PySpark - Read Parquet Files in S3

This code snippet provides an example of reading parquet files located in S3 buckets on AWS (Amazon Web Services). 

The bucket used is from New York City taxi trip record data.

S3 bucket location is: s3a://ursa-labs-taxi-data/2009/01/data.parquet.

To run the script, we need to setup the package dependency on Hadoop AWS package, for example, org.apache.hadoop:hadoop-aws:3.3.0. This can be easily done by passing configuration argument using spark-submit:

 spark-submit --conf spark.jars.packages=org.apache.hadoop:hadoop-aws:3.3.0

This can also be done via SparkConf:

conf.set('spark.jars.packages', 'org.apache.hadoop:hadoop-aws:3.3.0')

Use temporary AWS credentials

In this code snippet, AWS AnonymousAWSCredentialsProvider is used. If the bucket is not public, we can also use TemporaryAWSCredentialsProvider.

conf.set('spark.hadoop.fs.s3a.aws.credentials.provider', 'org.apache.hadoop.fs.s3a.TemporaryAWSCredentialsProvider')
conf.set('spark.hadoop.fs.s3a.access.key', <access_key>)
conf.set('spark.hadoop.fs.s3a.secret.key', <secret_key>)
conf.set('spark.hadoop.fs.s3a.session.token', <token>)

If you have used AWS CLI or SAML tools to cache local credentials ( ~/.aws/credentials), you then don't need to specify the access keys assuming the credential has access to the S3 bucket you are reading data from.