[英]Can I return a Tuple2 from an apache spark UDF (in java)?
I need a UDF2 that takes two arguments as input corresponding to two Dataframe columns of types String and mllib.linalg.Vector and return a Tuple2. 我需要一个UDF2,它需要两个参数作为输入,分别对应于String和mllib.linalg.Vector类型的两个Dataframe列,并返回一个Tuple2。 IS this doable? 这可行吗? if yes, how do I register this udf()? 如果是,我如何注册这个udf()?
hiveContext.udf().register("getItemData", get_item_data, WHAT GOES HERE FOR RETURN TYPE?);
the udf is defined as follows: udf定义如下:
UDF2<String, org.apache.spark.mllib.linalg.Vector, Tuple2<String, org.apache.spark.mllib.linalg.Vector>> get_item_data =
(String id, org.apache.spark.mllib.linalg.Vector features) -> {
return new Tuple2<>(id, features);
};
There goes a schema
which can be defined as follows: 有这样一种schema
,可以被定义如下:
import org.apache.spark.sql.types.DataType;
import org.apache.spark.sql.types.DataTypes;
import org.apache.spark.mllib.linalg.VectorUDT;
List<StructField> fields = new ArrayList<>();
fields.add(DataTypes.createStructField("id", DataTypes.StringType, false));
fields.add(DataTypes.createStructField("features", new VectorUDT(), false));
DataType schema = DataTypes.createStructType(fields);
but if all you need is just a struct
without any additional processing org.apache.spark.sql.functions.struct
should do the trick: 但是,如果您只需要一个没有任何额外处理的struct
,则org.apache.spark.sql.functions.struct
应该可以解决问题:
df.select(struct(col("id"), col("features"));
声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.