jsoncpy

Blazingly fast and fully typed deep copy utility for JSON-compatible data structures.

Features

• High Performance
  Outperforms popular libraries like klona/json, lodash.cloneDeep and even copying using the native JSON.parse or structuredClone method.

• Type-Safe
  Fully typed with TypeScript declarations.

Installation

# npm / node
npm install jsoncpy

# jsr / deno
deno add @observ33r/jsoncpy

# bun
bun install jsoncpy

Usage

import jsoncpy from 'jsoncpy';

export type JSONCpyValue =
  | number
  | string
  | boolean
  | null
  | JSONCpyArray
  | JSONCpyObject;

export type JSONCpyArray = JSONCpyValue[];

export interface JSONCpyObject {
  [key: string]: 
    | JSONCpyValue
    | undefined
}

/**
 *  Recursively creates a deep copy of a JSON-compatible value.
 * 
 *  This function supports only the following value types:
 *  number, string, boolean, null, arrays, and plain objects (no functions, symbols, etc.).
 *
 *  @param source - The input value to deeply copy.
 *  @returns A deep copy of the input value.
 */

export default function jsoncpy(source: JSONCpyValue): JSONCpyValue

Example

import jsoncpy from 'jsoncpy';

const source = JSON.parse('{ "a": 1, "b": ['2', null, false] }');

console.log(jsoncpy(source));

{ a: 1, b: ['2', null, false] }

Benchmark

Big JSON Object (~1.2 MiB, deeply nested)

Library	Time	Relative Speed
jsoncpy	677.72 µs	1.00x (baseline)
klona/json	842.93 µs	1.24x slower
JSON.parse	5.02 ms	7.41x slower
structuredClone	5.18 ms	7.64x slower
lodash.cloneDeep	6.30 ms	9.3x slower

Full benchmark result with hardware counters

clk: ~3.62 GHz
cpu: AMD Ryzen 5 3600 6-Core Processor
runtime: node 24.4.1 (x64-linux)

benchmark                   avg (min … max) p75 / p99    (min … top 1%)
------------------------------------------- -------------------------------
• Big JSON Object (~1.2 MiB, deeply nested)
------------------------------------------- -------------------------------
jsoncpy                      677.72 µs/iter 667.75 µs   █                  
                      (616.47 µs … 1.24 ms)   1.09 ms  ▇█                  
                    (181.60 kb … 802.63 kb) 570.57 kb ▇██▃▂▂▂▂▂▁▂▁▁▁▁▁▁▁▁▁▁
                   2.98 ipc ( 89.95% cache)   8.21k branch misses
          2.61M cycles   7.80M instructions 143.89k c-refs  14.47k c-misses

klona/json                   842.93 µs/iter 819.22 µs  █▄                  
                      (759.86 µs … 1.34 ms)   1.29 ms  ██                  
                    (370.55 kb … 575.42 kb) 570.39 kb ▃██▄▃▂▂▂▂▂▂▁▁▁▁▁▁▁▂▁▁
                   2.79 ipc ( 89.96% cache)   9.98k branch misses
          3.22M cycles   8.98M instructions 146.66k c-refs  14.72k c-misses

lodash.cloneDeep               6.30 ms/iter   6.35 ms  █▅                  
                        (5.95 ms … 8.46 ms)   7.59 ms  ██                  
                    (  2.63 mb …   6.30 mb)   4.96 mb ▆██▄▄▅▂▂▂▂▂▂▂▄▃▁▂▁▁▁▂
                   2.28 ipc ( 95.88% cache)  44.84k branch misses
         23.53M cycles  53.57M instructions   2.46M c-refs 101.33k c-misses

JSON.parse                     5.02 ms/iter   5.11 ms    █                 
                        (4.62 ms … 7.01 ms)   5.97 ms   ▄█▆▃▃              
                    (  1.53 mb …   1.54 mb)   1.53 mb ▂███████▃▅▆▄▃▁▂▃▂▁▁▄▃
                   2.51 ipc ( 88.98% cache)  88.40k branch misses
         17.95M cycles  45.04M instructions 454.77k c-refs  50.12k c-misses

structuredClone                5.18 ms/iter   5.16 ms  █▂                  
                        (5.08 ms … 6.15 ms)   5.72 ms ▃██                  
                    (  1.75 mb …   1.79 mb)   1.75 mb ███▆▄▂▂▃▁▂▂▂▁▁▁▁▂▂▂▁▂
                   2.49 ipc ( 93.15% cache)  68.73k branch misses
         19.95M cycles  49.73M instructions 718.22k c-refs  49.18k c-misses

summary
  jsoncpy
   1.24x faster than klona/json
   7.41x faster than JSON.parse
   7.64x faster than structuredClone
   9.3x faster than lodash.cloneDeep

---

Benchmark uses mitata to test performance with big JSON object to reflect a realistic real-world scenario.

You can run bechmark with:

npm run benchmark

Build

This package uses rollup to generate clean and optimized ESM/CJS builds.

To build package from source code, run:

npm run build

This will generate the output in the dist/ folder. Builds are handled via custom rollup config and exposed under appropriate exports in package.json.

Testing

All tests are written in Vitest with native ESM support and zero transform overhead.

You can run the full suite with:

npm test

Support

If you find this project useful, you can support it with a one-time donation over PayPal. Thank you!

Contributing

Feel free to open issues or submit pull requests on GitHub.

License

This project is licensed under the MIT License.