

The full name of the GHCNd dataset is the Global Historical Climatology Network daily. It is a global observation dataset publicly released by NOAA, with data starting from 1763, containing daily meteorological observation data from a total of 120,000 weather stations worldwide. The monthly summary version of the GHCNd dataset is the Global Historical Climatology Network monthly. During the weekend, I performed some statistical analysis on this dataset, and I’d like to record the general processing workflow. ...
A few months ago, I wanted to map Japanese coordinate data to administrative division levels. To achieve this, I developed a reverse geocoding tool reversejp using administrative boundary data from the Japan Meteorological Agency (JMA). The core functionality is implemented in Rust, and I packaged it as a Python library using PyO3 and Maturin. Although Japan’s land area is not very large, the JMA dataset subdivides the country into over 3,000 administrative units at its finest level. Considering Japan’s total land area, this results in extremely fine-grained coverage. ...
A local debugging MongoDB proxy
A simple example of how to group world cities by timezone
Developed and released pg-tzf, a PostgreSQL extension providing microsecond-level latitude/longitude to time zone conversion, based on the Rust library tzf-rs and the pgrx framework.
A record of blog improvements, including multilingual support, SEO optimization, and a personalized logo generated by ChatGPT.
A guide to implementing short URL redirects in Hugo static sites, featuring both manual HTML redirect pages and an automated Go-based solution for generating redirect pages.
Create a file: layouts/_default/_markup/render-image.html: 1 2 3 4 5 6 <figure class="align-center "> <img loading="lazy" src="{{ .Destination | safeURL }}#center" alt="{{ .PlainText }}" > {{ with .Text }} <figcaption><p>{{ . | safeHTML }}</p></figcaption> {{ end }} </figure> Then magic happens: 1  Render as: sample
Some historical fragments discovered while researching.
This article introduces two major pain points in Python project maintenance and their solutions: 1) Passwordless publishing using PyPI’s Trusted Publishers with GitHub Actions, and 2) Ultra-fast dependency locking with the uv tool. It also discusses version management when depending on machine learning frameworks like PyTorch, providing specific configuration examples.