From 6afde57ebf3d86ae3a0bb07cdb42875a7bbbd0ff Mon Sep 17 00:00:00 2001
From: Holden Karau <holden@pigscanfly.ca>
Date: Tue, 1 Sep 2015 23:09:26 -0700
Subject: [PATCH 01/13] switch to withColumn

---
 .../scala/org/apache/spark/ml/regression/LinearRegression.scala | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala b/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala
index 884003eb38524..f1d78516e58c1 100644
--- a/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala
+++ b/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala
@@ -300,7 +300,7 @@ class LinearRegressionModel private[ml] (
   private[regression] def evaluate(dataset: DataFrame): LinearRegressionSummary = {
     val t = udf { features: Vector => predict(features) }
     val predictionAndObservations = dataset
-      .select(col($(labelCol)), t(col($(featuresCol))).as($(predictionCol)))
+      .withColumn(col($(labelCol)), t(col($(featuresCol))).as($(predictionCol)))
 
     new LinearRegressionSummary(predictionAndObservations, $(predictionCol), $(labelCol))
   }

From 43420b7a14ff88f7d6aa3f288093537a8e52502f Mon Sep 17 00:00:00 2001
From: Holden Karau <holden@pigscanfly.ca>
Date: Tue, 1 Sep 2015 23:35:43 -0700
Subject: [PATCH 02/13] Fix it to work and add a test

---
 .../spark/ml/regression/LinearRegression.scala       |  2 +-
 .../spark/ml/regression/LinearRegressionSuite.scala  | 12 ++++++++++++
 2 files changed, 13 insertions(+), 1 deletion(-)

diff --git a/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala b/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala
index f1d78516e58c1..f06972bbb0f51 100644
--- a/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala
+++ b/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala
@@ -300,7 +300,7 @@ class LinearRegressionModel private[ml] (
   private[regression] def evaluate(dataset: DataFrame): LinearRegressionSummary = {
     val t = udf { features: Vector => predict(features) }
     val predictionAndObservations = dataset
-      .withColumn(col($(labelCol)), t(col($(featuresCol))).as($(predictionCol)))
+      .withColumn($(predictionCol), t(col($(featuresCol))))
 
     new LinearRegressionSummary(predictionAndObservations, $(predictionCol), $(labelCol))
   }
diff --git a/mllib/src/test/scala/org/apache/spark/ml/regression/LinearRegressionSuite.scala b/mllib/src/test/scala/org/apache/spark/ml/regression/LinearRegressionSuite.scala
index 2aaee71ecc734..7eb1b58aebd06 100644
--- a/mllib/src/test/scala/org/apache/spark/ml/regression/LinearRegressionSuite.scala
+++ b/mllib/src/test/scala/org/apache/spark/ml/regression/LinearRegressionSuite.scala
@@ -460,6 +460,18 @@ class LinearRegressionSuite extends SparkFunSuite with MLlibTestSparkContext {
     // Training results for the model should be available
     assert(model.hasSummary)
 
+    // Feature column should be equal
+    val expectedFeature = dataset.select("features").collect()
+    val resultFeature = model.summary.predictions.select("features").collect()
+    resultFeature.zip(expectedFeature).foreach{ case (r1, r2) =>
+      val result1 = r1.getAs[DenseVector](0)(0)
+      val result2 = r1.getAs[DenseVector](0)(1)
+      val expected1 = r2.getAs[DenseVector](0)(0)
+      val expected2 = r2.getAs[DenseVector](0)(1)
+      assert(result1 ~== expected1 relTol 1E-5)
+      assert(result2 ~== expected2 relTol 1E-5)
+    }
+
     // Residuals in [[LinearRegressionResults]] should equal those manually computed
     val expectedResiduals = dataset.select("features", "label")
       .map { case Row(features: DenseVector, label: Double) =>

From e633ac562792a2a302fb39af98da75b897bb7e28 Mon Sep 17 00:00:00 2001
From: Holden Karau <holden@pigscanfly.ca>
Date: Sun, 13 Sep 2015 19:39:47 -0700
Subject: [PATCH 03/13] CR feedback, use transform on the dataset when calling
 evaluate and als change our test to just check the new expected schema

---
 .../spark/ml/regression/LinearRegression.scala     |  6 +-----
 .../ml/regression/LinearRegressionSuite.scala      | 14 +++-----------
 2 files changed, 4 insertions(+), 16 deletions(-)

diff --git a/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala b/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala
index f06972bbb0f51..96a738f4f7779 100644
--- a/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala
+++ b/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala
@@ -298,11 +298,7 @@ class LinearRegressionModel private[ml] (
    */
   // TODO: decide on a good name before exposing to public API
   private[regression] def evaluate(dataset: DataFrame): LinearRegressionSummary = {
-    val t = udf { features: Vector => predict(features) }
-    val predictionAndObservations = dataset
-      .withColumn($(predictionCol), t(col($(featuresCol))))
-
-    new LinearRegressionSummary(predictionAndObservations, $(predictionCol), $(labelCol))
+    new LinearRegressionSummary(transform(dataset), $(predictionCol), $(labelCol))
   }
 
   override protected def predict(features: Vector): Double = {
diff --git a/mllib/src/test/scala/org/apache/spark/ml/regression/LinearRegressionSuite.scala b/mllib/src/test/scala/org/apache/spark/ml/regression/LinearRegressionSuite.scala
index 7eb1b58aebd06..60960dc2738bc 100644
--- a/mllib/src/test/scala/org/apache/spark/ml/regression/LinearRegressionSuite.scala
+++ b/mllib/src/test/scala/org/apache/spark/ml/regression/LinearRegressionSuite.scala
@@ -460,17 +460,9 @@ class LinearRegressionSuite extends SparkFunSuite with MLlibTestSparkContext {
     // Training results for the model should be available
     assert(model.hasSummary)
 
-    // Feature column should be equal
-    val expectedFeature = dataset.select("features").collect()
-    val resultFeature = model.summary.predictions.select("features").collect()
-    resultFeature.zip(expectedFeature).foreach{ case (r1, r2) =>
-      val result1 = r1.getAs[DenseVector](0)(0)
-      val result2 = r1.getAs[DenseVector](0)(1)
-      val expected1 = r2.getAs[DenseVector](0)(0)
-      val expected2 = r2.getAs[DenseVector](0)(1)
-      assert(result1 ~== expected1 relTol 1E-5)
-      assert(result2 ~== expected2 relTol 1E-5)
-    }
+    // Schema should be a superset of the input dataset
+    assert(model.summary.predictions.schema.fieldNames.toSet ===
+      dataset.schema.fieldNames.toSet ++ Set("prediction"))
 
     // Residuals in [[LinearRegressionResults]] should equal those manually computed
     val expectedResiduals = dataset.select("features", "label")

From 64e9f3f7a9404639b7be6f10f839f9737561221d Mon Sep 17 00:00:00 2001
From: Holden Karau <holden@pigscanfly.ca>
Date: Fri, 18 Sep 2015 14:01:02 -0700
Subject: [PATCH 04/13] murh

---
 .../org/apache/spark/ml/regression/LinearRegressionSuite.scala  | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mllib/src/test/scala/org/apache/spark/ml/regression/LinearRegressionSuite.scala b/mllib/src/test/scala/org/apache/spark/ml/regression/LinearRegressionSuite.scala
index 60960dc2738bc..bbfb08942060e 100644
--- a/mllib/src/test/scala/org/apache/spark/ml/regression/LinearRegressionSuite.scala
+++ b/mllib/src/test/scala/org/apache/spark/ml/regression/LinearRegressionSuite.scala
@@ -461,7 +461,7 @@ class LinearRegressionSuite extends SparkFunSuite with MLlibTestSparkContext {
     assert(model.hasSummary)
 
     // Schema should be a superset of the input dataset
-    assert(model.summary.predictions.schema.fieldNames.toSet ===
+    assert(model.summary.predictions.schema.fieldNames.toSet contains
       dataset.schema.fieldNames.toSet ++ Set("prediction"))
 
     // Residuals in [[LinearRegressionResults]] should equal those manually computed

From 9e96c133b6490eaabc0a099d5acfa82cb437930f Mon Sep 17 00:00:00 2001
From: Holden Karau <holden@pigscanfly.ca>
Date: Fri, 18 Sep 2015 14:02:03 -0700
Subject: [PATCH 05/13] use subsetOf for test

---
 .../apache/spark/ml/regression/LinearRegressionSuite.scala    | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/mllib/src/test/scala/org/apache/spark/ml/regression/LinearRegressionSuite.scala b/mllib/src/test/scala/org/apache/spark/ml/regression/LinearRegressionSuite.scala
index bbfb08942060e..47a030221b6e1 100644
--- a/mllib/src/test/scala/org/apache/spark/ml/regression/LinearRegressionSuite.scala
+++ b/mllib/src/test/scala/org/apache/spark/ml/regression/LinearRegressionSuite.scala
@@ -461,8 +461,8 @@ class LinearRegressionSuite extends SparkFunSuite with MLlibTestSparkContext {
     assert(model.hasSummary)
 
     // Schema should be a superset of the input dataset
-    assert(model.summary.predictions.schema.fieldNames.toSet contains
-      dataset.schema.fieldNames.toSet ++ Set("prediction"))
+    assert((dataset.schema.fieldNames.toSet ++ Set("prediction")).subsetOf(
+      model.summary.predictions.schema.fieldNames.toSet))
 
     // Residuals in [[LinearRegressionResults]] should equal those manually computed
     val expectedResiduals = dataset.select("features", "label")

From d8baa4784e3c512211ff63e2378f75ce64a1e419 Mon Sep 17 00:00:00 2001
From: Holden Karau <holden@pigscanfly.ca>
Date: Fri, 18 Sep 2015 14:02:58 -0700
Subject: [PATCH 06/13] Use single element add operation to set

---
 .../org/apache/spark/ml/regression/LinearRegressionSuite.scala  | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mllib/src/test/scala/org/apache/spark/ml/regression/LinearRegressionSuite.scala b/mllib/src/test/scala/org/apache/spark/ml/regression/LinearRegressionSuite.scala
index 47a030221b6e1..8d64d1e1a8e26 100644
--- a/mllib/src/test/scala/org/apache/spark/ml/regression/LinearRegressionSuite.scala
+++ b/mllib/src/test/scala/org/apache/spark/ml/regression/LinearRegressionSuite.scala
@@ -461,7 +461,7 @@ class LinearRegressionSuite extends SparkFunSuite with MLlibTestSparkContext {
     assert(model.hasSummary)
 
     // Schema should be a superset of the input dataset
-    assert((dataset.schema.fieldNames.toSet ++ Set("prediction")).subsetOf(
+    assert((dataset.schema.fieldNames.toSet + "prediction").subsetOf(
       model.summary.predictions.schema.fieldNames.toSet))
 
     // Residuals in [[LinearRegressionResults]] should equal those manually computed

From f6448971f7a3c82da8d763078ec0606784599361 Mon Sep 17 00:00:00 2001
From: Holden Karau <holden@pigscanfly.ca>
Date: Thu, 24 Sep 2015 18:26:16 -0700
Subject: [PATCH 07/13] Partial progress, handle missing or invalid prediction
 columns in one pass (todo handle in evaluate call as well)

---
 .../spark/ml/regression/LinearRegression.scala     | 14 ++++++++++++--
 .../ml/regression/LinearRegressionSuite.scala      |  9 +++++++++
 2 files changed, 21 insertions(+), 2 deletions(-)

diff --git a/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala b/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala
index 6c7c3174dac26..b10fa0001f045 100644
--- a/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala
+++ b/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala
@@ -267,9 +267,19 @@ class LinearRegression(override val uid: String)
     if (handlePersistence) instances.unpersist()
 
     val model = copyValues(new LinearRegressionModel(uid, weights, intercept))
+    // Handle possible missing or invalid prediction columns
+    val predictionColOpt = get(predictionCol).orElse(getDefault(predictionCol)).filter(_ != "")
+    val (summaryModel, predictionColName) = predictionColOpt match {
+      case Some(p) => (model, p)
+      case None => {
+        val predictionColName = "prediction_" + java.util.UUID.randomUUID.toString()
+        (copyValues(model).setPredictionCol(predictionColName), predictionColName)
+      }
+    }
+
     val trainingSummary = new LinearRegressionTrainingSummary(
-      model.transform(dataset),
-      $(predictionCol),
+      summaryModel.transform(dataset),
+      predictionColName,
       $(labelCol),
       $(featuresCol),
       objectiveHistory)
diff --git a/mllib/src/test/scala/org/apache/spark/ml/regression/LinearRegressionSuite.scala b/mllib/src/test/scala/org/apache/spark/ml/regression/LinearRegressionSuite.scala
index ce380442edb4b..48bf03372803f 100644
--- a/mllib/src/test/scala/org/apache/spark/ml/regression/LinearRegressionSuite.scala
+++ b/mllib/src/test/scala/org/apache/spark/ml/regression/LinearRegressionSuite.scala
@@ -461,13 +461,22 @@ class LinearRegressionSuite extends SparkFunSuite with MLlibTestSparkContext {
   test("linear regression model training summary") {
     val trainer = new LinearRegression
     val model = trainer.fit(dataset)
+    val trainerNoPredictionCol = trainer.setPredictionCol("")
+    val modelNoPredictionCol = trainerNoPredictionCol.fit(dataset)
+
 
     // Training results for the model should be available
     assert(model.hasSummary)
+    assert(modelNoPredictionCol.hasSummary)
 
     // Schema should be a superset of the input dataset
     assert((dataset.schema.fieldNames.toSet + "prediction").subsetOf(
       model.summary.predictions.schema.fieldNames.toSet))
+    // Validate that we re-insert a prediction column for evaluation
+    val modelNoPredictionColFieldNames = modelNoPredictionCol.summary.predictions.schema.fieldNames
+    assert((dataset.schema.fieldNames.toSet).subsetOf(
+      modelNoPredictionColFieldNames.toSet))
+    assert(!modelNoPredictionColFieldNames.find(s => s.startsWith("prediction_")).isEmpty)
 
     // Residuals in [[LinearRegressionResults]] should equal those manually computed
     val expectedResiduals = dataset.select("features", "label")

From f8758a04054109136400fde56ab3de6ea165d3b1 Mon Sep 17 00:00:00 2001
From: Holden Karau <holden@pigscanfly.ca>
Date: Fri, 25 Sep 2015 16:35:37 -0700
Subject: [PATCH 08/13] Do similar logic in the different places where we may
 be missing a prediciton column and want to generate one

---
 .../ml/regression/LinearRegression.scala      | 25 ++++++++++++++++---
 1 file changed, 22 insertions(+), 3 deletions(-)

diff --git a/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala b/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala
index b10fa0001f045..fa0a7a2603f62 100644
--- a/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala
+++ b/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala
@@ -179,9 +179,19 @@ class LinearRegression(override val uid: String)
       val intercept = yMean
 
       val model = new LinearRegressionModel(uid, weights, intercept)
+      // Handle possible missing or invalid prediction columns
+      val predictionColOpt = get(predictionCol).orElse(getDefault(predictionCol)).filter(_ != "")
+      val (summaryModel, predictionColName) = predictionColOpt match {
+        case Some(p) => (model, p)
+        case None => {
+          val predictionColName = "prediction_" + java.util.UUID.randomUUID.toString()
+          (copyValues(model).setPredictionCol(predictionColName), predictionColName)
+        }
+      }
+
       val trainingSummary = new LinearRegressionTrainingSummary(
-        model.transform(dataset),
-        $(predictionCol),
+        summaryModel.transform(dataset),
+        predictionColName,
         $(labelCol),
         $(featuresCol),
         Array(0D))
@@ -331,7 +341,16 @@ class LinearRegressionModel private[ml] (
    */
   // TODO: decide on a good name before exposing to public API
   private[regression] def evaluate(dataset: DataFrame): LinearRegressionSummary = {
-    new LinearRegressionSummary(transform(dataset), $(predictionCol), $(labelCol))
+    // Handle possible missing or invalid prediction columns
+    val predictionColOpt = get(predictionCol).orElse(getDefault(predictionCol)).filter(_ != "")
+    val (summaryModel, predictionColName) = predictionColOpt match {
+      case Some(p) => (this, p)
+      case None => {
+        val predictionColName = "prediction_" + java.util.UUID.randomUUID.toString()
+        (copyValues(this).setPredictionCol(predictionColName), predictionColName)
+      }
+    }
+    new LinearRegressionSummary(summaryModel.transform(dataset), predictionColName, $(labelCol))
   }
 
   override protected def predict(features: Vector): Double = {

From b8208552df93f5dc513fd39f06fdf9907e72431c Mon Sep 17 00:00:00 2001
From: Holden Karau <holden@pigscanfly.ca>
Date: Wed, 7 Oct 2015 11:59:16 -0700
Subject: [PATCH 09/13] Simplify checking for column with exists

---
 .../org/apache/spark/ml/regression/LinearRegressionSuite.scala  | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/mllib/src/test/scala/org/apache/spark/ml/regression/LinearRegressionSuite.scala b/mllib/src/test/scala/org/apache/spark/ml/regression/LinearRegressionSuite.scala
index 48bf03372803f..8f6ee557e3d67 100644
--- a/mllib/src/test/scala/org/apache/spark/ml/regression/LinearRegressionSuite.scala
+++ b/mllib/src/test/scala/org/apache/spark/ml/regression/LinearRegressionSuite.scala
@@ -476,7 +476,7 @@ class LinearRegressionSuite extends SparkFunSuite with MLlibTestSparkContext {
     val modelNoPredictionColFieldNames = modelNoPredictionCol.summary.predictions.schema.fieldNames
     assert((dataset.schema.fieldNames.toSet).subsetOf(
       modelNoPredictionColFieldNames.toSet))
-    assert(!modelNoPredictionColFieldNames.find(s => s.startsWith("prediction_")).isEmpty)
+    assert(modelNoPredictionColFieldNames.exists(s => s.startsWith("prediction_")))
 
     // Residuals in [[LinearRegressionResults]] should equal those manually computed
     val expectedResiduals = dataset.select("features", "label")

From 033f6bae777ba26f5ba4f05df42c9e387ad07ba1 Mon Sep 17 00:00:00 2001
From: Holden Karau <holden@pigscanfly.ca>
Date: Wed, 7 Oct 2015 12:04:57 -0700
Subject: [PATCH 10/13] Start making the linear regression panda

---
 .../spark/ml/regression/LinearRegression.scala      | 13 +++++++++++++
 1 file changed, 13 insertions(+)

diff --git a/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala b/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala
index fa0a7a2603f62..22e909e797a4d 100644
--- a/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala
+++ b/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala
@@ -299,6 +299,19 @@ class LinearRegression(override val uid: String)
   override def copy(extra: ParamMap): LinearRegression = defaultCopy(extra)
 }
 
+object LinearRegression {
+  /**
+   * Takes the current linear regression model and an option representing the
+   * prediction column. If the prediction column is set returns the current
+   * model and prediction column, otherwise generates a new column and sets
+   * it as the prediction column on a new copy of the input model.
+   */
+  protected def findSummaryModelAndPredictionCol(model: LinearRegressionmodel,
+    predictionColOpt: Option[String]): (LinearRegressionModel, String) = {
+
+  }
+}
+
 /**
  * :: Experimental ::
  * Model produced by [[LinearRegression]].

From e91cdc49ca53305031088f9bcca2b8a3857db441 Mon Sep 17 00:00:00 2001
From: Holden Karau <holden@pigscanfly.ca>
Date: Wed, 7 Oct 2015 12:37:49 -0700
Subject: [PATCH 11/13] Simplify the dynamic prediciton column creation code to
 be unified in a signle place

---
 .../ml/regression/LinearRegression.scala      | 49 +++++++------------
 1 file changed, 17 insertions(+), 32 deletions(-)

diff --git a/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala b/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala
index 22e909e797a4d..08de9d5a2419a 100644
--- a/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala
+++ b/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala
@@ -180,14 +180,8 @@ class LinearRegression(override val uid: String)
 
       val model = new LinearRegressionModel(uid, weights, intercept)
       // Handle possible missing or invalid prediction columns
-      val predictionColOpt = get(predictionCol).orElse(getDefault(predictionCol)).filter(_ != "")
-      val (summaryModel, predictionColName) = predictionColOpt match {
-        case Some(p) => (model, p)
-        case None => {
-          val predictionColName = "prediction_" + java.util.UUID.randomUUID.toString()
-          (copyValues(model).setPredictionCol(predictionColName), predictionColName)
-        }
-      }
+      val predictionColOpt = get(predictionCol).orElse(getDefault(predictionCol))
+      val (summaryModel, predictionColName) = model.findSummaryModelAndPredictionCol()
 
       val trainingSummary = new LinearRegressionTrainingSummary(
         summaryModel.transform(dataset),
@@ -278,14 +272,8 @@ class LinearRegression(override val uid: String)
 
     val model = copyValues(new LinearRegressionModel(uid, weights, intercept))
     // Handle possible missing or invalid prediction columns
-    val predictionColOpt = get(predictionCol).orElse(getDefault(predictionCol)).filter(_ != "")
-    val (summaryModel, predictionColName) = predictionColOpt match {
-      case Some(p) => (model, p)
-      case None => {
-        val predictionColName = "prediction_" + java.util.UUID.randomUUID.toString()
-        (copyValues(model).setPredictionCol(predictionColName), predictionColName)
-      }
-    }
+    val predictionColOpt = get(predictionCol).orElse(getDefault(predictionCol))
+    val (summaryModel, predictionColName) = model.findSummaryModelAndPredictionCol()
 
     val trainingSummary = new LinearRegressionTrainingSummary(
       summaryModel.transform(dataset),
@@ -299,19 +287,6 @@ class LinearRegression(override val uid: String)
   override def copy(extra: ParamMap): LinearRegression = defaultCopy(extra)
 }
 
-object LinearRegression {
-  /**
-   * Takes the current linear regression model and an option representing the
-   * prediction column. If the prediction column is set returns the current
-   * model and prediction column, otherwise generates a new column and sets
-   * it as the prediction column on a new copy of the input model.
-   */
-  protected def findSummaryModelAndPredictionCol(model: LinearRegressionmodel,
-    predictionColOpt: Option[String]): (LinearRegressionModel, String) = {
-
-  }
-}
-
 /**
  * :: Experimental ::
  * Model produced by [[LinearRegression]].
@@ -355,17 +330,27 @@ class LinearRegressionModel private[ml] (
   // TODO: decide on a good name before exposing to public API
   private[regression] def evaluate(dataset: DataFrame): LinearRegressionSummary = {
     // Handle possible missing or invalid prediction columns
+    val (summaryModel, predictionColName) = findSummaryModelAndPredictionCol()
+    new LinearRegressionSummary(summaryModel.transform(dataset), predictionColName, $(labelCol))
+  }
+
+  /**
+   * If the prediction column is set returns the current model and prediction column,
+   * otherwise generates a new column and sets it as the prediction column on a new copy
+   * of the current model.
+   */
+  private[regression] def findSummaryModelAndPredictionCol(): (LinearRegressionModel, String) = {
     val predictionColOpt = get(predictionCol).orElse(getDefault(predictionCol)).filter(_ != "")
-    val (summaryModel, predictionColName) = predictionColOpt match {
+    predictionColOpt match {
       case Some(p) => (this, p)
       case None => {
         val predictionColName = "prediction_" + java.util.UUID.randomUUID.toString()
-        (copyValues(this).setPredictionCol(predictionColName), predictionColName)
+        (copy(ParamMap.empty).setPredictionCol(predictionColName), predictionColName)
       }
     }
-    new LinearRegressionSummary(summaryModel.transform(dataset), predictionColName, $(labelCol))
   }
 
+
   override protected def predict(features: Vector): Double = {
     dot(features, weights) + intercept
   }

From a247aa6e8a703dac9eb890d1f05661cf9243c44c Mon Sep 17 00:00:00 2001
From: Holden Karau <holden@pigscanfly.ca>
Date: Wed, 7 Oct 2015 13:35:40 -0700
Subject: [PATCH 12/13] Remove some more dead code and simplify match

---
 .../apache/spark/ml/regression/LinearRegression.scala    | 9 +++------
 1 file changed, 3 insertions(+), 6 deletions(-)

diff --git a/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala b/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala
index 08de9d5a2419a..889589cde3e7e 100644
--- a/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala
+++ b/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala
@@ -180,7 +180,6 @@ class LinearRegression(override val uid: String)
 
       val model = new LinearRegressionModel(uid, weights, intercept)
       // Handle possible missing or invalid prediction columns
-      val predictionColOpt = get(predictionCol).orElse(getDefault(predictionCol))
       val (summaryModel, predictionColName) = model.findSummaryModelAndPredictionCol()
 
       val trainingSummary = new LinearRegressionTrainingSummary(
@@ -272,7 +271,6 @@ class LinearRegression(override val uid: String)
 
     val model = copyValues(new LinearRegressionModel(uid, weights, intercept))
     // Handle possible missing or invalid prediction columns
-    val predictionColOpt = get(predictionCol).orElse(getDefault(predictionCol))
     val (summaryModel, predictionColName) = model.findSummaryModelAndPredictionCol()
 
     val trainingSummary = new LinearRegressionTrainingSummary(
@@ -340,13 +338,12 @@ class LinearRegressionModel private[ml] (
    * of the current model.
    */
   private[regression] def findSummaryModelAndPredictionCol(): (LinearRegressionModel, String) = {
-    val predictionColOpt = get(predictionCol).orElse(getDefault(predictionCol)).filter(_ != "")
-    predictionColOpt match {
-      case Some(p) => (this, p)
-      case None => {
+    $(predictionCol) match {
+      case "" => {
         val predictionColName = "prediction_" + java.util.UUID.randomUUID.toString()
         (copy(ParamMap.empty).setPredictionCol(predictionColName), predictionColName)
       }
+      case p => (this, p)
     }
   }
 

From 7fb3b1c1bfeba2b8543a5db5fe51a37e3196a254 Mon Sep 17 00:00:00 2001
From: Holden Karau <holden@pigscanfly.ca>
Date: Wed, 7 Oct 2015 17:49:53 -0700
Subject: [PATCH 13/13] Remove extra braces (CR feedback)

---
 .../org/apache/spark/ml/regression/LinearRegression.scala      | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala b/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala
index b1a2496006791..dd09667ef5a0f 100644
--- a/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala
+++ b/mllib/src/main/scala/org/apache/spark/ml/regression/LinearRegression.scala
@@ -334,10 +334,9 @@ class LinearRegressionModel private[ml] (
    */
   private[regression] def findSummaryModelAndPredictionCol(): (LinearRegressionModel, String) = {
     $(predictionCol) match {
-      case "" => {
+      case "" =>
         val predictionColName = "prediction_" + java.util.UUID.randomUUID.toString()
         (copy(ParamMap.empty).setPredictionCol(predictionColName), predictionColName)
-      }
       case p => (this, p)
     }
   }