diff --git a/python/pyspark/sql/readwriter.py b/python/pyspark/sql/readwriter.py index 7fd7583972c7586d69bbf598b9d9ecbdb4951a8a..5cb186016e3db5c89f060e33a53e34a54d61c265 100644 --- a/python/pyspark/sql/readwriter.py +++ b/python/pyspark/sql/readwriter.py @@ -358,7 +358,7 @@ class DataFrameReader(object): >>> df = spark.read.csv('python/test_support/sql/ages.csv') >>> df.dtypes - [('C0', 'string'), ('C1', 'string')] + [('_c0', 'string'), ('_c1', 'string')] """ if schema is not None: self.schema(schema) diff --git a/sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/csv/DefaultSource.scala b/sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/csv/DefaultSource.scala index 948fac0d58ad759b16657a70eb02229866b124e9..f47ed76cba76520b661bd2a2f6b5f0852d165f6b 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/csv/DefaultSource.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/csv/DefaultSource.scala @@ -61,9 +61,11 @@ class DefaultSource extends FileFormat with DataSourceRegister { val firstRow = new LineCsvReader(csvOptions).parseLine(firstLine) val header = if (csvOptions.headerFlag) { - firstRow + firstRow.zipWithIndex.map { case (value, index) => + if (value == null || value.isEmpty || value == csvOptions.nullValue) s"_c$index" else value + } } else { - firstRow.zipWithIndex.map { case (value, index) => s"C$index" } + firstRow.zipWithIndex.map { case (value, index) => s"_c$index" } } val parsedRdd = tokenRdd(sparkSession, csvOptions, header, paths) diff --git a/sql/core/src/test/resources/cars-blank-column-name.csv b/sql/core/src/test/resources/cars-blank-column-name.csv new file mode 100644 index 0000000000000000000000000000000000000000..0b804b1614d60866e4a52e6970c76e6c8c3addab --- /dev/null +++ b/sql/core/src/test/resources/cars-blank-column-name.csv @@ -0,0 +1,3 @@ +"",,make,customer,comment +2012,"Tesla","S","bill","blank" +2013,"Tesla","S","c","something" diff --git a/sql/core/src/test/scala/org/apache/spark/sql/execution/datasources/csv/CSVSuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/execution/datasources/csv/CSVSuite.scala index b6cdc8cfabe3ca301bb9c48ea648eef244b4a4e1..ae91e0f606eccf46756e913e37419671bf35e062 100644 --- a/sql/core/src/test/scala/org/apache/spark/sql/execution/datasources/csv/CSVSuite.scala +++ b/sql/core/src/test/scala/org/apache/spark/sql/execution/datasources/csv/CSVSuite.scala @@ -38,6 +38,7 @@ class CSVSuite extends QueryTest with SharedSQLContext with SQLTestUtils { private val carsAltFile = "cars-alternative.csv" private val carsUnbalancedQuotesFile = "cars-unbalanced-quotes.csv" private val carsNullFile = "cars-null.csv" + private val carsBlankColName = "cars-blank-column-name.csv" private val emptyFile = "empty.csv" private val commentsFile = "comments.csv" private val disableCommentsFile = "disable_comments.csv" @@ -71,14 +72,14 @@ class CSVSuite extends QueryTest with SharedSQLContext with SQLTestUtils { if (withHeader) { assert(df.schema.fieldNames === Array("year", "make", "model", "comment", "blank")) } else { - assert(df.schema.fieldNames === Array("C0", "C1", "C2", "C3", "C4")) + assert(df.schema.fieldNames === Array("_c0", "_c1", "_c2", "_c3", "_c4")) } } if (checkValues) { val yearValues = List("2012", "1997", "2015") val actualYears = if (!withHeader) "year" :: yearValues else yearValues - val years = if (withHeader) df.select("year").collect() else df.select("C0").collect() + val years = if (withHeader) df.select("year").collect() else df.select("_c0").collect() years.zipWithIndex.foreach { case (year, index) => if (checkTypes) { @@ -224,6 +225,17 @@ class CSVSuite extends QueryTest with SharedSQLContext with SQLTestUtils { assert(cars.select("year").collect().size === 2) } + test("test for blank column names on read and select columns") { + val cars = spark.read + .format("csv") + .options(Map("header" -> "true", "inferSchema" -> "true")) + .load(testFile(carsBlankColName)) + + assert(cars.select("customer").collect().size == 2) + assert(cars.select("_c0").collect().size == 2) + assert(cars.select("_c1").collect().size == 2) + } + test("test for FAILFAST parsing mode") { val exception = intercept[SparkException]{ spark.read