# Strange @code\_warntype
**URL:** https://discourse.julialang.org/t/strange-code-warntype/14002
**Category:** New to Julia
**Created:** [August 24, 2018, 3:24pm UTC](https://discourse.julialang.org/t/strange-code-warntype/14002 "2018-08-24T15:24:52Z")
**Posts on this page:** 10
**Page:** 1
### Author:  [@tomtom](https://discourse.julialang.org/u/tomtom)
#### Post date: [August 24, 2018, 3:24pm UTC](https://discourse.julialang.org/t/strange-code-warntype/14002/1 "2018-08-24T15:24:52Z")
hello, I’ve tried to ensure type stability by using `@code_warntype`. However, something very strange happened: In the following, `closureB()` only differs from `closureA()` with a leading ` if-end` block, which is irrelevant actually. But, a lot of types in `gg()` becomes **`ANY`** , is it a bug??? I’m using JuliaPro-0.6.3.1.
```julia
function closureA(x::Int64)::Function
oneto3 = Base.OneTo(3)
function f()
for i in oneto3
println(i)
end
end
return f
end
function closureB(x::Int64)::Function
### the only difference with closureA()
if x < 0
error("abc")
end
### the only difference with closureB()
oneto3 = Base.OneTo(3)
function f()
for i in oneto3
println(i)
end
end
return f
end
ff = closureA(1)
gg = closureB(1)
julia> @code_warntype ff()
Variables:
#self#::#f#5{Base.OneTo{Int64}}
i::Int64
#temp#::Int64
Body:
begin
SSAValue(0) = (Core.getfield)(#self#::#f#5{Base.OneTo{Int64}}, :oneto3)::Base.OneTo{Int64}
#temp#::Int64 = 1
3:
unless (Base.not_int)((#temp#::Int64 === (Base.add_int)((Core.getfield)(SSAValue(0), :stop)::Int64, 1)::Int64)::Bool)::Bool goto 18
SSAValue(3) = #temp#::Int64
SSAValue(4) = (Base.add_int)(#temp#::Int64, 1)::Int64
i::Int64 = SSAValue(3)
#temp#::Int64 = SSAValue(4) # line 6:
$(Expr(:inbounds, false))
# meta: location coreio.jl println 5
SSAValue(2) = (Core.typeassert)(Base.STDOUT, Base.IO)::IO
# meta: pop location
$(Expr(:inbounds, :pop))
(Base.print)(SSAValue(2), i::Int64, $(QuoteNode('\n')))::Void
16:
goto 3
18:
return
end::Void
julia> @code_warntype gg()
Variables:
#self#::#f#8
i::Any
#temp#::Any
Body:
begin
SSAValue(0) = (Core.getfield)((Core.getfield)(#self#::#f#8, :oneto3)::Any, :contents)::Any
#temp#::Any = (Base.start)(SSAValue(0))::Any
3:
unless !((Base.done)(SSAValue(0), #temp#::Any)::Any)::Any goto 12
SSAValue(1) = (Base.next)(SSAValue(0), #temp#::Any)::Any
i::Any = (Core.getfield)(SSAValue(1), 1)::Any
#temp#::Any = (Core.getfield)(SSAValue(1), 2)::Any # line 12:
(Main.println)(i::Any)::Void
10:
goto 3
12:
return
end::Void
```
---
### Author:  [@Wenjie](https://discourse.julialang.org/u/Wenjie)
#### Post date: [August 24, 2018, 3:34pm UTC](https://discourse.julialang.org/t/strange-code-warntype/14002/2 "2018-08-24T15:34:23Z")
Tried your code on v1.0. It works fine:
```julia
julia> @code_warntype gg()
Body::Nothing
11 1 ── %1 = (Core.getfield)(#self#, :oneto3)::Base.OneTo{Int64}
│ %2 = (Base.getfield)(%1, :stop)::Int64 │╻╷╷ iterate
│ %3 = (Base.slt_int)(%2, 1)::Bool ││╻ isempty
└─── goto #3 if not %3 ││
2 ── goto #4 ││
3 ── goto #4 ││
4 ┄─ %7 = φ (#2 => true, #3 => false)::Bool │
│ %8 = φ (#3 => 1)::Int64 │
│ %9 = φ (#3 => 1)::Int64 │
│ %10 = (Base.not_int)(%7)::Bool │
└─── goto #10 if not %10 │
5 ┄─ %12 = φ (#4 => %8, #9 => %21)::Int64 │
│ %13 = φ (#4 => %9, #9 => %22)::Int64 │
12 │ invoke Main.println(%12::Int64) │
│ %15 = (Base.getfield)(%1, :stop)::Int64 ││╻╷ last
│ %16 = (%13 === %15)::Bool ││╻ ==
└─── goto #7 if not %16 ││
6 ── goto #8 ││
7 ── %19 = (Base.add_int)(%13, 1)::Int64 ││╻ +
└─── goto #8 │╻ iterate
8 ┄─ %21 = φ (#7 => %19)::Int64 │
│ %22 = φ (#7 => %19)::Int64 │
│ %23 = φ (#6 => true, #7 => false)::Bool │
│ %24 = (Base.not_int)(%23)::Bool │
└─── goto #10 if not %24 │
9 ── goto #5 │
10 ─ return │
```
---
### Author:  [@kristoffer.carlsson](https://discourse.julialang.org/u/kristoffer.carlsson)
#### Post date: [August 24, 2018, 3:36pm UTC](https://discourse.julialang.org/t/strange-code-warntype/14002/3 "2018-08-24T15:36:32Z")
The issue to track for this is [performance of captured variables in closures · Issue #15276 · JuliaLang/julia · GitHub](https://github.com/JuliaLang/julia/issues/15276).
---
### Author:  [@tomtom](https://discourse.julialang.org/u/tomtom)
#### Post date: [August 24, 2018, 3:40pm UTC](https://discourse.julialang.org/t/strange-code-warntype/14002/4 "2018-08-24T15:40:40Z")
thanks.
just a silly question: could I have more than one version of Julia installed? (wanna know before crashing everything)
---
### Author:  [@Wenjie](https://discourse.julialang.org/u/Wenjie)
#### Post date: [August 24, 2018, 3:49pm UTC](https://discourse.julialang.org/t/strange-code-warntype/14002/5 "2018-08-24T15:49:56Z")
> could I have more than one version of Julia installed? (wanna know before crashing everything)
I am not the perfect person to answer this question. I only know, in Linux, you can download the binaries for various versions, but I’m not sure whether the packages of various versions are by default installed at the same place (hopefully not).
---
### Author:  [@kristoffer.carlsson](https://discourse.julialang.org/u/kristoffer.carlsson)
#### Post date: [August 24, 2018, 3:52pm UTC](https://discourse.julialang.org/t/strange-code-warntype/14002/6 "2018-08-24T15:52:34Z")
Yes, there is no problem with multiple installations.
---
### Author:  [@tomtom](https://discourse.julialang.org/u/tomtom)
#### Post date: [August 24, 2018, 3:55pm UTC](https://discourse.julialang.org/t/strange-code-warntype/14002/7 "2018-08-24T15:55:00Z")
performance-wise, should I treat those “captured variables in closure” like global variables and declare them with `const` ? for example:
```julia
function closure(x::Int64)
const y = fill(x, 3) # "const" needed here for performance ???
function f(z::Int64)
y .+= z
println(y)
end
return f
end
```
---
### Author:  [@kristoffer.carlsson](https://discourse.julialang.org/u/kristoffer.carlsson)
#### Post date: [August 24, 2018, 3:57pm UTC](https://discourse.julialang.org/t/strange-code-warntype/14002/8 "2018-08-24T15:57:27Z")
`const` doesn’t do anything in local scope. There are some workaround in that issue including using `let` blocks and type annotations.
---
### Author:  [@tomtom](https://discourse.julialang.org/u/tomtom)
#### Post date: [August 24, 2018, 4:03pm UTC](https://discourse.julialang.org/t/strange-code-warntype/14002/9 "2018-08-24T16:03:53Z")
> [@kristoffer.carlsson](#):
>
> There are some workaround in that issue including using `let` blocks and type annotations.
could you kindly give an example (when u’re free) ?
I just don’t know how to let the compiler knows the type of those “captured variables in closure” when the function is “generated”.
---
### Author:  [@kristoffer.carlsson](https://discourse.julialang.org/u/kristoffer.carlsson)
#### Post date: [August 24, 2018, 4:11pm UTC](https://discourse.julialang.org/t/strange-code-warntype/14002/10 "2018-08-24T16:11:36Z")
```julia
julia> function closureB(x::Int64)::Function
### the only difference with closureA()
if x < 0
error("abc")
end
### the only difference with closureB()
oneto3 = Base.OneTo(3)
local f
let oneto3 = oneto3
function f()
for i in oneto3
println(i)
end
end
end
return f
end
closureB (generic function with 1 method)
julia> gg = closureB(1)
(::f) (generic function with 1 method)
julia> @code_warntype gg()
Variables:
#self#::#f#6{Base.OneTo{Int64}}
i::Int64
#temp#::Int64
Body:
begin
SSAValue(0) = (Core.getfield)(#self#::#f#6{Base.OneTo{Int64}}, Symbol("#4#oneto3"))::Base.OneTo{Int64}
#temp#::Int64 = 1
3:
unless (Base.not_int)((#temp#::Int64 === (Base.add_int)((Core.getfield)(SSAValue(0), :stop)::Int64, 1)::Int64)::Bool)::Bool goto 18
SSAValue(3) = #temp#::Int64
SSAValue(4) = (Base.add_int)(#temp#::Int64, 1)::Int64
i::Int64 = SSAValue(3)
#temp#::Int64 = SSAValue(4) # line 13:
$(Expr(:inbounds, false))
# meta: location coreio.jl println 5
SSAValue(2) = (Core.typeassert)(Base.STDOUT, Base.IO)::IO
# meta: pop location
$(Expr(:inbounds, :pop))
(Base.print)(SSAValue(2), i::Int64, $(QuoteNode('\n')))::Void
16:
goto 3
18:
return
end::Void
```