duckberg 0.0.3

No project description provided

duckberg

DuckBerg
query your iceberg data easily and efficiently

---

Table of Contents

• About
• Installation
• Features
• Playground
• Development
• License

About

Duckberg is a Python package that synthesizes the power of PyIceberg and DuckDb. PyIceberg enables efficient interaction with Apache Iceberg, a format for handling large datasets, while DuckDb offers swift in-memory data analysis. When combined, these tools create Duckberg, which simplifies the querying process for large Iceberg datasets stored on blob storage with a user-friendly Pythonic approach.

The underlying principle of the Duckberg Python package is to execute your SQL queries only on those data lake files that contain the necessary data for your results. To fully utilize the benefits of this package, it's assumed that your data is partitioned in a manner that suits your query and use case.

Iceberg catalog types

Duckberg supports the same Iceberg catalogs as PyIceberg, including REST, SQL, Hive, Glue, and DynamoDB. These catalogs are sources of information about Iceberg datasets, tables, partitions, etc. Before using Duckberg, ensure that you have access to an Iceberg catalog that can be utilized.

Installation

pip install duckberg

Features

Easy initialisation

Following initialisation is using the REST Iceberg catalog with Amazon S3 as a iceberg data storage.

from duckberg import DuckBerg

catalog_config: dict[str, str] = {
  "type": "rest", # Iceberg catalog type 
  "uri": "http://iceberg-rest:8181/", # url for Iceberg catalog
  "credentials": "user:password", # credentials for Iceberg catalog
  "s3.endpoint": S3_ENDPOINT, # s3 
  "s3.access-key-id": S3_ACCESS_KEY_ID,
  "s3.secret-access-key": S3_SECET_KEY
}

db = DuckBerg(
     catalog_name="warehouse",
     catalog_config=catalog_config)

Listing tables

db.list_tables()

Listing partitions for particular table

db.list_partitions(table="nyc.taxis")

Querying data to Pandas dataframe

query = "SELECT * FROM nyc.taxis WHERE trip_distance > 40 ORDER BY tolls_amount DESC"
df = db.select(table="nyc.taxis", partition_filter="payment_type = 1", sql=query).read_pandas()

Playground

You can run the playground environment running docker compose in the playground

cd playground
docker-compose up -d

The initial run could take additional time for jupyter docker image build. Then you can access

Iceberg data init

Once all the containers have been initiated run the Spark Iceberg Jupyter notebook that will init the Iceberg data and catalog.

Duckberg playground

Navigate to localhost:8888. Then select example Jupyter notebook you want to run and enjoy Duckberg!

Development

For the development, there is recommendation to use Python 3.10. If you manage your Python versions by Pyenv use

pyenv install 3.10.13
pyenv global 3.10.13

then create and activate virtual environment

python -m venv venv
source venv/bin/activate 

upgrade pip and install dependencies

pip install --upgrade pip
pip install .

then run dockers that contains Iceberg catalog and file storage containing iceberg files

cd playground
docker-compose up -d

init data by running Init Jupyter notebook and run/test Duckberg in the file tests/duckberg-sample.py

Style & Formatting

Use

hatch run lint:fmt
hatch run lint:style

Building package

The Duckberg project is managed by Hatch. Follow [Hatch docs] for an installation or just install by command

brew install hatch

or

pip install hatch

Increase package by

hatch version "x.x.x"

Build

hatch build

and publish

hatch publish

License

duckberg is distributed under the terms of the Apache 2.0 license.