pyspark指定schema

通过StructType对象指定DataFrame的Schema

• 没有嵌套结构的json

jsonString = [
"""{ "id" : "01001", "city" : "AGAWAM",  "pop" : 15338, "state" : "MA" }""",
"""{ "id" : "01002", "city" : "CUSHMAN", "pop" : 36963, "state" : "MA" }"""
]

jsonRDD = sc.parallelize(jsonString)

from pyspark.sql.types import *

#定义结构类型
#StructType：schema的整体结构，表示JSON的对象结构
#XXXStype:指的是某一列的数据类型
jsonSchema = StructType() \
  .add("id", StringType(),True) \
  .add("city", StringType()) \
  .add("pop" , LongType()) \
  .add("state",StringType())

jsonSchema = StructType() \
  .add("id", LongType(),True) \
  .add("city", StringType()) \
  .add("pop" , DoubleType()) \
  .add("state",StringType())

reader = spark.read.schema(jsonSchema)

jsonDF = reader.json(jsonRDD)
jsonDF.printSchema()
jsonDF.show()

• 带有嵌套结构的json

from pyspark.sql.types import *
jsonSchema = StructType([
    StructField("id", StringType(), True),
    StructField("city", StringType(), True),
    StructField("loc" , ArrayType(DoubleType())),
    StructField("pop", LongType(), True),
    StructField("state", StringType(), True)
])

reader = spark.read.schema(jsonSchema)
jsonDF = reader.json('data/nest.json')
jsonDF.printSchema()
jsonDF.show(2)
jsonDF.filter(jsonDF.pop>4000).show(10)