feat: add error if adding a vertices results in duplicated names attribute

[schochastics]

Fix #46 (check if #60 could be addressed)

Should this be an error now?

rigraph/R/operators.R

Lines 264 to 266 in 0386a24

	if ("name" %in% names(attr) && any(duplicated(attr$name))) {
	cli::cli_warn("Duplicate vertex names in disjoint union.")
	}

also need a strategy for this

library(igraph)
g1 <- g2 <- make_ring(4)
V(g1)$name <- c("A", "B", "C", "D")
V(disjoint_union(g1, g2))$name
#> Warning: Duplicate vertex names in disjoint union.
#> [1] "A" "B" "C" "D" NA  NA  NA  NA

^{Created on 2025-07-03 with reprex v2.1.1}

[schochastics]

re #60 and the disjoint_union() case:
For consistency, it might be safer to enforce consistent naming too. So we should probably also be throwing an error.

Cons: Might be annoying for quick prototyping

[schochastics]

Should also look at #1250

[maelle]

why is this needed? (genuine question)

[schochastics]

If the graph has no name attribute but we add a vertex with a name, we need to set all other names to NA (this is consistent with most of igraph but not really a good solution. Needs to be addressed elsewhere)

h <- make_full_graph(3)
set_vertex_attr(h, "name", 2, "B")
#>IGRAPH 4d3e0a7 UN-- 3 3 -- Full graph
#>+ attr: name (g/c), loops (g/l), name (v/c)
#>+ edges from 4d3e0a7 (vertex names):
#>[1] NA--B  NA--NA B --NA

[maelle]

I was more curious about the as.numeric().

[schochastics]

ah sorry. That was actually not needed and my mistake

[krlmlr]

Please resolve conversations, happy to review in a second iteration. The code looks complex.

[schochastics]

all comments addressed and I think I made it a bit simpler

[szhorvat]

Does this affect only disjoint_union? There are other ways to create duplicate names.

[schochastics]

Yes there should be more places. The question is, what should happen in this case. I think we reached a form of consensus last week, but I can't remember what we thought was the best way to deal with it. I think it might have been to throw an error if a named and unnamed object are combined

[schochastics]

Ok I realise I talked about combining unnamed and named objects only. The pr is about erroring for duplicated vertex names. I think this covers all places where this can happen or do I see any other place?