使用 Spark/Scala 過濾 RDD 的 csv 和 JSON 字段

Question

我正在研究 spark/scala，我需要按列上的特定字段過濾 RDD，在本例中為user 。

我想與用戶["Joe","Plank","Willy"]返回一個 RDD，但似乎無法弄清楚如何

這是我的RDD：

2020-03-01T00:00:05Z    my.local5.url   {"request_method":"GET","request_length":281,"user":"Joe"}
2020-03-01T00:00:05Z    my.local2.url   {"request_method":"GET","request_length":281,"user":"Plank"}
2020-03-01T00:00:05Z    my.local2.url   {"request_method":"GET","request_length":281,"user":"Willy"}
2020-03-01T00:00:05Z    my.local6.url   {"request_method":"GET","request_length":281,"user":"Plank"}
2020-03-01T00:00:05Z    my.local2.url   {"request_method":"GET","request_length":281,"user":"Plank"}
2020-03-01T00:00:05Z    my.local2.url   {"request_method":"GET","request_length":281,"user":"Tracy"}
2020-03-01T00:00:05Z    my.local6.url   {"request_method":"GET","request_length":281,"user":"Roger"}

預期 output：

2020-03-01T00:00:05Z    my.local5.url   {"request_method":"GET","request_length":281,"user":"Joe"}
2020-03-01T00:00:05Z    my.local2.url   {"request_method":"GET","request_length":281,"user":"Plank"}
2020-03-01T00:00:05Z    my.local2.url   {"request_method":"GET","request_length":281,"user":"Willy"}
2020-03-01T00:00:05Z    my.local6.url   {"request_method":"GET","request_length":281,"user":"Plank"}
2020-03-01T00:00:05Z    my.local2.url   {"request_method":"GET","request_length":281,"user":"Plank"}

我已經使用類似這樣的spark（偽代碼）提取了rdd：

val sparkConf = new SparkConf().setAppName("MyApp")
master.foreach(sparkConf.setMaster)
val sc = new SparkContext(sparkConf)

val rdd = sc.textFile(inputDir)
rdd.filter(_.contains("\"user\":\"THE_ARRAY_OF_NAMES_"))

Answer 1

您更容易使用數據框。

使用 from_json function 您可以將 json 列轉換為多列

val jsonSchema = StructType(Array(
    StructField("request_method",StringType,true),
    StructField("request_length",IntegerType,true),
    StructField("user",StringType,true)
  ))

val myDf = spark.read.option("header", "true").csv(path)
val formatedDf = myDf.withColumn("formated_json", from_json($"column_name", jsonSchema)
.select($"formated_json.*")
.where($"user".isin("Joe","Plank","Willy")

formatedDf.show

但是，如果您想要 RDD 方法，請告訴我。

使用 RDD 版本編輯：記住這是許多方法之一

//Define a regex pattern
val Pattern = """(?i)"user":"([a-zA-Z]+)"""".r
//Define a Set with your filtered values
val userSet = Set("Joe","Plank","Willy")
//Filter only the values you want
val filteredRdd = rdd.filter( x => {
    //Extract the user using the pattern we just declared
    val user = for(m <- Pattern.findFirstMatchIn(x)) yield m.group(1)
    //If the user variable is equal with one of your set values then this statement will return true and based on that the row will be kept
    userSet(user.getOrElse(""))
})

要查看結果是否正確，可以使用：

filteredRdd.collect().foreach(println)