How to work with Parquet Data in Apache Spark using SQL

Apache Spark is a fast and general engine for large-scale data processing. When paired with the CData JDBC Driver for Parquet, Spark can work with live Parquet data. This article describes how to connect to and query Parquet data from a Spark shell.

The CData JDBC Driver offers unmatched performance for interacting with live Parquet data due to optimized data processing built into the driver. When you issue complex SQL queries to Parquet, the driver pushes supported SQL operations, like filters and aggregations, directly to Parquet and utilizes the embedded SQL engine to process unsupported operations (often SQL functions and JOIN operations) client-side. With built-in dynamic metadata querying, you can work with and analyze Parquet data using native data types.

Install the CData JDBC Driver for Parquet

Download the CData JDBC Driver for Parquet installer, unzip the package, and run the JAR file to install the driver.

Start a Spark Shell and Connect to Parquet Data

1. Open a terminal and start the Spark shell with the CData JDBC Driver for Parquet JAR file as the jars parameter: $ spark-shell --jars /CData/CData JDBC Driver for Parquet/lib/cdata.jdbc.parquet.jar
2. With the shell running, you can connect to Parquet with a JDBC URL and use the SQL Context load() function to read a table.

   Connect to your local Parquet file(s) by setting the URI connection property to the location of the Parquet file.

   Built-in Connection String Designer

   For assistance in constructing the JDBC URL, use the connection string designer built into the Parquet JDBC Driver. Either double-click the JAR file or execute the jar file from the command-line.

   java -jar cdata.jdbc.parquet.jar

   Fill in the connection properties and copy the connection string to the clipboard.

   Configure the connection to Parquet, using the connection string generated above.

   scala> val parquet_df = spark.sqlContext.read.format("jdbc").option("url", "jdbc:parquet:URI=C:/folder/table.parquet;").option("dbtable","SampleTable_1").option("driver","cdata.jdbc.parquet.ParquetDriver").load()
3. Once you connect and the data is loaded you will see the table schema displayed.

4. Register the Parquet data as a temporary table:

   scala> parquet_df.registerTable("sampletable_1")

5. Perform custom SQL queries against the Data using commands like the one below:

   scala> parquet_df.sqlContext.sql("SELECT Id, Column1 FROM SampleTable_1 WHERE Column2 = SAMPLE_VALUE").collect.foreach(println)

   You will see the results displayed in the console, similar to the following:

Using the CData JDBC Driver for Parquet in Apache Spark, you are able to perform fast and complex analytics on Parquet data, combining the power and utility of Spark with your data. Download a free, 30 day trial of any of the 200+ CData JDBC Drivers and get started today.