hyperdrive

A file sharing network based on rabin file chunking and append only feeds of data verified by merkle trees.

npm install hyperdrive

For a more detailed technical information on how it works see SPECIFICATION.md. It runs in node.js as well as in the browser. Try a browser based demo here

Status

Almost ready for prime time 🚀

APIs/protocols might be still break.

Main things missing are:

• Storing downloaded files as actual files (not in leveldb doh)
• Full documention of the apis/protocols
• Tit-for-tat swarm logic
• pass in a hypercore instead of leveldb
• peer discovery events so a dht would know what to look for (in hypercore)
• Tests for internal logic (in hypercore)
• Move archive/file abstractions to new modules (moved core out to hypercore)
• A bit of refactoring

Usage

First create a new feed to share

var hyperdrive = require('hyperdrive')
var fs = require('fs')
var levelup = require('levelup')

var aLevelDB = levelup('./my-drive')
var drive = hyperdrive(aLevelDB)

var pack = drive.add()

var stream = pack.entry({
  name: 'my-file.txt',
  mode: 0644
})

fs.createReadStream('my-file.txt').pipe(stream)

pack.finalize(function () {
  var link = pack.id.toString('hex')
  console.log(link, '<-- this is your hyperdrive link')
})

Then to share it

var disc = require('discovery-channel')()
var hyperdrive = require('hyperdrive')
var net = require('net');
var levelup = require('levelup')
var aLevelDB = levelup('./mydb')
var drive = hyperdrive(aLevelDB)

var link = new Buffer({your-hyperdrive-link-from-the-above-example}, 'hex')

var server = net.createServer(function (socket) {
  socket.pipe(drive.createPeerStream()).pipe(socket)
})

server.listen(0, function () {
  function ann () {
    // discovery-channel currently only works with 20 bytes hashes
    disc.announce(link.slice(0, 20), server.address().port)
  }

  ann()
  setInterval(ann, 10000)

  var lookup = disc.lookup(link.slice(0, 20))

  lookup.on('peer', function (ip, port) {
    var socket = net.connect(port, ip)
    socket.pipe(drive.createPeerStream()).pipe(socket)
  })
})

If you run this code on multiple computers you should be able to access the content in the feed by doing

var feed = drive.get(link) // the link identifies/verifies the content

feed.get(0, function (err, entry) { // get the first entry
  console.log(entry) // prints {name: 'my-file.txt', ...}
  var content = drive.get(entry)
  content.get(0, function (err, data) {
    console.log('first block of the file', data)
    content.get(1, ...)
  })
})

API

Stability: UNSTABLE, likely to have major changes

var drive = hyperdrive(db)

Create a new hyperdrive instance. db should be a levelup instance.

var stream = drive.createPeerStream()

Create a new peer replication duplex stream. This stream should be piped together with another peer stream somewhere else to start replicating the feeds

var archive = drive.add()

Add a new archive to share.

var stream = archive.entry(fileInfo, [cb])

Add a new file entry to the file archive. fileInfo should look like this

{
  name: 'dir/filename', // required
  type: 'file or directory', // detected using the mode if not provided
  mode: 0666, // optional
  uid: 0, // optional
  gid: 0, // optional
  mtime: mtimeInSeconds // optional
  ctime: ctimeInSeconds // optional
}

The stream returned is a writable stream. You should write the file contents to that. If you are writing a directory the stream will be null.

Optionally you can provide a callback that is called when the content has been written

archive.finalize([cb])

Finalize the archive. After the callback has been called you can get the feed id by accessing archive.id.

var feed = drive.get(id_or_entry)

Access a feed by it's id or entry object. The entry object looks like this

{
  type: 'metadata or file',
  value: optionalValueCorrespondingToTheType,
  link: {
    id: feedId,
    blocks: blocksInFeed
  }
}

If you just pass in a feedId the type will default to metadata.

feed.get(index, callback)

Get a block from the the feed.

• If the feed is an metadata feed the return value will be an entry object.
• If the feed is a file feed it will be a buffer.

var stream = feed.createStream()

Create a readable stream of all entries in the feed. The values will be of the same type as described in feed.get

var cursor = feed.cursor()

If the feed is a file feed you can create a random access cursor by calling var cursor = feed.cursor().

cursor.read(byteOffset, callback)

Will return a buffer stored at that byte offset in the file.

cursor.next(callback)

Will return the next buffer at the current cursor position.

License

MIT