fix: mapping to preserve statically known types

[buschtoens]

Fixes #16.

[sindresorhus]

It sounds like from #16 that the existing type should be changed instead of just adding an overload.

[sindresorhus]

Friendly bump :)

[buschtoens]

Holy fuck — how do you manage to keep track of stale PRs? 😲

I'll try to get to it later today. Thanks for the ping. 😊

[sindresorhus]

how do you manage to keep track of stale PRs?

This one had open in an old tab. But I check https://github.com/search?o=asc&q=is:pr+is:open+archived:false+user:@me&s=updated&type=Issues once in a while.

[buschtoens]

Looking at this again, I think the original types need to be kept in order to keep full support for mapper. My type overload only kicks in, if no mapper was passed, as I don't think that there is a way to allow the mapper to be taken into consideration, when the return type is constructed.

I may be wrong though.

But as it stands, this PR already significantly improves the usability of the types for the happy path (no mapper). So even if the mapper could be somehow taken into consideration, I think this PR would still be good to merge as-is.

What do you think?

/cc @novemberborn in case you are interested as well

[buschtoens]

Tests are failing (yay!) and I will try to look at them locally some time today.

tbh this PR started as a mere drive-by quick fix via the GitHub web UI, but it seems I need to grant it some time in a real IDE. 😄

[sindresorhus]

Bump :)