Post

You can now run @apache.org DataFusion inside @cloudflare.social Workers using Rust and WebAssembly, with optional DuckLake support: github.com/tobilg/dataf... Features: - Query remote Parquet through native R2 bindings, S3-compatible storage, or HTTPS - Native R2 access - Read-only SQL API

github.comGitHub - tobilg/datafusion-ducklake-workers: Deploy DataFusion to Cloudflare Rust Workers via Emscripten, with optional DuckLake supportDeploy DataFusion to Cloudflare Rust Workers via Emscripten, with optional DuckLake support - tobilg/datafusion-ducklake-workers

4/10/26, 12:44 pm1 repost6 likes

The Projekt

@wearetheprojekt.bsky.social

Putting DataFusion inside Workers is a sharp boundary: remote Parquet stays remote while the query engine moves to the edge. The read-only SQL API keeps the first version constrained.

PDFops

@pdfops.dev

curious how the DataFusion WASM init cost looks on cold starts. we see something similar bundling pdf libs into workers, the module parse/instantiate tax dwarfs the actual query/fill time until you're warm. did you measure cold vs warm separately?