import { describe, expect, it } from 'vitest'; import { ParseJsonlStream } from '@apify/utilities'; const OBJS = [{ a: 123 }, { bb: 234 }, { ccc: 345 }, { dddd: 456 }, 5555, 'string', true, false]; describe('parse_jsonl_stream', () => { it('works on various objects', () => { const parseJsonl = new ParseJsonlStream(); parseJsonl.on('error', (err) => { expect(err).toBeInstanceOf(Error); }); let count = 0; parseJsonl.on('object', (obj) => { expect(OBJS[count++]).toEqual(obj); }); parseJsonl.write(''); parseJsonl.write(' \n \t \r '); OBJS.forEach((obj) => { parseJsonl.write(`${JSON.stringify(obj)}\n`); }); parseJsonl.end(''); }); it('handles JSON obj split over multiple lines', () => { const parseJsonl = new ParseJsonlStream(); parseJsonl.on('error', (err) => { expect(err).toBeInstanceOf(Error); }); let count = 0; parseJsonl.on('object', (obj) => { switch (count++) { case 0: expect({ aaa: 123 }).toEqual(obj); break; case 1: expect(true).toEqual(obj); break; case 2: expect(555666).toEqual(obj); break; case 3: expect('string').toEqual(obj); break; default: throw new Error(); } }); parseJsonl.write(' { "aaa" : '); parseJsonl.write(' '); parseJsonl.write(' 123 } '); parseJsonl.write(' '); parseJsonl.write(' \n '); parseJsonl.write(' true\n 555'); parseJsonl.write('6'); parseJsonl.end('66\n "string"'); }); it('handles multiple JSON lines', () => { const parseJsonl = new ParseJsonlStream(); parseJsonl.on('error', (err) => { expect(err).toBeInstanceOf(Error); }); let count = 0; parseJsonl.on('object', (obj) => { expect(OBJS[count++]).toEqual(obj); }); let json = ''; OBJS.forEach((obj) => { json += `${JSON.stringify(obj)}\n`; }); parseJsonl.write(json); }); it('fails on invalid JSON', async () => { const parseJsonl = new ParseJsonlStream(); parseJsonl.on('error', (err) => { expect(err.message.includes('Cannot parse JSON stream data')).toBe(true); }); parseJsonl.write(''); parseJsonl.write('{"invalid: json}\n'); // It seems that the ParseJsonlStream is implemented in a way that emits // events in a sync fashion (i.e. not using nextTick or setImmediate). // This makes all the other tests pass, but not this one, because some Node 14 // internals decided here that the error event should be emitted "later" anyway. // We need wait for next tick because of that. It would be great to rewrite the // stream to handle events properly, but given it's been like this since 2018, // I guess it would bring more trouble than benefit. await new Promise((resolve) => { process.nextTick(resolve); }); }); it('fails on unfinished JSON', async () => { const parseJsonl = new ParseJsonlStream(); let failed = false; parseJsonl.on('error', (err) => { expect(err.message).toMatch('Cannot parse JSON stream data'); failed = true; }); parseJsonl.write('{ "aaa" :'); parseJsonl.end('"aaa'); await new Promise((resolve) => { process.nextTick(resolve); }); expect(failed).toBe(true); }); });