@oresoftware/json-stream-parser

Transform stream

Transforms JSON stream to JS Objects

Installation

$ npm install '@oresoftware/json-stream-parser'

Import

import{JSONParser}from'@oresoftware/json-stream-parser';

Import (CommonJS)

const{JSONParser}=require('@oresoftware/json-stream-parser');

Usage

Right now, the library assumes each separate chunk of json is separated by newline characters.
In the future, we could attempt to use a different delimiting character, as a user-provided input variable.
Recommendations welcome.

Examples

Simple Node.js example:

Reading from stdin

process.stdin.resume().pipe(newJSONParser()).on('data',d=>{// now we got some POJSOs!});

Reading/writing to a tcp socket

import*asnetfrom'net';const[port,host]=[6970,'localhost'];constws=net.createConnection(port,host);ws.setEncoding('utf8').pipe(newJSONParser())// tcp connection is bidirection/full-duplex .. we send JSON strings each way.on('data',onData);// we receive data coming from the tcp server here// and we send data like this:ws.write(JSON.stringify({'some':'data'})+'\n','utf8',cb);// make sure to include the newline char when you write

Using bash shell

Simple bash example:

constk=cp.spawn('bash');k.stdin.end(`echo '{"foo":"bar"}\n'`);// make sure to include the newline char when you writek.stdout.pipe(newJSONParser()).on('data',d=>{// => {foo: 'bar'}});

Bash example with bash variables:

constk=cp.spawn('bash');k.stdin.end(` foo="medicine" cat <<EOF\n{"foo":"$foo"}\nEOF # make sure to include the newline char when you write`);k.stdout.pipe(newJSONParser()).on('data',d=>{assert.deepStrictEqual(d,{foo: 'medicine'});// should pass});

If your JSON has white space (newlines etc)

If you JSON has unescaped newlines, or the JSON is separated by some other character, then use the delimiter option.

newJSONParser({delimiter: '∆∆∆'});// use 3 alt-j's to separate json chunks, since newlines won't work

For other solutions to parsing JSON from CLIs, see: https://stackoverflow.com/questions/56014438/get-single-line-json-from-aws-cli

Other options

1. delayEvery: integer

JSONParser delays the transform callback every x chunks. LiveMutexJSONParser delays every x non-empty JSON records, which is better for long JSONL chunks that should not monopolize the event loop.

import{LiveMutexJSONParser,createLiveMutexJSONParser}from'@oresoftware/json-stream-parser';constparser=newLiveMutexJSONParser({delayEvery: 1024});constparserWithDefaultDelay=createLiveMutexJSONParser();

2. emitNonJSON: boolean

if there is a line of input that cannot be JSON parsed, it will be emitted as "string", but it will not pushed to output

3. there are some secret options in the code, have a look in lib/main.ts