Stable key order for map vectorizers (#88)

salesforce · Aug 24, 2018 · fb38b67 · fb38b67
1 parent 7135926
commit fb38b67
Show file tree

Hide file tree

Showing 7 changed files with 214 additions and 258 deletions.
diff --git a/core/src/main/scala/com/salesforce/op/stages/impl/feature/OPMapVectorizer.scala b/core/src/main/scala/com/salesforce/op/stages/impl/feature/OPMapVectorizer.scala
@@ -84,7 +84,6 @@ abstract class OPMapVectorizer[A, T <: OPMap[A]]
     val meta = if ($(trackNulls)) makeVectorMetaWithNullIndicators(allKeys) else makeVectorMetadata(allKeys)
     setMetadata(meta.toMetadata)
 
-
     val args = OPMapVectorizerModelArgs(
       allKeys = allKeys,
       fillByKey = fillByKey(dataset),
@@ -124,14 +123,14 @@ class IntegralMapVectorizer[T <: OPMap[Long]](uid: String = UID[IntegralMapVecto
   def setFillWithMode(shouldFill: Boolean): this.type = set(withConstant, !shouldFill)
 
   override def fillByKey(dataset: Dataset[Seq[T#Value]]): Seq[Map[String, Double]] = {
-    val size = getInputFeatures().length
-    val cleanedData = dataset.map(_.map(
-      cleanMap(_, shouldCleanKey = $(cleanKeys), shouldCleanValue = shouldCleanValues)
-    ))
-
     if ($(withConstant)) Seq.empty
     else {
+      val size = getInputFeatures().length
       val modeAggr = SequenceAggregators.ModeSeqMapLong(size = size)
+      val shouldCleanKeys = $(cleanKeys)
+      val cleanedData = dataset.map(_.map(
+        cleanMap(_, shouldCleanKey = shouldCleanKeys, shouldCleanValue = shouldCleanValues)
+      ))
       cleanedData.select(modeAggr.toColumn).first()
     }.map(convertFn)
   }
@@ -203,8 +202,6 @@ class TextMapHashingVectorizer[T <: OPMap[String]]
   def setHashSpaceStrategy(v: HashSpaceStrategy): this.type = set(hashSpaceStrategy, v.entryName)
   def getHashSpaceStrategy: HashSpaceStrategy = HashSpaceStrategy.withNameInsensitive($(hashSpaceStrategy))
 
-  def getFillByKey(dataset: Dataset[Seq[T#Value]]): Seq[Map[String, Double]] = Seq.empty
-
   def makeModel(args: OPMapVectorizerModelArgs, operationName: String, uid: String): OPMapVectorizerModel[String, T] =
     new TextMapHashingVectorizerModel[T](
       args = args.copy(shouldCleanValues = $(cleanText)),
@@ -230,14 +227,14 @@ class RealMapVectorizer[T <: OPMap[Double]](uid: String = UID[RealMapVectorizer[
   def setFillWithMean(shouldFill: Boolean): this.type = set(withConstant, !shouldFill)
 
   override def fillByKey(dataset: Dataset[Seq[T#Value]]): Seq[Map[String, Double]] = {
-    val size = getInputFeatures().length
-    val cleanedData = dataset.map(_.map(
-      cleanMap(_, shouldCleanKey = $(cleanKeys), shouldCleanValue = shouldCleanValues)
-    ))
-
     if ($(withConstant)) Seq.empty
     else {
+      val size = getInputFeatures().length
       val meanAggr = SequenceAggregators.MeanSeqMapDouble(size = size)
+      val shouldCleanKeys = $(cleanKeys)
+      val cleanedData = dataset.map(_.map(
+        cleanMap(_, shouldCleanKey = shouldCleanKeys, shouldCleanValue = shouldCleanValues)
+      ))
       cleanedData.select(meanAggr.toColumn).first()
     }
   }
@@ -282,7 +279,7 @@ trait MapVectorizerFuns[A, T <: OPMap[A]] extends VectorizerDefaults with MapPiv
     in.map(_.map(kb => filterKeys(kb, shouldCleanKey = shouldCleanKeys, shouldCleanValue = shouldCleanValues).keySet))
       .select(sumAggr.toColumn)
       .first()
-      .map(_.toSeq)
+      .map(_.toSeq.sorted)
   }
 
   protected def makeVectorMetadata(allKeys: Seq[Seq[String]]): OpVectorMetadata = {

diff --git a/core/src/main/scala/com/salesforce/op/stages/impl/feature/SmartTextMapVectorizer.scala b/core/src/main/scala/com/salesforce/op/stages/impl/feature/SmartTextMapVectorizer.scala
@@ -121,7 +121,7 @@ class SmartTextMapVectorizer[T <: OPMap[String]]
 
   def makeSmartTextMapVectorizerModelArgs(aggregatedStats: Array[TextMapStats]): SmartTextMapVectorizerModelArgs = {
     val maxCard = $(maxCardinality)
-    val minSup = ${minSupport}
+    val minSup = $(minSupport)
     val shouldCleanKeys = $(cleanKeys)
     val shouldCleanValues = $(cleanText)
     val shouldTrackNulls = $(trackNulls)

diff --git a/core/src/test/scala/com/salesforce/op/stages/impl/feature/BinaryMapVectorizerTest.scala b/core/src/test/scala/com/salesforce/op/stages/impl/feature/BinaryMapVectorizerTest.scala
@@ -55,9 +55,9 @@ class BinaryMapVectorizerTest
 
   val estimator = new BinaryMapVectorizer().setTrackNulls(false).setCleanKeys(true).setInput(m1, m2)
 
-  val expectedResult: Seq[OPVector] = Seq(
+  val expectedResult = Seq(
     Vectors.sparse(6, Array(1), Array(1.0)),
-    Vectors.sparse(6, Array(4, 5), Array(1.0, 1.0)),
+    Vectors.sparse(6, Array(3, 4), Array(1.0, 1.0)),
     Vectors.sparse(6, Array(), Array())
   ).map(_.toOPVector)
 
@@ -68,7 +68,7 @@ class BinaryMapVectorizerTest
     val expectedMeta = TestOpVectorMetadataBuilder(
       estimator,
       m1 -> List(IndColWithGroup(None, "A"), IndColWithGroup(None, "B"), IndColWithGroup(None, "C")),
-      m2 -> List(IndColWithGroup(None, "Z"), IndColWithGroup(None, "Y"), IndColWithGroup(None, "X"))
+      m2 -> List(IndColWithGroup(None, "X"), IndColWithGroup(None, "Y"), IndColWithGroup(None, "Z"))
     )
 
     transformed.collect(vector) shouldBe expectedResult
@@ -82,8 +82,8 @@ class BinaryMapVectorizerTest
     val transformed = estimator.setTrackNulls(true).fit(inputData).transform(inputData)
     val vector = estimator.getOutput()
     val expected = Array(
-      Vectors.sparse(12, Array(2, 5, 9, 11), Array(1.0, 1.0, 1.0, 1.0)),
-      Vectors.sparse(12, Array(1, 3, 7, 8, 10), Array(1.0, 1.0, 1.0, 1.0, 1.0)),
+      Vectors.sparse(12, Array(2, 5, 7, 9), Array(1.0, 1.0, 1.0, 1.0)),
+      Vectors.sparse(12, Array(1, 3, 6, 8, 11), Array(1.0, 1.0, 1.0, 1.0, 1.0)),
       Vectors.sparse(12, Array(1, 3, 5, 7, 9, 11), Array(1.0, 1.0, 1.0, 1.0, 1.0, 1.0))
     ).map(_.toOPVector)
 
@@ -93,9 +93,9 @@ class BinaryMapVectorizerTest
       m1 -> List(IndColWithGroup(None, "A"), IndColWithGroup(nullIndicatorValue, "A"),
         IndColWithGroup(None, "B"), IndColWithGroup(nullIndicatorValue, "B"),
         IndColWithGroup(None, "C"), IndColWithGroup(nullIndicatorValue, "C")),
-      m2 -> List(IndColWithGroup(None, "Z"), IndColWithGroup(nullIndicatorValue, "Z"),
+      m2 -> List(IndColWithGroup(None, "X"), IndColWithGroup(nullIndicatorValue, "X"),
         IndColWithGroup(None, "Y"), IndColWithGroup(nullIndicatorValue, "Y"),
-        IndColWithGroup(None, "X"), IndColWithGroup(nullIndicatorValue, "X"))
+        IndColWithGroup(None, "Z"), IndColWithGroup(nullIndicatorValue, "Z"))
     )
 
     transformed.collect(vector) shouldBe expected