InjestDB

A peer-to-peer database for dat:// applications. How it works

Example

Setup a database for social profiles, which can publish status updates and like other users' posts.

const Injest = require('injestdb')
var db = new Injest('social-profiles')
db.schema({
  version: 1,
  broadcasts: {
    primaryKey: 'createdAt',
    index: [
      'createdAt',
      '_origin+createdAt' // compound index. '_origin' is an autogenerated attribute which represets the URL of the authoring archive
    ],
    validator: record => {
      assert(typeof record.text === 'string')
      assert(typeof record.createdAt === 'number')
      return record
    }
  },
  likes: {
    primaryKey: 'createdAt',
    index: 'targetUrl',
    validator: record => {
      assert(typeof record.targetUrl === 'string')
      return record
    }
  },
  profile: {
    singular: true,
    index: 'name',
    validator: record => {
      assert(typeof record.name === 'string')
      return {
        name: record.name,
        description: isString(record.description) ? record.description : '',
        avatarUrl: isString(record.avatarUrl) ? record.avatarUrl : ''
      }
    }
  }
})

Next we add source archives to be ingested (added ot the dataset). The source archives are persisted in IndexedDB, so this doesn't have to be done every run.

await db.addArchives([alicesUrl, bobsUrl, carlasDatArchive], {prepare: true})

Now we can begin querying the database for records.

// get the first profile record where name === 'bob'
var bobProfile = await db.profiles.get('name', 'bob')

// get all profile records which match this query
var bobProfiles = await db.profiles
  .where('name')
  .equalsIgnoreCase('bob')
  .toArray()

// get the 30 latest broadcasts from all source archives
var recentBroadcasts = await db.broadcasts
  .orderBy('createdAt')
  .reverse() // most recent first
  .limit(30)
  .toArray()

// get the 30 latest broadcasts by a specific archive (bob)
// - this uses a compound index to filter by origin, and then sort by createdAt
var bobsRecentBroadcasts = await db.broadcasts
  .where('_origin+createdAt')
  .between([bobsUrl, ''], [bobsUrl, '\uffff'])
  .reverse() // most recent first
  .limit(30)
  .toArray()

// get the # of likes for a broadcast
var numLikes = await db.likes
  .where('targetUrl').equals(bobsRecentBroadcasts[0]._url) // _url is an autogenerated attribute which represents the URL of the record
  .count()

We can also use Injest to create, modify, and delete records (and their matching files).

// update bob's name
await db.profiles.update(bobsUrl, {name: 'robert'})

// publish a new broadcast for bob
var broadcastUrl = await db.broadcasts.add(bobsUrl, {
  text: 'Hello!',
  createdAt: Date.now()
})

// modify the broadcast
await db.broadcasts.update(broadcastUrl, {text: 'Hello world!'})

// like the broadcast
await db.likes.add(bobsUrl, {
  targetUrl: broadcastUrl,
  createdAt: Date.now()
})

// delete the broadcast
await db.broadcasts.delete(broadcastUrl)

// delete all likes on the broadcast (that we control)
await db.likes
  .where({targetUrl: broadcastUrl})
  .delete()

TODOs

Injest is still in development.

• [x] Indexer
• [x] Core query engine
• [x] Persisted tables and table reindex on schema change
• [x] Mutation methods (add/update/delete)
• [ ] Events
• [x] Multikey indexes
• [ ] Validation: filename must match primaryKey on non-singular tables
• [ ] Support for .or() queries
• [ ] Complete documentation

API quick reference

var db = new InjestDB(name)
InjestDB.list() => Promise<Void>
InjestDB.delete(name) => Promise<Void>
db.open() => Promise<Void>
db.close() => Promise<Void>
db.schema(Object) => Promise<Void>
db.addArchive(url|DatArchive, {prepare: Boolean}) => Promise<Void>
db.addArchives(Array<url|DatArchive>, {prepare: Boolean}) => Promise<Void>
db.removeArchive(url|DatArchive) => Promise<Void>
db.prepareArchive(url|DatArchive)
db.listArchives() => Promise<url>
db 'open' ()
db 'open-failed' (error)
db 'versionchange' ()
db 'indexes-updated' (archive, archiveVersion)

db.{table} => InjestTable
InjestTable#add(archive, record) => Promise<url>
InjestTable#count() => Promise<Number>
InjestTable#delete(url) => Promise<url>
InjestTable#each(Function) => Promise<Void>
InjestTable#filter(Function) => InjestRecordset
InjestTable#get(url) => Promise<Object>
InjestTable#get(archive) => Promise<Object>
InjestTable#get(archive, key) => Promise<Object>
InjestTable#get(index, value) => Promise<Object>
InjestTable#isRecordFile(String) => Boolean
InjestTable#limit(Number) => InjestRecordset
InjestTable#listRecordFiles(Archive) => Promise<Object>
InjestTable#name => String
InjestTable#offset(Number) => InjestRecordset
InjestTable#orderBy(index) => InjestRecordset
InjestTable#put(url, record) => Promise<url>
InjestTable#reverse() => InjestRecordset
InjestTable#schema => Object
InjestTable#toArray() => Promise<Array>
InjestTable#toCollection() => InjestRecordset
InjestTable#update(record) => Promise<Number>
InjestTable#update(url, updates) => Promise<Number>
InjestTable#update(archive, updates) => Promise<Number>
InjestTable#update(archive, key, updates) => Promise<Number>
InjestTable#upsert(url, record) => Promise<Void | url>
InjestTable#upsert(archive, record) => Promise<Void | url>
InjestTable#where(index) => InjestWhereClause
InjestTable 'index-updated' (archive, archiveVersion)

InjestWhereClause#above(lowerBound) => InjestRecordset
InjestWhereClause#aboveOrEqual(lowerBound) => InjestRecordset
InjestWhereClause#anyOf(Array|...args) => InjestRecordset
InjestWhereClause#anyOfIgnoreCase(Array|...args) => InjestRecordset
InjestWhereClause#below(upperBound) => InjestRecordset
InjestWhereClause#belowOrEqual(upperBound) => InjestRecordset
InjestWhereClause#between(lowerBound, upperBound, {includeUpper, includeLower}) => InjestRecordset
InjestWhereClause#equals(value) => InjestRecordset
InjestWhereClause#equalsIgnoreCase(value) => InjestRecordset
InjestWhereClause#noneOf(Array|...args) => InjestRecordset
InjestWhereClause#notEqual(value) => InjestRecordset
InjestWhereClause#startsWith(value) => InjestRecordset
InjestWhereClause#startsWithAnyOf(Array|...args) => InjestRecordset
InjestWhereClause#startsWithAnyOfIgnoreCase(Array|...args) => InjestRecordset
InjestWhereClause#startsWithIgnoreCase(value) => InjestRecordset

InjestRecordset#clone() => InjestRecordset
InjestRecordset#count() => Promise<Number>
InjestRecordset#delete() => Promise<Number>
InjestRecordset#distinct() => InjestRecordset
InjestRecordset#each(Function) => Promise<Void>
InjestRecordset#eachKey(Function) => Promise<Void>
InjestRecordset#eachUrl(Function) => Promise<Void>
InjestRecordset#filter(Function) => InjestRecordset
InjestRecordset#first() => Promise<Object>
InjestRecordset#keys() => Promise<Array<String>>
InjestRecordset#last() => Promise<Object>
InjestRecordset#limit(Number) => InjestRecordset
InjestRecordset#offset(Number) => InjestRecordset
InjestRecordset#or(index) => InjestWhereClause
InjestRecordset#put(Object) => Promise<Number>
InjestRecordset#urls() => Promise<Array<String>>
InjestRecordset#reverse() => InjestRecordset
InjestRecordset#toArray() => Promise<Array<Object>>
InjestRecordset#uniqueKeys() => Promise<Array<String>>
InjestRecordset#until(Function) => InjestRecordset
InjestRecordset#update(Object|Function) => Promise<Number>
InjestRecordset#where(index) => InjestWhereClause

API

db.schema(definition)

{
  version: Number, // the version # of the schema, should increment by 1 on each change

  [tableName]: {
    // is there only one record-file per archive?
    // - if true, will look for the file at `/${tableName}.json`
    // - if false, will look for files at `/${tableName}/*.json`
    singular: Boolean,

    // attribute to build filenames for newly-created records
    // ie `/${tableName}/${record[primaryKey]}.json`
    // only required if !singular
    primaryKey: String, 

    // specify which fields are indexed for querying
    // each is a keypath, see https://www.w3.org/TR/IndexedDB/#dfn-key-path
    // can specify compound indexes with a + separator in the keypath
    // eg one index               - index: 'firstName' 
    // eg two indexes             - index: ['firstName', 'lastName']
    // eg add a compound index    - index: ['firstName', 'lastName', 'firstName+lastName']
    // eg index an array's values - index: ['firstName', '*favoriteFruits']
    index: String|Array<String>,

    // validator & sanitizer
    // takes the ingested file (must be valid json)
    // returns the record to be stored
    // returns falsy or throws to not store the record
    validator: Function(Object) => Object
  }
}

How it works

InjestDB abstracts over the DatArchive API to provide a simple database-like interface. It is based heavily on Dexie.js and built using IndexedDB.

Injest works by scanning a set of source archives for files that match a path pattern. Those files are indexed ("ingested") so that they can be queried easily. Injest also provides a simple interface for adding, editing, and removing records on the archives that the local user owns.

Injest sits on top of Dat archives. It duplicates the data it's handling into IndexedDB, and that duplicated data acts as a throwaway cache -- it can be reconstructed at any time from the Dat archives.

Injest treats individual files in the Dat archive as individual records in a table. As a result, there's a direct mapping for each table to a folder of .json files. For instance, if you had a 'tweets' table, it would map to the /tweets/*.json files. Injest's mutators, such as put or add or update, simply write those json files. Injest's readers & query-ers, such as get() or where(), read from the IndexedDB cache.

Injest watches its source archives for changes to the json files. When they change, it reads them and updates IndexedDB, thus the query results stay up-to-date. The flow is, roughly: put() -> archive/tweets/12345.json -> indexer -> indexeddb -> get().