Error in forEach Spark Scala : value select is not a member of org.apache.spark.sql.Row

Question

I am trying to get avg of ratings of all json objects in a file. I loaded the file and converted to data frame but getting error while parsing for avg. Sample Request :

{
        "country": "France",
        "customerId": "France001",
        "visited": [
            {
                "placeName": "US",
                "rating": "2.3",
                "famousRest": "N/A",
                "placeId": "AVBS34"

            },
              {
                "placeName": "US",
                "rating": "3.3",
                "famousRest": "SeriousPie",
                "placeId": "VBSs34"

            },
              {
                "placeName": "Canada",
                "rating": "4.3",
                "famousRest": "TimHortons",
                "placeId": "AVBv4d"

            }        
    ]
}

so for this json, US avg rating will be (2.3 + 3.3)/2 = 2.8

{
        "country": "Egypt",
        "customerId": "Egypt009",
        "visited": [
            {
                "placeName": "US",
                "rating": "1.3",
                "famousRest": "McDonald",
                "placeId": "Dedcf3"

            },
              {
                "placeName": "US",
                "rating": "3.3",
                "famousRest": "EagleNest",
                "placeId": "CDfet3"

            },


}

{
        "country": "Canada",
        "customerId": "Canada012",
        "visited": [
            {
                "placeName": "UK",
                "rating": "3.3",
                "famousRest": "N/A",
                "placeId": "XSdce2"

            },


    ]
}

for this avg for us= (3.3 +1.3)/2 = 2.3

so over all, the average rating will be : (2.8 + 2.3)/2 = 2.55 (only two requests have 'US' in their visited list)

My schema :

root
|-- country: string(nullable=true)
|-- customerId:string(nullable=true)
|-- visited: array (nullable = true)
|    |-- element: struct (containsNull = true)
|    |   |-- placeId: string (nullable = true)
|    |   |-- placeName: string (nullable = true) 
|    |   |-- famousRest: string (nullable = true)
|    |   |-- rating: string (nullable = true)

val sqlContext = new org.apache.spark.sql.SQLContext(sc)
val df = sqlContext.jsonFile("temp.txt")
df.show() 

When doing :

val app = df.select("strategies"); app.registerTempTable("app"); app.printSchema(); app.show()
app.foreach({
  t =>  t.select("placeName", "rating").where(t("placeName") == "US")
}).show()

I am getting : 
<console>:31: error: value select is not a member of org.apache.spark.sql.Row t => t.select("placeName", "rating").where(t("placeName") == "US") ^

Can someone tell me what I am doing wrong here ?

Answer 1

Assuming app is a Dataframe (your code example isn't comprehensible... you create a df variable and query an app variable), you shouldn't call foreach in order to select from it:

app.select("placeName", "rating").where(t("placeName") == "US")

foreach would call a function on each record (of type Row ). That is useful mostly for invoking some side-effect (eg print to console / send to external service etc.). Mostly, you wouldn't use it for selecting / transforming Dataframes.

UPDATE :

As for the original question of how to calculate the average of US-only visits:

// explode to make a record out of each "visited" Array item, 
// taking only "placeName" and "rating" columns
val exploded: DataFrame = df.explode(df("visited")) {
  case Row(visits: Seq[Row]) => 
    visits.map(r => (r.getAs[String]("placeName"), r.getAs[String]("rating")))
}

// make some order: rename columns named _1, _2 (since we used a tuple),
// and cast ratings to Double:
val ratings: DataFrame = exploded
  .withColumnRenamed("_1", "placeName")
  .withColumn("rating", exploded("_2").cast(DoubleType))
  .select("placeName", "rating")

ratings.printSchema()
ratings.show()
/* prints:
root
 |-- placeName: string (nullable = true)
 |-- rating: double (nullable = true)

+---------+------+
|placeName|rating|
+---------+------+
|       US|   1.3|
|       US|   3.3|
|       UK|   3.3|
+---------+------+
 */

// now filter US only and get average rating:
val avg = ratings
  .filter(ratings("placeName") === "US")
  .select(mean("rating"))

avg.show()
/* prints:
 +-----------+
 |avg(rating)|
 +-----------+
 |        2.3|
 +-----------+
  */