Tootfinder

Deep Dive into #uv #Dockerfiles by #Astral: Image Size, Performance & Best Practices

So #Gentoo #Python eclasses are pretty modern, in the sense that they tend to follow the best practices and standards, and eventually deal with deprecations. Nevertheless, they have a long history and carry quite some historical burden, particularly regarding to naming.
The key point is that the eclasses were conceived as a replacement for the old eclasses: "distutils" and "python". Hence, much like we revision ebuilds, I've named the matching eclasses "distutils-r1" and "python-r1". For consistency, I've also used the "-r1" suffix for the remaining eclasses introduced at the time: "python-any-r1", "python-single-r1" and "python-utils-r1" — even though there were never "r0"s.
It didn't take long to realize my first mistake. I've made the multi-impl eclass effectively the "main" eclass, probably largely inspired by the previous Gentoo recommendations. However, in the end I've found out that for the most use cases (i.e. where "distutils-r1" is not involved), there is no real need for multi-impl, and it makes things much harder. So if I were naming them today, I would have named it "python-multi", to indicate the specific use case — and either avoid designating a default at all, or made "python-single" the default.
What aged even worse is the "distutils-r1" eclass. Admittedly, back when it was conceived, distutils was still largely a thing — and there were people (like me) who avoided unnecessary dependency on setuptools. Of course, nowadays it has been entirely devoured by setuptools, and with #PEP517 even "setuptools" wouldn't be a good name anymore. Nowadays, people are getting confused why they are supposed to use "distutils-r1" for, say, Hatchling.
Admittedly, this is something I could have done differently — PEP517 support was a major migration, and involved an explicit switch. Instead of adding DISTUTILS_USE_PEP517 (what a self-contradictory name) variable, I could have forked the eclass. Why didn't I do that? Because there used to be a lot of code shared between the two paths. Of course, over time they diverged more, and eventually I've dropped the legacy support — but the opportunity to rename was lost.
In fact, as a semi-related fact, I've recognized another design problem with the eclass earlier — I should have gone for two eclasses rather than one: a "python-phase" eclass with generic sub-phase support, and a "distutils" (or later "python-pep517") implementing default sub-phases for the common backends. And again, this is precisely how I could have solved the code reuse problem when I introduced PEP517 support.
But then, I didn't anticipate how the eclasses would end up looking like in the end — and I can't really predict what new challenges the Python ecosystem is going to bring us. And I think it's too late to rename or split stuff — too much busywork on everyone.

BrainLesion Suite: A Flexible and User-Friendly Framework for Modular Brain Lesion Image Analysis
Florian Kofler, Marcel Rosier, Mehdi Astaraki, Hendrik M\"oller, Ilhem Isra Mekki, Josef A. Buchner, Anton Schmick, Arianna Pfiffer, Eva Oswald, Lucas Zimmer, Ezequiel de la Rosa, Sarthak Pati, Julian Canisius, Arianna Piffer, Ujjwal Baid, Mahyar Valizadeh, Akis Linardos, Jan C. Peeken, Surprosanna Shit, Felix Steinbauer, Daniel Rueckert, Rolf Heckemann, Spyridon Bakas, Jan Kirschke, …

BrainLesion Suite: A Flexible and User-Friendly Framework for Modular Brain Lesion Image Analysis
BrainLesion Suite is a versatile toolkit for building modular brain lesion image analysis pipelines in Python. Following Pythonic principles, BrainLesion Suite is designed to provide a 'brainless' development experience, minimizing cognitive effort and streamlining the creation of complex workflows for clinical and scientific practice. At its core is an adaptable preprocessing module that performs co-registration, atlas registration, and optional skull-stripping and defacing on arbitrary multi-…

Software innovation might be freezing in place—and AI could be to blame. Theo Browne points out that Copilot and ChatGPT often return React-style code even for Solid or Elixir projects. Why? Because they’ve seen React a million times more. Python 3 took a decade to overtake Python 2. If that transition had to happen today, would our dependence on AI suggestions keep us from making the jump?

Bugs in the Shadows: Static Detection of Faulty Python Refactorings
Jonhnanthan Oliveira, Rohit Gheyi, M\'arcio Ribeiro, Alessandro Garcia
https://arxiv.org/abs/2507.01103

Bugs in the Shadows: Static Detection of Faulty Python Refactorings
Python is a widely adopted programming language, valued for its simplicity and flexibility. However, its dynamic type system poses significant challenges for automated refactoring - an essential practice in software evolution aimed at improving internal code structure without changing external behavior. Understanding how type errors are introduced during refactoring is crucial, as such errors can compromise software reliability and reduce developer productivity. In this work, we propose a stati…

Sound Interval-Based Synthesis for Probabilistic Programs
Guilherme Espada, Alcides Fonseca
https://arxiv.org/abs/2507.06939 https://…

Sound Interval-Based Synthesis for Probabilistic Programs
Probabilistic programming has become a standard practice to model stochastic events and learn about the behavior of nature in different scientific contexts, ranging from Genetics and Ecology to Linguistics and Psychology. However, domain practitioners (such as biologists) also need to be experts in statistics in order to select which probabilistic model is suitable for a given particular problem, relying then on probabilistic inference engines such as Stan, Pyro or Edward to fine-tune the param…

RewardBench 2: Advancing Reward Model Evaluation
Saumya Malik, Valentina Pyatkin, Sander Land, Jacob Morrison, Noah A. Smith, Hannaneh Hajishirzi, Nathan Lambert
https://arxiv.org/abs/2506.01937

RewardBench 2: Advancing Reward Model Evaluation
Reward models are used throughout the post-training of language models to capture nuanced signals from preference data and provide a training target for optimization across instruction following, reasoning, safety, and more domains. The community has begun establishing best practices for evaluating reward models, from the development of benchmarks that test capabilities in specific skill areas to others that test agreement with human preferences. At the same time, progress in evaluation has not…

Evaluating LLMs Effectiveness in Detecting and Correcting Test Smells: An Empirical Study
E. G. Santana Jr, Jander Pereira Santos Junior, Erlon P. Almeida, Iftekhar Ahmed, Paulo Anselmo da Mota Silveira Neto, Eduardo Santana de Almeida
https://arxiv.org/abs/2506.07594

Evaluating LLMs Effectiveness in Detecting and Correcting Test Smells: An Empirical Study
Test smells indicate poor development practices in test code, reducing maintainability and reliability. While developers often struggle to prevent or refactor these issues, existing tools focus primarily on detection rather than automated refactoring. Large Language Models (LLMs) have shown strong potential in code understanding and transformation, but their ability to both identify and refactor test smells remains underexplored. We evaluated GPT-4-Turbo, LLaMA 3 70B, and Gemini-1.5 Pro on Pyth…