Fastest JSON deep clone in JavaScript

This project is based on fast-json-clone.

fast-clone-json is a tiny library to clone JSON values in pure JavaScript focusing on the speed. According to the benchmark, this package seems the fastest among several deep-clone implementations.

import * as assert from "assert";
import { cloneJson } from "fast-clone-json";

const value = {
  str: "hello",
  num: 42,
  array: [true, null],
  object: { str: "hello", bool: true },
};

const cloned = cloneJson(value);

assert.deepStrictEqual(value, cloned);
assert.notStrictEqual(value, cloned);
assert.notStrictEqual(value.array, cloned.array);
assert.notStrictEqual(value.object, cloned.object);

Installation

This package is licensed with CC0-1.0. Directly copy src/clone.ts to your project.

Or install via npm package manager.

npm install --save fast-clone-json

API

This package exports cloneJson() and cloneJsonObject() functions.

cloneJson(x: JsonValue): JsonValue;

This function clones the given JSON value and returns it. There are some invariants for the parameter:

• The parameter must not contain circles as JSON does not allow it. This function will cause infinite loop for such values by design.
• The parameter must contain JSON values only. Other values like Date, Regexp, Map, Set, Buffer, ... are not allowed.

cloneJsonObject(x: JsonObject): JsonObject;

This function clones the given JSON object and returns it. There are additional invariants for the parameter:

• All invariants of cloneJson().
• Directly specified parameters must be arrays or objects. Primitive values such as string cannot be specified.

Benchmark

Here is the benchmark result with large (1.2MB) JSON value. This package is the fastest among the implementations. Please see the benchmark directory for more details and other results.

Native structuredClone             x 104 ops/sec ±0.40% (76 runs sampled)
structuredClone polyfill           x 105 ops/sec ±0.59% (77 runs sampled)
JSON serialize/deserialize         x 98.93 ops/sec ±0.69% (72 runs sampled)
Naive deep clone                   x 208 ops/sec ±0.18% (88 runs sampled)
lodash.clonedeep                   x 123 ops/sec ±2.42% (80 runs sampled)
clone-deep                         x 194 ops/sec ±0.38% (89 runs sampled)
rfdc (default)                     x 481 ops/sec ±0.19% (93 runs sampled)
rfdc (proto)                       x 551 ops/sec ±0.26% (92 runs sampled)
fastest-json-copy                  x 561 ops/sec ±0.15% (94 runs sampled)
fast-json-clone                    x 584 ops/sec ±0.15% (95 runs sampled)
fast-clone-json (cloneJson)        x 591 ops/sec ±0.24% (95 runs sampled)
fast-clone-json (cloneJsonObject)  x 592 ops/sec ±0.24% (93 runs sampled)

Fastest is fast-clone-json (cloneJsonObject),fast-clone-json (cloneJson)

FAQ

Why is this package the fastest?

Since this package is optimized for removing non-inline function calls in the hot loop as much as possible. rfdc is implemented with the same strategy.

Then why is this package faster than rfdc?

Since this package provides less functionalities. rfdc provides support for some non-JSON types (Date, Regexp, ...). It increases the number of branches in the hot loop so it causes trade-off in performance.

Why are my benchmark results different?

You are right. Improvements in this package are minor. In some cases it may be slower. (eg, differences in CPUs and Javascript engines.)

License

This package is distributed under CC0 1.0 (Public Domain). Feel free to copy and paste the implementation directly to your project without any copyright notice.