How to sort by column in descending order in Spark

问题:

I tried df.orderBy("col1").show(10) but it sorted in ascending order. df.sort("col1").show(10) also sorts in descending order. I looked on stackoverflow and the answers I found were all outdated or referred to RDDs. I'd like to use the native dataframe in spark.

回答1:

You can also sort the column by importing the spark sql functions

import org.apache.spark.sql.functions._
df.orderBy(asc("col1"))

Or

import org.apache.spark.sql.functions._
df.sort(desc("col1"))

importing sqlContext.implicits._

import sqlContext.implicits._
df.orderBy($"col1".desc)

Or

import sqlContext.implicits._
df.sort($"col1".desc)

回答2:

It's in org.apache.spark.sql.DataFrame for sort method:

df.sort($"col1", $"col2".desc)

Note $ and .desc inside sort for the column to sort the results by.

回答3:

The easiest way is to just add the parameter ascending=False:

df.orderBy("col1", ascending=False).show(10)

Reference: http://spark.apache.org/docs/2.1.0/api/python/pyspark.sql.html#pyspark.sql.DataFrame.orderBy

回答4:

df.sort($"ColumnName".desc).show()

回答5:

import org.apache.spark.sql.functions.desc

df.orderBy(desc("columnname1"),desc("columnname2"),asc("columnname3"))

回答6:

In the case of Java:

If we use DataFrames, while applying joins (here Inner join), we can sort (in ASC) after selecting distinct elements in each DF as:

Dataset<Row> d1 = e_data.distinct().join(s_data.distinct(), "e_id").orderBy("salary");

where e_id is the column on which join is applied while sorted by salary in ASC.

Also, we can use Spark SQL as:

SQLContext sqlCtx = spark.sqlContext();
sqlCtx.sql("select * from global_temp.salary order by salary desc").show();

where

• spark  -> SparkSession
• salary -> GlobalTemp View.