Spark的DataFrame调用write方法写入数据时有以下几点注意
- DF写入CSV,要求每一列的类型为String
- DF写入TEXT,要求DF只有一列,且类型为String
修改DataFrame列类型的方法可以参考
1 直接按列名修改
scala> a.select(a.col("id"),a.col("username"),a.col("birthday").cast("string").as("bir")).show
2 遍历DF的所有列依次修改,适用于修改成统一的类型
import org.apache.spark.sql.DataFrame
import org.apache.spark.sql.functions.col
val df: DataFrame = ...
val columns: Array[String] = df.columns
val df2: DataFrame = columns.foldLeft(df){
(currentDF, column) => currentDF.withColumn(column, col(column).cast("string"))
}
3 先修改列的类型,再用select函数查找
scala 有array:_*这样传参这种语法,就是就是把Array数组的参数转化成变长序列传入,而df的select方法也支持col*这样的变长序列。
这样看来,SELECT的COL只需要名字一样,struct可以任意指定,也就是说DF按名查找,再按COL的struct处理数据
val cols = colNames.map(f => col(f).cast(DoubleType))
df.select(cols: _*).show()
再写基于DF的ALS推荐算法时,遇到了DF的存储问题,解决代码如下
def main(args: Array[String]): Unit = {
//初始化
val sparkConf = new SparkConf().setMaster(config("spark.cores")).setAppName("OfflineRecommonder")
val spark = SparkSession.builder().config(sparkConf).getOrCreate()
import spark.implicits._
implicit val mongoConf = MongoConfig(config("mongo.uri"), config("mongo.db"))
val ratingDS = spark.read
.option("uri", mongoConf.uri)
.option("collection", MONGODB_RATING_COLLECTION)
.format("com.mongodb.spark.sql")
.load()
.select($"userId",$"productId",$"score" as("rating"),$"timestamp")
.as[ProductRating]
val Array(training,test) = ratingDS.randomSplit(Array(0.8,0.2))
val als = new ALS()
.setMaxIter(5)
.setRegParam(0.01)
.setRank(5)
.setUserCol("userId")
.setItemCol("productId")
.setRatingCol("rating")
val model = als.fit(training)
// Evaluate the model by computing the RMSE on the test data
// Note we set cold start strategy to 'drop' to ensure we don't get NaN evaluation metrics
model.setColdStartStrategy("drop")
val predictions = model.transform(test)
/**
* DF写入CSV,要求每一列的类型为String
* DF写入Text,要求DF只有一列,且类型为String
*
* 1.改写DF的列类型
*
* 2.写入csv
* */
predictions.columns.foldLeft(predictions){
(currentDF,column)=>currentDF.withColumn(column, col(column).cast("string"))
}.limit(10).coalesce(1)
.write
.mode(SaveMode.Overwrite)
.option("header",true)
.csv("ECommerceRecommendSystem/recs/predictions.csv")
/**
* DF写入CSV,要求每一列的类型为String
* DF写入Text,要求DF只有一列,且类型为String
* */
val userRecs = model.recommendForAllUsers(10)
val newCol = userRecs.columns.map(col(_).cast("string"))
userRecs.select(newCol:_*).limit(10).coalesce(1)
.write
.mode(SaveMode.Overwrite)
.option("header",true)
.csv("ECommerceRecommendSystem/recs/userRecs.csv")
//模型评估器RegressionEvaluator用于回归的评估器,主要是计算y与y^的插值
val evaluator = new RegressionEvaluator()
.setMetricName("rmse") //设置真实值与预测值误差的计算方式,RMSE 均方根误差 ---sqrt ( err(i)*eer(i) / n )
.setLabelCol("rating")//真实值
.setPredictionCol("prediction")//预测值
val rmse = evaluator.evaluate(predictions)
println(s"Root-mean-square error = $rmse")
spark.stop()
}