saprk的groupby和groupbykey的区别

Posted 2023-05-12

参考技术A

我们先来看一下spark源码中关于这两个API的解释。其实groupBy是Transformations，JavaRDD中的方法，可以操作RDD和PairRDD，而groupByKey是actions，JavaPairRDD中的方法，因此操作的是PairRDD，具体的看返回的结果有什么不同

groupBy

<U> JavaPairRDD<U,java.lang.Iterable<T>> groupBy(Function<T,U> f)


Return an RDD of grouped elements. Each group consists of a key and a sequence of elements mapping to that key.

Parameters:

f - (undocumented)

Returns:

(undocumented)

groupByKey

public JavaPairRDD<K,java.lang.Iterable<V>> groupByKey()


Group the values for each key in the RDD into a single sequence. Hash-partitions the resulting RDD with the existing partitioner/parallelism level.

Note: If you are grouping in order to perform an aggregation (such as a sum or average) over each key, using JavaPairRDD.reduceByKey or JavaPairRDD.combineByKey will provide much better performance.

Returns:

(undocumented)

本回答被提问者采纳 参考技术B 。

[Spark][Python]groupByKey例子

[Spark][Python]sortByKey 例子 的继续:

[Spark][Python]groupByKey例子

In [29]: mydata003.collect()

Out[29]:
[[u‘00001‘, u‘sku933‘],
[u‘00001‘, u‘sku022‘],
[u‘00001‘, u‘sku912‘],
[u‘00001‘, u‘sku331‘],
[u‘00002‘, u‘sku010‘],
[u‘00003‘, u‘sku888‘],
[u‘00004‘, u‘sku411‘]]

In [30]: mydata005=mydata003.groupByKey()

In [32]: mydata005.count()
Out[32]: 4

In [33]: mydata005.collect()
Out[33]:
[(u‘00004‘, <pyspark.resultiterable.ResultIterable at 0x7fcebe436b10>),
(u‘00001‘, <pyspark.resultiterable.ResultIterable at 0x7fcebe436850>),
(u‘00003‘, <pyspark.resultiterable.ResultIterable at 0x7fcebe436050>),
(u‘00002‘, <pyspark.resultiterable.ResultIterable at 0x7fcebe4361d0>)]

那么，对于这种：

(00004,sku411)
(00003,sku888)
(00003,sku022)
(00003,sku010)
(00003,sku594)
(00002,sku912)

理论上变成了这样形式的：

(00002,[sku912,sku331])
(00001,[sku022,sku010,sku933])
(00003,[sku888,sku022,sku010,sku594])
(00004,[sku411])

我们如何把它们都打印输出成如下的格式，我考虑需要用到函数，然后对RDD的每行的Value，看作list，再来遍历。
（等待下次编写）

00002
sku912
sku331

00001
sku022
sku010
sku933

00003
sku088
sku022
sku022
sku010
sku594

00004
sku411