# Constant propagation with a namedtuple with one field
**URL:** https://discourse.julialang.org/t/constant-propagation-with-a-namedtuple-with-one-field/121887
**Category:** General Usage
**Created:** [October 28, 2024, 5:31pm UTC](https://discourse.julialang.org/t/constant-propagation-with-a-namedtuple-with-one-field/121887 "2024-10-28T17:31:29Z")
**Posts on this page:** 7
**Page:** 1
### Author:  [@f.ij](https://discourse.julialang.org/u/f.ij)
#### Post date: [October 28, 2024, 5:31pm UTC](https://discourse.julialang.org/t/constant-propagation-with-a-namedtuple-with-one-field/121887/1 "2024-10-28T17:31:29Z")
I’m trying to use the type of a NamedTuple to statically compile some parts of functions. I’ve found some unexpected behavior that I don’t really understand: When I use a NamedTuple with only one field it doesn’t seem to want to do constant propagation.
Im using Julia 1.10.5
```julia
Function get_field_type(::Type{T}, field::Symbol) where T <: NamedTuple
names = fieldnames(T)
types = fieldtypes(T)
index = findfirst(==(field), names)
return index === nothing ? throw(ArgumentError("Field $field not found")) : types[index]
end
function get_field_type(@specialize(nt::NamedTuple), field::Symbol)
return get_field_type(typeof(nt), field)
end
function nttest(a, b , @specialize(nt))
if get_field_type(nt, :a) == Int64
return a+b
elseif get_field_type(nt, :a) == Float64
return a*b
end
end
function test1()
nt = (;a = 1)
@code_warntype nttest(1, 2, nt)
end
function test2()
nt = (;a = 1, b = 2)
@code_warntype nttest(1, 2, nt)
end
```
In this example for test1(), there is an if statement in the function, but for test2() it’s correctly compiled away. Why is this?
```julia
julia> test1()
MethodInstance for nttest(::Int64, ::Int64, ::@NamedTuple{a::Int64})
from nttest(a, b, nt) @ Main ~/Library/Mobile Documents/com~apple~CloudDocs/Documents/PhD/JuliaProjects/InteractiveIsing/Tests/NT.jl:24
Arguments
#self#::Core.Const(nttest)
a::Int64
b::Int64
nt::@NamedTuple{a::Int64}
Body::Union{Nothing, Int64}
1 ─ nothing
│ %2 = Main.get_field_type(nt, :a)::DataType
│ %3 = (%2 == Main.Int64)::Bool
└── goto #3 if not %3
2 ─ %5 = (a + b)::Int64
└── return %5
3 ─ %7 = Main.get_field_type(nt, :a)::DataType
│ %8 = (%7 == Main.Float64)::Bool
└── goto #5 if not %8
4 ─ %10 = (a * b)::Int64
└── return %10
5 ─ return nothing
julia> test2()
MethodInstance for nttest(::Int64, ::Int64, ::@NamedTuple{a::Int64, b::Int64})
from nttest(a, b, nt) @ Main ~/Library/Mobile Documents/com~apple~CloudDocs/Documents/PhD/JuliaProjects/InteractiveIsing/Tests/NT.jl:24
Arguments
#self#::Core.Const(nttest)
a::Int64
b::Int64
nt::@NamedTuple{a::Int64, b::Int64}
Body::Int64
1 ─ nothing
│ %2 = Main.get_field_type(nt, :a)::Core.Const(Int64)
│ %3 = (%2 == Main.Int64)::Core.Const(true)
└── goto #3 if not %3
2 ─ %5 = (a + b)::Int64
└── return %5
3 ─ Core.Const(:(Main.get_field_type(nt, :a)))
│ Core.Const(:(%7 == Main.Float64))
│ Core.Const(:(goto %12 if not %8))
│ Core.Const(:(a * b))
│ Core.Const(:(return %10))
└── Core.Const(:(return nothing))
```
---
### Author:  [@f.ij](https://discourse.julialang.org/u/f.ij)
#### Post date: [October 28, 2024, 5:51pm UTC](https://discourse.julialang.org/t/constant-propagation-with-a-namedtuple-with-one-field/121887/2 "2024-10-28T17:51:08Z")
Okay I’m guessing this is a bug.
If I add a `println` statement to the function `get_field_type` then suddenly it compiles correctly.
```julia
Function get_field_type(::Type{T}, field::Symbol) where T <: NamedTuple
println("Bla")
names = fieldnames(T)
types = fieldtypes(T)
index = findfirst(==(field), names)
return index === nothing ? throw(ArgumentError("Field $field not found")) : types[index]
end
```
Works fine.
---
### Author:  [@f.ij](https://discourse.julialang.org/u/f.ij)
#### Post date: [October 28, 2024, 11:22pm UTC](https://discourse.julialang.org/t/constant-propagation-with-a-namedtuple-with-one-field/121887/3 "2024-10-28T23:22:20Z")
Wrapping the symbol in a `Val` type also works. Still, I don’t know why this is necessary and what makes NamedTuples of a single field work different. Any help would be appreciated.
```julia
get_field_type(nt, field::Symbol) = get_field_type(nt, Val{field}())
function get_field_type(::Type{T}, field::Val{S}) where {T <: NamedTuple,S}
names = fieldnames(T)
types = fieldtypes(T)
index = findfirst(==(S), names)
return index === nothing ? throw(ArgumentError("Field $field not found")) : types[index]
end
```
---
### Author:  [@nsajko](https://discourse.julialang.org/u/nsajko)
#### Post date: [October 29, 2024, 9:26pm UTC](https://discourse.julialang.org/t/constant-propagation-with-a-namedtuple-with-one-field/121887/4 "2024-10-29T21:26:51Z")
Opened an issue:
>
>
> Reproducer:
>
> \`\`\`julia
> function get\_field\_type(::Type{T}, field::Symbol) where {T… \<: NamedTuple}
> names = fieldnames(T)
> types = fieldtypes(T)
> index = findfirst(==(field), names)
> if index === nothing
> throw(ArgumentError("Field $field not found"))
> end
> types\[index\]
> end
>
> function get\_field\_type(nt::NamedTuple, field::Symbol)
> get\_field\_type(typeof(nt), field)
> end
>
> function f(nt)
> get\_field\_type(nt, :a)
> end
>
> Core.Compiler.return\_type(f, Tuple{typeof((; a = 1))}) # inference fails: \`DataType\`
> Core.Compiler.return\_type(f, Tuple{typeof((; a = 1, b = 2))}) # OK: \`Type{Int64}\`
> \`\`\`
>
> This comes from Discourse: https://discourse.julialang.org/t/constant-propagation-with-a-namedtuple-with-one-field/121887
>
> Both return types were inferred as \`Type{Int64}\` back in v1.9.
Do you need a workaround or something?
---
### Author:  [@f.ij](https://discourse.julialang.org/u/f.ij)
#### Post date: [October 29, 2024, 11:20pm UTC](https://discourse.julialang.org/t/constant-propagation-with-a-namedtuple-with-one-field/121887/5 "2024-10-29T23:20:51Z")
It seems workaround with the Val type works for now, so it’s not a huge deal for me, assuming it will work otherwise. Still, I don’t understand why a `println` statement makes it suddenly compile correctly.
---
### Author:  [@Benny](https://discourse.julialang.org/u/Benny)
#### Post date: [October 30, 2024, 3:49am UTC](https://discourse.julialang.org/t/constant-propagation-with-a-namedtuple-with-one-field/121887/6 "2024-10-30T03:49:23Z")
Reproducible on 1.10.6 and 1.11.1 too
---
### Author:  [@nsajko](https://discourse.julialang.org/u/nsajko)
#### Post date: [November 1, 2024, 8:03am UTC](https://discourse.julialang.org/t/constant-propagation-with-a-namedtuple-with-one-field/121887/7 "2024-11-01T08:03:30Z")
Fixed already!