从RDD创建DataFrame
1.pandas df 与 spark df的相互转换
df_s=spark.createDataFrame(df_p)
df_p=df_s.toPandas()
import pandas as pd
import numpy as np
arr = np.arange(9).reshape(-1,3)
arr
df_p = pd.DataFrame(arr)
df_p
df_p.columns = ['A','B','C']
df_p
df_s = spark.createDataFrame(df_p)
df_s.show()
df_s.collect()
df_s.toPandas()
2. Spark与Pandas中DataFrame对比
http://www.lining0806.com/spark%E4%B8%8Epandas%E4%B8%ADdataframe%E5%AF%B9%E6%AF%94/
3.1 利用反射机制推断RDD模式
- sc创建RDD
- 转换成Row元素,列名=值
- spark.createDataFrame生成df
- df.show(), df.printSchema()
from pyspark.sql import Row
people = spark.sparkContext.textFile('file:///usr/local/spark/examples/src/main/resources/people.txt').map(lambda line:line.split(',')).map(lambda w:Row(name=w[0],age=int(w[1])))
sPeople = spark.createDataFrame(people)
sPeople.createOrReplaceTempView('people')
personDF = spark.sql('select name,age from people where age>20')
personRDD = personDF.rdd.map(lambda p:"Name:"+p.name+","+"Age:"+str(p.age))
personRDD.foreach(print)
sPeople.show()
sPeople.printSchema()
3.2 使用编程方式定义RDD模式
- 生成“表头”
- fields = [StructField(field_name, StringType(), True) ,...]
- schema = StructType(fields)
- from pyspark.sql.types import StringType,StructField,StructType
-
from pyspark.sql import Row
schemaString = "name age"
fields = [StructField(field_name, StringType(), True) for field_name in schemaString.split(" ")]
schema = StructType(fields)
fields
schema
- 生成“表中的记录”
- 创建RDD
- 转换成Row元素,列名=值
-
lines = spark.sparkContext.textFile('file:///usr/local/spark/examples/src/main/resources/people.txt')
-
part = lines.map(lambda w:w.split(","))
-
peoples = part.map(lambda p:Row(p[0],p[1].strip()))
-
peoples.collect()
-
- 把“表头”和“表中的记录”拼装在一起
- = spark.createDataFrame(RDD, schema)
schemaPeople = spark.createDataFrame(people,schema)
schemaPeople.show()
schemaPeople.printSchema()
4. DataFrame保存为文件
df.write.json(dir)
schemaPeople.write.json('file:///home/hadoop/schema_out')