datapipelab 0.3.5

A data pipeline library with connectors, sources, processors, and sinks.

Datapipelab

Overview

datapipelab is a lightweight, flexible data pipeline framework designed for building and orchestrating complex data workflows. It supports a modular node-based architecture, allowing users to plug in source, processor, and sink nodes using technologies such as Apache Spark, Google BigQuery, Hive, Delta Lake, and Microsoft Teams.

Installation

Clone the repository and install any required dependencies:

pip install -r requirements.txt

Or, if integrating as a module:

pip install datapipelab

Usage Guide

To run a pipeline, you typically follow these steps:

1. Define your pipeline configuration using a Python list or a JSON config.
2. Instantiate and execute the engine.

Example:

from datapipelab.engine import Engine

config = [
    {
        "type": "source",
        "format": "hive",
        "name": "load_customer_accounts",
        "options": {
            "query": "SELECT customer_id, enrollment_date FROM customer_account"
        }
    },
    {
        "type": "processor",
        "format": "spark",
        "name": "aggregate_active_users",
        "options": {
            "parents": ["load_customer_accounts"],
            "query": """
                SELECT 
                    YEAR(enrollment_date) AS enrollment_year, 
                    COUNT(*) AS active_user_count
                FROM load_customer_accounts
                GROUP BY enrollment_year
            """
        }
    },
    {
        "type": "sink",
        "format": "hive",
        "name": "store_active_user_report",
        "options": {
            "parents": ["aggregate_active_users"],
            "table": "report.active_user_summary"
        }
    },
    {
        "type": "sink",
        "format": "teams_notification",
        "name": "notify_report_ready",
        "options": {
            "parents": ["store_active_user_report"],
            "webhook_url": "{{{WEBHOOK_URL}}}",
            "message": "Active user report has been updated in Hive."
        }
    }
]

params = {"WEBHOOK_URL": "https://outlook.office.com/webhook/..."}
engine = Engine(config, spark, params)
engine.running_travelers()

Pipeline Configuration

Pipelines are defined using structured configuration objects or files that specify:

• Nodes (source, processor, sink)
• Dependencies and execution order via parents
• Parameters for each node, e.g., SQL queries, table names, paths

Available Node Types

Source Nodes

• spark_node

  • Executes a Spark SQL query to read data into the pipeline.

  • Example:

    {
      "name": "node_name",
      "type": "source",
      "format": "spark",
      "source": "spark",
      "options": {
        "query": "SELECT * FROM database_name.table_name"
      }
    }
    

• hive_node

  • Reads data from a Hive table.

  • Example:

    {
      "name": "node_name",
      "type": "source",
      "format": "hive",
      "source": "hive",
      "options": {
        "query": "SELECT * FROM database_name.table_name"
      }
    }
    

Processor Nodes

• bigquery_api_node

  • Executes a query via BigQuery API.

  • Example:

    {
      "name": "node_name",
      "type": "processor",
      "format": "bigquery_api",
      "options": {
            "credentials_path": "creadentials.json",
            "return_as_spark_df": false,
            "return_as_python_list": false,
            "return_as_is": true,
            "project_name": "project_name",
            "query": "select * from `project_name.dataset_name.table_name`"
      }
    }
    

    • *return_as_python_list and return_as_is are optional
    • *query can be any valid BigQuery SQL query including (SELECT/DDL/DML/Scripting/Control Flow/Stored Procedure Calls/Temporary Table Usage) statements.

• gcp_bucket_api_node

  • Deletes a bucket or a directory in a GCP bucket.

  • Example:

    {
      "name": "node_name",
      "type": "processor",
      "format": "gcp_bucket_api",
      "options": {
            "credentials_path": "creadentials.json",
            "project_name": "project_name",
            "bucket_name": "bucket_name",
            "subdirectory": "path/to/subdirectory"
      }
    }
    

    • *subdirectory is optional and can be used to specify a subdirectory within the bucket.

• bigquery_spark_node

  • Reads data from BigQuery using the Spark BigQuery connector.

  • Example:

    {
      "name": "node_name",
      "type": "processor",
      "format": "bigquery_spark",
      "options": {
            "parent_project": "parent_project_name",
            "materialization_dataset": "materialization_dataset_name",
            "query": "select * from `project_name.dataset_name.table_name`"
      }
    }
    

    • *query does not support DDL/DML/Scripting/Control Flow/Stored Procedure Calls/Temporary Table Usage statements. Only SELECT statements are supported.

• shell_node

  • Executes a shell command or script.

  • Example:

    {
      "name": "node_name",
      "type": "processor",
      "format": "shell",
      "options": {
        "query": "echo 'Hello, World!'"
      }
    }
    

• custom_node

  • Custom logic node written by user.

  • Example:

    {
      "name": "node_name",
      "type": "processor",
      "format": "custom",
      "options": {
        "module_name": "CustomModuleName"
        "module_path": "path/to/custom_module",
        "class_name": "CustomNodeClassName",
        "optional_param": "value"
      }
    }
    

Sink Nodes

• hive_node

  • Writes output to a Hive table.

  • Example:

    {
      "name": "node_name",
      "type": "sink",
      "format": "hive",
      "type": "spark",
      "options": {
        "parents": ["parent_node_name"],
        "database": "database_name",
        "table": "table_name"
      }
    }
    

• spark_node

  • Writes output to a Hive table.

  • Example:

    {
      "name": "node_name",
      "type": "sink",
      "format": "spark",
      "type": "spark",
      "options": {
        "parents": ["parent_node_name"],
        "database": "database_name",
        "table": "table_name"
      }
    }
    

• teams_notification_node

  • Sends a message to a Microsoft Teams channel.

  • Example:

    {
      "type": "sink",
      "format": "teams_notification",
      "name": "notify_report_ready",
      "options": {
         "parents": ["store_active_user_report"],
         "webhook_url": "{{{WEBHOOK_URL}}}",
         "message": "Active user report has been updated in Hive."
      }
    }
    

Extending the Framework

To create a custom node:

1. Subclass TNode from app/node/tnode.py
2. Implement the required methods (run, validate, etc.)
3. Register your node in the pipeline factory or configuration

Logging and Monitoring

Logging is centralized in logger.py. Logs are categorized by node and execution stage to assist with debugging and auditing.

Troubleshooting

---

For more advanced examples or integration guides, refer to the examples/ folder or reach out to the maintainers.