Why Whoosh Search Returns Zero Results and How the Analyzer Fixes It

·1 views

Whoosh is a pure-Python full-text search library where the analyzer — a pipeline of tokenizers and filters — determines how text is converted into searchable tokens at both index and query time. When the indexing and querying sides process text differently, searches return no results even when matching documents exist. Using a StemmingAnalyzer instead of the default StandardAnalyzer can resolve this, as it reduces word variants like 'connections' and 'connecting' to a common root, allowing a query for 'connect' to match both. Developers can debug analyzer behavior by running it directly on a string to inspect the tokens produced, without needing a full index. Additional filters such as CharsetFilter with an accent map can further improve matching by normalizing accented characters to their plain ASCII equivalents.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

React 19.3 Officially Releases with Stable View Transitions and Fragment Refs

React 19.3 was published to npm on September 9, 2026, promoting two long-experimental features — View Transitions and Fragment Refs — to stable status. View Transitions allow React to automatically animate elements as they appear, disappear, or move during state updates wrapped in startTransition. Fragment Refs introduce a FragmentInstance handle that lets developers attach ref-like behaviour to a group of sibling elements without needing a wrapper DOM node. The release also adds a new browser() utility to skip server-side rendering for components that rely on browser-only APIs, along with Trusted Types support for improved DOM injection security and native Context access inside Server Components. Several smaller bug fixes covering fullscreen events, form handling, focus behaviour, and independent transitions are also bundled in the update.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Backtest of 15,900 A-Share Stock-Days Finds Volume Confirmation Rule Unreliable

A developer backtested the popular Chinese retail trading rule '量价配合' (price-volume confirmation) using daily data from 584 liquid Shanghai and Shenzhen stocks, generating 15,900 stock-days without paid data vendors. The study found that while the rule appeared to hold in aggregate — strong-volume up days outperforming weak-volume up days — the edge collapsed when results were split by calendar month. Across three consecutive months in 2026, key setups including limit-up days and high-volume breakouts flipped between positive and negative returns, sometimes dramatically. The baseline next-day return across all stock-days was -0.063%, a benchmark most retail backtests omit, masking illusory edges. The author concludes that narrow, theme-driven market conditions — like the one observed on September 10, 2026, where only 20.7% of stocks advanced — are precisely when momentum rules appear most reliable yet are most likely to fail going forward.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Beginner Developer Seeks Guidance on Breaking Into AI Engineering

A beginner developer has posted on DEV Community asking for advice on how to get started in AI engineering. The individual is confused about the scope of the field and whether concepts like RAG, LoRA, fine-tuning, and LLMs represent its core areas or just a subset. They also have a background in low-level development, having studied C++, C, Assembly, and PostgreSQL. The developer is weighing whether to focus on AI, deepen their low-level skills, or pursue a combination of both. They are seeking guidance from experienced developers on how to prioritize their learning path given how rapidly AI is evolving.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Developer Open-Sources FastAPI Router to Unify Multiple LLM Provider APIs

A developer has released an open-source FastAPI template called fastapi-multi-llm-starter, designed to simplify integration with multiple large language model providers including Anthropic, OpenAI, Google, and DeepSeek. The tool exposes a single POST endpoint that accepts a unified request payload and returns standardized Server-Sent Events covering reasoning tokens, response text, tool calls, and stream completion. Model configurations are managed via an external JSON file, allowing new models to be added without modifying backend code. The project includes a built-in browser-based testing playground, pytest coverage, and Pydantic v2 validation schemas. It is available on GitHub under the MIT License, with users required to supply their own API keys.

0 comments Read more at DEV Community