`gridPathCells` can fail

[charleskawczynski]

From H3.jl (which simply calls the C binding):

import H3
p0 = (;latitude=41.141416500334635, longitude=-85.62262147756336)
p1 = (;latitude=36.678141479650726, longitude=-105.4336830442914)
cell(p, res) =
	H3.API.latLngToCell(H3.API.LatLng(deg2rad(p.latitude), deg2rad(p.longitude)), res)
c0 = cell(p0, 3)
c1 = cell(p1, 3)
@show H3.API.gridPathCellsSize(c0, c1) # returns error code object
@show H3.API.gridPathCells(c0, c1) # errors

yields

julia> import H3

julia> p0 = (;latitude=41.141416500334635, longitude=-85.62262147756336)
(latitude = 41.141416500334635, longitude = -85.62262147756336)

julia> p1 = (;latitude=36.678141479650726, longitude=-105.4336830442914)
(latitude = 36.678141479650726, longitude = -105.4336830442914)

julia> cell(p, res) =
               H3.API.latLngToCell(H3.API.LatLng(deg2rad(p.latitude), deg2rad(p.longitude)), res)
cell (generic function with 1 method)

julia> c0 = cell(p0, 3)
0x0832a96fffffffff

julia> c1 = cell(p1, 3)
0x08348dbfffffffff

julia> @show H3.API.gridPathCellsSize(c0, c1) # returns error code object

H3.API.gridPathCellsSize(c0, c1) = H3.API.H3ErrorCode(0x00000001)
H3ErrorCode(E_FAILED): The operation failed but a more specific error is not available

julia> @show H3.API.gridPathCells(c0, c1) # errors
ERROR: MethodError: no method matching Vector{UInt64}(::UndefInitializer, ::H3.API.H3ErrorCode)
The type `Vector{UInt64}` exists, but no method is defined for this combination of argument types when trying to construct it.

Closest candidates are:
  Array{T, N}(::Missing, ::Any...) where {T, N}
   @ Base baseext.jl:43
  Vector{T}(::UndefInitializer, ::Tuple{Int64}) where T
   @ Core boot.jl:591
  Vector{T}(::UndefInitializer, ::Int64) where T
   @ Core boot.jl:577
  ...

Stacktrace:
 [1] gridPathCells(origin::UInt64, destination::UInt64)
   @ H3.API ~/.julia/packages/H3/U8xMC/src/API.jl:384
 [2] macro expansion
   @ show.jl:1232 [inlined]
 [3] top-level scope
   @ REPL[8]:1

which shows that we're getting H3ErrorCode.

[Abhayindia]

still reproduces on 4.5.0 (via h3-py, same two cells): Couldn't find line between cells 0x832a96fffffffff and 0x8348dbfffffffff.
worth noting gridDistance on that same pair fails the same way, so it's falling over before gridPathCells gets involved - looks like the local IJ conversion rather than the line walk itself. taking a look if nobody's on it.

[Abhayindia]

had a closer look and i think this is the documented range limit rather than a bug in gridPathCells.

cellToLocalIjk says "Failure may occur if the index is too far away from the origin or if the index is on the other side of a pentagon", and gridDistance fails on this pair for that reason before gridPathCells is reached. measuring where it starts, from the reported c0 at res 3, testing every cell of each true ring:

  largest k where every cell resolves:  14
  first k with any failure:             15  (5 of 90)
  k=40:                                 201 of 225 fail

the reported c1 is around 40 cells out, so it's well past where local IJ stops working.

so nothing looks broken - the catch is that Couldn't find line between cells ... doesn't say "out of range", which is why it reads as a bug. would a clearer error, or a note about the local-IJ range in the gridPathCells docs, be worth doing? happy to close this otherwise.