[SPARK-21723][ML] Fix writing LibSVM (key not found: numFeatures)

## What changes were proposed in this pull request?

Check the option "numFeatures" only when reading LibSVM, not when writing. When writing, Spark was raising an exception. After the change it will ignore the option completely. liancheng HyukjinKwon

(Maybe the usage should be forbidden when writing, in a major version change?).

## How was this patch tested?

Manual test, that loading and writing LibSVM files work fine, both with and without the numFeatures option.

Please review http://spark.apache.org/contributing.html before opening a pull request.

Author: Jan Vrsovsky <jan.vrsovsky@firma.seznam.cz>

Closes #18872 from ProtD/master.

mllib/src/main/scala/org/apache/spark/ml/source/libsvm/LibSVMRelation.scala

@@ -76,12 +76,12 @@ private[libsvm] class LibSVMFileFormat
 
   override def toString: String = "LibSVM"
 
-  private def verifySchema(dataSchema: StructType): Unit = {
+  private def verifySchema(dataSchema: StructType, forWriting: Boolean): Unit = {
     if (
       dataSchema.size != 2 ||
         !dataSchema(0).dataType.sameType(DataTypes.DoubleType) ||
         !dataSchema(1).dataType.sameType(new VectorUDT()) ||
-        !(dataSchema(1).metadata.getLong(LibSVMOptions.NUM_FEATURES).toInt > 0)
+        !(forWriting || dataSchema(1).metadata.getLong(LibSVMOptions.NUM_FEATURES).toInt > 0)
     ) {
       throw new IOException(s"Illegal schema for libsvm data, schema=$dataSchema")
     }
@@ -119,7 +119,7 @@ private[libsvm] class LibSVMFileFormat
       job: Job,
       options: Map[String, String],
       dataSchema: StructType): OutputWriterFactory = {
-    verifySchema(dataSchema)
+    verifySchema(dataSchema, true)
     new OutputWriterFactory {
       override def newInstance(
           path: String,
@@ -142,7 +142,7 @@ private[libsvm] class LibSVMFileFormat
       filters: Seq[Filter],
       options: Map[String, String],
       hadoopConf: Configuration): (PartitionedFile) => Iterator[InternalRow] = {
-    verifySchema(dataSchema)
+    verifySchema(dataSchema, false)
     val numFeatures = dataSchema("features").metadata.getLong(LibSVMOptions.NUM_FEATURES).toInt
     assert(numFeatures > 0)
 

mllib/src/test/scala/org/apache/spark/ml/source/libsvm/LibSVMRelationSuite.scala

@@ -19,13 +19,16 @@ package org.apache.spark.ml.source.libsvm
 
 import java.io.{File, IOException}
 import java.nio.charset.StandardCharsets
+import java.util.List
 
 import com.google.common.io.Files
 
 import org.apache.spark.SparkFunSuite
 import org.apache.spark.ml.linalg.{DenseVector, SparseVector, Vector, Vectors}
+import org.apache.spark.ml.linalg.SQLDataTypes.VectorType
 import org.apache.spark.mllib.util.MLlibTestSparkContext
 import org.apache.spark.sql.{Row, SaveMode}
+import org.apache.spark.sql.types.{DoubleType, StructField, StructType}
 import org.apache.spark.util.Utils
 
 
@@ -44,14 +47,14 @@ class LibSVMRelationSuite extends SparkFunSuite with MLlibTestSparkContext {
       """
         |0 2:4.0 4:5.0 6:6.0
       """.stripMargin
-    val dir = Utils.createDirectory(tempDir.getCanonicalPath, "data")
+    val dir = Utils.createTempDir()
     val succ = new File(dir, "_SUCCESS")
     val file0 = new File(dir, "part-00000")
     val file1 = new File(dir, "part-00001")
     Files.write("", succ, StandardCharsets.UTF_8)
     Files.write(lines0, file0, StandardCharsets.UTF_8)
     Files.write(lines1, file1, StandardCharsets.UTF_8)
-    path = dir.toURI.toString
+    path = dir.getPath
   }
 
   override def afterAll(): Unit = {
@@ -108,12 +111,12 @@ class LibSVMRelationSuite extends SparkFunSuite with MLlibTestSparkContext {
 
   test("write libsvm data and read it again") {
     val df = spark.read.format("libsvm").load(path)
-    val tempDir2 = new File(tempDir, "read_write_test")
-    val writepath = tempDir2.toURI.toString
+    val writePath = Utils.createTempDir().getPath
+
     // TODO: Remove requirement to coalesce by supporting multiple reads.
-    df.coalesce(1).write.format("libsvm").mode(SaveMode.Overwrite).save(writepath)
+    df.coalesce(1).write.format("libsvm").mode(SaveMode.Overwrite).save(writePath)
 
-    val df2 = spark.read.format("libsvm").load(writepath)
+    val df2 = spark.read.format("libsvm").load(writePath)
     val row1 = df2.first()
     val v = row1.getAs[SparseVector](1)
     assert(v == Vectors.sparse(6, Seq((0, 1.0), (2, 2.0), (4, 3.0))))
@@ -126,6 +129,27 @@ class LibSVMRelationSuite extends SparkFunSuite with MLlibTestSparkContext {
     }
   }
 
+  test("write libsvm data from scratch and read it again") {
+    val rawData = new java.util.ArrayList[Row]()
+    rawData.add(Row(1.0, Vectors.sparse(3, Seq((0, 2.0), (1, 3.0)))))
+    rawData.add(Row(4.0, Vectors.sparse(3, Seq((0, 5.0), (2, 6.0)))))
+
+    val struct = StructType(
+      StructField("labelFoo", DoubleType, false) ::
+      StructField("featuresBar", VectorType, false) :: Nil
+    )
+    val df = spark.sqlContext.createDataFrame(rawData, struct)
+
+    val writePath = Utils.createTempDir().getPath
+
+    df.coalesce(1).write.format("libsvm").mode(SaveMode.Overwrite).save(writePath)
+
+    val df2 = spark.read.format("libsvm").load(writePath)
+    val row1 = df2.first()
+    val v = row1.getAs[SparseVector](1)
+    assert(v == Vectors.sparse(3, Seq((0, 2.0), (1, 3.0))))
+  }
+
   test("select features from libsvm relation") {
     val df = spark.read.format("libsvm").load(path)
     df.select("features").rdd.map { case Row(d: Vector) => d }.first