Yes, that's most likely a leftover artifact from an earlier design... I will definitely take another look at this and reconsider. Unless there is another reason that currently escapes me, I agree it should not use Option.

The other thing I dislike about the Router but could not figure out a way around is the requirement to pass some data, even just unit type (), to satisfy the generic type parameter. Haven't had time to revisit that, but if you're digging around in that part of the code and happen to have an idea, I'm all ears!

See Tide's ::with_state(x) vs ::new() for a way to hide the () from the caller who doesn't care. It'll still show up in types, of course.

https://github.com/http-rs/tide/blob/1d6f120c9e6c0e723bb8929dc0f601cd4f9d1449/src/server.rs#L59-L61

@tv42 - of course, seems so obvious in hindsight :) Thanks for pointing that out.

The only downside is that we still need to add Clone to the bounds on D for the data. I am still trying to get some notion of a "global" for the router to be created in the global scope, which would likely require this anyways. (Worker's Isolates can share global state if they are in the same colo and under other conditions)

I've got a branch up with this change if you would like to take a look: f95b3ad

This has been updated in #50 - thanks for the feedback @tv42 !

The only downside is that we still need to add Clone to the bounds on D for the data.

For posterity, Clone was not added as a bound to D here. Instead, a &D is returned from ctx.data() and if you need ownership, you can implement Clone on your types via derive or manually.