jsonpointer - an RFC 6901 implementation for Go
Package jsonpointer provides the ability to resolve, assign, and delete values of any type, including raw JSON, by JSON Pointers.
go get github.com/chanced/jsonpointer
package main
import (
"log"
"encoding/json"
"github.com/chanced/jsonpointer"
)
type Nested struct {
Str string
}
type Root struct {
Nested Nested
}
func main() {
r := Root{ Nested: Nested{ Str: "nested str" }}
// jsonpointer.Pointer is a string type so if you have a properly
// formatted json pointer then you can simply convert it:
// ptr := jsonpointer.Pointer(myPointer)
// err := ptr.Validate()
// Note: jsonpointer.New encodes each token's value.
// "/" encodes to "~1" and "~" encodes to "~0" in compliance with RFC 6901.
ptr := jsonpointer.New("nested", "str")
// Resolve
var s string
err := jsonpointer.Resolve(r, ptr, &s)
if err != nil {
log.Fatal(err)
}
log.Println(s) // outputs "nested str"
// Assign
err = jsonpointer.Assign(&r, ptr, "new value")
if err != nil {
log.Fatal(err)
}
log.Println(r.Nested.Str) // outputs "new value"
// Delete
err = jsonpointer.Delete(&r, ptr)
if err != nil {
log.Fatal(err)
}
log.Println(r.Nested.Str) // outputs ""
// jsonpointer can also Resolve, Assign, and Delete JSON in []byte format.
// This includes field values, such as those of type json.RawMessage.
r.Nested.Str = "str val"
b, err := json.Marshal(r)
if err != nil {
log.Fatal(err)
}
err = jsonpointer.Resolve(b, ptr, &s)
if err != nil {
log.Fatal(err)
}
log.Println(s) // outputs "str val"
}
Package jsonpointer provides 3 interfaces: Assigner
, Resolver
, and
Deleter
. Regardless of the operation, if Resolver
is implemented, ResolvePointer
will be
called. ResolvePointer
should not have side effects. If resolving for an assignment, utilize the
pointer to infer which type should be assigned.
AssignByPointer
is invoked on the way back from the leaf. DeleteByPointer
is invoked after resolving the current token.
All three methods are passed a pointer to the jsonpointer.Pointer
so that
it can be modified. If you do not modify it, jsonpointer will assume the current
token was addressed and continue on.
If you wish to only handle some cases with the interfaces, return jsonpointer.YieldOperation
to have the jsonpointer package resolve, assign, or delete as if the type did not implement the interface. Note that doing so results in changes to ptr
being dismissed.
All methods return new values rather than modifying the pointer itself. If you wish to modify the pointer in one of the interface methods, you will need to reassign it: *ptr = newPtrVal
func (mt MyType) ResolvePointer(ptr *jsonpointer.Pointer, op Operation) (interface{}, error) {
next, t, ok := ptr.Next()
if !ok {
// this will only occur if the ptr is a root token in this circumstance
return mt
}
if op == jsonpointer.Assigning && t == "someInterface" {
// maybe you need to know what comes after someInterface to
// determine what implementation of someInterface to assign
t, _ = next.NextToken()
switch t {
case "someIdentifier":
// you could modify ptr if you felt so inclined: *ptr = next
// but it is not needed in this scenario.
return SomeImplementation{}, nil
}
}
// otherwise hand resolution back over to jsonpointer
return nil, jsonpointer.YieldOperation
}
All errors returned from Resolve
, Assign
, and Delete
will implement Error
. A convenience function AsError
exists to help extract out the details.
Depending on the cause, the error could also be KeyError
, IndexError
, FieldError
with additional details. All have corresponding As{Error}
functions.
Finally, all errors have associated Err instances that are wrapped, such as ErrMalformedToken
, ErrInvalidKeyType
, and so on.
See errors.go for further details on errors.
Contributions are always welcome. If you run into an issue, please open a issue on github. If you would like to submit a change, feel free to open up a pull request.
This package is reflect heavy. While it employs the same caching mechanics as
encoding/json
to help alleviate some of the lookup costs, there will always be
a performance hit with reflection.
There are probably plenty of ways to improve performance of the package. Improvements or criticisms are always welcome.
With regards to raw JSON, json.Marshal
and json.Unmarshal
are utilized.
Ideally, in the future, that will change and the package will incoroprate the
encoding/decoding logic from encoding/json
directly, thus skipping the need to
run through unneccesary logic.