在Spark DataFrame中按数组值过滤

Question

I am using apache spark 1.5 dataframe with elasticsearch, I am try to filter id from a column that contains a list(array) of ids.

For example the mapping of elasticsearch column is looks like this:

    {
        "people":{
            "properties":{
                "artist":{
                   "properties":{
                      "id":{
                         "index":"not_analyzed",
                         "type":"string"
                       },
                       "name":{
                          "type":"string",
                          "index":"not_analyzed",
                       }
                   }
               }
          }
    }

The example data format will be like following

{
    "people": {
        "artist": {
            [
                  {
                       "id": "153",
                       "name": "Tom"
                  },
                  {
                       "id": "15389",
                       "name": "Cok"
                  }
            ]
        }
    }
},
{
    "people": {
        "artist": {
            [
                  {
                       "id": "369",
                       "name": "Carl"
                  },
                  {
                       "id": "15389",
                       "name": "Cok"
                  },
                 {
                       "id": "698",
                       "name": "Sol"
                  }
            ]
        }
    }
}

In spark I try this:

val peopleId  = 152
val dataFrame = sqlContext.read
     .format("org.elasticsearch.spark.sql")
     .load("index/type")

dataFrame.filter(dataFrame("people.artist.id").contains(peopleId))
    .select("people_sequence.artist.id")

I got all the id that is contains 152, for example 1523 , 152978 but not only id == 152

Then I tried

dataFrame.filter(dataFrame("people.artist.id").equalTo(peopleId))
    .select("people.artist.id")

I get empty, I understand why, it's because I have array of people.artist.id

Can anyone tell me how to filter when I have list of ids ?

Answer 1

In Spark 1.5+ you can use array_contains function:

df.where(array_contains($"people.artist.id", "153"))

If you use an earlier version you can try an UDF like this:

val containsId = udf(
  (rs: Seq[Row], v: String) => rs.map(_.getAs[String]("id")).exists(_ == v))
df.where(containsId($"people.artist", lit("153")))