more tuple/apply inference fun

[jrevels]

julia> g(args...) = h(args...)
g (generic function with 1 method)

julia> h(f, args...) = f(args...)
h (generic function with 1 method)

julia> f(T, y) = g(getfield, g(tuple, T, y), 1)
f (generic function with 1 method)

julia> @code_typed f(AbstractFloat, 1)
CodeInfo(
1 ─ %1 = Main.tuple::Const(tuple, false)
│   %2 = (%1)(T, y)::Tuple{DataType,Int64}
│   %3 = Main.getfield::Const(getfield, false)
│   %4 = (%3)(%2, 1)::DataType
└──      return %4
) => DataType

julia> k(T, y) = getfield(tuple(T, y), 1)
k (generic function with 1 method)

julia> @code_typed k(AbstractFloat, 1)
CodeInfo(
1 ─     return AbstractFloat
) => Type{AbstractFloat}

julia> versioninfo()
Julia Version 1.1.0-DEV.843
Commit a0bc8fdb75 (2018-12-08 18:35 UTC)
Platform Info:
  OS: macOS (x86_64-apple-darwin17.5.0)
  CPU: Intel(R) Core(TM) i7-7920HQ CPU @ 3.10GHz
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-6.0.1 (ORCJIT, skylake)

MWE from Cassette; I'd like the output of f(...) to be inferred as Type{AbstractFloat} (the same as k(...))

[jrevels]

To be more clear:

julia> @code_typed optimize=false f(AbstractFloat, 1)
CodeInfo(
1 ─ %1 = (Main.g)(Main.tuple, T, y)::Tuple{DataType,Int64}
│   %2 = (Main.g)(Main.getfield, %1, 1)::DataType
└──      return %2
) => DataType

julia> @code_typed optimize=false k(AbstractFloat, 1)
CodeInfo(
1 ─ %1 = (Main.tuple)(T, y)::PartialTuple(Tuple{DataType,Int64}, Any[Const(AbstractFloat, false), Int64])
│   %2 = (Main.getfield)(%1, 1)::Const(AbstractFloat, false)
└──      return %2
) => Type{AbstractFloat}