Fix Scala 2.13 compilation errors

- Fix ambiguous getAs method references by adding explicit type parameters
  - CLARA.scala: Specify Vector type and index for getAs calls
  - KMedoids.scala: Specify Vector type and index for getAs calls
  - SoftKMeans.scala: Specify Vector/Double types and column names for getAs calls
- Fix tuple pattern matching error in KMedoids.scala by simplifying assignment
- Remove unused imports to clean up compiler warnings
  - Remove unused scala.language.implicitConversions from compat package
  - Remove unused Vectors import from KMedoids.scala
  - Remove unused Random and DoubleType imports from CoresetKMeans.scala
  - Remove unused spark.implicits import from PersistenceRoundTripCoresetKMeans.scala

All changes maintain API compatibility and preserve existing functionality.
Verified compilation with both Scala 2.13.14 and 2.12.18 against Spark 3.5.1.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

Author: derrickburns

src/main/scala-2.13/com/massivedatascience/clusterer/compat/package.scala

@@ -1,7 +1,5 @@
 package com.massivedatascience.clusterer
 
-import scala.language.implicitConversions
-
 // Scala 2.13: Provide .par extension method via compat package
 package object compat {
   implicit class ParOps[A, CC[X] <: Iterable[X]](private val coll: CC[A]) extends AnyVal {

src/main/scala/com/massivedatascience/clusterer/ml/CLARA.scala

@@ -85,7 +85,7 @@ class CLARA(override val uid: String)
       .select($(featuresCol))
       .rdd
       .map { row =>
-        row.getAs
+        row.getAs[Vector](0)
       }
       .collect()
 

src/main/scala/com/massivedatascience/clusterer/ml/CoresetKMeans.scala

@@ -7,8 +7,7 @@ import org.apache.spark.ml.param.ParamMap
 import org.apache.spark.ml.util.{ DefaultParamsReadable, DefaultParamsWritable, Identifiable }
 import org.apache.spark.sql.{ DataFrame, Dataset }
 import org.apache.spark.sql.functions._
-import org.apache.spark.sql.types.{ DoubleType, StructType }
-import scala.util.Random
+import org.apache.spark.sql.types.StructType
 
 /** Core-set based K-Means clustering with pluggable Bregman divergences.
   *

src/main/scala/com/massivedatascience/clusterer/ml/KMedoids.scala

@@ -2,7 +2,7 @@ package com.massivedatascience.clusterer.ml
 
 import org.apache.spark.internal.Logging
 import org.apache.spark.ml.Estimator
-import org.apache.spark.ml.linalg.{ Vector, Vectors }
+import org.apache.spark.ml.linalg.Vector
 import org.apache.spark.ml.param._
 import org.apache.spark.ml.param.shared._
 import org.apache.spark.ml.util._
@@ -102,7 +102,7 @@ class KMedoids(override val uid: String)
       .select($(featuresCol))
       .rdd
       .map { row =>
-        row.getAs
+        row.getAs[Vector](0)
       }
       .collect()
 
@@ -351,8 +351,7 @@ class KMedoids(override val uid: String)
     val (_, _, costHistory) = swapPhaseWithHistory(data, initialMedoidIndices, maxIter, distFn)
     // last state is encoded in the returned medoid indices by swapPhaseWithHistory
     // but we need to recompute those indices; reuse the logic:
-    val distFnLocal         = distFn
-    val (_, _, _)           = (distFnLocal, costHistory) // keep params used, avoid warnings
+    val _                   = (distFn, costHistory) // keep params used, avoid warnings
     // For API compatibility, just re-run a lightweight swap without history:
     val n                   = data.length
     val k                   = initialMedoidIndices.length
@@ -568,7 +567,7 @@ class KMedoidsModel(
     df.select($(featuresCol))
       .rdd
       .map { row =>
-        val features = row.getAs
+        val features = row.getAs[Vector](0)
         val meds     = bcMedoids.value
         meds.map(m => distFn(features, m)).min
       }

src/main/scala/com/massivedatascience/clusterer/ml/SoftKMeans.scala

@@ -232,7 +232,7 @@ class SoftKMeans(override val uid: String)
       .sample(withReplacement = false, fraction, seed)
       .limit(k)
       .collect()
-      .map(_.getAs.toArray)
+      .map(_.getAs[org.apache.spark.ml.linalg.Vector](0).toArray)
   }
 
   /** Compute soft assignment probabilities for all points (Boltzmann distribution). */
@@ -280,10 +280,10 @@ class SoftKMeans(override val uid: String)
       var totalWeight                = 0.0
 
       membershipData.foreach { row =>
-        val features    = row.getAs.toArray
-        val probs       = row.getAs.toArray
+        val features    = row.getAs[org.apache.spark.ml.linalg.Vector](featCol).toArray
+        val probs       = row.getAs[org.apache.spark.ml.linalg.Vector]($(probabilityCol)).toArray
         val clusterProb = probs(clusterId)
-        val w           = weightColOpt.map(_ => row.getAs).getOrElse(1.0)
+        val w           = weightColOpt.map(col => row.getAs[Double](col)).getOrElse(1.0)
         val finalWeight = clusterProb * w
 
         if (weightedSum == null) weightedSum = Array.fill(features.length)(0.0)

src/main/scala/examples/PersistenceRoundTripCoresetKMeans.scala

@@ -22,8 +22,6 @@ object PersistenceRoundTripCoresetKMeans {
     val spark =
       SparkSession.builder().appName("CoresetKMeans Persistence").master("local[*]").getOrCreate()
 
-    import spark.implicits._
-
     mode match {
       case "save" => saveModel(spark, path)
       case "load" => loadModel(spark, path)