Jaromir Hamala

@jerrinot.bsky.social

48 Followers 75 Following

I play with distributed systems by day and poke various runtimes by night. Engineer @QuestDB, @Hazelcast alumnus.

screenshot from the article. text on the screenshot: Scenario 1: the Parquet tiling competition Let's start simple. There is a single Parquet file holding the dataset (about 14 GB), and we need to query it. We picked five engines that all read external Parquet directly: DuckDB, Polars, ClickHouse, DataFusion, and Salesforce Hyper. Four of them spin up a fresh CLI or Python process for each query invocation. Polars is the exception: ClickBench drives it as a long-lived Python session, which will matter in a moment.
screenshot of https://github.com/cmu-pasta/fray/blob/main/docs/bugs.md - bugs found by Fray