Spark数据帧否定过滤条件

Question

我正在尝试在DataFrame上应用过滤条件的否定,如下所示.

!(`Ship Mode` LIKE '%Truck%')

这是在下面抛出一个例外

Exception in thread "main" java.lang.RuntimeException: [1.3] failure: identifier expected

(!(`Ship Mode` LIKE '%Truck%'))
  ^
    at scala.sys.package$.error(package.scala:27)
    at org.apache.spark.sql.catalyst.SqlParser.parseExpression(SqlParser.scala:47)
    at org.apache.spark.sql.DataFrame.filter(DataFrame.scala:748)
    at Main.main(Main.java:73)

在MySQL中,相同类型的负过滤条件正常工作.如下请见

mysql> select count(*) from audit_log where !(operation like '%Log%' or operation like '%Proj%');
+----------+
| count(*) |
+----------+
|      129 |
+----------+
1 row in set (0.05 sec)

任何人都可以告诉我,如果计划在将来的版本中修复Spark DataFrames,或者我应该提出JIRA.

Answer 1

它看起来像你使用普通的SQLContext地方!是不支持:

import org.apache.spark.sql.SQLContext
val sqlContext = new SQLContext(sc)

val data = Seq(("a", 1, 3), ("b", 2, 6), ("c", -1, 2))

val df = sqlContext.createDataFrame(data).toDF("x1", "x2", "x3") 
df.registerTempTable("df")

sqlContext.sql("SELECT * FROM df WHERE ! (x2 >  2 OR x3 < 4)").show

// java.lang.RuntimeException: [1.25] failure: identifier expected
//
// SELECT * FROM df WHERE ! (x2 >  2 OR x3 < 4)
//                         ^

它可以很容易地替换为NOT:

sqlContext.sql("SELECT * FROM df WHERE NOT (x2 >  2 OR x3 < 4)").show

// +---+---+---+
// | x1| x2| x3|
// +---+---+---+
// |  b|  2|  6|
// +---+---+---+

如果您仍想使用,!您应该使用HiveContext:

import org.apache.spark.sql.hive.HiveContext

val hiveContext = new HiveContext(sc)

val df1 = hiveContext.createDataFrame(data).toDF("x1", "x2", "x3")
df1.registerTempTable("df")

hiveContext.sql("SELECT * FROM df WHERE ! (x2 >  2 OR x3 < 4)").show

// +---+---+---+
// | x1| x2| x3|
// +---+---+---+
// |  b|  2|  6|
// +---+---+---+