jsonstreamkit

Zero-dependency TypeScript streaming JSON parser and stringifier. Parse multi-gigabyte JSON files without loading them into memory.

Drop-in replacement for the abandoned JSONStream npm package (12.6M/week, archived 2018). Zero runtime dependencies, full TypeScript types, ESM + CJS.

Install

npm install jsonstreamkit

Quick start

import { createReadStream } from "node:fs";
import { parse } from "jsonstreamkit";

// Parse a large JSON array and process each element without loading the whole file
createReadStream("huge-data.json")
  .pipe(parse("rows.*"))
  .on("data", (row) => {
    console.log(row); // each row object, emitted as it arrives
  });

API

parse(path?)

Creates a Transform stream that:

1. Receives JSON as Buffer/string chunks
2. Emits matched values as JavaScript objects (no serialization overhead)

import { parse } from "jsonstreamkit";

// Match all elements of a top-level array
const stream = parse("*");

// Match nested array elements
const stream = parse("data.rows.*");

// Match a specific field
const stream = parse("metadata.title");

// Emit the whole document (path = "")
const stream = parse("");

// Array-form path (same as "data.rows.*"):
const stream = parse(["data", "rows", "*"]);

Path syntax:

| Path | Matches | |---|---| | "*" | Each element of the root array/object | | "rows.*" | Each element of json.rows | | "a.b.c" | The value at json.a.b.c | | "items.0" | json.items[0] | | "" | The entire document (emit once) |

stringify(open?, sep?, close?)

Creates a Transform stream that serializes objects into a JSON array string.

import { stringify } from "jsonstreamkit";

const stream = stringify();          // default: "[\n", "\n,\n", "\n]\n"
const stream = stringify("[", ",", "]");  // compact

stringifyLines()

Emits NDJSON / JSON Lines — one JSON object per line.

import { stringifyLines } from "jsonstreamkit";

const stream = stringifyLines();
stream.write({ id: 1 });  // → {"id":1}
stream.write({ id: 2 });  // → {"id":2}

Convenience helpers

import { parseString, stringifyArray } from "jsonstreamkit";

// Parse a complete JSON string (non-streaming)
const results = await parseString('[1,2,3]', "*");  // [1, 2, 3]
const results = await parseString(bigJson, "rows.*"); // array of rows

// Stringify an array of values (non-streaming)
const json = await stringifyArray([1, 2, 3], "[", ",", "]"); // "[1,2,3]"

Examples

Stream-process a large JSON file

import { createReadStream } from "node:fs";
import { createGunzip } from "node:zlib";
import { parse } from "jsonstreamkit";

let count = 0;
createReadStream("data.json.gz")
  .pipe(createGunzip())
  .pipe(parse("records.*"))
  .on("data", (record: Record<string, unknown>) => {
    count++;
    // process each record without memory pressure
  })
  .on("end", () => console.log(`Processed ${count} records`));

Transform and re-emit as NDJSON

import { createReadStream, createWriteStream } from "node:fs";
import { parse, stringifyLines } from "jsonstreamkit";

createReadStream("input.json")
  .pipe(parse("items.*"))
  .pipe(stringifyLines())
  .pipe(createWriteStream("output.ndjson"));

Stringify an async generator

import { Readable } from "node:stream";
import { stringify } from "jsonstreamkit";

async function* generateRecords() {
  for await (const row of db.cursor("SELECT * FROM events")) {
    yield row;
  }
}

Readable.from(generateRecords())
  .pipe(stringify())
  .pipe(response); // pipe to HTTP response

Drop-in for JSONStream

// Before:
const JSONStream = require("JSONStream");
const parser = JSONStream.parse("rows.*");

// After:
import { parse } from "jsonstreamkit";
const parser = parse("rows.*");

Known limitations

• Streaming null as an individual array element is not supported (Node.js object-mode streams use push(null) to signal end-of-stream). Use parseString() if your data contains top-level null items.

Contributors ✨

This project follows the all-contributors specification. Contributions of any kind are welcome — code, docs, bug reports, ideas, reviews! See the emoji key for how each contribution is recognized, and open a PR or issue to get involved.

Thanks goes to these wonderful people:

License

MIT