Spark中Json列的哈希值

Question

I have Cassandra table and in last column named "fullJson" is JSON log files. I need to Hash userID values in that JSON lines using MD5 . Here's my approach, but for some rason I always get stuck at some point. Loaded Cassandra table:

scala> val rawCass = sc.cassandraTable[cassFormat]("keyspace", "logs").repartition(200)
rawCass: org.apache.spark.rdd.RDD[cassFormat] = MapPartitionsRDD[73] at coalesce at CassandraTableScanRDD.scala:256

And I get:

scala> val cassDF2 = spark.createDataFrame(rawCass).select("fullJson")
cassDF2: org.apache.spark.sql.DataFrame = [fullJson: string]

scala> cassDF2.printSchema
root
 |-- fullJson: string (nullable = true)

My JSON file consists of "header" and "body" and I guess best approach is to get Data Frame , then select column userID and hash it with MD5 .

scala> val nestedJson = spark.read.json(cassDF2.select("fullJson").rdd.map(_.getString(0))).select("header","body")
nestedJson: org.apache.spark.sql.DataFrame = [header: struct<KPI: string, action: string ... 16 more fields>, body: struct<1MYield: double, 1YYield: double ... 147 more fields>]

scala> nestedJson.printSchema
root
 |-- header: struct (nullable = true)
 |    |-- KPI: string (nullable = true)
 |    |-- action: string (nullable = true)
 |    |-- appID: string (nullable = true)
 |    |-- appVersion: string (nullable = true)
 |    |-- context: string (nullable = true)
 |    |-- eventID: string (nullable = true)
 |    |-- interestArea: string (nullable = true)
 |    |-- location: struct (nullable = true)
 |    |    |-- lat: string (nullable = true)
 |    |    |-- lon: string (nullable = true)
 |    |-- navigationGroup: string (nullable = true)
 |    |-- sessionID: string (nullable = true)
 |    |-- timestamp: string (nullable = true)
 |    |-- userAge: string (nullable = true)
 |    |-- userAgent: struct (nullable = true)
 |    |    |-- browser: string (nullable = true)
 |    |    |-- browserVersion: string (nullable = true)
 |    |    |-- deviceName: string (nullable = true)
 |    |    |-- deviceResolution: string (nullable = true)
 |    |    |-- deviceType: string (nullable = true)
 |    |    |-- deviceVendor: string (nullable = true)
 |    |    |-- os: string (nullable = true)
 |    |    |-- osVersion: string (nullable = true)
 |    |-- userID: string (nullable = true)
 |    |-- userSegment: string (nullable = true)
 |-- body: struct (nullable = true)
 |    |-- OS: string (nullable = true)
 |    |-- active: boolean (nullable = true)
 |    |-- amount: double (nullable = true)
 |    |-- amountCritical: string (nullable = true)
 |    |-- beneficiary: struct (nullable = true)
 |    |    |-- beneficiaryAccounts: array (nullable = true)
 |    |    |    |-- element: struct (containsNull = true)
 |    |    |    |    |-- beneficiaryAccountBank: string (nullable = true)
...

Now to hash header.userID :

val newDF = nestedJson.withColumn("header.userID", md5($"header.userID"))

I want to save that in CSV file, but it can not be done because it's a struct.

newDF.write.format("com.databricks.spark.csv").option("header", "true").option("delimiter", "|").save("cass_full.csv")

Tried to avoid struct type but couldn't because of other nesting (eg location consists of lat, lon )

scala> val tempT = newDF.select($"header.*",$"body.*")
tempT: org.apache.spark.sql.DataFrame = [KPI: string, action: string ... 165 more fields]

scala> tempT.printSchema
root
 |-- KPI: string (nullable = true)
 |-- action: string (nullable = true)
 |-- appID: string (nullable = true)
 |-- appVersion: string (nullable = true)
 |-- context: string (nullable = true)
 |-- eventID: string (nullable = true)
 |-- interestArea: string (nullable = true)
 |-- location: struct (nullable = true)
 |    |-- lat: string (nullable = true)
 |    |-- lon: string (nullable = true)
 |-- navigationGroup: string (nullable = true)
...

Basic question What's the easiest and most preferable way to do this. Should I just change userID value for every line in JSON , or it can be done somehow different with Data Frames ? Reason to do it is because I have another CSV file from another database that needs to also be hashed with same algorithm and joined afterwards.

Answer 1

Please try saving this in parquet and move on with the second part of your logic to join.

Hope this helps!