python - PySpark使用RDD和json.load解析Json

Question

{

  "city": "Tempe",
  "state": "AZ",
  ...
  "attributes": [
    "BikeParking: True",
    "BusinessAcceptsBitcoin: False",
    "BusinessAcceptsCreditCards: True",
    "BusinessParking: {'garage': False, 'street': False, 'validated': False, 'lot': True, 'valet': False}",
    "DogsAllowed: False",
    "RestaurantsPriceRange2: 2",
    "WheelchairAccessible: True"
  ],
  ...
}

您好，我正在使用 PySpark，我正在尝试输出一个(状态，BusinessAcceptsBitcoin)的元组，目前我正在做:

csr = (dataset
        .filter(lambda e:"city" in e and "BusinessAcceptsBitcoin" in e)
        .map(lambda e: (e["city"],e["BusinessAcceptsBitcoin"]))
        .collect()
        )

但是这个命令失败了。如何获取“BusinessAcceptsBitcoin”和“city”字段？

Answer 1

您可以使用Dataframe和UDF来解析“attributes”字符串。

从您提供的示例数据来看，“属性”似乎不是正确的 JSON 或 Dict。

假设“attributes”只是一个字符串，下面是使用 dataframe 和 Udf 的示例代码。

from pyspark.sql import SparkSession
from pyspark.sql.functions import *
from pyspark.sql.types import *

spark = SparkSession \
            .builder \
            .appName("test") \
            .getOrCreate()

#sample data
data=[{

  "city": "Tempe",
  "state": "AZ",
  "attributes": [
    "BikeParking: True",
    "BusinessAcceptsBitcoin: False",
    "BusinessAcceptsCreditCards: True",
    "BusinessParking: {'garage': False, 'street': False, 'validated': False, 'lot': True, 'valet': False}",
    "DogsAllowed: False",
    "RestaurantsPriceRange2: 2",
    "WheelchairAccessible: True"
  ]
}]
df=spark.sparkContext.parallelize(data).toDF()

用户定义的函数来解析字符串

def get_attribute(data,attribute):
    return [list_item for list_item in data if attribute in list_item][0]

注册udf

udf_get_attribute=udf(get_attribute, StringType

数据框

df.withColumn("BusinessAcceptsBitcoin",udf_get_attribute("attributes",lit("BusinessAcceptsBitcoin"))).select("city","BusinessAcceptsBitcoin").show(truncate=False)

示例输出

+-----+-----------------------------+
|city |BusinessAcceptsBitcoin       |
+-----+-----------------------------+
|Tempe|BusinessAcceptsBitcoin: False|
+-----+-----------------------------+

例如，您也可以使用相同的 udf 来查询任何其他字段

df.withColumn("DogsAllowed",udf_get_attribute("attributes",lit("DogsAllowed"))).select("city","DogsAllowed").show(truncate=False)