Repository files navigation

Simple Extensions for Scala

We are using SBT to build the project. Although, you can build it just using Scalac.

Simple Build Tool is the most popular build tool for scala. We respect that. So for compiling the code:

$ sbt compile

And for testing:

$ sbt test:run

SBT respects our own SUnit tests.For making a jar out of the project, do:

$ sbt package

Currently we have implemented two thing:

Actually,ir.angellandros.scala.collection.PairedIterable. The main reason for such a data structure is to havereduceByKey.You can now do this:

importir.angellandros.scala.collection.Implicits._vall=List(1,1,1,2,2,3)l.map(_->1).reduceByKey(_+_)

The idea behindKeyedVector is that getting rid of dictionary. You don't need a dictionary, because the indices could beString or any other thing that you want. Although it could beInt and you can use it with some dictionary.Anyway,KeyedVector has a close relation withHashMap. EveryKeyedVector has an ID, and this is the most important difference:

valvector=newKeyedVector(id, map)

and withvector.id orvector.toMap you can get either one of them. You can also dovector.get(key) orvector.keySet.

The other class implemented here isKeyedVectors, that provides some tools forKeyedVectors:

vali1=KeyedVectors.dot(v1, v2)vali2=KeyedVectors.normInfinity(v1)vali3=KeyedVectors.normNN(v1, n)// Minkowski norm ^n, which is squared Euclidean norm for n=2vali4=KeyedVectors.distNN(v1, v2, n)// distance function induces from normvali5=KeyedVectors.norm(v1)// Euclidean normvali6=KeyedVectors.eucDist(v1, v2)vali7=KeyedVectors.sqEucDist(v1, v2)// squared Euclidean distancevali8=KeyedVectors.cosineSim(v1, v2)vali9=KeyedVectors.cosineDist(v1, v2)

Canopy algorithm on pure scala. There are two implementations for canopy algorithm. First, the simple implementation with O(n^2), and the other, the merged implementation with O(n^{3/2}) running time.

valcanopy=newCanopyDriver(0.95,0.5)valresults= canopy.mergedRun(vectors,KeyedVectors.cosineDist[String])

As you can see, the distance function is an input argument.

Simple unit testing class. Now hasassertEquals.