Sven Kreiss ←Articles

pysparkling

pysparkling is a native Python implementation of the interface provided by Spark’s RDDs. In Spark, RDDs are Resilient Distributed Datasets. An RDD instance provides convenient access to the partitions of data that are distributed on a cluster. New RDDs are created by applying transformations like map() and reduce() to existing RDDs.

pysparkling provides the same functionality, but without the dependency on the Java Virtual Machine, Spark and Hadoop. The original motivation came from implementing a processing pipeline that is common for machine learning: process a large number of documents in parallel for training a classification algorithm (using Apache Spark) and using that trained classification algorithm in an API endpoint where it is applied to a single document at a time. That single document has to be preprocessed however with the same transformations that were also applied during training. This is the task for pysparkling.

Removing the dependency on the JVM, Spark and Hadoop comes at a cost:

  • Hadoop file io is gone, but its core functionality is reimplemented in pysparkling.fileio. This by itself is very handy as you can read the contents of files on s3://, http:// and file:// and optionally with gzip and bz2 compression just by specifying a file name. The name can include multiple comma separated files and the wildcards ? and *.
  • Managed resource allocation on clusters is gone (no YARN). Parallel execution with multiprocessing is supported though.

It also comes with some advanced features:

  • Parallelization with any object that has a map() method. That includes multiprocessing.Pool and concurrent.futures.ProcessPoolExecutor.
  • It does provide lazy and distributed execution. For example, when creating an RDD from 50,000 text files with myrdd = Context().textFile('s3://mybucket/alldata/*.gz') and only accessing one record with myrdd.takeSample(1), pysparkling will only download a single file from S3 and not all 50,000.

Quickstart

Install pysparkling with pip install pysparkling. As a first example, count the number of occurrences of every word in README.rst:

from pysparkling import Context
counts = (
    Context().textFile('README.rst')
    .flatMap(lambda line: line.split(' '))
    .map(lambda word: (word, 1))
    .reduceByKey(lambda a, b: a + b)
)
print(counts.collect())

More examples including how to explore the Common Crawl dataset and the dataset of the Human Biome Project are in this IPython Notebook.

Further Reading

Get an overview of the API and more details from pysparkling’s documentation. If you like this project, star it on Github, tweet about it and follow me, @svenkreiss, on Twitter.

Related Posts

Go Top