C# has a convenient way to create iterators using the yield return statement. The package ResumableFunctions provides the same functionality for the Julia language by introducing the @resumable and the @yield macros. These macros can be used to replace the Task switching functions produce and consume which were deprecated in Julia v0.6. Channels are the preferred way for inter-task communication in julia v0.6+, but their performance is subpar for iterator applications. See the benchmarks section below.

ResumableFunctions is a registered package and can be installed by running:

using Pkg
Pkg.add("ResumableFunctions")

using ResumableFunctions

@resumable function fibonnaci(n::Int) :: Int
  a = 0
  b = 1
  for i in 1:n
    @yield a
    a, b = b, a+b
  end
end

for fib in fibonnaci(10)
  println(fib)
end

The following block is the result of running julia --project=. benchmark/benchmarks.jl on a computer with the processor: Intel(R) Core(TM) i5-5300U CPU @ 2.30GHz. Julia version 1.1.1 was used.

Direct:
  49.724 ns (0 allocations: 0 bytes)
ResumableFunctions:
  10.230 μs (281 allocations: 8.83 KiB)
Channels csize=0:
  305.877 μs (465 allocations: 8.84 KiB)
Channels csize=1:
  504.438 μs (379 allocations: 7.33 KiB)
Channels csize=20:
  87.889 μs (206 allocations: 5.06 KiB)
Channels csize=100:
  67.911 μs (198 allocations: 6.44 KiB)
Closure:
  2.437 μs (83 allocations: 1.31 KiB)
Closure optimised:
  245.676 ns (3 allocations: 64 bytes)
Closure statemachine:
  40.149 ns (0 allocations: 0 bytes)
Iteration protocol:
  63.635 ns (0 allocations: 0 bytes)

• Ben Lauwens, Royal Military Academy, Brussels, Belgium.

• To discuss problems or feature requests, file an issue. For bugs, please include as much information as possible, including operating system, julia version, and version of MacroTools.
• To contribute, make a pull request. Contributions should include tests for any new features/bug fixes.

• 2019: v0.5.1

  • inference problem solved: force iterator next value to be of type Union of Tuple and Nothing.

• 2019: v0.5.0 is Julia v1.2 compatible

• 2018: v0.4.2 prepare for Julia v1.1

  • better inference caused a problem;).
  • iterator with a specified rtype is fixed.

• 2018: v0.4.0 is Julia v1.0 compatible.

• 2018: v0.3.1 uses the new iteration protocol.

  • the new iteration protocol is used for a @resumable function based iterator.
  • the for loop transformation implements also the new iteration protocol.

• 2018: v0.3 is Julia v0.7 compatible.

  • introduction of let block to allow variables not te be persisted between @resumable function calls (EXPERIMENTAL).
  • the eltype of a @resumable function based iterator is its return type if specified, otherwise Any.

• 2018: v0.2 the iterator now behaves as a Python generator: only values that are explicitely yielded are generated; the return value is ignored and a warning is generated.

• 2017: v0.1 initial release that is Julia v0.6 compatible:

  • Introduction of the @resumable and the @yield macros.
  • A @resumable function generates a type that implements the iterator interface.
  • Parametric @resumable functions are supported.

• In a try block only top level @yield statements are allowed.
• In a finally block a @yield statement is not allowed.
• An anonymous function can not contain a @yield statement.