Move controversial instances to `Data.String.Conversions.Unsafe`? about string-conversions

[fisx]

yeah, not worth it. (-:

it might improve the library marginally, but i decided to not spend any more time on the idea.

thanks for the helpful feedback!

[soenkehahn]

Hi. :)

I'm not 100% sure but I think this wouldn't work well in practice, because of how instances are imported (or scoped) in Haskell. So if any of your transitive dependencies imports the Unsafe module, you'll get all the unsafe instances without importing them yourself. I think.

There could be a separate type-class for safe conversions though.

Apart from that I think that I would prefer to keep the existing behavior of Data.String.Conversions. I'm not completely against adding a Data.String.Conversions.Safe module that only exposes safe conversions however.

To be honest, I'm not fully convinced yet that this change is worthwhile though. Do you have a concrete example of a project where you do have ByteStrings that you are tempted to convert to either String or Text using cs, but where you would opt for a safer api? Or some other concrete example that motivates this change from your perspective?

[fisx]

No concrete use case, I just think that if there can be an error, that should be obvious from the types. If you want to make a case for using a library in security-sensitive code, it's nice if you don't have to admit that it sometimes it eats bytes from your strings when you clearly don't want it to.

You are right though, my idea was wrong. I like the extra type class. I'll think about it some more.