Hasty Briefsbeta

All tags

#python

387 stories total

Bilingual

Python lib generates its code on-the-fly based on usage
a year ago
- AutoGenLib is a Python library that dynamically generates code using OpenAI's API when importing non-existent modules or functions.
- Features include dynamic code generation, context-aware function creation, progressive enhancement, and automatic exception handling.
- No default caching ensures varied and creative results with each import, though caching can be enabled for consistency.
- Requires Python 3.12+ and an OpenAI API key set in the OPENAI_API_KEY environment variable.
- The library analyzes the importing code's context to generate relevant functions, ensuring compatibility with existing data structures.
- Includes optional initialization for hinting the library's purpose and supports runtime caching toggling.
- Generated code can be inspected using Python's inspect module, and prompts to OpenAI include comprehensive context for better consistency.
- Not suitable for production-critical code without review, intended for prototyping and experimentation.
- Contributions are not accepted as it's a proof-of-concept project, released under the MIT License.
PyPI Organizations
a year ago
- PyPI introduces organization accounts for better financial support and sustainability.
- Organizations on PyPI allow self-managed teams with branded web addresses.
- Free for community projects; corporate projects require a small fee.
- PyPI served 235.7 billion downloads, showing 57% annual growth.
- Increased revenue will help PyPI become a staffed platform for better support.
- New organization features are optional and won't affect existing users.
- Future features will focus on safety, security, and usability based on user feedback.
- Community and corporate teams can sign up for organization names now.
- Development funded by the Python Software Foundation's sponsorship program.
- Special thanks to Bloomberg and beta testers for their support.
LPython: Novel, Fast, Retargetable Python Compiler (2023)
a year ago
- LPython is a Python compiler that converts type-annotated Python code into optimized machine code.
- It supports multiple backends including LLVM, C, C++, WASM, Julia, and x86.
- Features include Ahead-of-Time (AoT) compilation, Just-In-Time (JIT) compilation, and interoperability with CPython.
- LPython uses Abstract Semantic Representation (ASR) for intermediate optimizations, shared with LFortran.
- Benchmarks show LPython's competitive performance against Numba and C++ in various tasks.
- Supports machine-independent optimizations like loop unrolling, vectorization, and dead code removal.
- JIT compilation in LPython is achieved using decorators, with backend-specific optimization flags.
- Interoperability with CPython allows calling functions from libraries like NumPy and Matplotlib.
- AoT compilation examples demonstrate LPython's speed in generating binary outputs.
- Visualization examples include generating Mandelbrot sets using Matplotlib via LPython.
CPython discussion on official support for adding virtual threads
a year ago
- Java has virtual threads, which are a better concurrency model than Python's async/await.
- Virtual threads combine the benefits of normal threads (no new syntax, intuitive execution) and async tasks (lightweight, structured concurrency).
- Unlike Python's coroutines, virtual threads do not split the language into 'colors' (sync vs. async).
- Virtual threads can be implemented in Python using continuations for context switching and non-blocking calls.
- Continuations in Python would allow stackful and symmetric coroutines, enabling better context switching.
- Non-blocking calls in virtual threads can be achieved by managing async operations via an event loop continuation.
Microsoft fires Faster CPython developers
a year ago
- Layoffs at Microsoft affected 3 Python core developers from the Faster CPython team.
- Affected developers include Eric Snow, Irit Katriel, and Mark Shannon.
- Requests for job opportunities for the affected developers are encouraged.
I Don't Like NumPy
a year ago
- NumPy is widely used for array computations in Python and influences machine learning libraries like PyTorch.
- NumPy makes simple operations elegant but struggles with more complex array manipulations, especially in higher dimensions.
- Broadcasting in NumPy can lead to non-intuitive behavior and requires careful dimension alignment.
- Advanced indexing in NumPy is complex and often leads to confusion, even among experienced users.
- Functions like np.linalg.solve have unclear conventions for handling multi-dimensional arrays.
- np.einsum is praised for its clarity and power in handling complex array operations.
- NumPy's design flaws stem from replacing explicit indices with broadcasting, leading to readability and usability issues.
- Multi-head attention in NumPy demonstrates the challenges of vectorizing operations without clear abstractions.
- The author proposes a prototype for a better array language that addresses NumPy's shortcomings.
Microsoft lays off faster CPython team
a year ago
- Microsoft canceled support for the Faster CPython project, leading to layoffs.
- The Python Language Summit proceeded despite the news, showcasing open-source resilience.
- Offers of support and job opportunities were extended to affected team members.
- Mike hinted at alternative open-source projects for speeding up Python.
- Various companies expressed interest in hiring the laid-off engineers.
Pyrefly:A fast type checker/IDE extension for Python written in Rust
a year ago
- Pyrefly is an open-source Python type checker and IDE extension written in Rust.
- It offers static type checking to ensure type consistency and catch errors before runtime.
- Supports IDE integration (e.g., VSCode) and CLI usage for flexible workflows.
- Built to handle large codebases with high performance (1.8M lines/sec).
- Designed with IDE-first principles for real-time feedback and type inference.
- Open-source under MIT license, encouraging community contributions via GitHub and Discord.
- Evolved from Pyre, Meta's earlier type checker, to address modern Python typing needs.
- Future goals include improving Python's type system and developer experience.
- Alpha release available now, with plans to remove the alpha label by Summer.
The first year of free-threaded Python
a year ago
- CPython 3.14.0b1 was released, marking progress in free-threaded Python.
- Free-threaded Python aims to utilize multicore CPUs and GPUs more effectively by removing the GIL limitations.
- Quansight contributed to enabling free-threaded Python in key tools and packages like NumPy, SciPy, and PyArrow.
- Significant improvements in CPython 3.14 include thread-safety in warnings, asyncio, and ctypes modules.
- A comprehensive guide for supporting free-threaded Python in existing packages has been created.
- The ecosystem has improved, but many packages still need auditing for thread safety.
- Challenges include performance tuning and maintaining legacy packages.
- Community involvement is encouraged via contribution guides and a Discord channel.
- A PyCon 2025 talk will share insights on porting packages to support free-threading.
Show HN: SQL-tString a t-string SQL builder in Python
a year ago
- SQL-tString enables t-string based SQL query construction without SQL injection risks.
- Basic usage involves passing a t-string to the `sql` function, which returns a query string and a list of values.
- Parameters in t-strings must be simple variable identifiers; expressions like `{a - 1}` are invalid.
- `sql_context` allows pre-defining valid column and table names, raising errors if values don't match.
- Special values like `Absent` remove parameters from queries, useful for optional updates or conditionals.
- `IsNull` and `IsNotNull` rewrite conditionals to handle SQL NULL values correctly.
- Supports `qmark` and `asyncpg` paramstyles, configurable via `set_context`.
- Compatible with Python 3.12 and 3.13, using standard strings if t-strings (PEP 750) are unavailable.
Community Stewardship of Faster CPython
a year ago
- Microsoft's support for Faster CPython has been cancelled, impacting Python's performance improvements.
- Python 3.14 is 20-40% faster than 3.10, thanks to contributions from various individuals and organizations.
- Options moving forward include finding another sponsor, shutting down experimental work, or continuing with community stewardship.
- Proposal to transfer Faster CPython to a community-owned model, forming a performance working group (WG) without decision-making power.
- Establish a PSF-owned benchmarking infrastructure and explore JIT design choices prioritizing maintainability over peak performance.
- Form a community council for conflict resolution and technical direction, modeled after the typing council (PEP 729).
- Nomination of initial community WG members, emphasizing diverse backgrounds and contributions to the project.
- Potential for community-led project success, citing examples like PyPy and existing contributors.
- Proposal for an informal WG to facilitate collaboration and discussion on CPython performance, replacing previous corporate-led mediation.
Building software on top of Large Language Models
a year ago
- Presented a three-hour workshop at PyCon US titled 'Building software on top of Large Language Models'.
- Workshop included interactive exercises with a detailed handout covering six topics.
- Topics included setup, prompting, Python API, text-to-SQL, structured data extraction, semantic search, and tool usage.
- Handout is available for self-study, even without attending the workshop.
- Some sections involved slides and presentations, now compiled into an annotated format.
- Private workshop versions can be arranged upon request.
Project Verona: Fearless Concurrency for Python
a year ago
- Project Verona is developing a new ownership model for Python called Lungfish to manage memory and concurrency safely.
- A toy language, FrankenScript, was created to prototype region-based ownership ideas dynamically.
- Engagement with the Python community includes discussions with the Faster CPython team and a presentation at the Python Language Summit.
- Steps towards a new ownership model start with deep immutability, divided into three parts: deep immutability, managing cyclic immutable garbage, and integration with message passing between sub-interpreters.
- The ownership model aims to support Python's dynamic typing and existing object graphs, differing from Rust's static approach.
- Project Verona has learned from dynamic ownership research, challenging static assumptions and exploring new programming patterns.
- Resources include publications, FrankenScript for experimentation, and Python forks on GitHub for those interested in contributing.
Show HN: Buckaroo – The data table UI for Notebooks
a year ago
- Buckaroo is a modern data table for Jupyter that simplifies exploratory data analysis.
- Features include a performant data table, sorting, value formatting, infinite scrolling, summary stats, histograms, smart sampling, auto-cleaning, and a low-code UI.
- Installation is simple with `pip install buckaroo` and works with Jupyter Lab, Jupyter Notebook, Marimo, VS Code notebooks, Jupyter Lite, and Google Colab.
- Supports DataFrame libraries like pandas, polars, and geopandas (optional).
- Includes extensive documentation, tests, and feature example videos on YouTube.
- Interactive styling gallery allows live editing and configuration.
- Core data grid is based on AG-Grid, enabling fast loading and customizable display.
- Numeric columns are formatted with fixed-width fonts and commas for better readability.
- Histograms and summary stats provide quick overviews of data distribution.
- Low-code UI with Python code generation and heuristic autocleaning for data.
- Pluggable Analysis Framework allows customization of summary stats.
- Development setup involves cloning the GitHub repo and installing dependencies.
- Contributions are welcome, with guidelines for issue reporting and code contributions.
Yahtzeeql – Yahtzee solver that's mostly SQL
a year ago
- Create a Python virtual environment named 'yahtzee-venv' and activate it.
- Install dependencies from 'requirements.txt' and run tests with './test.sh'.
- Execute 'yahtzee.py' with various strategies like 'random', 'random_greedy', 'all_yahtzee', 'prob', and 'prob_with_difficulty'.
- Use '--runs' to specify the number of games to simulate, '--interactive' for manual category selection, and '--show-plot' to visualize score distribution.
- Generate a SQLite database with 'build_prob_db.py' containing tables for all rolls, roll transitions, and expected values for strategies.
- Average scores for strategies: 'random' (84.5), 'random_greedy' (86.5), 'all_yahtzee' (87.5), 'prob' (88.5), 'prob_with_difficulty' (89.5).
Show HN: Cogitator – A Python Toolkit for Chain-of-Thought Prompting
a year ago
- Cogitator is a Python toolkit for experimenting with chain-of-thought (CoT) prompting in large language models (LLMs).
- CoT prompting improves LLM performance on complex tasks by generating intermediate reasoning steps before the final answer.
- The toolkit supports OpenAI and Ollama as LLM providers and includes Pydantic validation for structured model output.
- It provides a customizable benchmarking framework and implementations of popular CoT strategies like Self-Consistency.
- Installation can be done via pip or by cloning the GitHub repository and setting up a Python environment with Poetry.
- Example usage includes running Self-Consistency CoT with Ollama, demonstrated with a sample question and prompt.
- The project includes documentation, examples, and a benchmarking framework for evaluating CoT strategies on datasets like GSM8K.
- Contributions are welcome, and users are encouraged to cite the project if used in research.
- The toolkit is licensed under the MIT License.
Python Design Patterns (2020)
a year ago
- Brandon Rhodes is creating a guide on Python design patterns.
- The guide aims to consolidate his ideas about Python and design patterns in one place.
- The goal is to make design patterns more discoverable and easier to read than in scattered conference materials.
- Progress on the guide is intermittent due to other obligations.
- New updates can be tracked via the GitHub commit history or by selecting 'Watch' on the repository.
Bayesian Modeling and Computation in Python (2021)
a year ago
- Online version of 'Bayesian Modeling and Computation in Python' available.
- Physical copy can be purchased from the publisher or Amazon.
- Site includes all code used in the book, including for figures and tables.
- Updated code available in 'notebooks_updated' folder for latest library versions.
- Citation provided for using the book in academic work.
- Instructions for reporting mistakes via filing an issue.
- Environment setup guides provided for Conda and Google Colab.
- Encouragement to donate to ArviZ or PyMC for supporting open source.
- Book licensed under CC BY-NC-SA 4.0, code under GNU GPLv2.
Python for Mathematics
a year ago
- Book titled 'Python for Mathematics' covering various mathematical topics.
- Structured with tutorials, exercises, solutions, and further information sections.
- Topics include Algebra, Calculus, Matrices, Combinatorics, Probability, Statistics, Differential Equations.
- Includes building tools like Variables, Functions, OOP, Command Line usage, Modularisation, Documentation, Testing.
- Further information on Numpy, Matplotlib, Kernels, Installing libraries, Linear Programming, Working with files.
- Additional resources: Bibliography, About the book, Contributing, Citation, Software packages used.
Python Tooling at Scale: LlamaIndex’s Monorepo Overhaul
a year ago
- LlamaIndex manages a monorepo with over 650 Python packages, primarily Integrations and Packs.
- Each package has its own pyproject.toml file, defining dependencies, versioning, and tool configurations.
- The core package, llama-index-core, is a dependency for most other packages in the monorepo.
- Initial tooling included Poetry for individual package management and Pants for build orchestration.
- Challenges included slow dependency resolution, complex CI debugging, and high maintenance costs.
- The team migrated to uv (a faster Python project manager) and developed LlamaDev for monorepo-specific tasks.
- LlamaDev improved test speeds, logging clarity, and local reproducibility, reducing CI feedback loops.
- Contributors can now easily run LlamaDev locally, enhancing the development experience.

first prev3next

About|Login

#python

Python lib generates its code on-the-fly based on usage

PyPI Organizations

LPython: Novel, Fast, Retargetable Python Compiler (2023)

CPython discussion on official support for adding virtual threads

Microsoft fires Faster CPython developers

I Don't Like NumPy

Microsoft lays off faster CPython team

Pyrefly:A fast type checker/IDE extension for Python written in Rust

The first year of free-threaded Python

Show HN: SQL-tString a t-string SQL builder in Python

Community Stewardship of Faster CPython

Building software on top of Large Language Models

Project Verona: Fearless Concurrency for Python

Show HN: Buckaroo – The data table UI for Notebooks

Yahtzeeql – Yahtzee solver that's mostly SQL

Show HN: Cogitator – A Python Toolkit for Chain-of-Thought Prompting

Python Design Patterns (2020)

Bayesian Modeling and Computation in Python (2021)

Python for Mathematics

Python Tooling at Scale: LlamaIndex’s Monorepo Overhaul