从结构数组中选择Spark DataFrames中的特定列 [英] select specific columns in Spark DataFrames from Array of Struct

问题描述

我有一个具有以下架构的Spark DataFrame df:

I have a Spark DataFrame df with the following Schema:

root
 |-- k: integer (nullable = false)
 |-- v: array (nullable = true)
 |    |-- element: struct (containsNull = true)
 |    |    |-- a: integer (nullable = false)
 |    |    |-- b: double (nullable = false)
 |    |    |-- c: string (nullable = true)

是否可以在不执行map的情况下从df中选择v中的a, c?特别是df是从Parquet文件中加载的，我甚至不希望c的值被加载/读取.

Is it possible to just select a, c in v from df without doing a map? In particular, df is loaded from a Parquet file and I don't want the values for c to even be loaded/read.

推荐答案

这完全取决于您期望的输出结果，但您的问题尚不清楚.让我澄清一下. 你可以做

It depends on exactly what you expect as an output, which is not clear from your question. Let me clarify. You can do

df.select($"v.a",$"v.b").show()

但是，结果可能不是您想要的，因为v是一个数组，它将产生一个数组a和b每个.您可能想要做的是explode数组v，然后从爆炸的数据框中选择:

however, the result may be not what you want, since v is an array, it will yield an array for a and one per b. What you may want to do is explode the array v then select from the exploded dataframe:

df.select(explode($"v").as("v" :: Nil )).select($"v.a", $"v.b").show()

这会将v展平到一个表，其中所有值都展平. 在这两种情况下，spark/parquet 应该足够聪明，可以使用谓词下推而不加载c.

this will flatten v to a table with all its values flattened. In either case, spark/parquet should be smart enough to use predicate push down and not load c at all.

这篇关于从结构数组中选择Spark DataFrames中的特定列的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！