Polars and Parquet

Polars has been making waves in the Python data world — and for good reason. It’s fast, expressive, and built with performance-first principles. If you’re dealing with Parquet files in an S3 bucket and care even a little about speed or memory, this post is for you. Let’s talk about three powerful components working together: 🪣 S3 bucket storing Parquet files 📦 PyArrow-style datasets ⚡ Polars doing its thing — lazily and efficiently 🔍 The Case for Lazy Reading When you reach for read_parquet, you get everything. That’s fine… until it’s not. Instead, scan_parquet gives you a lazy frame — and that changes everything. ...

April 24, 2025 · 2 min · 327 words · Joost

Comparison of Python Clients for Object Storage

Basics: botocore is the low-level interface for practically all other clients. boto3 official Amazon Python client minio standalone alternative to botocore/boto3, does not natively support asynchronous operations as of now. Async: aiobotocore async support for botocore aioboto3 wrapper for aiobotocore s3fs wrapper for aiobotocore Boto3 import boto3 from botocore.config import Config # Initialize the S3 client with a custom endpoint s3 = boto3.client( "s3", endpoint_url="https://custom-endpoint.com", aws_access_key_id="your-access-key", aws_secret_access_key="your-secret-key", config=Config(signature_version="s3v4"), ) # Upload a file s3.upload_file("local/path/file.txt", "bucket-name", "destination/path/file.txt") # Download a file s3.download_file("bucket-name", "source/path/file.txt", "local/path/file.txt") Minio from minio import Minio # Initialize the Minio client with a custom endpoint client = Minio( "custom-endpoint.com", access_key="your-access-key", secret_key="your-secret-key", secure=True, # Set to False if not using HTTPS ) # Upload a file client.fput_object("bucket-name", "destination/path/file.txt", "local/path/file.txt") # Download a file client.fget_object("bucket-name", "source/path/file.txt", "local/path/file.txt") aioboto3 import aioboto3 # Create an async session and client session = aioboto3.Session( aws_access_key_id="your-access-key", aws_secret_access_key="your-secret-key", ) async with aioboto3.Session().client( "s3", endpoint_url=endpoint_url, aws_access_key_id=access_key, aws_secret_access_key=secret_key, ) as s3: # Upload the file with open(file_path, "rb") as file: await s3.upload_fileobj(file, bucket_name, object_name) S3fs Looks the most clean: ...

January 6, 2025 · 2 min · 219 words · Joost

Async Pandas

Pandas is great for Python because it offers efficient data manipulation and analysis capabilities, leveraging the speed of the underlying NumPy library. How does it behave with asyncio since I could not find much about it. Have an enourmnes dataset call an API with a throughput of 10call at once. The simple example pandas.DataFrame consists of 100 rows of lorem text: import lorem import pandas as pd df = pd.DataFrame({"Text": [lorem.text() for _ in range(100)]}) >>> df.head() Text 0 Labore quisquam neque adipisci labore non quae... 1 Aliquam etincidunt dolore dolore voluptatem. A... 2 Aliquam consectetur dolor dolorem dolorem ipsu... 3 Labore non aliquam numquam sed. Eius neque con... 4 Voluptatem ipsum modi amet tempora tempora eti... Asyncio If we want to sent every row to an API and that call takes about a second. Let’s consider this method reverses the text and returns the final three letters: ...

May 16, 2024 · 2 min · 340 words · Joost

Row-Level Security with SQLAlchemy

With Row Level security (RLS) you manage the access control at the row level within a database instead of the application. Row-Level Security allows you to define policies that determine which rows of data a particular user or role can access within a given table. Postgres Tables For this demonstration we create a simple setup with a User table and a Item table using SQLAlchemy 2.0: from sqlalchemy import Column, ForeignKey, Integer, String, create_engine, text from sqlalchemy.orm import declarative_base, relationship admin_engine = create_engine("postgresql://postgres:postgres@0.0.0.0:5432/postgres") Base = declarative_base() class Item(Base): __tablename__ = "items" id = Column(Integer, primary_key=True) name = Column(String) user_id = Column(Integer, ForeignKey("users.id")) user = relationship("User", back_populates="item_entries") class User(Base): __tablename__ = "users" id = Column(Integer, primary_key=True) username = Column(String) password = Column(String) item_entries = relationship("Item", back_populates="user") Base.metadata.create_all(admin_engine) Using the PostgreSQL superuser for application access is not a great idea due to its extensive privileges and security risks. It’s advisable to create a dedicated user with limited permissions tailored to the application’s requirements for improved security and operational control. ...

January 31, 2024 · 3 min · 562 words · Joost

Obfuscate Python

this post is under construction – I have the approaches here but need some time to also share the experience… How to obscure some Python code from anyone running the code? I am no expert here but I have tried a few things and will give my steps and recommendations here. Have a main.py with a simple helloworld FastAPI in this case. There is also an /error endpoint to see how much source code is returned in the logs. ...

November 6, 2023 · 2 min · 289 words · Joost