SShortSingh.
Back to feed

Eight-language HTML parser benchmark: Python and Rust lead, Go trails Java

0
·3 views

The author ported their HTML parser library to eight languages and benchmarked each on the same real-world pages. Tests ran one at a time on identical hardware, an AMD Ryzen 7 3800X, under Linux and Windows. Parsing a 51 KB books page showed Python (with Rust underneath), C#, Rust and Java finishing within the margin of error of each other. Go was slower, and Node.js, PHP and Ruby trailed well behind. Differences under about 10 percent were treated as ties because repeated runs varied that much.

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)

Log in to join the discussion and vote.

Log in

Related stories

0
ProgrammingDEV Community ·

Browser Back Button Testing: Can a Peer Return to the Same SaaS Task?

The article argues that Back button testing should verify whether users can resume the same useful task, not just whether the URL changes. In a SaaS review queue, that means the filter, sort order, selected item and list position should survive a return. Teams should define which state must persist before asking a peer to test a prototype. The author uses a fictional request-review app to illustrate a user opening a record and pressing Back to return to the filtered, sorted queue. The piece also separates essential state, such as filters, from less important details like exact scroll position, and suggests storing shareable state in the URL while treating unsent notes separately.

0
ProgrammingDEV Community ·

How to use AI tools without exposing customer data

AI chatbots such as ChatGPT, Claude and Gemini store what users send on the provider's servers, and retention periods and training use vary by plan. A March 2023 bug in ChatGPT exposed some other users' chat titles and the billing details of about 1.2% of Plus subscribers during a nine-hour window. Data can also leak when an employee enters customer or patient information into a personal chatbot account, which the Dutch regulator treats as a data breach. The recommended safeguards are replacing personal details with codes before sending, running models in-house, sharing only aggregated figures with the strongest models, and giving staff business accounts with clear rules. Terms can change, so organisations should check each provider's current policy.

0
ProgrammingDEV Community ·

Open-source project explores agent-based coordination for podcast sponsorship briefs

A developer is building Agent Market, an open-source service-discovery and task-coordination project, and is considering whether it could support an alternative to podcast sponsorship marketplaces at the discovery and coordination layer. The article describes a design scenario using fictional brands and creators, in which agents would reconcile episode timing, scripts, disclosures, revisions and delivery evidence before a human approves anything. Proposed checks include verifying the host-read segment, approved wording, disclosure and publication window, with discrepancies flagged for review rather than accepted silently. The author stresses that the experiment has not been run, that current trials are zero-price, and that payments, escrow and audience attribution are not implemented. The piece also notes that transcripts and live URLs do not fully prove an ad was heard or audience size, and it closes by asking which part of the process creators or brands would save the most effort on.

0
ProgrammingDEV Community ·

Two-label hostname shortcut merges unrelated .co.jp sites; ad-filter tool switches to Public Suffix List

A developer building a browser tool that checks ad-blocker filter rules found that the $third-party option produced wrong results. Their function treated the last two hostname labels as the site, so every .co.jp domain was grouped as one party. Testing in Node.js 24 showed that suffixes like co.jp, co.uk and github.io needed special handling, while a bare public suffix should have no owner. The author concluded that the Public Suffix List, which covers thousands of entries with wildcard and exception rules, is required for reliable third-party checks. The write-up was produced with AI assistance and describes the author's own side project.