[Discussion] Error instead of panic when unknown type

[Soulou]

In the different validators, the functions panic if the type is unknown, I think it would be more go-compliant to return an error. According to me, if the type is not supported, it should be simply a validation error.

What do you thing of that?

[rselbach]

The reason I did it this way is that I see the tags as being an integral
part of the struct definition and to be honest I wish I could make them
cause a syntax error at compile time if I could. Since I couldn't, I
thought they should fail as hard and soon as possible.

That said, I did this before I ended up exporting a second function that
allows rules as a function parameter, so yes, I can see how a panic might
not be appropriate there. I might have to change that.