Shouldn't typechecking have caught this bug?

The old function returned back an array of [string, number], where string was the Type and the number was the confidence pair. The new one follows the naming more literally and returns the 'Type' string. So taking the first index of a string technically means it is still a string just of length 1. In the example email for the bug all of the types that were visible started with the same letter 'p' so all color mapping was mapped to the same color. So it wasn't using the full name of the type for colors.

Shouldn't typechecking have caught this bug?

The old function returned back an array of [string, number], where string was the Type and the number was the confidence pair. The new one follows the naming more literally and returns the 'Type' string. So taking the first index of a string technically means it is still a string just of length 1. In the example email for the bug all of the types that were visible started with the same letter 'p' so all color mapping was mapped to the same color. So it wasn't using the full name of the type for colors.

I've been watching too many domain-driven design youtube videos, so my immediate thought is whether we can use the type system to enforce better constraints on these kinds of values. Something like TypeScript's type predicates or something like that... just something to think about.