Usage of spark DataFrame "as" method

Question

I am looking at spark.sql.DataFrame documentation.

There is

def as(alias: String): DataFrame
    Returns a new DataFrame with an alias set.
    Since
        1.3.0 

What is the purpose of this method? How is it used? Can there be an example?

I have not managed to find anything about this method online and the documentation is pretty non-existent. I have not managed to make any kind of alias using this method.

Answer 1

Spark <= 1.5

It is more or less equivalent to SQL table aliases:

SELECT *
FROM table AS alias;

Example usage adapted from PySpark alias documentation:

import org.apache.spark.sql.functions.col
case class Person(name: String, age: Int)

val df = sqlContext.createDataFrame(
    Person("Alice", 2) :: Person("Bob", 5) :: Nil)

val df_as1 = df.as("df1")
val df_as2 = df.as("df2")
val joined_df = df_as1.join(
    df_as2, col("df1.name") === col("df2.name"), "inner")
joined_df.select(
    col("df1.name"), col("df2.name"), col("df2.age")).show

Output:

+-----+-----+---+
| name| name|age|
+-----+-----+---+
|Alice|Alice|  2|
|  Bob|  Bob|  5|
+-----+-----+---+

Same thing using SQL query:

df.registerTempTable("df")
sqlContext.sql("""SELECT df1.name, df2.name, df2.age
                  FROM df AS df1 JOIN df AS df2
                  ON df1.name == df2.name""")

What is the purpose of this method?

Pretty much avoiding ambiguous column references.

Spark 1.6+

There is also a new as[U](implicit arg0: Encoder[U]): Dataset[U] which is used to convert a DataFrame to a DataSet of a given type. For example:

df.as[Person]