使用 Option 作为输入参数定义 Spark scala UDF答案

【问题标题】：Define a Spark scala UDF with Option as input param使用 Option 作为输入参数定义 Spark scala UDF
【发布时间】：2019-01-16 11:20:40
【问题描述】：

编写以下 UDF 旨在使其处理未定义参数的情况。下面是代码：

val addTimeFromCols: UserDefinedFunction = udf((year: String, month: String, day: String, hour: String) => {
      Option(hour) match {
        case None    => (List(year, month, day).mkString(DASH_SEP)).concat(SPACE).concat(defaultHour)
        case Some(x) => (List(year, month, day).mkString(DASH_SEP)).concat(SPACE).concat(hour)
      }
    })

 def addTimestampFromFileCols(): DataFrame = df
  .withColumn(COLUMN_TS, addTimeFromCols(col(COLUMN_YEAR), col(COLUMN_MONTH), col(COLUMN_DAY), col(COLUMN_HOUR)).cast(TimestampType))

我的目标是使此功能适用于所有用例（具有 HOUR 列的数据框和其他没有此列的数据框，在这种情况下，我默认定义一个值。不幸的是，当我再次测试没有列的数据框我收到以下错误：

cannot resolve '`HOUR`' given input columns

请知道如何解决这个问题

【问题讨论】：

标签： scala apache-spark user-defined-functions option

【解决方案1】：

如果该列不存在，则必须通过 lit() 函数提供默认值，否则会抛出错误。以下对我有用

scala> defaultHour
res77: String = 00

scala> :paste
// Entering paste mode (ctrl-D to finish)

def addTimestampFromFileCols(df:DataFrame) =
{
val hr = if( df.columns.contains("hour") ) col(COLUMN_HOUR) else lit(defaultHour)
df.withColumn(COLUMN_TS, addTimeFromCols(col(COLUMN_YEAR), col(COLUMN_MONTH), col(COLUMN_DAY), hr).cast(TimestampType))
}

// Exiting paste mode, now interpreting.

addTimestampFromFileCols: (df: org.apache.spark.sql.DataFrame)org.apache.spark.sql.DataFrame

scala>

+ve 案例

scala> val df = Seq(("2019","01","10","09")).toDF("year","month","day","hour")
df: org.apache.spark.sql.DataFrame = [year: string, month: string ... 2 more fields]

scala> addTimestampFromFileCols(df).show(false)
+----+-----+---+----+-------------------+
|year|month|day|hour|tstamp             |
+----+-----+---+----+-------------------+
|2019|01   |10 |09  |2019-01-10 09:00:00|
+----+-----+---+----+-------------------+

-ve 案例

scala> val df = Seq(("2019","01","10")).toDF("year","month","day")
df: org.apache.spark.sql.DataFrame = [year: string, month: string ... 1 more field]

scala> addTimestampFromFileCols(df).show(false)
+----+-----+---+-------------------+
|year|month|day|tstamp             |
+----+-----+---+-------------------+
|2019|01   |10 |2019-01-10 00:00:00|
+----+-----+---+-------------------+

scala>

【讨论】：