Writing

How we replaced the SQS queue and Lambda function behind AB Tasty’s Decision API with a small in-process queue, kept the old path as a safety net, and what it changed.

Setting up Xdebug’s profile mode in a Dockerised PHP 8.1 environment, triggering it from an HTTP request or the command line, and reading the callgrind files with PHPStorm and KCachegrind.

Why I stopped maintaining jQuery Mapael, what the download numbers still look like, and what happens to a small Open Source project when its author walks away.

My first attempt at the Titanic Kaggle challenge in R: exploratory analysis of the passenger data, then logistic regression, QDA and KNN models compared with leave-one-out cross validation.

A beginner’s mini-study in R of twelve years of French road accident data: automating the import of 48 open datasets, then looking at trends by year, by hour, by day of the week, by age and on a map.

Trying to reproduce a South China Morning Post rainfall infographic for the city of Rennes, using Météo-France monthly bulletins stored in an RDF triplestore and plotted with ggplot2.

A list of things open data producers should avoid, illustrated with a real dataset: disappearing columns, renamed columns, duplicated years, wrong coordinates and postcodes that are not postcodes.

Pulling a dataset of film actors out of Wikidata with SPARQL, then turning each row into a written paragraph with the TextGenerator add-on for Google Spreadsheet.