Spark的Dataset操作(四)-其他单表操作

最新推荐文章于 2022-04-29 10:59:16 发布

野男孩

最新推荐文章于 2022-04-29 10:59:16 发布

阅读量1.2w

点赞数 1

分类专栏： Spark Spark spark-sql dataset Spark SQL专栏文章标签： spark na drop withcolumn fill

本文链接：https://blog.csdn.net/coding_hello/article/details/75211995

版权

Spark 同时被 3 个专栏收录

19 篇文章 6 订阅

订阅专栏

Spark spark-sql dataset

10 篇文章 3 订阅

订阅专栏

Spark SQL专栏

8 篇文章 3 订阅

订阅专栏

Spark的Dataset操作(四)-其他单表操作

还有些杂七杂八的小用法没有提到，比如添加列，删除列，NA值处理之类的，就在这里大概列一下吧。

数据集还是之前的那个吧：

scala> val df = spark.createDataset(Seq(
  ("aaa",1,2),("bbb",3,4),("ccc",3,5),("bbb",4, 6))   ).toDF("key1","key2","key3")
df: org.apache.spark.sql.DataFrame = [key1: string, key2: int ... 1 more field]

scala> df.printSchema
root
 |-- key1: string (nullable = true)
 |-- key2: integer (nullable = false)
 |-- key3: integer (nullable = false)

scala> df.show
+----+----+----+
|key1|key2|key3|
+----+----+----+
| aaa|   1|   2|
| bbb|   3|   4|
| ccc|   3|   5|
| bbb|   4|   6|
+----+----+----+

下面来添加一列，可以是字符串类型，整型；可以是常量或者是对当前已有的某列的变换，都行：

/* 
新增字符串类型的列key_4，都初始化为new_str_col，注意这里的lit()函数 
还有人发消息说这个lit()，补一下说明吧。这里的lit()是spark自带的函数，需要import org.apache.spark.sql.functions
def lit(literal: Any): Column Creates a Column of literal value. The passed in object is returned
directly if it is already a Column. If the object is a Scala Symbol, it is converted into a Column
also. Otherwise, a new Column is created to represent the literal value. 
Since 1.3.0
*/
scala> val df_1 = df.withColumn("key4", lit("new_str_col"))
df_1: org.apache.spark.sql.DataFrame = [key1: string, key2: int ... 2 more fields]

scala> df_1.printSchema
root
 |-- key1: string (nullable = true)
 |-- key2: integer (nullable = false)
 |-- key3: integer (nullable = false)
 |-- key4: string (nullable = false)

scala> df_1.show
+----+----+----+-----------+
|key1|key2|key3|       key4|
+----+----+----+-----------+
| aaa|   1|   2|new_str_col|
| bbb|   3|   4|new_str_col|
| ccc|   3|   5|new_str_col|
| bbb|   4|   6|new_str_col|
+----+----+----+-----------+

/* 
同样的，新增Int类型的列key5，都初始化为1024 
*/
scala> val df_2 = df_1.withColumn("key5", lit(1024))
df_2: org.apache.spark.sql.DataFrame = [key1: string, key2: int ... 3 more fields]

scala> df_2.printSchema
root
 |-- key1: string (nullable = true)
 |-- key2: integer (nullable = false)
 |-- key3: integer (nullable = false)
 |-- key4: string (nullable = false)
 |-- key5: integer (nullable = false)

scala> df_2.show
+----+----+----+-----------+-----+
|key1|key2|key3|       key4|key5|
+----+----+----+-----------+-----+
| aaa|   1|   2|new_str_col| 1024|
| bbb|   3|   4|new_str_col| 1024|
| ccc|   3|   5|new_str_col| 1024|
| bbb|   4|   6|new_str_col| 1024|
+----+----+----+-----------+-----+

/*
再来个不是常量的新增列key6 = key5 * 2
*/
scala> val df_3 = df_2.withColumn("key6", $"key5"*2)
df_3: org.apache.spark.sql.DataFrame = [key1: string, key2: int ... 4 more fields]

scala> df_3.show
+----+----+----+-----------+----+----+
|key1|key2|key3|       key4|key5|key6|
+----+----+----+-----------+----+----+
| aaa|   1|   2|new_str_col|1024|2048|
| bbb|   3|   4|new_str_col|1024|2048|
| ccc|   3|   5|new_str_col|1024|2048|
| bbb|   4|   6|new_str_col|1024|2048|
+----+----+----+-----------+----+----+

/*
这次是用的expr()函数
*/
scala> val df_4 = df_2.withColumn("key6", expr("key5 * 4"))
df_4: org.apache.spark.sql.DataFrame = [key1: string, key2: int ... 4 more fields]

scala> df_4.show
+----+----+----+-----------+----+----+
|key1|key2|key3|       key4|key5|key6|
+----+----+----+-----------+----+----+
| aaa|   1|   2|new_str_col|1024|4096|
| bbb|   3|   4|new_str_col|1024|4096|
| ccc|   3|   5|new_str_col|1024|4096|
| bbb|   4|   6|new_str_col|1024|4096|
+----+----+----+-----------+----+----+

删除列就比较简单了，指定列名就好了

/*
删除列key5
*/
scala> val df_5 = df_4.drop("key5")
df_5: org.apache.spark.sql.DataFrame = [key1: string, key2: int ... 3 more fields]

scala> df_4.printSchema
root
 |-- key1: string (nullable = true)
 |-- key2: integer (nullable = false)
 |-- key3: integer (nullable = false)
 |-- key4: string (nullable = false)
 |-- key5: integer (nullable = false)
 |-- key6: integer (nullable = false)

scala> df_5.printSchema
root
 |-- key1: string (nullable = true)
 |-- key2: integer (nullable = false)
 |-- key3: integer (nullable = false)
 |-- key4: string (nullable = false)
 |-- key6: integer (nullable = false)

scala> df_5.show
+----+----+----+-----------+----+
|key1|key2|key3|       key4|key6|
+----+----+----+-----------+----+
| aaa|   1|   2|new_str_col|4096|
| bbb|   3|   4|new_str_col|4096|
| ccc|   3|   5|new_str_col|4096|
| bbb|   4|   6|new_str_col|4096|
+----+----+----+-----------+----+

/*
可以一次删除多列key4和key6
*/
scala> val df_6 = df_5.drop("key4", "key6")
df_6: org.apache.spark.sql.DataFrame = [key1: string, key2: int ... 1 more field]

/* 这里的columns函数以数组形式返回所有列名 */
scala> df_6.columns
res23: Array[String] = Array(key1, key2, key3)

scala> df_6.show
+----+----+----+
|key1|key2|key3|
+----+----+----+
| aaa|   1|   2|
| bbb|   3|   4|
| ccc|   3|   5|
| bbb|   4|   6|
+----+----+----+

再写几个null值等无效数据的一些处理吧
这次得换个数据集，null值的表用个csv文件导入，代码如下：

/*
csv文件内容如下：
key1,key2,key3,key4,key5
aaa,1,2,t1,4
bbb,5,3,t2,8
ccc,2,2,,7
,7,3,t1,
bbb,1,5,t3,0
,4,,t1,8 
*/
scala> val df = spark.read.option("header","true").csv("natest.csv")
df: org.apache.spark.sql.DataFrame = [key1: string, key2: string ... 3 more fields]

scala> df.show
+----+----+----+----+----+
|key1|key2|key3|key4|key5|
+----+----+----+----+----+
| aaa|   1|   2|  t1|   4|
| bbb|   5|   3|  t2|   8|
| ccc|   2|   2|null|   7|
|null|   7|   3|  t1|null|
| bbb|   1|   5|  t3|   0|
| null|   4|null|  t1|   8|
+----+----+----+----+----+

/*
把key1列中所有的null值替换成'xxx' 
*/
scala> val df_2 = df.na.fill("xxx",Seq("key1"))
df_2: org.apache.spark.sql.DataFrame = [key1: string, key2: string ... 3 more fields]

scala> df_2.show
+----+----+----+----+----+
|key1|key2|key3|key4|key5|
+----+----+----+----+----+
| aaa|   1|   2|  t1|   4|
| bbb|   5|   3|  t2|   8|
| ccc|   2|   2|null|   7|
| xxx|   7|   3|  t1|null|
| bbb|   1|   5|  t3|   0|
| xxx|   4|null|  t1|   8|
+----+----+----+----+----+

/*
一次修改相同类型的多个列的示例。
这里是把key3,key5列中所有的null值替换成1024。
csv导入时默认是string，如果是整型，写法是一样的，有各个类型的重载。
*/
scala> val df_3 = df.na.fill("1024",Seq("key3","key5"))
df_3: org.apache.spark.sql.DataFrame = [key1: string, key2: string ... 3 more fields]

scala> df_3.show
+----+----+----+----+----+
|key1|key2|key3|key4|key5|
+----+----+----+----+----+
| aaa|   1|   2|  t1|   4|
| bbb|   5|   3|  t2|   8|
| ccc|   2|   2|null|   7|
|null|   7|   3|  t1|1024|
| bbb|   1|   5|  t3|   0|
|null|   4|1024|  t1|   8|
+----+----+----+----+----+

/*
一次修改不同类型的多个列的示例。
csv导入时默认是string，如果是整型，写法是一样的，有各个类型的重载。
*/
scala> val df_3 = df.na.fill(Map(("key1"->"yyy"),("key3","1024"),("key4","t88"),("key5","4096")))
df_3: org.apache.spark.sql.DataFrame = [key1: string, key2: string ... 3 more fields]

scala> df_3.show
+----+----+----+----+----+
|key1|key2|key3|key4|key5|
+----+----+----+----+----+
| aaa|   1|   2|  t1|   4|
| bbb|   5|   3|  t2|   8|
| ccc|   2|   2| t88|   7|
| yyy|   7|   3|  t1|4096|
| bbb|   1|   5|  t3|   0|
| yyy|   4|1024|  t1|   8|
+----+----+----+----+----+

/*
不修改，只是过滤掉含有null值的行。
这里是过滤掉key3,key5列中含有null的行
*/
scala> val df_4 = df.na.drop(Seq("key3","key5"))
df_4: org.apache.spark.sql.DataFrame = [key1: string, key2: string ... 3 more fields]

scala> df_4.show
+----+----+----+----+----+
|key1|key2|key3|key4|key5|
+----+----+----+----+----+
| aaa|   1|   2|  t1|   4|
| bbb|   5|   3|  t2|   8|
| ccc|   2|   2|null|   7|
| bbb|   1|   5|  t3|   0|
+----+----+----+----+----+

/*
过滤掉指定的若干列中，有效值少于n列的行
这里是过滤掉key1,key2,key3这3列中有效值小于2列的行。最后一行中，这3列有2列都是null，所以被过滤掉了。
*/
scala> val df_5 = df.na.drop(2,Seq("key1","key2","key3"))
df_5: org.apache.spark.sql.DataFrame = [key1: string, key2: string ... 3 more fields]

scala> df.show
+----+----+----+----+----+
|key1|key2|key3|key4|key5|
+----+----+----+----+----+
| aaa|   1|   2|  t1|   4|
| bbb|   5|   3|  t2|   8|
| ccc|   2|   2|null|   7|
|null|   7|   3|  t1|null|
| bbb|   1|   5|  t3|   0|
|null|   4|null|  t1|   8|
+----+----+----+----+----+

scala> df_5.show
+----+----+----+----+----+
|key1|key2|key3|key4|key5|
+----+----+----+----+----+
| aaa|   1|   2|  t1|   4|
| bbb|   5|   3|  t2|   8|
| ccc|   2|   2|null|   7|
|null|   7|   3|  t1|null|
| bbb|   1|   5|  t3|   0|
+----+----+----+----+----+

/*
同上，如果不指定列名列表，则默认列名列表就是所有列
*/
scala> val df_6 = df.na.drop(4)
df_6: org.apache.spark.sql.DataFrame = [key1: string, key2: string ... 3 more fields]

scala> df_6.show
+----+----+----+----+----+
|key1|key2|key3|key4|key5|
+----+----+----+----+----+
| aaa|   1|   2|  t1|   4|
| bbb|   5|   3|  t2|   8|
| ccc|   2|   2|null|   7|
| bbb|   1|   5|  t3|   0|
+----+----+----+----+----+

ok，就到这吧，下次再写多表的部分了~~

喜欢这些内容的话，可以关注下公众号哈~