Path of the NodeBases Jedi

But First...

What is a database?

Which of these is a database?

• MongoDB
• Cat
• Filesystem
• PostgreSQL

Which of these is a database?

• MongoDB
• Cat
• Filesystem (almost!)
• PostgreSQL... Yes, but...

NodeBases Goal:

Do for databases what NPM did for web applications.

Small, composable pieces of functionality.

LevelDB

Right now the "canonical" NodeBases back-end is LevelDB.

It is a simple key-value store with 5 primitives:

• get
• put
• del
• batch (put/del)
• ordered iterator

LevelUP/LevelDOWN

LevelUP provides a consistent JS interface to Node.

LevelDOWN is the wrapper interface for DB engines.

Lots of options:

• LevelDB
• HyperLevelDB
• RocksDB
• LMDB
• IndexedDB (browser!)
• ... and more!

How to get it

Easiest:

npm install level

level provides a LevelUP/LevelDOWN bundle ready to use.

• level-hyper HyperLevelDB
• level-lmdb LMDB
• level-rocksdb RocksDB
• level-mem In-memory
• level-js IndexedDB (browser)

How to use it.

// var level = require("level")
var db = level("./vanilla")

// put  key    value
db.put("wow", "mom", function (err) {
  if (err) return alert("IO Error!?", err)

  // get  key
  db.get("wow", function (err, value) {
    if (err) return alert("Not found?")
    $("#vanilla .results").text("value: " + value)
  })
})

(press "r" to run example)

Using level-sublevel to add namespaces. (Think 'tables')

// var level = require("level")
// var Sublevel = require("level-sublevel")
var db = Sublevel(level("./namespaced"))
var stuff = db.sublevel("stuff")
var things = db.sublevel("things")

stuff.put("wow", "mom", function (err) {
  things.put("wow", "zzz", function (err) {
    stuff.get("wow", function (err, value) {
      $("#namespaced .results").text("value: " + value)
    })
  })
})

(press "r" to run example)

Using level-version to make writes versioned.

// var level = require("level")
// var Sublevel = require("level-sublevel")
// var Version = require("level-version")
var db = Sublevel(level("./versioned"))
var things = Version(db.sublevel("things"))

things.put("wow", "mom", {version: 0}, function (err) {
  things.put("wow", "zzz", function (err) {
    things.get("wow", {version: 0}, function (err, value, version) {
      $("#versioned .results")
        .text("value: " + value + " version: " + version)
    })
  })
})

(press "r" to run example)

Case Study: TimestreamDB

timestreamdb is a full time series database with a full Node streams-based query language.

How Timestreamdb Works

TimestreamDB itself is really simple, just 45 lines of code. All it actually does is combine level-version and timestream

Small, composable pieces of functionality.

Because of NodeBases

TimestreamDB is theoretically compatible with 100s of other features provided by other NodeBases modules.

• browserifiable level-js (and others...)
• namespacing level-sublevel (and others...)
• live streaming of changes level-live-stream
• map-reduce map-reduce
• networked multilevel (and others...)
• foreign keys level-assoc (and others...)
• replication level-replicate (and others...)
• graph database levelgraph
• etc...

// var level = require("level")
// var TimestreamDB = require("timestreamdb")
var db = TimestreamDB(level("./tsdb", {valueEncoding: "json"}))

var count = 0
var i = setInterval(function () {
  if (++count > 30) {
    clearInterval(i)
    db.ts("foo").mean().tail(function (r) {
      $("#tsdb .results").text(r.value)
    })
  }
  db.put("foo", {id: count, value: Math.random()})
}, 10)

(press "r" to run example)

var db = level("slidetimer", {valueEncoding: "json"})
window.SlideTimeDB = TimestreamDB(db)
function recordSlide(event) {
  SlideTimeDB.put("slide", {slide: event.indexh + "," + event.indexv})
}

var count = SlideTimeDB.ts("slide")
  .count()
  .rename("slide", "count")

var avg = SlideTimeDB.ts("slide")
  .elapsed()
  .numbers()
  .mean()
  .divide(1000)
  .round(1)

count.join(avg).tail(function(r) {
  $("#numslides").text(r.count)
  $("#avgtime").text(r.elapsed)
})

Bryce, you've visited ??? slides at an average of:

??? seconds per slide.

The Dark Side of NodeBases

• Incomplete experiments
• Competing efforts
• Land grab
• Over-promise
• Combinatorial compatibility
• Missing a solid Replication story

None of these things are evil or bad!

... but they can scare away potential users.

Help us realize the NodeBases potential

• Use it, try it, push the limits!
• Favor Science over Mad Science
• Be a good citizen!
  • Label project status
  • Point to similar projects
  • Use the modules you write!!!
• We need a rock-solid replication story
• Published performance metrics

Hype Cycle

http://en.wikipedia.org/wiki/Hype_cycle