Your Bug or Mine?

[cortner]

Hi - as mentioned I am trying to benchmark ResumableFunctions against other iteration protocols. The following below is something that is a bit on the simplistic side but reasonably close to a recent use-case I had. The println statements are for debugging since I got strange results.

using ResumableFunctions
using BenchmarkTools

@resumable function g(N)::Vector{Int}
   for n = 1:N
      println()
      println(n)
      for n1 = 1:10, n2 = 1:10
         print("*")
         # do something not terribly expensive
         x = rand(-10:10, 3)
         @yield x
      end
   end
   Int[]
end

f_resumable(N) = sum( sum(x) for x in g(N) )

f_resumable(100)

On Python-generators-v0.6 I get

1
*

as output which just means that it terminates after the first @yield. Why?

When I use the release for v0.6, the I get

1
*ERROR: UndefVarError: #temp# not defined
Stacktrace:
 [1] macro expansion at /Users/ortner/.julia/v0.6/ResumableFunctions/src/transforms.jl:14 [inlined]
 [2] (::##662)(::Void) at /Users/ortner/.julia/v0.6/MacroTools/src/utils.jl:12
 [3] next at ./generator.jl:44 [inlined]
 [4] mapfoldl_impl(::Base.#identity, ::Base.#+, ::Int64, ::Base.Generator{##662,Base.#sum}, ::UInt8) at ./reduce.jl:42
 [5] mapfoldl(::Base.#identity, ::Function, ::Base.Generator{##662,Base.#sum}) at ./reduce.jl:73
 [6] f_resumable(::Int64) at ./REPL[4]:1

which is probably related.

(just in case this turns out to be a stupid bug at my end, let me just apologise profusely in advance . . .)

[BenLauwens]

Duplicate of bug #14

[cortner]

Ups - sorry.