Apache Spark Tutorial: Actions in Spark on RDDs

In our last tutorials, we had a look at Transformations in Spark. Now, we look at Actions in Spark. One of them – collect – we already used frequently in our transformation samples. The reason for this is simple – transformations use late binding – so nothing happens – until you call an action. We used the most simple one – collect. Actions always do something with data and thus you should be prepared to use this. With the collect method, all the data is loaded into the memory. For our samples this wasn’t an issue, since we only had very small data. However, If datasets are larger, you need to re-consider this. Other functions might deliver better results for you. Let’s now have a look at the different options available

Reduce Action in Spark

Reduce calls a function on all items in a dataset that accumulates them. Basically, all binary operators can be used. For Python, please refer to this documentation: https://docs.python.org/3/library/operator.html. If we want to sum up all items by multiplying them in Spark, this would look like the following:

from operator import *
spark_data.reduce(mul)

The output would be this:

First and Count Actions in Spark

These Actions are easy to use – they either return the first item in an RDD or return the count of elements in the RDD. If you use count() on a very large RDD, it might take very long and your task could run into a timeout. There is another function called countApprox() that returns the approximate count in an RDD to prevent that.

ds_one.first()

Count is also used the same way like first – without applying a function or similar.

ds_one.count()

Saving data

One important thing to do is to store data eventually somewhere. In our case, we store it into the same directory like we have Jupyter notebooks. We can therefore check if it worked as expected. Normally, you would store data on S3 or any other storage you use. Spark RDDs provide several means to save to files, in our case we will use “saveAsTextFile”. This stores the content as part files that are in text format.

ds_one.saveAsTextFile("data/dsone.csv")

You can now navigate to the output at this folder: data/dsone.csv.

So, today we have learned about Actions in Spark and how to apply them to RDDs.

There is of course much more to learn about Spark, so make sure to read the entire Apache Spark Tutorial. I regularly update this tutorial with new content. Also, I created several other tutorials, such as the Machine Learning Tutorial and the Python for Spark Tutorial. The official Apache Spark page can intensify your experience. Your learning journey can still continue.