Scalars in acset transformations

[slwu89]

When defining an acset transformation from a domain acset with only 1 element in some Ob's set, sometimes a scalar will sneak into the components rather than a vector. When this hits coerce_component it results in a rather scary method error (see below). Because defining acset transformations can be rather complex, when seeing this kind of error a user might fruitlessly spend time looking for a bug in the mapping, instead of something "simple" like using a scalar rather than a vector in the components.

It would be nice to let coerce_component allow for scalars, or make the necessary conversion somewhere in the call stack, if it doesn't cause any problems with the rest of the code design.

ERROR: MethodError: no method matching (SetFunction{Dom, Codom} where {S, S′, Dom<:(FinSet{S}), Codom<:(FinSet{S′})})(::Int64, ::Int64, ::Int64)

Closest candidates are:
  (SetFunction{Dom, Codom} where {S, S′, Dom<:(FinSet{S}), Codom<:(FinSet{S′})})(::typeof(identity), ::Any...)
   @ Catlab ~/.julia/packages/Catlab/77h9f/src/categorical_algebra/FinSets.jl:186
  (SetFunction{Dom, Codom} where {S, S′, Dom<:(FinSet{S}), Codom<:(FinSet{S′})})(::AbstractDict, ::Any...)
   @ Catlab ~/.julia/packages/Catlab/77h9f/src/categorical_algebra/FinSets.jl:188
  (SetFunction{Dom, Codom} where {S, S′, Dom<:(FinSet{S}), Codom<:(FinSet{S′})})(::ACSets.Columns.Column{Int64, Int64}, ::Any, ::Any)
   @ Catlab ~/.julia/packages/Catlab/77h9f/src/categorical_algebra/CSets.jl:88
  ...

Stacktrace:
  [1] coerce_component(::Symbol, f::Int64, dom_size::Int64, codom_size::Int64; kw::@Kwargs{})
    @ Catlab.CategoricalAlgebra.CSets ~/.julia/packages/Catlab/77h9f/src/categorical_algebra/CSets.jl:465
  [2] #21
    @ ~/.julia/packages/Catlab/77h9f/src/categorical_algebra/CSets.jl:444 [inlined]
  [3] map(f::Catlab.CategoricalAlgebra.CSets.var"#21#24"{…}, t::NTuple{…})
    @ Base ./tuple.jl:294
  [4] coerce_components(S::Type, components::@NamedTuple{…}, X::WorkflowGraph{…}, Y::WorkflowGraph{…})
    @ Catlab.CategoricalAlgebra.CSets ~/.julia/packages/Catlab/77h9f/src/categorical_algebra/CSets.jl:443
  [5] (StructTightACSetTransformation{…} where {…})(components::@NamedTuple{…}, X::WorkflowGraph{…}, Y::WorkflowGraph{…})
    @ Catlab.CategoricalAlgebra.CSets ~/.julia/packages/Catlab/77h9f/src/categorical_algebra/CSets.jl:376
  [6] macro expansion
    @ ~/.julia/packages/Catlab/77h9f/src/categorical_algebra/CSets.jl:0 [inlined]
  [7] macro expansion
    @ ~/.julia/packages/CompTime/Ppb3B/src/CompTime.jl:137 [inlined]
  [8] comptime(::typeof(Catlab.CategoricalAlgebra.CSets._ACSetTransformation), 763::Type{…}, components::@NamedTuple{…}, X::WorkflowGraph{…}, Y::WorkflowGraph{…}, 764::Type{…})
    @ Catlab.CategoricalAlgebra.CSets ~/.julia/packages/CompTime/Ppb3B/src/CompTime.jl:137
  [9] _ACSetTransformation
    @ ./none:0 [inlined]
 [10] ACSetTransformation(components::@NamedTuple{…}, X::WorkflowGraph{…}, Y::WorkflowGraph{…})
    @ Catlab.CategoricalAlgebra.CSets ~/.julia/packages/Catlab/77h9f/src/categorical_algebra/CSets.jl:549
 [11] top-level scope
    @ ~/Desktop/git/dssi-dyve-gcsprod/dev/workflow/process-workflows.jl:223
Some type information was truncated. Use `show(err)` to see complete types.

[epatters]

In Julia, unlike in say R, scalars are not identified with one-element vectors and I think we should respect that. OTOH, I agree that that error message is scary and unhelpful. If it's a common mistake, I would suggest that we catch the mistake earlier in the call stack and throw an informative error message with a suggestion about what the user should have typed instead.

[slwu89]

@epatters if that's the design convention to stick to, that makes sense to me.

In that case, I could suggest changing coerce_component (defined here

Catlab.jl/src/categorical_algebra/CSets.jl

Lines 465 to 466 in b71a5b8

	coerce_component(::Symbol, f, dom_size::Int, codom_size::Int; kw...) =
	FinFunction(f, dom_size, codom_size; kw...)

) into two methods coerce_component(::Symbol, f::T, dom_size::Int, codom_size::Int; kw...) where {T<:Integer} and coerce_component(::Symbol, f::AbstractVector{<:Integer}, dom_size::Int, codom_size::Int; kw...), where the first will throw an informative error, assuming that by that point coerce_component will be getting integers or vectors of integers. Happy to make a small PR.

[epatters]

Thanks @slwu89! Happy to accept that PR.