },
{ "op": "move", "from": "/b", "path": "/d" }
]
Operations rationalization
The default method used to compare two JSON documents is a recursive comparison. This produce one or more operations for each difference found. On the other hand, in certain situations, it might be beneficial to replace a set of operations representing several changes inside a JSON node by a single replace operation targeting the parent node.
For that purpose, you can use the Rationalize() option. It uses a simple weight function to decide which patch is best (it marshals both sets of operations to JSON and looks at the length of bytes to keep the smaller footprint).
Let's illustrate that with the following document:
{
"a": { "b": { "c": { "1": 1, "2": 2, "3": 3 } } }
}
In order to obtain this updated version:
{
"a": { "b": { "c": { "x": 1, "y": 2, "z": 3 } } }
}
The expected output is one remove/add operation combo for each children field of the object located at path a.b.c:
[
{ "op": "remove", "path": "/a/b/c/1" },
{ "op": "remove", "path": "/a/b/c/2" },
{ "op": "remove", "path": "/a/b/c/3" },
{ "op": "add", "path": "/a/b/c/x", "value": 1 },
{ "op": "add", "path": "/a/b/c/y", "value": 2 },
{ "op": "add", "path": "/a/b/c/z", "value": 3 }
]
If we also enable factorization, as seen above, we can reduce the number of operations by half:
[
{ "op": "move", "from": "/a/b/c/1", "path": "/a/b/c/x" },
{ "op": "move", "from": "/a/b/c/2", "path": "/a/b/c/y" },
{ "op": "move", "from": "/a/b/c/3", "path": "/a/b/c/z" }
]
And finally, with rationalization enabled, those operations are replaced with a single replace of the parent object:
[
{ "op": "replace", "path": "/a/b/c", "value": { "x": 1, "y": 2, "z": 3 } }
]
Invertible patch
Using the functional option Invertible(), it is possible to instruct the diff generator to precede each remove and replace operation with a test operation. Such patches can be inverted to return a patched document to its original form.
However, note that it comes with one limitation. copy operations cannot be inverted, as they are ambiguous (the reverse of a copy is a remove, which could then become either an add or a copy). As such, using this option disable the generation of copy operations (if option Factorize() is used) and replace them with add, albeit potentially at the cost of increased patch size.
For example, let's generate the diff between those two JSON documents:
{
"a": "1",
"b": "2"
}
{
"a": "3",
"c": "4"
}
The patch is similar to the following:
[
{ "op": "test", "path": "/a", "value": "1" },
{ "op": "replace", "path": "/a", "value": "3" },
{ "op": "test", "path": "/b", "value": "2" },
{ "op": "remove", "path": "/b" },
{ "op": "add", "path": "/c", "value": "4" }
]
As you can see, the remove and replace operations are preceded with a test operation which assert/verify the value of the previous path. On the other hand, the add operation can be reverted to a remove operation directly and doesn't need to be preceded by a test.
Finally, as a side example, if we were to use the Rationalize() option in the context of the previous example, the output would be shorter, but the generated patch would still remain invertible:
[
{ "op": "test", "path": "", "value": { "a": "1", "b": "2" } },
{ "op": "replace", "path": "", "value": { "a": "3", "c": "4" } }
]
Benchmarks
Performance is not the primary target of the package, instead it strives for correctness. A simple benchmark that compare the performance of available options is provided to give a rough estimate of the cost of each option. You can find the JSON documents used by this benchmark in the directory testdata/bench.
If you'd like to run the benchmark yourself, use the following command:
go get github.com/cespare/prettybench
go test -bench=. | prettybench
Results
The benchmark was run 10x (statistics computed with benchstat) on a MacBook Pro 13", with the following specs:
OS : macOS Catalina (10.15.7)
CPU: 1.4 GHz Intel Core i5
Mem: 16GB 2133 MHz
Go : go version go1.15.5 darwin/amd64
Tag: v0.1.0
Output
name time/op CompareJSONOpts/default-8 26.5µs ± 1% CompareJSONOpts/invertible-8 27.4µs ± 1% CompareJSONOpts/factorize-8 31.6µs ± 1% CompareJSONOpts/rationalize-8 90.3µs ± 0% CompareJSONOpts/factor+ratio-8 95.5µs ± 1% CompareJSONOpts/all-options-8 126µs ± 1%
Credits
This package has been inspired by existing implementations of JSON Patch for various languages:
- cujojs/jiff
- Starcounter-Jack/JSON-Patch
- java-json-tools/json-patch
- Lattyware/elm-json-diff
- espadrine/json-diff
License
jsondiff is licensed under the MIT license. See the LICENSE file.
Documentation
¶
Index ¶
Examples ¶
Constants ¶
const ( OperationAdd = "add" OperationReplace = "replace" OperationRemove = "remove" OperationMove = "move" OperationCopy = "copy" OperationTest = "test" )
Variables ¶
This section is empty.
Functions ¶
This section is empty.
Types ¶
type Operation ¶
type Operation struct {
Type string `json:"op"`
From pointer `json:"from,omitempty"`
Path pointer `json:"path"`
OldValue interface{} `json:"-"`
Value interface{} `json:"value,omitempty"`
}
Operation represents a RFC6902 JSON Patch operation.
func (Operation) MarshalJSON ¶
MarshalJSON implements the json.Marshaler interface.
type Option ¶
type Option func(*differ)
An Option overrides the default diff behavior of the CompareOpts and CompareJSONOpts function.
func Factorize ¶
func Factorize() Option
Factorize enables factorization of operations.
Example ¶
package main
import (
"fmt"
"log"
"github.com/wI2L/jsondiff"
)
func main() {
source := `{"a":[1,2,3],"b":{"foo":"bar"}}`
target := `{"a":[1,2,3],"c":[1,2,3],"d":{"foo":"bar"}}`
patch, err := jsondiff.CompareJSONOpts(
[]byte(source),
[]byte(target),
jsondiff.Factorize(),
)
if err != nil {
log.Fatal(err)
}
for _, op := range patch {
fmt.Printf("%s\n", op)
}
}
Output: {"op":"copy","from":"/a","path":"/c"} {"op":"move","from":"/b","path":"/d"}
func Invertible ¶
func Invertible() Option
Invertible enables the generation of an invertible patch, by preceding each remove and replace operation by a test operation that verifies the value at the path that is being removed/replaced. Note that copy operations are not invertible, and as such, using this option disable the usage of copy operation in favor of add operations.
Example ¶
package main
import (
"fmt"
"log"
"github.com/wI2L/jsondiff"
)
func main() {
source := `{"a":"1","b":"2"}`
target := `{"a":"3","c":"4"}`
patch, err := jsondiff.CompareJSONOpts(
[]byte(source),
[]byte(target),
jsondiff.Invertible(),
)
if err != nil {
log.Fatal(err)
}
for _, op := range patch {
fmt.Printf("%s\n", op)
}
}
Output: {"op":"test","path":"/a","value":"1"} {"op":"replace","path":"/a","value":"3"} {"op":"test","path":"/b","value":"2"} {"op":"remove","path":"/b"} {"op":"add","path":"/c","value":"4"}
type Patch ¶
type Patch []Operation
Patch represents a series of JSON Patch operations.
func Compare ¶
Compare compares the JSON representations of the given values and returns the differences relative to the former as a list of JSON Patch operations.
Example ¶
package main
import (
"fmt"
"log"
"github.com/wI2L/jsondiff"
corev1 "k8s.io/api/core/v1"
)
func main() {
pod := corev1.Pod{
Spec: corev1.PodSpec{
Containers: []corev1.Container{{
Name: "webserver",
Image: "nginx:latest",
VolumeMounts: []corev1.VolumeMount{{
Name: "shared-data",
MountPath: "/usr/share/nginx/html",
}},
}},
Volumes: []corev1.Volume{{
Name: "shared-data",
VolumeSource: corev1.VolumeSource{
EmptyDir: &corev1.EmptyDirVolumeSource{
Medium: corev1.StorageMediumMemory,
},
},
}},
},
}
newPod := pod.DeepCopy()
newPod.Spec.Containers[0].Image = "nginx:1.19.5-alpine"
newPod.Spec.Volumes[0].EmptyDir.Medium = corev1.StorageMediumDefault
patch, err := jsondiff.Compare(pod, newPod)
if err != nil {
log.Fatal(err)
}
for _, op := range patch {
fmt.Printf("%s\n", op)
}
}
Output: {"op":"replace","path":"/spec/containers/0/image","value":"nginx:1.19.5-alpine"} {"op":"remove","path":"/spec/volumes/0/emptyDir/medium"}
func CompareJSON ¶
CompareJSON compares the given JSON documents and returns the differences relative to the former as a list of JSON Patch operations.
Example ¶
package main
import (
"encoding/json"
"fmt"
"io/ioutil"
"log"
"github.com/wI2L/jsondiff"
)
func main() {
type Phone struct {
Type string `json:"type"`
Number string `json:"number"`
}
type Person struct {
Firstname string `json:"firstName"`
Lastname string `json:"lastName"`
Gender string `json:"gender"`
Age int `json:"age"`
Phones []Phone `json:"phoneNumbers"`
}
source, err := ioutil.ReadFile("testdata/examples/john.json")
if err != nil {
log.Fatal(err)
}
var john Person
if err := json.Unmarshal(source, &john); err != nil {
log.Fatal(err)
}
john.Age = 30
john.Phones = append(john.Phones, Phone{
Type: "mobile",
Number: "209-212-0015",
})
target, err := json.Marshal(john)
if err != nil {
log.Fatal(err)
}
patch, err := jsondiff.CompareJSON(source, target)
if err != nil {
log.Fatal(err)
}
for _, op := range patch {
fmt.Printf("%s\n", op)
}
}
Output: {"op":"replace","path":"/age","value":30} {"op":"add","path":"/phoneNumbers/-","value":{"number":"209-212-0015","type":"mobile"}}
func CompareJSONOpts ¶
CompareJSONOpts is similar to CompareJSON, but also accepts a list of options to configure the diff behavior.
func CompareOpts ¶
CompareOpts is similar to Compare, but also accepts a list of options to configure the diff behavior.
Source Files