Scalany mongodb aug10

Scala with MongoDB

Brendan W. McAdams

Novus Partners, Inc.

New York Scala Enthusiasts
Aug. 8, 2010

B.W. McAdams (Novus Partners) Integrating Scala + MongoDB NY Scala Enthusiasts - 8/8/10 1 / 60

Outline
1 Introduction
What is MongoDB?
A Taste of MongoDB
MongoDB + Scala Drivers
2 Scala + MongoDB == Win
mongo-scala-driver
lift-mongo
casbah
STM + MongoDB via Akka
3 Interlude: Helping Scala + Java play nice together.
Java <-> Scala Basics
Implicits and Pimp Hats
4 Closing


Introduction What is MongoDB?

Outline
1 Introduction
What is MongoDB?
A Taste of MongoDB
mongo-scala-driver
lift-mongo
casbah
4 Closing



Introducing MongoDB
Categorized as a “Document-Oriented Database”
Features of both Key-Value Stores & RDBMS’
Rich query interface.
Works with JSON-like Documents
Favors embedding related data over “foreign key” relationships
Free license (A-GPL) cross-platform (Packages for Linux,
Windows, Mac OS X, Windows, FreeBSD & Solaris)
Cursor-based query results
ServerSide Javascript
Stored Javascript functions server-side
Powerful aggregation - Map/Reduce, Group Commands
JS Statements in queries (no indexes though)
Indexing system is much like RDBMS, includes Geospatial
support.
Scalable ﬁle storage with GridFS
Data scalability with Replica Sets & Autosharding


Programming with MongoDB
Provides a native API which allows interaction to adapt to the
programming language (rather than vice versa).
Ofﬁcial drivers for. . .
C
C++
Java
JavaScript
Perl
PHP
Python
Ruby
Community supported drivers include. . .
.Net: C# & F#
JVM: Clojure, Scala, Groovy
Erlang
Haskell
Go
. . . and many more.


But is anyone actually *using* it?!?

MongoDB is deployed in production at companies including. . .
New York Times
Foursquare
bit.ly
SourceForge
Etsy
Disqus
Github
. . . The Large Hadron Collider.


Introduction A Taste of MongoDB

Outline
1 Introduction
What is MongoDB?
A Taste of MongoDB
mongo-scala-driver
lift-mongo
casbah
4 Closing



Core Concepts
MongoDB’s equivalent to “tables” are called
“collections”;“collections” contain “documents” (individual pieces
of data)
Databases & Collections are lazy - they are created when first
inserted into.
MongoDB’s wire format/internal document representation is
BSON. . .
BSON is a binary optimized flavor of JSON; corrects JSON’s
inefficiency in string encoding (Base64).
Supports extras including Regular Expressions, Byte Arrays,
DateTimes & Timestamps, as well as datatypes for Javascript code
blocks & functions.
BSON implementation being split into its own package in most
drivers.
Creative Commons licensed http://bsonspec.org
Java driver represents BSON with a map-like DBObject (Which most Scala drivers use);
many dynamic languages (Perl, Ruby, Python, etc) use native dictionary objects.


The basics of Querying I

Find a single row with findOne(); returns the first document found
(by natural order).
You can find all documents matching your query with find(). No
query means you get the entire collection back.
Queries are specified as BSON documents to match against.
The find() and findOne() methods can take an optional second
DBObject specifying the fields to return.
If you have an embedded object (for example, an address object)
you can retrieve it with dot notation in the fields list (e.g.
“address.city” retrieves just the city value).
Use limit(), skip() and sort() on result objects (DBCursor in
Java-driver land) to adjust your results. These all return a new
cursor.



The basics of Querying II
distinct() can be used (on DBCollection to ﬁnd all distinct values
for a given key; it returns a list of values.
> db.routes.findOne({"route_short_name": "E"})
{
"_id" : ObjectId("4c5f755608c3693f59580f8c"),
"route_id" : "E",
"agency_id" : "MTA NYCT",
"route_short_name" : "E",
"route_long_name" : "8 Avenue Local",
"route_desc" : "Trains operate between Jamaica Center (Parsons/Archer),
Queens, and World Trade Center, Manhattan, at all times.",
"route_type" : 1,
"route_url" : "http://www.mta.info/nyct/service/pdf/tecur.pdf"
}

> db.routes.find({"route_long_name": /Local$/},
{"route_short_name": 1, "route_long_name": 1})

{ "_id" : ObjectId("4c5f755608c3693f59580f7f"), "route_short_name" : 1,
"route_long_name" : "Broadway - 7 Avenue Local" }
{ "_id" : ObjectId("4c5f755608c3693f59580f84"), "route_short_name" : 6,
"route_long_name" : "Lexington Avenue Local" }
{ "_id" : ObjectId("4c5f755608c3693f59580f86"), "route_short_name" : 7,
"route_long_name" : "Flushing Local" }
{ "_id" : ObjectId("4c5f755608c3693f59580f8a"), "route_short_name" : "C",
"route_long_name" : "8 Avenue Local" }



The basics of Querying III
{ "_id" : ObjectId("4c5f755608c3693f59580f8c"), "route_short_name" : "E",
"route_long_name" : "8 Avenue Local" }
{ "_id" : ObjectId("4c5f755608c3693f59580f8d"), "route_short_name" : "F",
"route_long_name" : "Queens Blvd Express/ 6 Av Local" }
{ "_id" : ObjectId("4c5f755608c3693f59580f91"), "route_short_name" : "J",
"route_long_name" : "Nassau St Local" }
{ "_id" : ObjectId("4c5f755608c3693f59580f92"), "route_short_name" : "L",
"route_long_name" : "14 St-Canarsie Local" }
{ "_id" : ObjectId("4c5f755608c3693f59580f93"), "route_short_name" : "M",
"route_long_name" : "Nassau St Local" }
{ "_id" : ObjectId("4c5f755608c3693f59580f96"), "route_short_name" : "R",
"route_long_name" : "Broadway Local" }
{ "_id" : ObjectId("4c5f755608c3693f59580f99"), "route_short_name" : "V",
"route_long_name" : "Queens Blvd/6 Av Local" }
{ "_id" : ObjectId("4c5f755608c3693f59580f9a"), "route_short_name" : "W",
"route_long_name" : "Broadway Local" }

> db.routes.distinct("route_short_name")
[
1,
2,
3,
4,
5,
6,
7,
"A",
"B",
"C",



The basics of Querying IV

"D",
"E",
"F",
"G",
"J",
/*... */
]



Query Operators I

MongoDB is no mere Key-Value store. There are myriad powerful
operators to enhance your MongoDB queries. . .
Conditional Operators: $gt (>), $lt (<), $gte (>=), $lte (<=)
Negative Equality: $ne (!=)
Array Operators: $in (SQL “IN” clause. . . takes an array), $nin
(Opposite of “IN”), $all (Requires all values in the array match),
$size (Match the size of an array)
Field Deﬁned: $exists (boolean argument)(Great in a schemaless
world)
Regular Expressions (Language dependent - most drivers support
it)
Pass Arbitrary Javascript with $where (No OR statements, so use
WHERE for complex range ﬁlters)
Negate any operator with $not
Using a query operator requires nested objects. . .



Query Operators II
> db.stops.find({"stop_lat" : {$lt: 40.6}, {"stop_lon": {$gte: -73.8}}})
{ "_id" : ObjectId("4c5f755608c3693f59580ef0"), "stop_lat" : 40.590927, "stop_lon"
: -73.796924, "stop_id" : "H06", "stop_name" : "BEACH 67TH ST - GASTON",
"location_type" : 0, "stop_geo" : { "lat" : 40.590927, "lon" : -73.796924 } }
: -73.788522, "stop_id" : "H07", "stop_name" : "BEACH 60TH ST - STRAITON AV",
: -73.776013, "stop_id" : "H08", "stop_name" : "BEACH 44TH ST - FRANK AV",
: -73.768175, "stop_id" : "H09", "stop_name" : "BEACH 36TH ST - EDGEMERE",

> db.trips.findOne({"route_id": {$in: ["E", "4", "5"]}})
{
"_id" : ObjectId("4c5f755708c3693f59583400"),
"route_id" : "E",
"service_id" : "B20100308W",
"trip_id" : "B20100308W_001350_E..S04R",
"trip_headsign" : "To World Trade Ctr",
"direction_id" : 1,
"shape_id" : 177710
}

> db.trips.find({"route_id": {$in: ["E", "4", "5"]}}).count()
928



Query Operators III

No syntactic sugar in Java to make it easier. . .



Insert/Update/Save I

Objects in MongoDB Collections have an “_id” field, which must
be unique.
Three ways to add/update data in MongoDB. . .
insert() always attempts to add a new row. If “_id” is present
and contains a value already in the collection, insert fails.
save() inserts if there is no “_id” field, otherwise it tries to update
the document with the specified “_id”.
update() takes a query and the new values to save. By default it
updates only the first document matching the query.
For update() you can specify two booleans whose default is false:
upsert, which indicates you wish to create a new document if the
query doesn’t match, and multi, which allows updating all
documents who match the query.



Insert/Update/Save II

> db.testData.insert({"userCount": 5})
> x = db.testData.findOne({"userCount": 5})
{ "_id" : ObjectId("4c607f48150c335a4e187f41"), "userCount" : 5 }
> x.userCount
5
> x.userCount = 20
20
> db.testData.save(x)
> db.testData.findOne({_id: x._id})
> db.testData.update({_id: x._id}, {$inc: {"userCount": 12}})
> db.testData.findOne({_id: x._id})
// upsert
> db.testData.update({"userCount": 5}, {"userCount": 209}, true)
> db.testData.findOne({"userCount": 209} )
{ "_id" : ObjectId("4c60800e08c3693f5962dda5"), "userCount" : 209 }



Geospatial Support I

MongoDB supports Geospatial indexing and distance based
queries
I loaded all of the NYC Subway data (in Google Transit format)
into MongoDB
Quick python code to index the “Stops” data.
connection = Connection()
db = connection[’nyct_subway’]
print "Indexing the Stops Data."
for row in db.stops.find():
row[’stop_geo’] = {’lat’: row[’stop_lat’], ’lon’: row[’stop_lon’]}
db.stops.save(row)

db.stops.ensure_index([(’stop_geo’, pymongo.GEO2D)])

“stop_geo” ﬁeld is now Geospatially indexed for each stop.
How hard is it to ﬁnd the 5 closest subway stops to Meetup HQ?



Geospatial Support I

db.stops.find( { stop_geo: { $near: [40.726021, -73.99617] } }, {’stop_name’:
1}).limit(5);

{ "_id" : ObjectId("4c5f755608c3693f59580e9b"), "stop_name" : "BROADWAY-LAFAYETTE
" }
{ "_id" : ObjectId("4c5f755608c3693f59580e29"), "stop_name" : "BLEECKER
STREET-LEXINGTON" }
{ "_id" : ObjectId("4c5f755608c3693f59580f50"), "stop_name" : "PRINCE STREET
" }
{ "_id" : ObjectId("4c5f755608c3693f59580e2a"), "stop_name" : "SPRING
STREET-LEXINGTON" }
{ "_id" : ObjectId("4c5f755608c3693f59580f4f"), "stop_name" : "8TH STREET (NYU)
" }

Further commands exist to deﬁne a rectangle or circle radius for
the search.



Finally, Data Scalability.

Traditional master-slave replication
Replica Sets (new in 1.6)
Replaces master-slave setup with 1-7 server clusters
Automatic failover and recovery
AutoSharding (new in 1.6)
Horizontal scaling - partition your collections & data across as many
nodes as necessary.
Multiple nodes can service the same shard, allowing for balancing
& failover.
Map/Reduce runs across multiple shards, allowing concurrency.


Introduction MongoDB + Scala Drivers

Outline
1 Introduction
What is MongoDB?
A Taste of MongoDB
mongo-scala-driver
lift-mongo
casbah
4 Closing



Using Scala with the ofﬁcial Java Driver I

JVM Object are JVM Objects. . .
import com.mongodb._

val conn = new Mongo()
val db = conn.getDB("test")
val coll = db.getCollection("testData")

val pies = new BasicDBList()
pies.add("cherry")
pies.add("blueberry")
pies.add("apple")
pies.add("rhubarb")
pies.add("3.14")

val doc = new BasicDBObject()
doc.put("foo", "bar")
doc.put("spam", "eggs")
doc.put("up", "down")
doc.put("pie", pies)

coll.insert(doc)

. . . Not terribly “Scala-ey”.



Using Scala with the ofﬁcial Java Driver II

The Java driver works, but doesn’t ﬁt well in Scala.
You need to convert your Scala objects to Java Objects, and get
nothing but Java Objects out.
Gets messy quickly.



The Scala Community Adapted. . . I

Compare the previous with various Scala drivers.
mongo-scala-driver wraps & enhances the Java driver:
import com.osinka.mongodb._

val coll = db.getCollection("testData").asScala

coll << Map(
"foo" -> "bar",
"spam" -> "eggs",
"up" -> "down",
"pie" -> List(
"cherry",
"blueberry",
"apple",
"rhubarb",
"3.14"
)
)



The Scala Community Adapted. . . II
.. Much better, although I was confused initially. Has a
object<->MongoDB mapping layer.
lift-mongodb has more than one way to do it. . . here’s just a taste:

import net.liftweb.mongodb._
import net.liftweb.json._
import net.liftweb.json.JsonAST.JObject
import net.liftweb.json.JsonDSL._

implicit val formats = DefaultFormats.lossless

MongoDB.defineDb(DefaultMongoIdentifier,
MongoAddress(MongoHost("localhost", 27017)), "test")

val json = JsonParser.parse("""
{ "foo": "bar",
"spam": "eggs",
"up": "down",
"pie": [
"cherry",
"blueberry",
"apple",
"rhubarb",
"3.14"



The Scala Community Adapted. . . III

]
}
""").asInstanceOf[JObject]

MongoDB.useCollection("testData")( coll => {
coll.save(JObjectParser.parse(json))
})

. . . Lift’s JS & JSON tools make it very flexible, as we’ll see later.
Also has an ActiveRecord style Object<->MongoDB Mapping
layer.
Casbah reflects my own attempt at creating a sane interface
between Scala & MongoDB. Influenced by pymongo:



The Scala Community Adapted. . . IV

import com.novus.casbah.mongodb.Imports._

val coll = MongoConnection()("test")("testData")

val builder = MongoDBObject.newBuilder
builder += "foo" -> "bar"
builder += "spam" -> "eggs"
builder += "up" -> "down"
builder += "pie" -> List("cherry", "blueberry",
"apple", "rhubarb", "3.14")

coll += builder.result

. . . The syntax is still growing but is meant to match Scala syntax
sanely. Object<->MongoDB Mapping coming soon.
We’re going to cover several tools, although I know Casbah best.


Scala + MongoDB == Win mongo-scala-driver

Outline
1 Introduction
What is MongoDB?
A Taste of MongoDB
mongo-scala-driver
lift-mongo
casbah
4 Closing



At its core mongo-scala-driver provides a few improvements
over the Java driver via wrappers.
Converting the Java objects to Scala requires explicit conversions,
converts DBCollection to DBObjectCollection.
DBObjectCollection implements Iterable, and provides
several operators . . .
« maps to “insert”
«? “insert”s, and checks for errors.
+= maps to “save”
-= maps to “remove”
Map objects can be used in place of DBObject (as previously
shown) due to implicit conversions.
Object mapping implemented via “Shapes”; provides Query
syntax.



mongo-scala-driver: Shapes for Object Mapping I

“Shapes” uses Scala code to deﬁne Object models.
Flexible, but syntactically verbose.
Object representing “stops” on the NY Subway. . .
// Base representation
class SubwayStop extends MongoObject {
var id: Int = _
var name: String = _
var locationType: Option[Int] = None
var latitude: Double = _
var longitude: Double = _

override def toString: String =
"NYC Subway Stop at %s [Lat: %d, Lon: %d]".
format(name, latitude, longitude)
}

// Factory object
object SubwayStop extends MongoObjectShape[SubwayStop] {
lazy val id = Field.scalar("stop_id", _.id,
(x: SubwayStop, v: Int) => x.id = v)
lazy val name = Field.scalar("stop_name", _.name,
(x: SubwayStop, v: String) => x.name = v)
lazy val locationType = Field.optional(



mongo-scala-driver: Shapes for Object Mapping II

"location_type", _.locationType,
(x: SubwayStop, v: Option[Int]) =>
x.locationType = v)
lazy val latitude = Field.scalar("stop_lat", _.id,
(x: SubwayStop, v: Double) => x.latitude = v)
lazy val longitude = Field.scalar("stop_lon", _.id,
(x: SubwayStop, v: Double) => x.longitude = v)
// per docs, you must define * with all fields or (de)serialization won’t work.
override lazy val * = id :: name :: locationType ::
latitude :: longitude :: Nil

override def factory(dbo: DBObject) = Some(new SubwayStop)
}

// retrieving items
val db = conn.getDB("nyct_subway")
val coll = db.getCollection("stops")

val stopsColl = coll of SubwayStop
/* stopsColl is a ShapedCollection[SubwayStop] */
SubwayStop where {
SubwayStop.latitude < 40.6,
SubwayStop.longitude >= -73.8} sortBy SubwayStop.name descending


Scala + MongoDB == Win lift-mongo

Outline
1 Introduction
What is MongoDB?
A Taste of MongoDB
mongo-scala-driver
lift-mongo
casbah
4 Closing



lift-mongo I

Formerly “scamongo”, integrated with Lift as of 2.0
Base code provides session wrappers to MongoDB, still utilizes
Java driver’s DBObject code.
MongoDB.defineDb(DefaultMongoIdentifier,
MongoAddress(MongoHost("localhost", 27017)), "test")

MongoDB.useCollection(collectionName) ( coll => {
val doc = new BasicDBObject
doc.put("name", "MongoDB")
doc.put("type", "database")
doc.put("count", 1)
// save the doc to the db
coll.save(doc)
})

// Alternately, do everything in a single thread...
MongoDB.useSession ( db => {
val coll = db.getCollection("testCollection")
val doc = new BasicDBObject
doc.put("name", "MongoSession")
doc.put("type", "db")
doc.put("count", 1)
coll.save(doc)
})



lift-mongo II

“lift-mongo-record” provides object mapping.
No native query syntax, but Foursquare is working on open
sourcing something they use internally.



lift-mongo-record & querying I

Object deﬁnitions are fairly straightforward. . .
class MainDoc extends MongoRecord[MainDoc] with MongoId[MainDoc] {
def meta = MainDoc
object name extends StringField(this, 12)
object cnt extends IntField(this)
object refdoc extends DBRefField[MainDoc, RefDoc](this, RefDoc)
object refdocId extends ObjectIdField(this) {
def fetch = RefDoc.find(value)
}
}
object MainDoc extends MainDoc with MongoMetaRecord[MainDoc] {
def createRecord = new MainDoc
}
class RefDoc extends MongoRecord[RefDoc] with MongoId[RefDoc] {
def meta = RefDoc
}
object RefDoc extends RefDoc with MongoMetaRecord[RefDoc] {
def createRecord = new RefDoc
}

// Querying appears limited to constructing Mongo DBObjects
val mdq1 = MainDoc.findAll(("name" -> "md1"))



lift-mongo-record & querying II
Foursquare’s query library allow for a saner way to query data. . .
// FSMongoRecord extends "MongoRecord" to add a few methods
class Venue extends FSMongoRecord[Venue] {
def meta = Venue

object venuename extends FSStringField(this, 255)
object keywords extends StringField(this, 255)
object userid extends LongField(this)
object closed extends BooleanField(this) with AuditableField[Venue]
object mayor extends LegacyForeignKey(this, User) {
override def optional_? = true
}
object mayor_count extends OptionalIntField(this)
object aliases extends MongoListField[Venue, String](this)
object popularity extends MongoListField[Venue, Int](this)
object popularityUpdated extends OptionalJodaDateTimeField[Venue](this)

object tags extends MongoListField[Venue, String](this)
object categories extends MongoForeignObjectIdList(this, Category)
}

object Venue extends Venue with FSMetaRecord[Venue] {
override def collectionName = "venues"
def createRecord = new Venue
override def mongoIdentifier = NamedMongoIdentifier.venue
}



lift-mongo-record & querying III

// Foursquare’s query engine allows for fluid queries in code
Venue where (_.venuename is "Starbucks")
Venue where (_.venuename nin ("Starbucks", "Whole Foods"))
Venue where (_.geolatlng near (40.72, -73.99))

Thank you to @jliszka for sharing this!


Scala + MongoDB == Win casbah

Outline
1 Introduction
What is MongoDB?
A Taste of MongoDB
mongo-scala-driver
lift-mongo
casbah
4 Closing



Shameless Self Promotion

Why Casbah?
Background in pymongo + MongoKit
Java driver too. . . “Java-ey”
Didn’t quite “get” scamongo and mongo-scala-driver early on
scamongo’s base didn’t ﬁx most of my issues w/ the Java Driver
(just helped connection management)
scamongo’s ORM libraries were dependent on Lift (now
scamongo is defunct and has become lift-mongo)
mongo-scala-driver’s shapes, etc were very confusing to me as a
newbie w/o much functional background



Casbah is Born

Borrowed bits I liked/understood from other places and built
something that felt comfortable to me
Early on, very pythonic
Query DSL, grown from wanting a feel close to the “metal” based
on generic MongoDB knowledge
Heavily inﬂuenced in structure by @jorgeortiz85’s libraries
Quickly grew as I used more and more MongoDB with Scala;
features have been grown organically from my own needs.



Interacting with DBObjects I

DBObject is far too structurally Java.
Sought to make them more usable & readable from Scala
Most recently - match Scala 2.8 collection Factory/Builders
Implicit conversions of Product (base for Tuple), Map. Explicit
method asDBObject for corner cases.
‘Pimped’ version of DBObject via MongoDBObject - lets
DBObject implement Scala’s Map trait.



Interacting with DBObjects II

import com.novus.casbah.mongodb.Imports._ // Only import needed - mongoDB type
aliases imported too

val coll = MongoConnection()("test")("testData")

// Map
val map: DBObject = Map(
"foo" -> "bar",
"spam" -> "eggs",
"up" -> "down",
"pie" -> List(
"cherry",
"blueberry",
"apple",
"rhubarb",
"3.14"
)
)

// ’Product’
val product: DBObject =
( "foo" -> "bar",
"spam" -> "eggs",
"up" -> "down",
"pie" -> List(
"cherry",
"blueberry",
"apple",



Interacting with DBObjects III
"rhubarb",
"3.14"
)
).asDBObject // Explicit conversion method

// "Factory" method
val constructed: DBObject = MongoDBObject(
"foo" -> "bar",
"spam" -> "eggs",
"up" -> "down",
"pie" -> List(
"cherry",
"blueberry",
"apple",
"rhubarb",
"3.14"
)
)

// We showed the builder before
val builder = MongoDBObject.newBuilder
builder += "foo" -> "bar"
builder += "spam" -> "eggs"
builder += "up" -> "down"
builder += "pie" -> List("cherry", "blueberry",
"apple", "rhubarb", "3.14")

val built: DBObject = builder.result



Interacting with DBObjects IV

// Also responds to the ’Map’ methods...
built += "x" -> "y"
built.getOrElse("x", throw new Error("Can’t find value for X"))
/* res15: AnyRef = y */

DBCollection behaves as a Scala Iterable, but interaction is
mostly the same (with addition of methods like +=).



Fluid Query Syntax I

My thought: Instead of keeping track of Yet Another API,
MongoDB’s Query Objects should “just work”.
Two kinds of Query Operators - ‘Bareword’ and ‘Core’.
Bareword Operators can be started as ‘bare’ statements:
val setMulti = $set ("foo" -> 5, "bar" -> "N", "spam" -> "eggs")
/* setMulti: DBObject = { "$set" : { "foo" : 5 , "bar" : "N" , "spam" : "eggs"}} */
val pushAll = $pushAll ("foo" -> (5, 10, 15, 20, 25, 38, 12, "bar", "spam", 86,
"eggs", "omg", 412, "ponies"))
/* pushAll: DBObject = { "$pushAll" : { "foo" : [ 5 , 10 , 15 , 20 , 25 , 38 , 12 ,
"bar" , "spam" , 86 , "eggs" , "omg" , 412 , "ponies"]}} */



Fluid Query Syntax II

Core Operators need to be anchored to the right of a DBObject
or a String (typically representing a ﬁeld name):
// Find any documents where "foo" is between 5 and 15
val findFoo: DBObject = "foo" $gte 5 $lte 15
/* findFoo: DBObject = { "foo" : { "$gte" : 5 , "$lte" : 15}} */
// Find any documents where "bar" contains 1, 8 or 12
val findIn: DBObject = "foo" $in (1, 8, 12)
/* findIn: DBObject = { "foo" : { "$in" : [ 1 , 8 , 12]}} */

Just a small taste - all MongoDB Query Objects are supported
(For 1.4.x syntax - 1.6.x ($or, etc. soon))



Other Features I

Custom converter implementations which allow most Scala types
to be serialized cleanly to MongoDB. (Joda time
serialization/deserialization support).
Improved GridFS Functionality (loan pattern, support for
scala.io.Source)
Wrapper objects for Map/Reduce system (Help parse results to
warn of errors, etc)



Coming Soon I

Max Afonov @max4f working on annotation driven object mapping.
Investigating ActiveRecord implementation, with ﬂuid query syntax
support.
Support for MongoDB 1.6.x features.


Scala + MongoDB == Win STM + MongoDB via Akka

Outline
1 Introduction
What is MongoDB?
A Taste of MongoDB
mongo-scala-driver
lift-mongo
casbah
4 Closing


Scala + MongoDB == Win STM + MongoDB via Akka

STM + MongoDB via Akka I

Akka has an implementation of STM inspired by Clojure’s; allows
datastructures such as Maps and Vectors to become
transactional.
Akka STM supports persistence to several backends including
MongoDB.
Allows you to setup relatively simple, code managed concurrent
transactions with state stored safely in MongoDB.
Supports JTA; not yet distributed (Dependent on Multiverse, which
is working on distributed STM)


Interlude: Helping Scala + Java play nice together. Java <-> Scala Basics

Outline
1 Introduction
What is MongoDB?
A Taste of MongoDB
mongo-scala-driver
lift-mongo
casbah
4 Closing



Helping Java + Scala Interact

Implicits, “Pimp My Library” and various conversion helper tools
simplify the work of interacting with Java.
Scala and Java have their own completely different collection
libraries.
Some builtins ship with Scala to make this easier.



Interoperability in Scala 2.7.x

Scala 2.7.x shipped with scala.collection.jcl.
scala.collection.jcl.Conversions contained some
implicit converters, but only to and from the wrapper versions - no
support for “real” Scala collections.
Neglected useful base interfaces like Iterator and Iterable
@jorgeortiz85 provided scala-javautils, which used “Pimp
My Library” to do a better job.



Interoperability in Scala 2.8.x
Scala 2.8.x improves the interop game signiﬁcantly.
JCL is gone - focus has shifted to proper interoperability w/ built-in
types.
scala.collection.jcl.Conversions replaced by
scala.collection.JavaConversions - provides implicit
conversions to & from Scala & Java Collections.
Includes support for the things missing in 2.7 (Iterable,
Iterator, etc.)
Great for places where the compiler can guess what you want
(implicits); falls short in some cases (like BSON Encoding, as we
found in Casbah)
@jorgeortiz85 has updated scala-javautils for 2.8 with
scalaj-collection
Explicit asJava / asScala methods for conversions. Adds
foreach method to Java collections.

Interlude: Helping Scala + Java play nice together. Implicits and Pimp Hats

Outline
1 Introduction
What is MongoDB?
A Taste of MongoDB
mongo-scala-driver
lift-mongo
casbah
4 Closing



So WTF is an ‘Implicit’, anyway?
Implicit Arguments
‘Explicit’ arguments indicates a method argument you pass, well
explicitly.
‘Implicit’ indicates a method argument which is. . . implied. (But you
can pass them explicitly too.)
Implicit arguments are passed in Scala as an additional argument
list:
import org.bson.types.ObjectId

def query(id: ObjectId)(implicit coll: DBCollection) = coll.findOne(id)

implicit val coll = db.getCollection("testData")

// coll is passed implicitly
query(new ObjectId())

// or we can override the argument
query(new ObjectId())(db.getCollection("testDataExplicit"))

How does this differ from default arguments?


So WTF is an ‘Implicit’, anyway?

Implicit Methods/Conversions
If you try passing a type to a Scala method argument which doesn’t
match. . .
def printNumber(x: Int) = println(x)

printNumber(5)
printNumber("212") // won’t compile

A fast and loose example, but simple. Fails to compile.
But with implicit methods, we can provide a conversion path. . .
implicit def strToNum(x: String) = x.toInt
def printNumber(x: Int) = println(x)

printNumber(5)
printNumber("212")

In a dynamic language, this may be called “monkey patching”.
Unlike Perl, Python, etc. Scala resolves implicits at compile time.



Pimp My Library
Coined by Martin Odersky in a 2006 Blog post. Similar to C#
extension methods, Ruby modules.
Uses implicit conversions to tack on new methods at runtime.
Either return a new “Rich_” or anonymous class. . .
import com.mongodb.gridfs.{GridFSInputFile => MongoGridFSInputFile}

class GridFSInputFile protected[mongodb](override val underlying:
MongoGridFSInputFile) extends GridFSFile {
def filename_=(name: String) = underlying.setFilename(name)
def contentType_=(cT: String) = underlying.setContentType(cT)
}

object PimpMyMongo {
implicit def mongoConnAsScala(conn: Mongo) = new {
def asScala = new MongoConnection(conn)
}

implicit def enrichGridFSInput(in: MongoGridFSInputFile) =
new GridFSInputFile(in)
}

import PimpMyMongo._

A note: with regards to type boundaries, [A <: SomeType]
won’t allow implicitly converted values. You can whitelist -them by / 60
B.W. McAdams (Novus Partners) Integrating Scala + MongoDB NY Scala Enthusiasts 8/8/10 58

Closing

Links

mongo-scala-driver http://github.com/alaz/mongo-scala-driver
lift-mongo http://www.assembla.com/wiki/show/liftweb/MongoDB
FourSquare’s Lift Mongo DSL Code . . . coming soon? @jliszka

Casbah http://novus.github.com/docs/casbah
Jorge Ortiz’ (@jorgeortiz85) Libraries
scala-javautils (Scala 2.7.x) http://github.com/jorgeortiz85/scala-javautils
scalaj-collection (Scala 2.8.x) http://github.com/scalaj/scalaj-collection
This presentation
http://github.com/bwmcadams/presentations/tree/master/scala_mongodb/scalany_aug10/


Closing

Contact Info

Twitter: @rit

Email: bwmcadams@gmail.com

Github: http://github.com/bwmcadams | http://github.com/novus
IRC - freenode.net #mongodb
MongoDB Mailing List http://groups.google.com/group/mongodb-user
Casbah Mailing List http://groups.google.com/group/mongodb-casbah-user
Boston MongoDB Conference - Sept. 20 (Cambridge, Mass.)
http://10gen.com/conferences/mongoboston2010

MongoDB NY Users Group
http://www.meetup.com/New-York-MongoDB-User-Group/


Scalany mongodb aug10

More Related Content

What's hot (20)

Similar to Scalany mongodb aug10 (20)

Scalany mongodb aug10