Add FAQ with correct solution(s) for "static" variables in functions

[dpsanders]

cf. #15056, #12627

[cossio]

Agree this should be documented. But what is the correct solution?

[dpsanders]

One is to use a global closure inside a let as follows:

julia> let a = 0
           global function f()
               a += 1
           end
       end
f (generic function with 1 method)

julia> f()
1

julia> f()
2

julia> a
UndefVarError: a not defined

I am not sure how to make a const static variable like this, though.

[JeffBezanson]

I am not sure how to make a const static variable like this, though.

Don't change its value :)

[dpsanders]

cf. https://discourse.julialang.org/t/const-static-function-parameters/1803/2

[cossio]

What if the static variable depends on a type parameter of the function? In this case each specific method should get its own static variable. For example, suppose I want to write something like,

global function f(x::T) where {T <: Real}
  static a = zero(T)
  return a + 1
end

In this scenario the above approach of defining a using let outside the function body doesn't seem to work because then I don't have access to type parameter T.

[pranshumalik14]

Why can't the compiler or even lowered code in Julia do the let based allocation of the static variable per method definition of the generic function? I have no knowledge about how any of the compilation and transformation of Julia code works but it seems pretty possible to me. Is the static keyword being planned in the (near) future? Because it is also pretty useful for recursive functions.