[SPARK-10835][ML] Word2Vec should accept non-null string array, in addition to existing null string array

## What changes were proposed in this pull request?

To match Tokenizer and for compatibility with Word2Vec, output a nullable string array type in NGram

## How was this patch tested?

Jenkins tests.

Author: Sean Owen <sowen@cloudera.com>

Closes #15179 from srowen/SPARK-10835.

mllib/src/main/scala/org/apache/spark/ml/feature/Word2Vec.scala

@@ -108,7 +108,8 @@ private[feature] trait Word2VecBase extends Params
    * Validate and transform the input schema.
    */
   protected def validateAndTransformSchema(schema: StructType): StructType = {
-    SchemaUtils.checkColumnType(schema, $(inputCol), new ArrayType(StringType, true))
+    val typeCandidates = List(new ArrayType(StringType, true), new ArrayType(StringType, false))
+    SchemaUtils.checkColumnTypes(schema, $(inputCol), typeCandidates)
     SchemaUtils.appendColumn(schema, $(outputCol), new VectorUDT)
   }
 }

mllib/src/test/scala/org/apache/spark/ml/feature/Word2VecSuite.scala

@@ -207,5 +207,26 @@ class Word2VecSuite extends SparkFunSuite with MLlibTestSparkContext with Defaul
     val newInstance = testDefaultReadWrite(instance)
     assert(newInstance.getVectors.collect() === instance.getVectors.collect())
   }
+
+  test("Word2Vec works with input that is non-nullable (NGram)") {
+    val spark = this.spark
+    import spark.implicits._
+
+    val sentence = "a q s t q s t b b b s t m s t m q "
+    val docDF = sc.parallelize(Seq(sentence, sentence)).map(_.split(" ")).toDF("text")
+
+    val ngram = new NGram().setN(2).setInputCol("text").setOutputCol("ngrams")
+    val ngramDF = ngram.transform(docDF)
+
+    val model = new Word2Vec()
+      .setVectorSize(2)
+      .setInputCol("ngrams")
+      .setOutputCol("result")
+      .fit(ngramDF)
+
+    // Just test that this transformation succeeds
+    model.transform(ngramDF).collect()
+  }
+
 }