YouTube video by PyTorch

Rust’s pattern matching feels simple enough: match on enums, destructure tuples, handle Option and Result.

This post was originally given as a talk for JJ Con. The slides are also available. Welcome to “stupid jj tricks”. Today, I’ll be taking you on a tour through many different jj configurations that I h...

How RisingWave built Foyer, a hybrid caching library in Rust to beat S3 latency

uv is a Python package and project management tool.1 Maturin is a build backend for Python extension modules implemented in Rust. How to best use them together? ## Why uv? First of all, if you’ve already got Maturin, why do you need uv? There are a couple of reasons: * uv takes care of the virtualenv management. * You can use uv to install development dependencies such as pytest. You could use another virtualenv manager such as tox. Myself, I wanted to use uv because I’m already familiar with it and it’s fast. ## Setting up the project Start by creating a new project the usual way: maturin new my_project Among other things, this creates a standard `pyproject.toml` that works with uv. If you run now `uv sync`, uv will build the project using Maturin’s build backend, create a virtualenv and install the package there. However, there is a problem: when you edit your Rust source files, there’s no easy way to rebuild the package. Running `uv sync` will not help because uv does not know that the Rust source files affect the package. Running `maturin develop` does rebuild the package, but if you do anything with uv that triggers a sync - for example, using `uv run` - then it will re-install the cached old version of the package. Below, I present three options for making it work. ## Option 1: Let uv handle the rebuild You can tell uv that the package needs to rebuild whenever the Rust source changes. To do that, you’ll need to set `tool.uv.cache-keys` in `pyproject.toml`: [tool.uv] cache-keys = [{file = "pyproject.toml"}, {file = "Cargo.toml"}, {file = "**/*.rs"}] Note that even if you have a mixed Python/Rust package, the package does not need to rebuild when the Python files change. `uv sync` by default installs the package as editable, so the Python interpreter uses the source files directly. There’s a downside to this approach: `uv sync` builds the package in the release mode which yields faster code but slower build times than the debug mode used by `maturin develop`. ## Option 2: Use maturin develop The alternative is to not install the project package at all when running `uv sync`. Edit `pyproject.toml`: [tool.uv] package = false Now if you run `uv sync`, it will install dependencies but not the project itself. You need to install it manually with Maturin: maturin develop You will have to re-run this command whenever you change the Rust files. ## Option 3: Use maturin import hook If you want to use `maturin develop`, but you also want the code to rebuild automatically whenever it changes, then the Maturin import hook is the way to go. It rebuilds the code if needed when you `import` it in Python. Like in option 2, you need to set `tool.uv.package` to false. You also need to set the minimum required Python version to 3.9 or higher - the default by `maturin new` is 3.8: [project] # ... requires-python = ">=3.9" Then, let’s install the import hook: uv add --dev maturin_import_hook uv run -m maturin_import_hook site install maturin develop You need to install the project with `maturin develop` once, but after that the import hook will take care of the rebuilds. **Gotcha:** If you’re on macOS and you use Python installed with Mac Homebrew, this will not work with Maturin 0.3.0. A workaround is to use uv-managed Python: `uv sync --managed-python` ## Managing Python dependencies You can manage Python dependencies with the usual `uv` commands. For example, to add pytest as a development dependency: uv add --dev pytest * * * 1. In my opinion, uv is the snappiest and the easiest-to-use package management tool for Python right now and it’s the one that I recommend to most people. I’ve previously recommended Poetry, but uv has surpassed it in features and it was always much faster. ↩︎

Modern columnar data format for ML and LLMs implemented in Rust. Convert from parquet in 2 lines of code for 100x faster random access, vector index, and data versioning. Compatible with Pandas, DuckDB, Polars, Pyarrow, and PyTorch with more integrations coming..

YouTube video by EuroRust

We keep saying that Rust is how we make software safer. In this blog, we'll tackle a real-world vulnerability, 'rewrite it in Rust', and show you the results of our empirical research, both as a h ...

Compiling Rust in GitHub Actions can be slow in unpredictable ways. Let's examine the underlying causes and explore sccache as an effective solution.

Open source free software. A vector graphics creativity suite with a clean, intuitive interface. Opens instantly (no signup) and runs locally in a browser. Exports SVG, PNG, JPG.