interpolate can handle all sorts of labels

src/MWT.jl

@@ -76,7 +76,7 @@ interpolate(x1::T, x2::T, lambda) where {T<:AbstractFloat} = (1 - lambda) * x1 +
 interpolate(x1::T,
             x2::T,
             lambda,
-           ) where {T<:AbstractArray} = @. (1 - lambda) * x1 + lambda * x2
+            ) where {T<:AbstractArray{<:AbstractFloat}} = @. (1 - lambda) * x1 + lambda * x2
 function interpolate(geometry1::T, geometry2::T, lambda) where {T<:Larva.SpineGeometry}
     x1, y1 = collect.(zip(coordinates.(Larva.vertices′(geometry1))...))
     x2, y2 = collect.(zip(coordinates.(Larva.vertices′(geometry2))...))
@@ -84,7 +84,8 @@ function interpolate(geometry1::T, geometry2::T, lambda) where {T<:Larva.SpineGe
                    Larva.PointSeries(interpolate(x1, x2, lambda),
                                      interpolate(y1, y2, lambda)))
 end
-function interpolate(tags1::T, tags2::T, lambda) where {T<:Larva.AbstractTags}
+function interpolate(tags1, tags2, lambda)
+    # default nearest-neighbor implementation for tags/labels
     lambda <= 0.5 ? tags1 : tags2
 end
 function interpolate(state1::T, state2::T, lambda) where {T<:NamedTuple}