fix: Use strict typing in EvaluationEngine and fix evaluation boolean handling

[tyiuhc]

Issue for reference.

Changes:

1. Update EvaluationEngine to use strict typing
2. Update EvaluationEngine to properly handle different boolean representations

[promptless]

✅ No documentation updates required.

[github-actions]

🎉 This PR is included in version 1.2.3 🎉

The release is available on GitHub release

Your semantic-release bot 📦🚀

[ruudk]

This breaks the old behavior where you had to use "1" or "0" for it to work.
So this fix is a breaking change 🚨

When I deployed this our feature flags stopped working.

I think booleans should still work with 1 or 0.

[tyiuhc]

@ruudk, thanks for raising this. To clarify, are you using "1" and "0" in your flag definition to support boolean matching?

[ruudk]

Yes and in PHP I feed a real Boolean. The old situation would convert the bool to "1" and thus it would match.

With your fix you broke that, and require me to match against true in amplitude. Fine, but others will get hit by this !

[tyiuhc]

@ruudk thank you for the clarification. To ensure the results of local and remote evaluation are the same, we recommend that existing flag configs using 1 and 0 for boolean targeting are updated to use true and false. Apologies for the unexpected behavior this has caused.

[ruudk]

Sure, I can change my Feature Flag conditions in Amplitude. But my point is, that you broke existing behavior in a patch version. You should be more careful next time. Especially with something important as what you're offering. People rely on the feature flags to keep on working, after a patch release. Your answer doesn't give me confidence for future updates tbh.