bug: canonical-link false positive on root URLs with multi-character TLDs (.app, .io, .dev, etc.)

[jettwayio]

Bug report

The nice-checkers/canonical-link rule incorrectly flags root canonical URLs when the TLD is multi-character (.app, .io, .dev, .ai, etc.).

Root cause

The extension check runs against the full URL string:

if (/\.\w+$/.test(href)) {

https://webscore.app matches because .app satisfies \.\w+$, even though the pathname has no extension.

Suggested fix

Test only url.pathname instead:

const url = new URL(href)
if (/\.\w+$/.test(url.pathname)) {

https://webscore.app has pathname /, so it passes. https://example.com/page.html has pathname /page.html, so extensions are still caught correctly.

Relates to #26

[jettwayio]

Also worth noting: the test suite has a coverage gap that allowed this bug to go undetected.

The only fixture testing a valid canonical (no-errors.html) uses http://localhost:9876/status/200 as the href which ends with /200, so the regex never fires against a TLD.

A fixture like this should be added to catch this case:

<link rel="canonical" href="https://example.com" />

Expected: no errors. Currently: false positive.

This would have caught the bug before it shipped.

[fulldecent]

Thank you for sharing. The first step here is to add that (failing) test case.

[jettwayio]

@fulldecent I've created PR, appreciate to take a look.

[fulldecent]

Thanks for sharing. We have a fix here in main. Not released yet. Can you please test in your own project and confirm before we release it?