abstract-log
v3.0.4
Published
An abstract interface to build data processing applications on a log-based architecture.
Downloads
10
Readme
abstract-log
An abstract interface to build data processing applications on a log-based architecture.
The idea would be to have implementations that would allow reading/writing to files, databases, kafka, redis, etc.
This is basically abstract-blob-store for append-only logs.
To understand all this append only log stuff and why it's useful to build data processing systems, watch this great talk by Martin Kleppman.
Basically Kafka is all good and well, but very heavy until you need that kind of scale. This project enables you to build on this kind of architecture and then scale up to bigger logging back-ends when you need to.
Usage
// There's a reference implementation of the memory based log included. Should be its own
// module
const memLog = require('abstract-log/mem-log');
// If this were a database log then you would pass through a connection string
let log = memLog();
// append to the log (currently the offset can be an object, not just an integer)
let offset = await log.append({ msg: 'hello world' });
// read from the log
let data = await log.get(offset);
// stream from the log
let offset = 0; // use 0 as initial offset as this is a memory log
log.createReadStream(offset)
.on('data', console.log);
// stream to a log
const arrayToStream = require('array-to-stream');
arrayToStream([0, 1, 2, 3, 4].map((i) => ({ msg: `hello world ${i}` })))
.pipe(log.createWriteStream());
abstract-log Implementations
- mem-log - Reference implementation
- fs-log - A very simple JSON text file log
- knex-log - Uses the
knex
database library to log to postgres, sqlite, mysql, etc
TODO
NB: Please feel free to contribute implementations. PRs welcome!
- levelup - log to leveldb
- kafka - log to kafka
- hyperlog - P2P distribute logs
- S3 - batch writes and log to S3, or even an
abstract-blob-store
- redis - log to redis
- ipfs - log to IPFS using scuttlebut logs for replication
Libraries helpful for implementations
Some logs don't have efficient push mechanisms for changes. So you have to poll for changes. Check out polling-stream which helps you create a perpetual stream of changes.
Docs
WIP. Currently - See tests.