unweighted graph returns error

[matei-ionita]

When supplying an unweighted graph (for example, the output of igraph::graph_from_adj_list) I get:

Error in find_partition(graph, igraph::E(graph)$weight, resolution, n.iterations) :
Not compatible with requested type: [type=NULL; target=double].

This is fixed if I manually set all weights to 1. It would be nice if this were the default behavior for unweighted graphs. Otherwise great implementation!

[evanbiederstedt]

Hi @matei-ionita

Thanks for using leidenAlg!

When supplying an unweighted graph (for example, the output of igraph::graph_from_adj_list)

Could you give me an example, saved as an RDS file perhaps? This sounds like something we could fix.

Thanks, Evan

[matei-ionita]

Hi Evan. Here's a quick example. Works fine if you un-comment the third line, returns error otherwise.

adj_list <- list("1"=c(2,3), "2"=c(1), "3"=c(1), "4"=c(5), "5"=c(4))
gr <- igraph::graph_from_adj_list(adj_list, mode = "all")
# igraph::E(gr)$weight <- rep(1, 3)
leidenAlg::leiden.community(gr)

[evanbiederstedt]

Thanks @matei-ionita, I'll have a look later in the week

[evanbiederstedt]

Hi @matei-ionita

I have something here: #3

The following now runs as you would expect:

adj_list <- list("1"=c(2,3), "2"=c(1), "3"=c(1), "4"=c(5), "5"=c(4))
gr <- igraph::graph_from_adj_list(adj_list, mode = "all")
leidenAlg::leiden.community(gr)

My only remaining question is....is there any rationale at all for throwing an error if a user inputs an unweighted graph? My understanding is that the leiden algorithm will handle the case of all edge weights being the same.

Would a warning be useful along with making all edge weights = 1?

Let me know what you think. I'm happy with the current behavior of course, and we could re-release.

Thanks, Evan

[evanbiederstedt]

Thinking about this more, I think think that the fix is right. I don't think a warning is necessary either.

adj_list <- list("1"=c(2,3), "2"=c(1), "3"=c(1), "4"=c(5), "5"=c(4))
gr <- igraph::graph_from_adj_list(adj_list, mode = "all")
igraph::is_weighted(gr) ## FALSE

leidenAlg::leiden.community(gr)
$membership
[1] 0 0 0 1 1
Levels: 0 1

$dendrogram
NULL

$algorithm
[1] "leiden"

$resolution
[1] 1

$n.iter
[1] 2

$names
NULL

attr(,"class")
[1] "fakeCommunities"

That should be work, I think.

I'll close the issue and re-version. Let me know if there's more here

Thanks for the issue! Best, Evan