Interpolation.piecewise sometimes fails tests

[Janiczek]

I was trying to use your test suite to stress-test my upcoming elm-explorations/test performance optimizations. Found a test was failing:

Running 191 tests. To reproduce these results, run: elm-test --fuzz 100 --seed 106499555100350

↓ InterpolationTests
↓ Interpolation
↓ piecewise
✗ never exceeds the range

Given (9007199254740996,[1],1)

    0
    ╷
    |atLeast
    ╵
    1

and

Running 191 tests. To reproduce these results, run: elm-test --fuzz 100 --seed 151805288162006

↓ InterpolationTests
↓ Interpolation
↓ piecewise
✗ never exceeds the range

Given (9007199254740994,[-1],1)

    -2
    ╷
    |atLeast
    ╵
    -1

[gampleman]

Yes these issues do seem to be there, but they do tend to show up at relatively extreme value ranges, so I suspect some floating point nastiness.

If someone can figure out how to fix this, I would be greatly obliged.

[Janiczek]

Would something like this make sense?

 piecewise : (a -> a -> Interpolator a) -> a -> List a -> Interpolator a
 piecewise makeInterpolator head tail =
     let
         n =
             List.length tail
+            
+        last =
+            List.Extra.last tail
+                |> Maybe.withDefault head
 
         interpolators =
             List.foldl (\item ( prev, accu ) -> ( item, makeInterpolator prev item :: accu )) ( head, [] ) tail
                 |> Tuple.second
                 |> List.reverse
                 |> Array.fromList
     in
     \t ->
+        if t <= 0 then
+            head
+
+        else if t >= 1 then
+            last
+
+        else
         let
             tn =
                 t * toFloat n
 
             i =
                 clamp 0 (n - 1) (floor tn)
         in
         Array.get i interpolators
             |> Maybe.map (\fn -> fn (tn - toFloat i))
             |> Maybe.withDefault head