Haskell-style map function for ACSets

[olynch]

Pretty simple, basically:

map(Int,s -> Int(s[1]),:Weight,g::WeightedGraph{String})::WeightedGraph{Int}

All attributes with a given codomain get converted using the supplied function. Partly addresses AlgebraicJulia/ACSets.jl#7.

[olynch]

@epatters btw this is ready to be reviewed.

[epatters]

Cool, thanks. I'll review it soon.

[epatters]

Thanks Owen. Two questions/comments:

1. Do you think there will be many use cases for this as is? My worry is that if you want to transform multiple attributes, you'd have to call map multiple times and end up reallocating the whole C-set each time. Also, maybe sometimes you want to map all data attributes with given codomain, as here, or just a particular data attribute. I think you could generalize this to something like mutate in dplyr without too much change to the structure of your code but a big boost in usefulness.

2. In idiomatic Julia you shouldn't usually have to specify types that can be inferred. So instead of supplying Int as a first argument, you compute map(s -> Int(s[1]), ...) in base Julia, which returns a Vector{Int} whose element type you can extract. I guess this would complicate the code a bit.

Putting these suggestions together, I would imagine an API supporting examples like

map(g::WeightedGraph{Symbol}, weight = s -> Int(s[1])) # `weight` attribute only
map(g::WeightedGraph{Symbol}, Weight = s -> Int(s[1])) # all attributes with codomain `Weight`

map(g::RelationDiagram{Symbol}, name = titlecase∘string, variable = uppercase∘string)

What do you think?

[olynch]

OK, so I guess then there is an error if you don't correctly map all of the attributes to a new type?

Yeah, can do!

[epatters]

Thanks Owen, I think this method will be very useful.

[epatters]

It'd be nice if this were built into Julia. I call unique!(sort!(x)) all over the codebase.

[epatters]

Awesome, merging now.