Julia: Can't infer types of elements of named tuples under getindex

[andyferris]

I had hoped that indexing a named tuple (#22194) with a Symbol would be a type-inferrable operation. I'm guessing that a tfunc or whatever, like for Tuple, may be needed here.

   _       _ _(_)_     |  A fresh approach to technical computing
  (_)     | (_) (_)    |  Documentation: https://docs.julialang.org
   _ _   _| |_  __ _   |  Type "?help" for help.
  | | | | | | |/ _` |  |
  | | |_| | | | (_| |  |  Version 0.7.0-DEV.2400 (2017-11-02 03:12 UTC)
 _/ |\__'_|_|_|\__'_|  |  ajf/empty/5fa163f (fork: 1 commits, 0 days)
|__/                   |  x86_64-linux-gnu

julia> nt = (a=1, b=2.0)
(a = 1, b = 2.0)

julia> f(x) = x[:a]
f (generic function with 1 method)

julia> @code_warntype f(nt)
Variables:
  x::NamedTuple{(:a, :b),Tuple{Int64,Float64}}

Body:
  begin
      return (Base.getfield)(x::NamedTuple{(:a, :b),Tuple{Int64,Float64}}, :a)::Union{Float64, Int64}
  end::Union{Float64, Int64}

[StefanKarpinski]

@vtjnash, I believe you've done some constant prop work that might soon help here?

[andyferris]

That would be even better :) I changed the OP to wrap in a function to ~better~ actually reflect the problem.

[StefanKarpinski]

That's the only way this could work. Note that the nt.a form is type inferred:

julia> f(x) = x.a
f (generic function with 1 method)

julia> @code_warntype f(nt)
Variables:
  x::NamedTuple{(:a, :b),Tuple{Int64,Float64}}

Body:
  begin
      return (Core.getfield)(x::NamedTuple{(:a, :b),Tuple{Int64,Float64}}, :a)::Int64
  end::Int64

[bramtayl]

You can usually get around constant propagation issues by wrapping things in a type. So Val{:a}() or Field{:a}() ala https://github.com/JuliaLang/julia/issues/1974#issuecomment-286961466 could avoid this issue. One reason why having a nice syntax for Field{:a}() would be nice, like .a. Alternatively @pure annotations seem to help.

[andyferris]

Yes, you can definitely work around this in a variety of ways - it's just that intuitively we can use getindex on a tuple so why not a named tuple?

[JeffBezanson]

Duplicate of #24011

[JeffBezanson]

See PR #24362