A Software Development Kit for the hypercore-protocol
Formerly known as "dat-sdk".
Hypercore-protocol and it's ecosystem consists of a bunch of low level building blocks for working with data in distributed applications. Although this modularity makes it easy to mix and match pieces, it adds complexity when it comes to actually building something.
The Hyper SDK combines the lower level pieces of the Hyper stack into high level APIs that you can use across platforms so that you can focus on your application rather than the gritty details of how it works.
The Hyper SDK can either work "natively", which means the full storage and networking stack runs directly within the SDK. Alternatively, it supports the experimental hyperspace daemon. In this mode, the SDK needs a hyperspace daemon running and will connect to it as a client.
- High level API
- Cross-platform with same codebase
- ✔ Node
- ✔ Web
- ✔ Electron
- React-Native (with nodejs-mobile-react-native?)
Node.js / Browserify workflows:
npm install --save hyper-sdk
const SDK = require('hyper-sdk')
The easiest way to get started is to run the build
command in this sdk
, then copy the bundle.js
into your own project. Here's how:
git clone git@github.com:datproject/sdk.git
cd sdk
# Compile the SDK into a single JS file
npm run build
# Copy `hyper-sdk-bundle.js` into your project
<script src="hyper-sdk-bundle.js"></script>
<script>
const SDK = window.hyperSDK
// Look at the examples from here
</script>
If the bundle above doesn't work for your setup, and you want to DIY in your own project, you'll need to mimic how the SDK generates the bundle, using:
- Browserify
- Babelify (babel for browserify)
- babel.config.json file
Combine Browserify with Babel (via Babelify) to make this work in the browser:
Dev Dependencies (must be a DevDependency):
npm install --save-dev browserify babelify util
and the regular dependencies
npm install --save hyper-sdk@latest @geut/sodium-javascript-plus hyperswarm-web
Add this as the build
command in your package.json
. It is important to add the transform (-t
) with babelify
to make it work. Babel will use the aliases in the babel.config.json file to change the code from nodejs to browser.
"build": "browserify -t [ babelify --global ] index.js > bundle.js"
Once you npm run build
then you can use the generated bundle.js
in your project!
These instructions are for Webpack 5.
Webpack 5 no longer polyfills Node.js core modules automatically,
so you need to polyfill quite a few of them to get it working in the browser.
I managed to get it working with the following config file.
This config file contains a few comments to better explain what happens.
If you just want to get it working quickly,
simply put this comment-less version in your .webpack.config.js
:
const path = require('path')
var webpack = require('webpack');
module.exports = {
'mode': 'development',
entry: './index.js',
target: 'web',
resolve: {
alias: {
hyperswarm: 'hyperswarm-web',
},
fallback: {
crypto: require.resolve("crypto-browserify"),
path: require.resolve("path-browserify"),
os: require.resolve('os-browserify/browser'),
stream: require.resolve('stream-browserify'),
constants: require.resolve('constants-browserify'),
assert: require.resolve("assert/"), // Note the trailing slash
buffer: require.resolve('buffer/'), // Note the trailing slash
process: 'process/browser',
fs: false
},
},
plugins: [
new webpack.ProvidePlugin({
Buffer: ['buffer', 'Buffer'],
process: 'process/browser',
}),
],
output: {
filename: 'bundle.js',
path: path.resolve(__dirname, 'dist')
},
}
First install all dependencies:
npm install webpack webpack-cli hyperswarm-web crypto-browserify path-browserify os-browserify stream-browserify constants-browserify buffer process assert```
Then create the bundle:
webpack
Then you can include `./dist/bundle.js` in your HTML page.
Note that you might need to add additional aliases or fallbacks
if they are used within your code.
See [webpack docs](https://webpack.js.org/configuration/resolve/#resolvefallback)
for the full list
of core Node.js modules which require a polyfill when used.
## API/Examples
```js
const SDK = require('hyper-sdk')
const sdk = await SDK({
// With this, all drive will disappear after the process exits
// This is here so that running the example doesn't clog up your history
persist: false,
// storage can be set to an instance of `random-access-*`
// const RAI = require('random-access-idb')
// otherwise it defaults to `random-access-web` in the browser
// and `random-access-file` in node
storage: null //storage: RAI
});
const {
Hypercore, // Create a new Hypercore
Hyperdrive, // Create a new Hyperdrive
resolveName, // Resolve hyper:// address to key using hyper-dns
close // Cleanup all hyper related resources
} = sdk
// Create a new Hyperdrive.
// If you want to create a new drive, pass in a name for it
// This will be used to derive a secret key
// Every time you open a drive with that name it'll derive the same key
// This uses a master key that's generated once per device
// That means the same name will yield a different key on a different machine
const drive = Hyperdrive('My drive name')
// You should wait for the drive to be totally initialized
await drive.ready()
const url = `hyper://${drive.key.toString('hex')}`
// TODO: Save this for later!
console.log(`Here's your URL: ${url}`)
// Check out the hyperdrive docs for what you can do with it
// https://www.npmjs.com/package/hyperdrive#api
await drive.writeFile('/example.txt', 'Hello World!')
console.log('Written example file!')
// Resolve a hyper:// address to a key
const key = await resolveName('hyper://blog.mauve.moe')
const drive = Hyperdrive(key)
// Pre-download the drive
await drive.download()
// Delete all the data
await drive.destroyStorage()
const SOME_URL = 'dat://0a9e202b8055721bd2bc93b3c9bbc03efdbda9cfee91f01a123fdeaadeba303e/'
const somedrive = Hyperdrive(SOME_URL)
// Download '/' from hyperdrive
console.log(await somedrive.readdir('/'))
// Create a hypercore
// Check out the hypercore docs for what you can do with it
// https://github.com/mafintosh/hypercore
// If you're using TypeScript, make sure to appropriately type the generic Hypercore
// e.g. Hypercore<string>(...) for a JSON encoded Hypercore
const myCore = Hypercore('my hypercore name', {
valueEncoding: 'json',
persist: false,
// storage can be set to an instance of `random-access-*`
// const RAI = require('random-access-idb')
// otherwise it defaults to `random-access-web` in the browser
// and `random-access-file` in node
storage: null // storage: RAI
})
// Add some data to it
await myCore.append(JSON.stringify({
name: 'Alice'
}))
// Use extension messages for sending extra data over the p2p connection
const discoveryCoreKey = 'dat://bee80ff3a4ee5e727dc44197cb9d25bf8f19d50b0f3ad2984cfe5b7d14e75de7'
const discoveryCore = Hypercore(discoveryCoreKey)
// Register the extension message handler
const extension = discoveryCore.registerExtension('discovery', {
// Set the encoding type for messages
encoding: 'binary',
onmessage: (message, peer) => {
// Recieved messages will be automatically decoded
console.log('Got key from peer!', message)
const otherCore = Hypercore(message, {
valueEncoding: 'json',
persist: false
})
// Render the peer's data from their core
otherCore.get(0, console.log)
}
})
// When you find a peer tell them about your core
discoveryCore.on('peer-add', (peer) => {
console.log('Got a peer!')
extension.send(myCore.key, peer)
})
const hypertrie = require('hypertrie')
// Pass in hypercores from the SDK into other data structures
// Check out what you can do with hypertrie from there:
// https://github.com/mafintosh/hypertrie
const trie = hypertrie(null, {
feed: Hypercore('my trie core', {
persist: false
})
})
trie.put('key', 'value', () => {
trie.get('key', (err, node) => {
console.log('Got key: ', node.key)
console.log('Loaded value from trie: ', node.value)
})
})
All available tests are run three times: For the native backend, for the hyperspace backend, and with one native and one hyperspace backend.
To run tests in Node.js simply run npm run test
in a checkout.
To run the tests in a browser, first run npm run build-test
to build the test bundle. Then, run npm run test-proxy
to run both a hyperswarm-web proxy and two Hyperspace servers that listen for clients on a websocket. Finally, open test.html
in a web browser and open the developer tools, where you should see the test results in the console.
The API supports both promises and callbacks. Everywhere where you see await
, you can instead pass a node-style callback.
Import the SDK contructor using the native backend. This means that the full storage and networking stack runs right within the current process. Works both in Node.js and in web browsers.
When running in a web browser, it needs a hyperswarm-web proxy for peer to peer connectivity.
Import the SDK contructor using the experimental hyperspace backend. Here, the SDK needs a running hyperspace server.
When running in NodeJS, this will attempt to connect to a hyperspace server running on the same machine. When running in a web browser, this will attempt to connect to a Hyperspace server over Websockets (experimental).
TODO: Document how to run a Websocket Hyperspace server.
NOTE: The hyperspace backend does not yet support the deriveSecret
function (will throw an exception if used).
const {Hypercore, Hyperdrive, resolveName, keyPair, deriveSecret, registerExtension, close} = await SDK(opts?)
Creates an instance of the Hyper SDK based on the options.
Options for the native backend:
opts.applicationName
: An optional name for the application using the SDK. This will automatically silo your data from other applications using the SDK and will store it in the appropriate place using random-access-applicationopts.persist: true
: An optional arg for whether data should be persisted. Set this tofalse
if you want stuff stored in memory. Ignored if you pass in a custom storage or corestore.opts.storage
: An optional random-access-storage instance for storing dataopts.swarmOpts
: This lets you configure hyperswarm and hyperswarm-webmaxPeers
: The maximum number of connections to keep for this swarm.ephemeral **Node**
: Set tofalse
if this is going to be in a long running process on a server.bootstap **Node**
: An array of addresses to use for the DHT bootstraping. Defaults to['bootstrap1.hyperdht.org:49737', 'bootstrap2.hyperdht.org:49737', 'bootstrap3.hyperdht.org:49737']
preferredPort: 42666 **Node**
: The port hyperswarm should try to bind on. You should allow it through your firewall on TCP/UDP for best results.webrtcBootstrap **Browser**: ['https://geut-webrtc-signal.herokuapp.com/'] **BROWSER**
: The WebRTC bootstrap server list used by discovery-swarm-webrtcwsProxy **Browser**: 'wss://hyperswarm.mauve.moe' **BROWSER**
: The Websocket proxy used for hyperswarm-proxy-ws
Options for the hyperspace backend:
opts.hyperspaceOpts
Options to initialize the connection to a hyperspace server.client
: An optional @hyperspace/client instance. If not set a client will be created automatically.protocol
: The protocol to use. Defaults tows
in browsers anduds
in Node.js.port
: If using thews
protocol: Port of the Websocket to connect to (default9000
)host
: Forws
protocol: Hostname of the Websocket to connect to (defaultlocalhost
). Foruds
protocol: Name of the socket (defaulthyperspace
).
Options for all backends:
opts.corestore
: An optional Corestore instance for using as hypercore storage.opts.corestoreOpts
: Options to pass into Corestore when it's initialized.masterKey
: Optional 32 byte Buffer with the master key that should be used to derive sercret keys for hypercores. Useful to restore from backupsack
: Whether you want there to be apeer-ack
event emitted when data has been uploaded to a peer.
opts.coreOpts
: This lets you configure the behavior of Hypercore instancessparse: true
: Whether the history should be loaded on the fly instead of replicating the full historyvalueEncoding: 'json' | 'utf-8' | 'binary'
: The encoding to use for the data stored in the hypercore. Use JSON to store / retrieve objects.
opts.driveOpts
: This lets you configure the behavior of Hyperdrive instancessparse: true
: Whether the history should be loaded on the fly instead of replicating the full history
opts.dnsOpts
: Configure the dat dns resolution module. You probably shouldn't mess with this.recordName: 'dat'
: name of .well-known fileprotocolRegex: /^dat:\/\/([0-9a-f]{64})/i
: RegExp object for custom protocolhashRegex: /^[0-9a-f]{64}?$/i
: RegExp object for custom hash i.e.txtRegex: /"?datkey=([0-9a-f]{64})"?/i
: RegExp object for DNS TXT record of custom protocol
This closes all resources used by the SDK so you can safely end your process. cb
will be invoked once resources are closed or if there's an error.
Resolve a DNS name to a Hypercore key.
url
is ahyper://
URL likehyper://blog.mauve.moe
key
will be the key that you can pass toHyperdrive
This gives you the public / private keypair used for the Noise protocol encryption when connecting to peers.
You can use this to identify peers in the network using peer.remotePublicKey
Derives a secret key based on the SDK's master key.
namespace
can be used to namespace different applications, and name
is the name of the key you want.
This can be used as a seed for generating secure private keys without needing to store an extra key on disk.
This function is currently only supported in the native backend.
Listens on extension messages of type name
on the feeds replication channels.
handlers.encoding
: The encoding to use for messages.json
,binary
, 'utf8'handlers.onmessage(message, peer)
: Function to invoke when a peer sends you a message for this extension type.handlers.onerror(err, peer)
: Function to invoke when a peer has sent you a mis-coded message on this extension.
You can respond to messages with extension.send(message, peer)
.
You can also broadcast out messages to all peers with extension.broadcast(message)
This initializes a Hyperdrive, the SDK will begin finding peers for it and will de-duplicate calls to initializing the same drive more than once.
keyOrName
: This must be provided. It's either ahyper://
URL / key or a string identifying the name. If you want to have a writable drive, you can use the name to generate one and use the name later to get the same drive back without having to save the key somewhere.opts
: These are the options for configuring the hyperdrive.sparse: true
: Whether the history should be loaded on the fly instead of replicating the full historysecretKey
: A secret key for granting write access. This can be useful when restoring backups.discoveryKey
: Optionally specify which discovery key you'd like to use for finding peers for this drive.lookup: true
: Specify whether you wish to lookup peers for this drive. Setfalse
along withannounce
to avoid advertisingannounce: true
: Specify whether you wish to advertise yourself as having the drive.
The rest of the Hyperdrive docs were taken from the Hyperdrive README. Note that we're wrapping over the APIs with Hyperdrive-Promise so any callback methods can be await
ed instead.
Get the current version of the drive (incrementing number).
The public key identifying the drive.
A key derived from the public key that can be used to discovery other peers sharing this drive.
A boolean indicating whether the drive is writable.
Emitted when the drive is fully ready and all properties has been populated.
Emitted when the drive has got a new change.
Emitted when a critical error during load happened.
Emitted when the drive has been closed
Emitted when a new peer has started replicating wiht the drive.
Emitted when a peer has stopped replicating wit the drive.
Checkout a readonly copy of the drive at an old version. Options are used to configure the oldDrive
:
{
metadataStorageCacheSize: 65536 // how many entries to use in the metadata hypercore's LRU cache
contentStorageCacheSize: 65536 // how many entries to use in the content hypercore's LRU cache
treeCacheSize: 65536 // how many entries to use in the append-tree's LRU cache
}
Download all files in path of current version. If no path is specified this will download all files.
You can use this with .checkout(version)
to download a specific version of the drive.
drive.checkout(version).download()
Clear the storage of all files in the path.
This is the opposite of the download
API.
Note that this doesn't delete the files from history, just clears the data locally.
You can use this with .checkout(version)
to clear a specific version of the drive.
Get a stream of all changes and their versions from this drive.
Read a file out as a stream. Similar to fs.createReadStream.
Options include:
{
start: optionalByteOffset, // similar to fs
end: optionalInclusiveByteEndOffset, // similar to fs
length: optionalByteLength
}
Read an entire file into memory. Similar to fs.readFile.
Options can either be an object or a string
Options include:
{
encoding: string
cached: true|false // default: false
}
or a string can be passed as options to simply set the encoding - similar to fs.
If cached
is set to true
, this function returns results only if they have already been downloaded.
Diff this drive with another version. version
can both be a version number of a checkout instance of the drive. The data
objects looks like this
{
type: 'put' | 'del',
name: '/some/path/name.txt',
value: {
// the stat object
}
}
Write a file as a stream. Similar to fs.createWriteStream.
If options.cached
is set to true
, this function returns results only if they have already been downloaded.
Write a file from a single buffer. Similar to fs.writeFile.
Unlinks (deletes) a file. Similar to fs.unlink.
Explictly create an directory. Similar to fs.mkdir
Delete an empty directory. Similar to fs.rmdir.
Lists a directory. Similar to fs.readdir.
Options include:
{
cached: true|false, // default: false
}
If cached
is set to true
, this function returns results from the local version of the drive’s append-tree. Default behavior is to fetch the latest remote version of the drive before returning list of directories.
Stat an entry. Similar to fs.stat. Sample output:
Stat {
dev: 0,
nlink: 1,
rdev: 0,
blksize: 0,
ino: 0,
mode: 16877,
uid: 0,
gid: 0,
size: 0,
offset: 0,
blocks: 0,
atime: 2017-04-10T18:59:00.147Z,
mtime: 2017-04-10T18:59:00.147Z,
ctime: 2017-04-10T18:59:00.147Z,
linkname: undefined }
The output object includes methods similar to fs.stat:
var stat = drive.stat('/hello.txt')
stat.isDirectory()
stat.isFile()
Options include:
{
cached: true|false // default: false,
wait: true|false // default: true
}
If cached
is set to true
, this function returns results only if they have already been downloaded.
If wait
is set to true
, this function will wait for data to be downloaded. If false, will return an error.
Stat an entry but do not follow symlinks. Similar to fs.lstat.
Options include:
{
cached: true|false // default: false,
wait: true|false // default: true
}
If cached
is set to true
, this function returns results only if they have already been downloaded.
If wait
is set to true
, this function will wait for data to be downloaded. If false, will return an error.
Similar to fs.access.
Options include:
{
cached: true|false // default: false,
wait: true|false // default: true
}
If cached
is set to true
, this function returns results only if they have already been downloaded.
If wait
is set to true
, this function will wait for data to be downloaded. If false, will return an error.
Open a file and get a file descriptor back. Similar to fs.open.
Note that currently only read mode is supported in this API.
Read from a file descriptor into a buffer. Similar to fs.read.
Close a file. Similar to fs.close.
Closes all open resources used by the drive.
The drive should no longer be used after calling this.
If you load this hyperdrive's key more than once at once, close()
will be a noop until all handles invoke it.
Closes all resources used by the drive, and destroys its data from storage. The drive should no longer be used after calling this.
Initializes a Hypercore (aka Feed) and begins replicating it.
keyOrName
: This must be provided. It's either ahyper://
URL / key or a string identifying the name of the feed. If you want to have a writable feed, you can use the name to generate one and use the name later to get the same feed back without having to save the key somewhere.opts
: The options for configuring this feedsparse: true
: Whether the history should be loaded on the fly instead of replicating the full historyvalueEncoding: 'json' | 'utf-8' | 'binary'
: The encoding to use for the data stored in the hypercore. Use JSON to store / retrieve objects.secretKey
: The secret key to use for the feed. Useful for restoring from backups.discoveryKey
: Optionally specify which discovery key you'd like to use for finding peers for this feed.lookup: true
: Specify whether you wish to lookup peers for this feed. Set tofalse
along withannounce
to avoid advertising.announce: true
: Specify whether you wish to advertise yourself as having the feed.
Append a block of data to the feed.
Callback is called with (err, seq)
when all data has been written at the returned seq
or an error occurred.
Get a block of data. If the data is not available locally this method will prioritize and wait for the data to be downloaded before calling the callback.
Options include
{
wait: true, // wait for index to be downloaded
timeout: 0, // wait at max some milliseconds (0 means no timeout)
valueEncoding: 'json' | 'utf-8' | 'binary' // defaults to the feed's valueEncoding
}
Callback is called with (err, data)
Get a range of blocks efficiently. Options include
{
wait: sameAsAbove,
timeout: sameAsAbove,
valueEncoding: sameAsAbove
}
Get the block of data at the tip of the feed. This will be the most recently appended block.
Accepts the same options
as feed.get()
.
Download a range of data. Callback is called when all data has been downloaded. A range can have the following properties:
{
start: startIndex,
end: nonInclusiveEndIndex,
linear: false // download range linearly and not randomly
}
If you do not mark a range the entire feed will be marked for download.
If you have not enabled sparse mode (sparse: true
in the feed constructor) then the entire
feed will be marked for download for you when the feed is created.
Cancel a previous download request.
Get a signature proving the correctness of the block at index, or the whole stream.
Callback is called with (err, signature)
.
The signature has the following properties:
{
index: lastSignedBlock,
signature: Buffer
}
Verify a signature is correct for the data up to index, which must be the last signed block associated with the signature.
Callback is called with (err, success)
where success is true only if the signature is
correct.
Retrieve the root hashes for given index
.
Callback is called with (err, roots)
; roots
is an Array of Node objects:
Node {
index: location in the merkle tree of this root
size: total bytes in children of this root
hash: hash of the children of this root (32-byte buffer)
}
Returns total number of downloaded blocks within range.
If end
is not specified it will default to the total number of blocks.
If start
is not specified it will default to 0.
Return true if a data block is available locally. False otherwise.
Return true if all data blocks within a range are available locally. False otherwise.
Clear a range of data from the local cache. Will clear the data from the bitfield and make a call to the underlying storage provider to delete the byte range the range occupies.
end
defaults to start + 1
.
Seek to a byte offset.
Calls the callback with (err, index, relativeOffset)
, where index
is the data block the byteOffset is contained in and relativeOffset
is
the relative byte offset in the data block.
Wait for the feed to contain at least minLength
elements.
If you do not provide minLength
it will be set to current length + 1.
Does not download any data from peers except for a proof of the new feed length.
console.log('length is', feed.length)
feed.update(function () {
console.log('length has increased', feed.length)
})
Create a readable stream of data.
Options include:
{
start: 0, // read from this index
end: feed.length, // read until this index
snapshot: true, // if set to false it will update `end` to `feed.length` on every read
tail: false, // sets `start` to `feed.length`
live: false, // set to true to keep reading forever
timeout: 0, // timeout for each data event (0 means no timeout)
wait: true // wait for data to be downloaded
}
Create a writable stream.
Fully close this feed.
If you loaded this feed more than once, close
will be a noop until all handles have invoked it.
Calls the callback with (err)
when all storage has been closed.
Closes the feed and deletes all of it's data from storage.
Audit all data in the feed. Will check that all current data stored matches the hashes in the merkle tree and clear the bitfield if not.
When done a report is passed to the callback that looks like this:
{
valid: 10, // how many data blocks matches the hashes
invalid: 0, // how many did not
}
If a block does not match the hash it is cleared from the data bitfield.
Listens on extension messages of type name
on the feeds replication channels.
handlers.encoding
: The encoding to use for messages.json
,binary
, 'utf8'handlers.onmessage(message, peer)
: Function to invoke when a peer sends you a message for this extension type.handlers.onerror(err, peer)
: Function to invoke when a peer has sent you a mis-coded message on this extension.
You can respond to messages with extension.send(message, peer)
.
You can also broadcast out messages to all peers with extension.broadcast(message)
Can we append to this feed?
Populated after ready
has been emitted. Will be false
before the event.
Can we read from this feed? After closing a feed this will be false.
Populated after ready
has been emitted. Will be false
before the event.
Buffer containing the public key identifying this feed.
Populated after ready
has been emitted. Will be null
before the event.
Buffer containing a key derived from the feed.key.
In contrast to feed.key
this key does not allow you to verify the data but can be used to announce or look for peers that are sharing the same feed, without leaking the feed key.
Populated after ready
has been emitted. Will be null
before the event.
How many blocks of data are available on this feed?
Populated after ready
has been emitted. Will be 0
before the event.
How much data is available on this feed in bytes?
Populated after ready
has been emitted. Will be 0
before the event.
Return per-peer and total upload/download counts.
The returned object is of the form:
{
totals: {
uploadedBytes: 100,
uploadedBlocks: 1,
downloadedBytes: 0,
downloadedBlocks: 0
},
peers: [
{
uploadedBytes: 100,
uploadedBlocks: 1,
downloadedBytes: 0,
downloadedBlocks: 0
},
...
]
}
Stats will be collected by default, but this can be disabled by setting opts.stats
to false.
Emitted when the feed is ready and all properties have been populated.
Emitted when the feed experiences a critical error.
Emitted when a data block has been downloaded.
Emitted when a data block is uploaded.
Emitted when the feed has been appended to (i.e. has a new length / byteLength)
Emitted every time ALL data from 0
to feed.length
has been downloaded.
Emitted when the feed has been fully closed
Emitted when a new peer has started replicating with the feed.
Extension messages and metadate about the remote peer isn't ready yet.
Emitted when a new peer has fully connected and shared it's metadata.
You should probably prefer this over peer-add.
Emitted when a peer has stopped replicating with the feed.
Emitted when a peer has acknowledged that it loaded some of your data.
Specify ack: true
in the corestoreOpts
to enable this.
You can get the have.start
and have.end
fields to see which portions the peer loaded. Alternately have.bitfield
will contain the bitfield of the blocks that got loaded.
If you are finding that WebRTC connections are not reliably made, you may get improved performance by using this:
https://github.com/webrtcHacks/adapter
E.g. In the browser code:
`<script src="lib/adaptor.js"></script>
<script src="lib/hyper-sdk-bundle.js"></script>`