为什么＆＃34; \\ s＆＃34;用＆＃34; rlike＆＃34;在Spark SQL中不起作用？

Question

＆＃34; \ s＆＃34;不起作用，如下面的代码：

scala> case class Item(id: Int, name:String)
defined class Item

scala> val df = Seq((1, "hello"), (2, " "))
df: Seq[(Int, String)] = List((1,hello), (2," "))

scala> df.toDF("id", "name").as[Item].filter("name rlike '\\s'").show()
+---+----+
| id|name|
+---+----+
+---+----+

Answer 1

作为LanguageManual UDF described，rlike需要双重转义模式\\s。

在火花壳中，Scala逃脱了一次。因此需要更多\：

scala> df.toDF("id", "name").as[Item].filter("name rlike '\\\\s'").show()
+---+----+
| id|name|
+---+----+
|  2|    |
+---+----+

scala> df.toDF("id", "name").as[Item].filter("name rlike '\\\\w'").show()
+---+-----+
| id| name|
+---+-----+
|  1|hello|
+---+-----+

或者使用没有转义的字符串，例如：

scala> val er = """name rlike '\\s'"""
er: String = name rlike '\\s'

scala> df.toDF("id", "name").as[Item].filter(er).show()
+---+----+
| id|name|
+---+----+
|  2|    |
+---+----+