both seem fine in principle. i think the implementation doesn’t adhere to transducer rules since state is created when you call map-prev instead of when it’s connected to a process

(def step (map-prev vec)) you're supposed to be able to reuse step, but the way it's currently written would reuse the same state

Oh yeah, that makes sense, so I guess that shortcut doesn't work and I need to actually put the state inside the outermost transducer function

but both seem useful

based on the usage, not sure if it would be better to have map-prev act more like: (map f coll (rest coll))

which only looks at values that have previous values

oh yeah that might be better, gotta think about that a bit

map-prev looks a lot like reductions 🙂