pyspark学习_dataframe常用操作_02

#回顾01常用操作
from pyspark import SparkSession,DataFrame

spark = SparkSession.builder.getOrCreate()

peopleDF = spark.read.json("people.json")

peopleDF.printSchema()#显示DataFrame的模式信息
peopleDF.show()#显示DataFrame的数据信息
peopleDF.foreach(print)#foreach方法参数为function,如print,每条数据输入function进行结果输出
def test_foreach(df:DataFrame):
    if df['age']>20:
        print(df)
    else:
        pass
peopleDF.foreach(test_foreach)
#以下操作返回新的dataframe
peopleDF.select(peopleDF['name'])#查询DataFrame部分列    
peopleDF.filter(peopleDF['age']>30)#按条件筛选数据         
peopleDF.groupBy(peopleDF['age']).count()#按列对数据进行统计   
peopleDF.sort(peopleDF['age'].asc(),peopleDF['name'].desc())#按age升序，如果一致，按name降序
peopleDF.select(peopleDF['name'].alias('username'))#修改列名
peopleDF.withColumn('test_col',peopleDF['age'])#新增列
peopleDF.replace('Bob','tom',subset=['name'])#替换信息，如果name列字段为Bob，替换为tom

2.11 withCoulumnRenamed

#withColumnRenamed:修改现有列名名称
"""
para1:existing:str   现有列名str类型
para2:new:str        修改后列名str类型
return:dataframe     返回dataframe
"""
peopleDF.withColumnRenamed("age","age_new").show()

输出如下:
+-------+----+
|age_new|name|
+-------+----+
|     12| tom|
|     22|jack|
|     33| Bob|
+-------+----+

2.12  join

#join:关联，类似于sql里面的join函数
"""
para1:other:DataFrame   另一个DataFrame
para2:on:               类似于sql中的on,两个dataframe的关联字段
how:                    left/right/inner/full
"""
peopleDF1 = peopleDF
peopleDF.join(peopleDF1,on = (peopleDF1['age']==peopleDF['age']) & (peopleDF1['name'] == peopleDF['name']),how='inner').show()

输出如下:
+---+----+---+----+
|age|name|age|name|
+---+----+---+----+
| 12| tom| 12| tom|
| 22|jack| 22|jack|
| 33| Bob| 33| Bob|
+---+----+---+----+

2.13 count

#count:计算数量--->返回int
peopleDF.count()  #返回dataframe的数量

输出如下:
3

 2.14 drop

#drop:删除dataframe的列,一次删除一列
"""
para:列名 str
return:dataframe
"""
peopleDF.drop('name').show()

输出如下:
+---+
|age|
+---+
| 12|
| 22|
| 33|
+---+

2.15 take

#take:获取dataframe的前N行数据
"""
para1:num   int类型
return  list(Row)
""" 
print(peopleDF.take(2)) #获取peopleDF前2行数据

输出如下:
[Row(age=12, name='tom'), Row(age=22, name='jack')]


解析Row数据
people_list = peopleDF.take(2)
for people in people_list:
    print(people['age'])

2.16 distinct

#distinct:输出dataframe不同的行数据  返回新的dataframe

peopleDF.distinct().show()

输出如下:
+---+----+
|age|name|
+---+----+
| 33| Bob|
| 12| tom|
| 22|jack|
+---+----+

2.17 union

#union:上下拼接两个不同的dataframe,要求两个dataframe有相同的列数
"""
para1: other:DataFrame 其他的DataFrame
return: DataFrame
"""
peopleDF.union(peopleDF).show()
+---+----+
|age|name|
+---+----+
| 12| tom|
| 22|jack|
| 33| Bob|
| 12| tom|
| 22|jack|
| 33| Bob|
+---+----+

2.18 first

#first:获取dataframe中第一个数据
print(peopleDF.first()) #返回Row

输出如下:
Row(age=12, name='tom')

2.19 createOrReplaceTempView

#createOrReplaceTempView:注册临时表
"""
para:name str类型,注册后表名
return:None
"""
peopleDF.createOrReplaceTempView("people")
result = spark.sql("select * from people")
result.show()

输出如下:
+---+----+
|age|name|
+---+----+
| 12| tom|
| 22|jack|
| 33| Bob|
+---+----+

2.20 repartition

#repartition:返回由给定分区表达式分区的新的DataFrame
"""
para1:numPartitions int:指定分区的数量，如果未指定，则默认分区数
para2:cols str或者Column  指定分区列
return:dataframe
"""
print('修改分区前:{}'.format(peopleDF.rdd.getNumPartitions()))  # 修改分区前分区数
peopleDF_new = peopleDF.repartition(3, 'name')
print('修改分区后:{}'.format(peopleDF_new.rdd.getNumPartitions()))  # 修改分区后分区数

输出如下:
修改分区前:1
修改分区后:3

 2.21 rdd

#rdd:将dataframe类型数据转为RDD类型数据
peopleRDD = peopleDF.rdd
print(type(peopleRDD))

输出如下:

2.22 toDF

#toDF:将RDD类型数据转换为DataFrame类型数据
"""
para1:schema:列名,由各个列名组成的list
para2:sampleRatio:采样率,用于推测各个列的数据类型,默认前100个数据
"""
peopleRdd = peopleDF.rdd
peopleRdd.toDF(schema=['name','age'],sampleRatio=0.5).printSchema()

输出如下:
root
 |-- age: long (nullable = true)
 |-- name: string (nullable = true)

2.23 collect

#collect:将DataFrame类型的数据转为list,同时会从远程集群拉去数据到driver端
print(peopleDF.collect())

#输出如下:
[Row(age=12, name='tom'), Row(age=22, name='jack'), Row(age=33, name='Bob')]

2.24 persist/unpersist

#persist:dataframe数据持久化   unpersist:数据释放持久化  cache:数据持久化，调用了persist
#is_cached:属性,是否缓存
"""
persist参数:
storageLevel:持久化策略 (useDisk,useMemory,useOffHeap,deserialized,replication=1)
默认持久化级别是:Memory_and_disk (True,True,False,False,1)
"""
persist_peopleDF = peopleDF.persist()
print(persist_peopleDF.is_cached)

persist_peopleDF_1 = persist_peopleDF.unpersist()
print(persist_peopleDF_1.is_cached)

输出如下:
True
False

2.25 fillna

#fillna:如果列为空，填充数据
"""
para1:value:要填充的值
para2:subset:列名,可以为多列
return:dataframe  返回新的dataframe
"""
修改原有数据如下:
{"name":"tom","age":12,"year":2012}
{"name":"jack"}
{"name":"Bob","age":33}

peopleDF.fillna(value=0,subset=['age','year']).show()

输出如下:
+---+----+----+
|age|name|year|
+---+----+----+
| 12| tom|2012|
|  0|jack|   0|
| 33| Bob|   0|
+---+----+----+