I think it might be a good idea to add a new wrapper method for this in spiceerrors called MustSafecast. Have it take in the generic type, call the safecast here, then if an error occurs, have it call spiceerrors.DebugAssert to fail with a panic at test time (but no-op at prod time) and then return the default value given to the function. That way, we have documented behavior everywhere.

I can see methods like what you describe here in spiceerrors.

So I could implement something, but first, I would like to check things with you guys, at least the maintainers:

• is there anyone else than Josh who wants to do it the way he described?
• do you expect me to do this ?

I'm asking because I'm unsure I'm the right person to update this.

Just in case, my lib has two methods you could consider:

• MustConvert that panics on failure
• RequireConvert that calls t.Error on failure and so invalidate the tests

Here I feel none of these matches what you described.

Also, I don't understand the need to panic on tests and passes on "production"
If the code differs, it cannot be tested with the same way. I dislike the idea.

That said, I can adapt my PR. But it goes a bit too far from what I expected to do on your project as the owner and maintainer of go-safecast.

I created this PR to facilitate the migration from v1 to v2 for you. I didn't plan to invest more.

Please let me know.

We have a general rule of not allowing panics at production time; instead, we prefer to return errors and panic only during tests to identify bugs but not cause production crashes.

We can change the behavior in a follow, I suppose

I would prefer you to apply something that suits your habit instead of having a PR that I have to update possibly iteratively because my changes might not face what you are expecting in term of changes.

Please note that go-safecast Convert method doesn't panics either in v1 or v2

So if you have an issue with the way you use my lib with v2, you had it in v1