static-analysis – Page 5 – Türker Yıldırım

The generic PHP does not have, and the analyser that does

Deep diveFebruary 27, 20239 min read

A collection class that returns mixed, and forty call sites asserting the type by hand. Docblock generics, and where they stop being worth it.

PHPUnit 10, and the listeners that no longer exist

Deep diveFebruary 20, 202314 min read

10.0 in February, a configuration file that has accreted since 2014, and a custom TestListener that the new version has no place for.

Extracting a module without extracting a service

Deep diveFebruary 3, 202314 min read

A billing module everybody agrees should be separate, a team of four, and 88 classes that reference it. The boundary, drawn in code rather than in conversation.

declare(strict_types=1) in every file, finally

NoteJanuary 30, 20232 min read

The codebase had it in about 60% of files, which is the worst of the three possible states because nobody can tell which behaviour applies where. The seven…

DNF types, and the signature that needed parentheses

NoteDecember 28, 20221 min read

8.1 gave pure intersection types and no way to make one nullable; 8.2 allows a union of intersections, which covers the case everybody hit first. The name describes…

Readonly classes and the DTO that stopped needing a constructor

Deep diveDecember 20, 202210 min read

8.2 on 8 December, and readonly applied to a whole class. readonly on every property, written out forty times.

The dynamic property that had been a typo since 2016

Deep diveDecember 13, 202214 min read

8.2 on 8 December deprecates dynamic properties, and the deprecation is the feature. $order->totl has been silently creating a property for six years.

A monorepo for four packages and one application

Deep diveOctober 5, 202210 min read

Four packages, four repositories, and a change that spans three. A version bump dance that takes an afternoon.

PHPUnit 9 to Pest, and the migration that was not worth it

Deep diveSeptember 30, 202210 min read

A trial on one suite, honestly reported. Nothing was wrong with the tests we had.

An ignoreErrors entry needs a reason, a baseline entry does not

NoteSeptember 2, 20221 min read

A baseline entry is historical debt and needs no justification; an ignore is a claim that the analyser is wrong and does. reportUnmatchedIgnoredErrors is on by default and…