I was recently confronted with an interesting conundrum when writing a complex data pipeline. It was an interesting problem that arose from my quest to reduce complexity in part of the design, which found itself creeping into another part, re-enforcing the classic idea of whether you can really make the complexity pea go away, or if you simply shuffle the pea somewhere else to hide it.
Being Data Analytics is a meat grinder, it’s the worst job ever. Horrible it is. It will crush you.
Back in March, I did a writeup and experiment called DuckDB vs Polars, Thunderdom, 16GB on 4GB machine challenge. The idea was to see if the two tools could process “larger than memory” datasets with lazy execution. Polars worked fine, DuckDB failed in spectacular fashion.
I also noted how many people had opened issues in GitHub about this very thing, but the issues were either ignored or closed. Someone on YouTube said some of these OOM issues were fixed in recent releases.
It’s probably what every single person wants to accomplish first after they’ve been writing code for a year professionally. How do I get to Senior Engineer? What skills do I need? “I am a good coder, give me the Senior Engineer title.”
Sadly, most Junior and Mid-Level Engineers think that being a Senior Engineer is all about coding, when in fact, it’s not. Being good at writing code is only the first step, the are many other way harder skills to learn to grow from Junior to Senior Engineer.
I recently did a post on Linkedin and Reddit about Databricks removing Standard Tier and forcing folks into Unity Catalog. The post got big traction and blew up, more than I thought. Enough for the Databricks folk to hunt me down at work and tell me I’m naughty.
I will be writing a more in-depth post soon on Substack about the downsides of Vendor Lockin and how Data Teams should think about such things.
I never thought I would live to see the day, it’s crazy. I’m not sure who’s idea it was to make it possible to write Apache Spark with Rust, Golang, or Python … but they are all genius.
As of Apache Spark 3.4 it is now possible to use Spark Connect … a thin API client on a Spark Cluster ontop of the DataFrame API.
You can now connect backend systems and code, using Rust or Golang etc, to a Spark Server and run commands and get results remotely. Simply amazing. A new era of tools and products is going to be unleashed on us. We are no longer chained to the JVM. The walls have been broken down. The future is bright.
It’s been a while since I wrote about Polars on this blog, I’ve been remiss. Some time ago I wrote a very simple comparison of switching from Pandas to Polars, I didn’t put much real effort into it, yet it was popular, so this is my attempt at trying to expand on that topic a little.
Recently, while laying flat on back on my sunporch soaking up the vitamin D beating down on me, dreaming about code, which I always do, it struck me.
Have you ever wondered at a high level what it’s like to build production-level data pipelines on Databricks? What does it look like, what tools do you use?
Ever wondered how to build and end-to-end project for an Open Source Python Package that gets published to PYPI? I built out lakescuman open-source package to help with Databricks Unity Catalog Delta Lake tables querying with Polars, DuckDB, or PyArrow. https://github.com/danielbeach/lakescum
Want to know how to grow to the Senior Engineering position? Take a look.