Parquet

Creating a data generator for Parquet. You will have the ability to generate and validate Parquet files via Docker.

Requirements

• 10 minutes
• Git
• Gradle
• Docker

Get Started

First, we will clone the data-caterer-example repo which will already have the base project setup required.

JavaScalaYAMLUI

git clone git@github.com:data-catering/data-caterer-example.git

git clone git@github.com:data-catering/data-caterer-example.git

git clone git@github.com:data-catering/data-caterer-example.git

Run Data Caterer UI via the 'Quick Start' found here.

Plan Setup

Create a new Java or Scala class.

• Java: src/main/java/io/github/datacatering/plan/MyParquetJavaPlan.java
• Scala: src/main/scala/io/github/datacatering/plan/MyParquetPlan.scala

Make sure your class extends PlanRun.

JavaScala

import io.github.datacatering.datacaterer.java.api.PlanRun;

public class MyParquetJavaPlan extends PlanRun {
}

import io.github.datacatering.datacaterer.api.PlanRun

class MyParquetPlan extends PlanRun {
}

This class defines where we need to define all of our configurations for generating data. There are helper variables and methods defined to make it simple and easy to use.

Connection Configuration

Within our class, we can start by defining the connection properties to read/write from/to Parquet.

JavaScala

var accountTask = parquet(
    "customer_accounts",                      //name
    "/opt/app/data/customer/account_parquet", //path
    Map.of()                                  //additional options
);

Additional options can be found here.

val accountTask = parquet(
  "customer_accounts",                      //name         
  "/opt/app/data/customer/account_parquet", //path
  Map()                                     //additional options
)

Additional options can be found here.

Schema

Depending on how you want to define the schema, follow the below:

• Manual schema guide
• Automatically detect schema from the data source, you can simply enable configuration.enableGeneratePlanAndTasks(true)
• Automatically detect schema from a metadata source

Additional Configurations

At the end of data generation, a report gets generated that summarises the actions it performed. We can control the output folder of that report via configurations. We will also enable the unique check to ensure any unique fields will have unique values generated.

JavaScala

var config = configuration()
        .generatedReportsFolderPath("/opt/app/data/report")
        .enableUniqueCheck(true);

execute(myPlan, config, accountTask, transactionTask);

val config = configuration
  .generatedReportsFolderPath("/opt/app/data/report")
  .enableUniqueCheck(true)

execute(myPlan, config, accountTask, transactionTask)

Run

Now we can run via the script ./run.sh that is in the top level directory of the data-caterer-example to run the class we just created.

./run.sh
#input class MyParquetJavaPlan or MyParquetPlan

Congratulations! You have now made a data generator that has simulated a real world data scenario. You can check the ParquetJavaPlan.java or ParquetPlan.scala files as well to check that your plan is the same.

Validation

If you want to validate data from a Parquet source, follow the validation documentation found here to help guide you.