fix: fix fallback not triggered for 0 measuredWidth

[wojtekmaj]

Currently fallback does not work: https://npmx.dev/api/registry/badge/version/nuxt?label=thisisatest

with this PR, fallback still sucks but at least does not appear very broken:

https://npmxdev-git-improved-fallback-npmx.vercel.app/api/registry/badge/version/nuxt?label=thisisatest

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
npmx.dev	Ready	Preview, Comment	Feb 22, 2026 8:25pm

2 Skipped Deployments

Project	Deployment	Actions	Updated (UTC)
docs.npmx.dev	Ignored	Preview	Feb 22, 2026 8:25pm
npmx-lunaria	Ignored		Feb 22, 2026 8:25pm

[Copilot]

Pull request overview

Updates badge SVG text measurement logic so the character-count fallback is triggered when the canvas text measurement returns 0, preventing badges from rendering with an incorrectly small width in that scenario.

Changes:

• Treat measuredWidth === 0 as an invalid measurement by requiring Number.isFinite(measuredWidth) && measuredWidth > 0 before using it.

[Copilot]

This change adjusts the fallback behavior when measureText() returns 0, but there’s no regression test covering the measuredWidth === 0 case (which is the bug being fixed). Consider adding a Vitest test that mocks @napi-rs/canvas (or getCanvasContext()) to return width: 0 for a non-empty string and asserts that the computed badge/text width uses the character-count fallback (i.e., grows beyond the minimum width for longer strings).

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ All tests successful. No failed tests found.

📢 Thoughts on this report? Let us know!

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

---

📝 Walkthrough

Walkthrough

This PR modifies the validation logic in the measureTextWidth function within the badge handler. The conditional check is tightened to verify that measuredWidth is both finite and positive (Number.isFinite(measuredWidth) && measuredWidth > 0) before returning the ceiling value. Previously, the function would return ceil(measuredWidth) as long as it was not NaN. The updated logic now explicitly excludes Infinity, negative Infinity, and non-positive values, with potential null return otherwise.

Possibly related PRs

• fix: improve badge text measurement accuracy #1486: Modifies the same measureTextWidth implementation in the badge handler, replacing the measurement approach with Canvas-based measurement
• fix: badge label with empty string should not occupy width #1095: Updates the same badge handler and affects label width computation logic, which interacts with the tightened measureTextWidth validation

Suggested reviewers

• danielroe
• ghostdevv

🚥 Pre-merge checks | ❌ 1

❌ Failed checks (1 inconclusive)

Check name	Status	Explanation	Resolution
Description check	❓ Inconclusive	The PR description is empty except for template content; the actual changes are not described by the author, only by the AI summary.	Provide a meaningful description explaining the bug fix, why measuredWidth of 0 requires a fallback, and any relevant test coverage or edge cases addressed.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing Touches

• 📝 Generate docstrings (stacked PR)
• 📝 Generate docstrings (commit on current branch)

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment
• Commit unit tests in branch improved-fallback

Tip

Issue Planner is now in beta. Read the docs and try it out! Share your feedback on Discord.

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}